EPRI
Electric Power
Research Institute
August 1997
TR-108683-V2
EPRI-DOE-EPA Combined Utility
Air Pollutant Control Symposium
The Mega Symposium
SO2 Control Technologies and Continuous
Emission Monitors
EPRI
Electric Power
Research Institute
Sponsored by
Electric Power Research Institute
U.S. Department of Energy
U.S. Environmental Protection Agency
August 25-29, 1997
Washington Hilton & Towers Hotel
Washington, DC
-------
EPRI-DOE-EPA Combined Utility
Air Pollutant Control Symposium
The Mega Symposium
S02 Control Technologies and Continuous Emission Monitors
August 25-29, 1997
Washington Hilton & Towers Hotel
Washington, DC
Conference Chairpersons
George Often, EPRI
Lawrence Ruth, U.S. DOE
David Lachapelle
Sponsored by
Electric Power Research institute
U.S. Department of Energy
U.S. Environmental Protection Agency
Prepared by
Electric Power Research Institute
-------
REPORT SUMMARY
This "Mega" Symposium combined several conferences that had been held separately
over the years to provide utilities and other interested parties with comprehensive
information on air pollution control technologies at a single time and place.
Emphasizing field experience, the conference showcased the state-of-the-art in the
measurement and reduction of NOx, SO2, and particulate /air toxic emissions.
Background
This first-ever "Mega" Symposium combines the SO2 Control Symposium, the Joint
Symposium on Stationary Combustion NOX Controls, the Particulate Control
Symposium, and the control technology portions of the EPRI/DOE International
Conference on Managing Hazardous and Particulate Pollutants. The Symposium also
includes sessions on Continuous Emissions Monitors.
Objective
To provide information on the latest developments and operational experience with
state-of-the-art methods for measuring and reducing NOx, SO2, and particulate/air
toxics emissions from fossil-fueled boilers.
Approach
EPRI, the U.S. Department of Energy, and the U.S. Environmental Protection Agency
cosponsored a "Mega" Symposium in Washington, DC on August 25-29,1997. Over 120
papers were presented with sessions grouped by pollutant, topical area, boiler type,
and/or process.
Key Points
The Symposium proceedings are published in three volumes: Volume I, NOx controls;
Volume II, SO2 Controls and Continuous Emissions Monitors; and Volume III,
Particulates and Air Toxics Controls. Topics covered during formal presentations and
poster sessions include:
• Combustion tuning/optimization
• Low NOX Systems for Coal-, Gas-, and Oil-Fired Boilers
-------
• Selective Catalytic Reduction
• Selective Noncatalytic Reduction
• Cyclones - Combustion NOx Controls
• Full-Scale Flue Gas Desulfurization (FGD) Experience
* FGD Conversions
• FGD Process Improvements
• Dry SO2 Control Processes
• Advanced SO2 Control Processes
• Continuous Emission Monitors
• New Technologies for Particulate Control
• Lab- and Pilot-Scale Research in Mercury Capture by Sorbents
• Mercury Capture by FGD
• Fligh Gas-to-Cloth Ratio Baghouses
- Engineering Studies in Particulate Control
• Postcombustion NOX/SO2 Reduction
TR-108683-V1-V3
Interest Categories
Air emissions control
Air toxics measurement and control
Emissions monitoring
Fossil steam plant performance optimization
Key Words
Nitrogen oxides Air toxics control
Flue gas desulfurization Particulates
SC>2 control Continuous emission monitoring
-------
CONTENTS
Tuesday, August 26; 1:00 p.m.
PARALLEL SESSION C: Full-Scale FGD Experience
• Petersburg Units 1 and 2 Flue Gas Desulfurization System Process Design and
Operation Experience
• Use of Stack Emissions To Control DBA Enhanced FGD Systems Burning
Fuel Highly Variable in Sulfur Content
• A Performance Review of Unit 4 FGD System At Cinergy's Gibson Station
• Henderson Municipal Power & Light: A Low-Cost Phase 1 Clean Air Act Retrofit
• Final Results: The EPRI-DOE-SCS Chiyoda Thoroughbred CT-121 Clean Coal
Project at Georgia Power's Plant Yates
• Wet Gypsum-Yielding FGD Experience Using Quicklime Reagent
• 3 into 1: First Multiple Boiler FGD Unit Started in Northern Alberta
• Design and Start-Up Of a Limestone FGD for an Oil Fired Boiler in
Werndorf/Austria
Wednesday, August 27; 8:00 a.m.
PARALLEL SESSION A: High Velocity FGD Systems
• S02 Compliance After 2000: Slam Dunk or Something Else Altogether?
• LS-2, Two Years of Operating Experience
• High Velocity Mist Elimination for Wet FGD Application
• Phase II: The Age of High Velocity Scrubbing
• Results of High Velocity Single Absorber Operation at OMU's Elmer Smith Station
• New Perspective of Wet Scrubber Fluid Mechanics in an Advanced Tower Design
-------
Wednesday, August 27; 11:00 a.m. 1:00 p.m.
PARALLEL SESSION A: FGD Conversions
• AEC Lowman Station FGD Conversion from Limestone to Magnesium-
Enhanced Lime Scrubbing
• Conversion of the 1600 MW Mill Creek Generating Station to Production of
Commercial-Grade Gypsum
• San Juan Generating Station Limestone FGD Conversion
• Conversion of NIPSCO'S Schahfer Dual Alkali FGD to a Limestone FGD
System Producing Wallboard Grade Gypsum
Wednesday, August 27; 2:00 p.m.
PARALLEL SESSION A: FGD Process Improvements
• Wet FGD Forced Oxidation: A Review of Influencing Factors and a Comparison
of Lance and Sparge Grid Air Introduction Methods
• Limestone Performance in a Pilot-Scale Forced Oxidation Scrubber
• Impact of Limestone Grind Size on WFGD Performance
• WFGD System Materials Cost Update
Wednesday, August 27; 1:00 p.m.
PARALLEL SESSION B: Dry S02 Control Process
• NID - A Mew Dry Flue Gas Desulfurization System
• FGD Experience at Poland's Rybnik Power Station: Dry Method With
Humidrfication
• First North American Circulating Dry Scrubber and Precipitator Remove High
Levels of S02 and Particulates
• B&W's E-Lids™ Process-Advanced SOx, Particulate, and Air Toxic Control
For the Year 2000
• Sodium Sorbents for S02 Trim in High-Ratio Baghouses
-------
Thursday, August 28; 8:00 a.m.
PARALLEL SESSION C: Advanced S02 Control Processes
• Superior FGD Cost-Effectiveness Via High-Volume, High-Value
By-Product Generation
• Operating Experience of the Simplified FGD Systems
• Clear Liquor Scrubbing with Anhydrite Production
• High Velocity Wet Scrubbing of S02 and NOx
• Simultaneous S02 and N02 Removal by Alkaline Solids
Wednesday, August 27; 8:30 a.m.
PARALLEL SESSION C: Continuous Emission Monitors
• The On-Line Real-Time Monitoring of Ammonia, NOx and S02 in Flue Gas
Using a UV-PDA Analyzer
• In situ Analyzer Using a Near Infrared Diode Laser for Process Control and
Environmental Monitoring
• Low Maintenance, High Performance CEMs for NOx, CO and 02 Monitoring
• Study of Stack Flow Test and Ultrasonic Monitor Fundamentals
• Overview of Developing Technologies for Continuous Emission Monitoring
• Development and Field Testing of a Continuous, Real-Time, Speciating Mercury
Analyzer
• Evaluation of Particulate Matter and Total Mercury CEMs for Compliance
Monitoring at Hazardous Waste Combustion Facilities
• Development of a Continuous Emissions Monitor for HCI and CI2
-------
Tuesday, August 26; 1:00 p.m.
Parallel Session C:
Full-scale FGD Experience
-------
PETERSBURG UNITS 1 AND 2
FLUE GAS DESULFURIZATION SYSTEM
PROCESS DESIGN AND OPERATION EXPERIENCE
C.P. Wedig
J.E. Martin
J.J. Youmans
Stone & Webster Engineering Corporation
245 Summer Street
Boston, Massachusetts 02210, USA
C.K. Rutledge
S.R. Wolsiffer
W.K. Watson
Indianapolis Power and Light Company
P.O. Box 1595
Indianapolis, Indiana 46206, USA
Abstract
This paper summarizes the process design and operation experience of the retrofit wet limestone
flue gas desulfurization system (FGDS) project of the Indianapolis Power & Light Company
(IPL), Petersburg Units 1 and 2. This project was initiated by IPL in response to the Clean Air
Act Amendments(CAAA) of 1990 and since mid-1996 has been commercially operating on two
base-load coal fired units with a combined capacity of approximately 700 MWe gross. The
paper will present the process design, use of EPRI FGD programs, and the process performance
testing results. The paper will address the issues of a full-scale retrofit FGDS, high sulfur
dioxide (SO2) removal efficiency, materials of construction, high reliability, continuous emission
monitors (CEMS), particulate removal, performance with and without chemical additives, and
new methods for flue gas and gypsum testing.
IPL-S&W.doc
-------
Description of the Petersburg Units 1 and 2 FGDS Project
Indianapolis Power & Light Company's Petersburg Units 1 and 2 were retrofitted with wet
limestone FGD systems in response to the Clean Air Act Amendments (CAAA) of 1990. A
major challenge of the project was to fit two high-efficiency FGD systems into a tight site that
was bordered by a river and existing plant equipment.
The Petersburg Generation station is located in southwest Indiana approximately three miles
north of Petersburg, IN or 125 miles southwest of Indianapolis. The station consists of four coal-
fired units with the following net generation capacity ratings and year of commercial operation:
Unit 1 239 MW 1967
Unit 2 418 MW 1969
UnitS 510 MW 1977
Unit 4 515 MW 1986
Units 3 & 4 also operate with wet limestone FGDS that were installed as part of the original
equipment.
The major retrofit FGD project activities for Units 1 and 2 included:
Evaluating FGD technologies
Evaluating FGD vendors
Preparing and issuing specifications and evaluating bids
Preparing and submitting documentation and participating in hearings for project
approval with the Indiana Utility Regulatory Commission (IURC)
Securing a contract for the sale of Byproduct Gypsum
Start-up and commissioning of the FGD units
Prepare performance testing specification, evaluate bids, supervise testing, and
review the results
Stone & Webster Engineering Corporation (S&W) was awarded the engineering and
construction contract for the project in February, 1991. The FGD specification was issued to five
bidders in June, 1991. The design, supply and erection of the FGD system was awarded to
General Electric Environmental Services, Inc. (GEESI) in December, 1991.
IPL-S&W.doc
-------
After the GEESI contract was awarded, IPL applied to the Indiana Utility Regulatory
Commission (IURC) for pre-approval of the project. S&W supported IPL in this process with
the necessary studies, cost estimates, schedules, and both written and oral testimony. The
process included public hearings and required 14 months to complete. In August, 1993, the
IURC issued an order approving the project, schedule and budget.
Full scale work on the project resumed in September of 1993, and the following schedule
milestones were confirmed:
Duct Tie in September, 1995
First Flue gas March 1, 1996
Complete Reliability Run May 1, 1996
Complete Owners Acceptance Test June 1, 1996
Commercial Operation June, 1996
IPL completed the project using a "team'1 concept. The "team1' consisted of companies that
provided services for the project. Major companies involved with the work included:
IPL (Owner & Operator)
S&W (Engineering, construction management, and balance of plant design)
GEESI (FGD Process & equip., gypsum dewatering, and LS grinding systems)
Pullman Power Products (chimney)
Radian (Consulting)
All milestone dates were achieved and the project was completed under budget.
Photographs 1 through 3, located at the end of this paper, show the Petersburg Units 1 and 2
FGD system during construction and after commercial operation.
Description of the Petersburg Units 1 and 2 FGDS Process
The Units 1 and 2 FGD process is a forced oxidation wet limestone system that produces
commercial grade gypsum byproduct. Each unit has a single absorber spray tower designed
capable of processing 100% of the flue gas volume for that specific unit. The design inlet flue
gas flow of the individual absorbers is approximately 1.2 and 2.24 million acfm at 340 F for
Units 1 and 2, respectively. The absorbers are open spray towers with each tower having five
levels of recycle slurry spray headers (normally four operating and one spare),oxidation air
headers, mist eliminators, and an absorber reaction tank.
IPL-S&W.doc
-------
Unit 1 employs a single booster fan and Unit 2 employs two booster fans located upstream of the
absorber. The "scrubbed" flue gas is discharged from the absorbers to a new chimney that
contains three flues. Each of two wet flues are dedicated to a respective absorber. No flue gas
reheat is utilized.
The existing Unit 1 and 2 common stack and the third flue in the new chimney can be used for
emergency bypass purposes for Units 1 and 2 respectively. The ability to "bypass" helps
alleviate the risk of boiler implosion to the boiler furnace and duct systems that were originally
designed to operate pressurized but were later converted to balance draft. In order to prevent the
"escape" of unscrubbed gas up the un-dampered bypass flue, a small percentage of ambient air is
drawn down through the bypass system during operation.
Absorber Bleed pumps transport gypsum slurry from the absorber Reaction Tank to Disposal
Feed Tanks (DFT). The gypsum slurry is then pumped 4500 feet to the dewatering area. At this
point, the slurry containing gypsum and water is separated using Hydroclones and horizontal belt
filters. The reclaimed water is reused to produce limestone slurry. A commercial grade gypsum
"cake" containing less than 10% moisture is transferred by conveyor belt to a gypsum storage
building. A front-end loader is used to fill trucks. The trucks are weighed and covered and the
gypsum is transported to the end customer for wallboard manufacture.
To date, 100% of the gypsum has met wallboard specification requirements and been sold to
customers. Reference 1 presents additional information concerning the FGD gypsum.
Photograph 4 shows the commercial grade gypsum piles and stockout area located at grade level
of the gypsum storage building.
A complete limestone handling and preparation facility was installed to provide the slurry needs
of the new FGD systems. The new limestone handling and slurry preparation area is common to
both Units 1 and 2. Limestone is delivered by truck, stored outside, reclaimed, belt conveyed to
limestone silos, and ground by two 75 percent capacity wet ball mill circuits to produce 90
percent through 325 mesh particles. The resultant slurry is stored in limestone slurry storage
tanks and circulated to the FGD absorbers.
A new continuous emission monitoring (CEM) system and data acquisition system was installed
New distributed control system (DCS) consoles for Units 1 and 2 were located in the existing
FGD system power and control building resulting in a centralized control area for all the station's
FGD systems.
IPL-S&W.doc
-------
Materials of Construction of the FGD System
The following is a description of the materials of construction for the Petersburg Units 1 and 2
retrofit FGD system.
The FGD system process design and material selection were based on the following equilibrium
FGD solution chloride concentrations (by weight):
Peak Excursion (10% of operating hours) 50,000 ppm
Average 30,000 ppm
Normal Minimum 10,000 ppm
Figures 1 and 2 present the materials of construction and the overall layout of the absorber area of
the FGD system.
The absorber inlet duct nozzles, included in the FGDS vendor scope, are solid alloy C-276, a
minimum of 1/4-inch thick including the wet/dry interface area. Unlined A-36 carbon steel is used
from that point back to the booster fan discharge.
The absorber inlet deflector plate and side shields are constructed of 3/16-inch solid alloy C-276.
The absorber reaction tanks are constructed of alloy C-276 clad carbon steel. The clad C-276
material is a minimum of 1/8-inch thick on a minimum of 1/4-inch thick carbon steel substrate.
Clad alloy material was hot mill bonded.
The absorbers are constructed of wallpaper alloy C-276 liner on carbon steel. The wallpaper alloy
liner material is a minimum of 1/16-inch thick on a minimum of 1/4-inch carbon steel substrate.
The absorber outlet duct is constructed of carbon steel lined with a borosilicate glass block lining
system, as provided by Elf Atochem Company.
Wallpaper alloy liner plug weld attachments were appropriately spaced apart, in order to minimize
vibration fatigue due to slurry impingement, and other potential mechanical vibration problems.
Welds were inspected to verify the correct filler metal was used and were leak tested.
Access doors are alloy C-276 materials of construction.
Absorber slurry nozzles are spiral reaction bonded silicon carbide tip producing a full cone spray
pattern with 150 pound fiberglass reinforced plastic (FRP) flanges.
IPL-S&W.doc
-------
Mist eliminators (ME) are constructed of FRP. The ME secondary supports are constructed of FRP
and the primary supports are constructed of rubber covered carbon steel. The ME wash water
nozzles, piping, and supports are constructed of FRP.
Absorber recycle slurry piping external to the absorber is rubber lined carbon steel. Chlorobutyl
rubber liners employ acid resistant fillers such as carbon black and silica flour. Note, acid-soluble
fillers such as calcium carbonate were avoided. Rubber has a cured hardness of at least Shore 60A.
Rubber liners are a minimum of 1/4 inch thick.
Absorber recycle piping which is internal to the absorber is erosion/corrosion resistant FRP. Both
the internal and external surfaces of the FRP have corrosion/erosion resistant coats. The absorber
recycle slurry piping which is internal to the absorber was designed to be capable of safely
supporting maintenance personnel and equipment as well as the weight of any scale or solids
without detrimental deformation or failure of the piping, nozzles, or support structure. Rubber-
covered carbon steel structural members were provided to adequately support the absorber recycle
slurry piping/nozzles.
Other slurry piping materials of construction are either chlorobutyl rubber lined carbon steel or
erosion/corrosion resistant FRP.
Forced oxidation air piping is carbon steel outside and erosion/corrosion resistant FRP inside the
absorber reaction tank.
Slurry tanks were lined with reinforced organic resin. Organic resin linings are either a
multifunctional bisphenol-A polyester or vinylester. The resin is reinforced with glass flakes.
Since the FGD system process design included the use of reclaim water in the limestone slurry
preparation system, the materials of construction of the limestone slurry preparation system were
designed to withstand the corrosive/erosive nature of the reclaim water.
The wetted parts of the top entry agitators are rubber lined. The agitator shaft material are either
carbon steel or high strength low alloy steel.
Side entry agitators wetted parts were constructed of corrosion/erosion resistant alloy.
Absorber recycle slurry pumps have rubber-lined casing and erosion/corrosion resistant alloy
impellers. All other slurry pumps have rubber lined casing and rubber lined impeller or rubber
lined casing and alloy impeller. Slurry pump impeller alloy material is resistant to chlorides to
50,000 ppmw and pH values as low as 3.
IPL-S&W.doc
-------
A corrosion/erosion resistant liner of flakeglass is employed in the floor trenches and sump pits.
The flue gas system for the Petersburg 1 and 2 FGD Project involves totally separate and open
(undampered) bypass ducts and flue gas liners for each unit. The Unit 2 bypass duct discharges to a
new bypass flue in the new chimney, while Unit 1 utilizes its existing ductwork and stack for
bypass.
As described earlier, the FGD booster fans have been sized to "draw down" a small percentage of
ambient air through the open and undampered bypass ducts and flues during normal operation.
While this separate bypass system eliminates the severe corrosion problems associated with mixing
unscrubbed gas and scrubbed gas in a common bypass flue, it does create the potential for localized
corrosion problems when mixing cool ambient air with hot unscrubbed flue gas.
The I.D. fan inlet ducts are unlined A-36 carbon steel.
The I.D. fan outlet to booster fan inlet ducts are 1/4 inch A-36 steel plate lined with borosilicate
block over its entire length. The block has an acid proof castible monolith over the bottom surface
for protection against damage by foot traffic and scaffolding.
Duct wall gusset plates for connections of turning vanes, trusses, etc, are stainless.
I.D. and booster fans have A-36 steel casings and A-514 steel wheels. The guillotine dampers at
the booster fan inlets have 317 LM blades and alloy C-276 for the internal seal assemblies and
internal wetted surface. A-36 carbon steel was used for the damper external frames.
The FGD system inlet duct up to the absorber inlet duct nozzle is unlined A-36 carbon steel plate.
The FGDS system outlet ducts extend from the absorber to the stack inlet breeching. These ducts
and the stack breechings are 1/4-inch A-36 carbon steel lined with field applied borosilicate block
over its entire length. The block on the ductwork floor has an acid proof castible monolith over the
surface for protection against damage by foot traffic and scaffolding. Ductwork internal stiffeners
are of C-276 alloy construction.
Expansion joints throughout the duct system consist of "fabric type" fluoroelastomer, "Viton - B"
material which is reinforced with both wire and fabric.
IPL-S&W.doc
-------
Both scrubbed gas flue liners and the Unit 2 bypass liner in the new chimney were constructed of
acid resistant brick and mortar (ARBM). The brick is fireclay, Type II chemical resistant
conforming to ASTM C980. The mortar is a potassium silicate material. The penetrations in the
scrubbed gas chimney liners are of alloy C-276 construction. The top of the scrubbed gas liners
were fitted with alloy C-276 chokes. The bypass liner is lined with alloy C-276 at the top. The
scrubbed gas liners are plumb and the bypass liner is tapered. The chimney is equipped with an
annulus pressurization system with air locks top and bottom. DynaFlow Systems Company
provided valuable assistance in the designing of the wet stack and the stack liquid collection
system. Reference 2 presents the engineering details concerning the new chimney at Petersburg
Units 1 and 2.
The bypass gas path for the Unit 1 FGD absorber utilizes the existing stack for Units 1 and 2. The
existing carbon steel chimney liner was lined with field applied borosilicate block lining system its
entire height.
References 3 and 4 present additional information concerning the materials of construction of the
Petersburg Units 1 and 2 FGD system.
Results of FGD System Performance Testing
The FGD systems have been extensively tested to document the performance, including
independent testing. The independent testing program was a joint team effort of all parties
including IPL, S&W, Radian, Mostardi-Platt, Ortech, Commercial Testing & Engineering
(CTE), GEESI, and others. Table I presents a summary of the results from the independent
testing program. In addition, IPL operates a chemistry laboratory that performs frequent testing
of the FGD process streams, including the gypsum product, in order to maintain a high quality
byproduct for its consumers and a reliable FGD system.
Table I values are the average of the results from the independent testing programs. The high
quality of the gypsum is the direct result of the team engineering, design, construction, operation,
and maintenance of the FGD systems.
As shown hi Table I, the gypsum oxidation fraction was over 99.9 percent on a molar basis,
moisture content less than 9.6 percent by weight as measured in the cake discharged immediately
from the filters and using ambient temperature filter wash water, the chloride content was less
than 5 parts per million on a dry weight basis, mass mean particle size by Micromeritics
Sedigraph instrument was greater than 35 microns, and the volume mean particle size by
Malvem laser diffraction instrument was greater than 58 microns. These physical-chemical
characteristics are indicative of a consistent high quality gypsum byproduct.
IPL-S&W.doc
-------
Of special interest in the testing program has been the low moisture content of the gypsum
product, for it has been found that the gypsum continues to dry as a result of conveyor handling
and temporary storage in the gypsum building and that the average moisture of the gypsum
placed in the transport trucks is consistently less than 9 percent by weight. Also, by using heated
wash water (instead of ambient temperature wash water) for washing the filter cake, the result is
an additional 10 percent decrease in the gypsum moisture content.
As shown in Table I, the stack liquid droplet carryover concentration was measured as less than
0.004 grains per actual cubic feet of scrubbed gas, which is considered to be a very low
concentration. The liquid carryover test method was based on the KOCH Phase Doppler Particle
Analyzer (PDPA) which measured the size and the velocity of individual droplets passing
through a "probe volume" formed by the intersection of two laser beams.
After the independent testing program, the FGD system was tested with the use of dibasic acid
(DBA) additive and the results indicated a slight increase in the overall sulfur dioxide removal
efficiency. A review of the test results showed that the presence of DBA in the scrubber liquid
had no negative effect on the composition of the FGD gypsum.
Use Of EPRI/NACE Programs In This FGD Project
Published documents of both the Electric Power Research Institute (EPRI) and National
Association of Corrosion Engineers (NACE) were used in the process design and selection of
materials of construction for the Petersburg Unit 1 and 2 FGD system.
The extensive research, supported by EPRI, in the development and optimization of FGD
technology was utilized in the design of the Unit 1 and 2 FGD systems. As one of many tools used
by the project team, the EPRI-developed FGDPRJSM, Version 1.1, computer program was used as
a tool to optimize internal absorber configuration. S&W developed the design inlet flue gas
properties and Radian developed recommended liquid to gas ratios, spray level spacings, nozzle
parameters, and other process design criteria. The FGDPRISM program was used in the bid
specification and comparison of bids phases of this project.
Results of an EPRI-/IPL-sponsored limestone particle size analyzer test at Petersburg Unit 3 were
referenced in the specification for the Unit 1 and 2 FGDS project. The Unit 1 and 2 limestone
slurry preparation system includes an on-line limestone grind analyzer to determine the percent by
weight limestone which passes through 325 mesh.
IPL-S&W.doc
-------
EPRJ Project RP2248-2, FGD Materials Failure Causes, provided materials performance
information supporting the materials selection decisions. This EPRI project has provided cause-of-
failure investigations of dozens of alloys, rubber liners, and organic resin coatings in diverse FGD
systems, and has greatly increased the understanding of the limitations and failure mechanisms of
different FGD construction materials.
NACE Task Group T-5F-5 of NACE Unit Committee T-5F on Corrosion Problems Associated
with Pollution Control, has published NACE Standard RP-0292-92, Installation of Thin Metallic
Wallpaper Lining in Air Pollution Control and Other Process Equipment, dated April 1992; this
document was used by the project and provided valuable information concerning lining design and
quality issues.
In addition to the aforementioned programs, other NACE-/EPRI-/EPA-/DOE-sponsored FGD
programs and reports were used to assist the project team.
Conclusions
Since commercial operation of the Petersburg Units 1 and 2 FGD system in June 1996, the
operation and maintenance of the FGD system has been performed well, with start-up related
problems being resolved. The reliability of the FGD system has been impressive and the project
success is considered the result of a well-planned team effort from all parties in the areas of design,
engineering, construction, start-up and testing, operations, and maintenance.
References
1. S. Wolsiffer and C.P. Wedig. "FGD By-Product Production at Petersburg Station", Ortech 5th
International Conference on FGDS and Synthetic Gypsum, Toronto, Canada (May 1997).
2. R.J. O'Donnell, P. McCord, D. Bird. "State of the Art Chimney for FGD Non-Reheat Service",
American Power Conference, Chicago, Illinois (April 1997).
3. C.P. Wedig, R.J. O'Donnell, C.K. Rutledge, S.R. Wolsiffer, P.P. Ellis. "Materials of
Construction for the FGD System at IPL Petersburg Generating Station Units 1 and 2.", NACE
Airpol 92 Seminar, Orlando, Florida (November 1992).
4. J.E. Martin, J.C. Khederian, R.J. O'Donnell, W.K. Watson. "Ductwork Materials of
Construction for FGD Systems, American Power Conference, Chicago, Illinois (April 1995).
IPL-S&W.doc 10
-------
TABLE I
SUMMARY RESULTS OF THE INDEPENDENT TESTING
PETERSBURG UNITS 1 AND 2 FGD SYSTEMS
FGD Parameter
Inlet gas flow rate
Inlet gas temperature
Sulfur dioxide inlet cone.
Chloride in absorber loop
DBA in absorber loop
Number of recycle pumps op.
Sulfur dioxide removal efficiency
System pressure drop
Absorber pressure drop
Stack liquid droplet carryover
Stack particulate
Oxidation ratio
Gypsum oxidation fraction
Reactive CaCO3 in limestone
Limestone consumption
Limestone stoichiometric ratio
Limestone slurry particle size
Limestone slurry solids cone.
DBA consumption
Makeup water consumption
Electrical power consumption
Measurement
million acfm
degree F
ppmv(dry)
ppmw
ppmw
Operating #
%
rwc
IWC
grains/acf
Ib/mmBtu
Ib air/lb SO2
sulfate molar
% by weight
tons/hour
molar
% passing 325 mesh
% by wt.
Ib/hour
gpm
kw
Unit 1
0.967
300
2250
850
0
4
96.6
5.0
3.6
0.0039
0.013
5.1
99.9
93
11.8
1.01
96
30
0
300
7,000
Unit 2
1.85
310
2160
1220
0
7
97.3
5.4
3.8
0.0032
0.028
4.3
99.9
93
20.9
1.01
96
33
0
660
11,000
IPL-S&W.doc
11
-------
TABLE I (continue)
SUMMARY RESULTS OF THE INDEPENDENT TESTING
PETERSBURG UNITS 1 AND 2 FGD SYSTEMS
Secondary hydroclone underflow solids
Gypsum cake moisture at filter discharge (using
ambient temperature filter wash water)
Gypsum bulk density, loose
Gypsum bulk density, packed
Gypsum CaSO4« 2H2O
Gypsum CaSO3» 1/2H2O
Gypsum CaCO3
Gypsum inerts
Gypsum sodium
Gypsum magnesium
Gypsum chloride
Gypsum ammonium
Gypsum organic carbon
Gypsum flyash (optical method)
Gypsum pH
Gypsum particle Malvern laser diffraction vol.
Mean diameter
Gypsum mean particle size (Sedigraph)
Gypsum Blaine surface area
Gypsum particle XxY axis (image analysis
microscope)
Gypsum particle aspect ratio of the X and Y axis
Special Test - Percent reduction of gypsum
moisture content in the filter discharge cake by
using hot water filter cake wash instead of using
ambient temperature wash water
% by weight
% by weight
Ib/ft3
Ib/ft3
dry wt, %
dry wt, %
dry wt, %
dry wt, %
dry , ppmw
dry, ppmw
dry, ppmw
dry, ppmw
dry, wt %
dry wt, %
pH
micron
micron
sq. cm/gram
sq. micron
ratio
percent less moisture
in the gypsum due to
hot water washing
54
8.6
70.6
86.4
95.1
<0.04
0.64
1.02
1.7
5.6
3.2
0.2
0.017
0.13
8.0
58
35
1020
2130
1.8
10
54
9.6
68.9
86.0
95.3
<0.03
0.59
1.1
2.9
8.3
1.7
0.4
0.012
0.13
7.9
66
37
1000
2210
1.9
10
IPL-S&W.doc
12
-------
Photograph 1 - Aerial Photograph of the Petersburg Staltoii During Construction of 'the Unit I and 2 FUD System.
IPl.-S&W.doc
-------
Photograph 2 - Petersburg Units 1 and 2 FGD Absorbers Looking Towards the Northwest.
Unit 1 on the Left and Unit 2 on the Right.
PL-S&W.dcc
-------
Photograph 3 - Units 1 and 2 New Stack, Looking From Grade Level Towards the South.
IPL-S&W.doc
15
-------
1'hotoj
Commercial Grade Gypsum Piles and Stockout Area.
Located at srade level of the Gypsum Storage Building.
1PL-S&W doc
-------
Booste r
Inlet Due
Fan
Absorber
Outlet
Ducts
Unit 2
Bypass
Duct
Unit 1
ID Fan
Booster
Inlet Due
Ou
Fan
t
Boros i I i cate
Block (TYP)
Unit 2
ID Fan Outlet/
Booster Fan
Inlet Duct
Figure 1
Side View of Petersburg Units 1 and 2
Retrofit FGD System Ductwork and New Stack.
IPL-S&W.doc
17
-------
Figure 2
Plan View of Petersburg Units 1 and 2
Retrofit FGD System Ductwork and Stack.
IPL-S&W.doc
18
-------
USE OF STACK EMISSIONS TO CONTROL DBA ENHANCED FGD
SYSTEMS BURNING FUEL HIGHLY VARIABLE IN SULFUR CONTENT
D.L. Hammack
Supervisor FGD System
Kentucky Utilities - Ghent Station
W. F. Gohara
Environmental Engineering
Power Generation Group
The Babcock & Wilcox Company
Abstract
pH is the most accepted method in limestone Wet Flue Gas Desulfurization (WFGD)
systems control. Systems using DBA to enhance SO2 removal operate smoothly in this
control mode, as long as a constant additive flow rate is fed to the tank. pH control
schemes become more difficult to handle when sudden and wide changes in the inlet sulfur
occur leading to changes in the FGD system sulfur loading. pH controls the flow of lime-
stone only and does not automatically monitor stack emissions. Operator intervention is
needed in order to maintain stack emissions in compliance if the inlet SO2 increases sud-
denly. A fully automated control system that monitors stack SO2 emissions and adjusts the
scrubber's operating conditions accordingly is described in this paper. This new control
concept was tested and implemented at the Ghent Generating Station. The use of the new
control scheme resulted in optimization of reagent consumption and significant reduction
in manual intervention by the FGD system operators. The new controls have been in opera-
tion since October, 1996.
Introduction
Seventy five percent of the operating costs of a power plant are spent on fuel purchases.
Any minor reduction in the unit fuel price translates into major annual savings in the
power plant's costs. To maintain flexibility, Ghent Station, as well as many other utilities,
are purchasing fuel on the spot market to realize savings in the fuel costs. Spot market
fuels are usually cheaper but they also vary widely in their composition and sulfur content
from one batch to the other. The FGD system at Ghent experienced variation in the sulfur
loading by as much as ±2 Ib per million Btu (MBtu) in periods as short as two hours.
-------
From 1994 until early 1996, Ghent Station burned fuel consistent in composition and sulfur
content. The system performed well in the automatic control mode and did not require any
extraordinary attention or operator intervention during this period. This situation changed
when coal purchased on the spot market was used and major swings in the fuel sulfur
occurred within a two hour period.
Adding to the above universal problem, as of January 1, 1997 the overall Kentucky Utilities
Ghent stack emissions were limited to 7599 tons of SO2 per year. To meet these stack emis-
sions limits, Ghent Generating Station had to use DiBasic Acid (DBA) to enhance the
system's SO2 removal efficiency from 90%, before January 1997, to as high as 96% particu-
larly when high sulfur fuel is used. The need to use DBA became more acute when the
potential for simultaneous changes in fuel sulfur and boiler load are considered. Changes
in the inlet sulfur required continuous changes in DBA to keep the system in compliance.
The organic acid additive buffered the slurry and instantly lowered the tank pH leaving the
control set point unsatisfied. The system responded by increasing the limestone flow.
Quick and frequent change in the FGD inlet sulfur loading required frequent changes in
either the pH set point, the DBA flow, or both to keep the system in compliance. Frequent
changes in the inlet sulfur by as much as ± 2 Ib SO2 per million Btu (MBtu) were experi-
enced regularly over a one to two hour period leading to corresponding change in the DBA
feed rate to the scrubber. The buffering effect of DBA depressed the tank pH and initiated
limestone flow to satisfy the controller set point. This situation led frequently to slurry
overflow from the tank unless the operator intervened and changed the pH set point. In
approximately two years of operation from 1994 until early 1996 the system operated
"without significant slurry overflow.
The wide and sudden change in fuel composition and the use of DBA to enhance the
system's SO, removal efficiency was beyond the flexibility of the original pH control
scheme. A change in the control philosophy to a system that tracks stack emissions and
adjusts the reagent flow to meet the required demand independent of the recirculation tank
pH seemed necessary.
pH Control
pH control philosophy is widely adopted in FGD systems because it is simple, reliable,
maintains tight control over limestone consumption, and reduces the potential for scale
formation particularly in the unoxidized zones of the scrubber. The control system is ca-
pable of adjusting to the operating pH set point to obtain the required stack emissions. The
control system is capable of adjusting changes in boiler load, limestone reactivity
and/or grind characteristics, and to a limited-extent, minor changes in fuel sulfur.
pH controlled FGD systems operate smoothly and effectively as long as the fuel sulfur is
reasonably constant, or the changes in the inlet sulfur are not abrupt and frequent. How-
ever, this control scheme relies on the operator's vigilance to monitor stack emissions and
to take corrective action to remain in compliance.
-------
In a pH controlled system, the operator adjusts the pH set point to obtain the required
stack emissions. Deviation in the measured tank pH, from the set point, by ± 0.05 pH unit
triggers a change in the limestone feed to restore the system's equilibrium.
The hierarchy of the pH control allows it to respond to small changes in the inlet SO2 concen-
tration. There is no link or a secondary loop that tracks the stack emissions status and
adjusts the system's operation accordingly. The only link between the Continuous Emis-
sions Monitor System (CEMS) and the absorber's SO, removal efficiency is the operator.
Effect of DBA on pH
While the addition of DBA improves SO, removal for a given limestone stoichiometry, the
DBA also depresses the tank pH and triggers demand for more limestone flow to satisfy the
set point. The buffering effect of the DBA and the magnitude of the pH depression depend
on the concentration of the DBA in the tank and can vary from 0.2 pH unit at 200 ppm to
more than 1 pH unit at DBA concentration greater than 3000 ppm.
If the fuel sulfur content increases suddenly, the set point may not provide the required
alkalinity to the system and stack emission violation may occur. Two options are available
to the FGD system operator at this point:
The first is to increase the recirculation tank pH so that more alkalinity is available in the
system. This option is limited by the threat of potential for scale formation in the absorp-
tion zone where adequate oxidation air may not be available.
The second option is to use DBA to enhance the system's SO2 removal. However, the addi-
tion of DBA causes a depression in the tank pH and triggers the flow of additional lime-
stone to satisfy the set point.
Successful application of pH control to an FGD system using DBA and variable inlet sulfur
requires a knowledge of the anticipated pH depression to bias the pH set point accordingly
and prevent wasteful slurry flow. This solution is feasible if the fuel sulfur content and the
DBA flow rate remain constant over long periods of time. On the other hand, continuous
changes in the fuel sulfur content result in frequent changes in the pH set point and the
DBA feed rate. This situation is manageable if the following two options are available to
the operator:
A. A relationship between DBA concentration in the system and the resulting depression
in the recirculation tank pH is known. This knowledge will allow the operator, or the
control system, to reset the pH set point accordingly. This option is practical if the
changes in the fuel sulfur are minor and occasional.
B. Constant monitoring of the recirculation tank DBA concentration to apply the proper
pH bias. This option requires frequent determination of DBA concentration in the
tank, every time the DBA flow changes in response to changes in stack emissions.
-------
The pH bias is then reset accordingly. This option is cumbersome and requires con-
tinuous availability of laboratory personnel to perform the DBA analysis.
These options may provide a solution to the effect of DBA on pH bias, however, they do not
provide a means for the control system to respond to changes in stack emissions. This link
is still an operator initiated function.
Recognizing the limitations of the pH control system under the prevailing operating condi-
tions, Kentucky Utilities identified the need for an ideal control scheme that links the
limestone flow to the changes in the inlet sulfur and to changes in stack emissions, and
dissociates the limestone feed from pH and the biasing effect of DBA. KU identified three
major goals for the new control scheme:
• The system should require minimum human intervention and follow the stack emissions.
• Improve the economy of the FGD system operation by reducing the use of DBA when
the conditions are suitable.
• Eliminate the dependence of limestone feed on the recirculation tank pH.
KU contacted The Babcock & Wilcox Company (B&W) to help develop a control scheme
that meets the above three requirements without interference with the daily operation of
the plant or the FGD system's availability.
Control Scheme Objectives
B&W conducted an engineering study to explore the available options and started by
identifying the following system objectives:
• Matches or exceeds the original pH control scheme.
• Field testing and implementation will not interrupt or interfere with the FGD system's
operation.
• Tracks and responds to fast and wide changes in the fuel sulfur content.
• Requires minimal operator intervention.
• Handles inlet SO2 loading ranging from 5.0 to 7.5 Ib/MBtu.
• Responds to maximum swings in fuel sulfur occurring in two hours period
• No stack emission violations over a 24 hour period.
• Compensates for changes in the boiler load.
-------
• Compensates for changes in limestone quality and grind.
• Maintains compliance without over scrubbing.
• Optimizes and maximizes the use of the low cost limestone reagent.
• Optimizes and reduces the use of the expensive DBA reagent.
• pH has no controlling function.
• pH is a process output used to monitor the system's chemistry.
• Fast acting and reliable.
Control Philosophy
After exploring the available options B&W outlined the system's operating philosophy as
follows:
• The limestone feed will be triggered by the inlet SO2 to satisfy the stack emission
limits based on a limestone/SO, relationship derived from the plant's historic data.
The baseline relationship is set to satisfy the full load boiler operation. If the boiler
load drops, the limestone flow is adjusted downward to optimize reagent consump-
tion and maintain stable chemistry conditions in the absorber. The boiler load correc-
tion is disabled below 60% to maintain adequate reserve alkalinity in the scrubber.
The excess residual alkalinity helps to control emissions during the anticipated in-
crease in the boiler load and possible simultaneous increase in sulfur loading during
the load raising periods. At equilibrium, the established limestone feed rate, the
reagent reactivity, and the resulting SO2 removal will establish the recirculation tank
equilibrium pH. The resulting pH acts as a process output indicative of the system's
chemistry status and the potential for scaling.
• DBA is used only, as necessary, as a polishing reagent and is stopped when the stack
emissions are in compliance. At full boiler load, the FGD system achieves 95% SO2
removal with inlet sulfur content less than 5.0 Ib SO2 per MBtu. DBA flow is triggered
when the inlet SO, exceeds 5.0 Ib per MBtu and the stack emission limits are ex-
ceeded. The primary control signal for the DBA flow is set according to a preset
schedule based on the inlet SO2 concentration and the secondary control signal from
the stack emission rates. The controller uses the SO2 hourly emission rate, reported
by the CEMS, to manipulate the FGD system variables as follows:
At the maximum emission limits of 1735 Ib SO2 per hour, DBA flow is continu-
ous at the maximum rate to achieve quick system recovery from emissions
infraction. The DBA flow continues until the stack emissions are brought under
control at 1700 Ib per hour.
-------
A normal operating range is set between 1650 and 1700 Ib SO, per hour. In this
range the limestone flow is set based on the system requirements dictated by the
fuel sulfur content and boiler load. DBA flow follows the established addition
schedule.
If the hourly stack emissions reach 1650 Ib SO, per hour, DBA flow is stopped
despite the addition cycle sequence or inlet SO2. The limestone flow is main-
tained unchanged as determined by the inlet SO2 and boiler load.
The low end of the emissions scale is 1625 Ib SO, per hour. When stack emissions
reach 1625 Ib SO, per hour, the limestone flow is reduced until the stack SO,
climbs up to 1650 Ib per hour again. Selection of this low limit was based on
review of the plant's actual operating data collected during the six weeks of
manual system trial.
Table 1 summarizes the control scheme set up and trigger points and the corresponding
system response at each level.
Table 1
Control Scheme Operating Range
Stack SO2 Emissions Operating pH Range
>1735 Ib/hr
1735 Ib/hr
1650 - 1700 Ib/hr
1625 - 1650 Ib/hr
<1625 Ib/hr
5.9 to 6.1
Limestone Flow
DBA Flow
Based on an inlet Continuous full
SO2and boiler load pump stroke
Per addition schedule
Interrupt cycling
and stop flow
Reduced Flow
No flow
At an inlet sulfur content less than 5.0 Ib per million Btu, DBA use is not allowed unless the
stack emissions exceed 1735 Ib per hour.
Table 2 shows the relationship between DBA addition schedule, the inlet SO,, and pump
stroke.
At a given inlet SO, the control system establishes and sustains the proper operating
sequence and cycle until the cycle is completed. Cycle time and pump strokes were set as a
function of the inlet SO, and based on the data collected during the manual test step. The
above DBA addition sequence may be overridden by any of the following conditions:
-------
• The CEMS shows an increase in the inlet SO2 concentration triggering a change in the
DBA schedule. Then, the pump "on" sequence is initiated using the new cycle param-
eters.
• Stack emissions exceed the maximum hourly stack emission value while all other
operating parameters, such as limestone flow and proper pH, are maintained and the
DBA addition sequence is in the "off" sequence.
On the other hand, if the inlet SO2 drops during an "off" sequence, the cycle is completed
without interruption and the next cycle starts using the new cycle parameters.
The DBA feed system design dictated the use of batch addition to maintain adequate
header pressure and to ensure equal flow to the operating absorbers. In the batch mode
operation, the minimum pump stroke at the high setting is 50% corresponding to a mini-
mum header pressure of 20 psig.
Table 2
DBA Addition Schedule
Inlet SO2,
Ib/Million Btu
<5.0
5.0 to 5.4
5.5 to 5.8
5.9 to 6.2
6.6 to 7.5
% Pump
High
0
50
50
75
100
Stroke
Low
0
15
24
35
40
Cycle Time, minutes
Total
0
60
60
60
60
High Low
0
30 30
35 25
35 25
40 20
NOTES
No DBA flow permitted
unless stack emission limits
are exceeded
DBA flow is permitted unless
stack emission limits are
met with limestone only and
the operating pH is between
5.9 and 6.1
Implementation Approach
B&W designed and implemented the new control philosophy in six weeks. During this
period, the control scheme concept was validated, tested, optimized, and integrated with-
out compromising the availability of the scrubber. The project was executed in the follow-
ing five steps:
1. Study of historical data
Study of the plant's historical and operating data collected over the past two years
and particularly during the periods when KU operations suffered the most problems.
-------
Data from this period provided the major input into the development of the prelimi-
nary relationships. This data was used by B&W to develop the control scheme concept
and specifications. The product of the study was reviewed by KU to ensure no inter-
ference with the other control loops currently in operation. The initial specifications
were broad and required further definition before the final installation of the new
control scheme.
The available plant historical data covered a narrower fuel sulfur content than that
intended for the new control scheme to handle. Some extrapolations beyond the
existing plant data base were necessary to develop a comprehensive system covering
the expected wide fuel sulfur content. The validity of the assumptions and extrapola-
tions were tested during the manual implementation step.
2. Manual testing step
The basic relationships established from the operating experience were tested in a
manual operating mode for six weeks. During this period actual operating data was
gathered and used to verify the conceptual assumptions and extrapolations. Manual
Operating mode was preferred by both B&W and KU operations for the following
reasons:
To allow the operators to maintain full control in case of an emergency.
To use the test period to familiarize the operators with the new control concept.
To obtain feedback from the operators about the system responsiveness
To allow implementation and testing of new ideas or remarks evolving from the
operators' response.
The FGD system operators were instructed to react and perform the same duties expected
from the automatic controller. The data collected within this period and the operators'
comments and suggestions were studied and analyzed.
3. Feedback step
In this step, the operators' comments and ideas were studied and incorporated in the
original concept.
4. Automation step
The final system configuration was automated and integrated with the other DCS
loops. After the automation, operator intervention in the daily operation of the FGD
system decreased considerably and the system has been operating steadily since
October 1996.
-------
5. Long term evaluation step
Evaluation of the system's performance to account for unforeseen long term operating
effects through the implementation of a quality control system. This step is a long
term objective and is monitored by KU operations with feedback to B&W of any
significant findings.
Description of the Control Scheme Loops and Function
The control scheme consists of four interacting loops. The interaction among the loops is
based on the system demand and in response to changes in fuel sulfur, boiler load, and
limestone reactivity/grind. The following describes the functions of these loops and their
response to changes in the stack SO2 emissions:
1. Fuel sulfur/limestone feed loop
The initial baseline limestone flow vs. inlet SO2 concentration relationship at pH 5.9
was established based on historical data spanning from the start up of the FGD sys-
tem in 1995. This relationship was further refined from the additional data collected
during the six weeks of manual testing to generate the pH 6.1 curve. Figure 1 shows
typical curves representing the relationship between the limestone feed rate and the
inlet SO2 at the two pH values.
A demand signal generated from the inlet SO2 content selects the proper baseline
limestone flow rate. The baseline limestone/inlet SO2 curves represent the properties
and the characteristics of the limestone used to establish this relationship. Deviation
of the limestone characteristics from that of the baseline is reflected in the tank's
equilibrium pH and automatically accounted for by making appropriate corrections to
the baseline curve.
2. Boiler load correction loop
After the limestone flow is established, based on the inlet SO2, the control system tests
for boiler load conditions and adjusts the hill load limestone flow rate to match the
demand required at the actual operating boiler load. Figure 2 shows a typical slurry
flow correction curve for the most common boiler loads.
3. pH loop
The pH of the scrubbing liquor is determined by equilibrium resulting from the
amount of soluble alkalinity available in the tank which is a function of the limestone
reactivity, the absorbed SO2, and the concentration of DBA in the tank. For the same
limestone flow rate, reactive or finer grind limestone provides a more alkaline scrub-
bing solution than less reactive or poorly ground limestone, hi this control scheme,
the scrubbing slurry pH is used as a benchmark to compare the reactivity or grind
-------
characteristics of the different limestones rather than being used to control the lime-
stone flow rate. An operating pH range of 5.9 to 6.2 was established as the equilib-
rium pH range based on historical data and the desire to reduce DBA consumption.
The upper limit of the pH range was established based on maintaining the highest
possible pH that reduces DBA consumption, and the potential of scale formation in
the unoxidized zones of the scrubber. The lower pH limit was established to ensure
that adequate alkalinity is available in the system when it responds to sudden in-
creases in the inlet SO,, to maximize the limestone scrubbing, and to maintain DBA
consumption at a minimum.
Effectiveness of the limestone reagent varies with two facts, limestone reactivity or
the limestone rate of dissolution in the scrubbing solution, and the stone grind or
fineness which is an indication of the ball mill performance. These two factors are
interdependent because finer limestone grind has a better chance to dissolve than the
coarse aggregates. The faster the stone dissolves, the higher is the system alkalinity at
a given flow rate, and the better the SO, removal.
At a given inlet SO, and proper limestone flow, a drop in the operating pH below the
lower limit is interpreted as a lower quality reagent and initiates a change in the
selected limestone feed curve. Likewise a rise in pH above the upper limit triggers a
corrective action in the opposite direction.
4. DBA feed loop
Figure 3 shows the effect of DBA on the SO, removal efficiency at Ghent Station as a
function of the inlet SO,. The figure shows that for limestone FGD systems not using
DBA additive, the change in SO2 removal efficiency is a strong function of both pH
and the inlet SO2 concentration. Meanwhile, DBA enhanced systems removal effi-
ciency is constant over the indicated inlet SO, range. However, at a given operating
pH, the DBA concentration is directly related to the inlet SO2. This plot and many
others of similar nature were essential to establish the operating pH range and the
required DBA flow rate at the corresponding sulfur loadings.
The DBA loop is inactive for inlet SO, concentrations less than 5.0 Ib/MBtu. Addition
of the acid is permissible but not mandated by the control system. If the inlet SO2 is
greater than 5.0 Ib/million Btu the stack emission limits are met with limestone only,
and the operating pH is within the established range. The FGD system will not re-
quire DBA addition despite the inlet SO,. On the other hand, if the emission limits are
violated with limestone only and the operating pH is at the upper limit, DBA addition
is started automatically according to a preset schedule which is solely dependent on
the inlet SO,.
-------
The Role of Quality Control and Continuous System Monitoring in Detecting Problems
The data evaluated during the manual operating phase was compiled, plotted, and com-
pared on quality control charts developed from the plant's historical data gathered during
the start up activities and the optimization test period. These quality control charts helped
in the detection of discrete changes happening such as the obstruction of the DBA feed line
and monitoring the ball mill performance. Table 3 shows some operating parameters estab-
lished from the quality control charts and the manual data gathering phase. Deviations
from these ranges warranted an investigation of the cause and the undertaking of proper
corrective action.
Table 3
Established Operating Ranges for the FGD System Two Main Reagents
Reagent Units Operating Range
Limestone Gallons per minute 0 to 240
DBA Gallons per hour 0 to 20
Figure 4 is an example of a daily quality control chart evaluating the stoichiometric ratio of
the FGD system during August. The plot shows a distinct drift pattern between the first
ten days and the rest of the month. The cause of this drift was traced to the change in
operating mode from the old pH control scheme to the new stack emission control scheme.
Subsequent data evaluation confirmed the formation of a new mean lower compatible with
the second part of the month.
Since the implementation of the new control scheme, the following parameters are moni-
tored and/or automatically calculated from the data acquisition system:
Reagent utilization ratio
Calculation of the reagent utilization ratio from reagent consumption and the corre-
sponding SO2 removed during a 24 hour period is more representative of the
system's reagent utilization ratio than a grab sample determined by analytical
method. Analytically determined system stoichiometry is influenced by transient
factors such as the sampling time relative to boiler load changes, the status of lime-
stone feed rate valve at the time of the sampling, the location of the slurry feed point
relative to the sampling location, and other inherent anomalies related to the sam-
pling procedure and tank mixing to name a few. A daily plot of the reagent utilization
ratio could detect subtle changes in the system performance and trigger immediate
corrective action.
-------
DBA consumption
DBA consumption is determined from a recording of the pump strokes, and the start
and stop times of the pump over a 24 hour period. The 24 hour data is summed and
an average hourly flow rate is compared to the expected performance. Daily calcula-
tion of DBA consumption provides a planning tool to plant operations to schedule
acid deliveries and storage. Table 4 shows the average acid consumption as a func-
Continuous monitoring of these two parameters during the six weeks of the manual testing
phase helped in assessing the ability and responsiveness of the control scheme to the
various operating conditions encountered. Any reason that caused an unusual change in
the consumption of limestone or DBA was investigated and corrected. The same quality
control system is continued now during the daily operation of the system pointing out the
presence of a problem or the development of an impending problem.
Table 4
DBA Consumption as a function of Inlet SO2
Inlet SO,, (Ib / million Btu DBA Usage, Gallons per hour
<5.0 0
5.0 - 5.5 20
5.6-6.1 23
6.2 - 6.8 25
6.9-7.5 28
Conclusions
The pH control problems encountered at the Kentucky Utilities Ghent station were solved
by applying a stack driven control system. Limestone is used as the primary reagent and
DBA is only added as needed. The limestone demand signal is initiated in response to the
inlet SO2 content and trimmed for boiler load. DBA feed rate is also based on inlet SO2 and
trimmed in response to changes in the stack emissions. The scrubber has been operating
successfully with this control scheme since October 1996.
Acknowledgment
The authors acknowledge and appreciate the efforts of the Kentucky Utilities FGD system
operators for their valuable contribution to the success of this project.
-------
240
200 —
160 —
120 —
Inlet SO2, Ib/MBtu
FIGURE 1 TYPICAL LIMESTONE SLURRY FLOW AS A FUNCTION OF INLET SO,
300
200 —
100 —
Inlet SO2, Ib/MBtu
FIGURE 2 BOILER LOAD CORRECTION FOR LIMESTONE SLURRY FLOW
-------
100
96 —
92 —
84
pH Range
5.5-5.6 No DBA
5.7-5.8 No DBA
5.9-6.1 No DBA
5.4-6.2 DBA
—i 1 , 1 , 1 1 1 -
3456 78
Inlet SO2, Ib/MBtu
FIGURE 3 S02 REMOVAL AS A FUNCTION OF INLET SO, WITH AND WITHOUT DBA
1.4 —
1.2 —
FIGURE 4 STATISTICALANALYSIS OFTHE FGD SYSTEM STOICHIOMETRIC RATIO OVER 30 DAYS
-------
A PERFORMANCE REVIEW OF UNIT 4 FGD SYSTEM
AT CINERGY'S GIBSON STATION
R. Richard
Superintendent, Materials / FGD
Cinergy's Gibson Generating Station
W.F. Gohara
Environmental Engineering
Power Generation Group
The Babcock & Wilcox Company
Abstract
The Gibson unit 4 flue gas desulfurization (FGD) system has been operating smoothly since
October 1994. The scrubber has responded favorably and consistently to the changes in the
Illinois basin high sulfur fuel that the plant has been firing. To improve system efficiency,
component reliability and to reduce maintenance cost, some changes have been made by
station operating and maintenance personnel to the original design.
Introduction
Early in the life of a project some of the project planners' original decisions and specifica-
tions are made based solely on economics with little regard to the operating aspects and the
commercial pressures imposed on the power plant operators to perform. As the project is
completed and the realities of life settle in, some changes are initiated by the plant opera-
tions and maintenance personnel to improve the operating aspects of the plant and correct
some of the decisions made based on economics alone. This paper discusses some of the
changes made at Gibson Station to the unit 4 FGD system which either fall under this
category or were made as pure improvements to facilitate the maintenance of the system.
The Gibson unit 4 FGD system is one of the early systems built to comply with the Clean
Air Act (CAA) Amendments and the Indiana State Implementation Plan (SIP). The system
provided a smooth start up during commissioning and performs with consistency and
predicted reliability during daily operation. The 97% SO2 removal efficiency exceeded the
-------
guaranteed 92% removal efficiency and the parasitic power consumption is about 1.5 % of
unit 4 output. The FGD system and all the supporting systems were built tc reflect the
existing state of the art in FGD technology and have operated reliably. However, the oper-
ating and maintenance personnel opted to implement some changes for convenience and
efficiency of operation and to reduce maintenance time and frequency of some compo-
nents. This paper presents background information on the unit 4 FGD system, a brief
discussion of the absorber's performance, and addresses some of the changes made to the
original design.
Background
To comply with the new environmental limits of the 1990 Clean Air Act (CAA) Amend-
ments and the Indiana State Implementation Plan (SIP), Cinergy installed a wet FGD sys-
tem on unit 4 at Gibson Station. The CAA required a 34% reduction in SO2 emissions dur-
ing phase I and a 37% reduction in phase II. To meet the SIP, Gibson Station had to phase in
emission rate reduction between 1991 and 1995.
The new unit 4 FGD system was selected to be an extension of Gibson Station personnel's
existing experience with the wet limestone scrubbing technology installed on unit 5. Like
unit 5, the unit 4 scrubber is an inhibited oxidation limestone system with sludge fixation.
Unit 4 has two 67% absorption tray scrubbers designed to remove 92% of the SO, incoming
with the flue gas. Each absorber is a counterflow limestone slurry scrubber contacting the
vertical gas flow stream (Figure 1). The scrubbing slurry is sprayed countercurrent into the
flue gas by four of the five levels of spray headers, each header level fed by a dedicated
pump. The absorption tray is located above the first spray header Two stages of mist
eliminator are located between the top spray header and the stack breach. The first stage is
a 3 pass chevron designed to use the gas and liquid inertia to remove the liquid droplets
from the gas stream. The first stage is located horizontally in the absorber allowing vertical
gas flow. The first stage mist eliminator is washed from above and below with a mixture of
fresh and process water from a series of 90 degree full cone nozzles. The second stage is a
horizontal flow 3 pass chevron located vertically in the outlet flue connecting the absorber
to the stack. The upstream face of the second stage mist eliminator is sprayed with a mix-
ture of fresh and process water through 90 degree full cone spray nozzles also. The reacted
slurry and the mist eliminator wash accumulate in the recirculation tank located at the
bottom section of the absorber. Fresh limestone slurry is added to the recirculation tank to
maintain the required system alkalinity and to satisfy the pH controller set point demand.
The slurry is maintained in suspension in the recirculation tank by means of four side-
entering agitators spaced around the perimeter of the recirculation tank.
The captured SO2 reacts with the limestone slurry to form hydrates of calcium sulfite and
calcium sulfate. An emulsified sulfur solution, consisting of 70% elemental sulfur, is added
to the fresh limestone slurry to inhibit oxidation and prevent excessive scaling. Slurry
blowdown from the scrubber is sent to a thickener and vacuum drum filters where it is
dewatered, fixated with fluidized bed boiler bed ash and unit 4 ash and sent to a land fill
for disposal.
-------
1 st Stage Mist Eliminator
Gas Inlet
Gas Outlet
2nd Stage Mist Eliminator
Absorber Spray Headers (typ)
Tray
FIGURE 1 GIBSON UNIT 4 TOWER DESIGN (ABSORBER MODULE)
Performance
The Absorber
The SO2 removal efficiency of the unit 4 FGD system can consistently be in the 97% range
exceeding the 92% guarantee point by 5% while the parasitic power consumption is about
1.5% of the unit's output. After extended periods of operation, the scrubber internals in-
cluding the tray, spray headers, spray nozzles, and the vertical flow mist eliminator have
-------
been clean and free of deposits. There has been a build up of friable material on the down-
stream side of the horizontal gas flow mist eliminator blades, however, the mist eliminator
build up has not affected the unit operation or forced a load reduction. A hard scale deposit
has also been found on the walls of the reaction tank located in the lower part of the tower.
pH Control
The heart of a limestone scrubbing system's performance is the available alkalinity. pH is
used as an indication of consumption of the alkaline scrubbing species in the FGD system.
A set pH point in the central control system detects deviations from the acceptable operat-
ing range and controls the fresh limestone feed.
Originally the pH electrodes were located in the suction pipes of recirculation pumps 2 and
3 of each absorber. Both pH meters sent signals to the control room, but the operator had
the choice to select either one to control the limestone feed. The benefit of this arrangement
was to maintain the pH electrode in a turbulent well-mixed area representative of the pH of
the scrubbing slurry reaching the headers while remaining in a fairly low velocity zone to
minimize the potential of physical damage to the electrode's sensitive parts. On the other
hand, the electrodes had to be placed in a position where they could be away from settled
solids. A problem with this arrangement was that it limited the operator's choice of which
sensor to use when either of the two pumps were out of service. Another limitation to this
arrangement was that the operator had to intervene to make the change as one pump was
taken out of service and the other put in.
In the new arrangement, the original pH probes were replaced with flow through type
probes and are now located in a common line at the sample sink. The pH line is also fitted
with a flush water connection to affect rinsing and cleaning of the probes while they are on
line without major disturbance to the FGD system operation. Another advantage to this
modification is that the operator does not have to swap probes when either pumps 2 and 3
are put into, or taken out of, service. However, the operators still have the option to select
either probes to control the process.
Absorber Level Transmitter
As the hot flue gas is quenched by the liquid falling off the inlet awning, water is evapo-
rated and carried out of the module. In addition, the captured SO2 forms solid hydrates
that consume some of the free water in the tank causing an increase in slurry density. The
density of the recirculated slurry is controlled by the addition of clarified recycle water to
the absorber and can be adjusted depending on load and fuel sulfur content. A signal from
the density meter controls the disposition of the make up water flow. The tank level detec-
tor also initiates the operation of the blowdown valves located on the discharge side of
pumps 2 and 3. To monitor the changes in the tank level, and to keep the process under
control, a tank level detection device is needed to prevent the unnecessary overflow of
unreacted slurry from the absorber, or to prevent the tank level from getting too low to
provide the proper net positive suction head for the recirculation pumps.
-------
The tank level was monitored and detected by a flange mounted ultrasonic level transmit-
ter which sent a signal to activate and control the position of the make up water flow and
the blowdown valves. On many occasions, the ultrasonic level detector failed to detect a
low tank level due to condensation or splashing on the protective glass window or due to
the presence of remnants of slurry on the walls of the measuring well. This erratic behavior
threatened the operation of the absorber and could have led to pumps tripping and inter-
ruption of the system's operation.
To measure the tank level reliably, the static pressure of the liquid in the tank was used as
an indication of the tank level. The height of the liquid in the tank was calculated by the
following equation:
Where: Tank Liquid Height = F*2.31 + H
SG
P liquid static pressure (psig)
2.31 feet of water column per psi
SG slurry specific gravity
H elevation of the pressure tap above the tank floor (feet)
A pressure transmitter was located on the line feeding the sample sink. Continuous flow of
slurry is maintained through this line to prevent the deposition of solids in the line which
would cause pluggage and a loss of the true level reading. A valve at the end of the line
controls the slurry flow to a small stream which allows the pressure to develop at the
transmitter connection. A water flush connection was also added to the root valve of the
pressure transmitter to allow on-line flushing and cleaning of the system. This system is
connected to the control room and provides the primary tank level indication while the
ultrasonic level transmitter is still in operation as a backup system.
Absorber Blowdown
Control of the recirculation tank level is achieved through operation of the blowdown
valves located on the discharge of pumps 2 and 3. Originally, when the tank level reached
the upper limit set by the operator, the controls selected and opened one of the blowdown
valves and sent the spent slurry to the absorber waste tank at a fixed rate. When the ab-
sorber tank liquid level reached the low limit, the absorber blowdown isolation valve
closed and a flush valve opened for a timed period to flush the line to the absorber waste
tank. Although this mode of operation was conducive to improving limestone utilization, it
did not provide steady flow of blowdown slurry to the thickener because it was dependent
on how fast the high tank level was reached which is a function of the absorber density and
boiler load. These high surges of absorber blowdown affected thickener clarity by over-
whelming the thickener for brief periods.
In the new mode of operation, the primary blowdown control is based on more frequent
operation of the blowdown valve based on a fixed time cycle trimmed for load and tank
level (i.e. as the load or tank level increases, the period between valve cycling decreases).
This approach allows for a more steady flow of slurry to the thickeners which has mini-
mized overloading and improved thickener clarity.
-------
Second Stage Mist Eliminator Deposits
The mist eliminators remove carryover mist from the flue gas by inertia] impaction. The
second stage mist eliminator, at unit 4, is a horizontal flow 3 pass chevron. The upstream
side of the second stage is washed by a four valve manifold that washes one quadrant of
the mist eliminator at a time. This strategy was adopted to minimize the instantaneous
demand for water and to be able to control the slurry density in the tank. It is imperative to
coordinate the wash cycles of the two stages of mist eliminator in order to maintain a good
water balance in the system. Heavy solids build up was observed on the second stage mist
eliminator more frequently due to materials drying on the blades, while the vertical flow
first stage performed without significant deposits build up.
To overcome this problem, the quantity of the water wash to the second stage was in-
creased by two fold and the wash frequency was also increased. In order to maintain the
water balance in the system, especially at lower boiler loads, the wash duration of the first
stage was reduced accordingly to maintain the proper solids content in the reaction tank.
The amount of wash for both stages is still adjusted based on boiler load as designed in the
original control and wash scheme.
Mist Eliminator Spray Flow Transmitters
The flow transmitters are located on the exterior of the towers. In the winter, the root lines
to the transmitters would freeze causing a loss of flow indication. The root lines have been
filled with antifreeze and have been heat traced. While this has helped the problem, it has
not completely eliminated it.
Mist Eliminator Wash System Valves
The original mist eliminator butterfly wash valves did not close properly and presented a
reliability problem during operation. These valves were replaced with AMRI valves which
have been more reliable.
Mist Eliminator Tie Rods
The original plastic mist eliminator tie rods broke frequently causing the blade spacers to
fall out, travel through the system, and plug the filter feed pumps to the waste fixation
system. Upon identification of this problem, Coastal changed the tie rods to C-276 material,
thus eliminating the rods breakage and release of the spacers into the absorber recirculation
tank and the downstream dewatering and fixation equipment.
Differential Pressure Indication
Differential pressure indications are used to determine mist eliminator pluggage or other
scaling in various parts of the tower. These readings have always fluctuated widely making
-------
it very difficult to draw any conclusions from the indications. Transmitters have been
calibrated and relocated to minimize sensing line length. While this has allowed some
improvement, the problem persists. It has been concluded that the design and location of
the pressure tap probes allows low pressure eddies to form at the tip of the probe creating
false differential pressure readings.
Agitators
The original FGD system design used rubber coated agitators, each with four flat blades
with bent tips. The agitators performed fairly well, however, the erosive nature of the
slurry, tramp material such as the mist eliminator spacers or hard pieces of scale, and the
forces exerted on the bent corners of the blade tips caused loss of the rubber coating at the
blade tips. The exposed carbon steel blade tips quickly corroded causing the loss of blade
tips and a resulting loss of agitator efficiency. Surprisingly, since all the blade tips corroded
at the same rapid rate, imbalance and vibration problems were not experienced. The rubber
coated blades have been replaced with duplex stainless steel blades to minimize the corro-
sive effect of the slurry and reduce the erosion of the blade tips. The shaft and hub are still
rubber coated.
There was considerable movement of the agitators as they were originally installed. This
was caused by the swirls and eddies in the reaction tank. Cross bracing was added to the
mounting frame to stiffen it and stabilize the agitators.
Another area of the agitators that needed attention was the gland seal water. Lake water
(service water) is used to supply the gland seal water. Due to the distance from the service
water pumps and the chlorination system, algae and pipe scale were blocking the flow of
the water to the shaft packing.
To address this problem, "Y" strainers were placed in the line to remove the debris from the
water and prolong the packing life. Differential pressure indicators were also located across
the strainers to detect flow obstruction and call for preventive maintenance.
Recirculation Pumps
The large slurry recirculation pumps had a minor amount of oil leaking past the oil seals.
GIW responded by providing a new design and replacing the oil seals on the slurry recircu-
lation pumps.
Another problem not directly related to the recirculation pumps, but interfering with their
performance, related to pressure bleeding from the hydraulic lines of some of the Clarkson
isolation valves located on the suction side of the pumps. When the valves were open,
pressure bleeding from the hydraulic lines allowed the valves to move away from their
respective limit switches and caused the pumps to trip as a result of the double light signal
being perceived by the control system as the valve not being fully open.
-------
The hydraulic lock module on each of these problem valves was replaced with new units.
There has been no reoccurrence of the problem.
Slurry Nozzles
The spray pattern of one of the spray nozzles on the under tray header fed from pump 3 on
absorber 4A impinged on the absorber wall more directly than the other nozzles. As a
result, this slurry impingement on the absorber wall caused thinning of the absorber wall
at this location and wore a hole through the shell. The hole was patched and the nozzle will
be redirected or eliminated from service during the next outage. This finding warrants
inspection of absorber 4B during the coming outage to correct the problem and reduce any
damage to the shell that may be in progress.
Ball Mill
The ball mill is a 50 TPH mill wet grinding " x 0" limestone into a 325 mesh product. While
the mill did a good job of producing the desired product, excessive oil leakage out of the
journal bearings was experienced during most mill start-ups. This caused a severe mess
running down the bearing pedestals and across the floor as well as wasting significant
quantities of oil.
Analysis indicated that when the heated oil entered the cooled journal areas, the viscosity
of the oil increased and the oil return lines weren't large enough to handle the thicker oil.
The seals on the journals were dust seals to keep dirt out and not oil seals to keep the oil in.
As the oil level rose in the bearing cavity, the oil would run past the seal and flow down the
pedestal. Once the journal area reached normal operating temperatures, the oil viscosity
and the oil level would return to normal.
Larger return lines were installed on one end of the mill (attached to the existing smaller
connections) and it has helped the situation. The fix would appear to be drilling larger
connection points into the bearing cavities and running larger return lines back to the oil
reservoir. This will be done during a future unit annual outage.
There was also a problem with significant amounts of rock being rejected to the floor out of
the trommel screen. A redesigned trommel screen with larger openings was installed. The
station personnel also installed a screen backwash header inside the hood covering the
trommel screen. Since that time, there has been very little rock discharged to the floor.
Flocculant Addition System
The thickener system was supplied with a typical flocculant addition system. Two aging
tanks were supplied so that a day's worth of diluted flocculant was mixed and aged in one
tank while the metering pumps drew suction from the other tank. The diluted flocculant
was pumped into the thickener system splitter box.
-------
A new system was installed that pumps directly from the flocculant container metering the
flocculant directly into a stream of water in a mixing pump. Additional water is then in-
jected after the mixing pump. By adjusting the rates of water flow before and after the
mixing pump, controlling the retention time in the piping, and introducing the diluted
flocculant directly into the thickener center well, effluent clarity has been maintained while
reducing flocculant usage nearly 50%.
Thickener Underflow Pumps
Each of the two thickeners was sized to handle 67% of the FGD system total spent slurry
design flow. Each thickener has two variable speed centrifugal thickener underflow pumps.
The thickener underflow pumps either transfer the thickened slurry to the filter feed tank
or recycle slurry back to the thickener solids bed or thickener distribution box.
The controls maintain constant thickener underflow density by comparing the density
signal from the thickener underflow to an operator adjustable density point. The resultant
error is applied to develop the required pump speed. The thickeners were designed on the
conservative side to account for any unknowns resulting from their application to an
inhibited oxidation system. As a result, the thickeners exceeded their rated performance
and one thickener can handle 100% of the FGD system total spent slurry design flow if the
underflow density is set at a higher than design value.
The resulting slurry density was too thick to pump at the rated capacity. The net pump
suction head was too low. The thickener underflow pipes were increased in size from 6 to 8
inches to accommodate the new slurry density and the pumps now operate at their rated
capacity.
Miscellaneous Changes
The packing on most of the small pumps were replaced by mechanical seals for reliability
and reduced maintenance.
The pneumatic air lines controlling the recirculation pump drain valves broke due to
vibration from the pumps. Lack of individual isolation valves for the recirculation pump
drain valves prevented the isolation of the air supply to a given pump without shutting off
the air to all the pumps which meant taking one absorber out of service. To remedy this
problem, individual isolation valves were installed on each recirculation pump drain valve
airline, thereby allowing the isolation of the broken line without interference with the
overall FGD system operation.
Design engineers have a difficult time visualizing the maintenance access to various pieces
of installed equipment. Once the equipment was installed, the maintenance personnel
immediately began to plan how they would remove various components when they even-
tually failed. It became obvious that in many locations there was no adequate means to
move components. Hundreds of thousands of dollars have been spent adding monorails,
-------
jib legs, and hoists at various locations to give the maintenance personnel adequate means
to rig and remove components and equipment.
Wrap-up
The Gibson unit 4 FGD system was put into service with a minimum of problems and has
operated reliably since that time. However, as with any system of this size, the operations
and maintenance personnel who deal with the system daily have found ways to improve
the efficiency and maintainability of the original design. This is an ongoing process that
will continue as long as the FGD system is in service.
-------
HENDERSON MUNICIPAL POWER & LIGHT
A LOW-COST PHASE I CLEAN AIR ACT RETROFIT
Lewis Benson
Dravo Lime Company
3600 One Oliver Plaza
Pittsburgh, PA 15222
Jeff Garner
Henderson Municipal Power & Light
100 Fifth Street
Henderson, KY 42420
James L. Murphy
Wheelabrator Air Pollution Control
441 Smithfield Street
Pittsburgh, PA 15222
Mike Thompson
Big Rivers Electric Corporation
201 Third St.
Henderson, KY 42420
Carl Weilert
Bums & McDonnell
9400 Ward Parkway
Kansas City, MO 64114
Abstract
The paper outlines the design and present results from two years of operation for one of the lowest
cost ($135/kW) Phase I wet scrubbing retrofits. Wet scrubbers using magnesium-enhanced lime
chemistry with controlled oxidation were retrofit to each of two 170 MW units at Henderson
Municipal Power and Light (HMPL) Station Two, located immediately next to Big Rivers Electric
Corporation's Green Station.
Several factors allowed very low cost. A single absorber, only 55 ft high, was retrofit to each unit.
By using magnesium-enhanced (Thiosorbic®) lime as reagent, the absorbers were designed to
achieve 95% SO, removal at a liquid-to-gas ratio of 30 gpm/1000 ACFM. SO2 removal and reagent
utilization results are presented. Research by EPRI and others on controlled oxidation and
crystallization was used in the design to produce calcium sulfite solids that settle quickly and filter
easily. This allowed the solids to be dewatered in the Green Station's existing thickeners and filters,
so that additional FGD solids dewatering equipment was not required. Data are presented that
demonstrate this improvement. Other operational experiences including periods of high chloride
concentration which required further optimization of process chemistry are also discussed.
-------
Introduction
Station Two is a two-unit coal-fired electrical generating plant located in Sebree, Kentucky which is
owned by Henderson Municipal Power and Light (HMPL) of Henderson, Kentucky, and operated
by Big Rivers Electric Corporation (BREC). Station Two shares a common site with the Green
Station which is owned and operated by BREC.
HMPL Station Two was targeted by the U. S. Environmental Protection Agency under Title IV of
the 1990 revision to the Clean Air Act Amendments to reduce SO2 emission levels. A magnesium-
enhanced lime process was chosen to scrub Station Two flue gas to meet SO2 emission
requirements.
Burns & McDonnell was chosen to perform process concept studies, prepare bid specifications, and
manage installation of the FGD system for the City of Henderson and Big Rivers. Wheelabrator Air
Pollution Control provided design and installation of absorbers and related FGD equipment. Dravo
Lime Company provided design parameters for the magnesium-enhanced lime process and supplies
Thiosorbic lime reagent for both the HMPL and Green FGD facilities.
To keep the project cost low. Burns & McDonnell developed a process concept for the HMPL FGD
system to share use of FGD process equipment available at the adjacent Green Station where extra
capacity was available. For example, lime unloading and storage capacity at Green is shared with
HMPL.
FGD solids thickeners and rotary drum vacuum filters needed for HMPL were available at Green,
but their original design capacity was only for the Green units. Since the actual capacity of the
dewatering equipment depends greatly on FGD solids characteristics, the dewatering properties of
the FGD solids had to be improved for this equipment to be shared.
Based on research and development by EPRI and Dravo Lime Company on ways to improve
dewatering properties of calcium sulfite solids, and on full-scale tests conducted by BREC at the
Green FGD system, the HMPL FGD system was specified to produce solids with good dewatering
characteristics.
To further control project costs, a single 100% absorber was specified for each HMPL unit. By
taking advantage of magnesium-enhanced lime process chemistry, he absorbers were small and
lower cost.
System Design
Two pulverized coal fired units each rated at a nominal 170 MW are in operation at HMPL Station
Two. The boilers were initially equipped with electrostatic precipitators for particulate control
which were not replaced during the FGD system retrofit. The FGD System has been in commercial
operation since May of 1995. Operating data for Unit 1 & 2 boilers is shown in Table 1.
-------
Table 1
Boiler Operating Conditions
Manufacturer
Type Firing
Unit Rating (MW)
Steam Flow (LB/HR)
Heat Input (MMBTU/HR)
UNIT1
Riley
PC
170
1,130,000
1,593
UNIT 2
Riley
PC
170
1,181,000
1,667
The HMPL FGD system includes one 100% booster fan and one 100% absorber for each unit. A
two-flue brick-lined chimney supplied with the FGD System discharges the scrubbed flue gas to the
atmosphere. A common reagent preparation system consisting of three - 50% capacity detention-
type slakers supplies lime slurry to both units. Dewatering of the absorber bleed slurry is
accomplished through the use of the existing Green Station thickeners and rotary drum vacuum
filters. The dewatered filter cake is then blended with flyash in pug mill mixers before being
transported to a landfill for disposal. The HMPL Station Two Wet FGD System design is based on
the gas flow information and performance requirements listed in Table 2.
Table 2
Specified System Requirements (Per Unit)
Inlet SO2 Loading, LB/MMBTU
Inlet SO2 Loading, LB/HR
Inlet Particulate Loading, LB/MMBTU
Inlet Flue Gas Flow, ACFM
Minimum SO2 Removal Efficiency, %
Minimum Particulate Removal, %
Maximum Absorber Velocity, ft/sec
Maximum Reagent Stoichiometry, Ib-moles Ca/lb-mole SO2
Minimum Absorber Reaction Tank Residence Time, minutes
7.59
12,653
0.21
745,400
95
50
10
1.03
3
Upon exiting the boilers, the flue gas flows through ductwork to dedicated electrostatic precipitators
where paniculate matter is removed. The flue gas is then directed to the booster fans and the
absorber inlet transition section. The original system ductwork and stack were not demolished
during the FGD System retrofit, and remain in place for flue gas to bypass the absorbers if
necessary. After exiting the absorbers, the scrubbed flue gas is directed through the outlet ductwork
to the two-flue chimney.
Fresh magnesium lime slurry is fed from the reagent preparation system to the absorber recycle tank
-------
and is controlled in response to pH in the recycle tank. The pH setpoint is maintained between 6
and 6.5. Suspended solids concentration in the recycle tank is controlled at 3 wt.%. Solids within
the recycle tank are kept in suspension by three side-mounted agitators.
Bleed slurry is pumped from the absorber recycle tank by variable speed bleed pumps to thickener
distribution boxes, which distribute the bleed slurry to four existing 125 ft. diameter thickeners.
The absorber bleed slurry is thickened to 30-37% solids by weight. The thickener underflow is then
pumped to rotary drum vacuum filters for dewatering to 45 - 50% solids. The filtrate is pumped to
thickener return water tanks for further use in the FGD system. A process flow diagram for the
HMPL Station Two absorber area, the Green Station thickener area, and the HMPL lime reagent
preparation area is shown on Figure 1.
Figure 1
FGD Process Flow Diagram for HMPL Station Two,
Shared Thickeners, and FIMPL Lime Preparation
Process Chemistry
Station Two utilizes magnesium-enhanced lime (Thiosorbic® lime) chemistry for SO2 removal.
Thiosorbic lime provided for the FGD System contains 3-6% magnesium oxide. In a magnesium
enhanced-lime system, soluble magnesium sulfite (MgSO3) in the absorber recycle slurry provides
high alkalinity that increases SO, removal capacity. The presence of high MgSO3 levels also
provides for low scaling potential.
-------
When absorber reaction tank slurry is sprayed into the absorber to contact the flue gas, dissolved
MgSO3 in the slurry reacts with absorbed S02. The MgSO3 immediately neutralizes acid produced
by absorbed SO2. By doing so, the pH of slurry contacting flue gas in the absorber spray zone and
on the tray stays high, which allows continued rapid absorption of SO2.
The reactions that take place in a magnesium enhanced lime system include absorption,
neutralization and regeneration. Within the absorption zone, sulfur dioxide is absorbed from the
gaseous into the liquid phase by the following reactions:
SO2 + H,O -» H2SO3 H2S03 •» H+ + HSO3"
Neutralization of the sulfurous acid formed in the above SO2 absorption reactions occurs through
the following reaction:
H+ + HSO3- + MgS03 •» Mg(HS03)2
In the absorber reaction tank, slaked magnesium-enhanced lime is added to precipitate the absorbed
SO2 as calcium sulfite and to regenerate MgSO3. Magnesium in the lime makes up for MgSO3 lost
with the filter cake when the calcium sulfite solids are dewatered.
Mg(HSO3)2 + Ca(OH)2 •*• CaSO3»l/2H2O + MgSO3 + 3/2H2O
Mg(HSO3)2 + Mg(OH)2 * 2MgS03 + 2H2O
The SO3= that builds up in solution gives the solution a high alkalinity. Alkalinity is usually
measured for magnesium-enhanced lime systems using an M-(for methyl orange endpoint)
alkalinity test. The pH end point of the M-alkalinity test is 4.2, which is the pH where all of the
SO3= present has been titrated to HSO3".
Oxidation of sulfite to sulfate reduces the alkalinity of the absorber recycle slurry, which in turn
affects SO2 removal capability and adversely affects solid settling and dewatering properties. High
oxidation can also lead to the formation of a hard gypsum scale which is difficult to remove from
the absorber internals. Magnesium-enhanced lime systems always operate sub-saturated for
gypsum, usually less than 10% of saturation. As a result, gypsum scaling is rare in these systems.
Laboratory tests performed daily during plant operation have demonstrated good settling
characteristics of the recycle tank solids, indicating controlled oxidation within the absorber recycle
tank.
Effect of Process Conditions on Solids Dewatering
Sulfite Oxidation. Lower levels of sulfite oxidation produce better dewatering of calcium sulfite
solids formed in magnesium-enhanced lime systems1' The calcium sulfite crystals that form in
magnesium-lime systems mostly consist of clusters of thin plates, caused by new plates growing
from the surface of others (surface nucleation), as seen in Figure 2. The clusters hold water in the
spaces between plates. When sulfite oxidation is reduced, the plates that form the clusters are
thicker, which reduces the amount of space and reduces the amount of water held.
-------
Figure 2
Calcium Sulfite Crystal from Magnesium-Enhanced Lime System
Control of Sulfite Oxidation Through Control of Magnesium Concentration. Full-scale
tests conducted at Big Rivers Green Station in 1990 showed that sulfite oxidation decreased when
absorber liquid magnesium concentration was decreased. Magnesium concentration was decreased
by using lime containing 3-4% MgO compared with lime with 5.5-6% MgO. Decreasing the
magnesium concentration reduced the liquid sulfite (SO3= and HSO3") concentration. Lower
oxidation with lower sulfite concentration might be expected, since oxidation rate depends in part
on dissolved sulfite concentration3. Reducing magnesium concentration also improves solids
dewatering4. Thickener underflow solids concentration increased to 37 wt.% and filter cake solids
to 49% during the test period. Laboratory thickening tests performed on absorber bleed samples
indicated that the solids would settle in a small thickener area, about 20 fWton (dry) solids per day.
Conversely, if dissolved sulfite concentration is too low, oxidation increases because of a large pH
drop in the absorber. When there is an adequate concentration of MgSO3 in the recycle slurry, it
buffers the pH inside the absorber so the pH drops very little, usually less than 1 pH unit. If the
sulfite concentration in recycle slurry is too low, so that the pH in the absorber drops to less than
-------
about 5, oxidation accelerates5, perhaps because the solubility of iron (Fe+3), a sulfite oxidation
catalyst, increases rapidly below pH 5.
Effect of Absorber Recycle Tank Suspended Solids Concentration. Pilot plant testing
and full-scale operating experience show that operation with low recycle slurry solids concentration
in the absorber recycle tank improves settling and thickening of calcium sulfite solids in
magnesium-enhanced lime systems.6 The absorbers at Green Station have operated successfully
with 3 wt.% suspended solids in the recycle tank since 1989.
The cluster-shaped calcium sulfite crystals are fragile, and operation at low solids concentration
reduces crystal breakage. Breakage is particularly severe as crystals pass through the recycle pump,
where shear is high. To maintain low suspended solids concentration, slurry is pumped to the
thickener at a higher flowrate. As a result, calcium sulfite crystals pass through the recycle pumps
fewer times before they flow to the thickener.
Also, mechanical shock caused by shear in the recycle pump causes precipitation of very small
crystals (secondary nucleation)7. Reducing suspended solids concentration appears to reduce the
amount of secondary nucleation in magnesium-lime systems. Low suspended solids concentration
in combination with small recycle tank slurry volume would be expected to increase supersaturation
(bulk supersaturation) of calcium sulfite in liquid in the recycle tank. This increase in
supersaturation appears to benefit crystal growth in magnesium-enhanced lime systems8.
Best dewatering operation occurs when low solids concentration is maintained along with moderate
to low sulfite oxidation levels.
Absorber Design
Each absorber in the HMPL Station Two system was designed to scrub flue gas from one unit.
Each Station Two absorber utilizes a Dual Flow Tray within the absorption zone to maximize mass
transfer of S02 from the gaseous to the liquid phase. The FGD system design specification for
HMPL Station Two allowed use of either an open spray tower absorber or a tray tower absorber. To
reduce the capital cost of the system, a Dual Flow Tray absorber design was proposed by
Wheelabrator. This reduced the FGD system cost by reducing the overall height of the absorbers.
By combining magnesium-enhanced lime chemistry with a Dual Flow Tray absorber, operating L/G
was reduced, which effectively reduced absorber recycle tank, recycle pump and recycle pipe sizing.
Flue gas flows vertically upward through perforations within the tray and restricts the downward
flow of liquid, causing a liquid froth to collect on the tray. Since the gas/liquid interface on the tray
is more efficient than that of an open spray tower, the operating L/G of the tray tower could be
reduced while maintaining a minimum 95 % SO2 removal.
With the use of a Dual Flow Tray instead of an open spray chamber absorber, the design L/G of the
tower was reduced from 45 to 30 GPM/1000 ACF as specified. This lower L/G resulted in a
reduction of slurry recirculation which allowed for the use of a single spray level instead of two. It
also allowed for use of a single recycle pump with a spare and a reduction in the operating liquid
level of the recycle tank from 12.5 to 10 ft. A comparison of the open spray chamber absorber and
-------
the Dual Flow Tray absorber is listed in Table 3. An arrangement drawing of the Dual Flow Tray
absorber design used for HMPL Station Two is shown in Figure 3.
Table 3
Open Spray Chamber Vs Dual Flow Tray Tower
Absorber Saturated Gas Flow, ACFM
Design L/G, GAL/ 1000 ACF
Total Recycle Rate, GPM
Recycle Pumps, (oper, spare)
Recycle Pump Capacity. GPM
Number Spray Levels
Overall Absorber Height, FT
Open Spray
590,500
45
26,600
3 (2+1)
13,300
3
67.5
Dual Flow Tray
590,500
30
17,800
2(1+1)
17,800
2
55
The absorber recycle tank is sized for four minutes retention time to allow for precipitation of the
calcium sulfite product to occur. The recycle tank liquid level allows for recycle pump NPSH
requirements to be met.
The suspended solids in the absorber recycle tank slurry are maintained at a nominal 3% by weight
in order to enhance growth of calcium sulfite crystals. Effective control of the slurry density is
maintained by controlling the absorber bleed flow via variable speed absorber bleed pumps, and by
the addition of reclaimed water makeup to the absorber recycle tank.
-------
, CD'
r^ i
o
CD
CO
CD
O
NORMAL
SLURRY
. LEVEL ,,
A 2r-o K
X ' ' \
r 36'-0 DIA.
MIST ELIMINATORS (2)
AND WASH HEADER (3)
RECYCLE HEADER
DUAL FLOW TRAY
Figure 3
Elevation of HMPL Absorbers (55 ft. total height)
By maintaining a low suspended solids content (3% wt.) and a short residence time in the absorber
recycle tank, growth of the calcium sulfite crystals is enhanced. The lower solids content provides
fewer crystals in the recycle tank which forces the calcium sulfite to precipitate out on these limited
sites, thereby resulting in larger calcium sulfite crystals.
The shorter solids residence time in the absorber recycle tank reduces the number of times a crystal
is pumped through the recycle system which could cause attrition of the crystals. This also results
in a larger mean particle size which enhances the settling and dewatering characteristics of the
produced calcium sulfite waste product.
Reagent Preparation
The Station Two reagent preparation area is located adjacent to the Green Station reagent
preparation/lime handling area. Station Two receives Thiosorbic lime from two of the existing four
Green Station lime silos. Lime handling equipment, including two lime silo discharge screw
conveyors and one lime transfer conveyor, was provided to transport reagent to the Station Two
lime feed distribution hopper.
Three 50% capacity detention-type lime slakers (Portec), each rated at 6.25 tons per hour, receive
lime from the distribution hopper. The lime is slaked with clarified river water to provide the slurry
required to operate the Station Two FGD systems. An additive hold tank and redundant additive
-------
feed pumps are supplied to provide slurry to the absorbers upon demand.
The Station Two additive hold tank is provided with flanged connections for transfer of slurry
to/from the Green Station additive hold tank if necessary.
Reagent Type
Thiosorbic lime used for FGD can be obtained with a magnesium content of either 3-4 wt.%
magnesium oxide (MgO) or 5-6 wt.% MgO. Thiosorbic lime is obtained at about the same cost per
ton as normal lime. As a result, the magnesium oxide that enhances SO, removal when using
Thiosorbic lime is obtained at no additional cost compared with normal lime.
Dewatering
Station Two was designed to utilize the available capacity of the dewatering facilities at the adjacent
BREC Green Station. Four existing 125 ft. diameter thickeners were operational at the Green
Station, and contained enough additional capacity for Station Two with the noted improvements in
solids dewatering properties. Three existing 12 ft. diameter by 20 ft. long rotary drum vacuum
filters located at the Green Station are used to process thickener underflow from the HMPL and
Green units. Filter cake solids content is 47-49 wt.% Filter cake is blended with about 0.7 Ibs of
flyash per pound of filter cake and dry quicklime fines to stabilize the mixture before it is hauled to
an on-site landfill.
Operating Data & Results
Performance testing was conducted on August 23 & 24, 1995. Table 4 lists the specified guarantee
values and the results for each parameter which was tested. Overall, the FGD System met
performance guarantees as required in the contract specification.
Table 4
HMPL Station Two Performance Test Results
SO, Removal Efficiency
Reagent Utilization
Particulate Removal
Moisture Carryover
UNITS
%
%
%
gr/acf
GUARANTEE
95
97
50
0.10
UNIT1
96
97
86
0.01
UNIT 2
97
97
72
0.01
SO2 Removal and Lime Utilization
Guaranteed SO, removal efficiency for Station 2 is 95%. The FGD system met this guarantee value
during performance testing and has consistently performed at this removal level during operation.
Figure 4 shows the monthly average SO2 removal and lime utilization for HMPL Unit 1 for 1996.
-------
100
% 95
94
93
92
91
90
-«—Sulfur Dioxide
Removal
-a-Reagent Utilization
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month in 1996
Figure 4
SO2 Removal and Lime Utilization for HMPL Unit 1
Reagent stoichiometry was guaranteed at 1.03 Ib-mole Ca(OH)2 / Ib-mole SO2 removed, equal to a
lime utilization of 97%. Performance test analysis on samples taken from the absorber bleed slurry
for each unit were analyzed by Dravo Lime Company. The test method measured wt.% CO2,
unreacted lime, and wt.% sulfur to obtain the stoichiometry results. Test results for each unit met
the guarantee value of 1.03, or 97% utilization.
Moisture Carryover
During performance testing, moisture carryover was measured with the KLD Labs AIMS droplet
measurement system. The highest moisture carryover value measured during testing was 0.01
gr/acf which was well below the guarantee value of 0.10 gr/acf. Consistent with the performance
test results, plant operators have not experienced problems with moisture carryover from the
absorbers in two years of operation.
Particulate Removal
A minimum particulate removal level of 50% across the FGD System was guaranteed for Station
Two. Particulate removal efficiency during performance testing averaged 85.72% for Unit 1 and
72.38% for Unit 2. These higher particle removal levels are another advantage of the Dual Flow
Tray absorber.
-------
Unit Area,
60
55
50
45
40
35
30
25
20
15
10
5
0
- Green Absorber 1B, !
4/18/957:30 f
-HMPLUnit 1 Absorber,;
4/27/95 8:50
20 25 30 35
Thickened Slurry Concentration, wt.%
40
Figure 5
Thickening Test Data from HMPL Start-up
Dewatering System Operation
Thickening characteristics of solids produced in the HMPL absorbers were monitored during start-
up of the HMPL Station Two FGD system. Thickening tests were performed each day on a slurry
sample taken from one of the HMPL units and on a sample taken from one of the Green units. For
the existing thickeners to accommodate the extra 350 MW-worth of solids from the HMPL units,
the absorber bleed solids had to settle and thicken in a thickener horizontal area less than 50 ft2 of
thickener area per ton of solids (dry basis). The thickening characteristics of the solids were
monitored using a 2-liter test recommended by Codan Associates. Figure 5 shows examples of test
data collected on HMPL and Green units samples. Samples showed required thickener areas (to
thicken to 30% solids) of about 10 ftVTPD for the HMPL absorbers and about 15 ftVTPD for the
Green absorbers. These required thickener areas were much less than the area of the existing
thickeners, so the tests confirmed that the thickeners would easily handle the additional HMPL
bleed solids.
High-Chloride Operation
As discussed earlier, the HMPL Station Two FGD system was designed to share solids dewatering
facilities with the Green Station FGD system. Most of the liquid in the FGD systems is contained in
the thickeners. As a result, the chemical composition of FGD liquor in the two systems tends to be
-------
nearly identical, and any operating condition in the Green units that effects FGD chemistry also
effects the HMPL FGD system.
The design of the combined HMPL and Green FGD systems assumed use of lime containing 3-4
wt.% magnesium oxide (MgO). Coal burned in the HMPL units was low chlorine coal. Coal
burned at Green usually contained 7.6 Ibs SO, per 106 Btu and less than 0.05 wt.% chlorine.
Because of the low chlorine content, FGD liquid chloride concentration in liquid in the Green FGD
system was moderate, about 3000 mg/1.
However, several months prior to start-up of the HMPL absorbers, Big Rivers began burning a
mixture of coals hi the Green units that included a substantial portion of lower sulfur, high-chlorine,
coal. Average coal sulfur decreased to about 6 Ibs SO2/106 Btu, and average coal chlorine increased
to 0.17 wt.%. As a result, FGD liquid chloride concentration gradually increased to near 10,000
mg/1 just prior to start-up.
This high chloride concentration has a negative effect on magnesium-lime chemistry. Chloride ties
up a portion of the magnesium in solution as MgCl2, which makes this portion ineffective for SO2
removal because it is unavailable to keep sulfite in solution. The portion of total magnesium not
tied up with chloride is called "effective magnesium". Effective magnesium is calculated as (total
Mg, mg/1) - (Cl, mg/l)/2.916.
To compensate for the increase in chloride concentration, a portion of higher magnesium lime,
containing 5.5-6 wt.% MgO, was shipped to the Green Station and stored in a separate silo. This
lime was added as needed to increase the effective magnesium concentration of the FGD liquid.
An effective magnesium concentration of 1500-2000 mg/1 in thickener overflow samples was
chosen as a control range. The effective magnesium concentration control range was selected based
on the dissolved sulfite concentration needed to achieve 95% SO2 removal and on the observed level
of sulfite oxidation. With 10000 mg/1 Cl", this required maintaining the total magnesium
concentration between 4900-5400 mg/1.
Since the volume of liquid in the thickeners is very large, the concentrations of magnesium and
chloride in the overflow changed very slowly. This made the overflow weir a good sample location
to monitor changes in magnesium and chloride concentrations in the FGD system. (In contrast,
absorber liquid concentrations can change rapidly in response to changes in fresh water addition.)
Initially, about one-fourth of the total lime used was the higher magnesium lime. The amount of
higher Mg lime used was adjusted based on week-to-week changes in magnesium and chloride
concentrations in the thickener overflow.
Table 5 shows typical HMPL absorber liquid chemistry during start-up and early operation of the
HMPL absorbers. The sulfite concentration that resulted from this level of effective magnesium
was high enough to produce desired SO2 removal efficiency in the HMPL and Green absorbers, but
low enough to reduce oxidation and produce good solids dewatering.
-------
Table 5
Typical FGD Liquid Chemistry in HMPL Station Two
During High-Chloride Period
Component and Units
pH
Total Mg, ppm
Chloride, ppm
Calcium, ppm
Effective Mg++, ppm
Total sulfites, ppm
Sulfate, ppm
M-alkalinity, as ppm CaCO3
Relative Gypsum Saturation
Value
6.7
4876
8248
17
2047
3547
4400
1661
0.03
During 1996, the average chlorine content of coal burned in the Green units decreased. As a result,
the chloride concentration of FGD liquor decreased from about 7000 ppm in January to 3000 ppm
in December, as seen in Figure 6. During this period, the fraction of lime with 5-6 % MgO was
gradually decreased to maintain 3000 ppm effective magnesium in HMPL absorber liquid.
Mg, ppm
Cl, ppm
Effective Mg, ppm
2000 - -
1000 ..
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month in 1996
Figure 6
Magnesium, Chloride and Effective Magnesium Concentrations in HMPL Unit 2 Absorber Liquid
-------
Summary
The retrofit of FGD to HMPL Station Two was a success. The absorbers installed on each unit
continue to perform as specified after 2 years of operation. The concept of sharing equipment with
the Green Station substantially reduced the HMPL system cost. Use of magnesium-enhanced lime
allowed the use of short absorbers. Operation at 3 wt.% solids in absorber slurry and optimization
of FGD liquid magnesium concentration improved solids dewatering characteristics. Solids
produced in the HMPL and Green units could be easily dewatered in the existing thickeners and
drum filters. An initial period of high-chloride concentration in FGD liquid was handled
successfully by increasing the liquid magnesium concentration.
Acknowledgements
The authors acknowledge the cooperation of Big Rivers Electric Corporation and Henderson
Municipal Power and Light in preparation of this paper.
References
1 R. Moser, '"Prospects for Inhibited Oxidation FGD Systems," presented at the 1993 SO, Control
Symposium, Boston, Mass. (August 1993).
2 J.H. Wilhelm, R. Moser, M. Stohs, B. Lani, D. Horn, "Results of Magnesium-Lime Liquor
Scrubbing Tests at the Miami Fort Station of Cincinnati Gas and Electric Company," presented at
the 1993 SO2 Control Symposium, Boston, Mass. (August 1993).
3 G.T. Rochelle, D.R. Owens, J.C.S. Chang, and T.G. Brna, "Thiosulfate as an Oxidation Inhibitor
in Flue Gas Desulfurization Process: A Review of R&D Results," presented at the Ninth
Symposium on Flue Gas Desulfurization, Cincinnati, Ohio (June 1985).
4 J.H. Wilhelm, R. Moser, M. Stohs, "Magnesium-Enhanced Lime FGD Reaction Tank Design
Tests at EPRI's High-Sulfur Test Center," presented at the 1991 SO2 Control Symposium, New
Orleans, Louisiana. (December 1991).
5 O. Hargrove, G.P. Behrens, and W.E. Corbett Studies of Flue Gas Desulfurization at Louisville
Gas and Electric's Paddy Run Station.. EPA-600/S7-82-032, October, 1982.
6 F. Baczek, L.B. Benson, R.M. Golightley, and J. Wilhelm, "Effect of Wet Lime FGD Operating
Conditions on Improving Particle Size and Dewatering of Solids,'' presented at the Tenth
Symposium on Flue Gas Desulfurization, Atlanta, Georgia (December 1986).
7 A. Randolph, et. al. "Improving Sludge Dewatering in Wet Lime FGD Systems," EPA/EPRI First
Combined FGD and Dry SO2 Symposium, St. Louis, Missouri (October 1988).
8 A. Randolph et. al.
-------
FINAL RESULTS
for the
EPRI-DOE-SCS Chiyoda Thoroughbred CT-121 Clean Coal Project
at
Georgia Power's Plant Yates
David P. Burford
Project Manager
SOUTHERN^
COMPANY
Erurgy u> Serve Your WarLf'
42 Inverness Parkway / Suite 340
Birmingham, Alabama 35242
ABSTRACT
The EPRI-DOE-SCS Yates Project tested the operational limits of Chiyoda's CT-121 wet
limestone SC>2 scrubbing system at Georgia Power's Plant Yates. Although the original test plan
called for a rather straightforward assessment, the CT-121 system proved robust, so it was tested
at widely varying conditions. Fuels ranged from 1.5 to 4.3 % sulfur, various limestone sources
and grind sizes were used, particulate removal and air toxics performance were measured and
gypsum soil amendment experimentation was conducted. In all cases, the CT-121 system with
Chiyoda's Jet Bubbling Reactor (JBR), gave encouraging results with predictably high SCh and
particulate removals at all conditions with high reliability. Closed loop operations called for the
extensive application of corrosion impervious, fiberglass reinforced plastics that was successful.
Gypsum proved to be significant as a soil enhancement and was granted a plant food license by
the State of Georgia. So far, the Yates Project has received four awards from industry and
environmental groups for its performance including PowerPlant of the Year in 1994 from Power
magazine.
Presented at the
EPRI - DOE - EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
Washington, DC
August 25-29, 1997
-------
Final Results David P. Burford
CfflYODA THOROUGHBRED CT-121 CLEAN COAL PROJECT Southern Company Services, Inc.
INTRODUCTION
The US Department of Energy's Clean Coal Technology effort was an opportunity to study
experimental uses of coal-to-energy processes that was cofunded by the US government. As one of
the largest electric utilities in the US with over 32,000 MW of generation and a significant user of coal
at over 50 million short tons per year, the Southern Company conducted four such projects in
partnership with the US DOE and the Electric Power Research Institute (EPRI). The largest of these
four Southern Company Clean Coal projects was an environmental control demonstration of a flue
gas desulfurization (FGD) removal system or 'scrubber', retrofit to a coal-fired boiler at Georgia
Power's Plant Yates near Atlanta.
The demonstration at Georgia Power's Plant Yates involved the retrofit of a Chiyoda Corporation
CT-121 wet-limestone scrubber system to an existing 100 MW pulverized coal-fired boiler. The
principal difference between the CT-121 process and more common spray tower-type FGD systems is
Chiyoda's use of a single process vessel, the patented Jet Bubbling Reactor® (JBR), in place of the
usual spray tower/reaction tank/thickener arrangement. Initial startup of the process at Plant Yates
occurred in October 1992 and the demonstration project was completed in December 1994. Process
operation continues today with the CT-121 scrubber as an integral part of the site's environmental
compliance plan.
Several of the latest evaluations that comprised the CT-121 demonstration project are discussed in this
paper. In the last trimester of testing, the CT-121 process was operated under moderate-ash inlet
loading conditions while process reliability and availability were evaluated. Exceptional paniculate
removal efficiencies were measured under these moderate-particulate loading conditions, which was
consistent with efficiencies observed in earlier measurements under both high- and low-particulate
loading conditions.
Parametric testing for SO2 removal was also conducted under moderate-ash loading conditions while
burning both high- and low-sulfur coals. The data gathered were regressed and multi-variable
regression models were developed to provide an accurate prediction of the scrubber's SO2 removal
efficiency under the most likely future operating conditions. As part of the moderate-particulate
removal evaluation, limited air toxics measurements were performed on two occasions. The purpose
of this testing was to evaluate air toxics removal across the CT-121 under elevated ash loading
conditions as well as to validate the findings of an earlier air toxics testing effort in June of 1993 '
A brief discussion of the properties of the gypsum stack following one year of dormancy is also
included in this paper. An analysis of the chloride content showed that chloride levels in the gypsum
Page 2
EPRJ -DOE-EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
Augus: 25-29. J997
-------
Final Results David P. Burford
CHIYODA THOROUGHBRED CT-121 CLEAN COAL PROJECT Southern Company Services, Inc.
declined over time from natural exposure to weather without any specific action by the project team.
This finding increases the possible uses of the unwashed gypsum produced by this CT-121 process.
An indicator of public acceptance was the granting of a Plant Food License for the non-ash gypsum
from Plant Yates, by the State of Georgia's Department of Agriculture in October of 1996.
In general, the Yates CT-121 process performed well, exhibiting excellent SCh removal efficiency,
paniculate removal and consistent reliability. In addition to these successes, several possible
process improvements were identified during the demonstration that could improve future designs
of an already superior process.
FACILITY AND OPERATING DESCRIPTION
The Yates plant site is comprised of seven coal-fired boilers with a total rated capacity of 1,250 MW.
All of the flue gas from Unit 1 is treated by the CT-121 wet FGD process with no provision for flue
gas bypass. During the low flyash phase of parametric testing in 1992 and 1993, the existing
electrostatic precipitator (ESP) for Unit 1 was fully energized for paniculate control with a design
efficiency of 98%. In March, 1994, the ESP was fully deenergjzed at the start of high-particulate
parametric testing and then partially re-energized to a target efficiency of 90% between June 1994 and
November 1994.
The central feature of the CT-121 process is Chiyoda's unique absorber, the Jet Bubbling Reactor®
(JBR), which combines simultaneous chemical reactions of limestone dissolution, SOj absorption /
neutralization, sulfite oxidation, gypsum precipitation and gypsum crystal growth, together in one
vessel. A cut-away view of the JBR is illustrated in Figure la. Large, easily dewatered gypsum
crystals are consistently produced in the CT-121 process since much of the undesirable crystal attrition
associated with the large centrifugal pumps and secondary nucleation in conventional FGD systems is
eliminated.
In the Yates installation (Figure la), the flue gas enters the scrubber system's JBR inlet gas cooling
section, down-stream of the boiler's induced draft fan. Here the flue gas is cooled and saturated with a
mixture of recycled pond water and JBR slurry. From the gas cooling section, the flue gas enters an
enclosed plenum chamber in the JBR formed by the upper deck plate and lower deck plate. Sparger
tube openings in the floor of the inlet plenum force the inlet flue gas downward so that it must 'bubble'
from underneath the level of the slurry-filled reservoir in the jet bubbling zone of the JBR as shown in
Figure Ib. After bubbling through the limestone slurry where all the simultaneous reactions occur, the
cleaned gas flows upward through large gas riser tubes that bypass the inlet plenum. Entrained liquor
in the cleaned gas disengages in the outlet plenum above the upper deck plate due to a drastic velocity
Page 3
EPRf - DOE - EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
August 25-29. 1997
-------
Final Results David P. Burford
CfflYODA THOROUGHBRED CT-121 CLEAN COAL PROJECT Southern Company Services, Inc.
reduction while the cleaned gas passes to a 2-stage mist eliminator and is then discharged through an
FRP chimney without reheat.
The underflow slurry from the JBR is pumped intermittently to a gypsum slurry transfer tank for JBR
slurry level and density control. The slurry is then diluted for high velocity pumping to a Hypalon®-
lined holding area or 'stacking area' for gravity dewatering by sedimentation and eventual storage.
Gypsum 'vertical stacking' is a disposal technique that uses its own product for containment. First,
gypsum slurry fills a diked area for gravity sedimentation .... over time, this area fills with settled solids.
The filled area is then partially excavated by dragging settled gypsum onto the existing dikes, which
increases the height of the new 'stack'. Fresh slurry is then diverted to fill this newly created diked area
where solid gypsum again settles out by gravity. The repetitive cycle of slurry addition / sedimentation
/ excavation / raising of perimeter dikes / fresh slurry addition continues on a regular basis as new
impoundment perimeters grow vertically in lifts, creating a pyramid. Process water is clarified by this
rapid gypsum sedimentation, naturally decanted and stored in a recycle surge pond to be returned to
the CT-121 process. There is no blowdown, water treatment or discharge from the CT-121 process at
Plant Yates.
During normal operation of the CT-121 FGD system, the amount of SO? removed from the flue gas is
controlled by varying the JBR pressure drop (AP) and slurry pH. This control variable responds more
quickly to changing conditions and is the preferential control variable since changing AP is done by
raising or lowering the set-point for the JBR slurry reservoir level. Higher liquid levels in the JBR
reservoir (i.e.; deeper sparging action) correspond to increasing SQz removal as the contact time
between the incoming flue gas and the scrubbing slurry is increased. The pH can also be varied to
affect SO2 removal although it is slower to respond due to the large slurry inventory in the JBR, with
higher pH's resulting in increased removal efficiency. Boiler load and inlet flue gas 863 concentrations
also affect CT-121 's SO? removal efficiency, but these are independent operational variables.
One of the most unique aspects of the CT-121 installation at Plant Yates is the extensive use of
fiberglass reinforced plastics (FRP) to avoid the corrosion damage associated with most other FGD
systems. Two of the main FRP vessels (the JBR and the limestone slurry storage tank) were
constructed on site, since their large size precluded shipment. The JBR inlet transition duct, where the
flue gas is cooled, is also made completely of FRP. Although the inlet transition was discovered to be
an area susceptible to erosion during high ash testing, homogeneous applique filler materials
(Duromar® and Duromix®), now provide robust protection to exposed FRP surfaces at Plant Yates.
A distinct advantage of the FRP construction at Yates was that it eliminated the need for a flue gas
prescrubber, traditionally included in flue gas scrubber systems to remove chlorides that cause
significant corrosion in alloys (fiberglass is unaffected by inorganic acid attack found in SO2 scrubbing
and rising chloride levels when operating FGD in a closed loop).
Page 4
EPRI -DOE- EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
August 25-29, 1997
-------
Final Results David P. Burford
CHIYODA THOROUGHBRED CT-121 CLEAN COAL PROJECT Southern Company Services, Inc.
PROJECT OBJECTIVES
To evaluate the effectiveness of the Yates CT-121 design advances, the following test objectives were
established for the two year demonstration program:
• Demonstrate long-term reliable operation of the CT-121 FGD system;
• Evaluate particulate removal efficiency of the JBR and
evaluate system operations at normal and elevated particulate loadings;
• Correlate the effects of pH and JBR pressure drop (AP) on system performance:
• Correlate the effect of limestone grind on system performance;
• Evaluate the impact of boiler load on system performance;
• Evaluate the effects of alternate fuels and limestones on system performance;
• Evaluate equipment performance and construction material reliability; and
• Monitor solids properties, gypsum stack operation and possible impacts of the
gypsum stack on local ground water.
Particulate and air toxics removal testing were also conducted during testing. The data from the
parametric portion of this test period was regressed to develop predictive performance models for the
conditions at which the testing was conducted, since these conditions are the most likely scenario for
post-demonstration operations.
RESULTS
The CT-121 scrubber at Plant Yates continued to prove itself a reliable and cost-effective technology
for retrofit or new-unit installation. CT-121 at Plant Yates exhibited excellent availability, maintained
greater than 97% limestone utilization, and demonstrated the ability to exceed 98% SO2 removal
efficiency with high sulfur coals, while at maximum boiler load as illustrated in Figure 2a.
Operating Statistics
The cumulative duration of the Yates CT-121 demonstration was 27 months (approx 19,000 hours).
The low-particulate test phase consisted of 11,750 hours (including shakedown), during which time the
scrubber was operated for 8,600 hours. The remaining 7,250 hours of the demonstration included
5,210 hours of operation at elevated particulate loading. Complete operating statistics for the entire
demonstration project are detailed in Table 1. The "high-ash" test period actually consisted of a period
in which the ESP was completely de-energized; the moderate-ash loading tests were conducted with
the ESP partially de-energized to simulate a more realistic scenario: a CT-121 retrofit to a boiler with a
marginally performing particulate collection device. The moderate-ash loading condition resulted in
slightly belter process availability than did the high-ash loading condition.
PageS
EPRI - DOE - EPA Combined Utility Air Potiatata Control Symposium (The Mega Symposium)
August 21-29,1997
-------
Final Results
CfflYODA THOROUGHBRED CT-121 CLEAN COAL PROJECT
David P. Burford
Southern Company Services, Inc.
Table 1
Summary of Operating Statistics
Total Hours in Test Period
Scrubber Available Hours
Scrubber Operating Hours
Scrubber Called Upon
Reliability1
Availability2
Utilization
Low-Ash
Test
Phase
11,750
11,430
8,600
8,800
98%
97%
73%
High-Ash
Test Phase
7,250
6,910
5,210
5,490
95%
95%
72%
Project
Duration
(Cumulative)
19,000
18,340
13,810
14,290
96%
97%
75%
1. Reliability = Hours scrubber operated divided by the hours called upon to operate.
2. Availability = Hours scrubber available divided by the total hours in the period.
3. Utilization = Hours scrubber operated divided by the total hours in the period.
Effect of Inlet 502 Concentration on Removal Performance
The SQz removal efficiency of the CT-121 scrubber was measured under five inlet SOj concentration
ranges. The coal burned by Unit 1 for a majority of the testing was the design basis fuel; a blend of
Illinois No.5 and No.6 bituminous coal that averaged 2.4% sulfur (as burned). However, there were
excursions of up to 3% sulfur in this fuel. During the Low-Particulate Alternate Coal Test block, a
4.3% sulfur bituminous coal was burned; while a 3.8% sulfur coal was burned for the High-Particulate
Alternate Coal Test block. The last test, the High-Particulate Alternate Limestone Test, coincided with
Plant Yates' transition to a low sulfur coal of approximately 1.2% sulfur.
The effect of inlet SQz concentration on SQz removal efficiency can be significant. Figure 2b illustrates
the decrease in SO2 removal as inlet SO2 concentration increased for the coal sources evaluated.
Conversely, performance of the scrubber was outstanding during the moderate ash / low-sulfur coal
bum. The test data from 1000 ppm (inlet SOa concentration) operations also indicate that SC>2 removal
efficiency did not decline at slightly lower pHs, needed to preclude aluminum-fluoride blinding when
significant amounts of ash are collected in the JBR reservoir.
The evaluation of five different inlet SC>2 concentrations demonstrates the flexibility of the CT-121
process as well as it's exceptional SO2 removal capability, even when burning fuels with a very high
sulfur content. This is impressive considering that the maximum designed sulfur content for the
Pages
EPRI - DOE - EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
August 25-29. 1997
-------
Final Results
CfflYODA THOROUGHBRED CT-121 CLEAN COAL PROJECT
David V. Burford
Southern Company Services, Inc.
demonstration unit was only 3.0%, and that this limit was exceeded by 43% during one test period.
Other test data shows that even higher SO2 removal efficiencies are achievable at higher pH values.
Particulate Removal Efficiency
Because of the torturous path taken by the flue gas during SO2 scrubbing in the JBR, it was anticipated
that significant particulate removal might occur in CT-121 systems. Consequently, the CT-121
process's ability to remove flyash was evaluated several times throughout the Yates demonstration.
Particulate loading measurements were made at the inlet and outlet of the scrubber under three
different conditions of inlet mass loading as summarized in Table 2. The discussion here will focus on
the particulate removal capabilities of the scrubber under the moderate-ash loading conditions.
Table 2
ESP Configuration during Particulate Testing
Condition
1
2
3
ESP
Energization
Full
Partial
Off
ESP
Collection Rate
High
Moderate
Low
JBR Inlet Mass Loading
(Ib/MMBTU)
Lmv (0.02-0.10)
Moderate (0.20-0.50)
High (5.00-5.50)
See
Figure
-
3a
3b
Measurements of particulate removal across the JBR were made near the minimum and maximum
nominal boiler loads (50 and 100 MW) and at low and high JBR AP settings (10 and 18 inches WC).
As shown in Table 3, at all tested inlet particulate loadings, boiler loads, and JBR pressure drops the
JBR exhibited excellent particulate removal efficiency, ranging from 97.7% to 99.3%.
Although the outlet particulate loading varied from 0.005 to 0.029 Ib/MMBTU, analytical results
indicate that from 20 to 80 percent of outlet particulate is sulfate (SO4). Based on the calcium
analyses performed on the same material, it is believed that the measured sulfate originated from
gypsum carryover and acid mist carryover, so it is scrubber-generated. This finding reduces the
estimate of actual ash mass loading at the outlet of the scrubber (actual fugitive emissions) to
approximately 70% of the amount captured, measured and recorded during outlet testing.
Page?
EPRJ - DOE ~ EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
August 25-29, 1997
-------
Final Results
CHtYODA THOROUGHBRED CT-121 CLEAN COAL PROJECT
David P. Burford
Southern Company Services, Inc.
Particulate Removal Testing
Table 3
Summary of Results (Moderate Ash Loading)
Test
I.D.
AL2-1
AL2-2
AL2-3
AL2-4
Approximate
ESP
Efficiency
(%)
90
90
90
90
JBRAP
(in. WC)
18
10
18
10
Boiler
Load
(MW)
100
100
50
50
JBR
Inlet Mass
Loading
(Ib/MMBTU)
1.288
1.392
0.325
0.303
JBR
Outlet Mass
LoadinglrZ
(Ib/MMBTU)
0.029
0.010
0.005
0.006
JBR
Removal
Efficiency
(%)
97.7
99.3
98.5
98.0
Federal U.S. NSPS is 0.03 Ib/MMBTU for units for which construction began after 9/18/78
2 Plant Yates Unit 1 's permitted emission limit for existing units is 0.241b/MMBTU (40% opacity)
Particulate Removal Efficiency by Particle Size
The particle size distribution of the scrubber inlet and outlet paniculate matter was measured at all four
test conditions shown in Table 3. The results of these analyses indicate that excellent particulate
removal efficiency occurred in most of the measured size ranges (cut-points). Figure 4a shows
surprisingly efficient particulate removal of the CT-121 scrubber at different particle size cut-points.
As was reported during earlier particulate removal tests and again observed in Figure 4a, the best
removal efficiencies were observed for particle sizes greater than 10|am. At all test conditions, there
was greater than 99% particulate removal efficiency of particles in this size bin. In some cases,
efficiency exceeded 99.99%. As the particle size decreased, there was a drop in observed particulate
removal efficiency, but over 90% efficiency was observed at all particle sizes between l|am and 10pm.
Between O.Sum and lum, the particulate removal dropped as low as 60%. In this range, it is believed
that acid mist carryover offsets the ash particulate removal, resulting in poor (calculated) particulate
removal values. As noted earlier, analyses of the outlet catch indicated that an average of 30% of the
outlet particulate can be attributed to gypsum and acid mist carryover. Below about 0.5|im, the
particulate removal efficiency again increases.
ATR TOXICS TESTING
The Yates CT-121 ICCT Project had two opportunities to measure its air toxics removal potential
(also referred to as 'hazardous air pollutants' or HAPs). In 1993, the Yates Project was chosen by the
DOE as one of its eight coal-fired sites for an air toxics study1 conducted on EPA's behalf, in support
Pages
EPRI - DOE - EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
August 25-29. 1997
-------
Final Results David P. Burford
CfflYODA THOROUGHBRED CT-ni CLEAN COAL PROJECT Southern Company Services, Inc.
of Clean Air Act Title LI requirements. In late 1994, the Yates Project expanded its scope of work to
duplicate portions of that 1993 effort, in an attempt to validate the DOE's 1993 results. The results are
both interesting and mutually supportive. However, the fossil fuel sources between the two tests were
very different and an exact comparison of results can not be easily made. In 1993, the DOE was
investigating three main issues;
• Air toxics characterizations / penetrations in fossil fuel systems (fuel / boiler / ESP);
• Air toxics removal potential for post-combustion equipment (ESP / wet scrubber);
• Air toxics emissions factor estimates, calculated in lb/1012 BTU.
The 1994 air toxics sampling conducted at Plant Yates by Radian, was performed to address the
technical difficulties encountered during the 1993 tests, specifically:
• Selenium sampling and analysis;
• Mercury partitioning and speciation;
• Flyash penetration of the FGD process; and
• Source apportionment (origin of exiting particulate matter),
In comparing the results of the two efforts, several observations emerge that may effect the use of air
toxics data in further rulemaking and health risk determinations:
• The Chiyoda CT-121 JBR is highly efficient at HAP removal;
• Sampling is very sensitive to ANY error at these near-minimum detection levels
(e.g.; sample container contamination, etc.);
• Source apportionment identifies a significant emission contribution from
particulate generated within the wet scrubbing process; and
• The 1994 effort saw significantly less measurement error than the 1993 effort.
The uncertainty in the 1994 testing data is generally lower than that of the 1993 testing data (i.e.,
sampling procedures improved). Due to the larger uncertainty evident in some species in 1993, the
accuracy of any calculated emission factors would likewise be suspect; fairly low uncertainty were
found for arsenic, vanadium, and lead. Conversely, antimony, chromium, manganese, and nickel all
had unacceptably large measurement confidence intervals in 1993; sometimes the confidence interval
was 10 times larger than the measurement itself. Calculated removal efficiencies for CT-121 's JBR
during the 1994 tests are shown in Figure 4b. It is prudent to remind ourselves that extrapolation of
admittedly 'uncertain' data does not produce 'certain' calculations for emission factor estimation or
subsequent health effects determinations. Caution should be emphasized in the use of the early 1993
data and any similar air toxics measurement data.
Page 9
EPRI - DOE - EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
August 25-29,1997
-------
Final Results David P. Burford
CHTYODA THOROUGHBRED CT-i 21 CLEAN COAL PROJECT Southern Company Services, Inc.
GYPSUM QUALITY
The gypsum stacking area at Plant Yates had three separate cells for segregated impoundment; a pure
gypsum area referred to as the "clean" gypsum stack area, a gypsum/ flyash stack area, and a recycle
water pond. During Phase I (the low-ash test phase) of this demonstration project, the "clean"
gypsum stack was used to dewater and store the pure-gypsum byproduct from CT-121; decanted clear
process water was collected in the common pond area and returned to the processThere was no
blowdown, discharge or treatment of scrubber process water. During the high-ash test phase (Phase
n), the segregated gypsum/ash area was used for stacking the ash/gypsum mixture. Since these stacks
are physically separated "cells", the original "clean" gypsum stack then, sat idle during the later
ash/gypsum phase of testing.
The gypsum slurry originally deposited in both areas was of a relative large crystal and had a high
chloride content, due to the closed loop nature of the Yates CT-121 system. Liquid phase chloride
concentrations were calculated to be as high as 35,000 ppm at equilibrium. Because of these high
chloride concentrations, any slurry-deposited gypsum solids would 'entrain' high chloride water and
would therefore normally require washing in order to satisfy limitations set by potential end-users; the
gypsum wallboard or cement manufacturing industries. Core samples of the "clean" stack at Plant
Yates that were taken after the stack had been idle for over a year indicated a surprising result: the
chloride concentration in the gypsum had decreased from about 6000 ppm, measured 3 months after
Phase I completion, to less than 50 ppm less than one year later.
Table 4 presents chloride data for the gypsum stack. The rate of chloride decrease over time, or as a
function of rainfall, was not measured at Yates because it was an unexpected development and an
unknown benefit of the gypsum stacking technique. There are two likely reasons for this decrease in
chloride concentration in the "clean" gypsum stack. First is that a majority of the chloride content in
the sedimented gypsum solids is due to the chlorides entrained in the moisture of the gypsum. Core
samples from the gypsum stack typically indicate that the solids content was approximately 83 wt.% on
average shortly after the stack was idled but that after one year, the solids content had increased to an
average of 90 wt.% at a depth of 3 feet. Secondly, the rainfall that occurred over the idle year washed
the gypsum thoroughly, time and time again which decreased chloride concentration in any remaining
entrained moisture. Although the decrease in entrained water played some role in decreasing the
chloride concentration in the gypsum, it is likely that rainwater washing of the stack was the dominant
factor. This is further evidenced by the data presented in Table 4 that shows free moisture did not
decrease at the 6 foot level, although chloride concentration did.
Page 10
EPRI - DOE - EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
August 25-29. 1997
-------
Final Results
CHIYODA THOROUGHBRED CT-121 CLEAN COAL PROJECT
David P. Burford
Southern Company Services, Inc.
Table 4
Declining Chloride and Moisture Levels in "Clean" Gypsum Stack
Dike
West
South
Inactive
Period
>90 days
>90 days
>90 days
>90 days
>400 days
>400 days
>400 days
>90 days
>90 days
>90 days
>400 days
>400 days
>400 days
Sample Depth (ft)
4
8
9.5
14.5
1
3
6
10
13.5
16.5
1
3
6
Chloride (ppm)
930
7610
5720
5540
60
40
20
5740
5610
6710
20
20
20
Moisture (%)
16.0
17.5
17.7
15.1
8.1
9.2
12.0
14.5
17.4
17.4
8.0
11.0
18.3
SUMMARY
Chiyoda's CT-121 FGD process was very successfully tested at conditions far beyond design
expectations. From an operating standpoint, the process was reliable, showed consistently high
removals (SQz, particulate, air toxics), was energy efficient and reagent efficient. From a chemical
engineering standpoint, the mass transfer interactions are robust and resilient, only limited at conditions
far beyond design parameters. This would allow a designer / operator to install a cost effective CT-121
system that would give consistently excellent service, even in periods of difficult operating conditions.
1. "A Study of Toxic Emissions from a Coal-Fired Power Plant Utilizing an ESP While Demonstrating the ICCT CT-121
FGD Project," Radian Corporation, Final Report for U.S. DOE, Contract No., DE-AC22-93PC93253. June 16, 1994.
Page 11
EPRJ - DOE - EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
August 25-29. 1997
-------
Chiyoda's Jet Bubbling Reactor
Air
Limestone
Slurry
Gypsum
Byproduct
Slurry
CT-121 Project at Plant Yates
SOUTHERN ML
COMPANY
Figure 1a
EPRI-DOE-EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
CT-121 Gas Sparger Action
Gas Gas Gas
Out In Out
JL f. J, Initial
^''" •'""- Liquid Level
Jet
Bubbling
'/li'-'i'' Submergence
•' " Depth
Gas
Sparger
CT-121 Project at Plant Yates
SOUTHERN
COMPANY
Figure 1b
EPRI-DOE-EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
-------
SO2 Removal vs JBR Deck AP
100
10 12 14
JBR Deck AP, In. W.C.
18
CT-121 Project at Plant Vales
SOUTHERN L
COMPANY
Figure 2a
EPRI-DOE-EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
High-Sulfur Coal Effects on SO2 Removal Efficiency
100
so,
Removal
Efficiency
85
CT-121 Project at Plant Yates
75 MWe/5300 mg/Nm
75 MWe/8300 mg/Nm
SOUTHERN^.
COMPANY
Figure 2b
EPRI-DOE-EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
-------
CT-121 Particulate Removal Performance
ESP - Field A Energized
Particulate Removals
Percent Removal
1
0.99
0.98
0.97
0.96
0.95
0.94
Test
I I50MW10"DP
BBS 50MW16'DP
CT-121 Project at Plant rates
I100MW 10"DP
I lOOMWIB'DP
Outlet Loadings
Test
-0.5 Lb/mm Btu
I I50MW10"
r^Eiq 50MW 16'
I100MW10'
I 100MW 16'
SOUTHERN^
COMPANY
Figure 3a
EPRI-DOE-EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
CT-121 Particulate Removal Performance
ESP Deenergized
Particulate Removals
Percent Removal
1-
0.995
0.99
0.985
0.98,
I 1 50MW 10-DP
rSRel 50MW 16"DP
|100MW10'DP
I 100MW16"DP
CT-121 Project at Plant rates
Outlet Loadings
0.04
Inlets: -5.0
I 1 50MW 10'
Test
-5.5 Lb/mm Blu
(100MW10-
I100MW16'
SOUTHERN^
COMPANY
Figure 3b
EPRI-DOE-EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
-------
FGD Dust Removal Efficiency
Removal Efficiency, %
100
80
40
20
CT-121
Conventional
FGD
0.1 0.2
0.5 1.0 2.0
Particle Size, pm
5.0 1C
CT-121 Project at Plant rates
SOUTHERN^.
COMPANY
Figure 4a
EPRI-DOE-EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
Air Toxics Removal
Yates CT-121 Project (JBR Components Only)
100
so
60
Toxics
40
20
mi 11
Chloride Fluoride Arsenic Cadmium Mercury Selenium Vanadium
CT-121 Project at Plant Yates
SOUTHERN M.
COMPANY
Figure 4b
EPRI-DOE-EPA Combined Utility Air Pollutant Control Symposium (The Mega Symposium)
-------
WET GYPSUM-YIELDING FGD
EXPERIENCE USING QUICKLIME REAGENT
H. Weiler
STEAG AG
Riittenscheider StraBe 1-3
45128 Essen, Germany
W. Ellison
Ellison Consultants
4966 Tall Oaks Drive
Monrovia, Maryland 21770
Abstract
Details are given of more than fifteen years of relevant commercial experience of STEAG,
Germany's largest IPP, in operation of wet lime-gypsum FGD serving low-sulfur bituminous coal
boilers. An assessment is provided of this substitution of high-calcium lime for limestone in
capacity aggregating 2100 mWe in respect to resulting improvements (compared with STEAG's
substantial, wet-limestone FGD experience) that include:
• Generation of gypsum byproduct fulfilling the most stringent quality requirements of German
gypsum wallboard manufacturers
• Suitability of scrubbing system materials of construction
• System availability
• Costs including maintenance.
History of STEAG's wet lime FGD orientation is outlined including an initial, six months long,
field test program that evaluated diverse limestone and lime sources; field testing of diverse
materials of construction, equipment components, etc.; and an ongoing quantification of process
parameters applicable to lime vs. limestone FGD design/operation.
-------
-2-
Introduction
When the GFAVO (Large Combustion Plants Ordinance) was enacted in Germany on July 1, 1983
new emission standards were enacted for large combustion plants. New plants, regardless of their
thermal capacity, had to meet these new values immediately. Existing plants with installed thermal
capacity greater than 300 MJ/s had to prove that these GFAVO values had been met within 5 years
at the latest, and those with installed thermal capacity of 50 to 300 MJ/s by 10 years at the latest. If
retrofitting was not intended then the plants had to be shut down after making use of a specified
period of remaining useful life.
By the end of 1996 desulphurization plants had been installed in bituminous coal-fired power
stations in Germany with a capacity of about 31,800 mWe, lignite-fired power stations with a total
capacity of about 11,500 mWe and oil-fired power stations with a capacity of 1,200 mWe. The
commissioning of retrofitted 500 mWe lignite-fired power stations in central Germany has raised
the contribution from lignite-fired units to 15,000 mWe.
Current Status of Wet Calcium Based FGD Plants in Germany
The experience gained in the development of flue gas desulfurization plants using quicklime in the
first phase started in 1978 and led to processes with an external sulfite oxidation stage. Later in
1983 most of the proposed SOx removal processes used limestone as absorbent. Large scrubbers
with flue gas volumes up to 2.5 Mio Nm3/h and with the sulfite oxidation stage incorporated in the
sump were erected.
Today a development is in progress via desulphurization plants that have smaller scrubber
dimensions because of their improved mass transfer coefficients. By the use of space-saving
components and appropriate design work these represent a compact, low cost, variant of the
plant [1]. Nowadays the absorber is usually an open spray tower without internal fittings so far as
possible. The single cycle scrubbing process is predominant in Germany. The plants offered differ
only in the dimensions and shape of the scrubber, in the chosen design parameters, and in
equipment and design details. As a rule the components in the flue gas path are not duplicated, but
redundant design is used in the rest of the equipment, such as pumps and belt filters, to improve
availability (Figure 1).
-------
3-
ean gaV
1
wgas
No.
1
2
J
>
ms/l?m>
>2700
<400
<400
Reheater
•
T
120-130
40-45
>75
Air
\
^ L Scrubber
/////////s-/
/f /i- /i>
Spray zone
v
Sump _„ 0
pH 5.0 -5.8
-_
Droplet collector ,,,
H Wqctpu/at^r
hydrocyclone
H^
_w<
>
^
^^
by
\
"
Limestone or slaked lime
Gypsum . -^— ^
drocyclone • ^
~Y
V/ s 9 IJ f^
T ( ) I ( )\b
[
Process water
^
),
acuum
:lt filter
Process water
(low chloride content)
^
Dewatered gypsum
^^ purity >95%
^"Wastewater
Figure 1
Lime/Limestone-Gypsum Desulphurization Process
The materials used for the surfaces in contact with the wash medium are rubber-lined carbon steel,
steel plated with a nickel-based alloy, or a pure austenitic material. In the past the higher operating
temperatures in oil- and lignite-fired plants resulted in extensive damage in scrubbers that were
lined with certain grades of rubber. The rubber linings now reach service lives of more than 5 years
by using modified, butyl-based, grades of rubber. Fibre reinforced plastic as well as pure plastic or
rubberised steel are used for the pipelines.
For supply of pumps the all-metal pumps have the greatest share of the market. Service lives of
over 35,000 running hours have now been achieved when limestone is used, with over 60,000
running hours for quicklime. The droplets in the cleaned flue gas are normally removed by coarse
and fine droplet separators. Droplet levels of less than 50 mg/Nm3 can now be achieved in
continuous operation. According to the 13th BImSchV the clean flue gas at the outlet of the stack
must have a minimum temperature of 72°C. The equilibrium temperatures in the scrubbers are
lower than this so the flue gases have to be heated. This heating can be carried out using external
energy, e.g. steam, or regeneratively by utilising rotating heat exchanger systems operating on the
Ljungstrom principle. It is possible to avoid the need to reheat the flue gas by discharging the
cleaned flue gases through an existing hyperbolic cooling tower.
Vacuum belt filters or centrifuges are mainly used for dewatering the gypsum slurry. In Germany
centrifuges account for about 35 % of the market for production of a marketable grade of gypsum.
Dictated by the processing requirements of the gypsum users, it is necessary to remove detrimental
impurities such as chlorides and magnesium salts from the gypsum dihydrate. The wash water
from the gypsum washing must be treated in a wastewater plant by the addition of milk of lime,
flocculants, or sulphides before it can be discharged into the public surface waters.
-------
-4-
Gypsum Quality Requirements in Germany
The recovery of gypsum is an indispensable part of the application of the German FGD technology
since there are generally no landfills available for ultimate disposal of such solids.
All existing FGD plants attain and exceed by far the gypsum purity required by the gypsum
industry, and thus the byproduct is generally superior to typical natural gypsum supply (Table 1).
The surface moisture content of the FGD gypsum is less than 10 % without use of thermal
treatment.
To meet the quality requirements of the gypsum industry, fresh-water washing of the gypsum filter
cake is necessary to displace the components, particularly the chlorides, that cause difficulty in
gypsum use in wallboard manufacture.
In meeting the stringent requirements specified by the gypsum industry, high quality FGD gypsum
can be used for:
• Wallboard and plaster production
• Self leveling poured floor toppings using anhydrite or alpha hemihydrate forms
• Replacement of fillers in reinforced plastics or paper coatings
• Special construction materials / products / shapes.
If the FGD gypsum is of intermediate quality, e.g. with chloride content up to 1,000 ppm, it can be
used as:
• Set retarder for Portland cement (drying and agglomeration is no longer found to be mandatory)
• Mining mortar using dehydrated gypsum forms.
Other areas for FGD gypsum consumption are:
• Road construction
• Landfillino and land recultivation.
-------
Table 1
Quality Parameters for Gypsum
Gypsum quality parameters Specifications for
Gypsum industry Cement industry
Governed by absorbent and
process design:
CaSO4x2H,O (purity)
Sulphite as SO2
Whiteness
Odour
Toxic constituents
Governed by gypsum washing
and dewatering system:
Surface moisture
cr
Na2O
MgO, water-soluble
pH value
> 95.0 %
< 0.25 %
> 80.0 %
neutral
none
< 10.0%
< 0.01 %
< 0.06 %
< 0.01 %
5-9
n.s.
< 2.5 %
n.s.
n.s.
n.s.
< 10.0%
< 0.3 %
< 0.5 % with K2O
< 2.0 %
neutral
With exception of the sodium content the requirements of the German and American gypsum
industries are similar. However, the requirements of the German and American cement industries
differ considerably [2].
The German cement industry has modified its facilities and is no longer dependent on use of lump
gypsum. Therefore, drying and agglomeration of the moist fine gypsum is no longer required.
As briefly shown in Table 1 a few parameters like purity, sulfite content, whiteness and odour can
be influenced by the quality of the absorbent used and by the process design. Other quality
parameters like surface moisture, chloride, magnesium and alkali content are nearly independent of
the absorbent quality. The content of water soluble impurities in the gypsum cake can be
controlled by the specific wash water consumption rate e.g. 0.5 t water/t gypsum.
Beside the a.m. criteria have the precipitation of considerable amounts of fly ash due to high fly
ash content in the flue gas can influence the gypsum quality significantly, especially via unburned
carbon. It is reported from some limestone FGD plants that, due to the content of organic matter,
sulfate reducing bacteria appeared, forming black metal sulphides like FeS, HgS, PbS, etc.. By
switching to quicklime this phenomenon disappeared. Using lime instead of limestone the
whiteness of gypsum produced is nearly equal to or slightly higher than the whiteness of the
quicklime.
A survey by the VGB in 1995 has shown that 99.8 % of the FGD gypsum from coal-fired plants is
utilised Some of the gypsum from lignite-fired power-stations in western Germany is still used for
back-filling open-cut mines, but through the development of new processes the gypsum is being
increasingly supplied to the market for building materials and fillers. Now that the retrofitted and
...6
-------
newly built power stations and desulphurization plants in central Germany (where 1.5 to 2.0
Mio t/a FGD gypsum is produced), have been commissioned there may be seasonal surpluses that
will have to be held in intermediate storage. A study by the UBA (Federal Environmental Office)
has shown that in the long term the demand for gypsum will be greater than the production of FGD
gypsum [3].
When the GFAVO was enacted in 1983 there were only 10 FGD plants in operation, typically
treating a slip stream of the total gas flow volume. At the end of the retrofit period in 1988 more
than 165 scrubbers were in operation in 72 power plants, representing a total capacity of nearly
38,000 mWe.
The trend in the application of the various FGD processes is shown in Table 2.
Table 2
Trend in FGD-Process Application in Power Stations
FGD Process
Wet scrubbing processes
. Lime/limestone-gypsum
. Wellman Lord
• Ammonia
• Desonox
Dry / semi-dry processes
. Spray absorption
. Direct injection incl. CFB-
boilers
• CFB-absorbers
. Activated coke
Total FGD capacity
Power station fueling
• Bituminous coal
. Lignite
. Oil
Market share
1990
84.94 %
3.51 %
0.19%
0.09 %
7.07 %
2.27 %
1.17%
0.76 %
46,600 mWe
72.41 %
26.48 %
1.11 %
1996
87.66 %
2.45 %
0.16%
0.15%
6.37 %
1.61 %
0.96 %
0.64 %
48,000 mWe
66.25 %
31.25%
2.50 %
According to Table 2, lime/limestone based FGD with a capacity of 42,100 mWe is in operation as
of 1996. Approximately 8,300 mWe are using quicklime as reagent. Roughly 6,900 mWe of FGD
capacity was designed during the first implementation phase for the use of quicklime. FGD with a
capacity of approximately of 2,400 mWe has been designed for limestone application and has been
switched after a short period of operation from limestone to quicklime.
-------
-7-
Use of Quicklime Instead of Limestone
If the FGD plant is designed and constructed for the use of quicklime as absorbent, a slaking
system has to be installed producing lime hydrate (slaked lime) from quicklime. Different lime
slaking techniques exist:
• Intermittent batch process for low throughput
• Continuously operating, two-tank slaking system for throughput up to 12 t/h
• Continuously operating, paste slaking system for throughput up to approx. 4 t/h
• Continuously operating, drum or ball mill slaking system with throughput up to 35 t/h.
In German FGD systems using quicklime two-tank and paste slaking systems are the preferred
type of installations.
Wet limestone FGD facilities in west Germany are typically supplied with ready-to-use, powdered
limestone. The preparation of the limestone slurry is normally carried out in a simple tank system
provided with agitators.
Such a limestone FGD facility can be easily converted to a quicklime FGD plant by modifying the
slurry preparation system.
If a limestone grinding system is in use in a large power station a drum slaking system has to be
used, especially for high sulfur fuel. The expenditure for such a retrofitted slaking plant is
comparatively high. Taking into account the capital cost of the grinding system already installed,
the economic benefit has to be assessed in depth.
Based on prior experience with medium size units and medium gross SO2 concentration, and
considering German regulations and site criteria, initial economic calculations led to the
conclusion that the use of quicklime instead of limestone could offer operational and cost
advantage. Nearly all values for the applied parameters are listed in Figure 2. Only the cost for
limestone gypsum, 2.00 DM/t, and the profit for quicklime gypsum, also 2.00 DM/t, and the
25% power reduction for the slurry recirculation pumps are not shown.
Another consideration is the lower capital cost of a quicklime-using FGD facility caused by the
smaller scrubber size and the reduced number of spray levels, pumps and auxiliary equipment
items, e.g. silos.
-------
350 M\V uml
4000 mg/Nm'SO:
lOOmg/Nm'HCI
VK,,.=90%
Parameter:
price ratio
CaO/CaCO,
100 200
Distance to FGD (km)
Figure 1
Reagent cost
Parameter: CaCC>3= 50 DM/t, power: O.lDM/kWh, solid waste: 150 DM/t,
limestone gypsum: +2 DM/t, quicklime gypsum: -2DM/t, Annual
operation: 5000 full-load hours, maintenance: 3%/year, freight: 0.2 DM/t.
Existing Operating Experience with Quicklime as Absorbent
The fist FGD plants installed in Germany used quicklime as reagent and reactors designed for
external oxidation with air injection and using sulfuric acid for the pH adjustment. After
commissioning of the second generation of FGD facilities with in situ sump forced oxidation in
Germany the existing FGD installations were investigated with regard to converting them to in situ
sump forced oxidation systems as a means of reducing the operating costs. After some test work to
find out the best kind of agitators and air injection systems the scrubbers of STEAG's first FGD
facilities have been converted to sump oxidation.
Due to the relatively small sump volume the residence time of the gypsum particles has decreased
and for this reason the average particle size has also decreased from 45 (i to 35 - 40 u. The quality
of the gypsum nonetheless fulfills the requirements of the users.
According to the GFAVO, STEAG's older units with a total capacity of 2,400 mWe had to be
retrofitted with FGD plants. The chosen single loop system could not fulfill the specified guarantee
for gypsum whiteness. During nearly 1 year of operation many types of limestone powder
including chalk were tested but the guarantee could not be met with any of the qualities tested.
STEAG agreed with the FGD supplier to add a lime slaking system to the FGD plant. The two-
tank slaking systems were able to use many parts of the existing limestone slurry preparation
-------
Table 3
Overview of STEAG-Power Stations and FGD -Installations
Power
Station
Bergkamen
Voerde A / B
Herne 1+2/3
Herne 4
Lunen 10/11
(until 10/94)
Siersdorf
Walsum 7 / 9
West 1 / 2
Leuna
Capacity
[mWe|
750
710/710
2x150/300
500
150/350
160
150/410
350 / 350
3x115
(thermal)
Furnace
System
dry ash
dry ash
slag tap
dry ash
slag tap
slag tap
dry ash
slag tap
oil
FGD Design Data
SOx-Inlet
Cone.
[mg/m3 normal cond.]
2,600
2,100
4,300
4,650
3,850
2,700
4,000
3,850
7,000
Process /
Absorbent
wet scrubber
lime
wet scrubber
lime
wet scrubber
lime/ limestone
wet scrubber
lime / limestone
wet scrubber
lime /limestone
semi-dry lime
(circulating fluid
bed)
wet scrubber
lime / limestone
wet scrubber
lime Limestone
wet scrubber
lime / limestone
Byproduct [t/h]
(Gypsum)
18.0
14.5/14.5
13.5/13.5
30.0
6.0 / 12.5
3.0
(spray absorption
product)
6.5/18.0
12.5/12.5
7.0
Operating
hours
(up to 12/96)
110,000
95,000 / 95,000
68,000 / 66,000
51,000
48,000 / 65,000
43,000
22,000 / 69,000
62,000 / 62,000
500
Operational
since
10/81
10/82
1/87
9/89
4/88
5/88
10/88
2/88
12/96
-------
-10-
systems. The necessary capital amounts to 1 - 2 Mio DM. The design and implementation of the
system allows the operator to switch from quicklime to limestone if required. One FGD plant is
still using limestone as absorbent because the high gypsum quality is not needed. The historical
development of the use of FGD processes and quicklime or limestone as reagent in STEAG's FGD
plants is further demonstrated in table 3.
It can be stated that adequate whiteness of the gypsum when retrofitting old plants can generally
only be achieved by the use of quicklime because of elevated dust content in the flue gas. With
equivalent conditions it is thus possible to achieve an improvement in the degree of whiteness of
15 to 20 % as compared with the use of limestone. The gypsum whiteness when quicklime is used
is the same or slightly better than the whiteness of the quicklime itself. When limestone is used the
whiteness of the gypsum is 5 to 15 % below that of limestone [4],
Concerning the transport costs the use of quicklime also offers the advantage that the absorption of
t S02 only requires 0.88 t quicklime instead of 1.56 t limestone. However, it is particularly
important that, with the available plant technology, the reactivity of the quicklime results in higher
removal efficiency. Beside the reduction of the scrubber slurry recirculation rate, which results in
energy savings, other potential savings are achieved as shown in Table 4.
Table 4
Comparison of Quicklime and Limestone Use in FGD Plants
Parameter
Power consumption of
recirculation pumps
Mol. stoichiometric ratio
Gypsum quality
. purity
. whiteness
Residues from wastewater
treatment
Maintenance cost
Transport cost for the
absorbent per t SO2 removed
Absorbent
quicklime limestone
CaO CaCO3
60 - 75 %
< 1.01
98 - 99 %
= 80
60 - 70 %
80 - 90 %
54%
100%
1.02- 1.04
95 - 98 %
«60
100%
100%
100%
Very low chloride and high sulfur content in fuels like residual oil, refinery residues or certain
lignite deposits leads to a change in FGD process control because the chloride level in the slurry
can no longer be used as a reference variable to prescribe the quantity of FGD effluent to be
discharged. Using limestone reagent and the usual method for wastewater treatment, large
quantities of difficult-to-dewater magnesium hydroxide sludge could be formed.
Using quicklime instead of limestone and keeping the pH value of the slurry above 6.3 the
magnesium oxide remains as inert matter in the slurry and does not form magnesium sulfate, later
to be neutralized in the wastewater treatment plant forming gypsum and magnesium hydroxide.
... 11
-------
11
Avoiding the dissolving of magnesium oxide in the FGD process results in better process control
and considerable cost reduction [5].
This was one important reason why a large utility in Germany made a decision to fit each of two
new 900 mWe lignite-fired boilers with a quicklime FGD facility.
A second important consideration was that the capital cost for a quicklime based FGD system is
significantly lower than for a limestone FGD plant.
The substitution of quicklime for limestone is carried out for both single loop and double loop
systems. Limestone FGD facilities of different suppliers of single loop design have been converted
to quicklime and tested. If the pH value in the scrubber sump is adjusted to within the range of 5.0
to 5.8, the oxidation rate is high enough to attain adequate oxygen efficiency in use of the injected
air.
Special attention has also to be paid to the pH value of the slurry entering the spray nozzles. PH
values above 8.5 - 9.0 could lead to the formation of scale and deposits.
The double loop FGD system was tested in a 1-month test run to check the applicability of
quicklime in that FGD process. The results were encouraging as the power consumption of the
scrubber recirculation pumps decreased considerably (30 %). The pH value in the quencher (first
scrubber stage) sump could not be adjusted to a value of approximately 6.3, and the surface
moisture of the dewatered gypsum was 3 % higher than that of the limestone gypsum. It is
expected that by optimization of the double loop process the results achieved will be similar to
those for the single loop process.
Transfer of Existing Experience With American FGD-Plants
The gypsum demand in Germany in 1992 was approximately 6.1 Mio short tons. About 3.5 Mio
short tons were produced in FGD plants of which 0.6 Mio short tons were exported e.g. to England
and Norway. Nearly 50 % of the gypsum used in Germany is produced by FGD plants. With
commissioning of new FGD plants in central Germany the market share served by FGD gypsum
will increase significantly. Large gypsum works with a total production capacity of approximately
2 Mio short tons per year have been erected in close proximity to the power stations in central
Germany.
The annual gypsum consumption in the US in 1994 amounted to 27 Mio short tons.
Approximately 9 Mio short tons are imported, transported long distance by water. Only 0.5 Mio
short tons, a market share of less than 2 %, are supplied by FGD plants. When available close to
points of large-scale use, FGD gypsum use is nonetheless economically attractive.
The physical form of the FGD gypsum is significantly different than that of natural gypsum, but
experience gained in Germany clearly shows that candidate users of FGD gypsum can adapt their
facilities to the moist fine gypsum filter cake. Only in special cases is there a need to dry and
agglomerate the gypsum. The purity of the FGD gypsum is normally equivalent to or greater than
the purity of natural gypsum depending on FGD design, type reagent, power station fuel use, etc.
In the US only 23 % of the existing wet FGD plants are designed for usable gypsum production.
Approximately 58 % of the plants use limestone and quicklime as reagent in producing sludge
... 12
-------
12-
requiring fixation. Approximately 11 % of the FGD plants are supplied with magnesia enhanced
lime yielding sludge that is fixed with dry fly ash.
Wallboard production constitutes the largest consumption of gypsum in the US (19.2 Mio short
tons per year) and requires high quality gypsum. Substituting high-calcium quicklime for
limestone is a potential means of enhancing gypsum quality when applicable.
In the US cement industry in 1994 4.75 Mio short tons of gypsum with lesser quality were used. A
large proportion of the required gypsum can potentially be supplied by limestone/gj'psum
facilities. Note the possibility of installing sulphite-oxidation systems in limestone- sludge FGD
plants to produce a gypsum slurry or filter cake of adequate quality for cement industry supply. A
major deterrent is the large unit size and high sulfur level at most FGD-equipped plants, requiring
very many cement-plant customers for each such power plant source.
Gypsum from flue gas cleaning is an environmentally sound byproduct, particularly in
comparatively small countries like Germany, The Netherlands and Japan with short shipping
distance and large gypsum consumption.
References
[1] Klingspor, I.S., Bresowar, G.E.: Innovative nasse Rauchgasentschwefelung mil Kalkstein.
ABB Technik 8 (1995), pp. 23-27.
[2] Luckevich, L.M.: Gypsum Market Saturation in North America. Proceedings Power Gen
International 96 Conference Dec. 1996.
[3] Haug, N.: Substitution von Naturgips durch Gips aus Abgasentschwefelungsanlagen.
Staub-Reinhaltung der Luft 54 (1994), pp. 309-312.
[4] Weiler, H.: Erfahrungen mit der Abgasreinigung mit Kalkprodukten bei Kohlekraftwerken
und Abfallverbrennungsanlagen. ZKG-International 50 (1997) No. 2, pp. 96 - 111.
[5] Kahl, D., Pfeiffer, I: Optimierung der Rauchgasentschwefelung fur die Qualitat der
Reststoffe beim Einsatz von Braunkohle mit hohem Schwefelgehalt an einer Pilotanlage.
VGB Kraftwerkstechnik 76 (1996) No. 2, pp. 147-150.
-------
3 INTO 1: FIRST MULTIPLE BOILER FGD UNIT
STARTED IN NORTHERN ALBERTA
Alex M. Runcie
John G. Mader
Glenn J. Lutz
Fluor Daniel Canada, Inc.
Calgary, Alberta, Canada T2W 3N2
ABSTRACT
The first commercial installation of the Chiyoda Thoroughbred 121 (CT-121) Fluegas
Desulphurization process in North America was recently commissioned at the Suncor Energy Inc.
plant site in Ft. McMurray, Alberta. Engineering, procurement and construction was performed by
Fluor Daniel Canada, Inc. The Chiyoda process has been installed in locations elsewhere in the
world, but never with one reactor serving three boilers simultaneously. The FGD unit is required
to remove up to 33,000 pph of SO2 from 3,430,000 pph of flue gas produced by three (3) delayed
coke fired boilers, using local limestone of low purity.
A number of unexpected problems were encountered during the design. These problems were
resolved without extension to the schedule, allowing the unit to be commissioned on time. Initial
operating results have proven the decisions made, and the performance of the process. The
requirement to maintain steam flow at all times has been met and the 95% SO2 removal
requirement has been exceeded.
INTRODUCTION
An EPCM (Engineering, Procurement, and Construction Management) contract was awarded to
Fluor Daniel Canada Inc. (FDCI) on March 3, 1994 to execute the detail engineering, material and
equipment procurement, and construction of an SO2 Reduction facility for Suncor Energy Inc., per
the defined design basis (DBM) of a third party.
Suncor is located near Ft. McMurray, Alberta, approximately 600 miles north of the Montana/
Alberta border. Suncor is a Canadian integrated oil and gas company with major interest in the oil
sand industry. Its Oil Sands Group has been producing light, sweet crude from the Athabasca
deposit since 1967. A byproduct of the upgrading process is delayed coke which contains
approximately 6%wt sulphur and has a heating value of 13,000 btu/lb. This coke is used as the
primary fuel for Suncor's three main utility boilers. In order to reduce the sulphur emissions, a
single common Flue Gas Desulphurization (FGD) unit has been installed to serve all three boilers.
Pagel
-------
The FGD unit uses the Chiyoda CT-121 process, which utilizes a large Jet Bubble Reactor (JBR).
in which the flue gas is bubbled through a limestone slum' to convert the sulphur compounds
(mostly SO,, with some SO,) to gypsum. Flue gas exits the unit into a new 450 foot wet stack. The
onainal "old" stack continues to be used for other gas fired utility boilers, and also serves as a
bypass stack for the coke fired boilers when the FGD unit is off line. A overview of the Suncor
utility plant and FGD unit can be seen in Figure 1.
Because of the severe winter weather conditions, as well as the importance of the utility plant to
oil production, maintaining the reliability of the steam generation is critical. The bypass stack is
one example of steps taken to ensure the reliability of the steam production. It allows the plant to
continue operating if the FGD unit is unavailable. The bypass system was designed to allow the
coke fired boilers to switch to. or back from, bypass, with no interruption in steam production
during the transfer.
This paper will describe some of the challenges met developing the design from the DBM. the
solutions applied, and results after one year of operation.
PROJECT REQUIREMENTS
The DBM identified a single SO, scrubber to be installed to scrub 91% of the 33,000 pph SO,
from the 3.43 million pph of flue gases produced from three delayed coke fired industrial boilers.
The FGD system description put forth in the DBM included:
1. The Chiyoda CT-121 Fluegas Desulphurization process be used.
2. A gas / gas heat (GGH) exchanger.
3. The boiler bypass and scrubbed gases go to the existing stack via the existing ducting.
4. One electric motor driven axial booster fan.
Page:
-------
5. The limestone reagent is supplied from the local formation underlying the oil sand.
6. One years supply of limestone is mined and sized during the summer months only.
7. Limestone is delivered to the FGD plant, and gypsum removed, via track, only during the
day shift (12 hrs/day).
8. The gypsum produced is mechanical dewatered.
The DBM also identified four issues that were unresolved and needed to be either finalized or
confirmed. These were as follows:
1. Perform a dynamic simulation to assess how best to control the booster fan and boilers.
2. Establish the heat balance around the GGH outlet temperature to allow the continued use of
the existing flues and stack, and to prevent the formation of SO3.
3. Negotiate the Purchase Order (P.O.) terms for the Chiyoda CT-121 process design and
auxiliary equipment supply.
4. Resolve the problem of dewatering the gypsum. Chiyoda had identified that there may be
problems mechanically dewatering the resultant gypsum, if native limestone was used.
Suncor also placed the following requirements on the design and construction of the plant:
1. Reduce SO2 emissions from the Utility Plant, to meet Federal and Provincial Air Quality
Standards, to the lesser of 33 tonnes/day or a ground level concentration of 0.17 ppm. To
accomplish these targets, 91% SO2 recovery was required. This translated to 93% recovery
in the reactor, a GGH leakage no more than 2%, and a minimum flue gas stack temperature
of325°F.
2. 98% on-line factor (7 days per year total down tune). To meet the 98% on-line factor, the
design basis included for redundant equipment. This high on stream factor was necessary
as Suncor's Energy Services provides all of the steam requirements for the process
(bitumen extraction and upgrading) and some of the overall plant electrical needs.
3. No increase in particulate.
Process Equipment Description and Discussion
Gas / Gas Heat Exchanger (GGH)
When the coke fired boilers were constructed (1965), no attempt was made to optimize the heat
recovery since the fuel is a byproduct of the process, and more coke is produced than combusted in
the boilers. Thus, the normal operating temperature of the flue gas leaving the ESP's (Electrostatic
Precipitators) is approximately 540°F. The DBM concept was to heat the scrubbed flue gas stream
against the hot incoming fluegas via a gas / gas heat exchanger (a regenerative airheater). If this
hot fluegas could sufficiently heat the wet scrubbed gas leaving the JBR, it would allow the "old"
refractory lined stack to continue to be used. Finalization of the conceptual design for the gas / gas
heat exchanger was required as part of the scope of work.
A review of the heat balance around the GGH showed that a balance could not be achieved that
would maintain a scrubbed gas outlet temperature above 325°F without supplemental firing. The
life cycle costs for the GGH and supplemental firing exceeded the estimated cost for a new wet
PageS
-------
stack, therefore, the GGH was abandoned. A consequence of abandoning the GGH was the
requirement for additional scrubbing in the JBR, from 91% to 95% recovery. The increase in
recovery was required to compensate for reduced dispersion of the cooler, less buoyant fluegas.
The "old" stack was also used by five (5) gas fired boilers in addition to the coke fired units. When
the decision was made to provide a wet stack, it was assumed that all the boilers would be
redirected to it, abandoning the "old" stack completely. The "hot" flue gases from the gas fired
boilers would add to the temperature of the scrubbed gases thereby adding lift which would help
meet the ground level concentration requirement. The consequence of this decision, however, was
that the liner material had to be suitable for a wide range of temperatures (from 600°F at bypass, to
130°F with the scrubber on-line), and the flue duct layout had to be changed. For a detailed
explanation of the impact on the new stack, refer to the "Wet Stack" section.
Chiyoda Scrubber
The Chiyoda scrubber has three mam parts, the gas cooling section, the jet bubble reactor (JBR),
and the mist eliminator section. The gas cooling section reduces the incoming flue gas temperature
and saturates the flue gas before it enters the JBR. There are four (4) banks of spray nozzles; one
for raw water (required during normal operation), one for emergency cooling, and two for gypsum
slurry (required during normal operation). The gypsum slurry is sprayed into the gas stream as the
primary cooling medium, and also provides humidification and a small amount SO2 removal. The
saturated, cooled flue gas is then bubbled into the reaction liquor in the JBR through sparger tubes.
Oxidation air, provided by three 50% centrifugal blowers, is introduced into the JBR reaction
liquor to complete the chemical reaction. As the fluegas bubbles through the liquor, the limestone
(CaCO3) is converted to gypsum (CaSOJ, leaving the scrubbed gas to exit the top of the vessel.
The oxidation air also increases the liquid level in the JBR (design stated 6" per blower). This
change in level results in an increased backpressure on the ID / Booster fan.
A detailed start-up and shutdown analysis of the system uncovered weaknesses in the design:
1. The control of the JBR liquid level is based on the pressure drop measured across the
sparger tubes. Therefore, the loss of gas flow results in liquid level control being lost.
With the loss of automatic control on the water input to the JBR, only operator intervention
can prevent the liquid level from eventually exceeding its maximum and the internal
pressure of the JBR exceeding its design.
2. The JBR maximum liquid height was 29'-10" with a normal operating level of 29'-6" It
was determined that starting the standby blower, without first stopping the operating unit,
would raise the liquid level in excess of the maximum liquid level, thereby increasing the
JBR internal pressure above the design point.
3. Should the mist eliminator start to plug, such that the pressure drop exceeds its maximum
design, the design pressure of the JBR would be exceeded.
Various mechanical and control system solutions were discussed, however the final decision was
to install a weir to limit the liquid level, and therefore pressure, in the JBR.
Page 4
-------
Since the commissioning of the FGD unit, it has been discovered that:
1.
2.
3.
4.
The process works very well. The 95% recovery of SO2 is exceeded on a consistent basis.
Tests have shown 98.9% achieved.
Each air blowers raises the JBR liquid level by 18", not 6" as stated, and therefore, have
had to be choked to reduce their individual contribution to the liquid level.
The installation of the weir was the right decision, as the JBR was overfilled on several
occasions during the commissioning of the unit.
Plugging of the gas cooling section nozzles has been a problem. Initially it was from
construction debris and coke contamination from the coke operation adjacent to the
limestone pile. Today it would appear it is scale and coke that are the causing the problem.
Scaling is not in excess of 1" (process guarantee), but it is of sufficient thickness to cause
plugging of the gas cooling spray nozzles. Remedial action has been taken by installing
suction screens for the gas cooling pumps. Based on previous applications in similar
installations, the screens are expected to eliminate the pluggage problem.
Limestone
The DBM included for a dump pocket, one limestone conveyor, and one limestone silo. It was
assumed the mining of the limestone would occur only in the warm summer months. A years
supply of limestone was to be mined, sized and stored at the mine. Daily it would be brought to
the plant from the mine in trucks and dumped into the drive over dump pocket. It would then be
conveyed from the dump pocket to the silo from which one of the two wet ball mills would be fed
for crushing and slurrying of the stone. The resultant limestone slurry was to be pumped through a
hydroclone into a limestone slurry tank from which the process was supplied. The limestone trucks
would then back haul the gypsum product to the mine to be used as landfill. This cycle was to
operate only on the day shift, hauling an equivalent 24 hours worth of presized limestone from the
mine and back hauling the same in gypsum. Thus, to meet the on-line factor, the limestone silo
and the limestone slurry tank were each sized for a 24 hour capacity.
It had been determined that the limestone underlying the oil sand was suitably reactive for
scrubbing SO2 . Core samples were taken to determined the quality of the limestone; the results
are as shown in Table 1. The bidders were warned that the limestone can contain impurities in the
form of oil sand, bitumen and clay. It was these impurities, the clay in particular, that caused
material handling problems with the limestone, and difficulties dewatering the gypsum.
Chemical Composition
Calcium Carbonate (CaCO3)
Silica (SiO2)
Aluminum (A12O3)
Iron (Fe,O3)
Chlorides (C1-)
Design, % by weight
57.0
25.6
7.5
5.1
0.11
Design Range, % by weight
55.0-91.1
3.1-28.4
0.5-8.3
0.0 - 5.7
0.02-0.11
Table 1
Composition of Limestone Native to Suncor
PageS
-------
Chiyoda had indicated in their proposal that there may be difficulties in dewatering the gypsum
produced using the native limestone. Therefore a test program was initiated using the EPRI's High
Sulphur Test Centre (HSTC), in Buffalo N.Y. It was recognized that the gypsum crystals produced
in the test facility simulating a tower scrubber would not be completely representative of what
would be generated by the CT-121 process, but it was felt that it would supply suitable
information on the ability to mechanically dewater the gypsum produced.
In the winter of 1994, limestone was obtained from the Suncor site, crushed and sized, and shipped
to the EPRI HSTC in bulk transports. When the trucks arrived in Buffalo, the limestone could not
be unloaded. It was thought the limestone was frozen in the trailer and steps were taken to thaw it
out. These efforts were not a complete success. One truck load had to be abandoned and brute
force got the other clear. Due to concern over this solidification, Jeneky and Johansson were
contracted to analyze the limestone and perform flowability tests. The results indicated that should
the limestone have a moisture content greater than 6% it would agglomerate, greater than 3%
moisture, and allowed to sit for 24 hours, it would also set up. With this knowledge, it was realized
that the batch crushing and silo configuration was not suitable, and another arrangement was
required.
It was determined that any hold up of the limestone was not appropriate for this type of limestone;
it would have to be conveyed to the process in a continuous manner. Working with Fluor Daniel
Wright, a scheme was proposed to deliver 6" minus rock to the FGD plant 24 hours a day, crush it,
then convey it into the silo. Unfortunately, this would have required adding more equipment,
would add significant cost to the project, and was not a sure way of eliminating silo plugging.
Eventually, Suncor's mining personnel determined that the stone could be crushed, sized (5/8"
minus), stockpiled in the mine, and transported 24 hours per day, thus eliminating the need for the
silo and long term storage at the FGD plant. Consequently, the slurry preparation equipment was
re-arranged to take advantage of the mine's capability to provide sized stone to the FGD facility.
Dual conveyors were installed, in place of the single one called for in the DBM. Each conveyor
feeds a surge bin located inside the building, instead of the limestone storage silo originally
envisaged. Trucks deliver the stone to an area near the conveyors and the coke pit front-end loader
piles the stone over the conveyors thus eliminating the drive over dump pocket. The tail end of the
conveyors is located in a tunnel under the pile of the crushed limestone. The stone is discharged
from the pile to either of the conveyors depending on which of the two ball mill trains are in use.
In emergencies, the stone can be deposited directly to either conveyor by the front-end loader. It
was expected the pile would crust over, but the inside of the pile, and the side being worked by the
loader, would remain loose.
The tendency for the limestone to agglomerate at low moisture levels produced a secondary
concern with respect to material bridging. With the knowledge that even a brief interruption in the
flow of limestone could allow the material to set up, measures needed to be taken to ensure
bridging did not occur.
Wherever possible, all chutework and transitions were oversized and designed to avoid any
restriction in flow area. A vibrating pile activator, generously oversized for the required duty, was
Page 6
-------
installed beneath the limestone stockpile to facilitate material transfer. Further, the stockpile
activator was designed without an internal pressure cone to further minimize the chance of
material bridging. Moreover, throughout the material handling equipment, all surfaces in contact
with the limestone were fabricated from stainless steel. Finally, air cannons were installed at both
the loading area beneath the limestone pile, and at the limestone surge hopper immediately
upstream of the wet ball mills, the areas believed to be most susceptible to bridging.
Field performance to date has indicated that the material handling system design has limited the
tendency of the limestone to bridge. Although the system has fouled on several occasions
throughout its first year of service, the root cause, in the majority of cases, has been the presence
of grossly oversized product. During several of the first system pluggages, uncrushed limestone as
large as 6" x 6" x 12" was pulled from the system. Improved quality control within the crushing
and stockpiling operation has since greatly reduced the frequency of system fouling. Operation
has also shown that the final arrangement of the system has proved to work, but is not without
operating difficulties. The limestone definitely does adhere together. The piles in the mine have
been difficult to fracture back into the sized material by use of the loading equipment.
Contamination of the limestone by the front-end loader and the proximity of the coke pit has
proven to be a source of plugging in the gas cooling section of the reactor.
Limestone Slurry
The final analysis of the native limestone revealed that it could contain up to 28%wt silica. This
extremely high content of abrasive particles proved to have a large impact on the selection of
materials for the limestone slurry piping. After reviewing the data, Chiyoda made the
determination that all limestone slurry piping must be constructed of cast basalt lined pipe.
The basalt pipe represented a significant deviation from the more typical selection of rubber lined
pipe often used in this application. Not only did the basalt pipe provide a substantial cost impact
for the raw piping materials, but it was also significantly heavier than conventional piping,
resulting in a large increase hi the quantity and size of structural support steel.
Operation of the FGD facility to date, has proven the merits of the basalt pipe. No indication of
wear has been found hi the pipe, even at transitions and abrupt changes of direction. Moreover,
two examples exist as to how alternatives materials would have faired. The feed chute into each
ball mill is standard rubber lined carbon steel construction. Even though each ball mill has only
seen service for approximately 50% of the tune until the first shutdown, the outage revealed that
the limestone slurry had completely worn through the rubber lining on one of the units. Both the
transition angle of the chute and the velocity of the slurry are typical of the other slurry piping
within the plant. To rectify the problem, new basalt feed chutes have been procured. The second
example was demonstrated in temporary piping modifications completed to rectify problems with
pluggage of the gas cooling nozzles. A stainless steel elbow that formed part of the temporary
piping, failed due to erosion / corrosion after less than one day in service. The slurry velocities in
the temporary piping were again similar to the remainder of the system.
-------
Vessel Material Selection
The original Request for Quotation issued for the Fluegas Desulphurization system, issued during
the Phase 1 Engineering (DBM), specified 317LM stainless steel for all process vessels. The
removal of the GGH resulted in an increase in the temperature of the fluegas to the Gas Cooling
Section, and therefore the operating temperature of the reactor. Moreover, the native limestone
selected for use as the reagent in the system, contained sufficient chlorides to result in a design
condition of 4500 ppm Ch The combination of chloride content, operating temperature, and
relatively low pH of the system (3-5) in the reactor, resulted in a re-evaluation of the materials
originally specified. As can be seen in Figure 2, 317LM is well below the acceptable limits for
operation at the specified design conditions. Only Inconel 625 or C-276 would be recommended
for this service.
After a comprehensive review of the design conditions, Fluor Daniel metallurgists recommended
that an alternative material should be selected. Several options were considered:
1. Carbon steel vessel with non-metallic coatings (glass, halar, teflon, etc.)
There were concerns with respect to the reliability of the lining, in addition to the
consequences to the vessel structure should the lining fail.
2. Carbon steel vessel clad with C-276.
There were concerns with respect to the application of the cladding, in addition to the
consequences to the vessel structure should the cladding fail.
3. Solid C-276 vessel.
Suitable for the service requirements, however, the cost was prohibitive.
4. Fiberglas Re-inforced Plastic (FRP) with a corrosion resistant veil.
Suitable for the service requirements, and less expensive than solid C-276.
An investigation was conducted to evaluate these material choices as well as any other
possibilities. Existing scrubber applications were surveyed with respect to successful applications,
failures, process conditions, ease of maintenance, and design life. The conclusion of the
evaluation indicated that the FRP construction would have that lowest installed cost, could be
easily maintained, and was suitable for the required service. As a result, all vessels for the project,
including the JBR, limestone grind tanks, limestone slurry tank, gypsum slurry tank, and return
water tank were all fabricated from FRP.
The JBR for the Chiyoda CT-121 process is 68 feet in diameter and 61 feet high. To fabricate this
vessel from FRP there were three available construction methods. The first method would involve
the fabrication of FRP panels off-site, transport them to the field, and assemble the panels into the
vessel. This process is very labor intensive, and was not cost effective for a vessel of this size.
The second method would be to spin a vessel "ring'1 or "can", lift it into place, spin another "can",
lift it onto the first, and continue until the desired height is obtained. This method of fabrication
requires a construction area equal to the finished vessel for construction of the "cans" in addition
to the area where the vessel is finally located. This additional space was not available at the
Suncor construction site. The final method of construction for an FRP vessel is continuous
winding of the vessel shell. As the first portion of the vessel shell is formed, it is raised as the
PageS
-------
51
100,000 -
10,000 -
2 1000 -
100-
winding process continues. As such, the top of the
vessel shell is wound first and is continually raised
until eventually the lowest portion of the vessel
shell is spun. The winding process is continuous,
and therefore, the vessel is fabricated directly on its
foundation.
At the time the contract for the JBR vessel
fabrication was awarded to Composite Construction
and Engineering (CC & E), the largest vessel
fabricated by continuous winding was less than 60
feet in diameter. Thus, upon completion, the JBR
became the largest spun in place vessel in the
world.
Due to the erosiveness of the service, the JBR
vessel was designed to have the components most
susceptible to wear either treated with an abrasion
resistant coating, and / or protected by C-276 wear
plates. It was observed during the shutdown that
areas where C-276 plates were installed suffered
the most deterioration of the coating, but no
obvious damage to the plates. Regions where just
the abrasion coating was applied, suffered minor
thinning of the coating. The design of the abrasion resistant coating and wear plates proved to be
very appropriate for the service.
The one drawback of FRP construction for the JBR, is its temperature limitations. The FRP can
only withstand short durations at temperatures above 450°F before damage occurs. The damage to
the FRP is a consequence of the evaporation of the styrene monomer in the resin. Moreover,
process requirements in the JBR also required the use of PVC components. These PVC
components have a maximum operating temperature of approximately 160°F.
At the Suncor facility, the fluegas leaves the coke fired boilers at approximately 540°F. The gas is
quenched in the gas cooling section to approximately 140°F immediately prior to entering the
JBR. As a consequence, the JBR is dependent upon the gas cooling section for protection.
Although several safeguards are in place to protect the JBR from an over-temperature situation,
one such instance did occur during the first year of operation. Due to localized plugging of a
portion of the nozzles in the gas cooling duct, a high temperature stream of fluegas entered the
JBR uncooled. The automatic safeguards, which have since been modified, did not react to just
one of the multiple temperature probes reporting an alarm condition.
During the brief over-temperature, several PVC components were damaged, and needed to be
replaced during the shutdown. There was also localized damage to a twenty four square foot area
of the FRP vessel wall at the entrance to the JBR. The damage to the vessel wall provided an
Figure 2
Limits of Usefulness of SS and Other
Alloys in Acid-Brine Systems(1'
Page 9
-------
insight into the forgiving nature of FRP. The damaged area was ground down to undamaged base
material and re-constructed with new resin and Fiberglas. The complete repair, including the
replacement of the corrosion veil after the FRP was repaired, was accomplished in one day.
In light of the changes made to the automatic safeguards, it is unlikely that such a temperature
extreme would again pass undetected. Moreover, additional measures have been taken to
minimize future plugging of the gas cooling nozzles. With these retrofits in place, it would appear
that the selection of FRP for the JBR vessel, and other vessels, was appropriate considering the
design of the entire system.
Gypsum
The DBM assumed zero discharge from the plant, with the exception of solid gypsum. The
process envisioned was to centrifugally dewater the gypsum, return the water to process, and
convey the solid gypsum to a silo from which it would be back hauled in the limestone delivery
trucks to the mine for landfilling. The silo was to have 12 hour residence time and should
something happen that would not allow the silo to be emptied, the gypsum would be deposited on
the ground to be removed by a front end loader into trucks for disposal.
A problem arose with this set up. Originally, trucks were to transport the limestone to the plant
site, and the gypsum from the plant site, 12 hours a day. Unfortunately due to the plant's location,
fog was a real threat to shutdown the trucking operation. With the limited capacity of the gypsum
silo, the requirement for the gypsum to be deposited on the ground was very real. The amount of
plot space available to deposit the gypsum was not sufficient to accommodate the quantity
produced in a 12 hour period should trucking stop. Moreover, in the winter months the dewatered
gypsum would freeze and would be difficult to remove.
Using EPRI HSTC, gypsum was generated using Suncor limestone. Codan Associates were
retained as a consultant to perform dewatering studies with various vendor equipment and
determine if there was a piece, or pieces, of equipment that could provide a dry gypsum product
for landfilling. It was determined that a pressure filter could perform that task but it was
imperative that there be two 100% trains for normal operation since the inerts would blind the
cloth after one use. Studies determined that to install the required equipment, the plot space
allotted was insufficient, and the cost was five (5) times that allotted in the budget.
The option of stacking the gypsum existed, however, there was significant concern as to whether
the slurry would completely dewater. Ardaman and Associates were contracted to determine if the
gypsum could be stacked. Gypsum produced at EPRI HSTC, from Suncor limestone, was used.
Their study concluded that the gypsum could be stacked, but the stack may not be stable to build
on. The decision was made to use ponding and stacking as the means of disposal. This relieved the
plot space constraints, but provided new challenges. The gypsum would have to be pumped to a
pond roughly three miles from the FGD plant and approximately 350 feet higher than the plant. It
would require the installation of two (2) trains of four (4) slurry pumps each, plus a gypsum slurry
tank and switching valves. Moreover, to minimize water intake, provisions needed to be made to
return the decanted pond water back to the FGD plant.
Page 10
-------
Operational results proved out the Ardaman findings. The gypsum does dewater, but there is a
sludge from the inerts in the limestone that does not dewater as readily as the gypsum. It would
appear that the CT-121 process produced gypsum actually dewaters better than the gypsum
produced at the HSTC and tested by Ardaman. This may be because of the larger gypsum crystal
size typically produced through the CT-121 process, compared to the tower design, or it could be
due to the handling of the gypsum liquor prior to Ardaman receiving the sample. Overall, the
operation of the gypsum facilities to date indicates that there is a clear supernatant that can be
returned to the FGD process for use as make up water. However, it is still too early to determine
the long term stability of the stacked material.
Though it is not part of this paper, it should be noted that the use of native limestone will bring
with it impurities that may have an environmental impact. Depending on their concentration in the
limestone, some elements can concentrate in the gypsum to hazardous levels. A study was
conducted on Suncor's limestone, and it was determined that no elements were concentrated
beyond the limits established in the Provincial Emissions Guidelines.
Wet Stack
NELS Consulting Services was commissioned to build and test a V12th scale model of the stack to
determine its flow characteristics to allow for the accurate placement of water collection devices.
As the project progressed, the scope of this model study increased to include all the new flues
complete with dampers, the JBR, and the existing stack. The model was also used for:
1. Obtaining flow profile information on the wet stack.
2. Including a flow straightening device to reduce recirculation in the inlet flue of the JBR.
3. Placement of the emission probes in the wet stack, inlet duct and at the ID / booster fans.
4. Verifying the suitability of using the existing stack for bypassing the FGD facility.
Due to the process requirements for the stack, it was determined that the most suitable material for
the liner was C-276. Solid C-276 was used in areas of anticipated high wear and the stack tip. It
was also used at the CEMS level liner penetrations where it was felt that the dilution of the welded
material would provide a site for corrosion. Carbon steel clad with C-276, was used throughout
the remainder of the stack liner.
Once the decision was made to use FRP for the reactor vessel, it was also selected as the material
for the ducting out of the reactor, and the mist eliminator. During a review of the startup and
shutdown of the unit, questions arose with regard to the design pressure of the ducting. The
subsequent investigation revealed that:
1 When bypassing all the hot flue gases to the wet stack, the draft generated was greater than
the vacuum design pressure of the FRP ducting and the JBR.
2. The hot bypass gases could migrate back into the JBR during a complete bypass.
One of the solutions proposed for the bypass problems identified was to leave the "old" stack in
service for the gas fired boilers and also use it for bypass. However, there was concern that the
Page 11
-------
"old" stack would not be suitable for servicing the gas fired boilers only, due to condensation in
the liner eventually causing it to fail. NELS study indicated that the "old" stack could be used for
the gas fired boilers only without fear for the integrity of the liner. With this knowledge, the
decision was made to bypass to the "old" stack. Bypassing to the old stack also removed the high
temperature process requirement for the liner, making alternative materials such as FRP suitable
for use. Unfortunately, this decision was made too late in the project to allow a change stack liner
materials and it remained the more costly combination of solid and clad C-276.
Operation of the FGD facility has shown that the water collection system within the stack operates
as intended. However, during the winter months, ice forms on the stack cap. This problem is
significant as the stack is located adjacent to the refinery. Temporary radiant heaters were installed
directly under the cap which have proven effective in controlling this problem. A more permanent
solution being installed for the 1997 / 1998 winter.
ID /Booster Fans and Dampers
The Energy services building at Suncor houses three (3) Foster Wheeler forced draft, delayed coke
boilers, each designed for 750,000 pph of steam at MCR. At the back of each boiler, there is an
airheater, and two (2) electrostatic precipitators (ESP). To accommodate the increased flue gas
head required by the addition of the FGD unit, some type of booster fan or fans had to be added
between the boilers and the scrubber.
The DBM included for the installation of one motor driven axial booster fan handling the flue gas
for all three boilers. As noted previously, a dynamic simulation was required to determine how
best to control this arrangement. The simulation was performed by the Fluor Daniel Irvine,
California office using parameters established by the Calgary office. The model indicated that this
single fan arrangement would be very difficult to operate and would most likely result in the
boilers tripping off line each time a bypass was attempted. It also indicated that the upset could be
of sufficient magnitude to cause a furnace implosion. As forced draft units, the ability of the
boilers to withstand a negative pressure in the furnace was minimal, only -7" w.c.
These results, along with a number of other factors, had to be considered in the selection of the fan
itself. The design flue gas temperature (585°F), the particulate loading (ESP efficiency is
approximately 85%), and the requirement for 98% reliability, also questioned the suitability of this
single fan arrangement. Furthermore, fan vendors were indicating that the service requirements
were pushing the design limits of their equipment. After debate and research, it was determined
that a better approach would be to use individual booster fans for each boiler. This arrangement
was proposed to Suncor identifying the reasons, and highlighting the opportunity made available
to convert the boilers to balanced draft operation. Balanced draft would eliminate the problem of
leakage of fluegas typical of forced draft boilers. The decision was made to proceed with the a
centrifugal fan for each boiler and to convert the boilers to balanced draft operation. For maximum
control, steam turbine drives were selected for two of the fans, and a VFD for the third.
Unfortunately, this new configuration caused a complete rework of the flue arrangement, a
reconfiguring of the dynamic simulation model, and the FGD project control scheme now had to
include converting the boilers to balanced draft.
Page 12
-------
The new simulation model still indicated that the transfer of flue gas into and out of the FGD unit
had to be handled carefully. The gas path through the JBR has a head loss of about 23" w.c., while
the bypass path has a few inches at most. If the transfer was not properly controlled, this large
difference in resistance could cause a large furnace draft excursion. The solution was to install a
control damper in the bypass duct which would permit the ID fan backpressure to be controlled.
To transfer the flue gas into the FGD unit, the ID fan backpressure is slowly raised by closing the
bypass damper until there is essentially no pressure differential across the JBR inlet dampers. The
JBR inlet dampers are then opened. The transfer is completed by raising the backpressure
controller setpoint. In response, the controller will continue to close the control damper in the
bypass duct. When it is fully closed, the flue gas transfer is complete and the bypass guillotine
damper can be closed. At this point, the gas path to the FGD unit is completely open, while the
bypass gas path is completely closed.
Transfer of flue gas flow out of the JBR is completed automatically after either an operator request
or an Emergency Shutdown signal. When a transfer is initiated, logic first ensures that the bypass
louvre damper is closed. The bypass guillotine damper is then opened. The ID fan backpressure
controller switches to transfer mode, in which it attempts to maintain the current backpressure
value. When the bypass guillotine damper is 50% open, the JBR inlet guillotine and louvre
dampers both start to close. As these dampers close, the ID fan backpressure will tend to increase
because the flow path is being closed off. The bypass damper will therefore open just enough to
keep the ID fan backpressure where it is. This will continue until both JBR inlet dampers are fully
closed. The flue gas path transfer is then complete. The bypass louvre damper will be partially
open and the ID fan backpressure should still be at the same value as when the transfer started.
The operator can gradually open the bypass louvre damper to allow the ID fan speed to decrease to
minimum.
Field results indicate:
1. The bypass dead leg portion of the ducting, (a portion of the original duct), is deteriorating.
There has been leakage past the guillotine damper allowing SO2 rich flue gas into the
"cold" space. The seals are being replaced and repairs to the ducting are being investigated.
2. The control system for the hydraulic guillotine dampers has not performed as well as
anticipated. Fabrication errors in the positioning system have since been discovered.
Replacement of the wrong components was undertaken during the shutdown.
3. Mechanical defects in the steam turbine drivers have limited the number of boilers that
could be scrubbed. For most of the first year, the JBR was handling only two boilers, due
to the turbine failures.
4. With two boilers through the JBR, numerous on line bypasses have be successfully
completed. The bump experienced on bypass is slightly less than predicted and easily
managed by the control system. The boilers have also been bypassed when there was a
higher than normal pressure drop in the JBR, and this also was handled without exceeding
the control limits. Unfortunately, there has not been the opportunity to perform a test on
three boilers due to the difficulties experienced with the fan drivers.
Page 13
-------
Structural
In the northern parts of Canada it is best to provide a shelter around equipment containing water.
This was done providing space not only for the process equipment but also electrical switch gear,
control racks, HVAC, and a maintenance and storage area. The equipment shelter was divided
into three process areas. These were the JBR shelter, the limestone shelter and the gypsum shelter.
Due to the enormous size of the components within the shelters, for example the JBR and
limestone slurry tank, an unobstructed floor plan was required for each the shelters. The result is
that each shelter is similar to an aircraft hanger. All internal building steel provides support for
only the equipment, platforms and piping.
Detailed Design / CADD
Originally, the contract with Chiyoda included the equipment and process piping, with FDCI
responsible for the buildings, the auxiliary equipment and utility piping. The decision was made to
use Intergraph PDS 3D CADD for the project to support project cost goals and the very tight
schedule. FDCI has used this CADD tool with a great deal of success and has seen significant
reductions in field rework and, therefore reduced field cost and schedule. Unfortunately, there was
no compatibility between Chiyoda's and FDCI's CADD platforms. It was, therefore, decided that
Chiyoda would provide routing guidelines and check the process piping, and FDCI would do the
detailing of the systems. This allowed the complete integration of the equipment, piping, structural
design and plot layout which proved to be a very important decision for the project. An example
of the benefits of the 3D CADD design was most apparent with the slurry piping.
Chiyoda stipulated that cast basalt pipe be used for all slurry piping because of the high inerts in
the limestone. They also included requirements for free draining of the slurry system along with
minimum radii of bends. Moreover, the size of the pipe, and the available fittings greatly defined
the way the pipe could be run. Further, basalt pipe can not be field welded to adjust for
misalignment or piping design errors. As a result, it was imperative that the piping design be
dimensionally precise. This extended to precision erection of foundations and equipment. As
fabrication tolerances in FRP construction, such as for the JBR, are much greater than those for
steel construction, the final nozzle positions were determined following installation of connecting
equipment and erection of the basalt piping, ensuring a precise fit up.
The field performance has proven the decision to use the 3D CADD for all civil / structural,
piping, equipment, and electrical. Rework in the field was minimal, with virtually no interferences,
greatly supporting the achievement of the project cost and schedule goals.
Conclusions
1. The Chiyoda Thoroughbred CT-121 process performs very well, even in the severe design
conditions at Suncor. The guaranteed 95% SO2 absorption has been met and continuously
exceeded. There have been some plugging problems of the gas cooling section that has
required the boilers to be bypassed and maintenance to be performed. Some of the
pluggage has been from foreign matter, but some has been a result of relatively thin scale
Page 14
-------
flaking off from equipment surfaces and plugging the nozzles. Pluggage due to scale is
becoming less prevalent as operating experience increases.
2. The equipment has performed well except for mechanical defects with the ID/ booster fan
steam turbine drives that have limited the FGD plant availability.
3. The wet stack has experienced icing problems. Suncor is installing a solution presently.
4. The FRP reactor has performed well, although the material's temperature limitation results
in the FGD system being dependent on the gas cooling section for fail safe operation.
Moreover, the inerts from the native limestone do not adversely affect the integrity of the
vessels or ducting.
5. Basalt piping has performed very well, showing no signs of erosion.
6. The FGD fluegas control system has performed very well. The bypassing operation can be
performed, with confidence, without tripping the boilers. As a result, even though the
FGD facility has not maintained its 98% availability, steam production was never lost
during the first year of operation.
7. Low quality, local limestone is suitable for use in a wet scrubber. The resultant gypsum
does dewater. The inerts do not settle out quickly, but the gypsum seems to have a positive
influence on their settling rate. The gypsum has also proven to be a useful product for the
consolidation of the extraction tailings; a benefit not initially anticipated.
8. The use of 3D CADD proved to be an effective tool to establish the design layout and
provide precise material take-offs. It helped maintain an aggressive construction schedule
and reduced field re-work.
9. The on-line factor (98%) has not been achieved during the first year of operation.
Mechanical problems with auxiliary equipment, not the process, has been the cause of the
poor availability.
References
1. C. M. Schillmoller, "Alloys for Flue Gas Desulphurization Scrubber Systems," Nickel
Development Institute Reference Book, Series No. 11001, 1986, page 97.
Page 15
-------
FOR AN OIL FIRED BOILER IN WERNDORF/A USTRIA
Klaus Bamthaler
Wolfgang Guggenberger
Austrian Energy & Environment
Waagner-Biro-StraBe 98
8021 Graz, AUSTRIA
Wolfgang Kindlhofer
STEWEAG
Leonhardgflrtel 10
8010 Graz, AUSTRIA
Abstract
In the year 1994 the order for the retrofit of the power plant Wemdorf in Austria was placed with
Austrian Energy. The job includes the upgrading of the boiler and the installation of a wet limestone
.FGD. The main requirements for the FGD were a SC>2 reduction of more than 96% and a maximum
emission of duct and aerosols (as SOi) of less than 5 [mg/Nm3] each. The product gypsum will be used
for the cement industry. To fulfill especially the very strict requirements for dust and aerosols emission
a system including Prescrubber, Scrubber and Wet ESP was chosen. The construction was finished in
December 1996, first flue gas will be generated by the end of February 1997. In this paper the design of
the FGD and results from the first three months of operation concerning SO?, SOs and dust removal of
the system will be presented.
Introduction
The "Styrian Water and Electricity Power Co." (STEWEAG) placed an order with AE in July 1994 for
revamping an existing oil boiler and for erecting a flue gas desulphurization plant including dust
separation for the Wemdorf Power Plant. The total value of the order for the desulphurization plant was
about 55 million US $ subject to a project handling time of 30 months.
Please refer to Table 1 for the most important data:
-1-
-------
Table 1
Boiler Wemdorf
Fuel: Heavy fuel oil (natural gas for auxiliary firing)
Sulphur content: 0.86 up to 3%
Fuel heat capacity: 480 MW
Electric power: 165MW
Discharged district heat: 200 MW
Live steam pressure: 186 bar
Live steam temperature: 540°C
The steam generator is executed as a two-pass forced circulation boiler - System Benson - with single
reheat. The boiler plant is operated with oil/gas mixed-fuel firing. Via the steam turbine a distant heat
decoupling is effected being tied into the existing long-distance heating system of the Mellach district
heating power station.
Design Basis
The plant has been based on a raw gas composition as shown in the Table below:
Table 2
Raw gas Werndorf
Volume flow: 485 000 (NmJ/h) wet
Temperature: 169°C
SO,: < 4 928 mg/Nm3 dry
SO3: < 350 mg/Nm3 dry
Ash: < 350 mg/Nm3 dry
CO2: 13.3 Vol.% dry
O2: 3.6 Vol.% dry
Due to the position of the power plant being situated in a local recreation area of the Styrian Capital
Graz and the sensibilization of the resident population for environmental affairs the operator was
required to met very stringent guaranteed values for the clean gas which are partially far below the limit
values demanded by the Authorities. The limit values for waste gas are shown in Table 3.
Table 3
Guaranteed values
SO2 from chimney: < 200 [mg/Nm3] dry
SO3 from chimney: < 5 [mg/Nm3] dry
Dust from chimney: < 5 [mg/Nm3] dry
Chimney temperature: > g5°
-------
Table 4
CaSO4 x 2H2O
CaSO3 x '/j H2O
CaCO3
Soot
Residual moisture
< 0.5%
< 0.3%
< 10%
The waste water of the FGD is passed via an outfall ditch to the nearby river, where the limit values
were defined as follows:
Table 5
Waste water limit values
Temperature
pH value
Substances which can be filtered out
Vanadium
Nitrite
Chloride
Fluoride
Sulphate
Sulphide
Sulphite
Iron
Mercury
Nickel
TOC
SCB
30°C
6.5 - 8.5
15mg/l
0.5mg/l
l.Omg/1
1 250 mg/1
S.Omg/l
2 500 mg/1
0.2 mg/1
20.0 mg/1
2.0 mg/1
0.01 mg/1
0.5 mg/1
45 mg/1
140 mg/1
SO3 and dust limit values
By demanding stringent limit values for dust and SO3 emissions the customer would have liked to ensure
that no aerosol plume - as is usually the case downstream of oil boilers with FGD - will be visible. The
high sulphur content in the fuel oil of up to 3% and the existing heavy metals, such as vanadium, have
the effect of a high conversion rate from SQz to SO3 in the boiler. This effect is additionally reinforced
by the installed DeNOx catalyst. Since, as is well-known, SQs and/or the sulphuric acid aerosols formed
of the same are separated to a very low percentage in the wet scrubber only, an additional technology
was required for SO3.separation. An analysis showed that, in principle, 3 technologies can be used for
the separation of SO3.
-3-
-------
Admixing of additives
Additives, such as Ca(OH)2 and or NH3 can be added at certain points to the boiler and/or to
the consecutive flue gas duct and bind the SO3 from the flue gas. In a subsequent filtering
equipment (ESP or baghouse) the formed product (calcium sulphate or ammonium sulphate)
will be separated.
2. Fine scrubber
By injecting a fine droplet mist downstream the FGD scrubber the formed sulphuric acid
aerosols can be bound and thus removed from the flue gas. An adequate technology has been
developed by VAI for fine dust separation downstream of sintering plants and has been adapted
by Austrian Energy for the separation of sulphuric acid aerosols.
3. Wet electrostatic precipitator (WESP)
For that the sulphuric acid aerosols are separated in an electrostatic precipitator being arranged
behind the flue gas scrubber.
Additive dosing was declined by the customer for fear of contamination and/or problems with
disagreeable smell caused by NH3. The fine scrubber has the disadvantage of a very high energy
demand for finest distribution of scrubbing fluid compared to the WESP, and moreover references for
this kind of application are lacking. Especially due to the proven technology a WESP has been chosen
for separating sulphuric acid aerosols and fine dust
Flue gas desulphurization plant
Process Design
For fulfilling the customer's requirements a plant was designed corresponding to Fig. 7 "General Flow
diagram" in the Annex.
The flue gas is cooled in a raw gas cooler to about 100°C. The energy gained from that is used for
preheating the combustion air and/or boiler condensate increasing, therefore, the entire boiler efficiency
by about 1%.
After cooling the flue gas is passed to a prescrubber. in which above all dust and soot, as well as a
small portion of SO3 are separated and where the flue gas is cooled down to saturation temperature.
Dust and soot separation was necessary to be able to achieve the required gypsum quality in the main
scrubber. Compared to the dry electrostatic precipitator the prescrubber stands out for lower investment
-4-
-------
Table 6
Main data of Prescrubber
Cross section
Height 19.075m
Spray effects 2
Nozzles / Spay level 30
Recirculated volume per spray effect 1 000 m3/h
Superficial velocity 4.0 m/sec
Sump volume 50m3
Recirculation pumps 1 x 2 000 m'fh
Especially for location-based advantages the prescrubber was designed as counterflow equipment with
2 spray effects. For the most important data of the prescrubber please refer to Table 6.
A two-layer rubber-lining of chlorine butyl caoutchouc was chosen for corrosion protection. At a level
of about the scrubbing slurry an approx. 1 m high PTFE foil has been applied to the rubber-lining. This
additional layer shall prevent that hydrocarbon constituents of the oil fly ash will attack the rubber-
lining. The recirculation lines are made of glass fibre reinforced plastic and the spaying effects are
coated with Arbosol (polyvinyl chloride). A single-layer mist eliminator prevents the entering of acid
and contaminated prescrubber slurry into the main scrubber.
The SO? scrubber proper consists of a cylindrical sump with an attached, rectangular absorption part.
Out of the 4 installed spray effects 3 effects are in operation at nominal load (100% boiler load with 2%
of sulphur in the fuel) to achieve the required clean gas concentration. At a sulphur content of 3% the
forth spray effect is also taken into operation.
The rectangular type of structure has the advantage that the spray header can be arranged in a simple
manner outside of the scrubber and that only spraying lances are projecting into the scrubber.
Table 7
Main data of scrubber
Diameter of sump llm
Sump height 11.698m
Agitators 3
Oxyair system Lance gas injection ahead of agitators
Cross section of absorption part 6.2 x 7.66 m
Height of absorption part 19.482m
Superficial velocity 3.6 m/sec
Spray effects 4
Recirculated volume per spray effect 2 300 m /h
Nozzle type Spiral hollow cone nozzles
Number of nozzles per spray effect 32
Mist eliminator 2 double roof mist eliminators
Scrubber material Carbon steel rubber-lined (in sump and near
the spray effects with double rubber lining)
Material of spray effect Coated with Arbosol
-5-
-------
The flue gas is subsequently passed to the wet electrostatic precipitator in which both the sulphuric
acid aerosols and the residual dust are removed from the flue gas. At the inlet of the wet electrostatic
precipitator up to 115 mg/Nm3 of sulphuric aerosols and about 30 mg/NmJ of dust are still existing. The
dust and aerosol particles are ionised by the discharge electrodes and separated at the collecting
electrodes being arranged in parallel with the gas flow. The cleaning of the electrodes is continuously
effected by injecting basic scrubbing slurry which is kept constantly at a level of pH = 12 * 13 by
dosing in a 10% slurry of Mg(OH)2. This relatively high pH value is required to protect the collecting
electrode against corrosion attacks.
Another criterion for the mode of function of the WESP is that the scrubbing slurry cycle should be kept
as far as possible free from any solids. This requirement is met by continuously adding fresh water and
discharging the same amount for mist eliminator flushing in the main scrubber. The last WESP-field is
flushed only temporarily with fresh water to minimize the droplet discharge from the system.
Hot air of about 320°C is used for reheating the about 55°C cold flue gas. This hot air is generated in a
tubular air heater and is mixed with the saturated flue gas in a venturi mixer.
The gypsum produced in the main scrubber is prehydrated to about 55% via a hydrocyclone stage and
subsequently dehydrated in the belt filter to 10% residual moisture.
In addition the belt filter is charged with the sludge from the prescrubbing stage and possibly by the first
separation stage of the waste water cleaning plant. The quantity of the additionally injected waster water
sludge depends on the requirements of the FGD gypsum. The entire filter cake is washed with fresh
water in order to remove the solvable constituents, especially S042" and Mg2+' from the filter cake. The
dehydrated gypsum is transported via a feeder to the wet gypsum silo, which was designed for a storage
time of 14 days.
The overflow of the first hydrocyclone stage is cleaned to about 1 % of solids content in a second
cyclone stage and passed into the cycle water tank. From there recycle water is reinjected into the
prescrubber thus compensating the water losses due to flue gas saturation and also the absorbent mixing
is being fed
Limestone is used as a absorbent having a grinding fineness of 90% < 63 urn (170 mesh). The
limestone is stored in a silo having a storing time of 14 hours and is mixed with cycle water to a slurry
density of 20%. The limestone slurry volume dosed into the scrubber is calculated from the emerging
SO2 concentration of the flue gas and volume flow as well as from the measured slurry density and
dosed in via a control valve.
The waste water cleaning plant has been designed for waste water treatment of the FGD, for the boiler
area (total demineralization, condensate treatment and various dehydration processes) and the DeNOx.
The cleaned waste water is passed into the river Mur (a nearby river).
-6-
-------
The scrubbing slurry from the flue gas cooler and prescrubber (lOSnrVday, feeding volume 4.5 m3/h),
the waste water from the condensate treatment (40m3 within 3 weeks, feeding volume 0.25 m3/h) and the
waste water from the annual boiler washing (100m Vyear, feeding volume 2.5m3/h) are treated in the
waste water cleaning plant
The waste water cleaning plant is designed as a precipitation plant consisting of three stages for removal
of the heavy metals and of the sulphate load provided with a pre-sedimentation stage. The sour waste
water pH = 1.5 from the prescrubber is stored in a buffer tank and pre-neutralised with limestone slurry
to a pH of 3.5. During this pre-neutralisation the sulphuric acid being separated in the raw gas cooler
and/or prescrubber is transformed to gypsum. The sludge from the pre-sedimentation thickener contains
mostly gypsum and can thus be fed to the belt filter without deteriorating essentially the quality of the
gypsum.
In the next two stages the heavy metals (especially Va, Ni, Co) and the sulphate (above all MgSO4) will
be precipitated by milk of lime and by rising in steps the pH-value to pH = 12. A stripping plant will
remove the existing ammoniac from the waste water and will produce ammonium sulphate of about
20%.
After ammoniac stripping a third sulphate precipitating stage is installed with calcium aluminate. This
stage is absolutely required to achieve the sulphate limit value (SC>4 < 2 000 mg/1), since the sulphate
load is considerably increased in this FGD plant by the Mg(OH)2 being introduced through the WESP.
The sludges from the individual precipitation stages are collected in a sludge stacking tank and -
depending on the required quality of the gypsum - partially admixed to the FOD gypsum or completely
.pressed out in the filter press and deposited.
Plant layout
The plant arrangement was essentially preset by the FGD main components, fans, flue gas cooler,
prescrubber, scrubber and electrostatic precipitator.
The FGD fan and the flue gas cooler are arranged at level + 0.0 m resulting in a place-saving
counterflow principle for the prescrubber and scrubber. An essential criterion for the design of the FGD
building was moreover the accommodation of the wet electrostatic precipitator, which had to be
constructed adequately large to achieve the low separating rates. However, these preset conditions
determined the size of the enclosure.
The limestone silo and the wet gypsum silo were positioned to the West to accord free access for trucks
for filling and discharging both tanks. The storing capacities of the limestone and wet gypsum silos were
designed for a period of 14 days at 110% load and 2% sulphur.
For revision work during operation a discharging vessel was provided for which can take up the total
content of liquids of the plant.
-7-
-------
Main components
FGD fan
Behind the boiler air heater the single-stage FGD fan is provided for overcoming the pressure losses of
the subsequent plant components in the FGD plant.
The axial I.D. fan is arranged upstream the flue gas cooler on foundations insulated against vibrations;
load control is effected by means of hydraulic blade-pitch variation. The fan is driven by an external
electric motor (P=1200 kW) via an intermediate shaft with coupling. An intake hub on the suction side
and straightening guide vanes being welded behind the impeller on the pressure side are providing for an
inflow and outflow with as little loss as possible. To avoid any vibrations the active fan pan is
connected on the suction and pressure side with the flue gas duct via expansion joints.
Raw gas cooler
The flue gas flow is fed to the raw gas cooler via the inflow hood. Three heat exchanger bundles are
arranged in the casing of the flue gas cooler one beside the other. The bundles being in contact with flue
gas are made of Teflon hoses, whereby conditioned deionized water is employed as heat carrying
medium. The casing is made of structural steel; the inside surfaces of the casing in contact with the flue
gas inclusive of the outflow hood and a part of the inflow hood are lined with PFA foil. Also the flue
gas duct from raw gas cooler outlet to prescrubber inlet are protected against sulphuric acid corrosion
by means of PFA foil.
The heat exchanger bundles are cleaned in certain intervals by means of a stationary arranged washing
system, which is also projecting into the interior of the bundle. Waste water from the wet electrostatic
precipitator is used for washing; the washing installations are rewashed with make-up water.
Fig. 1 Flue gas cooler
-8-
-------
Wet electrostatic precipitator (WESP)
The filter housing consists of a steel sheet structure, where the interior is provided with a vinylester
coating and the exterior is equipped with an insulation.
The separation of duct and aerosol is effected in two fields being arranged on behind the other. In the
fields the collecting electrodes and the discharge electrodes are alternatively arranged in parallel with the
direction of flow so that 44 lanes are formed in total. The discharge electrodes are suspended with post
insulators free from the WESP roof; the collecting electrodes are firmly attached to the filter housing.
The electrodes are made of special steel 1.4404/45.
Washing nozzles are mounted above the electrodes in the center of the lanes which facilitate cleaning of
the collecting electrodes.
The discharge electrodes are fed from two transformers positioned on the roof of the WESP and
rectifiers transforming 380 V alternating current into 60 kV direct current-
Fig. 2 Wet electrostatic precipitator
Prescrubber circulating pumps and scrubber circulating pumps
These centrifugal pumps are executed as single-stage, single-flow spiral casing pumps in process
construction with an open impeller. The process construction allows dismounting of the pump impeller
with the pressure-side wear insert, stuffing box body, shaft seal as well as bearing supports without
loosening the suction and pressure line.
The choice of a high-quality corrosion and acid resisting material (1.4460 Cu) for the pump casing,
impeller and wear inserts, the low relative velocity (low abrasion) as well as the wear inserts before and
behind the impeller has well proved in the commissioning phase.
-9-
-------
Product dehydration
Two cyclone stations and two downstream redundant belt filters are provided for the gypsum
dehydration.
The gypsum slurry withdrawn from the scrubber is fed to the product cyclone station with an inlet
pressure of about 1.8 bar. According to the supplied slurry volume individual cyclones of type Dorr
Oliver PZ 100/15 of the six in all installed cyclones can be interconnected or cut out via isolating
dampers provided with electric actuators. To protect the hydrocyclones against plugging a coarse grain
separator is arranged upstream the cyclone station.
The slurry emerging from the underflow discharge of the gypsum hydrocyclone being predehydrated to
a solids content of about 55% is passed to the respective belt filter via a distribution station supplying
the first feeding chute. Waste water sludge can be supplied to the second feeding chute according to the
requirements.
The liquid contained in the predehydrated gypsum slurry is sucked off through the belt filter (vacuum
chute). To decrease the chloride content in the filter cake to 100 ppm a single-stage cake washing with
two scrubbing slurry feeding chutes are provided for with water from the vacuum pump. In the
subsequent drying zone (residence time about 90 sec.) the scrubbing slurry is sucked off and the water
bound in the pores due to adhesion forces is entrained by the air flow up to a reminder of about 10%
and discharged
On the drive end the filter cloth is passed off from the carrying belt and the filter cake is dumped at the
crusher roll to the feeding chute of the feeder.
The feeder is a continuous steep conveyor with a shaft edge belt transporting the gypsum to the feeding
equipment of the wet gypsum silo. From there the gypsum is fed to a pipe screw conveyor via a rotary
chute which dumps the material to the telescopic chute. The telescopic chute ends in a distribution screw
which stores the resulting gypsum uniformly by turning round the silo axis.
For discharging the product the window being assigned by the respective bulk height will be opened at
the center column, the distributing screw is changing its direction of rotation and delivers the gypsum to
the center column. At the foot of the center column a loading equipment is provided for open trucks.
The design of this gypsum storage facilitates simultaneously feeding and discharge operation.
Project management
On October 24, 1994 the commodity contract was signed with the customer so that the detail design of
the plant could be started.
The distribution of work was effected according to the following pattern:
Civil works were made by STEWEAG, structural steelwork and components of the FGD plant as well
as boiler by AEE, a new chimney pipe of reinforced glass fibre was installed by Messrs. K.VS / Integral,
and Messrs. SIEMENS supplied the complete control and instrumentation equipment of the plant
-10-
-------
The very tight time schedule (about 14 months erection time) did not allow for any major delays in
engineering and erection. However, all setting work for the individual components has already been
started in December of 1996 so that the hot commissioning date of mid February 1997 was not
endangered seriously.
A survey of the erection work can be seen in Fig. 4 and 5 in the Annex.
Commissioning
Due to delays in erection the first flue gas from the boiler was available not before mid of March, 1997
with a deferment of one month. At the same time with the FGD the boiler was taken into operation so
that during the first month no longer periods of operation of the plant could be achieved since the boiler
was switched over every now and then to natural gas operation. From mid April the boiler was operated
according to the requirements of the FGD with about 2% heavy fuel oil (corresponds to about 3 500
mg/Nm3 of SO2). Due to the seasonal summer standstill of the Austrian caloric power plants 3 weeks
were available for commissioning and especially for optimization.
The first operating results showed that the SO^-separation of the scrubber did not comply with the
guaranteed values. Measuring the collecting efficiency of the individual spray effects showed following
results:
Spray effect
1
1+2
1+2+3
1+2+4
Collecting efficiency
60%
78%
87%
91%
Obviously the collecting efficiency of 3 recirculation pumps in operation did not meet the expectations
(>95%). Moreover, the measurements showed clearly that the third spray effect had an essentially worse
collecting efficiency than the other levels.
For this reason and due to a heavily fluctuating clean gas SQz concentration plugged spray nozzles were
soon detected to be the cause for the bad collecting efficiency. However, shut down and cleaning of the
plant was at that time dispensed with so that the commissioning of the other plant components would
not be endangered. Instead, the SCh. profile across the scrubber of mist separators was measured which
showed clearly a maximum in the rear comer of the scrubber (as viewed from the inlet duct). As already
known from other plants also in this scrubber an increased SOi concentration was detected in the
margin area where the decrease towards the center of the scrubber was not very marked. After finishing
the first commissioning phase the inspection of the plant showed that the GFR mountings of the oxyair
lances were broken and in sum up to 30% of the spray nozzles were plugged. The maximum number of
plugged nozzles was to be found in the area of the maximum SOi concentration.
-11-
-------
As a consequence of these results both the oxyair mountings were reinforced and also the recirculation
lines were equipped with strainer baskets.
In contrast to that the SO3-values with about 2 (mg/Nm3) measured in the chimney did come up to
expectations. However, so far SO3 raw gas concentrations of about 30 (mg/Nm3) were measured only
owing to the clean boiler condition and the DeNOx catalyst not in operation yet. However, the aerosol
plume was markedly visible when the WET-ESP was cut out which disappeared after taking the filter
into operation.
Tables 8 and 9 show the achieved gypsum quality and the measured limestone purity. The requirements
of the inspectors were complied with by far which on the one hand is due to the good limestone quality
and on the other to the satisfactory function of the prescrubber.
Table 8
Measured gypsum quality
CaSO4 x 2H2O
CaS03 x '/2 H,0
CaCO3, MgCO3
Inerts
Soot
97.4%
0.04
1.4
1.06
<0.1%
Table 9
Measured limestone quality
CaCO3
MgC03
Inerts
98.5
0.8
0.7
The water volume in the waste water plant discharged from the prescrubber is conspicuous of a high
concentration of soluble salts (see Table 10).
Table 10
Measured waste water composition
SO,2'
cr
Ca2'
Fe
Mg2"
Na+
Ni2-
V
[mg/1]
[rag/1]
[mg/1]
[mg/1]
[mg/1]
[rag/I]
[mg/1]
[mg/1]
Waste water cleaning inlet
38200
115
606
322
5790
176
101
284
Waste water cleaning outlet
1600
140
766
20.1
20
121
<0,1
0,2
This is mainly due to the additional SO2-separation in the WET-ESP and, as a consequence, high
magnesium supply into the system. In spite of that the required waste water limit values could be kept
after optimization of the plant and an increased operation of the calcium aluminate precipitation.
Moreover, it became obvious that sludge from the pre-sedimentation installation could be added without
deteriorating the quality of the gypsum. Thus, the sludge volume to be disposed could be essentially
reduced.
-12-
-------
Otherwise, no damages worth mentioning could be detected in the plant, where a final statement
especially regarding the corrosion protection could be made after a longer operating period only. After
installation of the scrubber filters and some optimizing work at the WET-ESP and the waste water plant
the FGD will be again taken into operation at the beginning of September 1997. PAC is intended for end
of October.
Fig. 3 Aerial view of boiler house and FGD
-13-
-------
Fig. 4 Wet electrostatic precipitator
-14-
-------
KCUwrmxJ
.nueju*>mp«r; '• FOD-fal • ilUwiucooler)
-------
o
3
o
a.
i wi'.-:::' .
>v/'.ID< PI Mft
'--' Trillin'.?! FF'I\!'-F.
HlliK'V l.-'JIli H' I'KIA i' Li'l
COOIIIIC WATFR
JOXUIIOH
-OiHTAir.R
Z?L
IM
:IUNE
jIL'J
REOC.i PUMPl?Kn,LE F'UM=- SLURPr FUHP
RtriROULAHOII
TMK
UllESIWIE
;LU:I:!I Bin.
^T^ >/ATCR "RCAIMCMF TLAJIT
H;-J.i,
-------
Wednesday, August 27; 8:00 a.m.
Parallel Session A:
High Velocity FGD Systems
-------
SO2 COMPLIANCE AFTER 2000:
SLAM DUNK OR SOMETHING ELSE ALTOGETHER?
J. Platt
Electric Power Research Institute
3412 Hillview Ave.
Palo Alto, California 94304
Abstract
Phase 1 SC>2 allowance prices have fallen below expectations, fueling enthusiasm for
market mechanisms and easing concerns about meeting Title IV requirements. SOz
compliance, for many, is no longer a pressing issue, replaced by NOx reductions,
major revisions to National Ambient Air Quality Standards and ongoing upheavals
of deregulation. But future SO2 reductions are not so far off or trivial that they can
be overlooked. This paper first summarizes how Phase 2 is likely to be very different
from Phase 1, especially once the bank of allowances built up during Phase 1
approaches exhaustion. Examining a variety of uncertainties, a recent EPRI study
shows that longer term compliance costs and additional scrubber retrofits hinge
primarily on how intensively existing coal plants will be used. High utilization is
more realistic than low, greatly increasing needed reductions, yet consumption of
banked allowances in the years after 2000 and uncertainty about allowance market
development create dilemmas for scrubber investment decisions. The paper next
indicates how tighter SC>2 restrictions proposed last year (or perhaps resulting from
this year's announced PMi.s standards) could lead to extreme costs, aided little by
efficiencies from emissions trading seen under more modest standards. Finally, the
paper questions comparisons between industry's early estimates of Title IV
compliance costs and today's allowance market. Such comparisons inflate what
industry said, and diminish what Phase 2 could cost by focusing on today's
conditions instead. The chief reference for this paper is SOz Compliance and
Allowance Trading: Developments and Outlook, EPRI TR-107897, April 1997.
Introduction
EPRI has conducted research since the mid-1980s on the implications of proposed
and actual clean air policies.* While frequently wrong, this succession of analyses
has been updated numerous times to keep pace with ongoing changes and it has
" Three studies were initiated since the Clean Air Act Amendments. See reference section.!/ 2/ 3
-------
clarified key drivers affecting coal prices and sulfur premiums, shifts in regional coal
demands, the use of retrofit flue gas desulfurization (FGD) units, and the emergence
of active emissions trading. The most recent EPRI report in this series is cited
above.l This report examined 1995 SC>2 allowance market developments', the
buildup of the bank of allowances during Phase 1, and other factors affecting
industry's SC>2 reduction requirements and marginal costs of compliance in Phase 2.
It explored possible impacts on SC>2 reduction requirements and costs associated
with much more stringent SC>2 restrictions, basing this analysis on EPA's 1996 Clean
Air Power Initiative but also offering insight into possible consequences of new
PM2 5 standards. The report also examined some financial and market dilemmas
associated with scrubber investments. This paper offers selected findings from this
report.
Competition between options in Phase 1. The period leading up to Phase 1
was one of intense and unprecedented competition and innovation, playing coal
against coal, coal and technology against one another, and technology against
technology. An EPRI study published in January, 1996 described the array and
interplay of compliance options during this period.6/7 Electric utilities were the
beneficiaries of this competition triggered by the Clean Air Act Amendments of 1990
(CAAA). This theme of intense competition is an important one to bring forward
while assessing Phase 2 developments - when tighter standards will inevitably
eliminate some of the compliance options that worked in Phase 1 and thereby
reduce the overall scope of competition among options. Beyond the effects of purely
technical considerations (such as mining productivity trends or advances in FGD
cost and performance), the prospect of a diminished range of compliance options
suggests the cost outlook may be more sharply influenced by difficult-to-predict
shifts in market power among fuel suppliers, carriers and utilities.
No crystal ball. A third EPRI report has too raised the question of shifting market
power, and while it may not be time for utilities to push the panic button, the
authors foresee greater volatility in coal and natural gas prices which translates into
volatility in power prices (and vice versa).8 If we throw into the mix the further
effects of business-driven utilities seeking to drive down costs in every part of their
business in a rapidly expanding power market, it becomes clear that the tasks of
forecasting regional and delivered fuel prices and estimating future compliance
costs and strategies are impossible to do with any real confidence - and we haven't
yet considered many far less subtle wildcards that influence electric generation
trends, fuel use and costs.* *
* EIA and MIT have issued recent, complementary reports analyzing the first year or so of Phase 1
compliance 4, 5
These include greenhouse gas measures, questions about nuclear plant license renewals, load growth
trends and load shaping under deregulation, different plans for and urgency of deregulation (including
stranded cost recovery), etc., all influencing power prices and fossil plant generation in ways ranging
from dramatic to obscure. Discussion in 9,10.
-------
Analysis is not pointless. This does not mean we should avoid analyzing SO2
compliance and its main drivers. Rather, we need to avoid single point forecasts and
instead develop a richer view of business planning risks and opportunities. This
comes from explicitly examining a variety of factors, considering how they interact,
and making judgments on how they may continue to change. "What if" analyses
help build perspective by revealing the consequences of specific developments that
are meaningful and plausible. Examples are calculating the effects of lowering
transportation costs of Powder River Basin coal by 25%, or invoking a rule that no
one will invest more that $150 ($/ton SC>2) to create allowances solely to sell them in
Phase 2, or tightening the payback period on FGD - all the kinds of "what if"
questions that one might like to understand better. It is this motivation to inform
business planning that has guided EPRI's analysis of industrywide environmental
issues from the start.
Organization. This paper discusses three main points in succession: the contrast
between Phase 1 and Phase 2; possible effects of much tighter restrictions (such as
could be associated with PMi.s standards); and the error of comparing early industry
projections with today's allowance prices. The final section provides a reference list.
How Phase 2 Might Be Very Different from Phase 1
Two of the principal factors affecting the outlook for compliance in Phase 2 are (a)
the use of allowances banked during Phase 1 and (b) the level of SOi reductions that
must be achieved, which is set by other things and partially offset by the use of
banked allowances during the early years of Phase 2. There is a great deal of
uncertainty about these and many other relevant factors, presenting a range of
scenarios in which allowance prices (really, marginal costs) initially rise not at all or
decline and nearly no further scrubber retrofits are needed, to cases in which prices
reach almost $300/ton by 2004-2005 and escalate to $450/ton or higher by about 2010
coupled with over 10 GW of FGD retrofits beyond those initially installed for Phase
1. These scenarios imply different swings in regional coal supply patterns. Faced
with such contrasting possibilities, it is important to understand what drives them
and which appear most plausible.
Size and use of "the bank". Estimates of the amount of allowances banked by
the end of Phase 1 range between about 10 and 15 million tons, and how long these
will last depends on how companies use their banked allowances to defer moderate
cost reductions vs. deferring only more expensive situations. High bank sizes seem
to depend on companies making new investments costing $150-200/ton to reduce
SO2 emissions during 1998 and 1999, the plausibility of which is questionable unless
substantially higher allowance prices are expected shortly thereafter. A complicating
consideration is how much companies will trade from these supplies, perhaps
selling/swapping them in early years and buying/swapping them back later. We
-------
have found that with modest expectations for electricity load growth, the time when
the bank is drawn down to a steady level such that it no longer effectively depresses
allowance prices is about 2005. At this point, Phase 2 begins to really look like Phase
2.
Required SO2 reductions. The primary driver of required SC>2 reductions is of
course the legislated cap. (A further change in the cap is considered in the next
section, not here.) Other than this, the level of required SC>2 reductions is set by a
combination of economic growth and electrification trends, changes in the
generation mix (i.e., relative shares of coal, natural gas, nuclear), pressures to
increase the capacity utilization of existing coal plants, and changes in the
underlying quality of coal used by the industry. In our analysis, the starting point
was "1995 status quo" with its compliance measures and coal flows in place (with
the added assumption that all recent and announced scrubber retrofits were
operating at least at design removals). From this starting point, we investigated two
cases of load growth/coal utilization - a "low" case comparable to a number of
recent government forecasts and a "high" case that incorporates about 2% per year
increase in generation from coal plants. The "high" case is actually somewhat
modest - it defined our base case.
The manner in which assumptions about coal utilization and banking interact to set
required SC>2 reductions is summarized in Table 1.
Table 1
Additional SOa Reductions Required in Phase 2
For Different Levels of Coal Generation and Bank Consumption
Coal Utilization Early Phase 2
Scenario (bank available)
Low
High
Emissions*(uncontrolled):
Effective Cap:** 9.4 + ~l.i
Req'd Reduction:
Emissions*(uncontrolled):
Effective Cap: 9.4 + ~l.i
Req'd Reduction:
11.9 MT
8** = 11.2 MT
0.7 MT
13.6 MT
8** = 11.2 MT
2.4 MT
Late Phase 2
("no" bank)
>11.9 MT
9.4 to 8.9 MT
>2.5 MT
>13.6 MT
9.4 to 8.9 MT
>4.2 MT
* These are hypothetical future emission levels under "1995 status quo" conditions. Load
growth would cause greater uncontrolled emissions in late Ph.2, indicated by ">" sign.
"Effective cap combines the legislated cap plus, during early Ph.2, a modest consumption
rate of banked allowances (-1.8 MT/yr).
Coal switching costs. Coal switching costs are also quite uncertain. Premiums
for coal sulfur quality will track changes in regional mine prices and rail rates. These
will reflect changing competitive circumstances, productivity trends in mining and
-------
transportation, and market power (the latter responding to changes in industry
structure, such as further mine and railroad consolidation"). In a chicken and egg
manner, coal sulfur premiums will also track allowance prices - they have become
quite closely aligned over the past year. We have looked at several coal price
scenarios, including a soft-price world with no upward movement in coal prices
from 1995 and an even softer case with low western transportation prices; upward
price movements are no less plausible. Additional considerations are capital and
operating costs associated with switching to non-design coals (e.g., changes in coal
inventory and handling systems and boiler modifications) and possible derates,
particularly during the 2005-2010 time period. In all cases, coal switching is the
primary route to compliance in Phase 2.
Scrubbing costs. Scrubbing costs are also expected to change. Without expecting
radical changes in technology, further declines in capital and operating costs are
expected by many, and removal efficiencies will likely remain high. We have
considered a case involving modest advances beyond those already achieved by the
most cost-effective Phase 1 scrubbers. In addition, existing scrubbers will play an
expanded role due to the likelihood that they will achieve significantly greater
removals by late Phase 2. The scrubbing option may not get meaningfully cheaper in
a financial sense, however, since a quicker return on invested capital may be
required than in the past. To capture this hurdle, we have assumed that scrubbers
must be amortized over 10 years rather than the usual fifteen, which tends to offset
the advances expected in technology costs and performance.
Allowance prices to date. A number of unique circumstances surround Phase
1, including the novel two-Phase nationwide cap on SC>2 emissions that is moderate
in Phase 1 and in many cases voluntary; the certainty of more stringent conditions
to come, encouraging banking of emission allowances for future use; awards of 1.3
million tons of allowances per year in 1995 and 1996 to encourage scrubbing (and
about 0.3 million tons per year the next three years); flexibility and intense
competition among compliance options, including the coincident (but not entirely
coincidental) surge in use of Powder River Basin subbituminous coals. These
circumstances led to substantial overcompliance and banking of allowances,
resulting in a glut of presently surplus allowances that has driven prices for the
most part below the cost of creating themes, just how different Phase 2 is likely to
be from Phase 1 requires an appreciation of the circumstances that have made Phase
1 unique, not just analysis of what can happen in Phase 2. Serving as a barometer of
all these developments is the record of allowance prices.
Several transactions in the $250-300 range took place in 199210. The first EPA
Auction, designed to kick-start allowance trading, took place in March 1993. The
* The extent of recent structural changes in coal production and rail transportation has prompted
several research studies and is a subject of ongoing interest to fuel buyers H/12.
-------
volume of transactions grew progressively ever since* , while allowance prices for
the most part declined, reaching their nadir early last year (1996). The next twelve
months saw a firming of prices that turned around once again several month ago.
Figure 1 portrays Emissions Exchange Corporation's index of allowance prices.
Allowances for short term use are valued slightly more than those that cannot be
used until 2000.
Phaseh Allowances - Current Years
Phase 2 Allowances - 2000
50
JMMJSNJMM
Figure 1
The Allowance Market To Date
Note: Emissions Exchange Corp. Exchange Values ($/ton), quoted last day of the month,
per "Exchange Value'' newsletter, various issues. Used by permission.
Results
Is there a future for FGD under Title IV of the CAAA? Or will FGD follow the fate of
the nation's nuclear plants? No new nuclear plants were ordered after 1977 (and
none ordered that also went to completion since 1974)."* Might the fleet of
scrubbers built for Phase 1 be the last? Such an outcome looks very unlikely, but a
-10 year hiatus can be expected. Results are summarized in Figure 2.
" It appears that 10-20% of allowance transactions recorded by EPA are utility-utility trades.
Anecdotally, there is evidence of a variety of forward transactions, swaps and leases. A high volume of
transactions is not felt to be required for emissions trading to accomplish its purpose, namely, to bring
flexibility, heighten competition and lower costs of compliance, discussed in MIT report and elsewhere.
"' J. Taylor, EPRI, personal communication. July 20,1997.
-------
Add'l SO2 Reductions
MTPY
Low Coal
Utiliz'n
$284
1.2GW
$/ton marginal cost
High Coal
Utiliz'n
$449
11.2 GW
Add'l FGD
X
$234
1.2GW
$306
3.6 GW
$86
1.2GW
$173
1.2GW
Add'l FGD
Incr. Removal At
Existing FGD
Switch, W. Coal
Switch, E. Coal
Med None High Med Low None
(1.76 MTPY) (2.42) (1.76) (1.05)
Bank Consumption Level
Figure 2
Effects of Coal Generation and Bank Consumption on
Additional SC>2 Reductions and Compliance Measures Needed in Phase 2
Source: EPRI TR-107897
This figure describes changes expected in marginal costs of compliance (the
economist's proxy for allowance prices) and the level of additional FGD retrofits (in
GW) that would be required under different combinations of load growth/coal
utilization and bank use. SC>2 reductions are characterized by method - switching to
eastern vs. western low-sulfur coals and additional scrubbing (in newly added vs.
previously built units).
The figure does not present a precise time-line. Rather, it characterizes the
consequences of different condition that could occur in the middle of the next
decade, possibly in sequence. A bank consumption level of "none" anticipates
conditions most likely to be approached (if not fully achieved) later in the decade.
With high coal generation, the time when bank consumption could greatly
diminish is not long after 2005. The industry is then looking at a "second wave" of
FGD retrofits about the same size as the first (at least initially). The fact that
conditions in Phase 2 appear to be so different with and without the bank presents
particular challenges to planning and investing in scrubbers. The Phase 1 experience
supports a "go slow" approach, and until allowance prices increase convincingly, the
scrubber option will look very risky. This caution may actually improve the
economics of those retrofits that do go ahead, however, since there is then less likely
to be a rush toward scrubbing with its price-depressing effects.
-------
Deregulation favors higher coal utilization. The only thing that might stall
this "second wave" this would be fundamental changes causing the industry to
follow the low coal utilization path, such as very costly environmental legislation
that greatly penalizes coal generation. This cannot be ruled out. Alternatively, the
industry might simply develops an aversion to scrubber investments and fail to add
any new retrofits. Such an aversion would be costly and was calculated, actually
raising the reference case allowance price by over $150/ton (vs. $449/ton shown in
Figure 1). However, the forces of deregulation suggest the high utilization case may
actually understate the extent of increased coal generation.
Electricity demand should increase somewhat in response to falling prices (whether
mandated or a natural outgrowth of deregulation), with baseload (including coal)
generation potentially benefiting from time of day pricing and load shifting. Even
more so, coal generation should benefit from industry's concerted campaign to
increase the performance and capacity utilization of power plants. A stated aim in
some instances is to lower the reserve margin from about 18% to 12% or less. Some
of this coal growth would occur at the expense of less efficient coal plants, but the
net effect favors coal. Moreover, developments and investments in transmission
along with expanded power trading will likely help the industry maintain reliable
operation at lower reserve margins.
The Startling Stringency of PM2.5
Revisions to the National Ambient Air Quality Standards are not a separate issue
from SC>2 compliance - SC>2 emissions would be affected by proposed standards on
fine particulates (PJMk.s). Just how close is this connection will likely take much
further scientific study and analysis, not to mention regulatory debate. Moreover, it
is not clear whether concern over improving ambient air quality will permit the
kind of nationwide flexibility to achieving SC>2 reductions that characterizes the SC>2
emission trading program. (This would be ironic, since market mechanisms are
touted for lowering future costs of compliance.) Recognizing that fundamental,
programmatic questions have not been resolved, we can still obtain clues into the
implications of PM2.5 standards from documents issued last year in support of EPA's
Clean Air Power Initiative. 13
In these and subsequent EPA analyses, the current legislated cap on SC>2 emissions
by 2010 was lowered by 50%, and in one analytical scenario by 60%. The presumption
is that fine sulfate particles could be controlled in a manner similar to Title IV
provisions by simply halving the SC>2 cap. In its study of compliance and allowance
markets, EPRI conducted a preliminary analysis of what such a reduction could
mean.i The results are summarized in Table 2.
-------
Table 2
Implications of Halving the Phase 2 SOi Cap By 2010
SC>2 reductions beyond 1995 measures: -11 MTPY
FGD retrofits beyond Phase 1 response: 124 G W
(yields 8.7 MTPY SO2 reductions)
Marginal SC>2 reduction cost: $1,470 per ton ($96)
Compliance costs beyond Phase 1
• Cost of full compliance with Title IV -$1.0 billion/yr
• Added cost of meeting 1/2 cap $4.0 billion/yr
These findings are not a prediction that allowance prices will be $1,470 per ton, or that
any of the other extreme impacts would occur. Instead, they provide an indication of
the stringency of meeting a cap that is half as large as that in late Phase 2.
Instead of investing in costly compliance measures on a vast population of existing
coal plants, it is likely that many other options would also be pursued, including
retirements and replacements of coal capacity. The costs of these other options
would probably be about the same or even more costly than SC>2 compliance, yet
they mean the cost of allowances and the numbers of retrofits would run up less. To
the extent natural gas generation receives a boost, the costs of that fuel would be
expected to run up higher.
The $4.0 billion/yr number does not correspond to either the "high" or the "low"
coal generation/utilization case, but a middle case in which the industry continues
with high utilization through 2004 and shifts to the low case from 2005-2010. This is
an effort to bring as much realism to the assessment as possible, but we do not
purport to have conducted a comprehensive analysis. Complex market adjustments
triggered by such stringent emission restrictions would need to be evaluated in
depth for very unfamiliar conditions. Careful analysis would also be needed of
impacts on regional coal supplies and demands, possible escalation of technology
costs under market pressures, and impacts on regional electricity prices and
demands.
It is interesting to stand back and place the potential SC>2 impacts in perspective. A
recent analysis by Reason Foundation/Decision Focus Inc. has attempted to compile
all relevant, available information of PM2.5 costs and benefits and bring
independent judgments on these questions.14 According to this study, EPA has
offered an estimate for "partial attainment" of PM2.5 provisions of $6.3 billion/year.
The authors estimate full attainment to cost $70-150 billion per year, after full
compliance with provisions of the Clean Air Act Amendments (with ozone
-------
provisions costing $20-60 billion per year).* Drawing on a model that calculates
impacts in different sectors, the study estimated that electric utilities were the
hardest hit sector, accounting for $20 billion per year out of a total cost $90 billion per
year (a low case for combined ozone and fine particulate compliance). If $20 billion is
approximately correct, the additional $4+ billion per year estimated in EPRI's SC>2
study represents a large fraction of the projected electric utility impacts.
A final point is that the efficacy of trading diminishes greatly when emission
limitations become very strict. Trading is most effective in moderating compliance
costs if the variation of circumstances among companies is such that some would
face high costs without trading while others would face low or no costs. However,
when the requirements are so stringent that virtually everyone faces high costs,
there isn't much diversity left to exploit via trading. Put another way, a "cap and
trade" system with a very stringent cap becomes increasingly indistinguishable from
a "command and control" system.
Exactly How Not to Interpret the Phase 1 Experience
Misinterpretation and rhetorical comparisons of the history of Phase 1 SO2
compliance have become commonplace in the media (e.g., USA Today, June 24,
1997). Accusing the industry of "crying wolf" the argument goes like this:
"Industry dramatically overestimated compliance costs in the past. Why should we
believe them now as they respond with a hue and cry to proposed tighter ambient
air standards?"
The March 10th remarks of US EPA Administrator Carol Browner capture this
sentiment succinctly:l5
Not only does law forbid us from considering costs in setting these standards, but
history and experience tell us we'd be foolish to try. Almost every time we have begun
the process to set or revise air standards, the costs of doing so have been grossly
overstated by both the industry and EPA. During the 1990 debate on the acid rain
program, industry initially projected the cost of an emission allowance to be $1500
per ton of sulfur dioxide; EPA projected it to be as much as $6000. Today, those
allowances are selling for less than $100. ... The predictions of economic chaos have
never come to pass. Why? Because industry always rises to the challenge, finding
cheaper, more innovative ways of meeting standards and lowering their pollution.
What's wrong with this perspective? The whole purpose of this paper is to explain
why Phase 2 compliance costs are likely to be very different and more costly than
Phase 1, and quite possibly extremely so - as well as to explain how Phase 2 breaks
into two parts (with and without availability of banked allowances) and what
extraordinary event has to happen to keep allowances relatively low (namely, very
* The main reason for the great discrepancy with EPA's estimate is the distinction between partial and
full compliance, although a careful bottom-up compilation of cost factors might reveal disagreements
over many other cost considerations as well.
-------
low growth in coal generation). To all these points are the added problems of
comparing apples and oranges.
<-/ O '
comparing apples and oranges
Being wrong doesn't mean you never learn from your mistakes. One choice is of
course to abandon any further analysis and rely on faith that everything will
work out, a consequence of C. Browner's argument. The other choice is to
sharpen one's pencils. EPRI's progression of analyses is an example of the latter
approach.
Industry estimates were wrong, but not as wrong as this suggests. $l,500/ton was
an estimate written into the CAAA through the political process. It was
Congress' number for a backstop price, in event utilities managed to hoard
allowances and EPA was required to redistribute a fraction of allowances
collected for this purpose. Industry estimates in my own experience ranged
between about $400-800/ton; the MIT study recalls a range of $500-700/ton. These
are consistent with public and private information on projections and surveys by
ICF, RDI, AER*X (now a unit of Enron), and others. The range of estimates
reflects, not just differences in opinion over single point estimates, but the great
range in individual company's compliance cost circumstances (such as retrofit
difficulties, sizes of affected units, delivered coal costs, and quality of coal
receipts).
The time frame is misplaced. For consistency, comparison should be between
either former estimates of Phase 2 compliance and current estimates of Phase 2
compliance, or former estimates of Phase 1 compliance and actual costs. The
industry cost estimates just cited typically assumed full compliance with Phase 2
provisions (for most, this was felt to be close to 2000, missing the effects banking
on the trajectory of prices during Phase 2).
- This attention on Phase 2 and assessing the full effects of the CAAA was
partly an outgrowth of habit, as the industry had been conducting policy
analyses of various acid rain proposals throughout the 1980s. It was also an
element of sound business planning to take into consideration the longer
term planning horizon. It was only as the full effects of cost competition
became apparent several years after 1990 that many companies were able to
redirect their compliance strategies toward optimization in the short run.
- The error of looking farther ahead, if it can be called that, was not the
exclusive domain of the power industry. Mining companies in Central
Appalachia, for example, positioned themselves early to meet an expected
surge in demand for very low sulfur coal.
A comparison against actual "costs" is problematic, since today's allowance prices
do not reflect industry's actual costs of compliance during Phase 1. MIT's
assessments of Phase 1 costs (the basis for EIA's report) found Phase 1 costs to
average ~ $200 per ton, although there is quite a spread between the averages of
-------
switching and scrubbing costs (e.g., ~$150 and $265 per ton). Comparable Phase 1
compliance costs were in fact foreseen in earlier studies, but greater weight was
given to developments after 2000. This does not mean the industry really was
right all along - there were real and unanticipated cost declines in fuel and
technology over the 1990-95 period - but much less wrong than suggested.
• The collapse of Phase 1 allowance prices reflects many factors:
- The prospect of much tighter restrictions encouraged overcompliance and by
definition a current allowance glut.
- Higher future prices are being heavily discounted, perhaps in part because of
heightened business uncertainties created by electric utility deregulation.
- Extraordinary productivity and favorable rail rates for Powder River Basin
coal brought serendipitous reductions.
- The many bonus allowances encouraged scrubber retrofits - the chief source
of allowances being added to the bank - which, once built, could generate
further reductions at relatively low variable costs.
- The relatively easy-to-meet Phase 1 limitation (2.5 Ibs SOi/mmBtu) that
broadened the array of eligible competing options.
It goes without saying that not all these factors will continue. One example: since
1990, both coal producers and railroad companies have consolidated to an
unprecedented degree, with as yet uncertain future consequences to competition
and prices.
• A particularly interesting analysis would be to estimate how much tighter the
markets for low-sulfur coal might have been if the top three FGD retrofits had
not been built (7,000 MW) and what the impacts might have been on allowance
prices and the build up of the bank during Phase 1.
The problem with such faulty comparisons is that bad rhetoric can lead to bad policy.
It can also suspend inquiry or spawn bad economic analysis, leading to bad business
decisions. The CAAA was indeed the first nationwide experiment in applying
market mechanisms to environmental policy, and much can be (and has been)
learned from this. Not least is an appreciation for technical innovation and for the
savings that come from flexible policies. Yet the role of emissions trading needs to
be carefully understood in the context of events, lest one lose sight of the all-
important role of the cap. And the role of fundamental cost factors too must be
understood in terms of what is feasible or might become feasible, and not merely
extrapolations of faith. This is a painstaking and careful process of building
incremental additions to understanding as these cost factors change in ways that are
often unpredictable. To expect surprises, and be aware of one's limitations in
predicting the future, is also part of the legacy of the Phase 1 experience. In the case
-------
of SO2, we need persistent but cautious analysis even when it appears that
environmental issues have moved on to other concerns.
Acknowledgments
The author appreciates the close reading and editorial suggestions offered by Keith
White; and the manuscripts of nearly completed and just completed studies offered
by Denny Ellerman of MIT and Anne Smith of Decision Focus Inc.
References
The EPRI publications in this list are reports prepared for EPRI's Fuel Supply Cost
Management Target, one of the areas of research in EPRI's Generation Group. The
scope of research in this target addresses changes occurring in fuel and power
markets and approaches to managing fuel supply and transportation costs and risks.
The work has a particularly broad focus when it comes to assessing impacts
associated with industry restructuring and clean air compliance, which are complex
integrated issues that combine technology, fuel and other assessments.
1. Keith D. White, T.A. Hewson, Jr., A.J. Van Horn and C.W. Miller, SC>2
Compliance and Allowance Trading: Developments and Outlook. Palo Alto,
Calif.: Electric Power Research Institute, April 1997.TR-107897.
2. K.D. White, T.A. Hewson, Jr., and A.J. Van Horn, The Emission Allowance
Market and Electric Utility SO2 Compliance in a Competitive and Uncertain
Future. Palo Alto, Calif.: Electric Power Research Institute, September 1995. TR-
105490.
3. A.J. Van Horn, K.D. White and T.A. Hewson, Jr., Integrated Analysis of Fuel,
Technology and Emission Allowance Markets. Palo Alto, Calif.: Electric Power
Research Institute, August 1993.TR-102510.
4. US Department of Energy. Energy Information Administration. "The Effects of
the Clean Air Act Amendments of 1990 on Electric Utilities: An Update," March
1997. Executive Summary on US DOE EIA website. Draws on material prepared
for MIT study by Ellerman and others (1997).
5. A.D. Ellerman, R. Schmalensee, P.L. Josko, J.P. Montero, and E.M. Bailey,
"Emissions Trading in the US Acid Rain Program," forthcoming, 1997. Center for
Energy and Environmental Policy Research, Massachusetts Institute of
Technology.
6. J.N. Heller and S. Kaplan, Coal Supply and Transportation Markets During Phase
One: Change, Risk and Opportunity. Palo Alto, Calif.: Electric Power Research
Institute, January 1996.TR-105916.
7. J.B. Platt, "Widening Uncertainties in the Utility Fuel Outlook," Preprint SME
97-21, presented at SME annual meeting, Denver, CO (February 1997). Society for
Mining, Metallurgy, and Exploration, Inc.: Littleton, Colorado.
-------
8. A.M. Stewart and M.J. Greenberg, Electricity Price Formation in the Western
United States, EPRI Report Series on Fuel and Power Market Integration. Palo
Alto, Calif.: Electric Power Research Institute, August 1997.TR-108475.
9. F.C. Graves, S.L. Thumb, L.S. Borucki, and A.M. Schall, Regional Impacts of
Electric Utility Restructuring on Fuel Markets, Palo Alto, Calif.: Electric Power
Research Institute and Gas Research Institute, April 1997.TR-107900-V1. Also
available from Gas Research Institute, Chicago, 111.
10. S.L. Thumb, W.R. Hughes and W. Glover, Framing Scenarios of Electricity
Generation and Gas Use, EPRI Report Series on Gas Demands for Power
Generation. Palo Alto, Calif.: Electric Power Research Institute, July 1996.TR-
102946.
11. S. Schwartz and W. Glover, Structural Change in the Coal Industry: Coal
Industry Concentration Trends, 1970-1994, Palo Alto, Calif.: Electric Power
Research Institute, May 1995. TR-105026.
12. S. Kaplan, J Heller, J. Price, S. Norwood, and D. Weishaar, Railroad
Consolidation and Market Power - Challenges to a Deregulating Electric Utility
Industry, Palo Alto, Calif.: Electric Power Research Institute, December 1996. TR-
107301."
13. US Environmental Protection Agency. Office of Air and Radiation. "Analysis of
Options for Air Emissions Control Under the Clean Air Power Initiative", April
1996. Washington, DC.
14. A.E. Smith, D.W. North, J.P. Bekemeier, N.Y. Chan, R.Glasgal, J.L. Welsh; Proj.
Dir. K.Green, Costs, Economic Impacts, and Benefits of EPA's Ozone and
Particulate Standards. Reason Foundation and Decision Focus Inc. Policy Study
226, June 1997.
15. C. Browner, "New Initiatives in Environmental Protection", in The
Commonwealth (newsletter), March 31, 1997. Commonwealth Club of
California.
-------
LS-2, TWO YEARS OF OPERATING EXPERIENCE
Jonas S. Klingspor
ABB Environmental Systems
1400 Centerpoint Boulevard
Knoxville TN 37932, USA
and
Charlotte Brogren
ABB Corporate Reserch
Department R
S-721 78 Vasteras, Sweden
ABSTRACT
In late 1995, ABB introduced the LS-2 technology, a limestone based wet FGD system which is
capable of producing high purity gypsum from low grade limestones. Drawing from 30,000 MWe
of worldwide wet FGD experience, ABB has incorporated several innovations in the new system,
designed to reduce the overall cost of SC>2 compliance. Collectively, these improvements are
referred to as LS-2. The improvements include a compact high velocity absorber, a simple dry
grinding system, a closed coupled flue gas reheat system, and a tightly integrated dewatering
system. The compact absorber includes features such as a high velocity spray zone, significantly
improved gas-liquid contact system, compact reaction tank, and a high velocity mist eliminator.
The LS-2 features can be used collectively or individually.
The LS-2 system is being demonstrated at Ohio Edison's Niles Plant at the 130 MWe level, and
this turnkey installation was designed and erected in a 20 month period. At Niles, all of the
gypsum is sold to a local wallboard manufacturer. The results and experience gained over the two
year operating period is presented below. Also described in detail are the mass transfer
enhancements utilized by the LS-2 system.
Many of the features included in the LS-2 design and demonstrated at Niles can be used to
improve the efficiency and operation of existing systems including open spray towers and tray
towers. The SC>2 removal efficiency can be significantly improved by installing the high efficiency
LS-2 style spray header design and the unique wall rings. The absorber bypass can be eliminated
or reduced by including the LS-2 style high velocity mist eliminator. Also, the LS-2 style spray
header design combined with wall rings allow for an increase in absorber gas velocity at a
maintained or improved performance without the need for costly upgrades of the absorber recycle
pumps.
The first upgrade using LS-2 technology is being done at CPA's Coal Creek Station (2x545
MWe). The absorbers are being retrofitted with new LS-2 spray headers and wall rings to
increase the performance of the system.
Page 1
-------
INTRODUCTION
The reduction in capital cost of the LS-2 FGD system has been achieved mainly by a significant
reduction in the size of the absorber and integration of critical subsystems such as limestone
grinding and gypsum dewatering. The absorber size has been reduced by operating at high gas
velocities, by reducing the reaction tank size, and by incorporating a high velocity horizontal flow
mist eliminator. Cost reductions in the additive preparation area have been achieved through the
use of a dry grinding system based on ABB roller mill technology. Also the primary dewatering
step has been optimized and from a process standpoint, integrated with the absorber system.
Simplification of the design has led to substantial cycle time reductions in engineering and
construction. The Niles demonstration plant was engineered and erected on a 20 month schedule,
far shorter than the current industry standard.
The capital cost savings have been achieved while reducing the operating cost. A reduction in
operating cost has been achieved through a reduction in power consumption.
LS-2 SYSTEM FEATURES
The wet FGD installation at the Niles Plant includes a number of innovative process
improvements; collectively these are referred to as LS-2. The spray tower has the ability to run at
velocities as high as 18 ft/sec (5.5 m/s), features a compact spray zone with ABB's patented co-
current and countercurrent staggered nozzle arrangement, and a compact reaction tank. The
reagent system is based on an ABB Raymond roller mill which is based on a completely dry
grinding circuit. This limestone grinding system is less costly, both to construct and operate, yet
produces a significantly finer grind. The primary dewatering system features fully integrated high
efficiency hydrocyclones followed by centrifuges for secondary and final dewatering. An
overview of the FGD system installed at the Niles Plant is shown in Figure 1.
The LS-2 system is designed to use fine grind limestone which allows the use of a significantly
smaller size reaction tank. By using fine grind limestone, fractionation of limestone and gypsum
in hydrocyclones becomes feasible. Hence, a gypsum purification step is included in the LS-2
system. The LS-2 grinding system inherently produces an inert fraction with a very fine particle
size distribution. Hence, the fine grind system also improves the fractionation of the inert portion
in the limestone and the LS-2 system lends itself to operating on relatively poor purity limestones
while producing high quality, wallboard grade gypsum.
Before exiting the absorber, the flue gas passes through the ABB high velocity mist eliminator.
The mist eliminator system consists of a bulk entrainment separator followed by two stages of a
two pass mist eliminator. The mist eliminator is capable of operating at velocities well above 30
ft/sec. Due to the compact spray tower, the LS-2 absorber makes use of a horizontal shaft rotary
Ljungstrom gas-to-gas reheat system. Produced gypsum is sold to a local wallboard company.
Page 2
-------
Figure 1 LS-2 installation at Niles, OH.
The compact absorber, achieved through the high velocity spray zone, high mass transfer spray
header design, and the compact reaction tank equates to a significant size reduction and reduction
in material requirements and capital cost. A summary of the reduction potential of the LS-2
system is shown in Table 1 below. Cost reductions are also achieved in the construction area,
resulting from the compact absorber and reduction in material requirements.
Table 1 LS-2 Size Reduction.
Parameter Savings
Reduction in Absorber Diameter
Reduction in Overall Height
Reduction in Plate Area
Reduction in Plate Weight
Reduction in Liquid-to-Gas Ratio
Reduction in Power Consumption
15-25%
20-30 %
25-35 %
35-45 %
30-50 %
10-20%
GYPSUM PRODUCTION
Production of gypsum in limestone based wet FGD systems is commonplace and was pioneered
by ABB in the early '70s. Several factors play an important role in the production of wallboard
grade gypsum including limestone preparation, reaction tank design, and gypsum purification and
dewatering. The features included in the LS-2 design which contribute toward optimum gypsum
quality are described below.
Page 3
-------
Additive Preparation. The additive preparation system features an ABB Raymond roller mill.
The millng system accepts a limestone feed stock less than 1.6 in (40 mm.). The limestone grind
will typically be 99.5 percent less than 325 mesh (44 |im). Untreated flue gas is used to dry and
convey the limestone during mill operation. The flue gas leaving the milling system is returned to
the absorber for processing. The limestone preparation and handling system is completely dry and
includes a wetting system prior to injection into the reaction tank.
The particle size of the ground limestone is
controlled by the dynamic classifier included in
the ABB Raymond roller mill. The speed (rpm)
of the classifier is inversely related to particle
size higher speeds produce smaller particles
exiting the classifier and mill. This is
accomplished by recirculating larger particles
back to the mill to be reground to the maximum
cut size. Recirculation of limestone therefore
increases as classifier speed is raised to produce
smaller mean particle size distributions.
Recirculation also increases the pressure drop
across the mill at a constant conveying air rate.
Figure 2 shows the relationship between grind
size and classifier speed. Both mass median
particle size (D50) and the 90% cut size (D90) Figure 2 Raymond roller mill performance.
are plotted.
Reaction Tank. The reaction tank is an integral part of the absorber and provides residence
time to complete a number of critical chemical reactions including:
Classifier Speed, rpm
limestone dissolution:
sulfite oxidation:
S03
'/202 -^ SO42
gypsum precipitation:
Ca2+ + SO42~ +2H20 -» CaS04-2H20
(a)
(b)
(c)
The size of the reaction tank at Niles has been reduced by about 60 percent compared with
conventional sizing criteria. ABB has determined that for solid residence times above 6 hrs, the
limiting reaction step is the limestone dissolution, not sulfite oxidation or gypsum precipitation.
At solid residence times above 6 hrs, the gypsum precipitation reaction is sufficiently fast as to not
affect the gypsum relative saturation. Also, the oxidation reaction is much faster than the
limestone dissolution rate and is not affected by a reduced reaction tank size.
Page 4
-------
With typical limestone grinds, a significant reduction in the reaction tank size will dramatically
lower the liquid phase alkalinity and increase the solid phase alkalinity. In order to reduce the
solid phase alkalinity, the operating pH must be lowered, which would result in a reduction in S02
removal efficiency. In order to maintain the SO2 removal efficiency, i.e., maintain the pH and the
liquid phase alkalinity at reduced residence time, the only option is to increase the limestone
surface area available for dissolution.
The LS-2 system overcomes the traditional limitation in reaction tank size by using an ultra-fine
limestone grind. The required limestone grind is achieved by using ABB Raymond roller mills.
Dewaterinq System. The dewatering system consists of a battery of 6 inch hydrocyclones for
primary dewatering and centrifuges for secondary dewatering. The hydrocyclones are tightly
integrated with the absorber loop to utilize their separation capabilities. Specifically, the
hydrocyclones are designed to optimize limestone-gypsum fractionation rather than dewatering as
shown in Figure 3. This is achieved by careful selection of hydrocyclone size, operating pressure,
and vortex and apex sizes.
Figure 3 Typical LS-2 hydrocyclone system.
The grind size used is 99.5 percent less than
325 mesh (44 u,m), or 90 percent less than 25
|im equal to a mass mean diameter of about
8 |im. This provides a clear advantage in the
primary dewatering step as fractionation can
be readily accomplished due to the size
difference between the gypsum and limestone.
Centrifuges are used to dewater the gypsum
byproduct down to a moisture content of 8
percent or less. The chloride concentration in
the gypsum cake is reduced to less than 50
ppm by means of water washing during the
centrifuge spin cycle..
SYSTEM PERFORMANCE
The LS-2 project entered the startup and tuning phase in late July 1995. A parametric test
program was started in early October and continued through 1996. All of the subsystems are
operational and have met or exceeded their design requirements. As of July 30, 1997, the system
had been on-line for approximately 15,000 hours and produced about 60,000 tons of on-spec
gypsum. The gypsum purity has consistently exceeded requirements in terms of purity, moisture,
and chloride content and is sold to a local wallboard manufacturer. The crystal size has lent itself
to easy dewatering and residual moisture levels down to 6 percent are achievable.
The high velocity spray zone is operating successfully at 18 ft/sec (5.5 m/s) at a SCh removal of
97 to 98 percent with two spray levels in service. The GGH has operated above expected heat
transfer rates, and the pressure drop has been controlled by regular soot blowing since the startup
Page 5
-------
of the system. However, the initial enamel coating on the heat transfer plates failed and had to be
replaced.
The roller mill has been operated successfully over a wide range of limestone grind sizes from 85
percent to above 99 percent less than 325 mesh (44 um). The grind size can be easily adjusted
online by adjusting the speed of the dynamic classifier.
Based on data collected during the test program, the LS-2 project has met or exceeded its design
targets as shown in Table 2:
Table 2 Typical LS-2 Performance Results
Parameter Design Test Results
Gypsum Purity
Gypsum Moisture
Gypsum Chloride
Sulfite Oxidation
Gypsum MMD
SO2 Removal
Limestone Grind
Gas Velocity
Reheat
>95%
<8%
< lOOppm
>99.5%
>30 urn
>90%
99% < 44 urn
>15 ft/sec
>200 °F
97-98%
6 - 8%
30-50 ppm
>99.9%
>50 jam
97%-99.5%
85-99% < 44 urn
18 ft/sec
>210 °F
MASS TRANSFER
Absorption of SOj can be expressed in terms of mass transfer coefficients and a driving force:
(1)
where NSo2 is the absorption rate in (mol/m3,s), a is the mass transfer area (m2 /m3), kg the gas
side mass transfer coefficient, k°i the liquid side mass transfer coefficient, H Henrys constant for
S02 and E the enhancement factor due to chemical reactions. The overall removal efficiency can
be calculated from the integrated value of equation (1):
J
"P
'so,
(2)
Essential for the performance of an open spray tower is the creation of conditions of high mass
transfer within the spray zone. This can be done by either increasing the mass transfer area, a,
and/or increasing the mass transfer coefficients. Increasing the mass transfer area can easily be
Page 6
-------
done by increasing the liquid flow rate and/or increasing the nozzle pressure and thereby
decreasing the droplet size. To increase the mass transfer coefficients, the fluid dynamics of the
scrubber must be altered, which is far more complicated.
To better understand and thereby be able to utilize different operating parameters to create high
mass transfer, ABB has developed both measurement techniques and test rigs as well as computer
models for in depth studies of the mass transfer within a spray scrubber.
Measurement Techniques. The mass transfer rate shows large spatial variation throughout
an open spray tower due to both changes in the partial pressure of SO2, in slurry pH and to
variations in the fluid dynamics. To localize the most active zones it has been necessary to develop
and/or adapt techniques to measure both velocity and concentration of SO2 within the spray
scrubber.
UV Differential Optical Absorption Spectroscopy (DOAS) has been developed to measure SO2
concentrations within a wet scrubber. The technique is based on penetration of UV light from an
emitter through an absorption path of defined length, as shown in Figure 4. The penetrated light is
detected by a receiver and converted by the analyzer. The DOAS technique is insensitive to water
vapor and to light attenuations resulting from dust, droplets or dirty window, as long as the light
level does not decrease below a certain limit. Hence, the DOAS analyzer can be used to
accurately measure SO2 concentration inside the spray zone without interference from the high
density spray of slurry droplets.
Figure 4 DOAS in-situ SCK probe
Figure 5 LDV in-situ velocity probe
The use of laser-based instruments has been applied to measure fluid dynamic properties within an
absorber. Within the WFGD R&D program a fiber-optical, two-dimensional (simultaneous
velocity in two directions) Laser-Doppler-Velocimetry (LDV), Figure 5, and Phase-Doppler-
Page 7
-------
Anemometry (PDA) instrument are used. LDV measures particle velocity, for gas velocity
measurements, the incoming gas flow is seeded by tracer- or seeding particles The velocity is
determined from the frequency of the scattered light, scattered by the particles. PDA is a direct
extension of the LDV and measures velocity in the same way but also measures simultaneously
the particle size. The size information is extracted from the phase difference between scattered
light in two different directions
Test Rigs. Along with the LS-2 demonstration plant at Niles, Ohio, ABB has two tests rigs
available for different types of absorption experiments. One smaller rig, which can be operated
with limestone as reagent. Figure 6, and a second with a cross section of 7x7 ft (2x2 m) where
positioning of nozzles, etc. can be studied in detail. Figure 7 In the latter test rig, a traversible
spray header has been installed in order to be able to study under operation the impact of distance
between inlet and first spray header, spray header spacing, etc. The smaller test rig can be
operated under a wide range of conditions, e g SO; concentration up to 3000 ppm, gas velocities
up to 40 ft/sec (12 m/s).
Figure 6 Chemistry test rie
Figure 7 Fluidynamics test rig.
To examine the effect of various operating conditions on the limestone dissolution rate, test rigs
for both batch- and continuous-operating mode have been developed. These test rigs have
especially been used for investigations of the limestone grind size on dissolution rate, utilization,
etc
Computer Models. A model based on the penetration theory has been developed to calculate
the dynamic absorption rate of sulfur dioxide into a droplet of limestone slurry The model
includes both instantaneous equilibrium reactions and reactions with finite rates, i.e. limestone
Paee 8
-------
dissolution, CC>2 hydrolysis, sulfite oxidation and gypsum crystallization. Due to the significance
of limestone dissolution on WFGD performance, a separate model for the limestone dissolution
rate has been developed. The model takes into account the limestone grind size, reaction tank
residence time as well as the chemical composition of the slurry.
Along with the chemistry modeling, ABB has started to apply computational fluid dynamics
(CFD) to study the processes within the scrubber. The modeling is done with a commercially
available CFD code, FLUENT v 4.4, using the k-e turbulence model. The model includes the
effect of the physical presence of the spray cone and droplets on the gas flow field and the
momentum transfer between the liquid and gas phases. To account for droplet-droplet interactions
within the WFGD open tower model, a subroutine was developed which stochastically predicts
droplet interactions in three dimensional space using Monte Carlo techniques. The probability of
interactions and the outcome of the interaction are determined on the present local droplet
concentrations, the average local droplet diameter, the diameter of the incident droplets, and the
traveling distance of the incident droplets being tracked during the dispersed phase calculation.
Kinetic submodels have been incorporated into the codes particle/droplet tracking algorithm. The
kinetic submodels are based on a modified film theory using equilibrium and diffusion data from
BMREQ / FGDPRISM. Since the SC>2 absorption in no way affects the aerodynamics of the
system, the fluid dynamics and the kinetics are solved separately. After convergence of the
system's fluid dynamics, the calculation of these parameters are turned off and the kinetic model,
which is tied to the droplet interaction model, is executed.
Further, CFD has been used for a full 3D model of the Niles LS-2 scrubber tower. This model is
used to investigate the scrubber from the macro scale.
MODELING AND EXPERIMENTAL RESULTS
Fluid Mechanics Within An Open Spray Tower. The two-phase dispersed flow in an
open spray tower is extremely complex. The gas flow field and the turbulence intensity have been
measured in various positions of the scrubber using an LDV instrument. Figures 8 and 9 show the
mean velocity and the turbulence intensity in the main flow direction 400 mm downstream of the
nozzle elevation. The measurements conclude that four different regions can be distinguished
from a fluid mechanics point of view:
• The region below the spray elevation where droplets are falling with terminal velocity due to
gravity.
• The spray region where liquid and droplets are ejecting the gas due to high momentum
transfer.
• The spray intersection zones where two orthogonal sprays are interacting, giving some degree
of droplet collision. In this region the ejected gas flow following the spray is deflected to the
stream wise direction and flowing upwards to the next spray elevation.
• The intersection zone where four sprays interact (both orthogonal and normal interaction)
resulting in droplet collision and gas flow deflection as above, but from four gas streams
accelerating.
Page 9
-------
Figure 8 Four nozzle configuration
Figure 9 Nine nozzle cofiguration.
In all of these zones the gas turbulence flow field directly affects liquid break-up and properties
related to the mass transfer process at the interface. Consequently, the mass transfer rates of these
four zones are quite different. This has, for example, been utilized with ABB's patented nozzle
configuration, which provides a higher mass transfer rate than a traditional nozzle configuration.
A CFD model is able to calculate the same type of flow profiles around the nozzles as measured
with the LDV-instrument. Figure 10 shows an example of the velocity profiles.
Figure 10 CFD simulation of the four nozzle case.
Mass Transfer Conditions Within An Open Spray Tower. To quantify the extent of
spatial variations of mass transfer within a scrubber, the SO? concentrations in various positions
were measured using the ABB DOAS analyzer To eliminate the gradients due to changes in pH-
value of the slurry, the experiments were made using NaOH as the absorbent. The measurements
clearly show that the absorption is most efficient close to the nozzles where high shear forces
Page 10
-------
3
K
X
S02 in NaOH
v = 15ft/s
between the gas and the liquid exists. For
example, in the zone 2-4 ft away from the
nozzle, the height of transfer unit was
about 3 to 4 times as high as in the zone
close to the nozzles < 2 ft , as shown in
Figure 11.
The droplet mechanics and its mass
transfer characteristics have been found to
be quite complex. Very often droplets are
regarded as stagnant spheres with no or
very little internal motion. There is,
however, considerable evidence that
internal circulation and mass transfer rates
are very large during the periods of drop
formation, release and acceleration, i.e.
close to the nozzles (Sherwood et. al., 1975).
Based on the measured and literature data of physical mass transfer properties, an analysis of gas
film resistance (GFR) in different positions of a limestone spray scrubber has been made The gas
film resistance is a measure of the portion of the total resistance to mass transfer located in the gas
phase.
Figure 11 Reactivity in the vicinity of spray nozzles.
0 =
The calculations show that the gas film resistance has its highest value in the top of the scrubber
where slurry with high alkalinity meets gas with low concentrations of SC^. For a pSC>2 of 200
ppm
Kg: 5x10-5 M: 6x10-3 to 6x10-4
35.8% Absorption
52.4% Absorption
Figure 12 CFD simulation of SOi concentration in the vicinity of spray nozzles.
Page 1!
-------
ppm the GFR is typically in the order of 60-70%. The GFR decreases very rapidly with increasing
distance from the nozzle as pH decreases and pSC>2 increases but also as the physical mass
transfer properties decrease. At the bottom of the absorber the gas film resistance is in the range
of 10-15%.
Figure 12 shows SC>2 profiles around the intersection point of two nozzles. The profiles have been
calculated with the combined CFD / chemistry subroutines. The results clearly show how most of
the absorption occurs in the region from the nozzle to the intersection point.
All test results, experimental as well as models, show that most of the absorption takes place close
to the nozzles. This also explains the lack of effect of tower height experienced in full-scale
installations. Since the LS-2 concept takes full advantage of these findings, it results in a compact
absorber tower with a dramatic reduction in absorber size and power consumption.
Gas Velocity. Extensive testing performed by ABB both in laboratory and full-scale units
shows the benefits of a high gas velocity on the SO2 removal. For example, by increasing the gas
velocity from 10 ft/sec (3 m/s) to 18 ft/sec (5.5 m/s) full-scale data shows that the liquid-to-gas
ratio can be reduced by more than 50 percent while maintaining a constant SO2 removal. This is
due to the fact that the increase in mass transfer almost compensates for the reduction in contact
time between the gas and the slurry.
It is well known that the gas side mass transfer coefficient is a function of the relative velocity
between the gas and the liquid. Further, the gas velocity will also increase the exposed mass
transfer area per volume scrubber of a countercurrent spray scrubber by decreasing the apparent
downward velocity of the droplets. To quantify the various effects, absorption experiments have
been performed under different conditions. From gas-film limited experiments (SO2 in NaOH),
the overall kga value was determined and from the liquid-film limited experiments (CO2 in NaOH)
values of mass transfer area a was determined. The correlation between kg and velocity was
thereafter calculated as the ratio between the two tests as shown in Figure 13. As expected, both
the mass transfer area and the mass transfer coefficient increases with increasing gas velocity.
the results indicate that kg is
_^- i ^ 0 6 - 065
proportional to v
For example,
which relates very well
to the general correlation between velocity and
kg found in the literature where the impact of
velocity on kg range from v°'5 (droplets) to v°'7
(packed columns). The results also relate very
well to results from EPRI testing at ECTC
where it was shown that during gas film limited
conditions, the decrease in contact was fully
compensated by the increase in kga.
The experimental results do not indicate any
significant impact of gas velocity on the liquid
side mass transfer coefficient, k°i, which is in
agreement with other studies, e.g. for packed
gas velocity (ft/s)
Figure 13 Mass transfer coefficents as a
function of gas velocity.
Page 12
-------
towers.
lowers. The LS-2 scrubber is designed to operate at a gas velocity of about 18 ft/sec (5.5 m/s).
Compared to a traditional scrubber operating at a gas velocity of 10 ft/sec (3 m/s), the mass
transfer properties of an LS-2 scrubber are about 80% greater.
Limestone Grind Size. Another operating
parameter which can increase the mass transfer
rate is the limestone grind size. Traditionally, it
has been believed that most of the limestone
dissolution takes place within the reaction tank
and that the contribution of solid phase alkalinity
to overall SC>2 removal in the spray zone is
limited. However, by using ultra fine grind
limestone the dissolution within the spray zone
becomes significant which improves the SC>2
removal rate. Figure 14 shows the chemical
enhancement factor of the slurry versus the
sulfite concentration for different limestone grind
sizes.
Figure 14 Effect of limestone grind on
limestone spray zone dissolution.
Initially, the chemical enhancement factor of the droplet is almost independent of the reactivity of
limestone. But as SO2 is absorbed and the pH-value drops the enhancement to mass transfer of
the chemical reactions decreases. For a limestone grind size of 90% < 325 mesh (44 urn), the
dissolution rate is so slow that the contribution to the mass transfer rate is insignificant. However,
with the ultra fine grind limestone used in the LS-2 process, 99.5% < 325 mesh (44 (im), the
limestone dissolution increases the mass transfer rate substantially
Nozzle Position. Besides producing contact between the gas and the liquid, the nozzle
configuration and the sprays have a large impact on the gas flow field. The gas flow field
measured using LDV shows very complex flow
profiles around the nozzles. The position of the
nozzles relative to each other is therefore
critical for the performance of the scrubber.
Both the gas-liquid contact and mass transfer
can be maximized by careful placement of the
nozzles, hence, leading to more compact
absorption towers. Different nozzle
configurations have been tested under various
conditions in the 7x7 ft (2x2 m) absorption test
rig. The testing has led to ABB's patented co-
current/countercurrent staggered nozzle
arrangement (egg-crate), Figure 15. With the
egg-crate configuration, the height of the spray
zone can basically be reduced by a factor of two
since each spray elevation has about twice the
L/G capacity compared to conventional spray
Figure 15 Flue gas flow field around adjacent
nozzles and headers.
Page 13
-------
head designs. But since the egg-crate configuration utilizes zones with high mass transfer by
positioning of the co-current nozzles in the intersection point of the countercurrent nozzles, the
performance of an egg-crate elevation is better than of two conventional elevations.
Wall Rings. Sneakage close to the wall in
open spray towers is a well known problem.
Most scrubbers have the nozzles positioned at
least two feet from the scrubber wall in order
to avoid erosion of the wall lining from the
limestone/gypsum slurry. This often results in a
reduced spray density, a lower pressure drop,
and lower SO2 removal close to the wall. The
less than optimum gas liquid contact close to
the wall has a negative impact on the L/G
required for a specific SO2 removal effeciency.
However, the the wall effects have not limited
the SO2 removal efficiency achievable in open
spray towers.
Conventional
V? '
Figure 16 Flue gas flow field with and without
wall rings.
To increase the efficiency of the outer part of open spray towers, ABB has developed wall rings
along the perimeter of the tower. Wall rings eliminate gas sneakages along the walls due to
blockage, but also redistributes the liquid along the walls back into the spray zone as shown in
Figure 16. In effect, the wall rings acts as if a nozzle was placed at the perimeter of the absorber
wall.
At the LS-2 demo plant, installation of wall
rings had a dramatic effect on the plant
performance. Wall rings were installed at all
spray levels to improve the gas/liquid contact
close to the wall. The nozzles at the LS-2
installation at Niles are located about 2 ft
(0.6 m) from the wall in order to safeguard the
lining. At Niles, 2 ft represent about 30
percent of the total cross sectional area.
Hence, for small size towers as the Niles
installation, wall effects can be noticeble. The
removal effeciency at Niles before and after the
installation of the wall rings is shown in
Figure 17. The removal effeciency increased
dramatically with the installation of the wall
rings, e.g. with two spray levels in service from
about 90 percent to 97.5 percent. With three
levels in service, the measured 862 removal
efficiency was 99.6 percent.
rjg' Penetration
\
\^
\
-*VUJI Rings
-»• ConvenBonal
^
X
Removal Efficiency, percent
0 0
o o» o
o o> « o
3123
Spray Levels in Service
7 Effect on SOi removal by wall
rings.
Page 14
-------
Hvdrocvclone Performance. At Miles,
the slurry from the reaction tank is pumped
to a battery of 6 inch hydrocyclones prior to
final dewatering in the centrifuges. The
purpose of the hydrocyclones is to increase
the solid content of the underflow slurry
stream sent to the secondary dewatering
system and to fractionate the feed solids; i.e.
recycle a higher proportion of limestone
particles and inerts in the overflow back to
the reaction tank compared to the
underflow. An efficient fractionation of
limestone particles from the feed slurry will
increase the limestone utilization and permit
reduced reaction tank volume, since higher
limestone concentrations within the reaction
tank can exist without reducing the gypsum
purity.
0 20 40 60 80 100 120
particle diameter (pm)
Figure 18 Frequency curves of limestone and
gypsum.
The measured and calculated particle size distribution curves of gypsum and limestone are shown
in Figure 18. The fine grind limestone provides a comfortable size difference between limestone
and gypsum in the reaction tank. The fine tail of the limestone size distribution curve can be
assumed to represent the inert fraction of the limestone feed.
Typical hydrocyclone performance data is
shown in Table 4. The limestone content in
the feed is reduced by about 60 percent to
produce a underflow stream with a high
quality gypsum. Also noticeable is the high
recovery of the inert fraction to the
hydrocyclone overflow. This feature allows
lower purity limestones to be used while
producing wallboard grade gypsum.
Table 4 Hydrocyclone Gypsum Purification
Parameter Feed Overflow Underflow
TSS
CaC03
Gypsum
Inert
D50
20.8
5.7
92.8
1.5
45.9
12.7
8.7
89.5
1.8
31.2
58.3
2.4
97.2
0.4
52.6
CONCLUSIONS
The LS-2 project represents an advanced wet FGD system which offers significantly improved
performance at reduced capital and operating costs. Projected cost savings for turnkey systems
are between 5 to 15 percent while cost saving between 15 to 30 percent can be expected for
limited scope projects. The LS-2 system has gone through an extensive test program and the
system comfortably meets all process performance goals. The gypsum quality has consistently
meet required specifications during the two year operation and is being sold for wallboard
manufacturing.
The experience from operating the LS-2 system in combination with extensive laboratory work
have provided unique insight into the fluid dynamics and mass transfer characteristics of open
Page 15
-------
spray towers. The work to date has clearly identified the high mass transfer zones inside open
spray towers. These high mass transfer features are incorporated into ABB's patented spray
header design.. The ongoing R&D work will lead to further improvements of the LS-2 design
The mass transfer characteristics close to the perimiter of open spray towers have been studied in
detail and can be greatly improved by the use of wall rings. Wall rings will significantly reduce the
L/G required for a given SO2 removal efficiency by eliminating the wall sneakage and by
reintroducing slurry lost on the absorber walls back into the spray tower.
Use of fine grind limestone provides an opportunity to efficiently separate gypsum from limestone
and inerts in hydrocyclones. High quality gypsum can be produced from relatively low purity
limestone due to the fractionation capability of hydrocyclones when fine grind limesone is used.
Further, the fractionation of limestone and gypsum elevates the limestone concentration in the
recycle tank which in turn reduces the L/G. Also, fine grind limestone contributes to limestone
dissolution in the spray zone.
The LS-2 system is protected by patents or pending patent applications.
ACKNOWLEDGMENT
The contribution to the paper and to the development of the LS-2 system by the ABB Wet FGD
R&D team including David Collins, Don Borio, Karl Hognefelt, David Anderson, Claes Halldin,
Rikard Hakanson, Ekkehard Schade and Marino Rota is gratefully acknowledged.
The cooperation of Ohio Edison through all stages of the LS-2 project is greatly appreciated.
The LS-2 demonstration project has been cofunded by the Ohio Coal Development Office
(OCDO) and by the Electric Power Research Institute (EPRI). ABB is grateful for OCDO's and
EPRI's participation in the LS-2 project.
REFERENCES
1. J.S. Klingspor, G.E. Bresowar, Advanced, Cost Effiective Wet FGD. Presented at the EPRI
SO2 Conference, Miami, Florida, May 1995.
2. J.S. Klingspor, G.E. Bresowar, Next Generation Low Wet FGD System. Presented at the
PowerGen 95 Conference, Anaheim, California, December 1995.
3. J.S. Klingspor, D.C. Borio, D.J. Collins, D. Gausmann, LS-2, A Performance Update.
Presented at the Powergen 1996 Conference, Orlando, Florida, December 1996.
4. C. Brogren and H.T. Karlsson, A Model for Prediction of Limestone Dissolution in Wet Flue
Gas Desulfurization Applications. Ind. Eng. Chem. Res.. 1997, Accepted for publication.
5. J. Klingspor, C. Brogren, High Purity Gypsum form low Purity Limestones. Presented a the
5th International Conference on FGD and Synthetic Gypsum, torontor, Canada, May 1997
6. C. Brogren, J. Klingspor, Impact of Limestone Grind Size on Wet FGD Performance.
Presented at the EPRI Mega Symposium, Washington, DC, August 1997.
Page 16
-------
High Velocity Mist
Elimination for Wet FGD Application
W. H. Kingston
ABB Environmental Systems
1400 Centerpoint Blvd.
Knoxville, TN 37932
D. K. Anderson
W. P. Bauver II
ABB Power Plant Laboratories
2000 Day Hill Road, P.O. Box 500
Windsor, CT 06095
Abstract
This paper discusses efforts and results to develop a mist elimination system capable of
operation at higher gas velocities than typically associated with FGD open spray towers.
Computational Fluid Dynamics (CFD) along with flow modeling were used to analyze and
focus the development efforts towards a high velocity design without a loss in collection
efficiency (due to re-entrainment) as usually experienced. Particular attention was directed
towards the development of a mist elimination system capable of controlling liquid/solids
carryover on a continuous operating basis, which necessitates effective on-line cleaning of the
mist eliminators.
Introduction
Wet Flue Gas Desulfurization (FGD) was initially introduced in the United States in the mid
1960's. Over the past 30 years the need for mist elimination in wet scrubbing has not changed
significantly. Today mist elimination is still a necessary key component of an SO2 absorber.
During the evolution of the wet scrubbing technology, mist eliminator (ME) pluggage has
been identified as the second most common cause of FGD system outages. However, this has
been improving incrementally, primarily because of two reasons. The first is that the
scrubbing technology has matured and the critical design factors necessary to control the
chemical process are better understood. Additionally, design parameters for mist eliminators
and their washing system have been refined to better meet the needs of the technology. Even
so, it is important to recognize that no ME design can operate satisfactorily in a system which
experiences chemical upsets such as heavy calcium sulfite or sulfate scaling. The second
reason is that the Electric Power Research Institute (EPRI) through various studies and
1
-------
reports on ME's. have provided plant owners with a valuable source of information on ME
performance and operating parameters. Due to limitations on the length of this report, it is
written with the assumption that readers have a working knowledge of wet scrubbing and mist
elimination.
Goals and Objectives
As the scrubbing technology continues to evolve, ABB has worked to reduce the cost of FGD
systems through both process development and component improvements. With the
development of the LS-2 absorber design which operates at velocities above 15 feet/second (
4.6 m/sec), it was felt that the traditional vertical flow ME's would be marginal in collection
efficiency. Therefore ABB selected a horizontal flow ME design which is capable of
operating at higher gas velocities (ref. Fig 1).
Mist Eliminators
BBS
Figure 1
ME Orientation
The same 4 pass ME vane design used in the vertical flow orientation was used in the LS2
absorber. However, the operating gas velocity was increased from 10-12 ft/sec (3-3.6 m/sec)
in the vertical flow section to 20 - 22 ft/sec (6 - 6.7 m/sec) in the horizontal flow section. The
collection efficiency in this orientation was very good , however there was a problem with
cleaning the rear passages of the vanes. This problem is related to the effect gravity has on
the drainage path of the liquid collected. In the LSI arrangement, the drainage path is counter
current to the gas flow as compared to 90 degrees to the gas flow for LS2. This permits the
liquid (with solids) after collection to travel deeper into the back passes of the ME vanes in
the LS2 orientation. The wash spray cleans the first two passes very well, but not the last two
passes, even with higher wash rates and 100 percent overlap. After a few months, as the back
passes of the vanes build up with solids, carryover significantly increases.
-------
Therefore the first objective of this development effort was to identify a vane design for LS2
which can be effectively cleaned while in service, with out the loss of collection efficiency.
Since the majority of droplets from spray nozzles in an absorber are in the 1000 micron
diameter range, capturing these relatively large droplets is not too difficult. The focus of this
development effort was to emphasize the need to control re-entrainment of the droplets after
initial collection. A second objective was to optimize the design for FGD use by maximizing
the operating velocity for an acceptable carryover efficiency. This then would lead to lower
cost ME sections and ductwork.
Development Approach
The following outlines the development approach used in addressing our stated objectives.
• Conduct a literature search of various designs. There was little information on horizontal
flow application in FGD service.
• Apply computational fluid dynamic modeling as a tool for systematic evaluation of
various ME vane designs.
• Test and establish the ABB 4-pass ME vane design as the baseline for minimum collection
performance.
• Concurrent with computational fluid dynamic modeling, physical flow model testing was
performed on promising vanes profiles to verify performance.
Design Development
Computational fluid dynamic or CFD modeling techniques were used to screen and evaluate
existing as well as candidate mist eliminator designs for both primary collection efficiency
and pressure drop. FLUENT, a commercially available CFD code was used for this study.
This code solves the governing equations of fluid flow, combustion, and heat transfer for a
user-specified geometry in three dimensions. The mesh, established in either Cartesian or
body fitted coordinates, consists of blocks or cells arranged to represent the geometry of the
duct or flow passage being modeled. At each cell, the code solves linearized forms of the
governing equations for the transport of mass, momentum, energy, and chemical species.
Each cell is treated as a control volume having uniform properties. Differencing techniques
are then applied over the domain to solve the equations to generate a converged solution. This
is done via iterative techniques until an acceptable tolerance over all the cells is attained.
Because of the nonlinear, coupled form of the equations, direct solution is not possible, hence
inefficient relaxation methods are required to solve these classes of problems. Other aspects of
the CFD solution are equally important. Turbulence was represented using the isotropic K-E
model, which calculates quantities for the generation and dissipation of turbulence kinetic
energy to provide closure for the Reynold's stress terms in the equations.
-------
In addition to solving transport equations for one or more continuous phases, FLUENT allows
the simulation of a dispersed second phase in a Lagrangian frame of reference. This second
phase consists of spherical particles ( droplets, particles or bubbles) evenly dispersed within
the continuous phase (gas or liquid). FLUENT computes the trajectories of these dispersed
phase entities as well as heat and mass transfer to/from them. Momentum coupling between
the phases and its impact on both the dispersed phase trajectories and the continuous phase
flow was included in the model.
Because of the lack of steep local mass flow gradients in the vertical direction of a horizontal
gas flow mist eliminator, two dimensional slice models of the ME blade assembly were
adequate for this study. 2D models had the added benefit of reducing the grid complexity as
well as the time for individual case convergence. A typical 2D grid used during this study is
shown in Figure 2.
Figure 2
ABB 4 Pass Mist Eliminator Computational Grid
Only a single passage was modeled, as the adjacent passages were considered to be identical.
The model starts 2 inches (50.8 mm) before the inlet to the assembly and extends past the
blade passage outlet. Gas is introduced through the inlet with a uniform distribution. For each
ME geometry evaluated, gas phase simulations were converged for velocities of 15, 20, and
25 ft/sec (4.57, 6.10, and 7.62 m/sec). Flow profiles, droplet trajectories and system pressure
drop information were then obtained from these simulations.
An extensive literature survey was performed to determine the state of the art in the
computational modeling of liquid films and droplet re-entrainment processes. It was found
that the process of liquid film formation and motion, as well as droplet formation from the
film are governed by a number of complex physical parameters and processes, ranging from
the local gas velocity to the viscosity of the liquid. Although a number of researchers are
currently working in this area, the direct modeling of droplet re-entrainment from a liquid
film, based on first principles is currently not within the ability of commercially available
CFD codes.
Because of this, mist eliminator CFD modeling was limited to the prediction of the ME's
primary collection efficiency and overall system pressure drop. Primary collection efficiency
is a measure of the ME's ability to collect droplets which enter the assembly entrained within
-------
the gas flow. Primary collection does not account for droplets which have been re-entrained
back into the gas, from liquid which has already been collected. To determine the primary
collection efficiency of an existing or candidate ME design, droplets of sizes ranging between
5 and 60^ were introduced isokineticaly, evenly spaced across the inlet of the model. These
droplets were then individually tracked through the model domain to determine which sizes
were capable of passing through the assembly uncollected. During this process, if a droplet
contacted a blade surface it was considered collected and removed from the simulation.
CFD Results
Mist Eliminators are inertial separation devices. That is, they rely on the fact that the droplets
entrained within the gas flow cannot change direction as rapidly as the gas in which they are
suspended. As the gas enters a typical ME assembly, it is forced to make a number of rapid
turns as it follows the passage defined by the ME blade profile. As a rule of thumb, droplets or
particles of around 10 microns or less are generally assumed to be capable of directly
following the gas flow. However, because of the greater mass and inertia of larger droplets, it
will take a finite amount of time for these droplets to react to the rapidly changing direction of
the gas. Many cannot turn quickly enough to follow the gas, before impacting a blade surface
and being removed from the gas flow. The maximum size which is capable of passing through
a particular ME design or arrangement uncollected is referred to as the top cut size. There are
a number of geometric parameters which control the assembly's top cut size, hence primary
collection efficiency: blade angle and spacing being the primary of these.
As mentioned elsewhere in this paper, the expected mean size of the droplets entering the ME
are expected to be in the order of lOOOu. In the CFD simulations, it was found that the ME
system pressure drop and primary collection efficiency was inversely proportional to the blade
spacing, (i.e. decreasing the spacing increases the AP and collection efficiency). At a pressure
drop of 0.70 "Wg (174.3 Pa) (corresponding to the design spacing of 1.5" (38.1 mm)) the
current ME blades will collect all droplets greater then 25|i. Because this is very close to the
droplets sizes which are generally assumed be capable of directly following the gas flow, little
additional improvement in primary collection efficiency will be achieved by spacing the
blades closer. It is also clear that because the size distribution of the droplets entering the
assembly has a very small fraction of the mass below this size, additional improvements in
the collection of these very small droplets will result in only very minor improvements in the
total system collection efficiency. On a mass basis, the liquid which make it through the
assembly is almost entirely from liquid re-entrainment off of the blades and their supports.
The processes of liquid film motion and droplet reformation require the expenditure of
energy. The source of this energy is the kinetic energy contained within the gas passing
through the assembly. The expenditure of this energy is manifested as additional pressure
losses across the ME assembly. Because the CFD models could not simulate these processes,
the calculated pressure drops were slightly lower than would be expected from the unit in
actual operation. In spite of this, good agreement between the calculated and measured
pressure drops was obtained. A comparison of these values is presented in Table 1.
-------
Table 1
Predicted Pressure Drop To Normalized Measured
Configuration 20 ft/sec (6.1 m/sec) 25 ft/sec (7.62 m/sec)
4 Pass 0.914 0.898
3 Pass 0.852 0.820
Commercial 0.831 no data
4 Pass Design
The majority of the CFD modeling was performed on the standard ABB 4 Pass ME design or
some variation of it. Figure 3 presents the results of the baseline 4 Pass ABB design
simulation at 20 ft/sec (6.1 M/sec). As can be seen in this figure, the gas flow accelerates
over the peaks formed by the intersection of the passes, while recirculation zones are formed
on the downstream side of the peaks. This high velocity region is also where the collected
liquid will be directed by the gas flow, leading to a condition with increased re-entrainment
potential. This is supported by observations in the wet model tests, where droplets are shed
off the peaks of the chevrons and liquid flows down the blades in this gas recirculation zone.
The fact that this primarily occurs at the first peak is good because subsequent collection of
the re-entrained droplets can be achieved in the remaining passes of the ME.
Figure 3
Contours of Velocity Magnitude Within 4 Pass Mist Eliminator
It was very evident from the results of the standard arrangement that majority of the droplets
were collected within the first 2 passes of the ME and that the final 2 passes had little impact
on the primary collection efficiency of the overall system, Figure 4. Various blade angles
were evaluated in an attempt to gain a better understanding of the role of this parameter on
droplet collection. The 45° blade angle with respect to the primary gas flow direction was
found, for example, to provide a high primary collection efficiency. For angles below 45° the
collection efficiency began to get worse while angles above this did not appear to result in any
significant increase, but did result in significant increases in the assembly pressure drop.
-------
Figure 4
Droplet Trajectories Through ABB 4 Pass Mist Eliminator
(25 micron/20 ft/sec (6.1 m/sec))
A number of different blade spacings were also investigated. It was found that for a given
blade geometry, the collection efficiency and pressure drop were inversely proportional to the
blade spacing.
3 Pass Design
Because it was found the majority of the primary collection was occurring in the first 2 passes
of the assembly, an obvious approach to reducing ME AP and manufacturing costs was the
complete removal of the assembly's last pass, Figure 5.
Figure 5
Droplet Trajectories Through ABB 3 Pass Mist Eliminator
(25 micron/20 ft/sec (6.1 m/sec))
As can be seen by comparing the previous two figures, the primary collection efficiency of the
system is not effected by the removal of the last pass of the assembly.
-------
Design Comparison
As part of this study, CFD modeling was performed on a commercially available 2 pass ME
design with and without their collection hooks ( referred to later as curved with hooks). The
results of the simulation with the hook are presented in Figure 6.
Figure 6
Contours of Gas Velocities Through Curved with Hooks Mist Eliminator Design
As can be seen in this figure, the hook has little aerodynamic effect, and in fact could be
considered more of a pocket than a hook. It is apparent that the hook should be effective in
reducing the quantity of liquid which would normally pass over the blade's peak due to
aerodynamic drag forces, although this is currently not capable of being modeled in the CFD
simulations. By keeping the liquid away from this relatively high velocity zone, liquid re-
entrainment should be minimized. This is how it should perform. In actuality, this
arrangement is ineffective in the inertial separation of fine droplets. As shown in Figure 7,
80% of the 25 micron droplets which are introduced into the ME assembly at 20 ft/sec (6.1
m/sec) manage to pass through uncollected.
Figure 7
Droplet Trajectories Through Curved with Hooks Mist Eliminator Design
(25 micron/20 ft/sec (6.1 m/sec))
One could speculate that this design recognizes that the fraction of the total material with a
very small droplet size, which enters the ME, is minor and that they have therefore
concentrated on the collection of the bigger droplets. In general, it appears that the sinusoidal
shape is effective in reducing system pressure drop but does so at the expense of primary
collection efficiency.
-------
Air Water Testing
Mist eliminator performance
evaluation was performed in the Air
Water Test Facility (AWTF) which
was constructed for this project. The
facility is shown schematically in Figure 8.
The facility can accept mist eliminators with
3 2
Collection Points
Figure 8
Air Water Test Facility
cross sectional dimensions of up to 3 ft.(0.91 M) wide by 6 ft. (1.82 M) high. This size was
selected to minimize wall effects and provide sufficient height to permit evaluation of the
liquid film build up along the height of the mist eliminators. An eight foot (2.44 M) section
upstream of the mist eliminators provides for the entrainment of liquid droplets and a uniform
air velocity distribution into the mist eliminators. Downstream of the mist eliminators is a 9
foot (2.74 M) section of ductwork which expands and transitions into a high efficiency wire
mesh collector. Liquid which is collected on the mist eliminators drains into a trough. Large
droplets leaving the mist eliminators are collected in a hopper upstream of the wire mesh
collector. A final hopper provides for accumulation of water from the wire mesh collector
itself.
Four types of tests were run in the AWTF: Droplet size distribution which passes through the
mist eliminators was determined using a Greenfield Instruments Video Droplet
Analyzer(VDA). The VDA was located approximately 18 inches (0.46 M) downstream of the
mist eliminators. Actual distance varied due to different mist eliminator dimensions.
Collection efficiency was determined by weight measurement of water collected from each
drainage point of the facility. Collection efficiency for the mist eliminators was defined as:
Where:
Eff.
Mme
Mco
Mfina|
Eff=Mme/(Mm
Collection Efficiency
Mass collected by mist eliminators
Mass recovered in duct downstream of mist eliminators
Mass recovered by final wire mesh collector
(1)
Pressure drop across the mist eliminators was determined by static pressure measurements at
wall taps ahead of and downstream of the mist eliminators. These measurements were taken
during collection efficiency tests.
Cleanability tests were run by first coating the mist eliminator blades with vegetable oil and
then flour. These materials were selected based on some preliminary tests to find materials
which would adhere well to the blades, require substantial wash flow for removal and be
environmentally benign, Prior to being coated, the blades were painted black. This provided
good contrast between clean and coated areas. After the wash cycle, the mist eliminator
-------
assemblies were removed from the facility and disassembled. Wash effectiveness was
determined based on removal of the coating from the blades. The results from this simulation
agreed closely with actual field experience.
A total of six mist eliminator configurations were evaluated in the AWTF Collection
efficiency, droplet penetration and pressure drop for the first five are summarized in Figure
9 and Table 2. These compare performance of the mist eliminators at comparable operating
conditions. The collection efficiency test results are based on a liquid loading of 1.16 gpm/ft
(0.79 ka/M2 sec) with wash flow provided by two Bete WL-12-90 nozzles These conditions
are typical of the LS-2 unit.
£ 0.001 '-*-
o
0.00001
20
(6.1)
Horizontal Flow Carryover
Inlet Loading 1.2 gpm/ft2
.7
.07
.007 '
.0007
—«— STD 4 pass
• 3 pass
—*—3 pass with ribs
—Q— Curved with hooks
-X—2 stage 2 pass
—•—2 stage 2 pass with
ext.
™Q—2 stage 2 pass with
ext. ribs
Velocity, ft/sec (m/sec)
25
(7.6)
Horizontal Flow Pressure Drop
Inlet Loading 1.2 gpm/ft2
—•— STD 4 pass
—••-3 pass
—4— 3 pass with ribs
-Q— Curved with hooks
X 2 stage 2 pass
—•—2 stage 2 pass with
ext.
—a—2 stage 2 pass with
ext. ribs
Figure 9
Horizontal Flow ME Performance
10
-------
The STD 4 pass design was tested first to obtain baseline performance information. This
design has provided acceptable mist elimination performance in operating units but has been
hard to clean , especially in the rear two passes. The rear pass of the 4 pass design was
removed to provide a 3 pass design. Tests of this design indicated reduced collection
efficiency, while cleanability of the final pass was still limited. Ribs were added at the
inflection points of the chevrons in an attempt to improve performance. These ribs resulted in
a further decrease in collection efficiency. The curved two pass ME vane with hooks was also
tested in a one stage configuration. Test results showed the collection efficiency of a single
stage (2 pass) ME vane was low compared to others considered, only, was poor. Results of
testing of similar designs ' in a two stage configuration indicate that collection efficiencies
are in the range of 0.00003 gpm/ft2 ( 0.00002 kg/M2 sec). Cleaning tests of this design
showed that the hooks effectively stopped the movement of liquid along the blade and
prevented wash water from reaching the 2nd pass.
Based on previous studies which showed good performance of two stage systems, the
standard 4 pass design was redesigned as a two stage, two pass system. CFD results had
indicated that the first passes of the 45 degree chevron design were effective in initial
collection of drops and it was known that the first two passes of this design were cleanable. A
straight trailing edge of 1.5" (38 mm) was used on both stages. The stages were separated by
a distance of 30" (762 mm) to permit access and wash nozzles for each stage, if required.
Tests of this design showed a slight decrease in collection efficiency compared to the standard
4 pass design. Observations during testing indicated that drops were being re-entrained off the
inflection point of the chevrons. The trajectories of the drops were being influenced by the air
flow in the rear of the blades such that the droplet passed behind the trailing edge of the
blades as shown in Figure 10. Extended
trailing edges were added to the blades as
shown, the length selected to intersect the paths
of re-entrained drops. The extended trailing
edges also provide additional surface area for
drainage. This design is also cleanable.
Previous tests and field experience had shown
that the first two passes of the four pass design
could be cleaned. The extended trailing edge
does not inhibit the movement of wash flow
along the blades.
1.5 inches (38 mm)
Orignial Length
4.5 inches (114 mm)
Extended Length
Figure 10
Vane Profile
As can be seen from Figure 9, the collection efficiency of the two stage design with extended
trailing edges (2x2+) was better than all other designs tested. The improved efficiency of the
two stage configuration is attributed to the gap between stages. This allows most of the
liquid captured by the first stage to drop out before reaching the second stage. This has the
effect of providing an extended drainable surface. In the four pass design, liquid is pushed
11
-------
along the blades by the air flow. This can be seen as droplet tracks and stream tubes of water
which move along the blades. In the two stage design, the drops which reach the rear of the
blades and are shed, are now in free fall. Most contact the floor before reaching the next stage.
The second stage therefore has a much lower liquid loading than the first stage. Since re-
entrainment is a function of liquid loading, it is greatly reduced for the second stage.
The 2x2+ blades were modified by the addition of 1/4" plastic tubing to the trailing edges.
This rib was added with the intent of further improving drainage and minimizing re-
entrainment. Tests results showed no definitive improvement with the ribs, however this may
be partially due to being at the limits of experimental accuracy. Furthermore this rib design
was not considered optimum. A rib with a flattened rear section would provide increased
surface area in a low velocity zone to improve drainage.
Ranking of mist eliminator performance in terms of droplet penetration as a function of size is
not as simple as ranking collection efficiency and pressure drop. In order to evaluate
penetration (vs. re-entramment) it was necessary to introduce a range of droplet sizes which
spanned the penetration topsize of the ME blades. Based on CFD results and previous
experience, this was expected to be in the range of 25 microns. A pair of twin fluid atomizers
was used to provide an inlet size distribution of 0 to 70 microns.
Table 2
Drop sizes Penetrating Mist Eliminator
Configuration
STD 4 pass
3 pass
Curved with hooks
2 stage 2 pass with ext.
2 stage 2 pass with ext.
and trailing edge ribs
2 stage 1 pass with ext.
20 ft/sec inlet velocity
Dmax
26.1
28.1
60
29.8
37.9
49.5
D90
24.8
24.6
60
27.5
26.8
36.8
D50
11.4
14.1
13.9
14.1
16.1
24.8
25 ft/sec inlet velocity
Dmax
17.6
30.8
23.8
34.3
24.5
D90
14.9
27.3
20
29.4
21.1
D50
10.7
15.2
11.8
21.2
9.1
The inlet loading was kept low (less than 0.1 gpm/ft?) to prevent re-entrainment. As
previously discussed, droplet size distribution downstream of the mist eliminators was
measured by a Video Droplet Analyzer (VDA). The VDA cannot distinguish between
droplets which pass through the blades and re-entrained droplets. The information shown in
Table 2 must therefore be regarded with this in mind. It is unlikely that 60 micron droplets
get through the curved mist eliminators with hooks. The single 60 micron droplet which was
12
-------
measured is most likely a re-entrained droplet. It can be surmised that none of the mist
eliminators tested permits droplets large than 50 microns to pass through the vanes. Past work
by others has shown that the performance of mist eliminators can be improved by tilting the
tops of blades into the flow. This enhances drainage along the blades due to the component
of air flow along the blades. A series of tests were run to evaluate the performance of the 2x2+
mist eliminator blades with a 15 degree forward tilt, as shown in Figure 11.
Based on their superior performance at 20
and 25 ft/sec (6.1 and 7.6 M/sec), the 2x2+
configurations were tested at higher gas
velocities. As shown in Figure 12a & b
collection efficiency is still excellent at 30
ft/sec (9.1 M/sec). At 35 ft/sec (10.7 M/sec)
collection efficiency is still comparable to
the standard four pass design at 20 ft/sec
(6.1 M/sec). At up to 40 ft/sec (12.2 M/sec),
there was no measurable carryover for the
forward tilted design. This was seen at inlet
loadings of up to 2.2 gpm/ft2(1.5 kg/M2 sec)
ME BLADES
Figure 11
Tilted ME Blades
0 10000
0 01000 -
Q.
01 0 00100 '.
I
£ 0 00010 .
ro
0
I
Two Stage Two Pass Carryover
Inlet Loading 1.2 gpm/ft2
o
;
—^ /
—* — 2 stage 2 pass
_g — 2 stage 2 pass with exL
—* — 2 stage 2 pass with
exlribs
— *— 2 stage 2 pass with ext
15deg
20 40
(6.1) Velocity, ft/sec (m/sec) (12 2)
Figure 12a
Two Stage Two Pass Results
13
-------
0 12'
CS
c
£
If)
2
(
Two Stage Two Pass Pressure Drop
Inlet Loading 1.2 gpm/ft2
X^ ^--X!i!ge „
,^ ^^~-^" m
^^jf^ ^^ ra
^^^^J^^^ 14s — *— 2stage;
i Sn — * — 2 stage;
ext.nbs
n ?«; rin a.s 40 ^
1) Velocity ft/sec (m/sec) (12"2>
pass with
Figure 12b
Two Stage Two Pass Results
Based on the success of the horizontal gas flow configuration, the air water test facility was
modified for, and the 2x2+ mist eliminators were evaluated in vertical flow. The lowest
velocity at which carryover was measurable was 18.9 ft/sec (5.76 M/sec) as shown in Figure
13. Based on results in EPRI reports (1,2), this performance is as good as or better than that
of other commercially offered two stage mist eliminators for vertical flow, including the
current ABB design. As with the horizontal gas flow application, these mist eliminators will
be cleanable on line. ABB is therefore commercializing the 2x2+ design for both horizontal
and vertical gas flow applications.
-07 j*
.s^ Tk \ \ ~a>
S / l
/ / I < .007
^ / / !
/ / ' nnn?
— t j; *. n; n i^ 00007
5 16 17 18 19
5> Velocity, ft/sec (m/sec) (5.8)
— < — ABB 2 Stage
—a — ABB 2x2+
Com2 2 Stage
— * — Com3 2 Stage
Figure 13
Vertical Flow Results
14
-------
Summary
The goal of identifying a mist eliminator design which can be washed on-line without the lost
of collection efficiency appears to be satisfied with the 2 stage 2 pass design with extended
trailing edge. This design will be field tested this summer as it is in the process of being
fabricated for installation in the LS2 absorber at the Ohio Edison Niles plant.
However, greater benefits are anticipated both in performance and operating velocity by
slanting the vanes in the horizontal gas flow orientations. This arrangement should have an
added benefit in handling greater gas distribution variances without liquid re-entrainment.
Some additional model testing of the 2x2+ ME vane is planned to complete evaluation of its
suitability as an improvement for vertical flow application. Capability of higher operating
velocities for this arrangement, needless to say, would be beneficial both in capital cost and
absorber performance.
References
1. FGD Mist Eliminator System Design and Specification Guide, Mclntush, K.E.,
Jones, A.F., Lundeen, J.E, Dec. 1993, EPRI TR-102864
2 Results of Mist Eliminator System Testing in an Air-Water Pilot Facility, Jones, F.
A., Mclntush, K.E., Rhudy, R.G., Bowen C.F.P., EPRI, Presented at the 1991 SO2
Control Symposium, Washington D.C. December 1991.
15
-------
PHASE II: THE AGE OF HIGH VELOCITY SCRUBBING
BRUCE W. LAM and MANYAM BABU
Dravo Lime Company
3600 Neville Road
Pittsburgh, Pennsylvania 15225
Abstract
The Clean Air Act Amendments require substantial reduction of SOa emissions from coal-fired
generating facilities and offer incentives to exceed targets by accumulation of excess emission
allowances. Although considered costly, wet scrubbing was the leading post-combustion
technology to achieve these reductions in Phase I. However, the emergence of other SC>2
control methods including the broad acceptance of low sulfur coal has prompted investigations
into reducing the capital and operating costs of scrubbing for those facilities evaluating this
option for Phase n compliance. One method to reduce the costs associated with scrubbing is
to increase the scrubber gas velocity as demonstrated at Northern Indiana Public Service
Company's Bailly Station and at Ohio Edison Company's Niles Generation Station. Scrubbers
designed for higher velocities will have a smaller cross sectional area and may also permit a
reduction in the number of required operating scrubbers. Dravo Lime Company in cooperation
with the Ohio Coal Development Office and Cinergy Corporation has conducted evaluations of
high gas velocity scrubbing in magnesium-enhanced lime FGD processes in a 4.5 MW pilot
plant. These studies have shown a significant decrease in the ratio of liquid volume to gas
volume required to scrub the same quantity of gas as the scrubber velocity increases. This
paper will discuss operating conditions for 98% S02 removal at gas velocities of 10 to 25
ft/sec in both vertical and horizontal scrubbers and potential cost impact, and comparisons
between magnesium-enhanced lime and limestone FGD systems.
Introduction
With the advent of the Clean Air Act Amendments and its emissions allowances, market-
oriented strategies to meet or exceed compliance emissions with respect to the regulations are
encouraged. The list of options to achieve compliance include, but are not limited to,
retirement of aging units, fuel switching, purchase of emission credits, and installation of FGD
systems. Therefore, each utility has the opportunity to mix or match the available strategies to
achieve the lowest cost plan to reduce SO2 emissions within their generating system.
-------
However, the market conditions which influence the economics of the various options continue
to change. The price of SO2 emission credits, currently below $90/ton of SO2, is far below the
pre-auction estimates of $300/ton of SO2. Predictions on the future price, near term or long
term, are clouded as a result of limited involvement of utilities in conducting trades and
potential limits placed on utility usage of these credits by regulatory commissions. Low sulfur
coal prices have been stable as a result of expanded production and improved rail
transportation. Limited long term contracts, the impact of Phase II units, and possible
obstacles posed by state and local government are unknown influences on the future cost of
this fuel. Upcoming regulations involving NOX, air toxics, C02, and possibly more stringent
SO2 emissions and the impact of "free-wheeling" electricity distribution provide additional
uncertainties in the selection of the "correct" emission control strategy.
It is not surprising in this market-oriented atmosphere for obtaining the desired Phase I
emission levels, that the majority of the reductions were achieved by fuel switching. Switching
to lower sulfur coals provided the advantages of both minimal operating and capital costs to
utilize the new fuel in order to achieve the necessary SO2 reductions. More than 62% of the
affected Phase I units opted for this method.
Calcium-based wet scrubbing, which continues to be the leading proven commercial post-
combustion FGD technology, captured 10% of the market based on the number of affected
units. Typically, these were large, high capacity units, firing medium to high sulfur coal which
could generate sufficient emission credits to off-set the necessary reductions from other
generating units within the utility system. The average capital cost for these units was
$233/kW and the estimated levelized cost was $350/ton of SO2.
For wet scrubbing to be competitive in today's S02 removal market, drastic reductions in both
capital and operating costs need to be realized. By utilizing conventional scrubber design
standards (300 MW facility, two operating plus one spare modules, 50% capacity each with
gas velocity of 10 ft/sec), the scrubber island can account for nearly on third of the capital
outlay of a FGD installation. The multiple recirculation pumps required to achieve today's
targeted S02 removal of+95% increase the parasitic load as a result of the pump horsepower
required for supplying the spray headers and the fan horsepower necessary to overcome the
corresponding pressure drop of the sprays. Improvements in the design and operation of the
scrubber will greatly improve the competitiveness of wet FGD.
Development work at Dravo Lime Company has focused on many aspects of decreasing the
related costs of wet scrubbing when utilizing magnesium-enhanced lime as the neutralizing
reagent. The ThioClear process, which generates magnesium hydroxide, an excellent waste
stream reagent, and wallboard quality gypsum as the by-products of the scrubbing process is
one example. This paper will focus on the development of high velocity scrubbers at the
Miami Fort pilot plant and the resulting benefits.
-------
Test Objectives
The advantages of successfully operating a FGD scrubber at higher flue gas velocities than the
current design standard are many. For greenfield and retrofit applications, there would be
smaller and/or fewer scrubber modules required to scrub the total flue gas from the facility,
greatly reducing the initial capital investment. FGD facilities which are currently operating
may desire to reduce or eliminate gas bypass so as to increase the quantity of SCh removed in
order to accumulate emission credits or offset emissions from other generating facilities.
Utilities which have scrubbed and nonscrubbed generating units at the same site may opt to
divert a portion of the nonscrubbed generating unit's flue gas to the existing scrubber as part of
a SC>2 control strategy.
The scrubbing liquor associated with a magnesium-enhanced lime FGD system is highly
alkaline due to the presence of magnesium sulfite. It rapidly neutralizes the captured SO2,
making the absorption process controlled by the rate at which the SC>2 transfers from the gas
phase to the liquid phase. To take advantage of the alkalinity of a magnesium-enhanced lime
FGD system, a mass transfer enhancement device such as a sieve tray is typically installed in
the scrubber to improve gas/liquid contact. Because the tray provides additional mass transfer
capability within the scrubber, the reliance on the mass transfer characteristics of spray droplets
alone is diminished. Thus the amount of liquor required to be sprayed can be reduced.
An alternative (or supplement) to a mass transfer device such as a tray can be the operation of
the scrubber at velocities greater than today's standards. Operating at higher gas velocities can
provide the following benefits in the spray zone with regards to mass transfer improvements:
• Higher gas velocities will increase the turbulence and interaction between the flue gas and
the spray droplets which decreases the gas-film resistance.
• Increasing the gas velocit • increases the shear forces between the up-flowing gas and the
counter-current spray droplets which enhances the movement of fluid within the droplet.
This fluid movement renews the liquid boundary film layer and decreases its corresponding
resistance.
• The increase in the gas velocity will decrease the rate at which droplets fall through the
scrubber and will increase the percentage of droplets which will be suspended within the
scrubber. Both of these factors contribute to an increase in the mass transfer surface area
per unit volume of the scrubber.
Similar phenomena resulting from high velocity operation can also enhance mass transfer of the
gas and liquid on the sieve tray.
-------
The relationship between the mass transfer coefficient and SO2 removal is defined by the
following equation:
H-K a-P / % SO, Removal
NTU = — J— = -
where: NTU SO2 removal as number of transfer units (dimensionless)
H height of scrubber (ft)
Ke overall mass transfer coefficient (— - - )
g ft -hr-atm
a mass transfer surface area per unit volume (ft2 / ft')
G flue gas flowrate ( — )
ft -hr
P total system pressure (atm)
Therefore, the benefits in mass transfer outlined above will have a positive impact on SO2
removal provided the product of the enhancements to the mass transfer coefficient (minimizing
film resistance) and the surface area (increasing droplet surface area per scrubber volume) is
greater than the detrimental offset due to the increased gas velocity.
To fully investigate the potential benefits of high velocity scrubbing in a magnesium-enhanced
lime FGD system, Dravo Lime Company established a test plan which included evaluations of
scrubber configurations being utilized at today's commercial FGD facilities. These scrubber
configurations included a vertical spray tower, a vertical spray/tray tower, and a horizontal
scrubber. During the evaluation of each of these configurations, various operating parameters
were monitored. These parameters included the SC>2 removal as a function of L/G and
velocity, the scrubber pressure drop as a function of L/G and velocity, and the carryover
escaping from the demister section as a function of velocity. For brevity, only the 862 removal
results are presented in this paper.
The scrubber configurations utilized for the vertical spray and spray/tray evaluations are typical
of today's commercial designs. This three foot diameter scrubber is equipped with a two stage
mist eliminator section. Three spray headers were available, each fitted with a Bete MP1 125
90° full cone whirl style nozzle. The tray, which was removed during the spray tower tests,
was a sieve-type tray with 1-3/8" diameter holes providing 40% open area. The mist
eliminator section housed two vertical flow chevron demisters. The three mist eliminator wash
headers were operated separately at 20 minute intervals to wash the targeted demister face for
90 seconds with process liquor. A two stage horizontal flow mist eliminator was installed in
-------
the outlet duct of the vertical scrubber to capture carryover escaping the vertical demist ers
during the high velocity tests.
Commercial application of vertical flow demisters at high velocities (greater than 20 ft/sec) is
not practical due to re-entrainment of accumulated liquor from the mist eliminator. This
limitation results from the accumulated liquor attempting to return to the scrubber due to
gravity while being opposed by the velocity of the flue gas which is exiting the tower. To
avoid this short-coming, Dravo Lime Company designed and installed a horizontal scrubber
which utilizes horizontal demisters which can effectively function at higher gas velocities than
the vertical design. Additional advantages of the horizontal scrubber include reduced height
requirements when compared to vertical scrubbers and minimal impact in connecting to
existing ducting. These advantages translate into greater flexibility in materials of construction
and lower power consumption by pumps for meeting hydraulic head requirements and by fans
due to lower inlet and outlet pressure losses.
Utilization of the high velocity horizontal scrubber in the crossflow mode of gas/liquid contact
is not new to the FGD industry. However, performance and reliability of these earlier
scrubbers were adversely effected as the length of the vessel was reduced. Nonuniform gas
flow distributions were created in the spray zone from top entry spray nozzles which
compressed the gas toward the scrubber floor and reduced the effectiveness for SC>2 removal.
The combination of the skewed velocity profile and spray headers installed immediately in front
of the demisters contributed to excessive carryover exiting the mist eliminators.
The horizontal scrubber evaluated at the pilot plant was designed to minimize the above
drawbacks of the prior commercial designs. The cross sectional dimensions of the scrubber
were 51" x 25" Each of the three available recycle pumps supplied two spray headers, each
equipped with two Bete TF40 fall cone spiral style nozzles. These spray headers were located
within the cross sectional area of the scrubber to minimize gas maldistribution and oriented
counter-current to the gas flow to enhance mass transfer. Taking advantage of the nonscaling
tendencies of the magnesium-enhanced lime FGD process, four layers of Kimre Kon-Tane
37/97 tower packing were installed downstream of the spray zone. The packing further
reduced gas maldistribution, enhanced mass transfer, and acted as a predemister. The demister
section consisted of two stages of horizontal impingement separators. As in the vertical
scrubber tower, three wash headers on the same time sequence were utilized. Due to
insufficient space downstream of the scrubber for the installation of a second set of demisters
for carryover collection, the downstream condensate drain was monitored as a relative
indication of demister performance.
One must be cautioned concerning the direct interpretation of pilot scale scrubber performance
to commercial size facilities. In general, the SC>2 removal achieved at the pilot plant is lower
than that of a utility FGD scrubber at similar operating conditions. The droplets exiting the
-------
spray nozzles in the scrubbers utilized for these studies impact the walls of the scrubbers within
1 to 1.5 feet from the nozzle. The liquor lost to the walls tends to have minimal beneficial
impact on SCh removal. This is not the case in the utility scrubbers where greater utilization of
the liquor is achieved due to minimal spray contacting the walls. Prior testing at the Miami
Fort pilot plant has shown that up to 50% of the liquor entering the scrubber does not
contribute to S02 removal when duplicating operating conditions of commercial facilities.
Vertical Spray Scrubber Test Results
Figure 1 details the results of parametric SOj removal studies utilizing the vertical spray
scrubber. The data points were generated by establishing a set velocity within the scrubber and
adjusting the flowrate among the three available spray nozzles to achieve the target removals
of 86, 95, and 98% which correspond approximately to 2, 3, and 4 NTU's. For these studies,
the scrubber liquor alkalinity averaged 2250 ppm (measure as comparable carbonate) and the
inlet flue gas S02 concentration was controlled at 2500 ppm by utilizing an on-site SC>2 spiking
system. In order to minimize the vast differences in the surface areas produced by droplet
formation at different flowrates, the flowrates to the individual nozzles were restricted between
60 to 90 gpm. Therefore, to increase or decrease the L/G to obtain the desired SO2 removal,
operation of one or two of the three recycle pumps may be initiated or terminated as required.
Even though the individual nozzle flowrates were limited between 60 and 90 gpm, the
corresponding Sauter mean diameters which correlate to droplet surface area varied between
2100 to 1600 microns according to the manufacturer specifications.
-------
Figure 1. VERTICAL SPRAY SCRUBBER - SO2 REMOVAL
Figure 1. VERTICAL SPRAY SCRUBBER - SO2 REMOVAL
Subsequent tests were performed to quantify SC>2 removal differences resulting from the
variance in the droplet size. These tests entailed operation of one nozzle at 120 gpm compared
to two nozzles at 60 gpm each at the velocities of 10 and 20 ft/sec. The removals at each
velocity were virtually identical whether one or two nozzles were utilized. Two phenomena
may explain these observations. The first is that within the droplet sizes observed, the
overriding driving force for SOz removal is the difference in the relative velocities of the flue
gas to the liquid as it exits the nozzle, irrespective of droplet sizes. The second explanation is
that by dividing the 120 gpm between two nozzles within the pilot scale scrubber, the decrease
in droplet surface area at the lower nozzle flowrates may be compensated by having two
distinct mass transfer spray zones. In either case, the implication of this observation is that
-------
multiple low pressure nozzles may be as effective as fewer high pressure nozzles at the same
header flowrate. The corresponding pump for such an application would have lower hydraulic
head requirements and consume less energy.
In examining the data presented in Figure 1, one trend is evident. Significant reductions in L/G
to achieve the same removal can be realized by increasing the gas velocity within the scrubber.
This trend indicates that the product of the mass transfer coefficient and the mass transfer
surface area is positively influenced by one or more of the potential benefits of enhancing mass
transfer at higher velocities described earlier. Furthermore, the proportional increase in this
product more than compensates for the proportional decrease of SC>2 removal which would be
expected with the reduced residence time resulting from the increased gas throughput within
the scrubber.
For example from Figure 1, 95% SO2 removal is achieved at a L/G of 60 for 10 ft/sec. If the
corresponding liquid flowrate was held constant and the flowrate of flue gas to the scrubber
was increased by 60% to obtain a scrubber velocity of 16 ft/sec, the resulting L/G would be
37.5 However, the pilot plant data shows that the L/G required for 95% removal at 16 ft/sec
is only 28 L/G, 25% below that which is theoretically available. These findings suggest that
existing magnesium-enhanced lime FGD facilities constructed with multiple scrubber modules
designed for 10 ft/sec and a fixed liquor flowrate can maintain or even exceed the design SC>2
removal efficiencies by reducing the number of operating modules in order to increase the flue
gas velocity within the remaining units. Obviously, sufficient fan capacity and adequate mist
eliminator performance at the higher velocities are prerequisites.
A second trend identified during the analysis of the parametric studies involved the curvature
of the trend lines. The S02 removal trend lines for scrubber velocities of 10, 14 and 16 ft/sec
have a slight curve to the right. The remaining trend lines at the higher velocities have a slight
curve to the left. Visual observations, as well as carryover determinations, have identified that
accumulation of scrubbing liquor occurs on the mist eliminators at gas velocities in excess of
18 ft/sec. This phenomena has the demisters effectively functioning as a packing. Therefore
substantial increases of SO2 removal occur as a result of the increase in the mass transfer
surface area from the demister as opposed to the removals at the lower velocities which are
dependent solely on the surface area of the spray zones. In commercial applications, the
resulting carryover emissions and increased pressure drop due to the accumulation would make
operating at velocities of 18 ft/sec or greater with this style of mist eliminators impractical.
Vertical Tray Scrubber Test Results
Because many of the magnesium-enhanced lime FGD systems in operation today utilize a tray
to enhance mass transfer, high velocity testing analogous to that of the vertical spray scrubber
evaluations was conducted in the three foot diameter tower with a 40% open area tray installed
-------
below the spray zone. The results of these studies are shown in Figures 2. During the initial
parametric evaluations, it was determined that operation above a gas velocity of 16 ft/sec was
impractical. Above this velocity, the hydraulic dynamics of the tray provided an unstable
operating condition. The liquid on the tray would cycle between loading and draining.
Corresponding to these two events, both the SO2 removal and scrubber pressure drop would
increase while liquor accumulated on the tray and both subsequently decreased as the tray
drained. Modifications which could be made to the tray to limit this phenomena were not
undertaken for these evaluations.
Figure 2 details the results of the vertical tray scrubber parametric SO2 removal studies. As
was expected, utilization of a tray enhances the mass transfer for SO2 removal above that
which was observed for the spray tower. Removals of 95% at velocities of 10, 14, and 16
ft/sec required a L/G of 44, 31, and 22, respectively. Whereas the spray tower operating at the
same velocities required a L/G of 60, 38, and 28, respectively. However with the utilization of
the tray, the incremental increase of the SO2 removal resulting from increasing the scrubber
velocity was not as great as that of the spray scrubber. For example from Figure 2, 95%
removal is achieved at a L/G of 44 for 10 ft/sec. If the flue gas flowrate was increased by 60%
to achieve an scrubber gas velocity of 16 ft/sec while maintaining a constant liquid flowrate,
the L/G would be reduced to 27.5. However, the data on Figure 2 indicates that a L/G of 22
would be sufficient. This value is 20% below that which is theoretically available whereas the
spray scrubber achieved a 25% decrease. Because of the enhancement of mass transfer on the
tray, the percentage of this theoretical reduction was not as great as that of the spray tower.
However, these reductions are significant and the overall L/G required when utilizing a tray is
substantially below that necessary in a spray tower to achieve comparable SO2 removals.
Figure 2. VERTICAL TRAY SCRUBBER - SO2 REMOVAL
-------
Horizontal Scrubber Test Results
As described earlier, horizontal scrubbers have many advantages including the utilization of
horizontal demisters which are more efficient than the vertical chevrons at velocities in excess
of 20 ft/sec. Therefore, to fully explore the feasibility of high velocity S02 removal, the
horizontal scrubber was required for evaluations at velocities of 20 and 25 ft/sec. Before these
parametric SO2 removal studies were initiated, detailed velocity traverses were conducted to
determine if the gas maldistribution common to commercial scale scrubbers existed. Figure 3
is representative of the resulting velocity profile between the packing and the first stage of the
demisters. The plug-like flow profile insures that the SO2 removal capabilities of this scrubber
and the performance of the mist eliminator will not be degraded by velocity variations within
the scrubber.
Figure 3. HORIZONTAL SCRUBBER VELOCITY PROFILE
The results of the parametric S02 removal studies are detailed in Figure 4. Each consecutive
data point on these trend lines represents the contribution of an additional 150 gpm to the
scrubbing process due to increasing the number of recycle pumps operating. The Sauter mean
diameter of the resulting droplets attributed to the nozzles was approximately 1000 microns.
As in the prior studies in the vertical tower, increasing the velocity of the flue gas permitted a
decrease in the L/G to achieve the same S02 removal. However, the large decreases in L/G to
achieve the same removals in the vertical tower evaluations were not realized in this testing.
Three explanations can be offered for this observation. The first is the realization that the
horizontal scrubber with counter-current oriented nozzles is not a true counter-current mass
transfer device. The effect of gravity on the droplets in a horizontal scrubber tends to remove
-------
these droplets from the gas stream and therefore limit their ability to absorb SC>2. The second
reason is due to the narrow walls (~2 feet apart) of the pilot plant horizontal absorber which
increase the potential for wall wetting. As with the effect of gravity, this mode of removing
liquor from the flue gas decreases the mass transfer surface area available from the spray
droplets. A third explanation is that the mass transfer process at these higher velocities with
the fixed alkalinity utilized in these evaluations may no longer be in the gas-film limited regime.
Decreasing the availability of excess alkalinity by operating at higher velocities and lower
L/G's can permit liquid-film resistance to influence the absorption process.
Figure 4. HORIZONTAL SCRUBBER - SO2 REMOVAL
Magnesium-Enhanced Lime And Limestone Pilot Plant Data Comparison
Operation of the Miami Fort pilot plant is limited to magnesium-enhanced lime FGD operation.
Therefore, a direct comparison of the data presented within this report with that comparable to
limestone operation utilizing the same equipment was not feasible. However, data was
presented at the 1995 SC>2 Control Symposium pertaining to pilot plant high velocity SCh
removal studies of a limestone forced oxidation (LSFO) FGD system. These pilot plant
evaluations were conducted at the EPRI Environmental Control Technology Center. The
vertical spray tower utilized in these studies has a diameter of five feet and was equipped with
three levels of spray headers that were each fitted with four Bete FST88 full cone spiral spray
nozzles. The larger scrubber diameter and the finer spray nozzles provide a slight advantage to
the mass transfer characteristics of this pilot plant when compared to those of the Miami Fort
facility.
Comparisons of the SO2 removal trend lines presented in the EPRI paper to those generated in
the vertical spray and tray towers from Miami Fort pilot plant operation are shown in Figure 5.
-------
Because the 15 ft/sec condition was not tested at the Miami Fort pilot plant, interpolation of
the 14 and 16 ft/sec data was conducted to provide a trend line for the comparisons. As
evident from the trend lines, the highly alkaline scrubbing liquor associated with magnesium-
enhanced lime and the resulting approach to gas-film limited mass transfer characteristics
enable exceptionally high SO2 removals with minimal L/G when compared to the limestone
scrubber operating at similar velocities. For example, the LSFO process required a L/G of 95
to remove 90% of the SOa at and scrubber velocity of 15 ft/sec. In the magnesium-enhanced
lime process, the required L/G to achieve comparable removals was 26 for the spray tower and
22 for the tray tower. These values represent more than a 3 to 4 fold reduction in the liquid
rate required to scrub the same volume of gas. As illustrated in earlier discussions, the gas-film
limited mass transfer characteristics of the magnesium-enhanced lime scrubber are greatly
enhanced by operating the scrubber at higher velocities. Comparison of the improved liquor
alkalinity utilization as the velocity of the spray scrubber is increased from 10 to 15 ft/sec
shows that a 41% decrease in the L/G was achieved with magnesium-enhanced lime compared
to only a 28% decrease with limestone.
Figure 5. SCRUBBER REAGENT EFFECT ON PERFORMANCE
LSFO Spray @1 Ofps
LSFO Spray @15fps
Mg-Ljme Spray @ 10fps
Mg-Lime Spray @ 15fps
Mg-Ljme Tray @ 10fps
Mg-Lime Tray @ 15fps
Economic Analysis
The implications of the magnesium-enhanced lime FGD pilot scale data from high velocity
testing on the design and operating criteria of FGD scrubbers and the corresponding capital
and operating costs are significant, especially when compared to LSFO FGD. For both new
plant construction and retrofit applications, capital cost savings can be realized as a result of:
• Fewer scrubbers and corresponding auxiliary equipment will be required in the construction
of a high velocity FGD facility.
-------
• Scrubbers with smaller diameters to accommodate the higher gas velocities and lower
height requirements due to fewer spray headers for lower L/G operation can be
constructed.
• Fewer and smaller pumps can be utilized to meet the reduced L/G and hydraulic head
requirements.
• The lower pressure drop resulting from the low L/G operation at the higher velocities will
enable fan sizing requirements to be minimized.
Associated with these capital costs savings will be the decrease in operating expenditures
realized from reductions in power consumption due to utilizing fewer scrubbers and smaller
and fewer pumps and fans. Many of the enhancements demonstrated in the pilot scale tests are
worthy of further investigations for commercial applications to existing FGD facilities.
Compliance strategies which may entail eliminating the scrubber flue gas bypass or connecting
an unscrubbed generating facility to an existing FGD system could be accomplished with
minimal capital investment when operating the scrubbers at higher velocities. FGD facilities
which have multiple scrubbers may find it beneficial to reduce the associated parasitic load by
reducing the number of operating scrubbers. Limestone-based FGD scrubbers which are over-
designed based on magnesium-enhanced lime requirements have the potential to substantially
reduce their scrubber power consumption costs by switching reagents and operating fewer
scrubbers at the higher velocities.
The options available to existing FGD facilities to improve the economics of SO2 removal by
investigating high velocity scrubber operation are many. However, the site specific details
which factor into this economic analysis are beyond the scope of this paper. Figure 6 has been
provided to illustrate the potential capital costs of retrofitting a high velocity scrubber island
(Area 20 - SO2 Removal System; EPRI publication GS-7193) to an existing generating unit.
The basis of the analysis is a 300 MW plant, utilizing 2.6% sulfur coal with a FGD system
achieving 95% SO2 removal. The four cases illustrated are: 1) LSFO with two operating and
one spare, spray modules, designed for 10 ft/sec; 2) LSFO with one operating and no spare,
spray module, designed for 15 ft/sec; 3) magnesium-enhanced lime with one operating and no
spare, tray module, designed for 15 ft/sec; 4) magnesium-enhanced lime with one operating
and no spare, horizontal module, designed for 25 ft/sec. Capital costs associated with the air
compressor requirements necessary for LSFO were not included in this analysis. As can be
seen in Figure 6, the high velocity LSFO FGD system can reduce the scrubber island capital
costs by 68% when compared to the conservative design standards. However, the superior
efficiency that magnesium-enhanced lime FGD scrubbers have in SO2 removal enable the tray
and horizontal scrubbers to further decrease this cost by 47% and 57% respectively, when
compared to the advanced LSFO scrubber.
-------
Figure 6. SCRUBBER ISLAND INSTALLED CAPITAL COSTS
LSFO 2+1
Spray 10 fps
LSF01+0 Mg-Lime1+0
Spray 15 fps Tray 15 fps
SCRUBBER CONFIGURATION
Mg-LJme1+0
Horizontal 25 fps
Summary
The conservative FGD design consideration of utilities which have been fairly common over
the past decades are not realistic in the new "market-oriented" atmosphere associated with the
SO2 compliance requirements of Phase II regulations and "free-wheeling" electricity
distribution. The compelling need to be a low cost producer of electricity will force utilities to
investigate new ways to achieve their objectives, especially when addressing SO2 compliance.
The testing conducted at the Miami Fort pilot plant to demonstrate that wet scrubbing can be a
cost effective method for SO2 compliance has shown that the design standard for the velocity
of a vertical scrubber should be increased to +15 ft/sec. When utilizing magnesium-enhanced
lime as the reagent for the scrubbing process, the highly alkaline scrubbing liquor enables the
scrubber to operate at 25% of the L/G which would be required in a comparable LSFO system.
Additional testing in the pilot scale horizontal scrubber was successful at maintaining +95%
SO2 removal while operating at gas velocities of 25 ft/sec. These results can be used as a
design basis for full scale applications and calibration of predictive models such as EPRI's
FGDPRISM. The benefits of reduced scrubber capital and operation costs from utilizing
magnesium-enhanced lime in high velocity applications can be realized in new plant
construction as well as existing lime and limestone FGD facilities.
Acknowledgments
The authors of this paper would like to thank the Ohio Coal Development Office and Cinergy
Corporation for their support of the project.
-------
References
1. Carey, T.R., Skarupa, R.C., Hargrove, Jr., O.W., and Moser, R.E., EPRIECTC Test
Results: Effect of High Flue Gas Velocity on Wet Limestone Scrubber Performance.
presented at 1995 S02 Control Symposium, Miami, FL, March 1995.
2. Keith, R.J., Ireland, P.A.., and Radcliffe, P., Utility Response to Phase I and Phase II Acid
Rain Legislation - An Economic Analysis, presented at 1995 SOa Control Symposium,
Miami, FL, March 1995.
3. Lani, B.W., College, J., and Babu, M., Results of ThioClear Testing: Magnesium-Lime
FGD with High SO2 Removals and Salable By-Products, presented at 1995 SO2 Control
Symposium, Miami, FL, March 1995.
4. Lani, B.W., Babu, M., and Johnson, H., Improvements in Wet Scrubbing Technology
Development at the Miami Fort Pilot Plant, presented at the Fifth International Conference
on Processing and Utilization of High Sulfur Coals, Lexington, KY, October 1993.
-------
Results of High Velocity
Single Absorber Operation
at OMU's Elmer Smith Station
Kevin D. Frizzell - Technical Services Superintendent
Owensboro Municipal Utilities
4301 Highway 60 East
Owensboro,Ky42302
Michael T. Hoydick - Process Engineer - FGD Applications
Wheelabrator Air Pollution Control
441 Smithfield Street
Pittsburgh, Pa 15222
Abstract
The Owensboro Municipal Utilities' (OMU) Elmer Smith Station is a two (2) unit, 441 MW coal
fired electric generating station located in Owensboro, Kentucky. An FGD system, comprised of
two (2) 67% capacity open spray chamber absorber modules, was installed for compliance with
Phase I requirements of the Clean Air Act Amendments. The FGD system supplied is of the
limestone forced oxidation type producing commercial grade gypsum. Since the commencement
of commercial operation, efforts have been made to utilize the absorber modules in an efficient
manner while maintaining performance and emission requirements. One example has been the use
of a single absorber module to scrub the flue gas that is normally treated by two operating absorbers.
This single absorber arrangement produces higher absorber gas velocities than conventionally
offered during Phase I FGD designs. This paper examines the performance results of the high
velocity absorber along with the additional flexibility the single absorber allows for plant operations.
Introduction
OMU's Elmer Smith generating station consists of two (2) coal fired boilers: Unit 1 is a 151 MW
cyclone fired boiler and Unit 2 is a 290 MW tangentially fired boiler for a combined power
production of 441 MW. As one of the electrical generating plants targeted under Phase I of the US
Environmental Protection Agency 1990 Amendments to the Clean Air Act of 1970, OMU's Elmer
Smith generating station selected a limestone based Wet FGD System to reduce S02 emission levels.
The flue gas from each unit is directed through, dedicated electrostatic precipitators and is then
Pagel
-------
combined to pass through the FGD system. A commercial quality gypsum byproduct is produced
and is sold to a wallboard manufacturer. The FGD system, consisting of two (2) 67% plant capacity
absorbers, has been in commercial operation since January 1, 1995. The flue gas flow diagram can
be seen in Figure 1.
Absorber Design Conditions
Each absorber is equipped with five recycle pumps which range in flow from 26,100 gpm to 30,000
gpm. The total recycle flow rate when all five recycle pumps are in operation is 140,000 gpm. An
integral recycle tank holds approximately 1,000,000 gallons of absorber liquor which produces a
liquid hold time of seven (7) minutes when five (5) recycle pumps are operating. Each absorber is
equipped with two (2) levels of WAPC designed horizontal flow three pass mist eliminator modules.
The schematic of each absorber is shown in Figure 2.
As was the norm during the design of many of the Phase I FGD projects in the United States, the
absorbers at Elmer Smith are sized to produce a maximum saturated gas velocity of 10 ft / sec at
design conditions. Based upon this sizing criteria, each absorber is 47' -0" in diameter and produces
absorber gas velocities of 9.0 - 9.5 ft / sec when operating at 67% plant capacity (approximately 290
mw). When an absorber is operating at 290 mw, all five recycle pumps must be in operation to
produce the design L/G ratio of 130 gpm / 1000 acfm. The SO2 removal during this condition ranges
from 95.0 - 97.0% when the plant is burning the design (6.0 Ib SO2 / mmbtu) coal.
Although each absorber is designed for 67% capacity, operation at these conditions is rare and
usually occurs only when Unit 1 is off line for maintenance and Unit 2 flue gas is directed through
one absorber module. The "normal" operating mode at the Elmer Smith plant is to operate at full
load with each absorber treating 50% of the combined flue gas from both operating boilers. This
scenario produces absorber gas velocities of 6.5 - 6.8 ft / sec. During these conditions, four of the
five recycle pumps are in operation to produce the 150 L / G and the design S02 removal.
Full Scale, High Velocity Absorber Operation
During performance testing of the FGD system in May 1995, Wheelabrator Air Pollution Control
(WAPC) and OMU discussed the possibility of treating all of the flue gas in one absorber. OMU
seriously considered a 100% capacity single module design early in the planning of the system
before deciding on the two (2) 67% module approach and therefore was interested in the
performance of the single module concept.
Initially, there were several concerns with the proposed high velocity operation. These concerns
included: excessive FGD system pressure drop, excessive moisture and solids carryover into the
outlet ductwork and the wet stack, the capability of the absorber blowdown system to maintain
recycle tank density, the capability of the absorber slurry delivery system to maintain recycle tank
pH, shifts in gypsum particle size distribution, and whether the mist eliminator modules would
Page 2
-------
remain in place with the expected increase in drag forces due to the higher gas velocities. After
careful review of the expected operating conditions, it was determined that single absorber operation
was possible at full load, and a three hour test run was conducted on May 25, 1995. The absorber
gas velocity during this period was 13.0 -13.5 ft / sec.
After this initial three hour test run, the absorber module was opened for inspection. The inspection
revealed that all the mist eliminator modules remained in place and there was no solid deposition
on the mist eliminator modules or in the outlet ductwork. Based upon these encouraging results,
it was agreed upon that a longer test run (approximately 2-3 weeks) was needed to determine if the
absorber bleed system and slurry delivery system could maintain the necessary set points for
absorber recycle tank density and pH. Initial process calculations revealed that the existing
limestone slurry and absorber blowdown systems would be able to handle the increased sulfur
loading to a single absorber, although slurry pipeline velocities in these systems would be on the
upper end of the traditionally acceptable range.
The second high velocity absorber test run occurred from September 16, 1995 to October 4, 1995.
During this three week period, the boiler load to the scrubber varied from 150 - 430 mw with
absorber velocities ranging from 5.5 - 13.5 ft / sec. Again, the single absorber module performed
well. This test run confirmed that absorber pH and density control could be maintained with the
existing limestone feed and absorber blowdown systems.
The absorber was again opened for inspection after the three week test run to assure that the mist
eliminator modules remained in place and were not damaged. The inspection revealed that the mist
eliminators were clean with no signs of gypsum scaling or pluggage. The outlet ductwork was also
very clean with no signs of slurry solids carryover into the outlet ductwork.
Since the three week test run revealed few problems with the FGD system operation, an extended
test run, a minimum of three months, was planned. This extended test run of the high velocity
absorber occurred from December 13, 1996 to May 13, 1997, with no restrictions placed on boiler
operations. OMU operated the boilers as needed with loads ranging from 150 - 440 mw and coal
sulfur ranges of 4.5 - 8.0 Ib SO2 / mm btu. The absorber velocities during this period varied as boiler
load varied and ranged from 5.5 - 13.5 ft / sec.
To determine the performance of the absorber module during the high velocity periods, a third party
emissions testing company was hired to perform independent testing to determine flue gas flow rate,
S02 removal, particulate removal, moisture carryover, limestone utilization, and gypsum particle
size distribution on February 26 & 27, 1997. Six (6) two hour test runs were conducted during the
two day period with the boilers operating near full load. The absorber velocities during this two day
period ranged from 12.8 ft / sec to 13.1 ft / sec.
During this extended operating period, WAPC & OMU also collected FGD system data on an hourly
basis. Included in this data were flue gas flow rates, sulfur dioxide inlet and outlet concentrations,
L / G ratios, oxidation air flow rates, recycle tank pH, level, and density, and absorber pressure
drops. Approximately 1,000 data points were collected with absorber saturated gas velocities in
Page 3
-------
excess of 10 ft / sec. The results of the extended, high velocity absorber test run along with the
results of the third party testing are summarized on the following pages.
SOj Removal
The SO2 removal data (inlet and outlet concentrations, saturated gas flow rates) was collected
utilizing OMUs EPA certified Continuous Emission Monitoring System (CEMs). The SO2 removal
data was also checked for accuracy by comparing the GEM'S output during the informational testing
on February 26 & 27, 1997 versus the independent testing results which were obtained utilizing EPA
Method 6C.
As expected, the SO2 removal fell as the absorber velocity increased. The SO2 removal during the
testing period(s) varied as the boiler and sulfur loading changed and ranged from 88% to 95% at
constant pH values. The results of the SO2 removal at full load, 12.8 - 13.0 ft/ sec absorber velocity,
are shown in Table 1.
Table 1
Absorber Efficiency Summary
Test
Run
1
2
3
4
5
6
Absorber
Velocity
ft / sec
12.83
12.92
12.90
12.83
12.95
12.93
Absorber
Inlet
Ib / mm btu
6.11
6.44
6.37
6.34
6.39
6.37
Absorber
Outlet
Ib / mm btu
0.74
0.76
0.81
0.68
0.72
0.68
SO2 Removal
Percent
87.9%
88.2%
87.3%
89.3%
88.7%
89.3%
There are two principal reasons which explain the decrease in efficiency of the high velocity
absorber. The primary factor is the loss in L/G ratio associated with the larger flue gas volume
which were treated. Since the recycle pump volumes are fixed, and all five available pumps
were operating, any increases in absorber volume resulted in a smaller L/G. The absorber L /
G ratio was 130 when the absorber gas velocity was 10 ft / sec (@ design case) but fell to 100 L /
G as the absorber velocity rose to 13.0 ft / sec. A typical curve of the SO2 removal vs the L / G
ratio at a selected inlet sulfur load can be found in Figure 3.
Page 4
-------
Another reason for the decrease in the absorber efficiency is the shorter time period that the flue
gas has in the reaction zone. WAPC defines the reaction zone as the area from the centerline of
the inlet ductwork to approximately four (4) feet below the uppermost spray level. In the case of
the Elmer Smith absorbers, any particular flue gas molecule will experience 24% less time in
the reaction zone when the absorber velocity is 13.1 ft / sec as opposed to 10 ft /sec.
Particulate Emissions
The paniculate emissions from the absorber were determined utilizing EPA Method Five (5) at
the absorber inlet and at the wet stack location. Four (4) of the six (6) test runs revealed a net
decrease in paniculate, however there was significant scatter in the results. The maximum
removal efficiency which was experienced was 55%. The remaining two test runs indicated a
net increase in paniculate matter from the absorber inlet location to the wet stack location.
Complete results of the paniculate emissions testing are shown below in Table 2.
Table 2
Particulate Efficiency Summary
Test
No.
1
2
3
4
5
6
Date
02/26/97
02/26/97
02/26/91
02/27/97
02/27/97
02/27/97
Average
Location: Absorber Inlet A
Temp
°F
282
282
287
289
284
283
285
02
(%)
5.7
6.0
5.7
6.4
6.4
6.4
6.1
Dust
Concentration
gr/dscf
0.0466
0.0346
0.0250
0.0215
0.0284
0.0202
0.0294
Emission
Rate
lbs/105
Btu
0.0901
0.0676
0.0481
0.0434
0.0572
0.0406
0.0578
Location: Wet Stack
Temp
°F
127
128
129
128
128
128
128
0,
(%)
6.1
6.2
6.3
6.1
6.1
6.2
6.2
Dust
Concentration
gr/dscf
0.0204
0.0189
0.0235
0.0234
0.0186
0.0246
0.0216
Emission
Rate
lbs/108
Btu
0.0405
0.0375
0.0471
0.0463
0.0366
0.0489
0.0428
Removed
Efficiency
*
%
55.05
44.53
2.08
-
36.01
-
25.95
Moisture Carryover
Moisture carryover from the absorber was measured at the test ports located at the absorber
outlet transition section in accordance with method JIS Z8808 EPA Method Five (5). This
method inertially separates entrained droplets from the sampled gas streams via a Mitsubishi
droplet separator. Four (4) of the six (6) tests produced entrained moisture figures which were
below the target of 0.001 gpm / ft2. The remaining two produced moisture in excess of the
PageS
-------
target, with a maximum carryover of 0.0014 gpm / ft2. The results of the entrained moisture
testing are shown in Table 3.
Table 3 - Entrained Moisture Summary
Test
Run
1
2
3
4
5
6
Ave
Grains/dscf
0.0384
0.0630
0.1371
0.0817
0.1022
0.0466
0.0781
Volumetric Flow
dscfm*
1,092,143
1,117,773
1,045,416
1,089,710
1,122,514
1,091,769
1,093,221
Ibs/hr
359.72
603.54
1228.90
763.12
983.57
436.38
729.21
gpm/ft2
0.00042
0.00070
0.00142
0.00088
0.00113
0.00050
0.00084
Limestone Utilization
One of the results of operating the high velocity absorber was a reduction of the solids retention
time within the recycle tank. Since the Elmer Smith Station's gypsum by-product is of
commercial quality and is sold to US Gypsum, the limestone utilization within the system was of
some concern due to the reduction in the solids retention. The limestone utilization was
monitored daily by OMUs laboratory. During the testing periods, the solids hold time within the
recycle tank varied from 16 -24 hours due to the assorted sulfur contents of the fuel and the load
changes which occurred.
The limestone utilization was determined by EPRI Method 45 which has an alkalinity detection
limit of 2%. A total of 115 absorber samples were analyzed during the high velocity operating
periods. Of these samples, 112 (97%) had alkalinity numbers below the detection limit which
indicates that the limestone utilization was adequate to maintain the required gypsum purity even
though the solid hold time was essentially reduced by 50% from normal operation.
Gypsum Particle Size Distribution
This decrease in solids hold time within the absorber recycle tank was also expected to affect the
Page 6
-------
gypsum particle distribution. However, only a very slight shift in median particle sizes was
observed. The median particle sizes decreased slightly from the usual 46 - 48 uM (40 -50 hour
solids hold time) to 41 - 43 uM but was well .above the US Gypsum requirement of 20 uM for
commercial quality. Figure 4 shows a comparison of the median particle sizes as analyzed by
SediGraph versus the solids hold time in the recycle tank.
One important point to be mentioned on this topic was that the dewatering characteristics of the
gypsum solids was negatively affected during the testing periods. The particle analysis revealed
a slight increase in gypsum fines (particles less than 10 uM) during the high velocity testing
periods. These gypsum fines tend to settle out in primary dewatering thickeners and are included
in the gypsum slurry which is pumped to the horizontal belt filters for final dewatering. The
increase in gypsum fines would explain why the product moistures increased during the test
periods.
Effects on Slurry Pipe, Mist Eliminator Modules, and Outlet Ductwork
Upon conclusion of the single module testing, the mist eliminator modules and outlet ductwork
were inspected for any indications of solids deposition. As was the case during the short test
runs, the extended operation revealed no solids buildup on either the mist eliminator modules or
in the outlet ductwork. Several small areas on the mist eliminators and the outlet ductwork were
coated with a thin layer of rust colored material, but this has also occurred at other times and is
not unique to the high velocity testing. No gypsum scale was found on the mist eliminators or
the mist eliminator wash system. Additionally, the rubber lined carbon steel slurry feed pipe was
inspected, with special attention paid to elbows, and no excessive wear was noted.
Impact of Single Absorber on Plant Operations
Single module operation was desirable to OMU for two reasons; to maximize operational
flexibility and to realize potential costs savings due to operation of less equipment. Single
module operation did not achieve a substantial economic benefit due to operation of less
equipment. During normal operation either three (3) or four (4) recycle pumps per module are in
service, depending on inlet SO2 concentrations. With single module operation, five (5) recycle
pumps are required to maintain comparable removal efficiencies. The total FGD system power
consumption was reduced due to the operation of fewer recycle pumps, but FGD balance of plant
equipment operation was not significantly affected. The overall FGD plant power consumption
did not decrease due to the additional booster fan power consumption. The additional power
required by the booster fans due to the higher absorber pressure drop did offset the recycle pump
savings.
The ability to operate at full load through a single module is a significant asset for plant
operations. With this capability, any absorber preventative maintenance can be scheduled as
needed. Initially, routine absorber maintenance could only be performed concurrently with a
Page?
-------
unit outage. This severely limited the ability to perform any preventive maintenance or to
inspect the absorber inlet and outlet ductwork. Without single module capability, one of the
units had to bypass the FGD to allow a module "to be removed from service when an absorber
inspection was required. For a plant that burns a medium to high sulfur coal, this resulted in
significant additional use of S02 allowances. With a 100% spare module, absorber inspections,
cleanings, and maintenance can be scheduled as needed. Non routine maintenance and repairs
can be performed as necessary - all without bypassing the FGD and using allowances.
Conclusions
Although the direct economic benefit was not significant, the additional operational flexibility
and ability to avoid excess allowance usage is a tremendous advantage to OMU. The results of
the near problem free operation over a five (5) month period of a full scale, high velocity
absorber at OMLPs Elmer Smith Station should be a benefit to many other power producers.
Utilities with existing multiple module FGD systems may now consider operating fewer modules
at higher velocities for periods when FGD system maintenance is required. Other power
producers, which plan to scrub for compliance with Phase II requirements, may reduce FGD
system capital costs by allowing the use of this high velocity absorber design.
Acknowlegements
We would like to acknowledge Mr. Stan Conn, OMLTs Director of Power Production, for his
cooperation along with the work of Mr. Tim Goetz, OMU's Plant Chemist, and the entire staff of
OMUs laboratory which contributed to many of the analytical results which were presented in
this paper.
PageS
-------
Figure 1
Flue Gas Flow Diagram
-------
UET ELJUMftTOR WASH
UBT ELUUMKTOR WASH
, RECYCLE EPHAY
RECYCLE CPRAY
RECYCLE SM»Y
RE CYCLE CPRAY
RECYCLE CPRAY
Figure 2
Absorber Schematic
-------
SO2 Removal vs L / G Ratio
10.0 -13.5 ft/sec Absorber Velocity
98.0%
96.0%
94.0%
> 92.0%
°J £> 90.0%
if. „>
-A
&-*-
* *
•
A^ 4
«»
^
>•
I
•
0 90 100 110 120 130 140 150 16
gpm / K acfm
L/G Ratio
Figure 3
S02 Removal vs Operating L/G Ratio
-------
Median Particle Size vs Solids Retention Time
OMU. Elmer Smith Station
-------
NEW PERSPECTIVE OF WET SCRUBBER FLUID MECHANICS
IN AN ADVANCED TOWER DESIGN
W.F. Gohara
Environmental Engineering
Power Generation Group
The Babcock & Wilcox Company
T.W. Strock & W.H. Hall
Thermal Hydraulics Section
Research and Development Division
McDermott Technology, Inc.
Abstract
The need to understand the gas/liquid hydraulic interactions in a Wet Flue Gas
Desulfurization (WFGD) absorber is important with the increased demand for compact
high velocity scrubbing systems. The interactions between the gas and liquid phases in
the absorption zone of the scrubber are largely influenced by and change with the gas
velocity. This changing relationship crosses several flow regime boundaries and goes
from a predominantly falling drop, to suspended and fluidized regimes, and ultimately to
a liquid carry-up regime. Gas/liquid interactions affect the design of the scrubber internals
from the inlet to the outlet and require ultimate care in applying the design criteria
compatible with the prevailing hydraulic regime in the scrubber. Gamma ray densitometry
was used to study the liquid and gas phase interactions and evaluate the relative
differences in the liquid hold-up in the absorption zone as the gas velocity crosses the
boundaries of the different liquid regimes.
Introduction
Counterflow spray towers are the most widely used scrubbers in the electric utility
industry. Intimate and even contact between the gas and liquid phases in the scrubber
provides the optimum means for SO2 removal. A patented absorption tray distinguishes
the Babcock & Wilcox (B&W) absorber tower from the other absorbers classified in the
same category. The tray provides positive means to distribute the gas and enhance SO2
removal through intimate gas/liquid contact. The even gas flow leaving the absorption
tray passes through the spray zone where a constant flux of refreshed slurry contacts the
gas. The resistance of the spray droplets to the gas flow provides additional mass transfer
surface area and the resistance of the liquid phase helps to deliver the even gas distribution
established at the tray to the face of the first stage mist eliminator section.
-------
The design of a functional trouble free Wet Flue Gas Desulfurization (WFGD) system
scrubber incorporates proper integration of chemical, hydraulic, and selection of the proper
material of construction. Both the chemical and material selection aspects of the FGD
system were studied extensively over the past decades and are well established. The
prevailing low gas velocity in the old vintage scrubbers was forgiving to minor infraction
of gas distribution. Therefore, it did not emphasize the importance of the hydraulic aspect.
However, as we develop the new generation of intermediate and high velocity scrubbers,
the role of gas/liquid interaction becomes very important. Proper gas/liquid distribution
and the hydraulic aspects of the WFGD scrubber become significant.
Close study of the liquid and gas interaction in the scrubber showed that the relationships
between the two fluids change with gas velocity. Several liquid flow regimes were identi-
fied as the gas velocity changed from 10 to 22 feet per second. Until recently, study of the
scrubber gas distribution was limited to one-phase modeling and occasionally, improper
application of two-phase modeling. At the low end of the velocity spectrum, the liquid
phase is more of a falling drop type then changes to fluidized regime at moderate gas
velocities, and is eventually carried out and entrained as the gas velocity exceeds 20 feet
per second. At low gas velocity, less than 12 feet per second, the imprecision in one-phase
modeling was covered up by the dominance of the chemical aspects of the system design.
As the gas velocity increased and the liquid regime in the scrubber changed from a falling
drop to a fluidized regime the pitfalls of one-phase modeling became more apparent and
the accuracy of two-phase modeling became important. Results derived from improper
application of two-phase modeling may lead the designer in the wrong direction.
Over the past five years, The Babcock & Wilcox Company embarked on a comprehensive
program to study the hydraulics of the FGD system over a wide range of gas velocities and
liquid flow. The understanding of the system dynamics helped the development of more
reliable design correlations and resulted in some fundamental changes in the design
formulae of the scrubber components. Figure 1 compares the measured pressure drop of
some commercial full scale FGD systems to the predicted values derived from extensive
testing at the hydraulic test facility. Previous publications discussed the basis, the scale
down rules and proper application of two-phase modeling to WFGD scrubbers.[1-2)
The work described in this paper studied the liquid loading and the gas/liquid interaction
in the zone between the absorption tray and the face of the first stage mist eliminator in
counterflow wet flue gas desulfurization systems.
The Tray
The high liquid loading at the tray is caused by the increased liquid retention, on the tray,
caused by the gas resistance to the down flow of the liquid. This is an advantage realized
only in absorbers equipped with absorption trays. When designed properly, the tray's
superior efficiency in removing SO2 is due to the pressure drop across it and its ability to
affect even gas distribution in the scrubber and the intimate contact between the gas and
the liquid on the tray. The tray pressure drop is a composite of three basic pressure drop
-------
3 —
B 1 —
I | I | I | I
01234
Measured Scrubber Pressure Drop (Inch Water)
FIGURE 1 WET SCRUBBER FIELD UNIT PRESSURE DROP COMPARISON
mechanisms 3: the dry porous plate, surface tension, and the static head of the bubbling
mixture which is the dominant factor among the three. A preferred dynamic scaling
approach is to geometrically scale the tray perforated plate geometry and baffle spacing.
Overall the tray pressure drop correlates well with Dillman's correlation.'3'
The Spray Zone
Assuming a uniform gas and liquid flows, the spray zone may be analyzed using a one
dimensional, two phase flow, drift flux model. The drift flux model'4' of the momentum
equation describes the two phase flow processes by modeling the difference between the
droplet and gas velocities as a diffusion process. In the spray region, the gas/liquid static
head and the nozzle momentum terms are the dominant effects. Separate effects testing to
measure the spray zone liquid fraction and static head pressure drops were conducted to
validate this approach. The validation was conducted over a wide range of gas and liquid
flows and various spay header spacing to verify the mixture static head effect. Figure 2
shows the scaled baseline configuration used to determine the pressure drop relationships
of the absorber.
Liquid Concentration / Fraction Measurements
The design of medium velocity scrubbers progressed steadily over the past few years.
A major contributor to this progress is the study and understanding of the hydraulic
behavior of the scrubber's various zones. B&W used gamma ray densitometry as a tool to
correlate the change in the absorber's liquid fraction, in the absorption zone and at the face
of the mist eliminator, to changes in gas and liquid flow rates.
-------
Outlet
Mist Eliminators
Interspacial Spray
Headers No. 1,2, 3, 4
Inlet Flue
FIGURE 2 BASELINE CONFIGURATION
Gamma rays are not significantly absorbed through a path length of a dry gaseous compo-
nent. On the other hand, water absorbs gamma rays and the attenuation of the emitted rays
is proportional to the density of the mixture. This concept is used to measure the liquid
fraction suspended in the flue gas as it passes through the FGD system.
The absorber's gas/liquid contact zone, located between the tray and the face of the first
stage mist eliminator, was studied to determine the liquid hold-up at various gas velocities
and variable liquid flow rates. The contact zone was divided into two regions to study the
water loading over a wide range of gas velocities. The first region is located between the
tray baffles and the spray headers and the second region is located between the top spray
and the face of the mist eliminator.
-------
The Absorption Zone Liquid Fraction Measurements
The absorption zone of a B&W scrubber consists of two components: the absorption/gas
distribution tray, and the spray headers. Understanding and manipulation of the interac-
tion of the gas and liquid phases in this zone is an essential part in the development of
medium and high gas velocity absorbers.
Theoretical calculations show that the trajectory of a single droplet changes directly with
the gas velocity surrounding the droplet. Figure 3 shows the behavior of various liquid
droplet sizes sprayed from a 12 psi nozzle, cocurrent and counter current, into a gas
traveling at 15 feet per second. Theory indicates that the residence time of large droplets is
increased when the droplets are sprayed cocurrent to the up flowing gas.
The dynamics of a single droplet is indicative of the prevailing trends in the absorption zone.
However, the interaction between the droplets, the effect of structural members, tray loading
and unloading characteristics, the drag forces around the nozzles and other complicating
factors limit the absolute application of these theoretical predictions to an FGD system.
To study the absorption zone's liquid fraction and hold-up characteristics, the scaled
baseline model configuration shown in Figure 2 was changed to the arrangement shown in
Figure 4 to provide reasonable distance between the measured boundaries and allow
reliable intermediate liquid fraction measurements between the tray and the spray headers.
Proper performance of the absorption zone liquid fraction measurements required that the
spay headers be raised away from the tray to evaluate the change in the zone's liquid
loading with height.
80
40 —
-40
Droplet Size (
50C
75C
microns)
/down
/down
1 000 / down
150
- - 750
100
150
^
'
-16 -1
0 / down
/up
/up
0/up
0/up
Nozzle Center Line
i i
< i
i i
. - "^ -="•'"•
-i
I
i
1 1
2 -8 -4
i
Nozzle Orifice
Elevation
^
0 t
Vertical Travel, feet
FIGURES DROPLET TRAJECTORY TRAVELING IN 15 FEET/SECOND GAS VELOCITY FROM A12 PSI NOZZLE
-------
Outlet
Tray
AAAAAAAAAAAA
. Mist Eliminators
\
XX XTX XKXK X'A' X'A
Interspacial Spray
Headers No. 3 and 4
Headers No. 1 and 2
Inlet Flue
FIGURE 4 ABSORPTION ZONE LIQUID FRACTION CONFIGURATION
Figure 5 shows the absorption zone (between the tray and the spray headers) relative
liquid fraction as a function of gas velocity and height above the tray. Two important
trends emerge from this graph. The first trend shows a distinct increase in the liquid
fraction, or the liquid retained, in the absorption zone with increase in gas velocity. The
second trend shows that the highest liquid density is close to the tray and the lowest is near
the spray header.
Liquid Loading at the Face of the Mist Eliminator
To determine the liquid fraction in the zone between the spray headers and the mist elimi-
nator, the configuration of the model was modified. The spray headers were raised an
-------
2
u.
0.4
8.0 12.0
Measurement Port Location
FIGURE 5 ABSORPTION ZONE LIQUID FRACTION TRAVERSE
16.0
adequate distance above the tray baffle to isolate the tray and spray effects. The distance
between the spray headers and the mist eliminator was extended to permit accurate liquid
measurements between the sprays and the face of the mist eliminator. Figure 6 shows the
spacial arrangements used to gather the liquid fraction data in the zone located between
the top spray level and the face of the first stage mist eliminator.
To study the liquid fraction in the zone between the sprays and the face of the mist elimi-
nator over a wide range of gas velocities, the total absorber liquid flux was also varied
from 40 to 80 gpm per square foot. In order to maintain a constant nozzle pressure for
some of these tests, different header combinations were used. The 30 and 80 liquid flux
measurements include the effect of a change in the nozzle pressure, due to the decrease and
increase in the liquid flow to the headers, respectively. All the header changes were made
in the under tray headers to maintain the liquid flux in the spray zone constant.
Figure 7 shows the liquid fraction above the spray headers for three and four headers in
operation at a 15 feet per second nominal gas velocity.
The plot shows that the liquid fraction above the sprays is affected by the total number of
operating headers regardless of the headers' location relative to the tray.
Figure 8 shows the liquid fraction at some selected gas velocities and height above the top
spray header in the zone between the headers and the face of the mist eliminator. The same
trends observed in the absorption zone are also evident above the spray headers but at a
different magnitude.
-------
Tray
Outlet
Mist Eliminators
Interspacial Spray
Headers No. 3 and 4
Headers No. 1 and 2
Inlet Flue
FIGURE 6 LIQUID FRACTION CONFIGURATION ABOVE THE SPRAYS
There is a cleat increase in the liquid fraction above the spray headers with an increase in
gas velocity and a higher mixture density close to the spray header. The mixture density
in the low velocity range is distinctly different in the zone above the sprays. At a gas veloc-
ity below 12 feet per second, the sprays clearly show a low mixture density characteristic of
the falling droplets regime. Between 12 and 13 feet per second, the liquid droplets sus-
pend around the header and then fluidization takes place above 13 feet per second.
Figures 9 and 10 are plots of the liquid fraction above the sprays vs. headers' liquid flux at
15 and 16 feet per second gas velocity respectively.
The plots show that the mixture density increased steadily with an increase in liquid flux
and decreased with increase in height above the top spray header until it reaches an almost
constant value near the face of the mist eliminator.
-------
2.0
1.6 —
1.2
SB 0.8 —
0.4 —
Liquid Flux / Number of
Headers in Service
1 30/3
1 40/4
- <&• -60/4
- 80/4
—\ 1
2.0 4.0
Measurement Port Identification
6.0
8.0
FIGURE 7 LIQUID FRACTION ABOVE THE SPRAY HEADERS AS A FUNCTION OF
NUMBER OF HEADERS IN OPERATION
6
Measurement Port Identification
FIGURE 8 LIQUID FRACTION ABOVE THE SPRAY HEADERS
-------
4 —
3
2 —
0 —'
30
15 ft/second, 4 headers
Near Spray Header
Port 2
Ports
Port 4
Ports
Near the ME face
55
Liquid Flux, gpm/ft2
80
FIGURE 9 RELATIVE LIQUID FRACTION ABOVE THE TOP SPRAY HEADER AT 15 FEET PER SECOND
4 —
2 —
Legend Title
_ <&. - Port 2
- C3 - Port 3
- A. Port 4
- -^ Port 5
- X - Port 6
- •» Near ME face
--"!---•;---- *;-
40
50
60
Liquid Flux, gpm/ft2
70
80
FIGURE 10 RELATIVE LIQUID FRACTION ABOVE THE TOP SPRAY HEADER AT 16 FEET PER SECOND
-------
Figure 11 shows the change in the liquid fraction magnitude with increase in gas velocity,
at a constant liquid flux, at the face of the mist eliminator. The data shows a significant
increase in the liquid fraction with gas velocity.
The Mist Eliminator
The most common type of mist eliminator used in FGD systems is of the inertial type.
Inertial mist elimination devices have been in commercial use since the beginning of FGD
systems. However, until recently, market demand limited the scrubber's gas velocity to 10
or 12 feet per second. However, gas velocity moved into the middle range of 13 to 18 feet
per second, which is equivalent to 20 to 22 critical mist eliminator velocities, a new genera-
tion of inertial mist eliminators emerged including vertical and horizontal flow mist elimi-
nators. While B&W commercial experience includes both types of inertial mist eliminators,
studies at the hydraulic facility are limited to the vertical flow type that successfully
contained the mist from the fluidized liquid regime. The B&W scrubber design achieves
even gas distribution at the bottom of the patented tray and carries the even flow to the
face of the mist eliminator. Theoretical calculations indicate that inertial type mist elimina-
tors will not be able to contain the mist evolving from high velocity ( greater than 19 feet
per second) scrubbers. Droplets as large as inch in diameter can be carried by the gas
stream. A new type mist elimination device or a new generation of mist eliminators is
0.4
0.8
Relative Gas Velocity
FIGURE 11 LIQUID FRACTION AT THE MIST ELIMINATOR FACE AT CONSTANT LIQUID FLUX
AS A FUNCTION OF ABSORBER GAS VELOCITY
-------
10.0000
I
LU
1.0000 —
0.1000
0.8
0.9
1.0
Relative Absorber Gas Velocity
FIGURE 12 TYPICAL INERTIAL MIST ELIMINATOR PERFORMANCE AT 60 GPM/FT2
needed to contain the heavy liquid loading and overcome the effect of high gas energy.
Figure 12 shows typical performance of an inertial mist elimination device. The y- axis
represents the change in carry-over order of magnitude.
The data shows an increase in liquid droplets carry-over on low and high ends of the
velocity scale.
On the low end of the velocity spectrum, mist eliminator carry-over increases due to the
low gas inertial forces and the inclusion of the liquid droplets within the gas stream. On the
other hand, at the high end of the gas velocity spectrum, the gas inertial forces can strip, re-
entrain, and carry the liquid droplets off the mist eliminator blades.
Conclusions
The results of the liquid fraction measurements indicate that the mixture density in the
absorption zone and the zone under the first stage mist eliminator are a function of gas
velocity, as well as, the liquid flux in the absorber. In the medium gas velocity range, the
tray as well as at the region above the spray headers form high liquid density zones in
which the liquid is in a fluidized state and suspended by the gas medium. The effect of gas
velocity is reflected on the magnitude of the liquid fraction retained in the gas phase pass-
ing through the absorption zone and the area between the sprays and the mist eliminator.
Generally speaking, the mixture density increases with an increase in liquid flux and
decreases with increase in height above the top spray header until it reaches an almost
constant value near the face of the mist eliminator.
-------
References
1. T.W. Strock, P. Dykchoorn, M.J. Holmes, and W.F. Gohara, "Experimental Approach and
Techniques for the Evaluation of Wet FGD Scrubber Fluid Mechanics" presented at the
EPRI/EPA/DOE 1993 SO2 Control Symposium, Boston, MA (August 1993).
2. T, W. Strock, and W.F. Gohara, "Use of Hydraulic Models to Identify and Resolve
Design Issues in FGD systems" presented at the EPRI1995 SO2 Control Symposium,
Miami, PL (March 1995).
3. V.V. Dillman, E.P. Darovshikh, M.E. Aerov, and L.S. Akselrod, "Hydraulic Resistance of
Grid Type and Hole Type Plates," Khim Prom, No. 3, pp 156 -161 (1956).
4. G.B. Wallis, One Dimensional Two-phase Flow, McGraw-Hill, New York, 1969.
-------
Wednesday, August 27; 11:00 a.m.; 1:00 p.m.
Parallel Session A:
FGD Conversions
-------
AEC Lowman Station FGD Conversion from Limestone
to Magnesium-enhanced Lime Scrubbing
William Inkenhaus Lonnie Loper
Dravo Lime Company Alabama Electric Cooperative
3600 Neville Road Charles Lowman Station
Pittsburgh, PA 15225 Carson Road
Leroy, Alabama 36548
Manyam Babu, Kevin Smith
Dravo Lime Company
3600 Neville Road
Pittsburgh, PA 15225
Abstract
AEC's Lowman Station is located in Leroy, Alabama. Units 2 and 3, with a total of 516
MW output capacity, were switched from limestone FGD operation in January of 1996.
Prior to switching, personnel from AEC and Dravo Lime Company conducted a four
week test on magnesium-enhanced lime and obtained scrubber performance data
including SO2 removal efficiencies on the modules while burning higher sulfur coal. It
was determined that the plant could take advantage of the higher SO, removal efficiency
of the magnesium-enhanced lime system.
Major benefits resulting from this conversion were AEC's ability to switch to a lower
cost high sulfur coal while meeting the stringent SO2 emission requirements. Power cost
savings resulted from the lower liquid to gas ratio required by the magnesium-enhanced
lime process. Three recirculation pumps per module were reduced to a single operating
pump per module, lowering the scrubber pressure drop. Significant cost reduction in the
operating costs of the ball mill was realized due to modifications made to slake lime
instead of grinding limestone. This paper discusses the plant modifications that were
needed to make the switch, cost justifications, and AEC's operating experiences to date.
AEC and Dravo Lime Company working together as a team conducted detailed cost
studies that followed with extended field tests and implementing plant modifications.
This plant continues to operate in the magnesium-enhanced lime FGD mode to date.
Introduction
Alabama Electric Cooperative owns and operates the Charles R. Lowman Station located
in Leroy, Alabama. The station consists of two 258 MW and one 85 MW coal fired
-------
boilers. The two 258 MW units, units 2 and 3, were equipped with flue gas
desulfurization systems. They were originally designed and constructed by Peabody as
once through natural oxidation limestone FGD systems. Each unit's FGD system
consists of two spray tower absorbers, one common agitated recycle tank with over
370,000 gallons of volume. The system was also designed with hydroclones to separate
calcium carbonate and calcium sulfite to decrease to the amount of lost reagent in the
exiting slurry.
Designed for an SO, removal efficiency of 85%, the spray towers are 24 feet in diameter
by 90 feet tall vessels filled with six spray headers and ceramic full cone nozzles. Mist
elimination was performed by an interface tray, as a bulk separator, and a chevron mist
eliminator.1 Each absorber was designed with a false bottom to direct the slurry to the
common recycle tank. Each spray tower is provided a liquid to gas ratio, L/G, of
approximately 70 GPM per 1,000 ACFM of flue gas flow by three recycle pumps capable
of 7400 GPM each. The absorber modules were originally designed to treat 70% of the
gas flow at a velocity of 10 feet per second.
To give the station more flexibility with their coal supply the common stack was
upgraded to allow wet operation and the bypass dampers were upgraded. The absorbers
were then capable of scrubbing 100% of the gas flow at a velocity of approximately 14
feet per second. Because of the change in velocity, the wash tray was removed and
another chevron mist eliminator level was added.2 At the top of each absorber the
chevron mist eliminators are washed with reclaimed water. The recycle tank blowdown
is discharged to a 57 acre waste disposal pond. In the pond calcium sulfite and sulfate
solids settle out and the supemate is pumped back to the plant as makeup and slaking
water.
One ball mill produced enough ground limestone for both units. The ball mill was
designed to grind 10 tph limestone to 70% passing 325 mesh. This was fed to
hydroclones to classify the material and then delivered to a storage tank for use by the
FGD system.
In the conversion from limestone to magnesium-enhanced lime, the reagent feed system
needed to be modified. In the limestone configuration, the ball mill would operate 22 out
of 24 hours per day. With the existing motor and size of the mill, it was capable of slaking
25 tph of lime. Therefore the mill only needed to operate less than 5 hours per day at a
similar sulfur coal. This resulted in less operating time and provided greater maintenance
flexibility with this part of the system. Unloading and conveying of lime however needed
to be addressed. Limestone had been conveyed to the storage silo for use in the system.
Due to lime's reactive nature, pneumatic conveying of the lime to the silo was required. A
pneumatic truck, the mode of lime transportation, offloads the reagent to the silo for the
plant. Thus, a feed pipe and baghouse were installed. The top plate and cone of the silo
were replaced to accommodate pebble lime. Also a new screw and mill feed chute were
installed. The discharge chute was modified to accommodate the new mill sump and a
-------
small wet scrubber. Also new sump pumps were installed to deliver the lime slurry to the
storage tank.
Background
Limestone chemistry can be shown by the following reactions:
CaCO3(s) <-» Ca++ + CO; Dissolution
SO2(g) + H2O(aq) ** H* + HSO; Absorption
HSO) 4* H* + so;
SO; + U2O2(aq) SO; Oxidation
Ca" + SO; -> CaSO3(s) Neutralization
Ca*+ + SO; -> CaSO4(s)
CaSO, + \I2H2O -> CaSO3»\/2H2O Precipitation
CaSOt + 2H2O -> CaSO4»2H2O
With this chemistry the driving force for SO2 removal is based on the dissolution of
calcium carbonate. Therefore, more liquor needs to meet the gas stream to remove more
SO2. Limestone demands high L/Q ratio to achieve high SO2 removals.
The Thiosorbic process uses slightly different chemistry to accomplish the same task.
Due to magnesium's higher solubility, it is able to neutralize more sulfites leading to
better absorption of SO2. The reactions are shown below:
CaO + H2O -> Ca(OH)2 Slaking
MgO + H2O -»• Mg(OH)2
Ca(OH)2 (s) <-» Ca^ + 2OH~ Dissolution
Mg(OH)2(s) o Mg" + 2OH-
SO2(g) + H2O(aq) <^> H" + HSO^ Absorption
HSO; *+ H* + so;
SO; + l/2O2(aq) <-> SO; Oxidation
Mg** + SO; <-> MgSO3 Neutralization
Mg" + so; +* Mgso4
MgSOi + Ca(OH)2 <-> CaSO3 + Mg(OH)2 Regeneration
MgSO, + Ca(OH)2 <* CaSO4 + Mg(OH)2
CaSO3 + \I2H2O ->• CaSO3 •\I2H2O Precipitation
CaSO, + 2H,O -> CaSO4»2H2O
Since magnesium sulfite is very soluble in water, there is no need for high L/G ratios,
reducing the number of pumps required to remove the same, if not more, SO2.
-------
Since Lowman station is a. one-pass limestone facility, calcium sulfate scale was found to
be a maintenance problem. Sulfur emulsion was introduced to the system in the 1991.
Sulfur emulsion inhibits natural oxidation in a limestone system. Inhibiting oxidation
reduces the amount of scale in the system by reducing the amount of sulfate in solution.
This kept the absorber pressure drops from increasing as rapidly which decreased the
amount of maintenance required by the absorbers. Thiosulfate levels were typically
controlled to between 500 to 1000 ppm.
Testing
Since the ball mill ran so frequently, Lowman Station was concerned about a forced
outage if the ball mill became inoperable for an extended period of time. A test of Unit 2
using hydrated lime as a reagent was conducted to determine if an alternative to limestone
could be utilized without drastically affecting their system. Lowman Station was
persuaded by Dravo Lime Company to compare lime to limestone as a primary reagent in
the FGD system while burning 1.8% sulfur coal for two days. Since hydrated lime
performed well, lime containing magnesium was tested at the plant. Since Lowman is an
open loop system, MgO levels of 8-12 weight percent were used in the lime feed. Dravo
provided a 25 tph portable slaker and operators to slake the lime for the testing of Unit 2.
Stack sampling was carried out while varying the number of recycle pumps in service.
Bypass dampers were positioned for the absorbers to treat as much flue gas as possible.
The results of the Thiosorbic testing are compared to normal limestone operation.
Results & Discussion
EPA Flue Gas Sampling Methods 5 and 6 were implemented for simultaneous sampling
of the flue gas inlet and outlet of the absorber modules. The first week of sampling was
performed while in normal limestone operation and three pumps per module were in
service, unfortunately the results from these tests were unable to be used because of
unacceptable confidence levels. Flue gas sampling was completed while Lowman Station
was in Thiosorbic operation. Simultaneous testing of the absorber inlet and outlet was
scheduled so that one pump per module would be taken out of service each day of testing.
Therefore the removals across the absorbers could be measured. For three pumps per
module (-70 L/G), the removals were found to be greater than 99.7%. With two pumps
per module (-47 L/G) in service the removals were around 99.2%. The last test of
running one recycle pump per module (-23 L/G) resulted in a removal of 96.7%. When
the removals are plotted as number of transfer units, NTU, the number of transfer units
increases linearly with respect to the number of pumps in service. (Figure 1) The
definition of NTU is as follows:
-------
Looking at the difference between Unit 1 and Unit 2 emissions, the removal for the FGD
system can be estimated. The overall removals for the testing days were found to be
90%, 89%, and 88% respectively. The resulting difference between scrubber removals
and emissions is attributed to bypass leakage. This leakage is between 9% and 10%,
depending on the pressure drop across the module. Using the same analysis as above, the
removal across the scrubber for lime hydrate was 95%.
When operating on limestone, the FGD system typically operates all three recycle pumps
per module, approximately 15% solids in the recycle tank, a recycle tank pH of 5.8, and
sulfur emulsion is added to maintain thiosulfate levels between 500 to 1000 ppm. The
FGD system removals were 80% with 3 pumps in service and 70% with 2 pumps in
service. Illustrated in Figure 2 is the effect of L/G on limestone and lime systems.
However, Thiosorbic scrubbing does give Lowman Station greater flexibility with their
coal feed. The Thiosorbic process allows this station to increase to a 3% sulfur coal and
still stay in compliance while limestone does not. (Figure 3) Thiosorbic scrubbing could
be accomplished with only one pump per module decreasing operation and maintenance
costs of the scrubber and ID fan horsepower requirements.
The lime used at Lowman Station was a blend of high calcium and dolomitic limes to
increase the amount of magnesium oxide in the lime. More MgO contained in lime is to
maintain reasonable alkalinity because the FGD system remains open loop. SO,
scrubbing was successful, as is represented in Figure 4, even though pH of the recycle
tank varied during the testing. A low pH can effect the alkalinity of the Thiosorbic
system as illustrated in Figure 5. Normal Thiosorbic operation sets the pH between 6.0
and 7.0, while limestone operation controls pH between 5.5 to 6.5. Even when the pH is
at 6.2 to 6.3 in limestone operation, the alkalinity of Thiosorbic operation is six times that
of limestone.
The stoichiometry for the Thiosorbic system was typically measured between 1.01 and
1.03. This compares to a stoichiometry of 1.08 normally seen in limestone operation.
The liquor chemistry of the FGD pond did not vary significantly during the testing
period. (Table 1)
Economic Justification
An economic analysis of the lime and limestone FGD operation reveals a large savings in
overall plant operating costs at Lowman Station by switching to lower cost, high sulfur
coal. Scrubbing a cheaper, higher sulfur coal can be the most cost affective route to lower
operating costs.2 Alabama Electric Cooperative and Dravo Lime Company independently
conducted economic comparisons of magnesium-enhanced lime and limestone FGD
operation and arrived at similar conclusions, that AEC could save up to 2.3 million dollars
a year in plant operating costs by changing from limestone to lime as their primary
reagent. The main component of these savings is switching from a 1.3% to a lower cost
3% sulfur coal without surpassing their emission limit.
-------
Parasitic load dropped from 0.75% to 0.20% of installed capacity as illustrated in Table 2.
This results in cost savings of 750 thousand dollars. Using a blend of high and low sulfur
coals, coal costs at Lowman station decreased by 3.1 million dollars. The savings in coal
and parasitic load costs more than offset the increased FGD variable operating cost of 1.7
million dollars a year, due to the higher cost of reagent lime. (Table 3) Taking all
operating costs into account, Lowman station saves about 1.4 million dollars a year by
changing the FGD system from limestone to the Thiosorbic process.
Summary
Lowman Station converted to lime as their primary reagent for the FGD system on
January 9, 1996. To keep Lowman Station from an outage to make the modifications to
the ball mill, the 25 tph portable slaker was operated to keep both units on-line. This kept
the plant operational while the ball mill was down for four months of retrofit to a lime
slaking ball mill. By improving the efficiency of the FGD system, Lowman changed from
a blend of 1.3% and 1.8% sulfur coals to a blend of 1.8% and 2.9% sulfur coals in the first
six months of Thiosorbic operation. Lowman did not experience any problems during this
time with the scrubber performance. The ball mill slaker runs approximately 10 hours per
day to keep the system supplied with lime slurry. The conversion of Lowman Station
FGD to the Thiosorbic process demonstrates that utilities gain the ability to decrease
operational costs with modest capital expenditures. With a minimal capital expenditure to
convert the ball mill, Lowman Station is capable of saving up to 2.3 million dollars per
year in operating costs.
Conclusions
The Thiosorbic system achieved about 10% higher SO2 removal at a lower L/G, without
the use of sulfur emulsion and at lower calcium to sulfur stoichiometry than that of the
limestone system at Lowman Station. The settling pond showed no signs of being
affected by the Thiosorbic chemistry during the testing period. With these advantages and
the ability to scrub higher sulfur coal, the higher cost of reagent lime is offset many times
by the savings of purchasing lower cost higher sulfur coal.
References
1. The Mcllvaine Scrubber Manual. 1980. vol. IV, p. 138.67. [report]
2. Steve Feeney, "Substitution: An FGD Vision Reaches Fruition," presented at the EPRI
SO2 Control Symposium, Miami, FL (March 1995). [conference paper]
-------
5.5
5.0
NTU
•
^^-
i
y = 0.0549X + 2.249
R2 = 0.9609
NTU
.Linear (NTU) |
40 50
UG(GPM/1000SCFM)
Figure 1
NTU for AEC Lowman Station in Thiosorbic Operation
100%
90%
SO2
Removal
Across
Scrubber
70%
60%
20 30 40 50 60
L/G(GPM/1000SCFM)
^ Testing Data Best Fit Limestone (estimated) i
70
Figure 2
Effect of L/G on SO2 Removal for Thiosorbic and Limestone Systems (1.3% S Coal)
-------
% S in Coal
Figure 3
AEC Lowman Station SO2 Emissions
Compliance Limit
11 1 Bypass
^^ Unit#1 (85 MW)
, Mg-Lime (47 L/C) '
, Limestone (70 L/G)
SO2
Emmission
(Ib/MMBTU)035
4/18/95 4/22/95 4/26/95 4/30/95 5/4/95 5/8/95 5/12/95 5/16/95 5/20/95
Date
Figure 4
Unit 2 Stack Emissions for Testing Period
-------
Alkalinity 2°°° l
CaC03
|^ Mg-Lime Alkalinity B Limestone Alkalinity
Figure 5
Relationship of pH and Alkalinity for the Thiosorbic Process
Table 1
FGD Pond Chemistry
Cl- (ppm)
SO4 (ppm)
Ca2+ (ppm)
M§2+ (PPm)
200-500
1000-1800
300-600
100-200
-------
Table 2
Parasitic Load
Number of Operating Pumps
Absorber Pressure Drop
Mill Power Consumption (kWh)
Pump Power Consumption (kWh)
ID Fan Power Consumption (kWh)
Total Parasitic Load (kWh)
Total Parasitic Load (%)
FGD Electricity Costs (MM$/year)
Limestone
3
5
178
2533
1144
3855
0.75
1.0
Thiosorbic
1
1.4
77
653
297
1027
0.20
0.25
Differential
101
1880
847
2828
0.55
0.75
Table 3
Economic Justification
Limestone
Sulfur content in coal (%)
Coal Costs (MM$/year)
FGD Variable Operating Costs
Reagent (MM$/year)
Electricity (MM$/year)
Water (MMS/year)
Total
1.3
47.7
0.4
1.0
0.1
49.3
Thiosorbic
2.4
44.6
2.9
0.25
0.1
47.9
Cost Differential
3.1
2.4
0.75
1.4
-------
CONVERSION OF THE 1600 MW MILL CREEK GENERATING STATION
TO PRODUCTION OF COMMERCIAL-GRADE GYPSUM
R.S. Straight, PE
S.D. Can-
Louisville Gas & Electric Company
J.D. Colley
T. Oolman, Ph.D.
M.L. Meadows, PE
Radian International LLC
Abstract
The four generating units at Louisville Gas & Electric Company's (LG&E) Mill Creek Station
are equipped with limestone-based FGD systems using first-generation spray towers. These
systems are being converted from inhibited-oxidation to forced-oxidation processes. The
inhibited-oxidation system produces calcium sulfite solids, which are fixated with lime and fly
ash before disposal in an onsite landfill. The existing landfill is approaching the end of its
storage capacity. The forced-oxidation system will produce gypsum solids, which will be used as
a raw material for the production of wallboard and other building materials. This paper discusses
the design of the proposed conversion, results of a pilot demonstration, LG&E economic
evaluation, and the proposed implementation schedule.
Introduction
The LG&E Mill Creek Generating Station is a 1600 MW (gross) facility located on the Ohio
River near Louisville, Kentucky. The station consists of four generating units. A summary of the
characteristics of the four FGD systems is presented in Table 1. All four units fire a variety of
medium-sulfur Kentucky and Midwestern coals. The average fuel characteristics produce 5.7
pounds of sulfur dioxide (SCh) per million Btu (Ib/MMBtu).
All four units are equipped with electrostatic precipitators followed by flue gas desulfurization
(FGD) systems that control SC>2 emissions to less than 1.2 Ib/MMBtu on a 3-hour rolling
average. Units 1 and 2 were originally equipped with FGD systems supplied by Combustion
Engineering (now ABB Environmental Systems); Units 3 & 4 with systems supplied by
07/22/97
-------
American Air Filter. LG&E extensively modified and upgraded all four FGD systems in the late
1980's.
Table 1
Generating Unit Characteristics
Parameter Unitl Unit 2 Unit 3 Unit 4
Gross generation, MW
Gross heat rate, Btu/kWh
No. of absorber modules
No. of reaction tanks
330
9,261
2
1
330
9,585
2
1
420
9,707
4
2
520
9,744
4
2
The FGD systems currently operate using the limestone-based, inhibited-oxidation process. The
limestone reagent being utilized is approximately 92% CaCOs, 3.2% MgCOs, and 3.5% SiO2 and
is supplied as a pre-ground slurry from a local cement plant. The slurry is stored onsite in two 1.8
million gallon limestone slurry storage tanks. Emulsified sulfur is added to the limestone reagent
storage tanks and sulfite oxidation is held to below 15%. The byproduct solids are dewatered by
gravity thickeners and rotary drum vacuum filters. The 52% solids filter cake is fixated by
mixing with fly ash and lime kiln dust and placed in an on-site landfill.
The objective of the current project is to convert all four FGD systems from the inhibited-
oxidation process to a limestone-based, forced-oxidation process that will produce a commercial-
grade of gypsum. There are several factors that make conversion of the existing FGD operation
to the production of commercial-grade gypsum both desirable and feasible. Although the current
inhibited-oxidation process is performing well, the production of commercial-grade gypsum
offers the following cost reductions:
• Elimination of emulsified sulfur feed:
• Elimination of byproduct fixation practices;
• Elimination of on-site landfill operations;
• Elimination of the need for future expansion of the byproduct solids landfill; and
• Increased sales of fly ash currently used to fixate the inhibited-oxidation process's byproduct
solids.
The elimination of the need for landfill expansion is a primary driver for the project, since the
station's existing landfill has limited remaining storage volume. Under current operation, the
development of a new landfill will be required in the near future. Conversion to the forced-
oxidation process saves significant capital cost for landfill construction, eliminates the need for
licensing a new landfill site, and improves relations with the community in the vicinity of the
plant.
The conversion of an existing waste material (the fixated byproduct solids) into commercial
products (gypsum and additional fly ash) also supports the environmental goals of LG&E. At
2 07/22/97
-------
design conditions, this conversion project will eliminate 950,000 tons per year (tpy) of solid
waste and free-up an additional 260,000 tpy of fly ash for off-site sales. Over a 10-year period,
this conversion will eliminate the need for over 7 million cubic yards of landfill.
LG&E has contracted with U.S. Gypsum to provide the gypsum solids produced by the converted
process. Conversion of the Mill Creek Station to the forced-oxidation process will generate a
consistent supply of more than 500,000 tons per year of wall-board-quality gypsum. Mill Creek
Station's location on the Ohio River provides the added benefit of barge shipment of gypsum to
the wall-board production facility.
Radian International LLC has operated as a consultant to the electric utility industry on FGD
systems for over 25 years. The conversion of the Mill Creek Station to the forced-oxidation FGD
process is similar to other projects in which Radian has been involved, most recently at the
Northern Indiana Public Service Company (NIPSCO) R.M. Schahfer Station (see companion
paper in this session). Like the Schahfer conversion project, Radian is taking a much more active
role in the project than simply as a process consultant. Radian is entering into an alliance
agreement with LG&E in which Radian's fee will be directly tied to the improvement in the
utilities cost for FGD and the percentage of gypsum that meets the wallboard specification.
The goals of this paper are to:
• Present the conceptual design of the conversion project;
• Present data from the full-scale pilot test;
• Outline the economic justification of the project; and
• Summarize the conversion schedule.
Conceptual Design
The objective of this project is to convert all four FGD systems at the Mill Creek Station from
the current inhibited-oxidation process to the proposed forced-oxidation process with a minimum
of new equipment. A process schematics for the existing system is presented in Figure 1.
Equipment that is to be abandoned is shaded. A process schematic for the converted system is
presented in Figure 2. New equipment is designated by a heavy outline.
The following existing equipment will be utilized with minimal modifications: the reagent
(limestone) feed system, scrubber modules, reaction tanks, slurry recycle pumps, mist
eliminators, flue gas reheaters, vacuum filters, solids conveyors/stackers, and return-water ponds.
Equipment to be abandoned include thickeners and pug-mill mixers.
07/22/97
-------
REACTION TANK
(TYP OF 6}
RETURN WATER
PUMP
•v
= -,;u
— £S
THICKENER
(TYPOF4)
1 ^
UNDERFLOW
SURGE TANK
_J
VACUUM
FILTER
(TYPOFS) CONVEYORS
o
-J
fcj
VO
Figure 1: Inhibited Oxidation System Schematic
-------
—sr
--&- ;c
Figure 2: Forced Oxidation System Schematic
-------
The process changes can be divided into the following areas:
• Addition of oxidation air blowers and spargers;
• Addition of a dibasic acid (DBA) feed system;
• Modifications to the existing primary dewatering system;
• Modifications to the existing secondary dewatering system;
• Installation of a new overload conveyor and barge loading system; and
• Construction of a new FGD laboratory building.
These changes are discussed below in more detail.
Addition of Oxidation Air Blowers and Spargers
Air will be sparge into each of the reaction tanks in order to provide the oxygen necessary for
conversion of sulfite to sulfate. To accomplish this, air spargers will be installed in each reaction
tanks and oxidation air compressors will be purchased.
Because the FGD systems are of three different designs, their reaction tank designs are quite
different. Units 1 and 2 each has a single, relatively deep reaction tank (76-ft diameter by 38-ft
deep). Unit 3 has a unique arrangement of two L-shaped, below-grade, concrete tanks, each made
up of four compartments. Each compartment is approximately 25-ft by 25-ft by 23-ft deep. Unit 4
has two, relatively shallow reaction tanks (63-ft diameter by 18-ft high). Because of these
physical differences and the generating units size differences, the requirements of each units'
oxidation air systems are significantly different.
Three separate oxidation air compressor systems will be provided: a common system serving
Units 1 and 2, one serving Unit 3, and one serving Unit 4. Each system will consist of two
oxidation air blowers. Each system will consist of two 100%-capacity blowers, each sized to
provide the combined oxidation air requirements for two reaction tanks at design conditions.
Air requirements for each reaction tanks were determined on the basis of a general mass-transfer
correlation that was calibrated to the results from the pilot demonstration and past project
experience.
Addition of a DBA Feed System
To meet the commercial-grade gypsum composition requirements, limestone reagent utilization
must be greater than 95%. Since limestone utilization on Units 3 and 4 in the existing inhibited-
oxidation system is marginal, the limestone utilization in these absorbers will be improved
through the addition of DBA. DBA chemically enhances the SOj removal process under higher
reagent utilization conditions by buffering the absorber slurry in the correct pH range. The
conceptual design includes a common DBA storage and feed system serving Units 3 and 4.
07/22/97
-------
Modifications to the Existing Primary Dewatering System
Hydrocyclones will replace the thickeners as primary dewatering systems on each of the four
units. Hydrocyclones are much smaller equipment, are easier to operate and maintain, and allow
for better segregation of the water and chemical balances on individual units.
Each hydrocyclone cluster will service one reaction tank (two scrubber modules). The overflow
from the hydrocyclones will be returned directly to the reaction tank from which the slurry was
received. Underflow from the hydrocyclones, which will be at approximately 50% solids, will
be discharge to the underflow transfer tanks. Two hydrocyclone clusters will discharge to each of
three underflow transfer tanks. Each tanks will be mechanically agitated to maintain solids
suspension.
Consideration was given to reutilizing the existing thickeners for primary dewatering in the
forced-oxidation process. It was determined that any one of the existing four thickeners could
provide sufficient settling area for the oxidized solids from all four units, and a second thickener
could be reused as a 100% spare. However, the higher density of the gypsum sludge would have
required replacement of the thickener rake mechanisms in both thickeners to handle much higher
torque. New thickener underflow pumps and piping to the underflow storage tank would also
have been required.
Modifications to the Existing Secondary Dewatering System
The five existing rotary-drum vacuum filters will be reutilized, with some minor modifications,
for gypsum dewatering. Due to the higher molecular weight of gypsum (CaSO4-2H2O) relative
to the presently produced sulfite solids (CaSO3-I/2 H2O), the mass of dry solids requiring
dewatering will increase approximately 33%. However, the higher concentration of the slurry
produced in primary dewatering and the superior dewatering characteristics of the gypsum solids
will increase the capacity of the filters by more than the increase in loading rate. New filter cake
wash equipment will be installed on each vacuum filter to wash soluble chlorides and DBA from
the cake, as required to meet the wall-board gypsum spec. The filter cake is expected to have a
solids content of greater than 87wt%.
Existing equipment associated with the filter cake fixation process will no longer be required.
The two existing pug mill mixers will be replaced with short sections of belt conveyor. The
existing screw conveyors for transferring fly ash and lime kiln dust from their respective storage
silos to the pug mills will be removed.
Installation of a New Overland Conveyor and Barge Loading System
A new barge loading system will be located on the Ohio River. The system will consist of a
reclaim hopper, 300 tph transfer conveyor, barge loading shuttle conveyor, and a barge
positioning system. An overland conveyor will be used to transport the dewatered gypsum to the
barge loading area. Upgrading of an existing road to transport gypsum between the stackout area
and the barge loading facility by truck was evaluated and rejected.
07/22/97
-------
A temporary gypsum storage area will be prepared for stockpiling of gypsum during short-term
disruptions in barge transportation operations and for providing gypsum supply during planned
and unplanned plant outages. The stockpile would be approximately 8 acres and provide 40 days
of storage.
Construction of a New FGD Laboratory Building
The existing chemical laboratory was determined to be inadequate for the expanded quality
assurance/quality control (QA/QC) test program imposed by production of commercial-grade
gypsum. A new laboratory building will be constructed to provide laboratory space for
performing tests required by the FGD system and several other plant systems. This building will
also include an office, break room, and rest room facilities.
Pilot Demonstration
A full-scale pilot demonstration of the forced oxidation process, co-funded by LG&E, Radian,
EPRI, and U.S. Gypsum, was conducted at Mill Creek Station. The stated objectives of this
demonstration were to show the viability of the forced-oxidation process, determine critical
design data, and produce a sufficient quantity of commercial-grade gypsum for a wallboard
production test.
The pilot demonstration was conducted on the Unit 4 reaction tank 4B, since this was considered
to be the severest test of the process. Unit 4 has relatively shallow reaction tanks (-16 feet of
water depth) and full oxidation at a reasonable O to SOj ratio might have been difficult to
achieve. Unit 4 also has the most difficulty in meeting SC>2 emission limits; thus, the DBA
requirement was expected to be the greatest among the four FGD systems.
The test equipment consisted of the following:
• Two, skid-mounted rotary air compressors rated at 9,000 inlet scfm and 0.65 psig;
• A T-shaped air sparger with supply air piping between the compressors and the sparger;
• A 4-in. hydrocyclone for dewatering a sidestream of absorber bleed slurry;
• A skid-mounted horizontal vacuum filter with a cloth area of approximately 10 ft2.
• A temporary stacker conveyor; and
• A 8,000-gal. tanker for DBA storage with a skid-mounted DBA feed control system.
A chemical test program was developed for the pilot test. This program consisted of monitoring
the following parameters:
• SC>2 removal;
• Sulfite in the absorber slurry;
• DBA concentration in the absorber slurry and washed filter cake;
• pH;
07/22/97
-------
• Gypsum quality in the absorber slurry and the filter cake as measured by:
- Sulfites/sulfates,
— Carbonates,
— Chlorides,
— Acid-insoluble inerts,
- Silica (SiO2), and
- Cake moisture.
The pilot plant was operated for 5-Vz weeks between November 25 and December 31, 1996.
During this time, the pilot plant experienced very cold weather operation, the temporary loss of
service water pressure, higher than desired fly ash loading due to ESP problems, horizontal
vacuum filter cloth blinding, and two catastrophic oxidation air sparger failures. In spite of these
challenges, the pilot plant met the following primary test objectives:
• The Unit 4 FGD system remained in compliance with its SO2 emission requirements during
the test period.
• Internal scaling on the absorber module surfaces, mist eliminator blades and reheater tubes
was the same or less than experienced in the two absorber modules operating in the inhibited-
oxidation mode.
• The relationships between DBA concentration, reagent utilization, and pH were established.
Reagent utilization above 95% is attainable at a pH of 5.0 and DBA concentration of 500
ppm. This DBA concentration was approximately one-half of what had been anticipated.
• Complete oxidation was achieved at an O to SC>2 ratio of 4.0 Ib-mole O per Ib-mole SC>2,
which was less than anticipated for this shallow tank.
• Approximately 160 tons of commercial-grade gypsum were produced and used in a
wallboard production test, demonstrating that the gypsum produced during the pilot test met
the wallboard manufacturer's physical and chemical requirements.
In addition to these primary results, the following conclusions were also drawn from the pilot
demonstration:
• Hydrocyclones were shown to be the preferred alternative for primary dewatering. The initial
conceptual design was based on reusing the existing gravity thickeners for primary
dewatering. However, a hydrocyclone was used as the primary dewatering device during the
pilot test, because it could be easily installed and greatly simplified the operation of the pilot
plant. As a result of the very good performance both in primary dewatering and in improving
the quality of the gypsum product, hydrocyclones have been incorporated into the conversion
design and the existing thickeners will be abandoned.
07/22/97
-------
It was demonstrated that the gypsum stackout area requires no weather protections. There
was some concern prior to the demonstration that precipitation could raise moisture content
in the dewatered filter cake to above wallboard standards. Although the accumulated gypsum
pile was exposed to several rain showers during the pilot test, the surface of the pile
developed a thin crust that shed rain water and prevented moisture from seeping into the pile.
As a result of this experience, no enclosure for the gypsum pile is included in the process.
design.
Economic Evaluation
Based on the above conceptual design scope of supply, a capital cost estimate was produced. The
capital and installation labor costs were estimated based on preliminary vendor quotes, Radian
experience on previous capital projects, and published reference unit costs (e.g. Means
Construction Cost Index and Richardson's Process Plant Estimating Standards).
A capital cost of $25.4 million was estimated. A significant portion of the costs reflect capital
expenditures that would be incurred at some point in the future even if the conversion project
was not implemented (slurry pipeline/pump replacement and tank relinings). The capital cost
components included the following:
• Material & Labor for:
- Oxidation Air System,
- DBA System,
- Primary Dewatering System,
- Secondary Dewatering System,
- Gypsum Handling System,
- Fly Ash Rotary Unloader, and
- QA/QC Laboratory;
• Process & Preliminary Design;
• Detailed Design Engineering;
• Construction Oversight & Management; and
• Commissioning/Startup/Initial Operation.
The estimates of annual O&M costs were calculated based on the LG&E projected annual
generation load factor for each unit over the evaluation period and the following cost categories:
• Avoided oxidation inhibition sulfur addition;
• Avoided solids fixation lime kiln dust costs;
• Avoided landfill costs;
• Reduced limestone reagent usage (due to higher utilization);
• Increased auxiliary power usage;
• Increased fly ash sales revenue;
• New DBA chemical cost; and
10 07/22/97
-------
• New gypsum on-site handling costs.
The existing Mill Creek Station generating units experience periodic forced outages as the result
of scale formation on the FGD systems' mist eliminator blades and reheater tube surfaces. These
scaling problems should be reduced by conversion to the forced-oxidation process, because of the
higher reagent utilization attainable. The inclusion of cost reductions from potential reductions in
generating unit forced outages was considered; however, this potential benefit was not used in
the cost analysis due to LG&E's desire to model the financial viability of this conversion on
conservative, defensible cost savings.
LG&E has an established, proprietary procedure for determining the cost effectiveness of
proposed projects. This procedure considered the initial capital cost and the predicted net savings
in annual O&M costs. The costs of the proposed conversion were contrasted with the costs of
expanding the existing fixated byproduct landfill. LG&E also performed sensitivity analyses
allowing for variations from the predicted capital costs, coal quality, station load model, and the
landfill operation and expansion costs. Even with the most conservative set of assumptions, the
economic analysis demonstrated a clear economic advantage of the process conversion over
expansion of the landfill. The project clearly met LG&E's return-on-investment goals.
Implementation Schedule
LG&E has approved the capital project for the conversion of their FGD systems to forced
oxidation. Detailed engineering design will commence in August '97. Construction will occur
in phases over a 12-month period, beginning in the fall of '97. Commissioning and startup of the
new process should be initiated late in the fourth quarter of '98. Full gypsum production will
begin the first quarter of 1999.
11 07/22/97
-------
SAN JUAN GENERATING STATION LIMESTONE FGD CONVERSION
H.S. Taylor
Public Service Company of New Mexico
San Juan Generating Station
W. Nischt
The Babcock & Wilcox Company
Abstract
The San Juan Generating Station (SJGS) limestone conversion project is unusual in several
aspects. First it is justified on an economic basis rather than regulatory requirements.
Second, it is a retrofit project utilizing as much existing equipment from the old FGD
system as possible. Third, it was designed and is being constructed under a teaming ar-
rangement rather than the traditional approach where a specification is developed and
competitive bids are solicited. The original Wellman-Lord (W-L) regenerative system was
installed in the late 70s. It was an economically competitive system in the beginning but
has become uncompetitive and costly due to age and outdated technology. The conversion
project is budgeted for 80 million dollars and it is expected to save about 20 million dollars
a year in operating and maintenance cost. The conversion was justified in cooperation with
Babcock & Wilcox (B&W) by reducing the project cost through collaboration on engineer-
ing, staffing and construction issues. The teaming arrangement gives both Public Service
Company of New Mexico (PNM) and B&W an incentive to be innovative in their approach
to the project. The project has been flexible throughout the design stage allowing both
organizations to contribute with ideas to improve operability and maintainability as well
as reduce cost. Cooperation is continuing in the construction phase and further cost reduc-
tions are anticipated.
Background
SJGS is an 1800 MW coal fired plant located in the four corners area of New Mexico. This is
the northwest corner of New Mexico and the plant is located close to large deposits of
subbituminous coal •with low sulfur content, 0.9%, and high ash, over 24%. Sulfur dioxide
-------
emissions are controlled with a Wellman-Lord (W-L) regenerative system and participates
are removed by hot side electrostatic precipitators. SJGS must meet New Mexico air emis-
sion regulations which have been more stringent than federal regulations. Cooling water
is supplied from the San Juan River to cooling towers. SJGS has been a "Zero water dis-
charge" location since 1983. This means no water discharge is allowed from the plant site.
All waste water must be recycled and the final brine must be stored in evaporation ponds.
The station is owned by nine different entities including investor owned utilities, munici-
pal utilities and cooperative organizations. Owners are located in New Mexico, Arizona,
Utah, Colorado and California. Power from the generating station is transmitted to all of
those states. Public Service Company of New Mexico (PNM) is the operating agent.
There are approximately 550 full time employees at the generating station. Two open pit
mines operated by Broken Hill Properties (BHP) supply coal to the station. One mine is
located next to the plant while the second is located about 20 miles away. All coal is trans-
ported by truck to the plant.
Introduction
In 1994 SJGS owners became concerned over the comparatively high operating and main-
tenance costs for the flue gas desulfurization (FGD) system. The existing system is a regen-
erable W-L system which is about 20 years old. The W-L system was originally chosen in
the mid seventies because it was competitive in operating and maintenance (O&M) costs
and because there was much concern about disposal of sulfur waste as a hazardous waste.
Early lime and limestone systems were having many difficulties in those early years. In
the past twenty years, the W-L system has aged and now requires high maintenance costs
while limestone systems have become reliable and efficient. In early 1995 SJGS contracted
a comparison study of costs for the W-L system with the most common modern systems.
The Electric Power Research Institute (ERPI) FGD Cost Model was used to make the cost
comparisons. Results of this comparison study confirmed the high cost of the W-L system
and indicated that for the plant location, a limestone forced oxidation system (LSFO) was
the best economic choice.
Babcock & Wilcox (B&W) developed the possibility for converting the existing W-L system
into a LSFO system. The EPRI Cost Model estimated the cost of a LSFO system for SJGS at
$110 million. Through the collaborative efforts of PNM and B&W, the budgeted cost of
the conversion was reduced to $80 million dollars. The major items which allowed this to
be accomplished were refining the design, utilization of existing equipment at the site,
utilization of plant personnel and facilities, value engineering and risk sharing. The LSFO
FGD system presented an opportunity for drastic reductions in O&M costs and energy
costs. Economic justification based only on the FGD system, however, was not satisfactory
SJGS has an extremely complicated waste water recovery system in order to meet zero
discharge requirements with sodium waste from the W-L process. A conversion to LSFO
could greatly simplify the waste water treatment system by using cooling tower and boiler
blow down directly. The project was justified and approved in 1996 on the basis of savings
-------
in the FGD system and on the savings from waste water treatment. Originally, additional
justification was thought possible in excess emission allowances but fluctuations in the
market have made income from this source doubtful. This conversion project is seen now
as an $80 million project which will provide a savings in O&M cost and fuel cost of $20
million a year.
The project is seen as unique in three ways. First it is justified on the basis of economics
with no regulatory requirements. The new system will, however, be more efficient and
remove more sulfur dioxide that the old system. Second it is a true retrofit project. Much
of the existing equipment will be converted and the station will remain in compliance with
the air emission regulations while it is being converted. A third unique thing about this
project is that it was designed and is being constructed under a teaming or partnership
arrangement between PNM and B&W. This simplifies organization, reduces cost and
involves the operator more deeply in the project.
Design
A flow schematic of the W-L FGD system is shown in Figure 1. The existing W-L system
consists of four absorbers per unit, one of which is a redundant spare. The W-L FGD sys-
tem operates with two liquid systems. There is a prescrubbing system to cool the gas,
remove residual flyash and remove chlorides. This liquid is very low pH and must be kept
separate from the absorbing solution. It also requires special waste water treatment. The
limestone system has only one circulating loop and the majority of water is recycled back
to the system.
The regeneration system of the W-L process is one of the main sources of the system's high
operating cost. The regeneration system is basically evaporators which use steam heat to
drive off the sulfur dioxide and convert it into sulfuric acid. SJGS has four trains of two
stage evaporators. These evaporator trains consume about 80,000 to 100,000 Ib/hr of low
quality steam desuperheated from the generating units. Elimination of the regeneration
system will save steam and allow increased generation from some of the units. The regen-
eration system also must purge out sulfates which are formed by oxidation in the absorbers
and cannot be regenerated. This purge system is complicated and also requires energy.
The acid plant has been relatively easy to operate but the market for sulfuric acid has
declined and very little cost is recovered. Elimination of the W-L system will eliminate
several hazardous chemicals such as sulfuric acid and ammonia from the plant site. This is
seen as one of the non-economic justifications for the project.
Retirement of the W-L system will allow simplification and reduction of the waste water
treatment systems. The current waste water treatment system consists of four brine concen-
trators or vapor recompression evaporators, a four bank reverse osmosis (RO) system, an
oxidation system and 105 acres of evaporation ponds. After the LSFO conversion, SJGS
will still require the use of two evaporators, however two evaporators can be eliminated as
well as the RO units and the oxidation system. The LSFO will also extend the life of the
existing evaporation ponds and may eliminate the need for additional ponds in the future.
-------
Five control rooms will be consolidated into one with the LSFO conversion. Control for the
north and south side waste water treatment systems will be performed from the new LSFO
FGD control room. The chemical plant control room and the two control rooms for the W-
L absorbers will be eliminated. A digital control system (DCS) will be used to control the
FGD system and the remaining waste water treatment systems. The control room consoli-
dation allows a considerable reduction in manpower.
A flow schematic of the LSFO FGD system is shown in Figure 2. Three of the four W-L
absorber vessels will be converted to LSFO. Each LSFO absorber will have the capacity to
handle 33% of the design gas flow. Each absorber is designed to remove 90% of the incom-
ing SO.,. The overall system for each unit is designed to achieve 75% SO, removal as some
by-pass is required for stack gas reheat.
The absorber vessels were originally constructed of concrete with a tile and acid brick inner
lining. Over time, some of the tile has fallen off and been replaced with various lining
systems. For the retrofit, the existing absorber internals will be removed and replaced with
B&W designed internals including the patented B&W absorber tray. In addition, two levels
of absorber sprays and two levels of mist eliminators along with the associated wash
systems will be added. Sides views of the W-L absorber and the B&W LSFO absorber are
shown in Figure 3 & 4 respectively.
The W-L absorber inlets were designed with inlet quenchers which have variable Venturis.
For the retrofit, the Venturis will be replaced with straight sections of ductwork on Ul and
U2. The venturi on U3 and U4 is less abrupt, therefore replacement will not be required.
On all units, the variable plunger will be removed which dramatically decreases the system
pressure drop. Physical model testing was performed to determine the proper configura-
tion to achieve even gas flow distribution through the absorber vessel. Two turning vanes,
an absorber inlet awning and the absorber tray yielded the desired level of gas flow
distribution with a minimum pressure drop.
Each existing W-L absorber vessel is equipped with a 5000 horsepower booster fan. Due
to the significantly lower pressure drop of the LSFO system, fan horsepower can be re-
duced to less than 3000 horsepower. On two of the units, new lower rpm fan motors will
be installed. The other two units will have new fan rotors with the same motors. The
energy savings from these fan modifications yield a substantial portion of the project
energy savings.
The W-L piping systems as well as the pumps will be removed. New absorber
recirculations pumps, absorber bleed pumps, mist eliminator wash pumps, and limestone
slurry feed pumps will be supplied with the appropriate tankage. Primary dewatering
will be accomplished with the use of hydorclones.
The absorber area equipment required for the LSFO system is considerably larger than for
the W-L. This made for a very tight arrangement and difficult erection. A plan view of the
absorber area for all four units, after the conversion, is shown in Figure 5.
-------
The reagent preparation and final dewatering will be accomplished in a new structure
(RPS/DWTR building) which has been located adjacent to the existing north side waste
water recovery building. A plot plan of the area is shown in Figure 6. The RPS/DWTR
building was integrated into this existing facility to allow reuse of an existing silo, control
room, tanks, pipe rack and electrical distribution equipment. The silo and three tanks
which were used in the waste water treatment process, will be modified to be used in the
LSFO system.
The limestone will be delivered to SJGS via belly dump truck. A drive over grizzly will be
utilized to load the limestone silo. The limestone will be ground by two by 100% capacity
wet ball mills.
Final dewatering will be accomplished with two 100% capacity vacuum filters. The gyp-
sum will be dewatered to greater than 80 % solids. The gypsum will be stacked out and
hauled away via truck back to the mine similar to the flyash and bottom ash.
A blend of cooling tower blowdown water and fresh water will be used to operate the
LSFO FGD system. The cooling tower blowdown, which was formerly recovered by the
waste water treatment system will now be used in the FGD system. This was a significant
contributor to the cost savings of the project.
Teaming Arrangement
PNM and B&W decided that the best approach for this project was for the two companies
to enter into an agreement which allowed both companies to work in a team environment
rather than the traditional supplier purchaser agreement. This arrangement allowed the
companies to work closely together toward the common goal of providing PNM with a
high quality FGD system which performs reliably to meet environmental requirements at
the lowest possible cost. This type of arrangement requires a large amount of communica-
tion and cooperation. It also required a great amount of trust on the part of both compa-
PNM , B&W and the major subcontractors participated in team building sessions prior to
the start of engineering and also prior to the start of construction. These team building
sessions went a long way in developing the trust which is required in this type of approach
to a project.
Monthly site meetings have been held which included PNM, B&W, and the major subcon-
tractors throughout the project to help communications and foster a team environment.
These sessions were attended by individuals from all levels in each organization. SJGS
engineers, mechanics, electricians, instrument journeymen, and operators were encouraged
to suggest design improvements and cost saving ideas. This approach greatly enhanced
the" buy in" to the teaming approach. Several brainstorming sessions were conducted in
order to solicit ideas for cost savings early in the project. Most of the ideas which came
from these meetings were incorporated into the project and have contributed significantly
-------
to the cost savings which have been realized to date. A few cost saving ideas could not be
incorporated into the project because of the "Fast Track" nature of the conversion. From
the time of award, the phase I equipment will be commissioned in 22 months.
In addition to the monthly meetings, weekly conference calls have been conducted
throughout the project. Much information has been transferred in these conference calls
which has kept the project on track.
The budget for the project was mutually developed by both companies through a detailed
estimate. Once this budget was established, both companies went to work on value engi-
neering improvements to lower the cost. All design issues including re-use of existing
equipment and new equipment selections were mutually agreed upon. Suggestions for
design improvements and cost saving methods have come from both parties. Other major
subcontractors have been given an incentive to generate ideas for design improvements
and cost saving methods also. The fact that the subcontractors have been invited to the
team building sessions as well as the monthly site meetings has done much to foster the
team approach.
Construction
B&W was released to begin engineering in May of 1996. Babcock & Wilcox Construction
Company (B&WCC) mobilized on the site one year later in May of 1997. Construction has
been organized in three phases to maintain environmental compliance throughout the
construction period. In the first phase, one redundant W-L absorber on each unit will be
converted to a LSFO absorber while the other three continue to operate. Also the limestone
preparation and dewatering area will be constructed. This phase is scheduled to last ten
months being complete in February of 1998. Once all of the equipment constructed in
Phase I is commissioned, the second W-L absorber per unit will be taken out of service for
conversion. The single LSFO absorber per unit will be operated in conjunction with two W-
L absorbers per unit during phase two. The second phase should last six months being
complete in August of 1998. After Phase II is constructed and the absorbers are commis-
sioned, two LSFO absorbers per unit will be available. At this point the W-L system can
be shut down. Existing compliance regulations should be attainable with two limestone
scrubbers per unit because it will be possible to pass more flue gas through the limestone
scrubbers than was possible through the W-L absorbers. The final construction phase will
be conversion of a third absorber on each unit. Phase III is also scheduled to last six
months being complete in February of 1999. The new emission requirement regulations
will require 75% SO2 removal or 0.46 Ib of SO2 per million Btu on a 30 day rolling average.
This phased approach to construction complicates the work dramatically as the new equip-
ment will be installed throughout the plant which must remain in operation. Also, many
of the piping systems which feed several cells will have to be installed and blanked off to
cells that are not in operation. The majority of new equipment will be powered from
existing power distribution equipment as it becomes available when the W-L absorbers are
decommissioned. This is also true for the distributed control system. This type of installa-
-------
tion requires a high degree of coordination, communication and team work. The teaming
arrangement being utilized for this retrofit is ideal for this type of project.
Eight PNM employees are part of the construction site organization. This was done to
reduce cost and take advantage of the plant personnel's knowledge of the plant. This
structure has improved the efficiency of the site team dramatically.
PNM is performing some of the design and construction work with site personnel. This
has reduced cost as well as enhanced SJGS ownership. PNM has the responsibility for
design and installation of the DCS as well as other complete parts of the construction
project. PNM will handle the start-up or commissioning of the system after each construc-
tion phase is complete. PNM in cooperation with B&W will perform all of the training
required for operations and maintenance personnel. This will accelerate PNM's learning
curve for operating the new system in addition to reducing the project cost.
Plant Transition
Transition to the new FGD system will involve the shut down of the old W-L FGD system
and part of the waste water treatment system. It is being accomplished through several
organizations. Start-up personnel have the responsibility not only of commissioning new
systems but ensuring operation of the old W-L absorbers through Phase II when both
systems operate together. Other subprojects will modify plant systems to accommodate
changes in waste water flows. A group was formed to plan the down sizing of personnel
since this conversion project will reduce manpower at SJGS by at least 110 positions. Vol-
untary separation packages were offered to union and non-union personnel both. Termi-
nated personnel, voluntary and selected, are being given enough incentives to remain in
their jobs until the W-L system and part of the waste water system are shut down. Another
group is looking at decommissioning and salvage of abandoned equipment. The training
department is busy training a new class of operator, Environmental Process Operator,
which replaced the old W-L Operators, Fuel and Ash Operators and Waste Water Opera-
tors. The newly trained operators will be required to operate any and all of the environ-
mental and fuel systems. Maintenance crews and supervision have been reorganized to
accommodate the new limestone system and continue to operate the old systems until they
are abandoned.
Economics
This conversion project was justified on the basis of O&M and power/fuel savings. The
reduction of personnel by 110 positions was a large percentage of the savings. The old W-L
and waste water treatment systems required 67 people in operations and the new lime-
stone/waste water system will require 26 operations people. The W-L and waste water
treatment required 70 people in maintenance. The new systems will require about 17
people in all crafts. The high cost for maintenance materials and equipment replacement
required for the twenty year old W-L system will be greatly reduced with the newer and
much less complicated LSFO system. Fuel savings for power and steam is estimated at
-------
almost $5 million per year. All of these savings are expected to add up to about $20 million
per year. In addition, there could be income from excess allowance credits.
Conclusion
This FGD project is truly unique in that it is driven entirely by economics, it is truly a
retrofit with extensive reuse of existing equipment, and it is being executed with a teaming
arrangement between the supplier and the end user. The teaming arrangement has
worked extremely well in that PNM, B&W as well as some of the subcontractors have
worked together to provide a system that will meet all of PNM's requirements at the lowest
cost possible. Both companies feel that the teaming approach offers many advantages over
the traditional contracting methods for these types of projects.
{ f Clean Flin
1 Orifice Contractor
2 Absorber Surge Tank
3. Absorber Feed Tank
4. Dump-Dissolving Tank
5- Evaporator Crystallizer
Cooling Water
Flyash
Purge to Pond
Treated Purge
Evaporator Stream
Crystallizer
Centrifuge /\ I Centrifuge
Purge Stream
To Sulfuric |Q
Acid Plant
S02
Vapors
Chiller Crystallizer
Steam
SO,
Compressor
Dried Sulfate
Product
FIGURE 1 W-L FGD SYSTEM FLOW SCHEMATIC
-------
Service Water
©Classifier
(2) Absorber Tower
(3) Ball Mill Product Pump
©Absorber Recirculation
(5) Pump
©Absorber Slowdown Pump
© Hydroclone Underflow Transfer Tank
(§) Hydroclone Underflow Transfer Pump
(D Hydroclone
©Filter Feed Pump
Filtrate Return Pump Filtrate
Tank
Unit 3 & 4 Cooling Tower Slowdown Water
Unil 1 & 2 Cooling Tower Slowdown Water •
FIGURE 2 B&W LSFO FGD SYSTEM FLOW SCHEMATIC
-------
FIGURE 3 W-LFGD ABSORBER SIDEVIEW
-------
FIGURE 4 B&W FGD ABSORBER SIDEVIEW
-------
FIGURES ABSORBER AREA PLAN VIEW
FIGURE 6 RPS/DWTR PLAN VIEW
-------
CONVERSION OF NIPSCO'S SCHAHFER
DUAL ALKALI FGD TO A LIMESTONE FGD
SYSTEM PRODUCING WALLBOARD
GRADE GYPSUM
J. D. Colley
S. M. Gray
T. E. Thomas
Radian International
9300 Shelbyville Rd.
Louisville, KY 40222
R. D. Cook
Northern Indiana Public Service Company
801 East 86th Street
Menillville, IN 46410
Abstract
The Northern Indiana Public Service Company (NIPSCO) Schahfer Station Units 17 and 18 have
used dual-alkali, flue gas desulfurization (FGD) technology for SO2 control since the startup of
these units in the mid-1980s. Due to chronic reliability problems and high operating and
maintenance costs, NIPSCO decided in 1995 to replace these systems with a more conventional
limestone technology. This paper describes the innovative approach that was taken to convert
the dual-alkali system to a limestone forced oxidation system that produces a high-quality
gypsum product for wallboard manufacture.
NIPSCO has worked with Radian International over the past two years to develop the technology
and demonstrate it at the pilot and full-scale with the assistance of EPRI. NIPSCO and Radian
also designed and constructed the new system, as well as operated and maintained it to meet
specific SO2 removal, cost, reliability, and gypsum quality goals.
-------
Introduction
This paper describes the recent conversion of the Northern Indiana Public Service Company
(NIPSCO) Schahfer dual-alkali flue gas desulfurization (FGD) systems to a new innovative
limestone-gypsum process. First, background is provided that presents the rationale for the
process conversion. The paper then describes a unique alliance developed between NTPSCO and
Radian International to develop the process technology, convert the FGD system, and operate the
new process to minimize operating costs and maximize reliability. The approach used to execute
the fast-track conversion is also described along with several of its innovative features. Finally,
actual performance results for the new process are presented, in addition to the status of
NTPSCO's gypsum marketing efforts.
Background
NTPSCO operates the R. M. Schahfer Generating Station located in Wheatfield, Indiana. The
station has four coal-fired boilers and two combustion turbines. Two boilers (Units 17 and 18)
operate with FGD systems and have a rated capacity of 393 MW gross each. The units bum a
coal with a typical FGD inlet SO2 loading of 5.3 Ib/MMBru and operate in a load-following
manner at an annual capacity factor of about 70 percent. The dual-alkali FGD systems were
installed in 1982 and 1984. Recent operation of these systems is described below.
Dual-Alkali System Performance
The dual-alkali FGD systems at the Schahfer station first began operation in the early 1980s.
Over the last decade of operation, the performance and reliability of these systems degraded and
operating costs increased. Higher operating costs were attributed to increased consumption of
the two reagents, lime and soda ash, and increased maintenance and materials costs.
To address the performance problems associated with the dual-alkali process, NTPSCO began a
program to evaluate potential solutions. As part of this program, NTPSCO contracted with
Radian in 1992 to evaluate alternative processes that would provide improved performance and
reliability while simultaneously lowering operating costs, but without the need for significant
capital expenditures. Radian's analysis indicated that optimization of the existing dual-alkali
process offered the most promise. As a result, Radian provided on-site engineering support in
the fall of 1992, and again in the fall of 1994, to implement process improvements. The primary
focus of these efforts was to reduce the high consumption rate of soda ash.
Figure 1 shows the soda ash molar consumption ratio (moles of soda ash consumed per mole of
SO2 removed) for each quarter from 1992 to 1995. As shown, the soda ash consumption ratio
well exceeded the original design guarantee of 0.035. This caused an increase in annual
operating costs of more than $2 million over design conditions. The figure also shows that while
process optimization programs conducted during the fall of 1992 and 1994 were successful in
dramatically reducing the consumption of soda ash, these improvements were not sustainable
because of further degradation and poor reliability of critical process equipment.
-------
(0
en
O.
I
w
o
o
J5
o
n
•o
o
CO
0.25
0.20
0.15
0.10
0.05
0.00
Design Consumption Rate =0.035
1Q 2Q 3D 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 40 1Q 20 3Q 4Q
1992 1993 1994 1995
Figure 1
Historical Soda Ash Consumption
Evaluation of Process Alternatives
In the spring of 1994, NIPSCO formed a task force to identify and implement the best long-term
solution to the problems plaguing the Schahfer FGD systems. The task force identified the
following objectives for the improved FGD systems:
• Continuous SC>2 compliance;
• High reliability;
• Low operating cost;
• Salable byproduct; and
• Low capital expenditures.
To achieve these goals, it was clear that a change in the FGD process technology would be
required and that unreliable process equipment would have to be eliminated. To this end,
NIPSCO prepared and issued a Request for Proposals (RFP) to several companies specializing in
FGD technology and services.
A review of the proposals NIPSCO received showed that while the primary objectives could be
met, the capital costs to do so were excessively high and the proposed project schedules
exceeded 24 months in most cases. As a result, NIPSCO deemed the proposals unacceptable.
Based on past experience with the Schahfer dual-alkali process and recent R&D results, Radian
was able to propose an innovative technical approach to solving NIPSCO's FGD problems that
offered several advantages to NIPSCO. These advantages included:
-------
• Use of lower cost reagents;
• Production of wallboard grade gypsum;
• Elimination of unreliable process equipment;
• Reuse of the majority of existing equipment;
• Simplified process operations;
• Lower capital costs; and
• Shortened project schedule.
On the basis of these advantages, NIPSCO evaluated the innovative approach and entered into an
alliance with Radian to convert the Schahfer FGD systems.
FGD Alliance
Development of the Alliance
The development of an alliance between NIPSCO and Radian was a unique feature of this
project. The key components of the alliance are described below.
Shared Risks/Rewards. The strength of the Alliance rests in the concept of shared risks and
shared rewards. NIPSCO and Radian were familiar with this approach from prior projects at the
Schahfer FGD system. In 1994, Radian and NIPSCO conducted an intense four-month
optimization program of the dual-alkali FGD system. As part of the program, specific cost
reduction goals were established. If met, Radian was to receive the full contract amount for the
support program; if not, Radian's compensation would be significantly less. Thus was born the
concept of shared risks and rewards. This principle formed the foundation of the Alliance.
Facilitated Process. Another important step in developing the Alliance was the use of a
facilitator. NIPSCO contracted with a third-party management consulting firm to facilitate the
development of the Alliance. This process allowed the two companies to cooperatively define
and discuss their needs so that they could develop a mutually beneficial plan to meet these needs.
Integrated Staff. Another key component of the Alliance was integration of the staff of the
two companies. This concept extended from project inception to project development to project
implementation to long term operations and maintenance (O&M). By establishing a common
team working cooperatively to achieve the goals of the Alliance, project milestones could be
achieved more consistently.
Long-Term Commitment. NIPSCO management expressed a strong interest in a business
relationship different from that of the typical customer-vendor relationships they have had in the
past. The Alliance provides the framework for a relationship in which the two companies work
together over the long term to ensure that their mutual needs are meet, rather than just having
product guarantees met during initial operation only to see performance and reliability degrade
with time.
Continuous Improvement Incentives. The last key component of the Alliance was the
inclusion of incentives for continuous improvement. With the approach of deregulation,
NIPSCO realized that continuous improvement would be critical to their success in being a low-
-------
cost provider of electricity. As a result, the Alliance includes provisions that motivate both
companies to maintain their relationship and provides rewards to both companies for
improvements in the operation and maintenance of the FGD system.
Project Analysis
Once the framework of the Alliance was developed, a thorough business analysis of the project
was performed. This analysis addressed three critical areas; technical, economic, and schedule.
The Alliance team worked together to perform the analyses.
Technical. Conversion of the existing dual-alkali process to a limestone-gypsum process
presented numerous challenges. The analysis focused on whether the process technology
identified by Radian could be adapted to the Schahfer FGD system and achieve the goals of SC>2
removal compliance, high reliability, low operating costs, and low capital costs. The Alliance
team concluded that pilot-scale testing was warranted to better simulate the chemistry and
physical equipment constraints of the Schahfer FGD system and absorbers. In addition, the team
felt that a full-scale demonstration was needed to both minimize technical risks and to optimize
final equipment design, thus reducing capital costs.
Economic. Conversion of the dual-alkali process to a new process technology represented a
significant capital investment. As a result, a thorough economic analysis was performed to
ensure that the project met NIPSCO financial requirements. This analysis considered all cost
components associated with operation and maintenance of the FGD system including:
• Reagent and additive;
• Waste stabilization;
• Waste landfilling;
• Emergency pond cleaning;
• Operating and maintenance labor;
• Maintenance materials; and
• Byproduct sale revenues.
Results of the analysis indicated a potential savings of $7-10 million annually and a capital
payback period of only a few years. Based on these favorable results, NIPSCO management
elected to move forward with the project contingent upon successful full-scale demonstration of
the process technology and the development of a detailed project capital cost estimate.
Schedule. Project schedule was also critical to NIPSCO management. The pending
deregulation of the utility industry meant that NIPSCO needed to reduce operating costs rapidly.
The lengthy process of issuing an RFP, evaluating bidder qualifications, and reviewing bidder
proposals, only to find the costs unacceptable, had delayed NIPSCO in achieving its objectives.
As a result, it was critical that the project be able to meet planned outages for the two units
during the winter of 1996/1997.
The alliance team prepared an aggressive project schedule to meet this deadline. Figure 2 shows
the major activities associated with the project. As shown, it was necessary to conduct several
activities concurrently.
-------
Project Activities
Eccrcrric Analjss
Rlcl-Scale Testing
KD Alliance E&dqranl
Rjll-Scale Efenmsaadcn
EesigiyBigineering
Equipment F
Gusttuaian
Training
UritCttages
Stamp
Figure 2
Conversion Project Timeline
Process Technology Development
Conversion of a dual-alkali FGD process to a gypsum-producing process presented many
technical challenges. The absorbers at Schahfer had disc and donut internals for gas-liquid
contacting, a low liquid-to-gas (L/G) ratio of -15 gal/macf, and very small reaction tanks. As a
result, the Alliance team had to develop and adapt existing technologies to the site-specific
conditions at the Schahfer station. The pilot-scale and full-scale testing conducted to develop
and demonstrate the technology are briefly described below
Pilot-Scale Testing. Pilot-scale testing was conducted at EPRI's Environmental Control
Technology Center (ECTC) located near Buffalo, New York, during the spring and summer of
1995. Testing was conducted on the five-foot-diameter pilot wet scrubber (4 MW). The testing
focused on the use of both lime and limestone reagents in conjunction with in-situ oxidation in a
shallow reaction tank. Organic acids were used during the testing to enhance SQz removal
performance while operating at relatively low L/G ratios.
Test results showed that good SO2 removal, reagent utilization, and gypsum purity could be
achieved using these reagents. Furthermore, gypsum dewatering properties with the limestone
reagent were superior to those produced during operation with lime under Schahfer test
conditions
Full-Scale Demonstration. Following successful testing at the pilot-scale, a full-scale
demonstration test was conducted at the Schahfer station in the fall of 1995. The test program
was jointly funded and conducted by NIPSCO and Radian. As part of the demonstration, one
absorber module was converted from dual-alkali to limestone-gypsum operation. Five weeks of
-------
testing were conducted to examine the effects of slurry pH, oxidation air rate, dibasic acid
concentration, and reaction tank slurry density and level on 862 removal performance, reagent
utilization, and gypsum quality.
Test results showed that the Schahfer absorbers could successfully be converted to produce a
wallboard-grade gypsum. As part of the demonstration, the gypsum produced in the single-
module test was dewatered and washed using the existing rotary vacuum drum filters and then
used for a production test run at a nearby wallboard plant. The test run showed that the gypsum
material met the required specifications for production of a high-quality wallboard.
Conversion Project
Schedule and Approach
The project schedule was driven by major maintenance outages planned for the generating units
late in 1996 for Unit 17 and early in 1997 for Unit 18. Based on the timing and completion of
the full-scale demonstration test, this created a fast-track schedule. In essence, there were only
13 months from the start of process engineering to full gypsum production for Unit 17 and 15
months for Unit 18. Typically, a project of this complexity would be expected to take 24 months
or longer to complete.
The fast-track schedule demanded a project approach that required all equipment lead times to be
analyzed and incorporated into the detailed project schedule. Several pieces of critical process
equipment (i.e., oxidation blowers and absorber recycle pumps) were procured immediately
following the demonstration test based on preliminary engineering. Most other equipment and
components were procured for just-in-time delivery and installation during the two six-week unit
outages.
Roles and Responsibilities
Based on the importance of this strategic project, and its complexity and schedule constraints,
NTPSCO recognized that a traditional project approach was not the best one and asked Radian to
perform total project implementation responsibilities. Radian was responsible for developing
and implementing the entire conversion process including:
• Overall project management;
• Process engineering;
• Detailed design engineering;
• Procurement;
• Construction and construction management;
• Equipment commissioning;
• Startup; and initial operation (including training);
• Routine operation; and
• Process optimization.
Even though Radian was given total project responsibilities, it was essential that NIPSCO
maintained its involvement throughout the project. NIPSCO, acting in an oversight role,
-------
provided key direction, with a consistent emphasis on project objectives, and performed vital
functions during the commissioning and startup of the converted systems.
Process engineering included all of the pilot- and full-scale demonstration testing that NIPS CO
and Radian performed, with support from EPRI. During the process engineering phase, Radian
defined the basis of the converted FGD process design by developing the process flow diagram,
material balances, and piping and instrumentation design, and combining these into a process
design package.
Radian subcontracted with Duke/Fluor Daniel, a full-service engineering company, for the
detailed design engineering, procurement of new process equipment, and construction
management. The detailed design engineering consisted of specifying all of the equipment
needed to implement the process design. To cost effectively construct the converted FGD system,
Duke/Fluor Daniel competitively bid the installation work to local contractors that had
experience with the local labor force.
Radian was responsible for commissioning the new FGD system, a process involving field
verification of the newly constructed system and a mechanical checkout of the system's
components (e.g., electrical system, instrumentation, loops, etc.). During system startup, Radian
engineers introduced limestone, dibasic acid, and flue gas into the system to test its ability to
meet design specifications.. Radian also managed the initial operation of the system and trained
NTPSCO's FGD operators. The system is now in its routine operation phase; Radian's process
optimization efforts continue to take place and will throughout the life of the alliance.
Innovative Design Features
Figure 3 presents a simplified process flow diagram for the new process and shows newly
installed equipment. Numerous innovations were incorporated into the conversion project and
are briefly discussed below.
Reuse of Equipment. Reuse of major process equipment was a key feature of the project.
Reagent silos and storage tanks, absorbers, drum filters, conveyors, and other process vessels
were reused to minimize capital costs. Equipment with a history of poor reliability and high
maintenance costs, including lime transfer and slaking equipment and thickeners and underflow
pumps, were abandoned to ensure high reliability and minimum operating costs for the new
process.
Limestone Slurry Preparation. The new process uses a powdered limestone supplied by a
third party and prepared using dry roller-mill grinding equipment. To minimize capital costs, the
existing lime storage silos were modified to store the powdered limestone. In addition, ejector
mixing technology was used in combination with existing plant equipment to provide a highly
reliable, cost effective limestone slurry preparation system.
Absorber Oxidation. The existing absorber reaction tanks were very small and operated at a
shallow depth. To minimize capital costs, minor modifications were made to the absorber to
increase the slurry operating depth, thus permitting effective in-situ oxidation.
-------
Mist Eliminator
Service __
" Water Rue Gas In
Oxidation Air
Blowers (3+1 Total)
Chloride
Slowdown Service
Absorber Gypsum
Recirc. Bleed
Pumps Tank(l)
(2)
Limestone Feed
Pumps (1+1)
Service
Water
Stacker (1)
Water
1
, J
Recycle
Water
Tank(l)
Return
Pumps
(1+1)
Gypsum Scackout
Pile
COOOOOQ
Transfer Conveyor (2)
Conveyor (1)
FigureS
Simplified Limestone-Gypsum Process Flow Diagram
Absorber Design. The dual-alkali process used a very low L/G ratio and disc and donut
absorber internals for gas contacting. To minimize the capital costs of the new limestone-based
process, the absorber internals were left unchanged and new slurry recirculation pumps and
piping were installed to increase the L/G ratio. In addition, a removal-enhancing additive,
dibasic acid, was used to enhance SO2 removal performance to desired levels.
Absorber Agitation. A cost-effective alternative to traditional absorber reaction tank
mechanical agitation systems was used to minimize capital costs. The oxidation air distribution
and sparging systems were designed to effectively provide a high degree of agitation, thereby
eliminating the need for conventional mechanical mixers.
Single-Step Dewatering. The new process was designed to operate at an elevated slurry
solids concentration in the absorber sumps. As a result, the need for an intermediate dewatering
step, before final dewatering by drum filters, was eliminated. This significantly reduced the
capital costs of the project and eliminated the troublesome thickener equipment and operation.
-------
FGD Performance Results
Because the project involved the conversion of an existing FGD system, it was critical that the
new process achieve the required performance immediately upon startup. This was made more
challenging by the complete upgrade of the existing controls with a new PLC-based system
which required that all new and reused equipment be integrated during the relatively short unit
outages. Both FGD systems were successfully converted on schedule and were available to
support boiler operation by the end of each unit's outage. Presented below are both technical and
economic performance results during the first six months of operation.
SO: Removal
A key criteria for the conversion project was the ability to maintain SC>2 removal compliance and
minimize environmental compliance risks. Figure 4 compares the daily average SOj emissions
Dual Alkali
Process
Limestone-Gypsum
Process
Figure 4
SC>2 Removal Performance Comparison
of the dual-alkali process and the new limestone-gypsum process. As shown, SC>2 emissions of
the new process are well within compliance requirements. In addition, the figure illustrates the
dramatic improvement in SO2 removal control and stability with the new process compared to
past operation with the dual-alkali process, thereby reducing environmental risks and unit derates
caused by poor SO2 removal performance. While daily average SO2 emissions are shown, the
permit limit for the FGD system is based on a 30-day weighted rolling average.
10
-------
Reagent Utilization
The primary cause of high operating costs of the dual-alkali process was poor utilization of the
lime and soda ash reagents. As a result, it was important to achieve good utilization of the
limestone reagent to minimize operating costs as well as ensure production of high-purity
gypsum. Figure 5 presents the monthly average limestone utilization for the new process. As
shown, utilization ranges from 94 to 96% compared to the goal of 95%.
100
-?98
ll 96
5 94
| 92
JE 90
ffi 88
2 86
10
I84
'J 82
80
Goal>95%
mini
Jan Feb Mar Apr May Jun
Figure 5
Limestone Utilization Results
Gypsum Quality
Because NBPSCO planned to market the gypsum produced by the new process, it was important
it meet the quality specifications of potential buyers. As a result, quality specifications were
developed for the three key parameters; purity, chloride content, and moisture content. Table 1
summarizes the desired and acceptable ranges for these key parameters.
Table 1
Gypsum Quality Specifications
Desired
Acceptable
Purity (wt. % Gypsum) > 95
Chloride Content (ppm, dry) < 100
Moisture Content (wt. %) <15
>93
<150
-------
To ensure production of high quality gypsum, performance goals were established that were
consistent with the desired quality levels. Figure 6 presents the monthly average gypsum purity
for the first six months of operation. As shown, gypsum purity ranged from 95 to 97%,
exceeding the goal of 95%.
100
99
^ 98
!„
^- 96
>.
£ 95
3
0- 94
93
O
92
91
90
Goal>95 Wt.%
Jan
Feb
Mar
Apr
May
Jun
Figure 6
Gypsum Purity Results
Figure 7 presents the monthly average gypsum chloride content for the first six months of
operation. As shown, gypsum chloride levels ranged from over 200 ppm during initial operation
to less than 50 ppm during recent operation. The high levels during initial operation were a
result of cold weather-related startup problems associated with the filter cake wash system.
Figure 8 presents the monthly average gypsum moisture content for the first six months of
operation. As shown, gypsum moisture levels ranged from 13 to 15%. Recent improvements in
the operation of the drum filters and filter cake wash system have resulted in reduced moisture
levels.
Reliability
Improving the reliability of the Schahfer FGD systems was a very important goal for the project.
Operation of the dual-alkali process had resulted in significant derates and forced outages of the
generating units. As a result, the Alliance established an FGD Equivalent Forced Outage Rate
(EFOR) goal of less than 2% for operation of the new process. EFOR is calculated as the
weighted sum of forced outages and derates caused by the FGD system relative to the hours the
FGD system was called upon to operate.
12
-------
300
Jan Feb Mar Apr May Jun
Figure 7
Gypsum Chloride Content Results
20
18
£16
I"
0 12
I10
i 8
I •
to
Goal<15 Wt.%
0
mm
Jan Feb Mar Apr May Jun
Figure 8
Gypsum Moisture Results
13
-------
Figure 9 presents the quarterly EFOR results of the FGD system during dual-alkali process
operation in 1996 and limestone-gypsum operation during the first half of 1997. As shown,
conversion to a limestone-gypsum process and elimination of unreliable high maintenance
equipment has significantly improved FGD system reliabDity.
£
0)
01
CO
I 6
a 4
£
§•
LLI
Dual-Alkali
(1996)
Limestone-Gypsum
(1997)
1Q
2Q
3Q
4Q
1Q
2Q
Figure 9
FGD Reliability Results
Operating Costs
Probably the most significant performance result of the conversion is the reduction in the O&M
costs of the FGD system. Figure 10 presents a comparison of the projected O&M costs for the
dual-alkali process compared to the new limestone-gypsum process. The projected costs are
based on actual performance during the first six months of operation projected through the end of
1997. As shown, the process conversion has resulted in a projected reduction in O&M costs of
approximately $9 million/year. Most of these savings have been in the areas of reagent and
waste disposal. Some savings have also been realized as a result of reduced overtime labor and
maintenance materials due to improved and simplified process operations and elimination of
high-maintenance equipment. It should be noted that while the savings reflect the elimination of
the need for waste stabilization and landfilling, they do not reflect revenues generated by the
future sale of the gypsum byproduct.
Gypsum Marketing
In 1995, NIPSCO began an assessment of the gypsum market in relation to the potential process
conversion at the Schahfer Station. This assessment concluded that the wallboard industry was
the most appropriate target for Schahfer's synthetic gypsum. As a result, NIPSCO prepared and
14
-------
issued an RFP to potential buyers to establish interest and the potential market value of the
gypsum.
*
o
o
eS
O
•o
o
2"
Q.
Net Savings~$9.0 M/yr
(Excludes Gypsum Sale Revenue)
Dual-Alkali
Limestone-
Gypsum
Figure 10
Projected O&M Cost Savings
The credibility of NIPSCO's gypsum offering was enhanced by actual production of high-quality
gypsum. Approximately 200 tons of high-quality gypsum were produced during the full-scale
demonstration test conducted in late 1995. This material was successfully used by an interested
wallboard company to make a full-scale production run of wallboard. Following successful
startup of the new limestone-gypsum process in early 1997, an interested party conducted another
successful full-scale wallboard production test using approximately 1700 tons of gypsum. Again,
NIPSCO demonstrated that the Schahfer FOB system could produce a high-quality synthetic
gypsum.
Based on the successful conversion of the Schahfer FGD system to limestone-gypsum operation
and the production of high-quality gypsum, a large wallboard manufacturer recently announced
plans to construct a wallboard production facility in Wheatfield, Indiana. Design and engineering
of the $65 million facility are underway with an expected startup date in late 1999. The facility
will employ up to 100 people at full production and will use 100% of the gypsum produced by
the Schahfer FGD systems.
Summary
NIPSCO had experienced increasing operating costs and decreasing reliability of the dual-alkali
15
-------
FGD systems operated at the company's Schahfer Station. As a result, they evaluated alternative
process concepts to reduce costs and improve reliability. Then- evaluation revealed that while
these objectives could be achieved, the capital cost estimates provided by FGD vendors were
unacceptably high.
As a result, NIPSCO began working with Radian International to develop a process technology
that would provide the desired benefits, but at a much reduced capital cost. This relationship
developed into an alliance between the two companies based on the principle of shared risks and
shared rewards.
FoDowing a thorough economic analysis and successful demonstration of an innovative process
technology with the support of EPRI, NIPSCO initiated a project to convert the Schahfer FGD
system from dual-alkali to limestone-gypsum operation. To ensure that the integrity of the
process technology was maintained through commercial operation, Radian was awarded full
project responsibility including design, engineering, procurement, construction, training, startup,
and process optimization.
Innovative design features of the converted FGD process include the use of pre-ground limestone
reagent, use of ejector-mixer technology for reagent slurry preparation, in-situ oxidation in
shallow absorber sumps, absorber slurry agitation without mechanical agitators, and single-step
dewatering.
Performance goals for the new process have been met including SC>2 removal performance,
reagent utilization, gypsum quality, and reliability. In addition, projected O&M cost savings
attributed to the process conversion are approximately $9 million/year, based on actual
performance during the first half of 1997.
NIPSCO conducted a thorough market analysis of potential buyers for FGD gypsum. Following
successful operation of the new limestone-gypsum process and production of high-quality
gypsum, a large wallboard manufacturer announced plans to install a nearby wallboard plant that
will use 100% of the Schahfer gypsum. The plant is expected to be operational by late 1999.
Acknowledgments
The authors acknowledge the contributions of the NIPSCO project team to the success of this
project including members of Major Projects, FGD Operations and Maintenance, and Electric
Production Business Planning and Environmental Affairs.
The authors also acknowledge EPRI for its assistance and support in demonstrating the process
technology and Radian's Engineering and Operations project team members for their support.
16
-------
Wednesday, August 27; 2:00 p.m.
Parallel Session A:
FGD Process Improvements
-------
Wet FGD Forced Oxidation:
A Review of Influencing Factors
and a Comparison of Lance and Sparge Grid
Air Introduction Methods
Kevin J. Rogers
The Babcock & Wilcox Company
20 S. Van Buren Avenue
Barberton, Ohio 44203
Paul D. Dwelle
EKATO Corporation
700C Lake Street
Ramsey, New Jersey 07446
Abstract
The dominant worldwide wet Flue Gas Desulfurization (FGD) design is the limestone forced-
oxidized system, whereby limestone is utilized as the reagent and the process incorporates air
injection to ensure a fully oxidized gypsum product. The method chosen for the introduction of
oxidation air has an influence on plant costs and system operating requirements.
Air sparge grids and air lances with mechanical agitators are two generally applied methods of
introducing oxidation air into the process. The sparge grid is a multiple air header arrangement
with near even spacing of bubble stations across the vessel plan area. The lance system consists
of air pipes directed to a definite region in the liquid jet created by side entry mixers. The
performance of the lance system is influenced by the energy of the fluid jet (mixer power) and the
submergence depth (compressor power). The performance of the sparge grid is less dependent
on the mixer power and is, to a much greater degree, influenced by submergence depth.
A general review of the parameters influencing mass transfer, as well as the design attributes of
sparge grids and lances, is made to allow for a comparison of each method of air introduction.
Introduction
The oxidation characteristics and performance capabilities of a given FGD system are influenced
by a variety of factors. Vessel geometry, liquid depth, natural oxidation, dissolved and
suspended solids, system temperature, pH, and energy input are among those that affect the
results. Further, the method used to introduce the air into the process influences the equipment
arrangement, operating requirements, and overall efficiency.
Various methods exist to introduce and disperse oxidation air into the wet FGD process.
Examples include jet mixer/aerators, sparging beneath radial impellers, perforated plates, sparge
grids, porous diflusers, rotating sparge/impeller arms, and agitator air lance assemblies.
-------
Two primary methods for typical wet limestone forced oxidized FGD systems are sparge grids
and agitator air lances. Each have their own attributes which influence both suitability and cost
effectiveness. This paper is a general discussion of these two system types.
Factors of Influence
The oxidation requirements and capabilities of a given system are influenced by numerous factors.
The degree of oxidation occurring naturally is of importance. An increase in the natural oxidation
component can allow forced oxidation air flow requirements to decrease. A known degree of
natural oxidation can also indicate other factors. For example, the presence of an inhibitor may be
indicated if low natural oxidation is observed at relatively high O2 and low SO2 inlet flue gas
conditions.
Liquid depth provides an influence through several mechanisms. At a constant bubble rise
velocity, the air residence time increases with increasing depth. At constant air flow, the system
power input increases and thus the power dissipation per unit volume is affected by depth. Air
hold-up and average oxygen solubility are similarly influenced by liquid depth through variations
in hydrostatic head.
Vessel geometry can influence design options and system performance. The effects of changes in
basic shape, square versus round for example, are mostly to vary header arrangements, agitation
requirements, and bulk fluid flow profiles. For a constant volume, changes in the liquid height to
tank diameter ratio can influence the values for oxidation air hold-up and the corresponding liquid
level rise.
Dissolved solids can affect the process in multiple ways. Their presence can increase the density
of the solution and thus influence hydrostatic pressure and energy input. Dissolved species with
oxidation inhibiting and/or catalytic effects can be present. Dissolved impurities can improve
mass transfer by lowering static surface tension to allow smaller bubble formation and retard
coalescence. However, impurities concentrating at the interface can also give rise to increased
resistance, retard circulation near the interface, and lead to a reduction in mass transfer.
Suspended solids can influence the process in ways similar to dissolved solids. The effects on
slurry density or specific gravity must be considered. Ultra-fine, insoluble impurities, can produce
similar interfacial effects. The aqueous volume of a fixed vessel geometry is decreased as
suspended solids concentrations increase.
Viscosity can affect shear, turbulence, bubble size stability and terminal rise velocity. It can also
influence molecular diffusivity.
The air bubble diameter has a significant effect on mass transfer, as it directly influences the
available interfacial surface area. Bubble formation diameters are typically much larger than the
final average diameter. Localized fluid flow profiles and power dissipation significantly influence
the break-up of the bubbles formed and the final stable diameter.
Vessel fluid flow profiles indicate of the degree of mixing and will impact oxidation performance.
-------
Flow profiles that increase air hold-up can improve mass transfer.
Changes in system temperature will influence reaction rates, oxygen molecular diffusivity and
solubility. As temperature increases, molecular difiusivity increases. However, oxygen solubility
decreases with increasing temperature.
Hydrogen ion concentration is known to affect sulfite phase equilibria and the chemical form of
aqueous sulfite. As a result, oxidation rates are shown to reduce at increasing levels of pH.
System energy input can have a significant effect. The major contributors are mechanical
agitators, recirculation pumps and the compressor power at the injection point. The relative
influence of each power component may change as the system design varies. Lances are more
dependent and scalable to agitator power than compressor power at the injection point. Sparge
grid performance is more dependent and scalable to compressor power at the injection point than
agitator power.
Method of Air Introduction
Sparge Grid System
Sparge grids are an effective method of dispersing air into an absorber vessel. They typically
consist of a multiple header arrangement with an approximate equal spacing of bubble stations
across the vessel plan area. (Figure 1)
The performance of one sparge grid design from that of another can vary significantly. To
optimize mass transfer capabilities, a well designed sparge grid takes into account grid geometry,
bubble station design, submergence depth and air flow rate. Additionally, hydrostatic pressure
and the grid air pressure drop characteristics must be properly analyzed to ensure adequate
distribution of air to the bubble station points.
The mass transfer efficiency is dependent on the air-liquid interface area available and the contact
residence time. The mass transfer surface area is. dependent on the final average stable bubble
size, while the residence time is dependent on the average effective bubble rise velocity.
Reaction tank flow profiles produced by air sparging are dependent on vessel diameter, sparger
geometry, bubble station spacing, air flow rate, slurry level, and grid submergence depth. These
flow profiles can influence bubble break-up and residence time within the vessel slurry volume.
With the typical high air flow rates through wet FGD sparge grid systems, bubbles will form with
a near constant frequency. The specific formation diameter and frequency is dependent on the
sparger design, air flow rate per bubble station, and the physical properties of the air and liquid.
Above the initial bubble formation zone, turbulent liquid circulation causes the bubbles to break
up into an average size which is balanced by surface tension and shear forces. This final average
bubble size will, to a large extent, dictate the oxygen mass transfer potential of the system.
In a stagnant system, the residence time would be solely dependent on the bubble terminal rise
-------
velocity. The terminal rise velocity is dependent on bubble size and surrounding fluid properties.
The terminal rise velocity of air in water, over the range of expected bubble sizes, is estimated at
0.25 m/s. The actual rise velocity is influenced by the state of surrounding fluid motion. The fluid
motion is governed by vessel geometry, slurry level, air flow rate, sparger design, the slurry
recirculation rate and mechanical agitator induced flows.
FGD units incorporating sparge grids will often have agitators arranged to produce a specific flow
profile. The flow profile corresponds to the one the agitator supplier recommends for the best
solids suspension process result. The influence agitator induced fluid flows can have on sparge
grid performance is not always fully considered; primarily due to the difficulty in properly
characterizing the effect.
Conversely, the physical presence of a sparge grid across the tank cross-section, as well as the air
plume induced fluid flow, can disturb the desired flow profiles attempted by agitators designed for
solids suspension only. Harmonizing agitator and sparge grid designs, in the most efficient and
reliable way, is not always practical. The independent design approach can potentially lead to
increased error in sparge grid performance prediction. However, when a sparge grid is located far
above the agitators and the compressor input power is much greater than the agitator input
power, the influence agitators have on the oxidation air induced flows can often be considered
negligible.
Sparge grids require a minimum level of fluid above the grid elevation for acceptable
performance. Clearance requirements between the vessel floor, the agitator impeller tips, and
the sparge grid can set a minimum sparge grid centerline elevation at nearly 20% of the vessel
diameter (T). With the sparge grid oxidation performance tied to submergence depth, the
minimum un-aerated slurry level for sparge grid systems can easily reach 0.4T and greater. To
provide optimum efficiency, sparge grid systems can require normal operating slurry levels of
0.5T and greater. (Figure 2)
In general, the mass transfer performance of a given sparge grid design is proportional to air flow
rate and the submergence depth. The basic relationship can be described as;
Sparger Mass Transfer Performance <* C(H)(Q)/V (1)
Where,
H = Submergence depth, (m)
V = Volume, (m3)
Q = Air Flow, (mVhr)
C = Empirically derived constant
To ensure proper air distribution and to minimize sparge grid plugging, the typical maximum
oxidation air flow turndown capability is approximately 30%. This turndown capability can be
improved somewhat by increasing the design bubble station pressure drop. The drawback is an
undesirable power penalty at the full load condition.
The ability to increase oxidation air flow, beyond the full load design point, is limited by the total
system resistance and the capability of the compressed air source.
-------
Lance System
Lances are open ended pipes that extend in front of an agitator impeller face. The high localized
fluid flow, turbulence and energy dissipation, forces a high degree of bubble break-up and overall
air dispersion into the entire volume of the vessel. (Figure 2 & Figure 3)
Pilot and full scale operation experience suggests lances have a reduced dependence on
hydrostatic head versus that of a sparge grid. The basic reasons being the production of smaller
bubbles and an agitator flow that improves residence time by imparting horizontal motion. This
trait makes lance designs more effective on FGD systems operating with low reaction tank levels.
In terms of operation, lances have an infinite air flow turndown capability. With respect to
increasing air flow, the compressor size and the dispersion capability of the fluid jet become the
governing factors. As the fluid jet dispersion capabilities are exceeded, the mass transfer
capability can deteriorate. If air flow is in significant excess of the dispersion capability, a
flooding condition can result.
Flooding occurs when the air flow rate is sufficient to penetrate the envelop of the fluid jet and
becomes caught in fluid currents flowing towards the suction side of the impeller. Under these
conditions, the apparent density of the fluid being pumped decreases and the pumping rate falls.
As the fluid jet deteriorates, a greater degree of air penetrates, further reducing pumping rates.
When pumping rates drop substantially, there are no longer sufficient currents to cany air to the
impeller suction. Thus, fluid flow is re-established and the cycle repeats until the air flow rate is
reduced.
In general, the mass transfer performance of a given lance system is proportional to air flow rate
and the energy input. The basic relationship can be described as;
Lance Mass Transfer Performance « (P/V)a (VSG )b (2)
Where,
P/V = Agitator Supplied Power/Unit Volume, (W/m3)
VSG = Superficial Gas Velocity, (cm/s)
a, b = Empirically derived exponents
Comparisons:
The design requirements and attributes of each method yield variations in capital cost, energy
efficiencies and operating requirements. These can be generally categorized as follows.
Design and Operating Limitations
To minimize plugging, sparge grid operation typically requires air flow at all times, whenever
slurry is covering the grid. Lance designs are essentially immune to plugging. Thus, they do not
typically have this plant operating requirement and/or the control interlocks and alarms associated
with it.
-------
To ensure optimum sparge grid efficiency and reliability, good air distribution across the grid is a
necessity. This distribution requirement will typically limit the sparge grid turndown capability to
about 30%.
On the other hand, lances have 100% turndown capability. However, in order to take full
advantage of the turndown capability, the compressor design must be evaluated in that respect.
Increased compressor control logics may be required. More costly inlet guide and diffuser vane
designs may be justified to maintain higher energy efficiency across the compressor turndown
range. Alternate approaches which consider the use of more smaller compressors, that can be
brought on or off-line, may adequately accomplish the desired turndown range. This again can
influence the plant design philosophy on spare compressors, equipment and piping arrangements,
electrical requirements, controls and logics.
In contrast to the case for sparge grids, the lance agitator is designed considering air dispersion.
The harmonizing of the agitator design for solids suspension and air dispersion occurs by default.
As such, the performance predictions for lances can potentially be more predictable. Agitator
requirements for lance systems will typically provide a design that is conservative with regard to
the solids suspension task.
Compared to sparge grids which have practical limitations on operating slurry levels, the lance
system is more tolerant of designs with low tank liquid levels. They can work well at fluid levels
too low to be practical for vessels containing agitators for solids suspension and sparge grids for
oxidation.
Incorporation of lance systems can produce equipment arrangement difficulties at the base of the
absorber tower, as the quantity of agitators is increased to accomplish the needed air dispersion.
High sulfur loadings, which can dictate relatively high oxidation air flows, may lead to a
substantial increase in agitator size and/or quantity. These same conditions typically require
increased quantities and/or sizes of absorber recirculation pumps. This can quickly lead to
arrangement competition between the two. While achievable, the task of providing an acceptable
arrangement from a process and maintenance standpoint for all the required agitators,
recirculation pumps, recirculation piping, oxidation air piping, miscellaneous process and
instrument connections, can become significant.
Capital Costs
Wide variations in FGD system design requirements has made it difficult to fully evaluate and
clearly predict which system is the most cost effective from an initial capital cost standpoint.
For the same basic absorber design, sparge grid systems will often operate at a lower compressor
discharge pressure and a higher air flow than the lance system. A lance system, while often
requiring less air flow, will typically require a higher compressor discharge pressure due to the
increased submergence depth at the agitator/lance assembly. As a result, capital costs can be
influenced through possible alterations in compressor design, type, and spare equipment
philosophy.
If the oxidation air flow requirement is such that few or no additional agitators are required, then
the capital cost of the sparge grid typically forces it to be the most expensive. However, when the
-------
lance system agitator requirement significantly exceeds that needed for solids-suspension only,
capital cost economics can begin to favor the sparge grid.
Material of construction is important. A lined carbon steel absorber tower versus alloy
construction, will alter the economics. The effect is more pronounced for sparge grids, as the
total mechanical support requirements are typically greater than those for lances. Additionally, as
the tower diameters increase, sparge grid support systems will become more elaborate and
complex.
A useful capital cost analysis needs to include variations in costs associated with compressors,
piping, agitators, mechanical supports, motor control centers, electrical wiring, control system
input/output, etc.
Maintenance
The primary maintenance consideration with sparge grid designs is plugging potential. Sparge
grids can be designed to tolerate a high degree of plugging and remain functional. A well
designed sparge grid can be capable of operating without inspection or cleaning, except at
scheduled outages. Operational upsets, where oxidation air flow is lost, can increase the need for
grid inspection and/or cleaning.
Compressor design can affect the ability of a sparge grid to provide the needed mass transfer
while operating under partially plugged conditions. Dynamic compressors, which can have a
reduction in air flow as system resistance increases, are not always as forgiving as positive
displacement machines.
Similarly, if the margin on the required oxidation air flow rate is low, lance systems can become
susceptible to the loss of an agitator. Air flow can remain directed to the lance of an agitator
which is down for maintenance. However, the mass transfer capability of that particular unit will
be significantly degraded. With additional valving, the air could be directed towards the agitators
remaining in service if their dispersion capabilities are not exceeded by doing so.
If extra agitators are required for the lance system, there is the additional routine maintenance
associated with them. Spare parts inventory may also be affected, especially if it is found that all
agitators must be in service to maintain adequate oxidation.
Mass Transfer Capability
At a constant superficial gas velocity, the mass transfer capability of the lance typically exceeds
that of a sparge grid. This is especially the case when compared to grid designs which have not
been optimized for mass transfer. Figure 4 represents an EKATO study that highlights the mass
transfer capability of lance aeration, as compared to sparger aeration. This comparison was based
on the review of various sparge grid performance information found in the literature, EKATO
lance experience, and the retrofit of EKATO lance systems to poorly performing FGD sparge grid
installations.
Babcock & Wilcox experience has typically shown improved performance over the sparge grid
-------
representation indicated by figure 4. An analysis of discrepancies in the reported performance of
various sparge grid installations has not been made. It is surmised the apparent differences in the
performance of separate sparge grid installations could be related to the level of attention given to
all of the influencing factors.
Energy Efficiency
Energy efficiency is always an important consideration. Efficiency can be described in multiple
ways. A useful method for forced oxidation systems is to divide the oxygen up-take by the
system power usage. The oxygen up-take is the amount of oxygen chemically consumed by the
forced oxidation of sulfite to sulfate.
Some analyses will add the total agitator power to the compressed air power, for a total power
consumption value. This approach includes the portion of agitator power that can be considered
attributable to the solids suspension function.
On this basis, the resultant energy efficiency values suggest some sparge grids and lances can
have similar performance. When considering the total power consumption, the performance of a
well designed sparge grid appears in the range of 3-5 kg O2/kWh. The corresponding
performance of lance systems surfaces in the range of 2-4.5
The total agitator power input can be used to model and predict the mass transfer capabilities of
the lance. However, when making comparisons to sparge grid systems, dividing power
consumption into components can be useful.
The agitator energy input required due to the presence of lance air injection, over and above that
required for solids suspension only, can be considered a system air dispersion energy requirement.
This energy requirement is somewhat analogous to the pressure drop of the sparge grid, which
can be said to provide an air distribution or dispersion function. The analogy is not perfect, since
this energy component is typically higher than that required to achieve air dispersion similar to the
sparge grid. The additional energy goes towards improving the utilization of compressor power
beyond that of the sparge grid.
For either system, the energy required as a result of the air flow and hydrostatic head at the
injection point is considered an injection energy component.
When these components are divided by the mass quantity of oxygen uptake, a specific energy
value is obtained. The specific energy values are expressed in units of kilo-joules per gram of
oxygen uptake (kJ/g). Ei is the injection specific energy component, Ed is the dispersion energy
component, and Et is the sum of the two or the total specific energy of the oxidation process.
Et = Ei + Ed (3)
Table 1 highlights these energy components for two confidential FGD installation site
measurements and for a hypothetical design which makes a direct comparison between the sparge
grid and lance system while holding the basic absorber design constant.
-------
Table 1
Power & Energy Case Comparisons
Case Grid Site Lance Site Grid Lance
Reaction tank volume, m3 3172 650 735 735
Oxygen up-take, kg Oj/hr 1159 1607 1220 1220
Compressed air power at injection point, kW 326 212 317 243
Dispersion power consumption, kW 41 192 49 106
Injection specific energy (Ei), kJ/g 1.01 0.47 0.94 0.72
Dispersion specific energy (Ed), kJ/g 0.13 0.43 0.14 0.31
Total oxidation system specific energy (Et), kJ/g 1.14 0.90 1.08 1.03
With this approach, most cases analyzed have shown the lance to be more energy efficient than
the sparge grid. A few specific cases have predicted a properly designed sparge grid having an
energy efficiency exceeding the lance. These cases appear to occur at high sulfur loadings and
submergence depths greater than four meters.
Figure 5 provides a representation of the energy components versus submergence depth. The
curves represent the sparge grid and the vertical bars represent two specific lance system designs.
In terms of energy efficiency, the lance will typically outperform the sparge grid in the region
below an approximate 4m submergence depth. At increased submergence depths, the grid system
has the potential to exceed the energy efficiency of the lance system.
For the grid system, the dispersion energy component is always less than the injection energy
component. For the lance system, as the submergence depths decrease, it is possible for the
dispersion energy to exceed the injection energy component.
Summary
Large quantities of comparative data, allowing clear comparisons between lances and sparge
grids, is not readily available. Additionally, with the final FGD system design being governed by
many factors, a wide variety of system designs and operating parameters are the norm.
In that environment, universal formulae to accurately predict, estimate and compare the benefits
of one approach to the other become intricate and difficult to develop. However, certain
generalities can be acknowledged to better assess the advantages and disadvantages of each
approach.
Energy efficiencies are often comparable. In cases where there is a significant difference, the cost
of power could dictate the method chosen.
From a capital cost standpoint, high sulfur projects with greater allowable submergence depth can
move an evaluation towards a sparge grid. Low sulfur projects, with correspondingly low
oxidation air flows, appear ideally suited towards the lance system.
System energy efficiency and capital cost are not the only criteria to evaluate when integrating
oxidation air flow to an FGD absorber vessel. Proposal strategies and customer preferences can
-------
influence the path chosen. Factors such as slurry level and plant turndown requirements could
dictate which method is ultimately more cost effective or appropriate. The lance system has the
advantage of increased system design and operational flexibility. The plant load profile,
variations from actual versus design coal, and variations from actual versus predicted natural
oxidation, are prime examples of variables which the lance system can adjust to on a day to day
basis.
Table 2 provides a summary of the differences between design aspects of the sparge grid and
lance system relative to each other.
Table - 2
Summary of Grid and Lance System Relative Differences
Design Aspect Sparge Grid System Lance System
Air Flow Typically Increases Typically Decreases
Air Pressure Typically Decreases Typically Increases
Compressor Power Typically Increases Typically Decreases
Agitator Quantity Potential Decrease Potential Increase
Agitator Power Decreases Increases
Piping Material & Supports Increases Decreases
Electrical-MCC-I/O Potential Decrease Potential Increase
System Turndown Typically < 30% 100%
Operating Slurry Level Typically > 0.5T Typically ) 0.25T
In the end, either method can be designed to perform adequately. High sulfur, base loaded
projects may find the sparge grid to be the most economical approach. Other projects may find
the lance system most appropriate with the potential for improved energy efficiency across a
wider possible turndown range.
REFERENCES:
AstaritaG. Mass Transfer with Chemical Reaction. Elsevier, Amsterdam, 1966.
Perry R.H., C.H. Chilton. Chemical Engineers' Handbook. 5th ed., McGraw-Hill Inc.,
1973.
-------
EKATO. Handbook of Mixing Technology EKATO, Ruhr-und Mischtechnik GmbH. 1991
Villermaux J. "The Role of Energy Dissipation in Contacting and Mixing Devices"
Chemical Engineering Technology. 11 (1988) pp.276-287.
Loomis A.W. Compressed Air and Gas Data. Ingersoll-Rand, Woodcliff Lake,
N.J. 1982.
Stephenson Revis L., Harold E. Nixon. Centrifugal Compressor Engineering.
Hoffman. East Syracuse N.Y. 1974.
Geankoplis Christie J. Transport Processes and Unit Operations. Ohio State
University. Allyn and Bacon. Boston MA. 1978.
Smith J.M. Chemical Engineering Kinetics. 2nd ed. McGraw-Hill Inc. New York, NY
1970.
Oldshue James Y. Fluid Mixing Technology. McGraw-Hill Inc. New York, NY.
1983.
Tatterson GaryB. Fluid Mixing and Gas Dispersion in Agitated Tanks. McGraw-
Hill Inc. New York, NY. 1991.
Cheremisinoff N.P. Practical Fluid Mechanics for Engineers and Scientists.
Technomic Publishing. Lancaster, PA. 1990.
Bolz R.E. and George Tuve. Handbook of tables for Applied Engineering
Science. 2nd ed. CRC Press. Cleveland, OH. 1976.
-------
Sparger Headers
with Bubble Static
Header Supports
Figure 1
TYPICAL SPARGER GRID PLAN VIEW
Approximate minimum sparger grid system slurry level
Approximate mm gnd elevation or lance system slurry level
Figure 2
INFLUENCE OF SYSTEM DESIGN ON
PRACTICAL OPERATING SLURRY LEVELS
-------
Figure 3
AGITATOR / LANCE EQUIPMENT ASSEMBLY
-------
EKATOCOMPARISON OF LANCE AND SI'ARCER AERATION
1000 —
100
10
0.1
1
Vs [cm/s]
FIGURE 4
10
-------
1.60 —i
1.20 —
W
o
iS
a,
C/D
0.80 —
0.40 —
0.00
0.00
--Et
— tt Ei erid
Et Grid
Lance
+Ed
Xs
Lance
• Ei
--Ed
Ei = Injection Specific Energy
Ed = Dispersion Specific Energy
Et = Total Specific Energy
- — Ed Grid
2.00 4.00 6.00
Submergence Depth, m
8.00
Figure 5
Specific Energy versus Submergence Depth
-------
Limestone Performance in a Pilot-Scale Forced Oxidation Scrubber
Sharon Falcone Miller, Bruce G. Miller, and Alan W. Scaroni
Coal Utilization Center
The Pennsylvania State University
C211 Coal Utilization Laboratory
University Park, PA 16802
Introduction
With the promulgation of the 1990 Clean Air Act Amendments (CAAA), many utilities are required to
further reduce SO2 emissions. Title IV of the CAAA calls for a two-step program to reduce SO2
emissions nationwide by 10 million tons from 1980 levels by the year 2000 [1].
The dominant commercial technology for SO, capture is wet scrubbing with lime or limestone. New and
retrofitted flue gas desulfurization (FGD) applications in Pennsylvania and nearby states represent a
sizable market for the Pennsylvania stone industry. Manufacturers of wet flue gas desulfurization
(WFGD) systems typically require the use of high-purity limestones, i.e., > 95 wt.% CaCO3. Within
the state of Pennsylvania there are limited reserves of high-purity limestone, but abundant quantities of
intermediate-purity limestones. In addition, limestones of similar chemical composition may not perform
similarly in a scrubber due to differences in mineralogy and petrography. Consequently, a study was
conducted to provide a comprehensive data base of highly characterized Pennsylvania limestones, with
respect to their grindabilities and dissolution behavior, and to document their relative performance in a
pilot-scale WFGD facility. Of interest was to identify those limestone characteristics and operating
parameters that significantly affected, and allowed for the prediction of, limestone behavior and
performance.
In this study, the chemical, mineralogical and petrographic composition, grindability performance, i.e.,
Bond Work Index (BWI), dissolution behavior in a bench-scale reactor, and performance in a pilot-
scale, forced-oxidation WFGD test facility of twenty-five limestones, representing the major limestone
stratigraphic intervals in Pennsylvania, were measured. The specific operational and slurry parameters
that were studied were the liquid-to-gas (L/G) ratio, SO2 concentration, gas-liquid contact time in the
scrubber column, residence time in the reaction tank, solids loading, limestone grind, and pH of the
slurry.
Limestone Selection
Twenty-five limestones were selected on the basis of chemical composition, stratigraphic interval,
anticipated petrographic variations, and geologic history as inferred from the physiographic province.
General information including stratigraphic interval, supplier, and quarry identification for each
limestone is given elsewhere [2]. Limestone suppliers provided the approximate chemical compositions
of the various stratigraphic intervals that are, or can be made, commercially available. The limestones
contained 79 to 100 wt.% CaCO3 and 1 to 20 wt.% MgCO3. The carbonate composition is based on
assigning all calcium and magnesium to their respective carbonates. Three of the twenty five limestones
had CaCO3 contents of less than 80 wt.% and were included because they represent large limestone
reserves in Pennsylvania and are in close proximity to power plants affected by Phases I and n of the
CAAA.
Limestone Characterization
The characterization of each limestone consisted of elemental, mineralogical, petrographic, and lithologic
analyses. Table 1 lists the carbonate and inert contents and percent of spar-sized crystals of the
limestones along with their stratigraphic intervals. The elemental, mineralogical, petrographic, and
-------
Table 1. Stratigraphk Interval, Carbonate and Inert Contents (in Weight Percent).
and Volume Percent of Spar-Sized Crystals of Study Samples
Sample ID
SORB-21
SORB-::
SORB-23
SORB-24
SORB-25
soRB-26
SORB-27
SORB-2S
SORB-:9
SORB-30
SORB-31
SORB-32
SORB-33
SORB-34
SORB-35
SORB-36
SORB-37
SORB-3S
SORB-39
SORB-40
SORB-41
SORB-42
SORB-13
SORB-U
SORB-45
Stratigraphic
Interval
Annville
Ann\ ille
Annville
Ben\vood
Bossardville
Centre Hall
Gettysburg
Greenbnar
Jacksonbure
Kevser
KeNSer
Kinzers
Kinzers
Lovsburs/Coburn
Lovsbure/Coburn
Lovsbure/Ccburn
Lovsbure/Coburn
Nealmont
St. Paul Group
Tonolowav
Vallenline
\'alle\ View
Vanport
Vanport
Vanrjort
Calculated
CaCOi i MgCO,
85. S6 1 9. SI
96.93 i 2. 87
99. 7S i 1.49
54.09 i I9.5S
S3. 72 : 3. IS
SS 90 ! 3.22
S2.ll 1 5.19
79. OS i 312
7943 1 5.00
93.36 ; 2.36
S6.04 i 4 «s
8S.71 i 12.87
93.00 i S.24
9S.1S i 2.20
85. 6S ; 7.13
93 IS ! 5.54
9460 : 3.31
92 29 • 3.41
97.64 i 3.60
79.00 6.15
9S 71 1.00
96.57 2.30
93.71 1.3S
94.07 1 49
90. 87 1.53
Inert"
3.37
0.68
0.95
IS. 60
10.90
1 "7 ~l
14.40
1 1.90
13.90
4 29
7.S5
0.64
0.5S
1.31
6. S3
2.S7
A 5 ""
5.24
1.30
12.30
2.49
:.4S
3.22
2. S3
4.94
Measured
Total
CO,
58.0
59 '-,
5S.5
45.3
50.5
53. S
42 :
47.5
49.5
S 5 . ^
5:. 9
61.0
59.7
59. 1
54.5
57.9
56 4
56.1
59.0
50 4
58.1
57.6
56.5
56.9
55.1
Percent
Spar
SO
100
05
S
40
20
40
40
30
45
30
100
100
20
10
45
15
15
45
30
10
50
40
T S
1 ^
Quartz and Minor Clay
''Geologic Age and Stnuigraphic Interval Relationships arc Unclear
lithologic analyses are provided elsewhere [2]. In addition, the grindability performance and dissolution
behavior of the limestones are given elsewhere [2]. Semiquantitative measurements of CaCO, and
MgCO, by X-ray diffraction were also made. It is recognized that some limestones may contain high
magnesium calcium carbonate (MgCa(CO,)2)- No dolomites were included in the study.
Limestone Performance in a Pilot-Scale WFGD Test Facility
Description of the Pilot-Scale WFGD Test Facility and Operating Conditions
The design and operating conditions of the WFGD facility and the character of the limestone slurry were
based on the conditions typical of utility-scale FGD systems and their slurry feed. It was not possible to
directly duplicate all aspects of utility-scale designs and processes in the pilot-scale scrubber; however,
every attempt was made to take these differences into consideration when interpreting the data to provide
a relative index of limestone performance.
A schematic design of the WFGD test facility is shown in Figure 1. The unit is connected to a 400,000
Btu/hr down-fired combustor (DFC) capable of firing natural gas and solid and liquid fuels. The DFC
provided the heated flue gas to the scrubber.
A single BETE Fog spiral nozzle (ST-12FCN), operated at 20 psig and located in the center of the
scrubber, was used for slurry injection. Because spiral nozzles with spray angles of less than 90° are not
-------
Gas Sample Port
Materials of Construction:
316 Stainless Steel
304 Stainless Steel
Water Flush for Demister
Demister
Limestone
Slurry
Atomization / Inspection
Port 4" Half-Coupling
Conditioned Flue Gas Outlet
Continuous Mixing
Inlet gas C
analyzer
sample port]
I <3
Rubber expansion joint
used on inlet of reactor
Density output
ft capabilities
Continuous
Mixing
Flue Gas from
Down Fired Combustor
, , 'HTT <
L Reagent Reorc Loop
Figure 1. SCHEMATIC DIAGRAM OF THE WFGD PILOT-SCALE TEST
FACILITY
-------
available at this scale, trays were installed at two levels in the upper portion of the scrubber to reduce
sheeting of the slurry on the walls.
During a test run, samples were taken from the reaction tank every two hours and analyzed for specific
chemical and physical characteristics. The specific analyses performed on the limestone, the solid
residue, and the liquid portion of the samples collected from the reaction tank are shown in Table 2.
The unit can be run in batch or continuous mode. In the continuous mode, makeup slurry is added and
removed from the tank. The continuous mode of operation is appropriate when the scrubber has reached
steady state. In this study, the tests were conducted in a batch mode and steady state or equilibrium
conditions were not reached during a test period.
The baseline operating conditions and slurry characteristics are given in Table 3. The DFC was fired
with natural gas and doped with SO, during the tests. During the baseline tests, no additives were used
to modify the pH of the slurry. A series of tests was also conducted in which selective operating
parameters and slurry characteristics were varied to determine their effect on limestone performance in
the scrubber. The operational parameters and slurry characteristics that were varied and their ranges are
given in Table 3.
Tesf Results and Discussion
The objective of the pilot-scale testing was to generate a data base of limestone behavior under WFGD
conditions in a pilot-scale facility to be used as a tool for determining the relative performance of low- to
high-purity limestones and relate measurable limestone composition or petrographic features to
performance. Limestone performance is presented in three ways: the extent and rate of sulfur capture as
a function of time and the total carbonate-to-sulfur molar ratio at the completion of each test. Forty-five
tests were conducted in the WFGD Test Facility, twenty-five baseline tests were conducted to determine
the relative performance of each limestone and fifteen tests were conducted using high- and intermediate-
purity limestones, SORBs 22 (Annville) and 32 (Kinzers), respectively, in which changes were made in
the test conditions and slurry preparation.
Percent of Sulfur Capture as a Function of Time. The percent of sulfur removed from
the flue gas decreased as a function of time for all twenty-five limestones. Because the tests ranged in
duration from nine to eleven hours, the percent sulfur capture at nine hours is used for comparison
purposes. In addition, the change in percent sulfur capture from the beginning to the end of the test was
used to compare the relative performance of the limestones.
A comparison of the percent sulfur capture as a function of time for selected low- intermediate-, and
high-purity limestones is given in Figure 2. For the purposes of this study, limestones with CaCO3
contents of < 80, 80-95, and > 95 wt.% were classified as low, intermediate, and high purity. SORB
26, an intermediate-purity limestone, maintained the highest level of sulfur removal at the end of the
baseline test compared to the other limestones tested. Interestingly, SORBs 22 and 41, both high-purity
limestones, performed differently in the pilot-scale tests with SORB 22 exhibiting more sulfur capture
than SORB 41 after nine hours.
Statistical analysis of the entire sample population (baseline) produced a correlation coefficient between
the percent sulfur capture at nine hours and the CaCO3 content of 0.552. Moreover, the relative
performance of the limestones did not correlate with their relative dissolution rates measured in bench-
scale tests based on a reactor design and test procedure developed by Radian Corporation [2]. Variations
in mineralogy (i.e., clays, quartz, calcite, high-magnesium calcite, and dolomite) were also unable to
explain the differences in dissolution rates [2], therefore, the petrographic features of the limestones
were examined for a possible relationship with dissolution rate and pilot-scale performance. The
primary difference between limestones observed during petrographic analysis was crystal size.
Limestones contain calcite crystals classified as sparite (>4 |im) or micrite (<4 um). A particular
limestone may contain predominately one crystal size or a combination of the two. The volume percent
of spar-sized crystals for each limestone is given in Table 1.
-------
Table 2. Analyses Conducted on Limestone, Residue, and Filtrate Samples
Limestone
S
Ca
Mg
CO3
Particle size distribution
(PSD)
Solids content
Petrography
I Residue
i s
ICa
jMg
i'cos
I PSD
I Solids content (Comb, sample)
! Scanning electron microscopy
! (SEM)
I X-ray diffraction
Filtrate
Anions: SO3", SO4~
Ca-H-
Mg++
-
—
—
Anions: Oh, NO2-, NO3-, PO4"
SEM (Crystal Morphology)
Table 3. Operating Conditions and Slurry Characteristics for Baseline and Modified
Test Runs
Operating Conditions
Natural Gas Firing Rate (1,000 Btu/hr)
Flue Gas Temp, at Scrubber Inlet (°F)
Liquid-to-Gas Ratio (L/G)
SO2 Concentration in Flue Gas (ppm)
Gas - Liquid Contact Time (s)
Volume of Slurry Charge (gal)
Temp, of Slurry in Reaction Tank (°F)
Slurry Injection Pressure (psig)
Test Duration (hrs)
Air Injected into Reaction Tank
(moles O, : mole CaSO,)
Slurry Characteristics
Particle Size Distribution
(% minus 325 mesh)
Solids Loading (Weight %)
PH
Baseline Tests
380-400
280-300
100
2,000
5
100-110
135-140
20
9-11
2
90-98
3
5.6-7.5
Modified Tests
380-400
280-300
60-138
1,000-2,500
1.5-5
60-110
135-140
20
2-11
2
72-99
1-5
4.5-7.5
-------
100'
95-
90-
3
Q.
85-
80-
75
D SORB 24 (54.1)
O SORB 26 (88.9)
....O-— SORB 22 (96.9)
& SORB 41 (98.7)
10
11
12
Figure 2.
012345678
Time (hrs.)
PERCENT SULFUR CAPTURE AS A FUNCTION OF TIME FOR
SELECTED LIMESTONES
(Benwood (24), Centre Hall (26), Annville (22), and Vallentine
(41) Limestones; Weight Percent CaCO3 given in Parenthesis)
Figures 3 A and B are photomicrographs of two limestones having similar mineralogical compositions
but different textural character. SORB 22 is composed of 97 wt.% calcite and consists entirely of spar-
sized crystals (Figure 3A) whereas SORB 41, which is composed of 99 wt.% calcite, has a micritic
texture with a few spar-sized crystals (Figure 3B). It was found that the majority of the lower-purity
limestones had micritic textures with occasional spar-sized crystals in localized areas of neomorphism or
healed fractures.
Although SORB 22 and SORB 41 contain similar amounts of calcite, their dissolution rates differ by an
order of magnitude (SORB 22 = 2xlO'10 mol/cnrs and SORB 41 = 5.58x10'" mol/cm2s). The
differences in the fabrics between the limestones, especially in the higher purity limestones, suggest that
the texture, i.e., the grain/crystal size of the calcite may affect the rate of dissolution. Multiple regression
analyses of the results from the bench-scale dissolution tests showed that 86.5% of the variation in the
log of the dissolution rate was explained by the percent spar and the weight percent carbonate. An
equation that predicted within ±2.5% the measured dissolution rate as a function of the weight percent of
carbonate and the volume percent spar was:
Log Dissolution Rate = 0.065X,+0.006X-,-14.166
(D
where X, is the weight percent of carbonate and X, is the volume percent spar in the limestone.
If the rate of calcite/carbonate dissolution were the rate determining step in the removal of sulfur dioxide
from flue gas in the WFGD system, then among limestones with high carbonate contents, those having a
significant percent of spar-sized crystals should outperform other htgh-purity limestones. However,
while textural character may explain the difference in the performance between high-purity limestones,
i.e., SORBs 22 and 41, it does not explain the high performance of intermediate-purity SORB 26 (88.9
wt.% CaCO3, 2x10"" mol/cnrs) which contained only 20% spar. In none of the baseline tests was the
carbonate totally dissolved. The reduction in sulfur capture over nine hours ranged from 5 to 14% and
averaged 8.7%. There was no clear relationship between carbonate content and textural features and the
percent sulfur capture as a function of time. Therefore, other parameters in addition to calcium carbonate
content, texture, and dissolution rate, as it is currently defined, must be considered in order to predict
-------
limestone performance in the pilot-scale scrubber. Work is underway to refine the manner in which the
dissolution rate is expressed to better reflect the range in composition of the limestones used in the study.
(B)
Figure 3. TRANSMITTED LIGHT PHOTOMICROGRAPH OF
(A) THE ANNVILLE LIMESTONE (SORB 22) AND
(B) THE VALLENTINE LIMESTONE (SORB 41)
-------
Rate of Sulfur Capture as a Function of Time. In general, the rate of sulfur removal
(Ib/hr) decreased with time. The rate of sulfur removal ranged from 0.58 to 0.72 Ib/hr and averaged
0.65 Ib/hr. Specific limestones had statistically significant higher or lower rates of sulfur removal but
there was no clear relationship between the carbonate content of a limestone and the rate of sulfur
removal at a given time during the pilot-scale tests. Furthermore, none of the limestones having the top
four rates of sulfur capture were ranked among the top ten in dissolution rates.
In Figure 4. the rate of sulfur capture as a function of time is shown for selected low- to high-purity
limestones. The rate curves for SORBs 24 (low-purity limestone) and 41 (a high-purity limestone) are
very similar. SORB 26 had the highest rate of sulfur capture at the completion of the test compared to
the other limestones tested.
Carbonate Utilization. Limestone performance in a WFGD system can also be described by the
calcium-to-sulfur molar ratio. Theoretical!}' one mole of calcium is required to capture one mole of
sulfur to form CaSO- or CaSO4. Typically, utility-scale forced oxidation WFGD systems operate at a
calcium-to-sulfur molar ratio of 1.1. The solid residue at the end of each test run was analyzed for total
carbonate content.
The carbonate-to-sulfur molar ratios are given graphically in Figure 5. All of the limestones, except for
SORBs 23 and 28, had a carbonate-to-sulfur molar ratio between 1.0 and 1.3 and compare well with
typical carbonate-to-sulfur molar ratios of efficient utility-scale WFGD systems. SORBs 23 and 28 had
carbonate-to-sulfur molar ratios of 0.96 and 0.90. respectively. This is probably due to experimental
error.
Solution Chemistry. A significant reduction in the extent of sulfur capture and rate of sulfur
capture was observed for limestones containing significant amounts of soluble magnesium, e.g., SORB
24 (Benwood; 54.1 wt.% CaCO,,, 19.6 wt.% MgCO,)- This is attributed to the slow rate of dissolution
of calcium due to inhibition by magnesium in solution. At the completion of the test, 91 % of the
magnesium had been removed from the solid phase. The level of sulfite was still increasing at the
completion of the test suggesting that dissolution of SO, was continuing; however, the calcium in
solution was depleted. The concentration of SO4": in solution steadily increased from 1.080 ppm (two
hours into the test) and peaked at 2,630 ppm at the completion of the test. The saturation concentration
of CaS04 in water at 100°C is 1,620 ppm.
The Vanport limestone (SORB 44) has a calcium carbonate content of 94.1 wt.% and a magnesium
carbonate content of 1.5 wt.%. The concentrations of calcium and sulfite are almost equal nine hours
into the test. They remain relatively unchanged for the last 1.3 hours of the test. The magnesium
concentration increased at an even rate from the beginning of the test. It should be noted that the
maximum level of magnesium observed during this test is one-fourth that measured in the Benwood test
described above. The percent sulfur capture gradually decreased during the test (2.9% during the latter
seven hours of the test). The steady removal of sulfur from the system is reflected in the steady sulfite
concentration. The concentration of SO4"~ peaked at 1,840 ppm 3.5 hours into the test and gradually
declined to 1,500 ppm. The rise then decline in the calcium concentration corresponds to an increase
then decline in the rate of sulfur removal at six hours.
In general, limestones that maintained fairly constant and equitable sulfite and calcium concentrations and
lower magnesium (0 to 100 ppm) concentration in solution maintained higher levels of percent sulfur
capture and rates of sulfur removal.
Effect of Operating Parameters and Slurry Conditions on Limestone
Performance. Tests were conducted using high- and intermediate-purity limestones, SORBs 22 and
32. respectively, in which operating parameters and slurry conditions were varied to determine their
effect on performance.
The rate controlling step in the removal of SO, from the flue gas is the dissolution of the limestone,
dissolution of the SO, in the slurry, or a combination of the two. The operational parameters that affect
-------
Q.
n
O
0.80
0.75 -
0.70 -
0.65 H
0.60
;~0-0-
o
O-.
'•a
D-O..
O
%*••<>•<
'O.
10 11
12
Time (hrs.)
Figure 4. SULFUR CAPTURE AS A FUNCTION OF TIME FOR SELECTED
LIMESTONES
(Benwood (24), Centre Hall (26), Annville (22), and Vallentine
(41) Limestones; Weight Percent CaCO3 given in Parenthesis)
I
£ -a
B «,
£ r
ra o
o ij
8 ™
.2 "5
0 $
° o
o o
1 °
0
n7 -
25 39
§ • 33 I "
31 B 37
2 T " " " • 36™» -V. -fliJ
_ _ 27 B 39 _ »
24 • ^ 34 " 38 • •
• 26 " 35 " 40 • 43
• 28
•
Figure 5. RATIO OF MOLES OF TOTAL CARBONATE DISSOLVED TO
TOTAL MOLES OF SULFUR DIOXIDE CAPTURED DURING
BASELINE TESTS
(Molar Ratio of 1.1 Typical of Efficient Utility-Scale WFGD
Systems - Sorbent Numbers above Boxes)
the dissolution of the SO2 are the liquid-to-gas ratio, inlet SO2 concentration, and contact time in the
scrubber column while the operating parameters and slurry conditions that affect the limestone
dissolution rate are pH, limestone particle size, residence time of the slurry in the reaction tank, and
slurry solids loading. The effect of varying each of these parameters on limestone performance is
discussed in the following sections.
-------
Effect of Liquid-to-Gas Ratio on the Percent and Rate of Sulfur Capture.
The L/G ratio in the scrubber column affects the total droplet/liquid surface area exposed to the flue gas.
The dissolution of SO, occurs at the droplet surface. Increasing the surface area increases the SO,
dissolution rate. In a utility-scale scrubber, the L/G ratio is generally around 0.096 gpm/acfm
(saturated). In the baseline tests, a L/G ratio of 0.10 gpm/acfm was used. A series of tests was
conducted using high (SORB 22; Annville; 96.9 wt.% CaCO,) and intermediate-purity (SORB 32;
Kinzers; 88.7 wt.%"CaCO3) limestones in which the L/G ratio was varied from 0.06 to 0.14 gpm/acfm
by increasing the slurry flow rate (see Figure 6).
A B C D
•C- 0.70 -
.c
"55
=,
u
3
'S.
Q 0.60 -
>_
•2
3
in
^^
o
| 0.50 -
0.40 -
V
""LJ^O- —
• D""---.~~~
_ H
^ B
•
•Q^
D* *r~v -Q.*-,
~""^y~~
1 •
^.
"-"-•----.-_- n..
• • -- y
H
•
rf *n
^-g;;;;;"^1-...
-o.
• .
• 0
--D-
— o. D
-o
i i i i I i i I i
- 90
g
o
3
- 80 1
O
^
3
(fi
- 70
- 60
01 23456789 10
A- L/G changed from 0.08 to 0.06 gpm/acfm.Tlme*hrs-> Rate % Sulfur Capture
B- L/G changed from 0.06 to 0.10 gpm/acfm. m Modified Test • Mod fied Test
C- L/G changed from 0.10 to 0.12 gpm/acfm. _ ,. ,. „,.,..
D- L/G changed from 0.1 2 to 0.1 4 gpm/acfm. — O" Baseline Test -O-~- Baseline Test
Figure 6. EFFECT OF L/G RATIO ON THE EXTENT AND RATE OF SULFUR
CAPTURE BY THE KINZERS LIMESTONE (SORB 32)
An increase in the L/G ratio resulted in an increase in the percent of sulfur capture. The rate of sulfur
capture is directly related to the L/G during the first six hours of each test. After six hours, the rate of
sulfur capture decreased regardless of the L/G ratio. For SORB 32, decreasing the L/G ratio from 0.08
to 0.06 gpm/acfm resulted in a 12% decrease in the percent sulfur removed and a 21% decrease in the
rate of sulfur removal. Increasing the L/G ratio from 0.06 to 0.10 gpm/acfm resulted in a 24% increase
in the percent sulfur capture and a 15% increase in the rate of sulfur removal.
SORB 22 showed a 13% increase in the percent sulfur capture and a 5.6% increase in the rate of sulfur
capture with an increase in the L/G from 0.06 to 0.08 gpm/acfm. Increasing the L/G from 0.08 to 0.10
gpm/acfm resulted in only a 4% increase in the percent of sulfur capture. The performance of the
intermediate-purity limestone was more sensitive to the variations in L/G.
Effect of Inlet Sulfur Dioxide Concentration on Limestone Performance.
During the baseline tests, the concentration of SO, was maintained at 2,000 ppm. In this series of tests,
the inlet SO, concentration was varied from 1,000"to 2,500 ppm using the Kinzers and Annville
-------
limestones (SORBs 32 and 22, respectively). Increasing the concentration of SO, increased the rate of
sulfur removal but decreased the percent of sulfur removed as a function of time In both tests. Figure 7
illustrates the results of testing with SORB 32. The pH decreased slightly with time, then more
significantly when the concentration of SO2 was increased from 1,000 to 1,500 ppm. This is a reflection
of the effect of increased SO, concentration on the rate of dissolution of SO,. With the increased rate of
dissolution of SO2, there is a corresponding increase in the pH decline due to the increased IT
concentration in the slurry solution. The concentration of dissolved calcium increased from 145 to 608
ppm upon injection of SO2, then to 718 ppm when the SO, concentration was increased to 1,500 ppm,
then decreased to 618 ppm as the SO, concentration was increased from 1,500 to 2,500 ppm. The data
suggest that the calcium concentration in solution is fairly constant, therefore the increase in the sulfur
capture is due to the increased rate of dissolution of SO,.
1.00
•C- 0.80-
^
0)
¥
3
Q.
5 0.60-
.»
'a
CO
i^
0
a>
« 0.40-
^-'o.....
q
\%
VQ--.
"O- —
•
i i i i
"•-•n..
• D»""n---ji...n.
• ff-—
• • • •
o — ~ ^
^O — _^
-0— -
1 1 1 1 1 I 1 1
V
••o- a- o__
«
•
•
-o-o
-------
0.9
0.8-
2 0.7 -
•5. 0.6 -{
to
O
1 0.5 H
a 0.4-
0.3 -
0.2
0-0-.,
°o
100
- 90
- 80
- 70 £
0
3
- 60 g-
- 40
- 30
20
Figure 8.
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.51010.51111.5
Time (hrs.)
Bate (%) Sulfur Capture
• Modified Test ----^---- Modified Test
D Baseline Test —o— Baseline Test
EFFECT OF GAS-LIQUID CONTACT TIME IN THE SCRUBBER ON
THE EXTENT AND RATE OF SULFUR CAPTURE BY THE KINZER
LIMESTONE (SORB 32)
Effect of pH on Limestone Performance. In general, during the baseline tests, the
pH and percent sulfur capture decreased with time. In some tests, there was a linear relationship
between pH and the percent sulfur capture while in other tests the relationship between pH and sulfur
capture was less clear [2]. Test runs were conducted to determine the extent to which pH affects the
percent sulfur removal and rate of sulfur removal as a function of time. The pH was modified by
additions of adipic acid, i.e., hexanedioic acid (HO,C(CH,)4CO,H), to the reaction tank to achieve a
stepwise reduction in pH.
An example of the results is shown in Figure 9 for SORB 32. The rate of sulfur removal was increased
by the reduction in pH with the addition of adipic acid. At nine hours into a test, the Kinzers limestone
slurry with pH modification had a sulfur capture rate of 16.8% (pH 4.39) greater than for the baseline
test case (pH 5.8). This is due to the increased rate of limestone dissolution.
The shape of the rate curves are irregular (steplike) and mimic the pH curves for the pH modified tests.
After each addition, the pH would stabilize and then quickly increase. The increased concentration of H+
in solution results in an increase in the dissolution rate of the carbonate portion of the limestone. The pH
modified test results are consistent with research conducted by Rochelle et al. who observed that
additions of adipic acid resulted in increased rates of limestone dissolution [3]. Although an increase in
the concentration of H+ ions in solution inhibits the absorption of SO,, it appears that the increased rate
of limestone dissolution outweighed the reduction in SO, absorption." Similar tests with the higher-
purity SORB 22 (Annville) limestone showed that it was"less sensitive to the pH reduction than the
intermediate-purity Kinzers limestone.
-------
0.80'
0.70-
8
10 11
12
01 234567
Time (hrs.)
Rate pH
D pH Modified Test ^....... pH Modified Test
O- — Baseline & Baseline
Figure 9. EFFECT OF pH MODIFICATION ON THE RATE OF SULFUR
CAPTURE BY THE KINZER LIMESTONE (SORB 32)
Effect of Limestone Particle Size on Limestone Performance. A slurry was
prepared using a coarser grind, i.e., 72% minus 325 mesh, of the Annville limestone (SORB 22). In
Figure 10, the performance of the coarser grind slurry is compared to that of the finer grind slurry, i.e.,
96% minus 325 mesh, used in the baseline test. The percent sulfur capture and rate of sulfur capture of
the coarser grind limestone were less than for the finer grind limestone. This agrees with studies
conducted by Thompson and Burke of Radian Corporation, who observed that finer grind limestone
performed better than coarser-ground limestone over a wider range of process conditions [4]. In their
study, a fine-ground limestone was defined as having 85-90 wt.% of particles minus 325 mesh.
Coarse-ground limestone was defined as having 50-70 wt.% of particles minus 325 mesh. This effect is
caused by the increased rate of limestone dissolution with reduction in particle size.
Effect of Residence Time of Slurry in the Reaction Tank on Limestone
Performance. The volume of the limestone slurry was also reduced during the coarser grind
test run. The percent sulfur capture and rate of sulfur capture were significantly reduced when the
volume of slurry was reduced from 100 to 60 gallons (see Figure 10). Reducing the volume of the
slurry in the system effectively reduces the residence time of the slurry in the reaction tank. In the
studies of Thompson and Burke, changing in the volume of the reaction tank did not affect limestone
performance. This was not the case in the pilot-scale WFGD Test Facility, where reducing the residence
time in the reaction tank decreased the extent of limestone dissolution.
Effect of Solids Loading on Limestone Performance. Two test runs were
conducted with the Annville limestone (SORB 22) in which the slurry solids of 1 and 5 wt.% were used
(see Figure 11). The rate of sulfur removal and percent of sulfur removed by the limestone slurry were
lower in the test conducted with the 1 wt.% solids slurry. This again illustrates the importance of
limestone dissolution.
-------
0.40
0.5 1 1.5 2 2.5
3 3.5 4 4.5 5 5.5 6 6.5
Time (hrs.)
Rate
A- Reaction tank volume changed from - m - Coarser Grind
1 00 gal. to 60 gal. . . . a. . . Finer Grjnc|-
Baseline Test
7.5 8
% Sulfur Capture
...... * ...... Coarser Grind
— O — Finer Grind-
Baseline Test
Figure 10. EFFECT OF REACTION TANK VOLUME AND LIMESTONE GRIND
ON THE EXTENT AND RATE OF SULFUR CAPTURE BY THE
ANNVILLE LIMESTONE (SORB 22)
0.70 - »..
£ 0.50
Time (Mrs.)
1 1% Solids
---D--- 5% Solids
% Sulfur Capture
» 1% Solids
—O— 5% solids
Figure 11. EFFECT OF SOLIDS LOADING ON THE EXTENT AND RATE OF
SULFUR CAPTURE BY THE ANNVILLE LIMESTONE (SORB 22)
-------
Concluding Remarks
The objectives of the study were to develop a comprehensive data base of highly characterized
limestones, with respect to their grindabilities, dissolution behavior, and relative performance in a pilot-
scale WFGD facility and relate measurable limestone composition or petrographic features to their
performance. Of interest was to identify those limestone characteristics and operating parameters that
significantly affected limestone behavior and performance.
Twenty-five Pennsylvania limestones of high, intermediate, and low purity were studied. From a
technical perspective, the study suggests that limestones containing as low as 82 wt.% calcium carbonate
may be suitable as reagents. Relative performance was not predicted from calcium carbonate content.
Two limestones having almost identical chemical composition performed quite differently in the WFGD
Test Facility. Textural variations between high-purity limestones were shown to be significant in
affecting dissolution rates and pilot-scale performance.
Although textural variations between high-purity limestones explained differences in sulfur capture
performance, no measurable limestone component or petrographic feature successfully explained
differences in performance between high-, intermediate-, and low-purity limestones. Performance in the
WFGD Test Facility was determined by both the rate of SO, absorption in the scrubber column and the
rate of limestone dissolution in the slurry tank. Changing those operating parameters or limestone
properties which increased either or both rates, increased"scrubber performance. The operational
parameters that affected the dissolution of the SO, were the liquid-to-gas ratio, inlet SO? concentration,
and contact time in the scrubber column while the operating parameters and slurry conditions that
affected the limestone dissolution rate were pH, limestone particle size, residence time of the slurry in the
reaction tank, and slurry solids loading. Work is continuing at Penn State to predict pilot-scale
performance from easily measured chemical and/or physical properties.
Acknowledgments
Financial support for the project was provided by the Pennsylvania Energy Development Authority and
the Pennsylvania Electric Energy Research Council. The limestone suppliers are acknowledge for their
contribution of material to the project. The faculty and staff of The Energy Institute are acknowledged
for their contributions, particularly Joel L. Morrison for assisting in selection and collection of the
limestone, Frank J. DiGnazio for performing the petrography of the limestones, and Roger L. Poe and
Sarma V. Pisupati for design and fabrication of the WFGD Test Facility. Susan Brantley is
acknowledged for her contribution to the dissolution study. The Electric Power Research Institute
(EPRI) Environmental Control Technology Center (formally the EPRI High Sulfur Test Center) is
acknowledged for contributing materials and design recommendations for the construction of the WFGD
Test Facility.
Literature Cited
1. J. Makansi, "Clean Air Act Amendments: The Engineering Response." Power, Vol. 135, No. 6,
1991, p. 11.
2. S.F. Miller, F.J. DiGnazio, J.M. McConnie, A.D. Stempkowski, R.S. Wasco, and A.W. Scaroni,
Evaluation of Pennsylvania Limestone Products for Feedstock for Wet Flue Gas Desulfurization
Systems. University Park, PA: The Pennsylvania State University, July 1996. Final Report for the
Pennsylvania Energy Development Authority, Grant Agreement #9103-4035.
3. G.T. Rochelle, "Limestone Dissolution in Flue Gas Desulfurization Processes," prepared for
Industrial Environmental Research Lab, Research Triangle Park, NC, Texas University at Austin,
Dept. of Chemical Engineering, U.S. Department of Commerce, NTIS, August 1983.
4. P.A. Thompson and J.M. Burke, EPRI High Sulfur Test Center Report: Factors in Limestone
Reagent Selection. Final Report TR-100137, Radian Corporation, Research Project 1877-1, 1991.
-------
Impact of Limestone Grind Size on WFGD Performance
Charlotte Brogren
ABB Corporate Reserch
Department R
S-721 78 Vasteras
Sweden
and
Jonas S. Klingspor
ABB Environmental Systems
1400 Centerpoint Boulevard
Knoxville TN 37932
USA
ABSTRACT
Limestone based wet flue gas desulfurization, has for about 30 years been the most
cost effective and reliable method for control of sulfur dioxide emissions from coal- and
oil-fired power plants. Considerable experience has been accumulated from full-scale
installations since the installations of the first system in 1968. One of the most
important process parameters for the performance of a WFGD plant is the limestone
grind size. To better understand the impact of the limestone grind size and thereby be
able to find optimal design criteria's, ABB has investigated the impact of limestone
grind size on various parameters including S02 removal, gypsum purity and
hydrocyclone operation. The study shows that fine grind limestone can lead to
decreased size of the reaction tank, increased SOa performance, as well as increased
gypsum purity. A model based on basic mass transfer theories has been developed to
predict both limestone utilization as well as the particle size of the unreacted limestone
in the slurry as a function of various parameters. Very good agreement has been
obtained between the model and experimental results. Further, the model has been
used to examine the potential for gypsum and limestone separation in a hydrocyclone.
Page 1
-------
INTRODUCTION
ABB's wet FGD system is based on the conventional open spray tower technology
which has been the market leader for over two decades. Limestone based wet flue gas
desulfurization (FGD) has been the dominating control technology since the
introduction of the Clean Air Act in the US and is projected to be the preferred FGD
technology globally for the foreseeable future. Recently ABB introduced its new wet
FGD system which represents a significant advance in the state-of-the-art of limestone
wet scrubbing, and can be offered at substantially reduced capital, operating and
maintenance costs (1-3).
Drawing from 30,000 MW of world-wide wet FGD experience, ABB has incorporated
several innovations in the new system designed to reduce the overall cost of S02
compliance. Collectively, these improvements are referred to as LS-2. The LS-2 system
is being demonstrated at Ohio Edison's Niles Plant at the 130 MWe level and this
turnkey installation was designed and erected in a 20 month period.
The LS-2 project at the Niles Plant includes a number of innovative process
improvements.
• Spraytower The spray tower which has the ability to run at velocities as high
as 18ft/sec, features a compact spray zone with ABB's patented co-current and
countercurrent staggered nozzle arrangement and wall rings. The high gas
velocity and the staggered nozzle arrangement leads to effective mass transfer
between the gas and the liquid and thereby to high removal efficiency.
. Grinding system The reagent system employs an ABB Raymond roller mill
which is based on a completely dry grinding circuit. This limestone grinding
system is less costly, both to construct and operate, yet produces a significantly
finer grind.
• Reaction tank The LS-2 system is designed to use fine grind limestone which
allows the use of a significantly smaller size reaction tank.
. Dewatering The primary dewatering system features fully integrated high
efficiency hydrocyclones followed by centrifuges or belt filters for secondary
and final dewatering. By using fine grind limestone, fractionation of limestone
and gypsum in hydrocyclones becomes feasible. Further, the LS-2 grinding
system inherently produce an inert fraction with a very fine particle size
distribution. Hence, the fine grind system also improves the fractionation of the
inert fraction in the limestone. Therefore, the LS-2 system lends itself to
operating on relatively poor purity limestone's while producing high quality
wallboard grade gypsum.
• Mist Eliminator Before exiting the absober, the flue gas pass through the ABB
high velocity mist eliminator. The mist eliminator system consists of a bulk
entrainment seperator followed by two stages of a two pass mist eliminator.
The mist eliminator is capable of operating at velocities well above 30 ft/sec.
Page 2
-------
• Reheat Due to the compact spray tower, the LS-2 absorber makes use of a
horizontal shaft rotary Ljungstrom gas-to-gas reheat system.
Cost reductions
The reduction in capital cost of the Niles FGD system has been achieved mainly by a
significant reduction in the size of the absorber and integration of critical subsystems
such as limestone grinding and gypsum dewatering. Cost reductions in the additive
preparation area have been achieved through the use of a dry grinding system based
on ABB roller mill technology. Also, the primary dewatering step has been optimized
and from a process standpoint integrated with the absorber system. Simplification of
the design has led to substantial cycle time reductions in engineering and construction.
The Niles demonstration plant was engineered and erected on a 20 month schedule,
far shorter than the current industry standard. A picture of the installation at Niles, OH
is shown in Figure 1.
Figure 1
3D View of the LS-2 Installation at Niles, OH
FINE GRIND LIMESTONE
The use of fine grind limestone within the LS-2 design has an impact on several of the
critical process parameters, e.g.:
• size of reaction tank
• gypsum purity
• SO2 removal efficiency
PageS
-------
Limestone preparation
The additive preparation system features
an ABB Raymond roller mill, as shown in
Figure 2. The mill system accepts a
limestone feed stock less than 40 mm (1.6
in.) Untreated flue gas is used to dry and
convey the limestone during mill operation.
The flue gas leaving the milling system is
returned to the absorber for processing.
The limestone preparation and handling
system is completely dry and includes a
wetting system prior to injection into the
reaction tank. The particle size of the
ground limestone is controlled by the
dynamic classifier included in ABB
Raymond roller mill. The speed (rpm) of
the classifier is inversely related to particle
size higher speeds produce smaller
particles exiting the classifier and mill.
This is accomplished by recirculating
larger particles back to the mill to be
reground to the maximum cut size. F'9ure 2 LS-2 Roller Mi"
Recirculation of limestone therefore increases as classifier speed is raised to produce
smaller mean particle size distributions. Recirculation also increases the pressure drop
across the mill at a constant conveying air rate. The limestone grind will typically be
99.5% less than 44 |im (325 mesh) as shown in Figure 3. The data in Figure 3 were
determined with the use of an on-site Malvern Mastersizer Micro laser-based particle
size analyzer.
PSD of LS-2 limestone
10
particle size (pm)
Figure 3 Typical Limestone Size Distribution
Page 4
-------
Reaction tank
The reaction tank is an integral part of the absorber and provides residence time to
complete a number of critical chemical reactions including:
limestone dissolution:
CaCO3 + H* -» Ca2+ + HCO3' (1)
sulfite oxidation:
S032' + y202 -» S042' (2)
gypsum precipitation:
Ca2+ + S042" +2 H2O -> CaS04*2H2O (3)
The size of the reaction tank at Niles has been reduced by about 60 percent compared
with conventional sizing criteria. ABB has determined that for most cases, the limiting
reaction step is the limestone dissolution, not sulfite oxidation or gypsum precipitation.
The gypsum precipitation reaction is sufficiently fast not to affect the gypsum relative
saturation. Also, the oxidation reaction is much faster than the limestone dissolution
rate and is not affected by a reduced reaction tank size.
With typical limestone grinds, a significant reduction in the reaction tank size will
dramatically lower the liquid phase alkalinity and increase the solid phase alkalinity. In
order to reduce the solid phase alkalinity, the operating pH must be lowered which
would result in a reduction in SO2 removal efficiency. In order to maintain the S02
removal effeciency, i.e maintain the pH and the liquid phase alkalinity at reduced
residence time, the only option is to increase the limestone surface area available for
dissolution. The LS-2 system overcomes the traditional limitation in reaction tank size
by using an ultra-fine limestone grind.
Gypsum purity
The optimal operating conditions of the reaction tank is a compromise between the two
opposite requirements of a reactive liquid for S02 absorption and the production of a
gypsum product with high purity. As the LS-2 concept is based on a compact absorber
with a compact reaction tank the difference in optimal operating condition between
these two requirements will be more pronounced than in an absorber with a larger
reaction tank. However, the LS-2 concept overcomes this problem by using fine grind
limestone produced by an ABB Raymond roller mill and by using hydrocyclones for
gypsum purification. The potential for purification is equal to the difference in particle
size distribution between the produced gypsum and the ultra fine limestone produced in
the ABB Raymond roller mill.
Page 5
-------
The dewatering system consists of a battery of
6 inch hydrocyclones for primary dewatering
and centrifuges for secondary dewatering. The
hydrocyclones are tightly integrated with the
absorber loop to utilize their separation
capabilities. Specificaly, the hydrocyclones are
designed to optimize limestone-gypsum
fractionation rather than dewatering as shown
in Figure 4, This is achieved by careful
selection of hydrocyclone size, operating
pressure, and vortex and apex sizes.
Figure 4 LS-2 hydrocyclone system
Centrifuges are used to dewater the gypsum byproduct down to a moisture content of 8
percent or less. The chloride concentration in the gypsum cake is reduced to less than
50 ppm by means of water washing during the centrifuge spin cycle.
The separation principle of hydrocyclones is based on the centrifugal forces acting
upon particles within the cyclone. Higher density and larger-sized particles, as
gypsum, migrate to the cyclone wall and down the length of the cyclone and out
through the apex. This flow is generally termed the underflow. Lighter and smaller
particles, as limestone and inerts, tend to follow the fluid flow and are forced into the
vortex of the cyclone. This flow is generally termed the overflow. The degree of solids
separation depends on the feed composition and characteristics, the hydrocyclone
equipment configuration and size and its operational parameters.
The optimal hydrocyclone set-up for maximal gypsum purity depends on the difference
in distribution of limestone, inerts and gypsum. The cut between the particles in the
overflow and underflow can be adjusted to minimise the amounts of limestone and
inerts in the underflow. There are a number of hydrocyclone equipment related
parameters that can be tuned to optimize the solids separation, e.g. geometrical
features such as nominal diameter, vortex finder and apex size; operational parameters
such as feed pressure and special devices such as cyclo wash (4).
Liquid to gas ratio
The liquid to gas ratio is primarily determined by the SO2 concentration of the incoming
flue gas, required removal efficiency, gas velocity and slurry conditions in the reaction
tank. Important parameters of the slurry are the liquid and solid alkalinity, respectively.
During forced oxidation conditions the liquid alkalinity of the slurry is very low since
sulfite is oxidized to sulfate The liquid alkalinity can therefore be considered to be
represented by the concentration of bicarbonate and carbonate in the slurry. Figure 5
shows the bicarbonate concentration as a function of pH under the assumption of a
constant back pressure of C02 of 0.1 atm. From Figure 5 it is clear that the solid
alkalinity must play an important role in neutralizing the absorbed SO2 within the spray
zone. Not until the pH reaches very high values, greater than 6.0, the magnitude of the
liquid alkalinity become important.
Page6
-------
To what extent the solid alkalinity can be
accessed within the spray zone is
determined by the limestone dissolution
rate and the residence time of the droplets
within the spray zone. The physical
parameter of the limestone itself which
can impact the dissolution rate is mainly
the surface area which is a function of
limestone grind size and limestone
concentration.
Figure 5 Bicarbonate concentration as a function of pH
Since it very difficult to vary the solid alkalinity without changing the liquid alkalinity and
vice versa it is often difficult to design experiments where only one of the parameter is
studied. However, tests performed at EPRI HSTC with different limestone grinds
concluded that SO2 removal mainly is a function of the limestone concentration within
the slurry and not the pH itself. However, for the same operating conditions the pH-
value is an indirect measure of the limestone concentration. The tests also showed that
for the same limestone concentration, finer grind increased the removal efficiency. This
is taken advantage of in the LS-2 system by using fine grind limestone and by using
hydrocyclones for fractionation of limestone and gypsum and thereby allowing higher
concentration of limestone in the reaction tank. Further, the high gas velocity in the
LS-2 system increases the residence time of the droplets in the spray zone allowing
more limestone to dissolve in the spray zone.
MODEL
The solids in the reaction tank consists of a mixture of gypsum, limestone and inerts
where the latter originates from impurities in the limestone feed and from fly ash
collected in the absorber. The main part of the inerts are present within the system as
fines with diameters below 10 urn, whereas the gypsum particle size distribution will
depend on process parameters such as reaction tank size, type of recirculation pumps,
etc. The limestone particle size distribution of the unreacted limestone in the reaction
tank is a complex function of the limestone feed size, limestone utilization and reaction
tank conditions. Within the wet FGD system, the limestone particles dissolve and
change size. The size of each limestone particle within the slurry therefore depends on
the initial size and the age of the particle within the system. Since small particles
dissolve faster than large particles, the particle size distribution will be influenced
greatly by the coarse end of the feed distribution. Inerts can often easily be removed
from the gypsum due to the large differences in particle size whereas the degree of
separation between limestone and gypsum requires fine grind limestone and well tuned
hydrocyclone operation.
Page?
-------
ABB has developed a model to predict both limestone utilization and the particle sizes
of unreacted limestone in the reaction tank. The model is being used to determine the
impact of various grinds on limestone utilization and to be able to select an optimal
hydrocyclone set up for limestone / gypsum fractionation. The model takes into account
both the mixing conditions of the reaction tank as well as the chemical conditions of the
scrubber(5).
The model consists of two different parts: one to model the flux from the limestone
surface area which determines the shrinking rate of different limestone particles and
one model to account for the effects of the continuous system of the reaction tank. The
flux of calcium ions from the limestone surface is calculated by a steady-state mass
transfer theory, the film theory, assuming the limestone particles to be spheres
surrounded by a stagnant film. The differential mass balance in spherical coordinates
around a limestone particle is:
-^•V(jt.b*) = rt (4)
where k indicates component k, b is the distance from the centre of the particle and r
is the reaction rate of component k.
Solving the differential mass balances all acid base reactions of sulfite and carbonate
species are considered as well as the presence of ion pairs. Equilibrium constants,
activity coefficients and diffusivities are calculated by correlation's from the Bechtel
Modified Radian Equilibrium Program, BMREQ (6) with the exception of the ionpairs:
CaHSCV and MgHS03+ and the calcite solubility (7):
=-171.9065-0.077993-7+ - j + 71.595- Iog7* (5)
The model also takes into account the diminishing effect of sulfite on the limestone
dissolution rate by surface kinetics. A rate expression of the surface kinetics has been
derived by Gage and Rochelle (7):
(6)
where kc is a surface rate constant dependent on the limestone type, subscript eg.
corresponds to the activities at the limestone surface when the solubility product of
CaCO3 is met and the subscript s corresponds to the actual activities at the limestone
surface.
To calculate the dissolution rate the material balance equation (5) is integrated and
solved for the boundary conditions given by the bulk concentrations and equation (6).
Page 8
-------
As a particle dissolves it should dissolve more and more quickly per unit area due to
variations in film thickness with shrinking particle diameter. However, in earlier
investigations a decrease in the flux as the limestone dissolves has been measured
which is the opposite result compared to mass transfer theory (8). This behavior was
especially significant when sulfite was present and was referred to as an aging effect.
The decrease in the flux was experimentally examined and it was claimed that the
aging effect could be modeled as a function of the fraction dissolved. Other factors that
could effect the dissolution over time might be inhibition, accumulation of inerts at the
interface, changes in effective area, etc. In this model we have choose to assume the
increase in flux with decreasing diameter to be equal to the decrease in flux from
inhibition, changes in effective area, etc. Hence, the flux of a particle is assumed to be
constant and only a function of the initial diameter.
J(d) = J(d0) (7)
The scrubber together with the reaction tank can be regarded as a continuous stirred
tank reactor, CSTR, due to the large recirculation ratio of slurry over the scrubber. Thus
the particle size distribution, and thereby also the limestone mass transfer area of the
slurry, will be determined by the conditions of the scrubber system, such as residence
time distribution and the overall limestone utilization. The residence time distribution
function of a CSTR is:
(8)
where 9 is the mean residence time of the tank.
The particle size distribution of the limestone feed has been modeled with the log-
gamma density function (9). The cumulative distribution curve, P(Y), is found by
integrating the density function:
(9)
where Y = n (10)
The cumulative distribution curve is characterised by three parameters, A, p and oW
Gage (1989) showed by statistical analysis that the 2 value could be adequately
represented by assuming integer values of 4 and 6. Using a fixed A, the ft and dioo
values can be determined by two sieve measurements.
Page 9
-------
The limestone particles are assumed to be spherical. By expressing the mass and the
area of the particle in terms of the diameter, the following relationship can be derived
for a particle:
(11)
where M is the molar weight of limestone and p is the density of limestone. The
limestone conversion, f, of the system can be expressed in terms of the ratio between
the limestone of the slurry, Mtot, and of the feed, M0. tot'-
A4
Mn,.
(12)
The mass of one fraction pf the limestone feed, MQ., consists of n, particles with the
mass of m0., and a diameter of da. /. The remaining mass of the i:th fraction in the slurry,
Mo.!, can be calculated according to the following equation:
M>.,' = J«, -m^ E(t)-dt
0
Expressing m, and n, by the following equations:
(13)
(14)
(15)
and by adding up all fractions, the total conversion of limestone can be calculated as a
function of average residence time, 6, and the particle size distribution:
(16)
Substituting ds with d0ii using equation (11) the integral can be solved analytically:
*"•"-!
(17)
Page 10
-------
where
2-M
(18)
To calculate the particle size distribution of the unreacted limestone, the change in
distribution of each fraction of the limestone feed must be calculated. The mass of the
rth fraction of the feed with a residence time longer than t is equal to:
t (19)
The mass, m,, can be expressed with equation (14) and n, with equation (15). If t is
substituted with d, using equation (11), the remaining mass of the i:th fraction of the
feed with a diameter less than dy, WirJ, can be calculated by the following equation:
X.-0.
-3
+ 6
d.
-6
(20)
Adding the contribution from all the feed fractions the total mass of the /.th fraction of
the limestone is obtained:
(21)
The particle size distribution of the unreacted limestone is calculated as follows:
M.
+jr. ™
RESULTS
Conversion model versus laboratory data
A continuous stirred tank reactor, CSTR, was
used to simulate a wet FGD reaction tank,
Figure 7. A limestone slurry with a solids
concentration of 10 weight-% was fed to the
reaction tank. The CSTR had a fixed volume
and by varying the limestone slurry feed rate,
solids residence times ranging from 2 to 8 h
were simulated. pH stat equipment was used
to maintain the pH value of the reaction tank
at a constant value by adding 18 M H2S04.
The added H2S04 reacted with the dissolved
Limestone
Feed
Absoiber
Reaction
Tank
Figure 7 Test apparatus for a CSTR
Page 11
-------
limestone, forming gypsum, CaSO-i 2 H2O, which precipitated in the CSTR. Each
experiment was run at least three times the mean residence time, 6, in order to obtain
steady-state before any analysis was made. A computer connected to the pH stat
recorded the time, the pH-value, the volume of added acid and the flow rate of acid. At
the end of each experiment, a slurry sample was taken from the reaction tank for solids
analysis. The solids concentration was measured by gravimetric analysis. The content
of limestone was analyzed by dissolving a dry slurry sample in excess hydrochloric acid
followed by back-titration with sodium hydroxide. The dissolution rate of the slurry was
measured by the flow rate of added acid.
A limestone with three different particles
sizes (A, B, C) was used in the tests.
Limestone A: dso = 22.3 urn and dc,o= 46.6
urn; limestone B: dso = 7.1 Mm and d9o=
37.2; and limestone C: dso = 6.0 urn and
d9o= 18.1. Figure 8 shows the
corresponding values between limestone
conversion and residence time for
limestone A, B and C in a solution with a
pH value of 5.8. The agreement between
the model and the measured values is
very good. The limestone conversion
increases with increasing residence time
and, as anticipated, as the conversion
approaches unity the rate of conversion
decreases.
Residence time (h)
Figure 8 Limestone utilization versus solids
residence time
Conversion model versus field data
At HSTC different limestone grinds have
been tested under various conditions (10).
As anticipated the limestone conversion
was found to decrease with increasing pH
and decreasing solids residence time.
Further, it was found that the utilization
increased with as finer grind limestone
was used. The derived model have been
used to predict the limestone utilization.
Figure 9 shows the calculated values
versus measured values of limestone
utilization. As for the laboratory data, the
agreement is very good.
3 0.1 -•
• 98% through 325 mesh
o 90% through 325 mesh
' ultra fine grind
(1 - limestone utilization) measured
Figure 9 Modeled versus measured limestone
conversion
Page 12
-------
Particle size of unreacted limestone
Figure 10 shows the calculated particle
size distribution of a fine ground
limestone used at the LS-2 with d^ equal
to 10 urn and dgo equal to 25 urn. As the
conversion increases, the particle size
distribution is shifted towards larger
diameters since the small particles
dissolve first. The particle size distribution
within a continuous system is
consequently determined by the large end
of the cumulative distribution curve.
Figure 10 PSD as a function of limestone utilization
The use of the ABB Raymond roller mill with adjustable classifier rate which allows the
duo to be fixed at a low value is therefore beneficial for the limestone gypsum
separation.
EFFECT OF LIMESTONE GRIND SIZE ON HYDROCYCLONE
PERFORMANCE
The slurry from the reaction tank is pumped to a battery of 6 inch hydrocyclones prior to
final dewatering in the centrifuges. Six hydrocyclones are employed, each with optional
Cyclowash™ available. The purpose of the hydrocyclones is both to increase the solid
content of the underflow slurry stream sent to the secondary dewatering system and to
fractionate the feed solids; i.e. recycle a higher proportion of limestone particles and
inerts in the overflow back to the reaction tank compared to the underflow. An efficient
fractionation of limestone particles from the feed slurry will increase the limestone
utilization and permit reduced reaction tank volume, since higher limestone
concentrations within the reaction tank can exist without reducing the gypsum purity.
Testing has been conducted to quantify the performance of the hydrocyclones, and to
optimize operation with respect to gypsum/limestone separation by varying operating
parameters and internal mechanical components. The parameters studied have been:
operating pressure, vortex size, and apex size (11). The intent during optimization was
to obtain high gypsum content in the underflow and high limestone recovery in the
overflow.
The model described above was used to predict the particle size distribution of the
unreacted limestone particles in the reaction tank slurry, Figure 11. The frequency
curve of gypsum and inerts has been calculated as the difference between the
frequency curve of the slurry and of limestone. The portion of gypsum curve below 10
urn is considered as inerts.
Page 13
-------
The standard configuration of the
cyclones a 6" hydrocyclone
corresponds to a dso cut of around 15
urn. Based on the particle size
distributions curves in Figure 11 it was f
obvious that a cut size of dx equal to
15 urn would leave a substantial
amount of unreacted limestone in the
gypsum in the underflow. To increase
the gypsum purity dso needs to be
moved to larger diameters. This could
be done by changing the geometrical Figure 11
features of the cyclone and reducing
the feed pressure.
frequeny curves of limestone and gypsum
0 20 <0 60 80 100 12
particle diameter ((jm)
Predicted frequency curves of limestone
and gypsum/inerts of the reaction tank
slurry
To increase the dso, the feed pressure was reduced with about 40%, the vortex finder
was increased with 50% and the apex reduced with 50% compared to the original set
up. After these modifications, dso was increased from 15 urn to 35 urn and it was
possible to increase the concentration of CaC03(s) in the reaction tank from about
3.5% up to 6%. This increase in limestone concentration had a substantial effect on the
S02 removal.
Following optimization, the
compositions of the feed, overflow,
and underflow streams are as shown
in Table 1. The overflow solids
content is higher than initially
projected, but has produced the best
limestone recovery. Future testing
and designs will aim towards lower
overflow solids while maintaining
fractionation. A second stage hydro-
cyclone to treat the overflow is
optional but would add complexity
and cost.
The predicted particle size
distribution curves in Figure 11 were
used to estimate the gypsum purity
for different cuts that were obtained
by changing the geometrical features
and the operating conditions of the
hydrocyclone. Figure 12 includes an
analysis of the gypsum purity for both
Table 1 Hydrocyclone Performance
Parameter Feed Overflow Underflow
TSS
CaCO3
Gypsum
Inert
Dso
20.8
5.7
92.8
1.5
45.9
12.7
8.7
89.5
1.8
31.2
58.3
2.4
97.2
0.4
52.6
Measured w % CaCO3
Figure 12 Predicted and measured carbonate content of
the overflow and underflow at LS-2
Page 14
-------
the underflow and overflow from the hydrocyclone. The agreement between modeled
and analysed gypsum purity, as shown in Figure 12, was exceptional.
CONCLUSIONS
The LS-2 project represents an advanced wet FGD system which offers superior
performance at significantly reduced capital and operating costs. Many of the LS-2
design concepts is based on the use of fine grind limestone, e.g. small reaction tank,
high gypsum purity and high S02 performance. The limestone preparation system of
LS-2 features an ABB Raymond roller mill. The limestone grind will typically be 99.5%
less than 44 urn.
A model has been developed to predict limestone utilization and the particle size
distribution of the unreacted limestone in the reaction tank. Good comparison between
predicted and measured values have been obtained. The model is able to predict
effects of limestone grind size, slurry chemistry and reaction tank design on limestone
utilization. The calculation of the particle size of the unreacted limestone is important to
choose an optimal hydrocyclone for limestone / gypsum fractionation. Hence, the model
will be an important tool in optimization of WFGD plants.
ACKNOWLEDGEMENT
The co-operation of Ohio Edison through all stages of the LS-2 project is greatly
appreciated.
The LS-2 demonstration project has been cofunded by the Ohio Coal Development
Office (OCDO) and by the Electric Power Research Institute (EPRI). ABB is grateful for
OCDO's and EPRI's participation in the LS-2 project.
REFERENCES
1. J.S. Klingspor, G.E. Bresowar, Advanced, Cost Effective Wet FGD. Presented at
the EPRI S02 Conference, Miami, Florida, May 1995.
2. J.S. Klingspor, G.E. Bresowar, Next Generation Low Wet FGD System. Presented
at the PowerGen 95 Conference, Annaheim, California, December 1995.
3. J.S. Klingspor, D.C. Borio, D.J. Collins, D. Gausmann, LS-2, A Performance
Update. Presented at the Powergen 1996 Conference, Orlando, Florida, December
1996.
4. C. Bessendorfer, Exert the Forces of Hydrocyclones. Chem. Eng., Sept. 1996.
Page 15
-------
5. C. Brogren and H.T. Karlsson, A Model for Prediction of Limestone Dissolution in
Wet Flue Gas Desulfurization Applications. Ind. Eng. Chem. Res. 1997, accepted
for publication.
6. M. Epstein, EPA Alkali scrubbing test facility: Summary of testing through October
1974, U.S. EPA 600/7-7-105, 1977
7. C.L. Gage, G.T Rochelle, Limestone dissolution in flue gas scrubbing: effect of
sulfite. J. Air Waste Manage. Assoc. 1992, 42, p. 926
8. J.B. Jarvis, F.B. Meserole, T.B. Selm, G.T. Rochelle, C.L. Gage, R.E. Moser,
Development of a predictive model for limestone dissolution in wet FGD systems.
Presented at EPA/EPRI combined FGD and dry S02 control symposium, St Louis
1988
9. C. Gage, Limestone dissolution in Modeling of Slurry Scrubbing for Flue Gas
Desulfurization. Ph. D. thesis The University of Texas at Austin, 1989
10. EPRI ECTC Fine grind limestone (FGL) test block report, 1997
ll.J.S. KLingspor, C. Brogren, High purity gypsum from fine grind limestone.
Presented at the 5th international conference on FGD and synthetic gypsum,
Toronto, May 1997
Page 16
-------
WFGD SYSTEM MATERIALS COST UPDATE
M.G. Milobowski
Utility & Environmental Power Division
The Babcock & Wilcox Company
20 S. Van Buren Avenue
Barberton, Ohio 44203
Abstract
This paper is an update of the report "Economic Comparison of Materials of Construction of Wet
FGD Absorbers and Internals" which was presented at the 1991 EPRI/EPA/DOE symposium. An
economic comparison of the materials used for fabrication of wet FGD spray towers will be
presented, with focus placed on the latest experience/developments with alloy, lined carbon steel,
and lined concrete material selections. Costs for various materials of construction for such
absorber components as spray headers, moisture separators, and gas distribution devices will also
be addressed.
Introduction
Materials of construction are a significant design consideration for wet FGD systems. While
much has been learned about the materials for absorbers and their components during the period
1970 through 1990, it is still true that corrosion is not an exact science. Wet FGD system
operating requirements such as seawater make-up or closed loop/zero discharge only serve to
"push the envelope" of corrosion potential. This paper will review some of the options for
construction materials for wet FGD absorbers and their internals, and will compare current pricing
with that from 1991.
This paper also discusses Babcock & Wilcox's (B&W) experience with FGD system materials of
construction. The following components are addressed:
Inlet fluework and dampers
Absorber inlet wet/dry interface
Absorber recirculation tank
Absorber spray zone
Absorber moisture separator and outlet zones
1-
-------
• Headers
• Absorber tray/gas distribution devices
• Absorber moisture separators
• Absorber internal supports
A typical countercurrent spray tower is shown in Figure 1. The countercurrent spray tower is a
commonly used design in the wet FGD industry. Shown in this illustration are B&W's patented
absorber tray which promotes uniform gas flow distribution across the absorber when coupled
with a centered cone outlet, and interspatial spray headers which reduce absorber height and save
capital cost.
Inlet Fluework and Dampers
Prior to utilization of regenerative gas-gas heaters (GGFf), the inlet fluework and dampers of the
wet FGD system were exposed only to dry, hot (3 OOF) flue gas. The 300F temperature was well
above the acid dewpoint of the gas, so ASTM A3 6 carbon steel was an acceptable and
economical material choice for the fluework and damper structural elements located upstream of
the absorber. If a guillotine damper was utilized, the blade was typically fabricated from ASTM
A242 corten steel since it primarily remains raised in the open position and exposed to nothing
more corrosive than the atmosphere. 300-series stainless steel was used to fabricate the blade seal
strips. More exotic materials were used only if slurry spray-back from the absorber was a
The probability of slurry spray-back into the inlet fluework depends primarily upon inlet
configuration. B&W's absorber tower design incorporates an awning over the inlet and side
shields to minimize spray-back. To protect against corrosion resulting from the tower spray that
does penetrate the inlet, B&W's practice is to extend the materials used for the wet/dry interface
back into the inlet flue from the tower.
The use of regenerative GGHs has made material selection for flue and damper components more
costly because flue gas temperature is now below the acid dewpoint. The fluework between the
GGH dirty gas outlet and the absorber inlet must now be solid alloy, or carbon steel typically
having a resin or alloy wallpaper lining. Absorber inlet damper components exposed to the flue
gas must be alloy, usually ASTM B575 Alloy C-276 or C-22.
Absorber Inlet Wet/Dry Interface
The absorber inlet wet/dry zone is exposed to both the incoming dry hot flue gas (if no GGH is
present) and the absorber's slurry spray. The corrosive condition at this interface is most
formidable; not only is it very acidic, but if the inlet is improperly designed, this wet/dry interface
can foster solids' deposition and the possibility of under-deposit corrosion. Due to evaporation
-2-
-------
at the wet/dry interface, concentrations of chloride ion in excess of 100,000 ppm are possible.
A variety of materials have been used in this wet/dry interface zone. Prior to the development of
high temperature-resistent glass-flake filled resin (flakeglass) linings, resin linings overlaid with
refractory for thermal protection were used. Alloy C-276 or C-22 is specified when reduced
maintenance is desired. Carbon steel substrate with an alloy wallpaper lining (typically 1/16"
thick) is a very cost effective compromise between resin-lined carbon steel and solid alloy
construction. And borosilicate glass block, having a hard glaze surface layer to deter abrasive
wear, has also been used, although rather infrequently. This type block lining is more typically
used in the absorber outlet fluework.
As with any lining system, proper application is paramount to system integrity. Wallpaper
systems are no different than non-metallic lining systems in that proper applicator training,
supervision and QA inspection/testing are essential. It should be noted that while alloy wallpaper
lining provides greater longevity than a non-metallic lining system, there is the drawback that any
application flaws not detected by final inspection will, more often than not, go unnoticed until
substrate material damage becomes apparent. Application flaws in non-metallic linings are more
readily apparent prior to severe substrate damage, via lining blistering, cracking or delamination.
Absorber Recirculation Tank
In most FGD systems today, the absorber recirculation tank is integral to the tower structure,
rather than being an independent vessel. This tank is an agitated vessel in which side-entry
agitators are used to keep the slurry solids in suspension. The two highest areas of abrasive wear
in this tank are the floor, due to solids' "sweeping" by agitator-induced and recirculation pump
suction-induced currents, and the pump suction nozzles themselves. This abrasion is not quite as
severe as that experienced in the tower's spray zone, since the "cutting" action of spray
impingement is not present, but it is significant none-the-less. Floor linings must be protected
from mechanical damage caused by unit outage inspections and maintenance work.
Acid-resistent brick or tile can be used to line a carbon steel recirculation tank's floor and walls,
but while these ceramics are highly resistent to abrasion and corrosion, the mortar can become
permeable with time, necessitating an elastomeric or resin membrane. Also, vibration and/or
thermal cycling may cause the bricks or tile to disbond from any vertical surfaces.
Stainless or alloy steels would provide excellent service in this application, with stainless steels
being the economic choice when slurry chloride concentrations are less than 20,000 ppm. As
shown in Table 1, the next lowest capital cost options are flakeglass resin-lined or elastomer-lined
carbon steel, both of which are suitable for 100,000+ ppm chlorides' service. It should be noted
that theoretically calculated chloride levels are seldom realized in actual operation. The principal
-3-
-------
drawback of non-metallic linings is their limited life expectancy, typically 10-15 years. Both
linings are also susceptible to mechanical damage, so the installation of an acid brick overlay on
the floor's surface is recommended.
Absorber Spray Zone
Slurry spray nozzles must be positioned within the absorber tower to provide complete cross-
sectional area coverage, eliminating the possibility of untreated flue gas circumventing slurry
spray contact. Spray impingement against the tower shell does therefore occur, and since the
reagent slurry is typically 15-20% solids, shell abrasion can result. Most FGD system designers
try to position their spray nozzles to minimize slurry spray impingement force against the shell,
without sacrificing spray coverage.
Non-metallic lining products have improved dramatically over the past twenty-five years.
Elastomer lining manufacturers have offerings which now have excellent permeability resistance
to complement their products' superior abrasion resistance. Manufacturers of high permeability-
resistent resin linings have overcome erosion concerns by offering specially formulated abrasion-
resistent linings that are typically 1.5-2.0 times thicker than their standard lining products.
While it is true that the expected life of non-metallic linings may be only one-third to one-half the
life of the FGD system, the use of solid stainless steel or alloy in the spray zone area is not
without maintenance costs. Operators of solid stainless absorbers that have been in service for a
considerable length of time are finding some shell thinning at slurry spray impingent points,
necessitating installation of metallic wear plates or non-metallic linings. If alloy hardness is used
as an indicator of abrasion-resistence, one finds that except for the harder duplex stainless steels,
there is not a significant difference in hardness between the austenitic stainless steels, 6% moly
stainless steels, and the high nickel alloys.
For customers wary of non-metallic linings, but unable to economically justify a solid alloy plate
tower, alloy wallpapering of carbon steel substrate is an option. Although not recommended for
liquid containment service such as the absorber recirculation tank, alloy wallpaper is a viable
option for tower lining material. The number of alloys now being offerred as wallpaper has
increased dramatically in recent years. In a direct slurry spray impingement zone, however,
wallpaper thicker than the minimum 1/16" is definitely recommended.
Absorber Moisture Separator and Outlet Zones
The absorber moisture separator and outlet zones are exposed to a different environment than
those previously discussed. The abrasive action of the recycle slurry no longer dominates.
Instead, corrosive attack by sulfurous acid, resulting from the combination of moisture separator
wash water with flue gas residual sulfur dioxide, is the main problem. Typically, the same
-4-
-------
materials and/or linings used for the absorber recirculation tank's shell wall are also used for the
tower's moisture separator zone, with few problems being reported.
Elastomer linings, while not required for abrasion-resistence, have also been utilized successfully
in this area of the tower. B&W has found that when transitioning from an elastomer lining in the
moisture separator zone to a flakeglass resin lining in the outlet fluework, the best interface is
achieved by conditioning the resin lining and overlapping the elastomer lining onto it.
While the absorber's outlet fluework was a major maintenance headache in early FGD systems,
advances in FGD components and materials have siginificantly reduced problems in this area.
More efficient moisture separators have reduced the amount of solids' carryover, and the
accompanying underdeposit corrosion problems, in this fluework. The introduction of improved
flakeglass resin and epoxy linings have eliminated the maintenance headaches associated with
refractory overlays that were prone to damage as a result of mechanical impact. Availability of
alloy wallpaper linings, with better experience-driven application procedures, has reduced the cost
of being able to use the more corrosion-resistent alloys.
With regard to dampers, those damper components in the gas stream will typically require high
nickel alloy materials, such as ASTM B575 Alloy C-276 or C-22. A lower grade alloy, such as
317L stainless steel, could be utilized for a guillotine damper's blade since it spends the majority
of its time out of the gas stream.
Headers
While stainless steel, alloy, fiberglass reinforced plastic (FRP), and elastomer-lined and covered
carbon steel pipe have all been used as absorber spray headers, the trend has been to FRP headers
on the majority of recent wet FGD projects. The two most common reasons for this are: the
unavailability of most alloys, needed because of the high chlorides' concentration, in pipe form;
and the increasingly successful track record of FRP in this service. Spray header materials' cost
comparisons are presented in Table 2.
In the absorber, three different types of FRP pipe are typically specified. For the slurry spray
headers, the FRP piping must be chemically-resistant and have abrasion-resistent inner and outer
surfaces. For the mist eliminator wash headers, the FRP can be specified as chemically-resistent
only, provided the wash water's solids concentration is below 5%. And for the oxidation sparge
piping located in the recirculation tank, chemically-resistent FRP, having an abrasion-resistent
outer surface layer only, is what's typically required.
This is not to say that FRP piping is without its drawbacks. To minimize piping failure from
abrasion/erosion in slurry service, the limestone grind has to be strictly controlled. Also, slurry
spray nozzles have to be properly connected to the headers, since a flange leak can grow into a
-5-
-------
fan spray causing erosive damage. FRP headers may require more maintenance during their
useful life; fortunately, typical maintenance involves cutting out the bad section of pipe and
adhesively bonding in a replacement pipe section.
Absorber Tray/Gas Distribution Devices
All of B&W's wet FGD absorbers utilize a patented absorber tray for more uniform flue gas
distribution across the absorber. A portion of this perforated and compartmentalized tray is
shown in Figure 2. The tray is typically 14 guage stainless steel or alloy, with slurry pH and
chlorides' concentration being the key corrosion factors considered in material selection.
An economic comparison of absorber tray materials is shown in Table 3. The relative pricing is
essentially equivalent to the materials' cost.
Absorber Moisture Separators
Moisture separators have been commonly fabricated from fiberglass reinforced plastic (FRP),
glass-coupled polypropylene, austenitic stainless steels, and high nickel alloys. The most
commonly used material for moisture separators in the U.S. is FRP, with polypropylene coming in
second. Polypropylene's acceptance in the U.S., first in the replacement market and now in the
original equipment market, was initially based on successful experience with this material in Japan
and Europe. Depending upon process conditions and moisture separator design, polypropylene
moisture separators can last as long, or as little as one half as long, as FRP moisture separators.
Acid attack and underdeposit corrosion are the main problems with using austenitic stainless
moisture separators. Unless all stainless surfaces are thoroughly cleaned and passivated prior to
start-up, and water washed at frequent intervals (i.e. once an hour) during tower operation, low
pH sulfurous acid and chloride deposits will decimate the thin gauge metal. Further, stress and
crevice corrosion can be accelerated in areas where the austenetic stainless has been mechanically
formed into multiple-pass chevron shapes. While high nickel alloys are considerably more
resistent to this environment, they are also more expensive, as illustrated by Table 4.
Absorber Internal Supports
Most internal supports are constructed from material identical to that used for the absorber shell.
If internal supports are to be lined, box beams are typically utilized to simplify the lining
application. If the supports are to be stainless or alloy steel, then I-beams are routinely used.
-6-
-------
Localized failures of internal supprts are most often caused by spray impingement abrasion. These
areas are repaired and then protected by increasing the lining thickness, or installing a replaceable
metallic or non-metallic shield. Such problems are not applicable if self-supporting headers are
utilized; however, this construction is not economical with the large diameters of today's towers.
New Materials
The Utility Power industry is very conservative in its approach to using new materials in wet FGD
service. Typically, if the material has not been in service for five years minimum, the Power
industry considers it an unproven commodity. And material tests conducted in "simulated"
scrubber environments usually have little persuasive worth.
Since 1991, B&W has seen only two "new" materials accepted by its customers for use in wet
FGD systems. Korea Electric Power Company (KEPCO) specified 6% molybdenum stainless
steel (AL-6XN/254-SMO) as the material of construction for the ten new absorbers at Hadong
and Taean. Preliminary word is that Taiwan Power will also be specifying 6% molybdenum
stainless steel for its Hsinta 3 & 4 units' absorbers. At Pembroke in the United Kingdom, a
fluoroelastomer lining was to be used for fluework that was being exposed to somewhat high
levels of sulfur- trioxide condensate - since this project is on "hold", so is final material approval.
Cost Comparison with 1991
For absorber tower construction, Table 1 indicates that the 1997 total cost ratios (compared to
the unlined carbon steel base case) are only slightly higher overall than those of 1991. The
maximum increase observed is 7.5% for a flakeglass-lined carbon steel absorber. In a few cases,
the cost ratios are even less than the 1991 values, with the greatest reduction being -3.4% for an
Alloy 255 duplex stainless steel absorber. For 1997, the procurement/fabrication costs, when
compared to carbon steel, have decreased compared to the 1991 values. And while the erection
cost ratios for 1997 have not changed significantly, B&W's data indicates that erection costs now
account for a higher percentage of the total cost than was the case in 1991. It should also be
noted that there have been some reverses in the overall rankings of the type of absorber
construction by total cost ratio: a 317LMN stainless steel absorber now has a lower cost ratio
than a flakeglass-lined carbon steel absorber; and a 255 duplex stainless steel absorber now has a
lower cost ratio than a concrete/block absorber.
The fact that 1997 cost ratios have not changed significantly from those of 1991 is also observed
for absorber internal components. Tables 2 through 4 indicate that absorber components' 1997
cost ratios are basically unchanged, if not slightly lower overall, from those of 1991. The most
significant decreases seen were in absorber tray construction, where cost ratios are approximately
10% lower than in 1991.
-7-
-------
Conclusion
There are a large number of materials available for construction of wet FGD systems; however
very few new materials have gained the Utility Power industry's approval in the last six years.
Preferred materials include elastomer and flakeglass-lined carbon steel, alloy wallpapered carbon
steel, austenitic stainless steel, 6% molybdenum stainless steel, high nickel alloys, FRP and
polypropylene. While the total cost for absorber construction is greater in 1997 than it was in
1991, the increase can be considered minimal. This cost trend is also observed for absorber
internal components, where the 1997 cost ratios are unchanged, if not slightly lower, as
compared to those of 1991.
-------
Table 1
FGD Absorber Cost Comparison
The following absorber plate (with stiffeners) cost comparisons are based on a 0.25 inch thick
nlate and knockdown construction
plate and knockdown construction.
1997
Material
Carbon Steel (CS)
CS Elastomer-Lined*
CS Flakeglass-Lined*
316L Stainless Steel
317L Stainless Steel
317 LMN Stainless Steel
AL-6XN/254-SMO
255 Duplex Stainless Steel
Alloy C-276
Alloy C-22
CS w/C-22 Cladding
CS w/C-22 Wallpaper
CS w/Tile Lining**
Concrete/Block
Procurement/
Fabrication
1.00
5.00
4.60
4.00
4.40
5.30
7.50
7.00
14.00
14.00
9.50
4.50
6.20
10.00
Erection
1.00
.28
.28
.00
.00
.00
.05
.05
1.19
1.19
1.14
1.43
1.00
Included
Total
1.00
2.26
2.15
1.79
1.89
2.13
2.71
2.58
4.55
4.55
3.34
2.30
2.37
2.63
1991
Total
1.00
2.22
2.00
1.70
1.89
2.11
N/A
2.67
4.59
4.59
3.40
2.30
2.30
2.63
* Lining/coating cost varies based on surface area and site location.
** Lining material cost includes field installation.
Note: The least expensive option is assigned a base cost of 1.0. Costs of all higher priced
options are presented relative to the base cost of 1.0.
-9-
-------
Table 2
Cost Comparison of Slurry Spray Headers
1997 1991
Material Cost Ratio Cost Ratio
1.0 1.0
1.5 1.5
2.1 N/A
2.4 N/A
2.5 2.5
7.0 6.8
Abrasion-Resistent FRP
Carbon Steel, Rubber-Lined & Covered
316L Stainless Steel
317L Stainless Steel
317LMN Stainless Steel
Alloy C-276/C-22
Table 3
Cost Comparison of Absorber Trays
Material
3 16L Stainless Steel
317L Stainless Steel
3 17LMN Stainless Steel
AL-6XN/254-SMO
255 Duplex Stainless Steel
Alloy C-276/C-22
1997
Cost Ratio
1.0
1.1
1.3
1.9
1.8
3.5
1991
Cost Ratio
1.0
1.2
1.4
N/A
2.0
3.9
Table 4
Cost Comparison
Material
Glass-Coupled Polypropylene
Fiberglass Reinforced Plastic (FRP)
317LMN Stainless Steel
Alloy C-276/C-22
of Moisture Separators
1997
Cost Ratio
1.0
2.5
3.0
7.7
1991
Cost Ratio
1.0
2.6
3.0
7.6
Note: The least expensive option is assigned a base cost of 1.0. Costs of all higher priced
options are presented relative to the base cost of 1.0.
10-
-------
Moisture
Separator
Water
Wash
Nozzles
Moisture Separator Level
Agttuor
Interspatial Spray Level
(Plan View)
Figure 1 Absorber Cutaway View
Patented Alloy
Perforated Tray
Figure 2 Absorber Gas Distribution Device
-------
Wednesday, August 27; 1:00 p.m.
Parallel Session B:
Dry SO2 Control Processes
-------
NID-
A New Dry Flue Gas
Desulfurization System
Stefan Ahman
ABB Flakt Industri AB
S-351 87 Vaxjo, Sweden
John Buschmann
ABB Environmental Systems
1400 Centerpoint Blvd.
Knoxville, TN 37932-1966
Abstract
This paper describes a New Integrated dry flue gas Desulfurization (NID) System which has been
developed by ABB. By the integration of the desulfurization and paniculate removal into one unit it is
possible to lower costs and to lower the space requirements of the DFGD system.
In addition to the design simplification, the system shows better SO2 removal performance and lime
utilizations than existing DFGD systems. The new technology is modular, allowing for flexible system
design in any size required. The process was developed at the ABB R&D facilities in Vaxjo, Sweden
during 1994-96.
Tests in pilot plants and a demonstration plant at a Polish power plant were very successful. The first
commercial installation (2x 120 MW) was started up in the autumn of 1996. The paper reports on the
results of the development efforts and highlights the first commercial experience.
-------
Introduction
The need for a simple and reliable, low cost desulphurization system was identified by ABB as a
strategic issue, especially for the emerging markets in Asia and East Europe. A dry flue gas
desulphurization (DFGD) technology was assessed to be the preferred method of choice.
The method should be able to achieve at least 80 % S02 removal on low and medium sulfur coals. The
system should further be easy to retrofit at existing sites; it should thus have minimum space
requirements. An important feature of the DFGD technology, sometimes not highlighted enough, is the
fact that particulate collection of fly ash is facilitated by the FGD system at no extra capital charge. The
flue gas temperature after a DFGD system also often allows the flue gas to be passed on to an existing
stack without reheat.
All these factors were weighed in as ABB decided to develop the new DFGD system "NTD" "NTD" is
an acronym for "Novel Integrated Desulphurization" indicative of the innovative nature of this FGD
technology enabled by the integration of several process steps into one unit.
The development work has followed a very fast track. A decision was made in the spring of 1994 to go
ahead with the development after initial conceptual studies.
In lune 1994, the Polish power company "Elektrownia Laziska" placed an order with ABB for a fabric
filter (FF) after its boiler # 2 for high efficiency collection of fly ash. At this time an agreement was
signed between Elektrownia Laziska and ABB to install and test the new NTD concept for flue gas
desulphurization on one of the compartments of the new fabric filter.
Remaining development work was made at the ABB R&D facilities at Vaxjo, Sweden, concurrent with
designing the demonstration unit The demo plant was started up in February 1995 and has been in
operation for several test periods since that time
Based on the satisfactory results from the demo plant, Elektrownia Laziska placed orders with ABB for
the extension of the NID technology to their two 120 MW boilers (#1 & #2). Both full scale FGD units
were commissioned during 1996.
-------
The NID Process
The NID process is based on the absorption of SO2 by a dry absorbent containing quicklime (CaO) or
dry hydrated lime, Ca(OH)2 Either of these absorbents may be used or e. g. a fly ash containing an
appropriate amount of alkali
The key parameter to be controlled in any dry FGD process is the humidity of the flue gas. At a relative
humidity of 40 - 50 % the hydrated lime becomes activated and absorbs SO2. The relative humidity of
the flue gas is increased by the injection of water into the flue gas. In a conventional DFGD process,
water and lime is supplied to the flue gas as a slurry (with or without recycle) with a solids content of
35 - 50 %. In the NID process, the same amount of water is injected into the flue gas, but it is
distributed onto the surface of dust particles at a water content of only a few percent.
Thus, the amount of absorbent which is recycled is much bigger than in a conventional DFGD process.
This means that the surface available for the evaporation is very large. Thus, the time for drying of the
dust added to the flue gas is very short, which in turn makes it possible to use very small reactor vessels
compared to conventional spray dryer technology. The resulting increase of the relative humidity of the
flue gas is sufficient to activate the lime for absorption of SO2 at typical DFGD operation temperatures
of 20-40 deg F above saturation. Figure 1 below illustrates the steps of wetting and drying the recycled
absorbent in a conventional DFGD and a NID system.
% H20
70
60
Thin slurry
50
40
Thick slurry
10
Free flowing dust I
Adsorbed equilibrium
moisture
Conv. DFGD Process NID Dry Process
Figure 1. Illustration of difference between slurry based vs. wetted dust based DFGD processes
-------
Water is added to the absorbent in a humidifier prior to its introduction into the flue gas. A unique
feature of the NTD technology is the fact that all recycled absorbent is being subject to wetting and
activation in the humidifier, which maximizes the utilization of the recycled absorbent. After the
activation/drying step, the dried recycle dust is separated from the flue gas in an efficient dust collector,
preferably a fabric filter (FF). From here the dust is again fed to the humidifier, to which make up lime is
added as well. Water is fed to the humidifier in a quantity sufficient to maintain a constant outlet flue
gas temperature. The control system uses a feed forward signal with back trim, based on the in- and
outlet flue gas temperatures, supplemented by a signal indicating the gas flow. The outlet SO2
concentration is controlled in a similar way; the in- and outlet SO2 concentrations plus the flue gas flow
determines the lime flow to the system. The main features of the NTD process are shown in figure 2
below. As indicated, the FGD waste product accumulates in the hopper of the dust collector, and as the
maximum level of the hopper is reached, the waste product leaves the system by overflow.
The NID process is thus characterized by a very high recycle rate, which in turn means-that the
utilization of the reagent is maximized. As indicated above, the high recycle rate also means that there is
a large surface area available for the rapid evaporation of water, which in turn means that the volume of
the reactor/dryer for the NID process is an order of magnitude less than the corresponding size of a
conventional dry flue gas cleaning system based on spray dryer technology.
Further, the need for sophisticated special equipment is minimized in the NTD process. There is no
rotary atomizer with its high speed machinery; nor are there any dual fluid nozzles with their need for
compressed air. Power requirements for the mixing of the recycle/reagent in the humidifiers are much
lower than for the corresponding items in a conventional dry flue gas cleaning system: by comparison
rotary atomizers and dual fluid nozzles appear much more complex than the NTD humidifier. An
important consequence of using humidifiers rather than nozzles or rotary atomizers is that all equipment
that needs operator's attention is placed near ground level, in an enclosure common with the fabric filter.
This arrangement means lower cost and increased ease of maintenance.
Finally there is no slurry handling with requirements for special pumps etc., since water is added directly
to the NTD humidifier. The high recycle rate also means that only dry material is handled in the system.
The system is thus free from build-ups in gas ducts etc., since there is no wet slurry that can impinge on
surfaces in the installation.
-------
Water
Reagent End Product
Figure 2. The main features of the NID process: High recycle rate means rapid drying and a small
reactor size, paired with a high utilization of the reagent.
The Fabric Filters after Laziska #1-2 boilers
The boilers # 1 - 2 at Laziska are pulverized coal fired ,each rated at a power output of 120 MW. The
fuel is domestic bituminous coals from nearby sources.
In the initial phase, the FFs were supplied for the collection of fly ash only, without the capability for
desulphurization. The FFs handle all of the flue gas from the boilers and are capable of treating a
nominal gas flow of 2 x 420,000 ACFM.
The FF is the high ratio type using ABB's Optipulse LKP design. The filtration is from the outside of
the filter bags, inwards through a dust layer deposited on the surface of the bags and eventually through
the felted fabric itself. Cleaning is by sending a pulse of compressed air backwards through the bags,
which make the dust fall off the bag into the hopper below. The bags are cleaned to maintain a constant
differential pressure drop over the FF. The filter media used is Acrylic which can withstand a max.
temperature of 280 deg F. In case of a higher gas inlet
-------
temperature, the flue gas is cooled by ambient air dilution to protect the bag material from high
temperature excursions. The guaranteed dust outlet emission is a maximum of .012 gr./ACF.
NID Demonstration Plant
The NID demonstration unit was installed on one of the fabric filter compartments of the FF on unit # 2.
Initial supportive testing in the ABB R&D Laboratory at Vaxjo, Sweden, was much focused on the
issues of dust wetting and operating performance of the full scale unit. Efficient and homogenous dust
wetting is a key issue for the success of the NID process. The wetting aspects were studied separately in
a semi-commercial scale humidifier, utilizing a mixture of fly ash and lime. After concluding this study,
the humidifier was added to an aerodynamic flow model, which allowed for testing of the combined
systems for wetting and dust dispersion into the flue gas. Computational fluid dynamics (CFD) modeling
of the process was performed in parallel with the lab work. Information gained from the laboratory
testing was then utilized for the design of the demo- and full scale units.
The nominal gas flow of the demo unit was 36,000 ACFM. Depending on the fuel fired, the SO2
content of the flue gas varied between 600 - 1200 ppm The flue gas temperature was approximately
250 deg F.
The principle of the NID demo installation is shown in figure 3 below.
taziska. Unit 2, 120MW
Fabric Filter
Penthouse
iuegass>
Irom bSfjr
f Flue gas
'
I
i
v~
W\ ^-. Lime ••
t- { '-iW-^..: : : ;• ; : ;•;-: : i
! i ki;™il* End product : .
i-q S Rcftary valve
jj-'j .••> :. •. ___ ' - ;;;
^i Moisturizetr^:;
ti/ • ; DOOf-
B
1
o
HI
^L
•: p
: ; ' -'K :
bbser
; 3otaryvalve
fi| Compressor
Figure 3: The NID Demo Plant at Laziska Power Station, Poland
-------
The flue gas is taken from a common inlet flue gas manifold into a vertical duct, making up the inlet to
the MID reactor. In the reactor, the flue gas is intimately mixed with wetted dust, which consists of a
mixture of absorbent and recycled material, i. e. reaction products of SO2 and absorbent, mixed with fly
ash. The NTD reactor is connected directly to the FF and the gas thus enters into the set of bags in a
horizontal flow pattern. Upon reaching the bags, the particles are separated from the gas. The flue gas
passes through the filtration media and is then ducted to the flue gas ID fan. The FF is cleaned by pulses
of compressed air, which dislodges dust into the FF hopper. Here make up absorbent is added before the
recycled material is sent to the humidifier, placed immediately below the FF hopper. In the humidifier,
water is added to the recycled material in a controlled proportion to maintain the desired outlet flue gas
temperature. Lime is stored in a 60 ton silo. From the silo the lime powder is transported pneumatically
to the FF hopper, and thus introduced into the flow of recycled material. The lime addition is controlled
via a signal from an SO2 meter at the outlet duct which sets the speed of a rotary feeder at the silo
discharge,
Demonstration test program
Testing of the demonstration plant started in late February 1995. During the test campaigns a set of
different operation cases and parameters were investigated. Performance with CaO vs. Ca(OH)2 was
explored for various coal qualities. An important task was also to confirm turndown capabilities and
pressure drop of the dust dispersion reactor and the mixer performance at high dust throughputs.
Special testing of the fabric filter operation and its performance was carried out as well.
After testing was concluded, the demo plant was thoroughly inspected, searching for corrosion, signs of
dust build-ups etc. or any other type of possible damages to the process equipment etc..
The results of the intense operation and testing of the NTD demo were very encouraging.
The goal stated before the pilot tests started, was to reach at least 80 % SO2 removal efficiency at
realistic operation conditions. This efficiency target was reached with good margins at normal test
conditions; efficiencies in the 90 - 95 % range were easily and consistently obtained.
The fabric filter in the NID demo plant-acts both as a particulate collector and as an additional SO2
absorber. The tests indicated that the fabric filter withstood the very high inlet dust concentration very
well and that it would operate without any problems. Measurements of the outlet dust concentration
showed that the dust emissions are still maintained at a very low level, in spite of the
-------
operation of the MID process. The operation of the humidifier, where lime, recirculated material and
water are mixed together, was confirmed to be satisfactory. The system to disperse the absorbent/dust
mixture in the inlet duct was also working well; no deposits or other signs of malfunctions were found
when inspecting the equipment after operation.
The reagents tested were dry hydrated lime , Ca(OH)2, and quicklime, CaO. There is an advantage to
use CaO directly rather than dry slaked Ca(OH)2, due to the additional cost of the latter absorbent. The
slaking of the quicklime into the hydrate is integrated in the NTJD process and no extra slaking equipment
is needed. However, if used directly in the process, slightly more CaO is needed than when using
Ca(OH)2 The active specie is, however, Ca(OH)2 in both cases, The slaked lime reacts with the sulfur
dioxide in the flue gas and the reduction of S02 occurs both in the inlet duct and in the dust cake
formed on the filter bags of the fabric filter.
Figure 4 shows the SO2 removal efficiency at typical test conditions and when using quicklime, CaO, as
the absorbent. The fluctuations in the SO2 reduction are due to differences in the inlet SO2
concentrations. (The peaks round 13.20 are due to a deliberate stop of the NTD pilot plant.)
Q.
•inn -
•''
ft
i -! •/
>v w\ i
vV\
s A, v/-v"-~*~"''- ^•"-'••v^ -'"ii
1 S02reOurtion ^
312 reduction l'
'
. , _ ,,, - , j
\y '* "-v'" "'*•' ^-v' ~< / ' •"°- *:<>^'?-'\ '•
' V '"' \|N
302 in
SO2out
>•'""" V-^r^,.x>
1
1
•' '.'' ^ '
'"X- ••'
on
^ in
\
- SO2 out, ppm -
2 in, ppm SO2 reduction, c
Figure 4 Example of S02 in- and outlet concentrations and SO2 removal efficiency when using CaO
as the absorbent.
-------
During the pilot test campaign, runs were made with both process water (cooling tower blow-down)
and drinking water quality. The results show that process water has some negative impact on the slaking
process compared to the use of drinking water. The difference can, however, be compensated by adding
some extra lime.
The operation and testing of the NED demo plant at the Laziska power plant was an extremely important
step in the development of the NID process for desulphurization of flue gases from coal fired boilers.
The goal of a minimum SO2 absorption of 80 % was achieved with very good margins. Likewise, the
dust emissions were very low, according to the testing done at site.
The experience gained also strongly indicated that the process can be accomplished in the equipment as
designed; the remarkably small reactor volume and the integrated features of the process were
demonstrated to take place without any secondary problems like dust build ups etc..
Full Scale Application of the NID Technology
General
For the full scale plant, it was decided to install a commercial type dry lime hydrator. Although the
operation on quicklime alone was proven in the demo, it was felt that this additional new process feature
would add unnecessary risk to the project. A flow sheet of the foil scale installation is shown in figure 5
below.
Certain other scale-up and modifications of equipment were necessary for the full scale plant.
The flue gas from the boiler to the FF is transported in two main flue gas ducts, each with its individual
ED flue gas fan. Each main flue gas duct then branches off into two FF compartments, which each can be
isolated by inlet and outlet dampers. This means that there are four separate FF compartments; each of
which is equipped with a hopper for recycled dust from the FF bags. From each of these four hoppers
the recycled dust is fed with rotary feeders to the humidifiers. Finally, after wetting in the humidifiers,
the recycled dust is fed back into the inlet ducts via short air slides. This arrangement allows for partial
load operation by taking compartments/mixers offline. See figure 6 below.
-------
Quicklime is fed from a silo into the dry hydrator, from which the dry hydrated lime is pneumatically
transported to the two FGD units.
^v \ --
Boiler V
Fuel
End
Product
Figure 5. The NID process as configured at the Laziska Power Station in Poland.
Lime is also added to the each of the humidifiers and the waste product is taken out from the FF hopper
via an overflow weir into mechanical conveyors, which also pass through the bottom part of the inlet
ducts The latter as a precaution, should any fallout of dust occur below the injection point of the
recycled material. The conveyors transport the waste material to a end product silo, from which it is
pneumatically conveyed to a landfill or is collected by silo trucks for utilization as a filler in coal mines.
All humidifiers are placed on a common level below the FF hoppers. This means that they are all easily
accessible for inspection and service.
-------
There are two main control loops, one for the addition water to the humidifier for temperature/moisture
control of the outgoing flue gas, the other one for the outlet SO2 concentration based on a feed
forward signal with back trim.
Recycle
Reagent
Fig 6. NTD installation at Unit # 2, Laziska Power Station, Poland
Operational Experience
The commissioning and testing of the full scale plant took place in several steps. In February 1996, the
first full scale humidifier was installed at the NID system at unit #1. A series of aerodynamic testing and
optimization efforts followed. It was concluded that the operation of the FF-hopper/rotary
feeder/humidifier was as expected. Changes were made in the automatic start-up sequence to facilitate a
smooth initial operation after bringing the NID on line. It was further confirmed that the pressure drop
over the unit was fully within expectations from the laboratory model testing and the CFD-modeling. In
August 1996 all four gas paths of unit # 1 were successfully brought on line. The start up of unit # 2,
followed later in the autumn of 1996.
As always when applying a new technology in full scale, some operational problems have appeared.
When analyzing the flow pattern at the FF inlet, it was concluded from the laboratory flow model
testing, that some high velocity areas existed. As a precaution, wear protection plates were added to
-------
minimize the possible adverse effect of dust erosion of the front bags at the FF inlet. However, after
some 1000 hours of operation in the NDD mode, it became apparent that the protected area had to be
increased, since some front bags were found to be damaged by erosion. The problem was alleviated by
installing additional bag shielding; since this arrangement was installed, no further bag damages have
been reported.
Another area which has caused occasional problems is the dry lime hydration with subsequent storage
and pneumatically transport. The dry lime hydrator is per se commercially available, however, the
application of this type of equipment in a flue gas desulfurization system is very scarce.
The problems encountered are caused by the fact that lime quality often varies and that typical operation
of a lime hydrator in the lime industry is manual. Several modifications in this area have been made.
First, the control system of the hydrator has been modified, and is now supplemented by regular manual
sampling of hydrated product. In addition to this, the silo for the hydrated lime has been modified
including a change out of the rotary feeders, supplying the hydrated lime to the pneumatic transport
lines. The rotary feeders originally provided did not seal tightly enough to accurately meter the lime
feed.
Guarantee performance testing was carried out in the spring of 1997 and all guarantee points were met
successfully.
Summary
The NTD dry flue gas desulphurization system has been developed through the phases of idea,
conceptual designs, laboratory testing, field demo, and the first full scale application in a very short
period of time. In spite of the time pressure the targets set for the development work have been
exceeded and the technology is now being commercially implemented. A next generation of NED with
still further optimized features is also under way. Recently ABB obtained an order for a NTD system to
be installed in a 35 MW diesel power plant to be fired with high sulfur oil. This is only one of many
additional applications where the NTD technology will be used in the future.
The NID technology offers a simple, low cost desulphurization system with high performance, minimum
space requirements and easy access. The fact that the combined full scale plant for the removal of fly
ash and SO2 can be fitted into the space of the existing electrostatic precipitators may also be of
particular interest for further application of the NID technology for retrofit situations.
Acknowledgment
The kind support by Elektrownia Laziska in all matters, irrespective if small or big, is gratefully
acknowledged by all ABB personnel involved in the NID development effort.
-------
FGD EXPERIENCE AT POLAND'S RYBNIK POWER STATION: DRY
METHOD WITH HUMIDIFICATION
L. Pinko
Energopomiar
ul. Sowinskiego 3
44-101 Gliwice, Poland
J. Chachula
Rybnik Power Plant
ul. Podmiejska
44-207 Wielopole, Poland
W. Ellison
Ellison Consultants
4966 Tall Oaks Drive
Monrovia, MD 21770, USA
J. Podkanski
Institute of Chemical Engineering, Polish Academy of Sciences
ul. Baltycka 5
44-100 Gliwice, Poland
Abstract
Details are presented of low-cost FGD technology retrofit application and operation on low-
sulfur, high-ash, bituminous coal-fired, 200 MWe units in Poland, erected in the 1970s.
Operating and maintenance experience is presented in achieving 80%+ SOi removal using an
in-tandem system comprising furnace limestone injection coupled with a Polish-developed gas
humidification reactor that employs lime and water supply and a fly-ash and sorbent
recirculation technique. After favorable performance at the 330,000 m3(STP)/h flue-gas flow
level over a three-year period another, full-scale, 200 MWC installation (for 800,000
m'(STP)/h flue gases) for the newly developed gas treatment process begins operation in
1997. The study on the system cost-effectiveness and operational advantages includes focus
on applicability in absence of adequate plot space for conventional wet and semi-dry types of
SOa removal equipment, simple installation and operation, low capital and total costs.
-------
Rybnik Power Plant
Rybnik Power Plant is located in the southern part of Katowice Voivodship, Poland, in the
center of bituminous coal-mining region. It was built in two phases: phase one - 4 units x 200
MWe built from 1972 to 74, and phase two - 4 units x 200 MWe built from 1977 to 79. Table
1 presents the current values of the main parameters describing the plant:
Table 1
Power Plant in Rybnik, Poland. 1996
capacity. MWe 1,625 coal calorific value, MJ/kg 21.41
gross electricity production, GWh 9,042 coal ash content, % 22.90
net electricity production, GWh 8,416 coal sulfur content, % 0.84
operation time of the unit, h/year 6,088 coal humidity (total), % 10.20
availability factor. % 85.40 SO2 emission, t/year 54,515
gross efficiency, % 36.92 NOX emission, t/year 22,706
coal consumption, mln t 4.096 fly ash emission, t/year 11,588
The power plant was designed to bum low-quality fuel. Up to the year 1990, the power plant
burnt coal of calorific value around 20.8 MJ/kg, sulfur content 1.3% and ash content up to
35%.
Since 1990, the quality of the fuel burnt in the power plant has improved significantly. Its
calorific value rose to 22.0 MJ/kg and sulfur content went down to 0.83-0.72%. Right now,
the plant bums around 4 mln tons of coal per year, thus leading to major pollution and threat
to the environment, mainly because of the emissions of fly ash, sulfur dioxide and NOxs.
Better-quality fuel and higher content of water in flue gases due to the FGD plant increased
electrostatic precipitators' efficiency from 97.0 to 99.6%. New burners and optimization of
the burning process allowed the emission of NOxs to be decreased significantly (below 400
mg/Nnr').
The most difficult problem was to curb the emission of S02. The original design of the power
plant did not include any FGD system and there was no space available for additional
installations. However, taking into consideration the fact that the modernized power plant
would work for at least 20-30 more years, the plant's management decided to build an FGD
installation. The main problem was caused by the lack of space, and after a thorough analysis,
the dry method with additional humidification of flue gases was chosen. During the first phase
of the project, the limestone injection into the furnace was used in four 200 MWe units (1992).
Tab. 2 gives overall emissions of SO2 for the period from 1989 to 1996.
Table 2
Emissions of SO2 from 1989 to 1996 in the Rybnik Power Plant, Poland
Year 1989 1990 1991 1992 1993 1994 1995 1996
Emission of S02[tons/year] 96,932 70,176 64,494 45,497 46,434 47,622 48,675 54,515
-------
The increased emission in 1996 in comparison with the year 1995 results from the higher
production of electric energy.
The humidification reactor for 330,000 mJ(STP) of flue gas/h started operation three years
ago. The full-scale reactor for 800,000 m3(STP)/h covering the needs of the entire 200 MWe
unit has begun its operation this year, another one is under construction. The ultimate goal of
the project is to desulfurize flue gases produced in all eight 200 MWe units.
The process
Fig. 1 presents a block diagram of the installation. Limestone is ground in the existing system
using the stand-by capacity of coal mills. Next, the limestone is blown into a furnace also
using the existing system of nozzles. In the furnace, the limestone is decomposed into the
calcium oxide and carbon dioxide. CaO reacts with SC>2 and then, together with the flue gases
and fly ash, passes through a system of air reheaters. Due to increased heat and air
requirements in the system (desulfurized flue gases are heated by the addition of hot air) a
part of flue gases entering the reheaters is taken from above water superheaters in a boiler.
After the reheaters, the flue gases (for 200 MWe units - around 800,000 m3(STP)/h) are
directed to the upper part of a cocurrent reactor 50 m high. Ducts are specially modeled to
ensure uniform flow of the flue gases, fly ash and sorbent particles. Hydrogenated lime and
recycled ash are blown into the duct just before the reactor. At the top of the reactor the flue
gases pass through a Venturi-type tube. In the tube's throat there are several water-spray
nozzles. After the humidification and desulfurization, at the bottom of the reactor, the flue
gases are mixed with hot air and directed to electrostatic precipitators. The fly ash and sorbent
from the bottom of the reactor and a part of the fly ash and sorbent from the precipitators are
recycled.
The processes which occur in the furnace, ducts and in the reactor were thoroughly studied in
order to obtain the most suitable system and an optimum set of operating parameters. A
theoretical analysis of the processes led to a number of design and operational guidelines;
some of them are listed below:
• It is important to blow limestone into the furnace at a place where the temperature is high
enough to ensure fast decomposition of the limestone, but low enough to allow SOa
molecules to enter the pores of CaO particles and react inside them instead of reacting just
on the particle surface (which results in the blocking of the pores).
• Hydrogenated lime which is injected into the duct before the flue gases reach the reactor
should have the highest possible specific surface area.
• The process in the reactor should be conducted at a temperature and humidity close to the
adiabatic saturation conditions.
• It is better to humidify the flue gases with water spray nozzles than by using steam.
• The longer time of droplet evaporation (i.e. the bigger is the mean droplet diameter), the
better is the efficiency of desulfurization in the reactor. The droplet diameter is, however,
-------
limited by the condition of the complete evaporation of the droplets before they leave the
reactor.
The hiaher value of Ca/S ratio, the better is the SC>2 removal; above certain value, however.
an increased Ca/S ratio has little influence on the process efficiency.
The recirculation of the particles allows the use of only unblocked parts of their surface;
much better results of desulfurization can be obtained when the recirculated particles are
regenerated. There are several processes of regeneration available, like grinding the
particles or regeneration with steam; the regeneration can increase the process efficiency
by several per cent.
Design and start-up experience
Several severe limitations had to be taken into consideration during the design process. The
most important of them are as follows:
• Lack of space. The reactor had to be built between a boiler and electrostatic precipitators.
The space available was limited to several meters.
• Ensuring the uniform flow of flue gases, fly ash and sorbent particles. This issue was
important at the inlet to the reactor with ducts placed asymmetrically, inside the reactor and
at the inlet to the electrostatic precipitators. In order to solve these problems detailed
modeling of flue-gas flow was carried out. The investigations resulted in a number of
patented solutions.
• Ensuring the uniform mixing of the desulfurized flue gas with hot air at the bottom of the
reactor and avoiding the rise of the hot gases to the top of the reactor. This was
accomplished by a special construction of the nozzles for hot air injection. Their task is to
mix the hot air with the flue gases and to keep the temperature of the walls at the bottom of
the reactor high enough to prevent the sedimentation of sorbent particles.
• The temperature of the flue gases entering the existing electrostatic precipitators should
always be kept above 95°C (without FGD system this temperature was around 110°C). This
is accomplished by mixing the flue gases with a part of hot air which comes out from the
air reheaters. A detailed heat balance which included kinetic modeling of the reheaters was
derived in order to show that the technique proposed is feasible and that the boiler's heat
and mass balances can still be closed.
• Maximum flue gas pressure drop over the FGD installation should not exceed 1100 Pa.
Such a pressure drop would allow the existing fans to be used. This limitation was in
contradiction to the requirement of flow uniformity and resulted in a special construction
of ducts and mixing devices.
• H2O and H2SO4 should not condense inside the existing stack. Detailed calculations
demonstrated that even under severe winter conditions no condensation would occur.
-------
• Water droplets introduced at the top of the reactor should evaporate before reaching the
reactor's bottom. Special nozzles were chosen which produce droplets of maximum
diameter 150 micrometers (the reactor height is around 50 m).
• It was necessary to use contaminated waste water in the reactor spraying system. Two
streams of water are used: waste water from the ash handling system and brackish water
from the cooling towers.
• It was decided to couple the automatic control of FGD with that of the power production
units - it was therefore necessary to rebuild and enlarge the system of monitoring and
control for the entire power unit.
During the start-up period, several problems were encountered:
• Fly ash recirculated from below the reactor has around 4-5% of moisture, which normally
does not cause any problems. However, if the moisture content increases to 10%, severe
transport problems occur. This issue is directly connected with the temperature imposed on
the reactor by the amount of water sprayed at the top of the reactor.
• Sometimes sedimentation of sorbent particles on the reactor walls occur; special devices
will be installed to avoid this phenomenon.
It was decided to start the R&D work on sorbent activation using various techniques. At
present, the recirculated sorbent particles are not activated.
Optimum variant of the technique proposed and its costs
A cost analysis of several possible techniques for different operating conditions was
performed in order to find the best option for Rybnik Power Plant units. The following main
possibilities were analyzed:
1. Limestone injection into the furnace, humidification with water in the reactor, fly ash and
sorbent recirculation (no injection of hydrogenated lime into the duct).
2. Hydrogenated lime injection into the duct, humidification with water in the reactor, fly ash
and sorbent recirculation (no injection of limestone into the furnace).
3. Limestone injection into the furnace, hydrogenated lime injection into the duct,
humidification with water in the reactor, fly ash and sorbent recirculation.
Options comprising humidification with limestone slurry or sorbent recirculation using the
slurry were excluded from the analysis because of their poor economic and technical
parameters for the power plant studied.
For each option, similar investment outlays were taken, however, different constant and
variable operating costs were assumed. For every option detailed mass and energy balances
were prepared including mass and heat balance for the boiler, air reheater and reactor. The
analysis revealed that the optimum option is option No. 3. Below, we give some details on the
-------
costs of the optimum solution. The prices concern the year 1996 and were recalculated to USS
using the exchange rate USS 1.00 = 2.80 PLN.
200 MWe unit 800.000 m3(STP) of flue gas/h:
coal supply - 94 t/h; S content - 0.8%, 95% of S leaves the boiler with flue gases: coal
calorific value - 21.5 MJ/kg: investment outlays - USS 9.0 mln (for Rybnik Power Plant
conditions: it is estimated that other retrofit installations built under not so favorable
conditions would cost up to USS 14.0 mln); discount rate r - 10%; installation lifetime - 20
years; FGD load factor - 0.66 (5800 hours of operation per year); limestone consumption -
6500 kg/h; hydrogenated lime consumption - 1830 kg/h. electric energy consumption - 0.5
MW; maintenance costs - 0.05 x investment outlays: number of personnel - 7.
For the discount rate r=10% and 20 year lifetime, the levelised annual capital charge factor is
0.117 (no correction for inflation is made) and the levelised annual capital charge is USS
1.053 mln. The sum of levelised annual capital charge along with the fixed and variable
operating costs comprises the total annual costs.
For the variable operating costs, the calculations give USS 1.774 mln/year. The variable
operating costs consist mainly of the costs of sorbent purchase and transport - USS
932,500/year, cost of the heat losses in the boiler (flue gases are heated with the hot air
withdrawn from the reheaters) - USS 203,600/year, calcination loss in the boiler - USS
85.000/year. waste handling system - USS 345,700/year, additional electric energy
consumption by air and flue gas fans - USS 29.300/year, air compression (for water spraying
system) - USS 128.200/year. Fixed operating costs (including maintenance of furnace
injection system) are around USS 369,900/year.
With 80% of desulfurization efficiency, the unit cost of the FGD process is around USS
0.48/kg SCb or 2.8 mills/kWh. Such a relatively low cost could be achieved because of many
factors, of which some were site-specific:
• It was possible to use the existing equipment to unload, transport, ground and blow
limestone into the boiler.
• Spare power of the existing air and flue gas fans was sufficient to cover the FGD
installation's requirements.
• Electrostatic precipitators did not have to be replaced.
• Sorbent reactivity and cost of transport were optimized to select the best vendor.
• Both the dry desulfurization in the furnace and the humidification in the reactor were
optimized in terms of the process conditions, efficiency and costs.
• No royalties were paid and the installation was designed by the power plant's staff.
FGD efficiency - experimental results
-------
The installation was tested under various operating conditions. The efficiency of the dry
process in the furnace and humidification in the reactor were tested separately. The main
results for the coal of around 0.8% sulfur content are as follows:
• The desulfurization efficiency for the dry process in the furnace, for Ca/S from 1.5 to 2.8
falls within the range 20 to 27 %; for Ca/S>2, an increase in the FGD efficiency is slow.
• For the reactor alone, with Ca(OH)2 injection into the duct, with no limestone injection into
the boiler, and for Ca/S from 0.7 to 2.0, the FGD efficiency falls within the range from 60
to 66%; fly ash and sorbent recirculation is here of major importance and can increase the
rate of SC>2 removal by as much as 7%.
• The Ca(OH)2 injection into the duct coupled with the limestone injection into the furnace
can increase the FGD efficiency in the reactor by around 10% in comparison with the
Ca(OH)2 injection into the duct alone.
• If the temperature in the reactor is decreased from 71-71°C to 68-69°C, the FGD efficiency
in the reactor increases by 7-8%.
• For Ca/S from 2.5 to 3.5, the total desulfurization efficiency rises from 71 to 80% (for both
the dry FGD in the furnace and the humidification in the reactor).
Fig. 2 presents the results of the experiments recalculated for a constant temperature in the
reactor (71°C).
Similar tests were performed for coal with the sulfur content 1.3 to 1.6%. For Ca/S from 3.0
to 3.5 the FGD efficiency was around 75 % at 72°C at the bottom of the reactor. Higher
efficiency is possible for an increased Ca/S ratio.
Summary
One of the most recent full-scale, retrofit FGD installation in operation in Poland is described.
The method comprises the limestone injection into the furnace with further humidification of
the flue gases in the reactor. Fresh sorbent (calcium hydroxide) is added to the flue gases
before they enter the reactor. Fly ash and sorbent are recirculated. A full-scale plant designed
to process 330,000 m3(STP)/h of flue gas was built and tested. The first FGD installation for a
200 MWC unit (800,000 m3(STP)/h) is currently being commissioned, another unit is under
construction. The cost optimization was performed to identify the best option for the process.
The FGD efficiency achieved was 80% with costs around US$ 0.48/kg SO2.
The method was developed and implemented by the Rybnik Power Plant's staff with the
participation of the following institutions:
Silesian Technical University, Wroclaw Technical University, Polish Academy of Sciences,
Institute of Non-Ferrous Metals, and ERPRO Rybnik.
-------
BOILER
LIME
BALL MILL "*>
WATP.R
COMI'RKSSKIJAIR
Figure I
FGD Installation at Rybnik Power Station
-------
100-j
90-E
Efficiency of SO2 removal
80 -.
70 \
60-I
50 -_
30 -.
20-.
10-.
Total POD efficiency (furnace + reactor)
FGD efficiency in the reactor (no limestone injection into the furnace)
AAAAA FGD efficiency in the reactor (with limestone injection into the furnace)
FQJJ efficiency in the furnace
\ 2
Ca/S
Figure 2
Experimental Results
-------
FIRST NORTH AMERICAN CIRCULATING DRY SCRUBBER AND PRECIPITATOR
REMOVE HIGH LEVELS OF SO2 AND PARTICULATES
Lloyd L. Lavely
Black & Veatch
Kansas City, Missouri USA
Vera Schild
Black Hills Power and Light Company
Rapid City, South Dakota USA
John Toher
Environmental Elements Corporation
Baltimore, Maryland USA
Abstract
Stringent sulfur dioxide (SO2) and particulate emission limits are required by recent Best
Available Control Technology (BACT) determinations, the 1990 Clean Air Act Amendments,
and state environmental regulations. Semi-dry lime flue gas desulfurization (FGD) systems,
including the Circulating Dry Scrubber (CDS), have become viable options to achieve 95 to
98 percent SO2 removal and particulate emissions less than 0.012 Ib/MBtu for short averaging
periods. Ideally, an FGD system should also achieve high pollutant removal levels for other
regulated, and potentially regulated pollutants, such as mercury.
Earlier studies conducted by Black & Veatch and nearly 2 years of commercial operation of a
CDS and electrostatic precipitator (ESP) at Black Hills Power and Light's (BHPL) Neil
Simpson Unit 2, demonstrate that semi-dry lime CDS/ESP scrubbing technology can meet
these ambitious objectives at low capital and operating costs.
This paper describes the Black & Veatch project for BHPL and the air quality control system
(AQCS) purchased for Neil Simpson Unit 2, which is the first North American CDS and
precipitator.f The paper also describes operating problems encountered and the results of
system operation.
•[The detailed evaluation, vendor selection process, European site visit observations, and supporting related
studies and technologies are described in earlier papers: "Circulating Dry Scrubber (CDS): Cost Effective FGD
for Clean Coal Plants," by Lavely, L. L., K. I. Mastalio, and T. M. Ohlmacher, presented at American Power
Conference, Chicago, Illinois, April 1994, and "A Fully Flexible Air Quality Control System for Stringent SO2 and
Particulate Emission Limitation," by Mastalio, K. I., L. L. Lavely, and T. M. Ohlmacher, presented at Power-Gen
Americas, Dallas, Texas, November 17-19, 1993. These papers also contain information on the provision to add a
limestone furnace injection system to potentially lower reagent costs.
-------
Project Description
BHPL constructed a baseload, mine mouth, 80 MW net coal fueled steam electric generating
unit (Unit 2) at the Neil Simpson Station located at the Wyodak Coal Mine near Gillette,
Wyoming. The plant began commercial operation in August 1995, 4 months ahead of
schedule.
The Babcock & Wilcox (B&W) pulverized coal boiler is designed to burn Wyodak coal and
has a maximum continuous rating (MCR) of 813,466 Ib/h of steam at 1,620 psi. The specified
primary coal properties pertinent to the CDS and ESP, which were furnished by
Environmental Elements Corporation (EEC), are shown in Table 1.
The AQCS design inlet flue gas conditions are shown in Table 2. The turbine generator,
boiler, and AQCS plan and elevation views are shown on Figure 1. The views indicate the
compact size of this AQCS.
Table 1
Wyodak Mine Coal Properties
Typical
Range
Proximate Analysis, As
received,
% by weight
Moisture 30.00 24.00 - 34.00
Ash 7.50 4.00 - 10.00
Volatile matter 30.40 28.09 - 36.99
Fixed carbon 32.10 29.45 - 37.40
Total 100.00
Sulfur 0.85 0.20 - 1.20
Heating Value
Btu per Ib, as received
7,950
Btu per Ib, dry 11,500
Ultimate Analysis, As
received,
% by weight
H2O 30.00
Ash 7.50 4.00 - 10.00
Sulfur 0.85 0.20 - 1.20
Nitrogen 0.70
Carbon 46.70
Hydrogen 3.10
Oxygen 11.14
Chlorine 0.01 0.00 - 0.03
Total 100.00
-------
Table 2
AQCS Design Typical Inlet Flue Gas Conditions
Steam Generator Operating
Conditions, percent of MCR
25 60 80 100
Fuel Heat Input,
million BtU/h
Flue Gas Mass, Ib/h
Volume, acfm
Temp, " F
Nominal
Minimum
Maximum
Paniculate, Ib/h
Maximum
Typical
Particulate, gr/acfm
Maximum
Typical
Sulfur Dioxide,
Ib/h
Maximum
Typical
Minimum
275 635
443,520 829,628
148,105 287,339
200 226
175 201
520 660
2,895 6,684
2,075 4,792
2.280 2.714
1.635 1.946
868 2,005
587 1,356
131 302
821
993,602
354,851
248
223
720
8,642
6,196
2.841
2.037
2,593
1,754
391
1,003
1,187,532
435,973
268
243
770
10,558
7,570
2.825
2.026
3,167
2,143
478
-------
Ume
Storage
Silo
o
Fan
\
v&r
/^BQ.
k
ELECTROSTATIC
PRECIPITATOR
1
Circulating
Dry
Scrubber
"s^ y
NHf^
/ — ^ —
' Inlet
Duct
Tu rbins/Generator,
Boiler Building
B295'-StACk
El 175
Figure 1
Turbine Generator, Boiler, and AQCS Layout and Elevation Views
Emission Limits for Sulfur Dioxide, Nitrogen Oxides, Particulate, and Opacity
Wyoming has a stringent SO2 limitation of 0.20 Ib/MBtu on a 2 hour rolling average. The
Wyodak Mine method of batching coal deliveries to the existing power plants at the Wyodak
site has often provided the complete range of coal quality to the boilers in a span of minutes.
The wide range of sulfur in the Wyodak coal (0.20 to 1.20 percent), the wide variation in coal
quality on a short-term basis, and the stringent SO2 limitation dictated that substantial
operating margins be required in the specifications to ensure permit compliance. The
emission limits for SO2, nitrogen oxides (NOX), paniculate, and opacity at the facility are
shown in Table 3.
-------
Table 3
Emission Limits
Pollutant
Sulfur Dioxide, lb/106 Btu
Nitrogen Oxides, lb/106 Btu
PC Boilers
Particulate, lb/106 Btu
Opacity, percent
Specification
0.17
0.17
0.015
20
Air
Permit
0.20
0.23
0.02
20
Low NOX burners and overn're air systems were considered to be the Best Available Control
Technology (BACT) for the control of NOX in pulverized coal (PC) boilers.
CDS/ESP Process
The CDS/ESP process can achieve SO2 removal greater than 98 percent.1 Figure 2 shows the
simplified CDS absorber with the feed points for hydrated lime, recirculated solids, and
cooling water. The absorber operates as an evaporator and as a chemical reactor for
absorbing gaseous contaminations. The lime addition is independent of the evaporative
capacity of the flue gas. Lime utilization and cost limit the amount of lime addition.
Flue gas is directed to the CDS for scrubbing of acid gases and is then cleaned of particulate
matter by an ESP or fabric filter (FF). It is a dry process which normally introduces hydrated
lime as a dry, free flowing powder and produces a dry, free flowing disposal product. The
utilization of lime is improved by increasing the evaporative cooling of the flue gas and the
recirculation of the calcium in the process for 30 minutes or longer. The lime utilization also
improves as the CDS is operated closer to the adiabatic saturation temperature, and as the
chloride content is increased. Brine or chloride containing water can be used in the
evaporative cooling water to provide the necessary chloride addition for low chloride fuel
applications, such as those at the BHPL Neil Simpson Unit 2. Typically, this temperature is
maintained at a minimum of 30° F above adiabatic saturation to improve material handling
and minimize cold spots and corrosion.
Over 90 percent of the ESP discharged solids (which contain unreacted lime) are recirculated
to achieve high reagent utilization. The recirculated solids are fed to the CDS by using gravity
air slides. This is the key to low cost solids transfer. Fresh hydrated lime is fed into the air
slides by a rotary screw feeder.
Using gravity air slides to recirculate fly ash to the CDS requires a higher ESP elevation than
is typical, which is advantageous for retrofit projects as well as new installations with
significant space constraints. The space below the ESP can be used to enclose other plant
processes, such as lime hydration, boiler water treatment systems, and related equipment such
as air compressors and pumps.
-------
The spray of cooling water into the CDS is independent of the introduction of lime and
recirculated solids from the paniculate collector hoppers. Therefore, surface humidity of
solids within the CDS can be held nearly constant. Low quality water can be used for this
purpose.
The inlet flue gas enters the bottom of the absorber and then flows upward through an
internal venturi nozzle grid composed of either one or several Venturis. More Venturis are
used for larger capacity.
Ash and Absorption
Products to Particulate
Collection Device
Recirculated Ash and
Absorption Products
from Particulate
Collection Device
Bed Removal to
Ash Silo at Shutdown
Figure 2
Circulating Dry Scrubber
Process Control
The simplified process control (Figure 3), which effectively consists of three major control
loops for fully automatic operation, is in principle simple and reliable. All three control loops
operate independently of the other.
• SO2 Control-The initial feed rate of hydrated lime is determined by the amount of
SO2 in the inlet flue gas. The actual feed rate of hydrated lime is constantly
controlled by an SO2 signal from the outlet flue gas.
• Temperature Control-The gas temperature leaving the absorber directly controls the
amount of flue gas cooling water which will be injected into the CDS through high-
pressure single fluid flow nozzles.
• Solids Discharge-The amount of ash discharged from the system is controlled by the
solids loading of the absorber. This is measured by the differential pressure across
the absorber height. Basically, the resultant solids are discharged from the system at
the same rate that hydrated lime, fly ash, and SO2 enter the system. This maintains a
constant mass of solids in the system at all times.
-------
Operating Results
The EEC contract agreement with BHPL required emissions testing during the initial
commercialization of the unit and repeat testing 1 year later. Operating costs were to be
determined during the initial tests, but the true evaluation of these costs versus respective
guarantees would not occur until the repeat tests. This allowed for a grace period of 1 year
for system tuning and required equipment modifications. The year of operating and
equipment optimization before a final assessment of operating costs was agreed to by all
parties because this plant introduced this technology in North America.
Initial Operating Results
Commercial operation was achieved in August 1995, and initial emissions tests were conducted
in September. The emissions tests identified a problem associated with a construction error in
the interior of the ESP. This caused unusually high emissions of particulate matter. A
modification quickly remedied the problem, and a retest was passed successfully in October
1995. Results of the October retest are summarized in Table 4.
Absorber Pressure Drop Controls Ash Recirculation
Gas Temperature Controls Humidrfication Water Row Rate
Gas Inlet and Stack SO, Concentration Controls Fresh Reagent AddrSc
Figures
CDS/ESP Control Schematic
To Disposal
-------
Table 4. Results of October 1995 Test for Emissions
Parameter
Paniculate Matter
Sulfur Dioxide
Opacity
ESP 100%
Energized
< 0.010 Ib/MBtu
<0.141b/MBtu
<5%
ESP 10% De-
energized
< 0.010 Ib/MBtu
<0.141b/MBtu
<5%
ESP 20% De-
energized
< 10%
The total operating cost was within the guarantee value, and individual consumables were
significantly below the guarantee values. However, lime consumption was higher than had been
predicted and became the focus of considerable efforts during the next year. A 6 Month availability
demonstration run followed a successful initial performance retest. This test consisted of a 6 month
duration test to demonstrate 100 percent availability of the scrubbing system when the boiler
operated. The availability run began in November 1995 and was scheduled to end with a low load
demonstration test in April 1996.
Meanwhile, a close examination of the SO, control logic began, including the continuous emissions
monitoring system (CEM). Figure 3 shows the SO2 control logic, which is comprised of a
combination feed-forward and feed-back signal from the CEMs to regulate the introduction of fresh
hydrated lime into the flue gas. Each CEM extracts a sample of flue gas and draws it through an
umbilical cord to a remotely located analyzer. This resulted in a 4 minute lag time in the SO2
signal to the controller, which caused process instability. Compensation for the lag time proved
impossible; therefore, a new analyzer was ordered which would reduce this lag time to less than
60 seconds. It would take 4 to 5 months to install and begin using the new stack analyzer.
On a parallel path, coal samples were analyzed for chlorine content and were found to be near the
zero end of the specified range. A simple brine additive system was installed to study the effect on
lime consumption. This additive system was operating by February 1996. Lime consumption
improved with the brine addition, but was limited by the degree of increased tackiness of the
circulating solids within the reactor. An extensive data collection system was initiated as a joint
effort of BHPL and EEC. All pertinent plant signals were collected in a data file by the plant
distributed control system, and EEC was allowed on-line access to this data file via their own
desktop PC and software. A modem allowed on-line analysis of this data by office personnel in
Baltimore. This team approach to resolving the lime utilization problem proved to be effective in
determining the multiple sources of difficulty.
Meanwhile, problems developed with the water spray nozzles in the CDS vessel. A defective seal
design allowed a dribble of water between the body and the end cap of the nozzle. This in turn
caused deposits of moist material to build up at the tip of the water nozzles, which would
eventually interfere with the water spray pattern from the nozzle. On one occasion, the CDS
reactor had to be cleaned during an outage to remove solids buildups. Afterward, routine on-line
cleaning of the nozzles prevented further buildup from shutting down the system. The manufacturer
designed modifications with the assistance of EEC and BF1PL. Various modifications were tested,
and a final design was successfully implemented in February 1996. Operations throughout February
and March 1996 proved that the moist ash deposits had been eliminated and the water nozzles were
operating reliably.
-------
The CDS/ESP had to operate successfully at low level before the end of April 1996 to complete the
demonstration run requirements. Previous limited testing showed no CDS operating difficulty at
low load, but the boiler had been tuned extensively over the past several months, and the quantity
of flue gas flowing to the CDS had been reduced significantly. The April test proved that the CDS
could not reliably operate under the new gas flow rates associated with low load. The low load
operating restrictions and lime consumption were the only major disappointments in an otherwise
successful 6 month run that demonstrated 100 percent availability on the part of the CDS
technology. A low load corrective plan was devised for implementation during the scheduled
outage beginning in May 1996. The plan consisted of modifying the scrubber vessel to effectively
increase the gas velocity through the Venturis at low load.
Modifications were completed on schedule during the May-June 1996 outage, and the unit was
returned to duty in mid-June. Operations at low load improved, but operational flow patterns within
the CDS vessel changed. Relocation of the water spray nozzles proved nominally successful, but
efforts to re-establish operational flexibility of the order that existed prior to the outage were
unsuccessful. The location of the water spray nozzles was and remains critical to the successful
operating flexibility and performance of the CDS. The vessel modifications were reversed in
September 1996, and a new plan was devised to deal with low load operation. This plan was
implemented during the next scheduled unit outage in May 1997. After September 1996, the plant
returned to the baseload operation that was established shortly after startup.
The new stack CEM became operational after the May 1996 outage. The combined impacts of the
new CEM and the brine addition showed a lime consumption improvement of 20 percent or more
since the initial measurements were taken in November 1995. This was accomplished with low
brine injection rates of less than one gpm of saturated solution. However, significant fluctuations in
lime utilization were still being experienced, which seemed to correlate well with the coal sulfur
content. As a mine mouth unit, the coal analysis varied as the mine worked the coal seam from
north to south. The nature of the mining operation produced long periods of stable coal analysis
followed by abrupt changes in the CDS inlet SO2 between 1.2 and 2.2 Ib/MBtu. Subsequent to the
changes in coal sulfur content, the CDS process would experience a similar and abrupt change in
lime utilization. These abrupt changes occurred routinely about every 2 to 3 months.
Final Operating Results
While efforts were underway to better understand the nature of these fluctuations in lime utilization,
the end of year 1 performance tests were scheduled for December 1996. Results of these tests are
summarized in Table 5.
-------
Table 5
Results of Year-End Performance Tests
Parameter
Particulate Matter
Sulfur Dioxide
Opacity
ESP 100%
Energized
< 0.010 Ib/MBtu
< 0.14 Ib/MBtu
<3%
ESP 10% De-
energized
< 0.010 Ib/MBtu
< 0.14 Ib/MBtu
<5%
ESP 20% De-
energized
< 10%
Except for lime, emissions and operating costs were well below guaranteed values. Additionally,
there is no unusual corrosion evident in the CDS/ESP system through the May 1997 scheduled
outage. The December 1996 test coincided with an abrupt increase in coal sulfur content, and lime
utilization was unusually high. On an individual basis, the I.D. Fan static pressure loss was
28 percent below guarantee, parasitic power load was 24 percent below guarantee, and water
consumption was 6 percent below guarantee. While efforts continued to determine the cause of
poor Lime utilization, EEC and BHPL mutually agreed that a 1 hour test on any given day was truly
not representative of the average cost of lime for the unit. It was mutually agreed to conduct a
30 day test period in March 1997. The test concluded on April 5 and the 30 day average was
documented to be 15 percent above the guarantee on a utilization basis. When the calculated lime
cost, based on this 30 day test, was added to the other operating costs, as documented during the
December 1996 tests, the contractual obligation between BHPL and EEC for the total consumables
cost was met.
Two issues now remained: low load operation and variations in lime utilization. EEC, BHPL, and
Black & Veatch teamed again to investigate and propose solutions. The product knowledge of
EEC, operational knowledge of BHPL, and engineering expertise of Black & Veatch were applied
to these issues.
Low Load
A means to provide supplemental gas flow to the CDS reactor inlet at low loads was installed
during the May 1997 outage. A low load test in June 1997 was successful. It proved to be cost
effective, simple to operate, and involved use of existing plant equipment with minor modifications.
EEC plans to propose this solution for future plants where unit turn-down is an issue.
Lime Utilization
The lime utilization variations were eventually traced to the hydration subsystem. A portion of the
lime being fed to the hydrator was entering the desulfurization process through the hydrator vent
system as un-hydrated lime (CaO). Lime in this form does not react with SO2 as readily as
hydrated lime [Ca(OH),]. The problem worsened as the load on the hydrator was increased, or as
the machine ran more often. This confirmed the earlier correlation with increased coal sulfur
content. Design modifications to further improve lime utilization are the subject of ongoing
discussions between EEC and BHPL.
Operating and Maintenance (O&M) Labor
10
-------
O&M labor costs are less than originally predicted. BHPL has maintained records for all time
charged against operating and maintaining the equipment. These records cover the CDS, ESP, ash
recirculation airslides and blowers, as well as the lime handling equipment. Time spent by
operations in the calendar year 1996 was equivalent to 1.35 man-years based on 2,080 hours in a
man-year and operation of 24 hrs per day for 7 days a week. Time spent by the maintenance staff
in calendar year 1996 was equivalent to 1.25 man-years based on 2,080 hours per man-year
(40 hours per week). The combined O&M hours are 16 percent below the guaranteed values for
the first year of operation. The first year of operation was expected to require more attention from
operations and maintenance personnel due to learning curve activities with this technology.
Records to date indicate the O&M labor rate for 1997 is less than that in 1996. This was one of
the more attractive features of this technology when BHPL was deciding whether to use a spray
dryer/baghouse or the CDS/ESP in 1992. The low O&M cost predictions have been fully realized
through the first 2 years of commercial operation.
MERCURY CONTROL
The CDS technology has other applications that are being studied by the EEC advance air quality
research department. This work was completed under a U.S. Department of Energy SBIR grant.ft
The work shows enhanced effectiveness of iodine impregnated activated carbon in a fluidized bed
in removing elemental mercury. A 200 CFM laboratory unit was used in the evaluation. EEC has
now received the Phase II award to conduct a 1,100 CFM pilot test using a slip stream from Public
Service Electric and Gas Company's Mercer Station.
Fine particles and elemental mercury vapor can be significantly reduced by passage through a
fluidized bed of fly ash and activated carbon. The elemental mercury vapor concentration can be
effectively reduced to near zero by adding about 1 percent of iodine impregnated carbon to the bed
weight. The laboratory and field pilot units simulate EEC's commercial size circulating dry
scrubber process.
During the DOE study, the inlet elemental mercury concentration was 50 |ig/m3. After initial iodine
impregnated carbon addition, the outlet mercury dropped to almost zero for more than 2 hours,
without additional carbon injection. In addition, the bed held the outlet mercury to below 20 ng/m3
after the mercury had saturated the carbon, due to continued adsorption by the fly ash. The fly ash
alone reduced the outlet mercury concentration to 25 ug/m3. After the activated carbon was added,
the outlet mercury decreased to 3 ug/m3. The carbon utilization was 1250 gm of mercury removed.
If the bed fly ash adsorption is also counted, the utilization was 770 gm carbon per gram of
mercury removed.2
These pilot results and the associated costs need to be determined on the commercial size units.
However, a confidential test report on a full scale 2 x 85 MW commercial utility plant with a
ttFor more information on this topic, read "A Circulating Fluid Bed Fine Paniculate and Mercury Control
Concept," by D. J. Helfritch, P. L. Feldman, and S. J. Pass, published in the proceedings of the August 1997 EPRI-
DOE-EPA Combined Utility Air Pollutant Control Symposium.
11
-------
circulating dry scrubber/ESP on each unit, showed an inlet mercury concentration of about
34 |jg/dscm that was reduced to about 13ug/dscm with a normal lime/fly ash bed used for SO2
removal.
CONCLUSION
The commercialization of the CDS technology in North America has been successful due to a
strong teaming environment among all parties involved in the Neil Simpson Unit 2 project. After
nearly 2 years of commercial operation, most operating issues have been resolved. EEC remains
actively involved in these efforts. Many adjustments of the modifications were completed either
online or during the scheduled annual outages. The emissions performance capability is better than
guaranteed and significantly better than specified. O&M costs are below guaranteed values except
for lime. The presence of chlorides is beneficial to the lime consumption of the CDS process. The
Lime consumption may be further reduced through the joint efforts of EEC and BHPL.
The power generation industry is facing unprecedented pressure to reduce O&M cost as a result of
ongoing deregulation. New and pending environmental regulations may pose further challenges to
the industry as emissions of NOX, SOX, particulates, and mercury are regulated more heavily. This
CDS technology, with its unique characteristics, provides a cost-effective way to meet these
challenges.
REFERENCES
1. Moore, S. R., J. G. Toher, and Dr. Harold Sauer, Lurgi Gmbh, "Update of the EEC/Lurgi High
Efficiency, Dry Circulating Fluid Bed Scrubber Experience," presented at the EPRI Conference
on Improved Technology for Fossil Fueled Power Plants—New and Retrofit Application,
Washington, DC, March 1, 1993.
2. DOE SRIR Grant No. DE-FG02-95ER81968 - Phase I Final Report.
12
-------
B&W'S E-LIDS™ PROCESS - ADVANCED SOx, PARTICULATE,
AND AIR TOXICS CONTROL FOR THE YEAR 2000
Deborah A. Madden*
Michael J. Holmes
McDermott Technology, Inc
Research and Development Division
1562 Beeson Street
Alliance, Ohio, 44601
Abstract
Babcock & Wilcox (B&W) is currently developing the Enhanced Limestone Injection with Dry
Scrubbing (E-LEDS™, patent pending) system to be capable of reducing SOx and particulate emissions
significantly below that allowed under the current New Source Performance Standards (NSPS)
while addressing the concerns of solid waste generation and air toxics regulation. The work is
being performed as an integral part of B&W's development of an advanced low-emission boiler
system in a project entitled, "Engineering Development of Advanced Coal-Fired Low Emission
Boiler Systems (Combustion 2000 - LEBS)." The program is sponsored by the U.S. Department of
Energy. The overall goal of the DOE's program is to dramatically improve environmental
performance and thermal efficiency of conventional coal-fired power plants. The B&W E-LIDS™
process is a limestone-based, furnace injection/dry scrubbing SO, removal process. The process
comprises the cost-effective integration of three commercially-proven flue gas cleanup technologies:
furnace limestone injection, dry scrubbing, and pulse-jet fabric filtration. Specific LEBS goals that
are addressed by the E-LIDS™ system are l)sulfur dioxide emissions less than 0.10 Ib/SO, MBtu,
2)particulate emissions less than 0.005 Ib particulates/MBtu, 3)air toxics emissions significantly
reduced, and 4)solid by-product minimized and/or utilized. This paper introduces results of 10
MWe E-LIDS™ testing performed in B&W's world-class Clean Environment Development Facility
to demonstrate the LEBS project SO2 removal goal of 98% under cost-effective operating conditions.
Air toxics measurements of mercury, trace metals, and acid gases are also covered.
B&W'S Enhanced Limestone Injection Dry Scrubbing (E-LIDS™) Process
The B&W E-LIDS™ process - shown in Figure 1 - is a limestone-based furnace injection/dry
scrubbing SO, removal process. The process actually comprises the cost-effective integration of
three commercially-proven flue gas cleanup technologies: furnace limestone injection, dry scrubbing,
and pulse-jet fabric filtration. Sulfur dioxide removal occurs in the boiler furnace and convection
pass, in the dry scrubber, and in the fabric filter. Limestone is pulverized and injected as a dry
powder into the flue gases in the upper furnace cavity of the boiler. Upon injection, the limestone
undergoes calcination to form lime, a portion of which reacts with SO2 in the flue gases forming
calcium sulfate. The flue gases exiting the boiler then pass through a dry scrubber reactor where
they are contacted by a slurry containing calcium hydroxide. In the dry scrubber, the flue gases are
-------
cooled and humidified to conditions near the water saturation temperature - commonly referred to
as operation at a "low approach to saturation temperature" Under these conditions, a portion of the
SO, in the gas reacts with the calcium hydroxide. Water contained in the slurry droplets evaporates
as the droplets pass through the reactor vessel, and the products leave the dry scrubber as a dry
powder still suspended in the flue gases.
Participate
Collection
Limestone
Boiler
Figure 1 - The E-LIDS™ Process
Finally, the flue gases enter the pulse-jet fabric filter (baghouse) where coal fly ash, spent sorbent,
and unreacted sorbent particles are collected. The use of a baghouse is a key feature of the E-
LIDS™ process because of the additional SO2 removal it yields as the flue gases pass through the
sorbent-containing filter cake on the filter bags. The portion of the solids from the particulate
collection device are recycled to a reagent preparation system to produce calcium hydroxide slurry
for the dry scrubber. The remainder of the solids collected are conveyed to disposal and/or utilized.
A new feature of the process is the addition of a simple particulate collection device upstream of the
dry scrubber to remove particulate matter from the flue gases. The solids are sent to the dry scrubber
feed slurry reagent preparation system.
The furnace limestone injection process facilitates the cost-effective use of a dry scrubber for
downstream SO2 removal by: l)permitting the use of limestone as the sorbent (as opposed to the
more expensive lime used in most dry scrubbing processes), and 2)by reducing the inlet SO2
concentration to the dry scrubber through the in-furnace removal of SO,. This latter fact permits
the E-LIDS™ process to be applied to units firing high-sulfur coals by lessening the amount of
calcium needed in the dry scrubber, thereby reducing the heat needed to evaporate the water contained
in the sorbent slurry fed to the dry scrubber.
-------
10 MWe (100 MBtu/hr) E-LIDS™ INTRODUCTION
The main objective of the 100 MBtu/hr test series was to successfully scale-up the E-LIDS™ system
from 5 MBtu/hr and confirm its SO2 and air toxics removal capabilities. Previously it had been
shown that at 5 MBtu/hr the E-LIDS™ system achieved 98% SO, removal (Figure 2) and greater
than 90% mercury removal (Table 1) under cost-effective conditions.[I>2J E-LIDS™ scale-up was
carried out in B&W's 100 MBtu/hr Clean Environment Development Facility (CEDF) in June and
September 1996.
Furnace
Furnace +• Scrubber
Figure 2 - 5 MBtu/hr E-LIDS™ SO, Removal
Stream
Furnace Outlet
Dry Scrubber Outlet
Baghouse Outlet
Furnace Outlet
Dry Scrubber Outlet
Baghouse Outlet
Furnace Outlet
Dry Scrubber Outlet
Baghouse Outlet
Phase I
5 MBtu/hr LIDS
April 1994
Total Hg
Kg/Mm5
5.00
2.49
0.047
4.91
2.33
0.067
6.19
2.17
0.40S
Hg Removal
Across System
%
99.1 %
98.6%
93.4 %
Phase II
5 MBtu/hr E-LIDS™
January 1995
Total Hg
jig/Nm1
19.9
4.90
2.73
16.5
3.49
1.16
17.9
3.58
O.SOO
Hg Removal
Across System
%
86.3 %
93.0 %
95.5 %
Table 1 - 5 MBtu/hr E-LIDS™ Mercury Removal
-------
B&W's Clean Environment Development Facility (CEDF)
B&W's CEDF - Figure 3 - located at the McDermott Technology, Inc, Research & Development
Division in Alliance, Ohio, is an integrated state-of-the-art combustion and emissions testing facility
that offers unique research, development and demonstration capabilities. The CEDF's prime directive
is to improve the environmental performance of current and future power plants.
LJme slurry tank
Figure 3 - B&W's Clean Environment Development Facility
The CEDF is designed for a heat input of 100 MBtu/hr (approximately 10 MWc equivalent) when
burning a wide range of pulverized coals, fuel oil, and natural gas. The facility integrates combustion
and emissions testing capabilities to facilitate the development of the next generation of power
generation equipment. The furnace is designed to operate with a single 100 MBtu/hr burner or
multiple wall-fired burners. It has been carefully designed to yield combustion zone temperatures,
flow patterns, and residence times representative of commercial boilers. In order to provide maximum
flexibility and control, separate fans and air heaters are used for supplying the pulverizer, primary
(coal conveying) air, and secondary air. The use of an indirect pulverized coal feed system in
conjunction with the separate air supplies decouples pulverizer and burner operation, and permits
operation over a wide range of coal types, air-to-fuel ratios, fuel moisture contents, and coal particle
size distributions.
Boiler convection pass and air heater simulators maintain representative conditions through the
entire boiler system to facilitate studies of air toxics capture in the E-LIDS™ dry scrubber and
baghouse. Representative gas phase time-temperature profiles and surface metal temperatures are
maintained throughout the convection pass. Convection pass metal temperatures are maintained by
way of a novel double-walled tube design.
Following the air heater, the flue gas enters a vertical dry scrubber. An atomized slurry is introduced
through a single B&W DuraJet™ atomizer located to provide uniform spray coverage in the vessel.
The B&W DuraJet™ atomizer is used in commercial dry scrubbing and humidification systems.
The atomizer not only provides a finely atomized slurry, but also acts as a mixer to ensure intimate
-------
contact between the hot entering flue gas and slurry, maximizing SO, removal and drying. Flue gas
along with dried paniculate exiting the dry scrubber is ducted to a pulse-jet fabric filter baghouse.
The baghouse consists of six modules each containing 42 full-size bags for a total of 252 bags in the
baghouse. The air-to-cloth ratio is adjustable by blanking off modules. The pulse-jet cleaning
system is designed to permit either on-line or off-line cleaning in either manual or automatic operating
modes. For additional flexibility, in the automatic mode, the fully adjustable cleaning cycle may be
initiated on either baghouse pressure differential, timed, or combined pressure differential/timed
basis. The solid by-product from the bags is transferred to disposal and/or utilization.
In order to complete E-LIDS™ testing at 100 MBtu/hr a furnace limestone injection system,
mechanical particulate collector, and a recycle slurry preparation system were added to the CEDF.
These systems along with the existing furnace, dry scrubber, and pulse-jet fabric filter completed
the 100 MBtu/hr (10 MWe) E-LEDS™ system.121
Successful 10 MWe (100 MBtu/hr) E-LIDS™ Scale-up
The first 100 MBtu/hr E-LIDS™ SO, test series was completed in June 1996. The objectives of the
test series were to achieve the SO, emission target level of 0.10 Ib/MBtu (98% SO, removal) and
perform testing to obtain SO2 performance curves. Both of these objectives were accomplished
during the test series. Figure 4 shows the total SO2 removal that was achieved at the outlet of each
of the E-LDDS™ unit operations. At the outlet of the furnace limestone injection system, 38% SO2
removal was achieved. At the outlet of the dry scrubber, 77% SO2 removal was achieved. At the
outlet of the baghouse (or the E-LIDS™ system outlet) 99% SO2 removal was achieved. The
approximate operating conditions for this test included: l)furnace stoichiometric ratio (Ca/S) -1.4
mol Ca/mol S, 2)approach to saturation temperature - 10°F, 3)slurry percent solids - 30-35%, and
4)limestone grind - <10 microns mass mean diameter.
Another way to express E-LIDS™ SO, capture is by the percentage of the total SO, (created from
the coal) that is removed with each unit operation or that goes up the stack. Figure 5 shows that for
2578 ppm SO2 produced from the coal, SO, capture occurred as follows (for the same conditions as
the data shown in Figure 4): 38.0% (967 ppm) in the furnace, 38.8% (989 ppm) in the dry scrubber,
22.6% (575 ppm) in the baghouse, and 0.7% (17 ppm) went up the stack.
Furnace & Dry Scrubber E-UDS
Figure 4 -100 MBtu/hr E-LIDS™ Total System SO2 Removal - %
Fine Limestone
-------
Figure 5 -100 MBtu/hr E-LIDS™ SO2 Removal by Unit Operation - %
Fine Limestone
Figure 6 shows the SO, removals obtained versus approach to saturation temperature (T^) for the
dry scrubber, baghouse and total E-LIDS™ systems. The figure clearly shows that low operating
temperatures enhance SO, capture in the dry scrubber, baghouse, and therefore, the overall E-LIDS™
systems. The high baghouse SO2 removal achieved at low dry scrubber/baghouse operating
temperatures is a necessary ingredient for the E-LIDS™ final SO2 emissions trim.
IUU
90
80
SS 70
15 60
>
§50
HI
K 40
-------
A second round of 100 MBtu/hr E-LEDS™ tests was completed during the Fall of 1996. Goals of
the test series included:
Coarser limestone grind. It is necessary to achieve 98% SO, removal while using a coarser
limestone for injection. This goal is important to the process in order to improve the E-
LIDS™ system economics and operability. A coarser limestone will do the following:
l)decrease the auxiliary power necessary to grind the limestone, hence, increase the cost-
effectiveness of the E-LIDS™ system and 2)decrease the amount of convection pass fouling
associated with limestone adhering to the tubes.
• Air toxics measurements. Air toxics measurements were made at the inlet to the dry scrubber
and at the stack during E-LIDS™ operation. Measurements of mercury, trace metals, and
acid gases were be collected to further characterized the E-LIDS™ system's air toxics removal
capabilities.
Ultra-high SO2 Removal Achieved with Coarser Limestone
The LEBS SO2 emissions goal of 0.10 Ibs SO2/MBtu (98% SO2 removal) was achieved during 100
MBtu/hr E-LIDS™ tests performed with a coarser limestone with similiar operating conditions.
Figure 7 shows the total percent of SO2 removed at the outlet of the furnace, dry scrubber, and
baghouse. At the outlet of the furnace limestone injection system, 32% SO, removal was achieved.
At the outlet of the dry scrubber, 62% SO2 removal was achieved. At the outlet of the baghouse (or
the E-LIDS™ system outlet) 99% SO, removal was achieved. The approximate operating
conditions for this test included: l)furnace stoichiometric ratio (Ca/S) 1.40 mol Ca/mol S,
2)approach to saturation temperature - 10°F, 3)slurry percent solids - 30%, and 4)limestone grind >
15 microns MMD.
Furnace Furnace & Dry Scrubber
Figure 7 -100 MBtu/hr E-LIDS™ Total System SO2 Removal - %
Coarser Limestone
-------
Figure 7 shows that for 2363 ppm SO, produced from the coal, SO, capture occurred as follows:
31.9% (854 ppm) removed in the furnace, 30.5% (699 ppm) removed in the dry scrubber, 37.2%
(801 ppm) removed in the baghouse, and 0.40% (9 ppm) went up the stack.
Figure 8 -100 MBtu/hr E-LIDS™ SO2 Removal by Unit Operation - %
Coarser Limestone
Figure 9 shows the SO, removal obtained versus approach to saturation temperature (T J for the dry
scrubber, baghouse and total E-LIDS™ systems while using a coarser limestone. Once again, the
figure clearly shows that low operating temperatures enhance SO2 capture in the dry scrubber,
baghouse, and therefore, the overall E-LIDS™ system.
100
90 -
80 -
70 -•
; 60 r
i
: 40
] 30
20
10
5 10 15 20 25 30 35
Approach to Saturation Temperature (Tas) - °F
40
Figure 9 -100 MBtu/hr E-LIDS™ SO2 Removal % versus T^
Coarser Limestone
-------
E-LIDS™ Air Toxics Removal
The CEDF is an excellent test bed for air toxic measurements due to the careful control of the flue
gas cooling rate through the boiler convection pass and air heater simulators which provides a gas
time-temperature profile that is representative of commercial units. A two-stage flue gas cooling
process is used to emulate the desired time-temperature profile and tube metal surface temperatures.
CEDF benchmarking measurements were performed during B&W's Advanced Emission Control
Development Program (AECDP)'31 testing to characterize the emissions from the boiler. Air toxics
emission verification was achieved through comparison of the air toxics measured from the CEDF
with the emissions predicted by the trace element content in the coal and the draft emission
modification factors (EMF) established by the U. S. Environmental Protection Agency (EPA).141
The predicted emissions based on the coal analysis and EPA EMFs for each trace species are compared
in Figure 10 to the emissions measured from the CEDF boiler firing a bituminous, high-sulfur Ohio
coal under full load conditions. The similarity between the predicted and measured emissions
indicates that the HAPs generated from the CEDF are representative of commercial units firing
bituminous coal.
As Be Cd Co Cr Hg Mn Ni Pb Sb Se
D Measured • EPA EMF Predictions
Figure 10 - CEDF Representative Uncontrolled Hazardous Air Pollutant Emissions
Mercury. Mercury, in particular, is the subject of intensive research due to its known buildup in the
atmosphere, subsequent deposition in lakes, and potential human health impacts. Because it seems
likely that mercury emissions will be regulated in some manner in the future, measurements of
mercury emissions from the E-LIDS™ process have been characterized throughout the LEBS
project. Initial mercury characterization of the LIDS and E-LIDS™ processes were conducted by
Frontier GeoSciences (MESA method) at 5 MBtu/hr (Refer back to Table 1). These previous LEBS
pilot-scale results showed that the E-LIDS™ system was capable of removing greater than 95% of
the mercury entering the dry scrubber.
E-LIDS™ mercury measurements were made with the Ontario Hydro method during 100 MBtu/hr
testing in the CEDF. The Ontario Hydro method applies to the determination of particulate and
-------
gaseous metals emissions from industrial, utility, and municipal sources. The measurements were
made before the E-LIDS™ dry scrubber and at the stack. Table 2 lists the 100 MB tu/hr E-LIDS total
mercury removal results. The greater than 95% mercury removal obtained at 5 MBtu/hr with the
MESA method was confirmed at 100 MBtu/hr with the Ontario Hydro method. These results
showed that the E-LIDS™ mercury removal is not an artifact of the facility or the measurement
method. Figure 11 presents the E-LIDS™ mercury emissions graphically.
Total Mercury Data Summary Avg = 99.4
OH#1
OH #2
OH#2A
OH #3
OH#3A
OHM
Ib/trillion Btu
Uncontrolled
12.378
12.378
12.378
12.378
12.378
12.378
Ib/trillion Btu
DS Inlet
0.94777
2.88872
4.98302
1 .93703
1 .60578
0.76554
Ib/trillion Btu
Stack
0.09065
0.09006
0.09006
0.05488
0.05488
0.06671
E-LIDS Total Removal
based on Coal
%
99.3
99.3
99.3
99.6
99.6
99.5
Table 2 -100 MBtu/hr E-LIDS™ Mercury Removal by Ontario Hydro Method
1000.00
0.01
Figure 11 -100 MBtu/hr E-LIDS™ Mercury Emissions
Trace Metals. Many of the trace elements targeted as potential Hazardous Air Pollutants (HAP's)
of interest condense onto the particulate as the flue gas cools through backend emissions control
equipment. Therefore, particulate removal systems have the potential to remove large amounts of
air toxic substances associated with the particulate matter. A pulse-jet fabric filter operating at low
temperatures is an integral part of the B&W E-LIDS™ system. This gives the E-LIDS™ process two
advantages in terms of air toxics removal: l)fabric filtration has been shown to be extremely
-------
effective in removing particulate and therefore many trace metals, and 2)the E-LIDS™ dry scrubber
and baghouse are operating at low temperatures creating a higher potential for trace metals to
condense out of the flue gas.
Although high levels of trace metal removal was predicted for the E-LIDS™ process, actual
measurements were not made at 5 MBtu/hr. Therefore, it was necessary to make measurements at
100 MBtu/hr in the CEDF to confirm that the predicted high trace metal removal did indeed occur
with the E-LIDS™ process. The Ontario Hydro method was used to obtain trace metal
concentrations. Measurements were made at the cyclone exit (dry scrubber inlet) and the stack.
Acid Gases. Acid gases have come under scrutiny primarily because of the large quantity of these
substances emitted from electric power utilities. If acid gas regulations are required, coal-fired
plants equipped with wet or dry FGD have been shown to achieve over 90% removal of HC1.
Another form of HC1 and HF control may be furnace sorbent injection. The E-LIDS™ acid gas
capture efficiency was investigated at 100 MBtu/hr through the use of EPA Method 26A.
Figure 12 summarizes the trace metals and acid gas removals for the E-LIDS™ process.
As Ba Be Cd Co Cr Mn Ni Pb Hg Se Cl Fl
Figure 12 - E-LIDS™ Average HAP's Removals
E-LIDS™ versus State-of-the-Art Wet Flue Gas Desulfurization (WFGD)
B&W's E-LIDS™ system provides superior SO2 removal at a lower reagent cost, lower capital
cost for equipment and with lower auxiliary power requirements than conventional technologies.
Although regenerable scrubbing processes are being developed, none are expected to be
commercially ready until well after the year 2000 primarily due to difficulty in developing a
reagent material which is both economical and tough enough to provide an acceptable useful life.
Currently, the most widely applied technology for the desulfurization of coal flue gases is non-
regenerable wet scrubbing, with better than 83% of the installed base (1995), and more than 90%
-------
of the identified new scrubbing capacity. Limestone and lime are the two most common non-
regenerable wet scrubbing reagents, with about 71% using limestone and 16% using lime. Over
the last five years, approximately 65 GWe of wet flue gas desulfurization (WFGD) systems have
been sold around the world. Since the passage of the 1990 U.S. Clean Air Act Amendment, new
capacity in the U.S. has been dominated by the limestone forced oxidation wet scrubbing (LSFO)
process. Phase I FGD installations were 79% LSFO. LSFO also won 78% of the projects in
recent non-Phase I selections. It is therefore appropriate to compare B&W's E-LIDS™ process
with a state-of-the-art LSFO WFGD system.
E-LIDS™ is superior to a state-of-the-art WFGD system for the following reasons:
SO., Emissions - E-LIDS™ achieves higher sulfur removal without additives at a lower cost.
Work at both the 0.6 MWe and 10 MWe scales in Phases I and II has shown that E-LIDS™ can
meet the stringent project SO, removal target of 0.1 Ib SO2/MBtu at a calcium stoichiometry near
the most favorable economic case for the E-LIDS™ process used in the Phase I evaluation. A
state-of-the-art wet scrubber can only achieve about 95% SO2 removal; the 98% removal
required for the LEBS application would require an advanced wet scrubber — possibly requiring
costly performance-enhancing additives.
Particulate Emissions - E-LIDS™ has superior paniculate removal performance compared to a
state-of-the-art wet scrubber preceded by a precipitator or baghouse. Since spent reagent is
carried from the dry scrubber into the baghouse creating a cake on the bags, the filtration
efficiency is enhanced. E-LIDS™ may also yield lower emissions of PM 2.5 precursors due to
its superior acid mist control (see below). Testing at B&W supports the expectation that
paniculate emissions from the E-LIDS™ system can be as low as 0.005 Ib/MBtu.
Capital Cost - The capital cost of an E-LIDS™ system is projected to be as much as 40% less
than the cost of a state-of-the-art WFGD system.
Auxiliary Power - E-LIDS™ requires as much as 70% less auxiliary power than WFGD. Since
E-LIDS™ is a dry FGD system, no slurry recirculation pumps are required. An additional
power savings is gained because the ID fans are located downstream of the FGD system.
Levelized Cost - The lower capital cost and auxiliary power consumption for E-LIDS™ offset its
slightly higher limestone usage resulting in a lower levelized cost relative to a state-of-the-art wet
scrubber.
Mercury Removal - Mercury capture using E-LIDS™ has surpassed all expectations. The 95%
mercury removals consistently achieved using E-LIDS™ are a clear advantage of the process. A
review of mercury capture across commercial WFGD systems indicates average removals of
about 50%. However, some wet scrubbers perform better than the average.
"Particulate Metals" - The high paniculate collection efficiency exhibited by the E-LIDS™
process during Phase n testing results in excellent control of "paniculate" trace metal
(nonvolatile trace metal) emissions.
-------
HClandHF - Testing has shown E-LIDS™ is capable of removing almost all of the HC1 and
HF from the flue gases. This directly reduces the emissions of air toxics and may also impact
paniculate emissions by removing PM 2.5 precursors.
Acid Mist Control - The E-LIDS™ process has an advantage in greater acid mist control due to
the limestone furnace injection process. Normally, airheater outlet temperatures must be kept
above the SO3 acid dew point temperature to avoid condensation and corrosion in the airheater
and downstream flues. Removing the SO3 upstream of the air heater allows the air heater flue
gas outlet temperature to be reduced by 40°F or more. This reduction in temperature has a
favorable impact on boiler efficiency.
Net Plant Efficiency - The overall impact of E-LIDS™ on the net plant efficiency is noticeably
favorable. This is primarily due to the significantly lower auxiliary power consumption of the E-
LIDS™ process relative to the state-of-the-art wet scrubber.
Solids Handling - E-LIDS™ has the advantage that the solid by-products produced by the
process are dry and can be handled with conventional dry solids handling equipment. B&W
continues to actively pursue potential utilization schemes for the E-LIDS™ by-product.
Waste water - The E-LIDS™ process does not produce a waste water stream.
Figures 13 and 14 show a E-LIDS™ versus WFGD comparison for SO2 and mercury removal,
capital and levelized costs, and auxiliary power. B&W is currently pursuing the initial
commercial retrofit of the E-LIDS™ system.
SO2 Removal
Mercury Removal
Figure 13 - SO2 and Mercury Removal; E-LIDS™ versus WFGD
(note that WFGD mercury removal can range anywhere from 10-90% total removal)
-------
Capital Cost
Aux. Power
Levelized Cost
Figure 14 - E-LIDS™ versus WFGD; Capital Cost, Aux Power and Levelized Cost
Proof-of-concept Demonstration
In Phase IV of the B&W LEBS project, (beginning October 1997) a POC facility will be constructed
and operated in order to prove the readiness of the technology for commercial application. During
the POC demonstration, the E-LIDS system will be integrated with B&W's DRB-4Z™ burner and
an advanced controls and sensors system to demonstrate the fully-integrated B&W LEBS technology.
Acknowledgments
The authors express their thanks to the U. S. Department of Energy for supporting the B&W Low
Emissions Boiler System development (Contract Number DE-AC22-92PC92160). Special thanks
goes to Anthony Mayne as technical monitor.
References
1. Madden, D. A. and Amrhein, G. T, "Performance of Babcock & Wilcox's Limestone Injection
with Dry Scrubbing (LIDS) Process," Presented at the Second North American Conference
and Exhibition, Clean Air 1996, Orlando, Florida, November 19 - 22, 1996.
2. Madden, D. A. and Musiol, W. E, "10 MWe Prototype Testing of LIDS as Part of the
Babcock & Wilcox Low Emission Boiler System," presented at the 21s1 International
Technical Conference on Coal Utilization & Fuel System, Clearwater, Florida, March 18 -
21, 1996.
-------
3. Evans, A. P., Redinger, K. E., and Farthing, G. A., "Air Toxics Benchmarking Tests on a 10
MW Coal-Fired Utility Boiler Simulator," presented at the 21st International Technical
Conference on Coal Utilization & Fuel System, Clearwater, Florida, March 18 - 21, 1996.
4. Study of Hazardous Air Pollutant Emissions from Electric Utility Steam Generating Units
Pursuant to Section 112(n)(l)(A) of the Clean Air Act, Draft, June 1995.
-------
Sodium Sorbents for SO2 Trim in High-Ratio Baghouses
Sharon Sjostrom, James Butz
ADA Technologies, Inc.
304 Inverness Way South Suite 365
Jean Bustard
ADA Environmental Solutions, LLC.
7931 S. Broadway, #349
Littleton, CO 80122
Ramsay Chang, PhD.
Electric Power Research Institute
P.O. Box 10412
Palo Alto, CA 94303
Introduction
A significant number of utility power plants are switching to low-sulfur coals in response to the
1990 Clean Air Act Amendments mandate to reduce SO2 emissions. Title I (National Ambient
Air Quality Standards) identified sulfur oxides and particulate matter as Criteria Pollutants. Title
IV (Acid Rain Program) established limits for plants through the SO2 Emission Allowance
Program and initiated a SO2 permit system. Further reductions in SO2 may be required as a result
of potential revisions to the fine particulate standards in Title I because sulfates condensing from
stack gases likely contribute to the ambient PM 2.5 particulate levels. Recent measurements of
filterable and condensible particulate emitted from several utilities burning coal in the Unites
States showed that from most sources tested, the majority of the total particulate emitted was
condensible particulate.'
To allow more flexibility in the sulfur levels of coal purchases and to meet potentially more
stringent SO2 emission requirements, a low cost method of trimming SO2 levels in the flue gas is
desirable. One option is the injection of dry sorbents such as sodium bicarbonate and sodium
sesquicarbonate upstream of a particulate collector such as an ESP or a baghouse. These
sorbents decompose in the elevated temperature environment of a flue gas duct to form porous
sodium carbonate particles while evolving carbon dioxide gas and water vapor. The porous
sodium carbonate particles then react with flue gas SO2 and oxygen to form a solid sodium
sulfate by-product.
Dry sorbent injection technology is particularly attractive with COHPAC (Compact Hybrid
Particulate Collector), a patented EPRI technology where a pulse-jet baghouse operates at high
air-to-cloth ratios downstream of an electrostatic precipitator. In this configuration, sorbent can
be injected after the ESP and before the pulse-jet baghouse so that the sorbents and flyash are
1
-------
collected separately and can be disposed of independently. Sodium-based sorbents introduce
potential disposal concerns when collected simultaneously with flyash because the sodium in the
ash may leach into the groundwater.
Flue gas desulfurization with dry sodium based sorbents has been investigated extensively since
the 1960s in conjunction with both ESPs and low-ratio baghouses. The Electric Power Research
Institute (EPRI) has sponsored several pilot and full-scale tests over the past four years to
evaluate the cost effectiveness and limitations of dry sodium compounds injected for SO2 trim in
high-ratio, pulse-jet baghouses (including the COHPAC configuration). This paper offers an
analysis of the test data and compares the pulse-jet and COHPAC results to sorbent performance
from reverse-gas baghouses and ESPs. A stepwise regression statistical analysis was applied to
the COHPAC and pulse-jet data to quantify the significance of the several parameters which are
believed to impact the removal efficiency and sorbent utilization rate of the sodium sorbent SO2
removal process.
Sodium Reaction Chemistry with SO2
The chemistry by which the sodium products remove the SO2 is relatively straightforward,
consisting of sequential gas-solid reactions. First, sodium sesquicarbonate or bicarbonate
decomposes in the elevated temperature environment of the flue gas to form porous sodium
carbonate particles while evolving carbon dioxide gas and water vapor. Sodium sesquicarbonate
decomposes by the following reaction:
Na,C03 • Na HCO3(s) ->• 3/2Na2CO3(s) + 1/2 H2O(g) + l/2CO2(g)
The porous sodium carbonate particles then react with flue gas SO2 and oxygen to form a solid
sodium sulfate by-product. More carbon dioxide gas is evolved in this reaction:
Na2C03(s) + S02(g) + l/202(g) -> Na^SO^s) + CO2(g)
The sodium sulfate by-product is removed in the downstream particulate control device, an ESP
or baghouse.
There is an additional, unwanted reaction that typically occurs when sodium sorbents are injected
into the flue gas for S02 removal: a net oxidation of flue gas NO to NO,. Because NO2 is a
brown-colored gas, this conversion can lead to the appearance of a visible plume. Both the
amount of NO converted to NO2 and the minimum concentration of N02 in the flue gas which
results in plume visibility tend to be site specific. The amount of conversion of NO to NO2
typically depends on variables such as the molar ratio of SO2 to NOX in the flue gas, the sodium
sorbent type, and the sodium sorbent injection ratio. The concentration of NO2 which results in
plume visibility typically depends upon the stack diameter and background conditions. Under
"ideal" viewing conditions (e.g. clear blue sky behind the stack exit, calm winds), this brownish-
orange coloration can be seen as a product of stack diameter times concentration as low as 100
ppm-meters. Thus, under some circumstances, with a 10-meter stack this plume visibility could
-------
be seen at an NO2 concentration as low as 10 ppm. The injection of ammonia has proven to be
an effective means of reducing NO2 concentrations to alleviate this problem. There are several
possible reaction products of ammonia and NO2 in the flue gas including an ammonia/nitrogen
dioxide compound, an ammonia/nitrogen dioxide/sulfur dioxide compound, or simply nitrogen
and water. The addition of ammonia to reduce NO2 may also result in ammonia discharge from
the plant, or "slip" 2
Critical Parameters
A uniform distribution of sorbent particles into the flue gas is important to achieve high SO2
removal efficiencies, as is a high sorbent specific surface area (i.e. small sorbent particle
diameter). The optimum range for sorbent particle size is 20-25 micron diameter, which presents
fairly high surface area while maintaining good solids handling and feeding properties. The
reagent becomes less effective as the particle size increases. Differences in particle size are a
result of processing, and are therefore vendor-specific; however, additional pulverizing can be
performed at the point of use to obtain smaller particle sizes. Particle size is typically reported as
mass median diameter (MMD), the diameter at which 50% of the mass of particles in the sample
are smaller than the value, and 50% are larger. Most measurement techniques for particle sizing
use light scattering to determine the diameters of a sample in a range of size bins. The data is
plotted on a cumulative basis to find the MMD. It must be recognized that particulate materials
are not of a single size, but rather the particles span a size range that is characterized in a single
value, the MMD.
Heat from the flue gas to thermally decompose the sorbent plays an important role in achieving
high SO2 removal efficiencies and high sorbent utilization. Carbon dioxide and water vapor are
released from the sodium sesquicarbonate or bicarbonate particles and form voids when the
sorbent is exposed to elevated temperatures, and a very porous sodium carbonate particle
evolves. The thermally decomposing sorbent particle has a much higher specific surface area
than could be achieved by grinding sodium carbonate particles to smaller particle diameters.
Sodium sulfate by-products have a higher molar volume than the original sodium
sesquicarbonate or bicarbonate, and tend to plug the newly formed pores generated by the
thermal decomposition. However, since the thermal decomposition step occurs simultaneously
with the sulfation process, new sodium carbonate surface area is continually exposed to allow the
reagent to be converted to the sulfate form. Previous testing has indicated that flue gas
temperatures greater than 250-270°F promote more effective sodium bicarbonate reagent
utilization. Temperatures above 270 °F do not significantly improve sodium sesquicarbonate
performance.
It is believed that the sorbent reacts to some depth, x, beneath the surface, regardless of initial
sorbent size. Therefore, if the irregularly shaped sorbent is modeled as a spherical particle, the
volume unreacted is 4/3 n (R-x)3, where R is the sorbent particle radius. Figure 1 presents the
unreacted volume of spherical particles for reaction depths limited to 5 and 10 microns for
sorbent diameters ranging from 10 to 45 microns. Although the sorbent particles are in fact not
spherical, the trend is clear. If the reaction depth is limited to 5 microns, this calculation
-------
indicates that a significant fraction (nearly 40% for spherical particles) of the sorbent volume
would be unavailable for a 40 |j.m diameter sorbent particle.
50
Unreacted
Volume
20
30 40 50
Sorbent MMD (urn)
Figure 1. Volume unavailable for reaction on a spherical particle if reaction depth is limited to 5
or 10 )im.
SO2 Removal in Pilot and Full-Scale COHPAC and Pulse-Jet Baghouses
The Electric Power Research Institute has sponsored several evaluations of SO2 removal with
sodium-based sorbents in pulse-jet and COHPAC baghouses over the past four years. The data
set evaluated for this paper consisted of 38 test configurations from three 3700 acfm pilot
COHPAC baghouses, a 5000 acfm pilot pulse-jet baghouse and a 145 MW full-scale COHPAC
baghouse, representing operations at four sites.3'4 The pressure drop across the baghouse filter
varied from 2.5 - 7.8 inches H2O and the air-to-cloth ratio ranged from 3.5 to 16 ft/rnin.
Temperatures at which testing was performed varied from 230 to 340 °F. The mass median
diameter of the un-pulverized sorbents ranged from 32 to 43 p.m.
A multi-variable least squares model fit to predict SO2 removal was developed from a statistical
analysis of the data collected at the four sites. Flue gas temperature, sorbent type, sorbent mean
diameter, sorbent mean diameter squared (a variable proportional to surface area), air-to-cloth
ratio, pressure drop across the fabric, and sorbent injection ratio were evaluated for a regression
effect. The sodium sorbent injection ratio is typically expressed in terms of normalized
stoichiometric ratio (NSR), which is a ratio of moles of sorbent injected per mole of SO2 in the
flue gas. All parameters EXCEPT air-to-cloth ratio, pressure drop, and sorbent mean diameter
squared showed a statistically significant effect in a stepwise linear regression with S02 removal
as the dependent variable, based on a student's t test. The R2 coefficient (an estimate of the
proportion of the variation in the response around the mean that can be attributed to variables in
the model rather than random error) is 0.92 for the model fit.
-------
In a stepwise linear regression, the analysis specifies the order of independent variables in their
impact on the dependent variable. This model shows that four variables statistically effect SO,
removal and in order of importance they are:
1. Normalized Stoichemetric Ratio (NSR);
2. Sorbent type (bicarbonate or sequicarbonate);
3. Mass median diameter of the sorbent; and
4. Flue gas temperature
The actual SO, removal data as a function of NSR and the corresponding model prediction for
each data point are shown on Figure 2. The data includes bicarbonate and sesquicarbonate
evaluations in the pulse-jet and COHPAC configurations. In most cases, the model predicts SO2
removal rates within a few percent of the measured removal rate.
O
E
i?
«f
5!
100
90
80
70
60
50
40
30
20
10
0
4. X I
A
,#•> ««i»:« " -
* A* *A *
"1*
D 0.5 1 1.5 2
NSR
j.COI^AC
4 Pulse-Jet !
i x Model Prediction
Figure 2. SO, removal as a function of normalized stoichiometric ratio (NSR) at four test sites.
Figure 3 shows the empirical model prediction for SO, removal as a function of sorbent size for
both sorbents. The model prediction shown is for a fixed temperature of 300 °F and a NSR of 1.
The model shows that sodium bicarbonate is expected to result in nominally 20% more SO2
removal than sodium sesquicarbonate for similar operating conditions. The model also shows
that 20% lower SO, removal is expected for a 50 pm particle versus a 20 p.m particle. Although
it is known that sodium bicarbonate has better utilization than sodium sesquicarbonate,
bicarbonate is approximately 4 times the price of sesquicarbonate. If twice as much
sesquicarbonate is required to remove equivalent SO2 as bicarbonate, the sorbent costs for
sesquicarbonate are still less.
-------
30
20
10
0
T = 300 OF, NSR = 1
Sorbent Type
Bicarb
_ —Sesqui
10 20 30 40
Sorbent MMD (micron)
50
Figure 3. Empirical-based model prediction of SO2 removal as a function of sorbent particle size
for sodium bicarbonate and sesquicarbonate.
The influence of temperature on SO2 removal at the test sites was significant, but not as dramatic
as sorbent particle size. Figure 4 presents the model prediction for SO2 removal as a function of
flue gas temperature. The NSR for this model run was 1 and the assumed particle MMD was 30
u.m. Increasing the temperature from 250 to 350 °F results in an increase in SO2 removal of 7%.
50
40
30
20
10
30mm MMD, NSR = 1
Sorbent Type
Bicarb
— _ Sesqui
250
270
290 310
Temperature (°F)
330
350
Figure 4. Empirical-based model prediction of S02 removal as a function of flue gas temperature
for sodium bicarbonate and sesquicarbonate.
-------
Results from Previous ESP and Reverse-Gas Evaluations
Data representing several years of operating history have been published for SO2 removal in
ESPs and reverse-gas baghouses using sodium-based sorbents. The information plotted in Figure
5 includes studies at six sites with ESPs.5'6'7 The upper shaded region on the figure shows SO,
removal for sodium bicarbonate and the lower shaded region for sesquicarbonate, where the
sorbent MMD was 11 - 18 urn and flue gas temperatures ranged from 250 to 400°F. A slightly
larger bicarbonate sorbent (20-25 (irn) at 380 °F showed an SO2 removal fraction approximately
15% below the shaded region for bicarbonate (data point shown as x in square). Data from an
un-pulverized, 43 um MMD bicarbonate sorbent at 390 °F resulted in SO2 removals 40% below
the shaded zone (data points shown as triangles) . The importance of particle size is illustrated in
this data set.
100 -
90 {
80 |
70 I
60 |
50
40
30
20
10 -r
-& Bicarb" Sf?;0
'
NSR
Figure 5. Summary of ESP results from previous tests.
5,6,7
Results with sodium bicarbonate and sesquicarbonate injected upstream of reverse-gas
baghouses8 are shown in Figure 6. The shaded zones in this figure include pulverized and un-
pulverized sorbent and particle MMDs from 10 urn - 45 urn MMD. The data represented by the
shaded zones was collected at temperatures from 230 to 400°F. SO2 removal with sodium
sorbents was more effective upstream of reverse-gas baghouses than upstream of ESPs.
-------
>
o
as
OL
O
Figure 6. Summary of reverse-gas SO2 removal results.l
The shaded regions in Figure 7 show model predictions for SO2 removal as a function of NSR for
pulse-jet and COHPAC at sorbent sizes from 20 - 50 p.m MMD. Actual test data are also plotted
in this figure (plotted points). This figure shows predicted and actual removal rates lower than
shown in Figure 6 for reverse-gas baghouses, but higher than shown in Figure 5 for ESPs with
comparable sorbent sizes. It is noteworthy that the sodium sorbents used in this model ranged in
size from 32 to 43 jj.m, which is a larger size than much of the data in Figures 5 and 6. It is likely
that higher SO2 removal rates are possible with pulse-jets and COHPAC if sorbents with smaller
particle diameters are used. However, sufficient data is not available to determine if the lower
SO, removal was due to larger sorbent sizes, or some other effect.
-------
100
90
80
70
60
50
40
30
20
10
0
Bicarb
' v'A-rstM?.-^--.-.- '
^^'f^&J^s-^r-
tiuUXiStiS
Sesqui
« Sesqui COHPAC
B Bicarb COHPAC '
i o Sesqui P-J ;
: A Bicarb P-J !
Figure 7. Summary of pulse-jet and COHPAC SO2 removal results
Conclusions
Data from a series of tests of dry sorbent injection of sodium compounds for SO2 control was
analyzed to identify significant process variables. Test data were used from pilot-scale and full-
scale installations of pulse-jet and COHPAC baghouses at power plants in the US. A stepwise
linear regression was used with a student's T test. The following conclusions were reached:
• The predominant factors influencing the rate of SO2 removal with sodium dry sorbent
injection in pulse-jet baghouses and COHPAC are 1) normalized stoichiometric ratio, 2)
sorbent type (bicarbonate or sequicarbonate) 3) sorbent particle size, and 4) flue gas
temperature. Pressure drop across the filter and the air-to-cloth ratio at which the baghouses
were operated did not produce statistically significant effects on SO2 removal.
• S02 removal rate in pulse-jet baghouses and COHPAC with sodium sorbent injection is
higher than in ESPs for similar flue gas temperatures and with comparable sorbent particle
sizes.
• S02 removal rate in pulse-jet baghouses and COHPAC appears to be lower than in reverse-
gas baghouses for similar temperatures and with comparable sorbent sizes.
• It is believed that this model can be used to predict SO2 removal for a pulse-jet or COHPAC
installation considering sodium injection for SO2 control.
-------
References
1. Corio, Louis and John Sherwell (1997). "In-stack Condensible Particulate Matter
Measurement and Permitting Issues" Presented at the Air & Waste Management
Association's 90th Annual Meeting & Exhibition, Toronto, Ontario, June 8-13.
2. Bland, V.V, C.E. Martin (1990). "Full-Scale Demonstration of Additives for NO2 Reduction
with Dry Sodium Desulfurization" EPRI Report Number GS-6852.
3. S.M. Sjostrom, CJ. Bustard, R.H. Slye, T. Hunt, H. Noble, G. Schott, S. Thomas, R. Chang
(1993). "Pilot-Scale Demonstration of the Compact Hybrid Particulate Collector
(COHPAC)." Presented at the Tenth Particulate Control Symposium & Fifth International
Conference on Electrostatic Precipitation, Washington, D.C, April 5-8.
4. EPRI project reports.
5. Sonnichsen, Tim, (1987). "Dry Sodium Sorbent Injection Upstream of an Electrostatic
Precipitator for Retrofit SO2 Control on Coal-Fired Utility Boilers" ASME Publications 87-
IPGC-Pwr-51 American Society of Mechanical Engineers. Presented at the Joint
ASME/IEEE Power Generation Conference, Miami Fl. October 4-8.
6. Hooper, Richard (1988). "Full Scale Demonstration of Flue Gas Desulfurization by the
Injection of Dry Sodium Bicarbonate Upstream of an Electrostatic Precipitator'1 7th
Symposium on the Transfer and Utilization of Particulate Control Technology, Nashville,
TN, March 22-25.
7. Coughlin, Terry, P. Schumacher, D. Andrew and R. Hooper (1990). "Injection of Dry
Sodium Bicarbonate to Trim Sulfur Dioxide Emissions." Presented at EPRI/EPA 1990 S02
Control Symposium, New Orleans, LA, May 8-11.
8. Muzio, L and G. Offen (1987). "Assessment of Dry Sorbent Emission Control Technologies,
Part I" APCA vol 37 no. 5, p 642-654.
Acknowledgments
The authors would like to thank several individuals for their insight and expertise: Cam Martin
and Gary Anderson of ADA Technologies, Inc., Terry Hunt of Public Service Company of
Colorado, and Gary Blythe of Radian Corporation.
10
-------
Thursday, August 28; 8:00 a.m.
Parallel Session C:
Advanced SO2 Control Processes
-------
SUPERIOR FGD COST-EFFECTIVENESS
VIA HIGH-VOLUME, HIGH-VALUE BYPRODUCT GENERATION
GJ. Emish
Krupp Wilputte Corporation
1370 Washington Pike
Bridgeville, Pennsylvania 15017
W. Ellison
Ellison Consultants
4966 Tall Oaks Drive
Monrovia, Maryland 21770
Abstract
Current worldwide advancements in application and field development of
ammonia FGD, largely justified by significant revenues from ammonium
sulfate byproduct generation, are detailed. This major new direction in
optimization of FGD selection and the manner of design of such systems
encompasses major performance advancements that will be reviewed:
• Potential for highest efficiency in removal of acid gases: difficult to
collect, SO3, as well as SO2 and HC1
• Conversion of all captured acid gases into agriculturally usable ammonium
compounds
• No discard or long-term, outdoor storage of sulfurous waste byproducts
-------
• No liquid effluent
• Negligible stack opacity.
Achieved at a capital-cost penalty, in part because of need in
medium/high sulfur service for and use of ultra-high-efficiency
particulate/aerosol removal capability downstream of pollutant removal
facilities, cost effectiveness is quantified and shown, nonetheless, lo be
favorable. Attractiveness of process economics is made possible because of
the substantial value added in conversion of ammonia reagent supply to sulfur
blending stock, i.e. granulated ammonium sulfate, highly valued and much in
demand in worldwide agriculture. Major/growing size of the potential
international market for ammmonium sulfate is quantified, with a detailing of
multiple factors that have, in recent decades, elevated this commodity, once
the "poor-man's fertilizer", to the status of optimum, high-valued, NPKS
fertilizer blending stock that serves to address critical long-term, worldwide
agricultural soil sulfur shortage.
Introduction
Ammonia-base FGD lias become a new, major, I'"CD option1 and, because of
byproduct revenues and other performance advancements, can be seen to be a
new, prime, cost-effective alternative to wet lime/limestone scrubbing2
Critical goals in FGD process selection and use, reflected by Table I3, are
advantageously addressed by ammonia reagent use.
Table 1
Objectives of System Design/Operation
Function Broadest Aims
Stack Gas Cleaning Simultaneous removal from dedusted
raw gas of SO2, S03, HC1, HF and fine
participates, including air toxics (trace metals)
Cost Effectiveness Minimal total annual cost
-------
Characteristics of Catch Broadly usable liquids/solids, preferably
of intrinsically high sales value
Solid Waste Management High-volume, agricultural-
fertilizer/chemical yield, or generation of
pozzolanic material supply, with
utilization unconstrained by contaminants
Liquid Waste Management No effluent outfall or leachate formation
creating pollution of natural water bodies
(surface water and groundwater).
Stack Gas Cleaning Achieving Multi-Pollutant Removal
Performance and cost-effectiveness goals call for efficient removal by the
FGD system of the array of significant gasborne pollutants present, while at
the same time achieving a low stack opacity.
Cost Effectiveness That Minimizes Total Cost Per Ton of SO2 Removal
With worldwide focus on abatement programs for economically driven, S02
emission reduction, FGD-process selection calls for the use of a technology
that incurs a minimum total cost per ton of SO2 removal. The supplemental
effect on overall/virtual economics brought about by adverse externalities,
such as extent of consequential environmental pollution brought about in
management of gas cleaning residues, should be taken into account. This will
influence the FGD process choice, made on economic grounds, toward
alternatives such as ammonia FGD that avoid adverse impact on natural water
bodies, particularly potable surface water and groundwater resources.
SO2-Catch Characteristics That Provide Potential For High-Volume
Utilization
Sulfurous FGD reaction product, either in the form generated or
after cost-effective reprocessing, should have properties that, in practical
terms, are compatible with recycle or reuse objectives, i.e. taking foil account
of the level of byproduct yield-volume al the site-specific FGD installation.
Ammonia FGD well meets these criteria. When practical, complete,
commercial utilization of a high-volume FGD byproduct is doubtful, a yield
-------
that, (in the final form of output from the powerplant), is significantly water
soluble is undesireable.
Overview of Ammonia FGD
With a long history of commercial development and early full-scale
applications by IHI (Japan) and Krupp Uhde (Germany), wet ammonia FGD
has been brought to new prominence with the 1996 FGD startup at Basin
Electric Cooperative's Dakota Gasification Plant, Beulah, North Dakota of a
uniquely high sulfur and high capacity retrofit installation. The process
design utilizes flue-gas sensible heat in a pre-scrubber stage to render the
ammonium sulfate byproduct in a crystalline solids form suitable for
mechanical dewatering to a cake yield. This, in turn, is converted into
screened, crushed briquettes for use in agriculture. Ammonia-base FGD
provides a means to gain substantially improved gas cleaning economics, (i.e.
reduced annual cost of abatement), with superior waste/byproduct
management and gas treatment effectiveness. Notable advancements offered
by present day ammonia FGD technology, originating domestically or
overseas, are as follows:
• System-design remedies for the historic ammonia-FGD blue
plume emission problem
• Major revenues from byproduct ammonium sulfate generation and sale
ensuring the greatest available cost-effectiveness in large-capacity
applications with high gross SO2 concentrations
• Conversion of all collected acid-gas species including S02, S03, HC1,
etc. into a catch comprising ammonium compounds that are commercially
usable, collectively, as a high-value, sulfur/nitrogen blending stock in
large-scale NPKS fertilizer production
• No requirement for elimination from the system of collected secondary
pollutants, e.g. by liquid purging, even in coal-fired service, provided that
typical, dedusting facilities for coal fly-ash removal are provided
upstream of FGD
• Avoidance of wastewatcr discharge through zero-diluent Hue-gas
treatment system operation. Condensate generated in any of the
-------
byproduct agglomeration processes may be recovered and utilized, either
as water diluent in forming aqua ammonia feed to the FGD system or as
FGD process makeup water. Thus, there is no wastewater outfall from the
full, integrated, gas-desulfurizatioi^yprocluct-aggloineration operation
encompassing ammonia-using SO2 removal.
Comparative Cost-Effectivene.ss of FGD Processes
The paper4 given by Electric Power Research Institute, (EPRI),
at the 1995 SO2 Control Symposium graphically details the superior
economics of wet ammonia FGD in high-sulfur, electric utility plant service
i.e. applied in conjunction with construction of a hypothetical, new, high-
capacity unit. Despite a higher capital cost, wet ammonia FGD is shown to
incur a total/levelized cost per ton of SO2 removal that, routinely, is 25%
lower than that of wet limestone forced-oxidation FGD yielding gypsum
byproduct. Moreover, EPRI data indicates that ammonia FGD economics
further improves with increase in fuel sulfur level. In this hypothetical, new-
boiler application cited by EPRI, wet ammonia FGD performance and
economics have been stated as follows:
• Powerplant unit capacity: 300 mWe
• Coal sulfur level: 2.6%
• SO2 removal efficiency: 90%
'• Capital cost: 1990 $223/kWe
• Levelized total cost (net): 1990 $322/ton SO2 removal
Byproduct Properties/Value
Overview
Substantial byproduct revenues in medium and high-sulfur service,
(accompanying sustained sales value of agglomerated ammonium sulfate,
typically 2 to 4 times that of the ammonia from which it is formed), is the key
factor accounting for the highly attractive, low overall cost of ammonia FGD
system ownership and operation. Byproduct price and attractiveness are
supported by the high ammonium byproduct quality typically available when
treating flue gas that has been dedusted or has low participate concentration.
See Table 2, which, based on European commercial experience, reflects the
-------
very low trace metal content as contrasted with governmental standards and
with typical commercial NPK fertilizer production.
Comparative Trace-Metal PPM-Content of FGD Ammonium Sulfate
Impurities
Average
Ammonium
Sulfate
Byproduct
From
Commercial FGD
(Germany)
Allowable
Concentration
Per EU Regul-
ation (European
Union)
Commonplace
NPK
Fertilizer
Brands
(Germany)
Arsenic
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Zinc
10.4
0.61
5.21
3.27
12.50
0.67
3.99
32.1
500
20
670
670
670
10
335
2000
<100
4 to 9
80 to 110
20
<100
0.5
40
210 to 240
Evolution of the Market To Date
In worldwide byproduct marketing of ammonium sulfate it has for many
decades been viewed as a "poor man's" fertilizer because it contain's only
21% nitrogen. When used alone, it includes no P (phosphorous) or K
(potassium) nutriment, while creating a high degree of soil pH depression, an
unfavorable impact in common acid soils. However, the evolution of process
technology utilized in manufacture of massive amounts of chemical fertilizer
derived from phosphate rock in the 1960s and 1970s led to sharply
diminishing sulfur level in the NPK (NPKS) fertilizer supply. This
shortcoming has created increasing demand for available ammonium sulfate
production as an attractive sulfur additive in formulation of NPKS blends.
This sulfur sourcing has strong endorsement by the U.S. fertilizer industry2
-------
This conventional fertilizer supply can be tailored for requirements of regional
markets so as to meet plant nutrient sulfur, (PNS), needs of the local
agricultural soils. As a result, ammonium sulfale byproduct has consistently
sold in midwest U.S. for approximately 3 or more times the market value of
the contained anhydrous ammonia. See Table 3. A comparatively low price
is commanded by crystallized product, (Figure I6), and compacted material,
(screened crushed-briquettes, Figure 26), the commercial forms of
agglomerated ammonium sulfate most commonly supplied in North America
to date. Maximum revenue and cost effectiveness in byproduct generation
and marketing is available through agglomeration by granulation from an
ammonium sulfate solution, e.g. 30% strength process liquor, discharged from
the FGD system.
Basis for Preferred Supply as Granules
Typically, all producers ol" ammonium sulfate prefer lo refer to (heir output as
granules, which are near-spherical, agglomerated particles. They are
optimally 2 to 4 mm in diameter and, by technical definition, are formed from
solution by deposition/buildup of successive solids layers onto an initial,
small particle or seed crystal using hot air as (lie heal source for water
evaporation. True granules are relatively free of dusting tendencies and are
the ideal product- form for extended storage and/or long-distance transport.
However, ammonium sulfate liquors are known by the fertilizer industry to be
typically difficult to efficiently granulate e.g. through the use of simple,
conventional, pelletizing equipment.
Preferred Granulation Means
To establish commercial granulation capability, Europeans have developed
the Fluidized Drum Granulation Process, (Figures 3 and 4), of Kaltenbach
Thuring S.A., Beauvais, France, a full-scale installation of which processes
spent scrubbing liquor output of the 70 mWe-equivalent wet ammonia
scrubbing facility at the coal-fired, municipal powerplant in Karlsruhe,
Germany. In this unique agglomeration process, a combination of drum
granulation and fluidized bed technology, the FGD scrubbing-solution is
'concentrated in an evaporator, (Figure 3). Additives such as filler,
micronutrients or other components may be mixed with this concentrated
byproduct solution, if required, before it is passed to the granulator. The
solution is sprayed onto the moving bed of growing granules in the rotating
-------
granulating drum, (Figure 4), a cylindrical horizontal vessel that turns on its
axis. The granulating drum is filled with guide vanes, i.e. unique anti-
clogging lifters, so that the granules, conveyed upwards, lull into the lop of a
fluidized bed dryer, integral to the granulator, to be dried with hot air
subsequently drawn off by an evacuating fan. From the fluidized bed the dry,
partially grown granules flow back into the granulating drum where they are
once again sprayed with solution: and (in this cyclic manner) continue to
grow in diameter, step by step. The granulating drum discharge is separated
by size on a screening device. The oversize, after size reduction in a crusher,
is returned to the granulator in the form of fines that provide granulation seed
material. Those screened granules meeting specification size are passed to
the product dryer for final drying prior to cooling. Agglomerate hardness
achieved is greater than that of any comparably simple, available
Table 3
Midwest Com Belt
Pricing of Ammonia and Ammonium Sulfate,
(Green Markets Publication, Bethesda, MD)
Calendar
1984
1985
1986
1987
1988
1989
'1990
1991
*M.W.
2xM
S/s.ton Price:
Anhydrous
year Ammonia (A)
196
172
126
118
105
130
128
140
of(NH4)2 SO4
.W. of NH3
S/s.ton Price:
Ammonium
Sulfate (AS)
106
110
104
106
96
112
117
125
132
2x17
Pcr-Ton Price
Ratio: AS/A
0.54
0.63
0.83
0.90
0.91
0.86
0.91
0.89
.... the 3. 88
Ratio Nitrogen-Price
X
X
X
X
X
X
X
X
mi
Conversion Ratio:
3.88*= 2.1
2.4
3.2
3.5
3.5
3.3
3.5
3.5
ultiplier, above
AS/A
-------
agglomeration process. To improve storage characteristics the granulate is
treated with an anti-coagulation agent. All gas flow is dedusted using a
scrubbing device, and the resulting scrubbing-liquid discharge is recycled to
the process.
Characteristics of Commercially Supplied Granules
Product specifications for the granules shipped from Karlsruhe are as follows:
* Nitrogen content, min. 20 wt.%
• Water content, max. 0.3 wt.%
• 'Particle size distribution (diameter):
- <2 mm 5%, max.
2-4 mm 90%, min.
>4 mm 5%, max.
• Hardness (referenced to the amount of
test weight that can be supported by
a sample particle of size 3 mm dia.) 2.5 kg, min., and typically 3 kg,
(4 kg after 8 days aging)
• Bulk density, mm. 800 kg/m3, (50 Ib/cu ft)
The Role of Ammonium Sulfate Supply in Worldwide Agriculture
Addressing Soil Sulfur Shortage
The current increased availability and application of ammonia FGD porlcnds
a substantial increase in world supply and use of byproduct ammonium
sulfate. This latent demand is verified by the sustained, high, market price in
relation to that of anhydrous ammonia commanded by ammonium sulfate.
Utilized as a combined nitrogen and plant-nutrient-sulfur (PNS) blending
stock, i.e. sulfur additive, this growing output derived from slack gas cleanup
will help alleviate the major and worsening worldwide deficiency in
commercial agronomy due to significant, widespread, soil-sulfur shortage in
agriculture. This impairment is primarily the result of past and present
overworking of agricultural lands, as well as long-term diminishing of sulfur
-------
level in generic NPK fertilizer output prior to the 1980s. Sulfate is the
preferred/optimal sulfur additive, most readily assimilated by crops, and
ammonium sulfate is the ideal byproduct sulfate; it combines sulfate with the
cation, ammonium, the most advantageous nitrogen fertilizer form.
Expansion of Tillable Agricultural Lands
A growing international supply of ammonium sulfate can be a potentially key
factor, along with development of untapped sources of irrigation water, in
enhancing the amount of food harvested, worldwide, through increased
utilization of land area available for agriculture. Such application of growing
worldwide FGD output of ammonium sulfate byproduct to virgin, naturally
alkaline, semi-arid land would be uniquely effective as crop nutriment
because:
• Its very high, soil-acidification effect would beneficially moderate soil-pH
to an acceptable, productive level.
• Substantial reduction of the pH of highly alkaline soils by the use of
ammonium sulfate (in conjunction with the production of repeated
annual/seasonal harvests) would substantially liberate, and make available
as additional nutriment, plant foods such as phosphate, present in the soil
in a natural form.
Need For Expanded Agricultural Productivity
This promising outlook for significant growth in worldwide ammonium
sulfate use is encouraged by 1994 findings of the environmentalist-forum, The
Worldwatch Institute, (U.S.A.), which reports that:
• After nearly four decades of unprecedented growth in both land-based and
oceanic food supplies, the world at the end of the 20lh Century is
experiencing a massive loss of momentum in expansion of such
production.
• From 1984 until 1993, grain output per person fell 11%.
• The backlog of unused agricultural technology is shrinking, in industrial
and developing countries alike, slowing the continued rise in cropland
-------
productivity.
''• At present, there is judged to be nothing in sight that would reverse the
worldwide decline in grain output per person.
• 1984 was the last year in which a large increase.in fertilizer use (in the
1 form and manner in which it is produced and supplied) led to a
comparable gain in world grain output.
• Most future growth in the grain output must therefore come from
exploiting technologies not presently being fully used.
The Foreseen Crisis in World Food Production
The Institute further emphasizes imminent world food shortage through its
authorship of an August 14, 1994, Washington Post, news article indicating
that:
• Sharp population increases coupled with declining growth in grain
production are likely to cause serious food deficits in the developing world
during the next 40 years.
• The worldwide output of rice and other grains will increase only by 369
million tons over the next 40 years, approximately one-third of the rate of
increase since 1950.
• Food scarcity will become extremely dire in regions and countries already
experiencing shortages, i.e. Africa, India and China.
Weed for Improved Practice in Fertilizer Formulation
Forecasts of specialists at the World Bank and FAO, i.e. the United Nations
Food and Agriculture Organization, differ from those of the Institute and
expect worldwide grain output other to climb steadily into the next century.
However, projections of these other groups are said by the Institute, to be
based on extrapolations that do not take into account important negative
trends that include decreasing effectiveness with lime of synthetic fertilizer in
the prevailing manner of its formulation and use. FAO reported on April 14,
-------
1997 (Reuter) that its regular Food Outlook (its first forecast for 1997-98)
put world cereal output at approximately 1.880 billion metric tons, compared
to the record 1.873 billion tons produced in the previous year.
Extent of Worldwide Byproduct Market
Amount of Soil Sulfur Shortage
The world soil-sulfur shortage as quantified by The Sulphur Institute,
Washington, D:C., offers a partial measure of both the current and future
market demand for ammonium sulfate. The pattern of agricultural shortage of
soil sulfur, i.e. PNS (plant nutrient sulfur), has become a major problem,
particularly in Asia and North America, as shown for year 1985 in Table 4.
Moreover, world PNS requirements (annual nutrient sulfur shortfall) is
expected, on a "business as usual" basis, to grow by a factor of two by the
year 2000 to 11 million annual metric tons of sulfur equivalent.
Table 4
Shortfall of Plant Nutrient Sulfur, (PNS), 1985 Basis,
Million Annual Metric Tons Sulfur Equivalent
North
Item Asia America Balance World Total
A. Crop Removal of 21 0.9 2.5 5.5
Sulfur
B. Total Sulfur 64 2.7 7.4 16.5
Fertilizer
Requirements*
C. Sulfur Fertilizer 2.7 1.5 6.9 11.1
Consumption
D Net Shortfall in
Sulfur Fertilizer 3.7 1.2 0.5 5.4
Use, B minus C
Basis 33% efficiency in S fertilizer use, i.e. B = A -^ 0.33
-------
FGD Capacity Corresponding to Agricultural Sulfur Supply Shortage
The 1985, North American, 1.2 million annual metric ton shortfall
in agricultural application of sulfur (elemental sulfur equivalent) per Table 3 is
equivalent to 5 million metric tons of ammonium sulfate, which is the annual
ammonia-FGD capacity in 3% sulfur coal service, (0.65 plant capacity factor
with 90% SO2 removal by FGD). The 1985 worldwide shortfall, equivalent
to more than 20 million annual metric tons of ammonium sulfate, (which
byproduct yield that would be available from approximately 20,000 mWe of
can be compared to current and long-term, annual, world, ammonium sulfate
production of 10 million metric tons, approximately half of which is sold in
international trade), corresponds to byproduct available from 90,000 mWe of
future ammonia FGD facilities!
References
1. W. Ellison, "Ammonia-Based FGD Emerging as Alternative for SO2
Removal", ELECTRIC 1'OWli.R International, Fourth Quarter, 1994
2. W. Ellison, "FGD Contenders Challenge Supremacy of Wet Limestone",
ELECTRIC POWER International, Second Quarter, 1997
3. W. Ellison, "FGD Design for Simultaneous SO2/NOx Removal With
Usable Byproduct", presented at the International Joint Power
Generation Conference, Minneapolis, MN (October 1995).
4. R.J. Keeth, P.A. Ireland, and P. Radcliffe, "Utility Reponse to Phase I and
Phase II Acid Rain Legislation - An Economic Analysis", presented at
the EPRI/DOE/EPA 1995 S02 Control Symposium, Miami, FL (March
1995).
5,. D.E. Mattick, "Ebara S02/.NOx Control By-Product", presented at the
Joint ASME/IEEE Power Generation Conference, Milwaukee, WI
(October 1985).
6. Hermine N. Soud, and Mitsuru Takeshita. FGD Handbook. London,
U.K.: TEA Coal Research, IEACR/65, Second Edition, January, 1994, pp
174 and 164.
7. R.J. Morris, "Sulphur The Fourth Major Plant Nutrient", presented at
The Sulphur Institute's International Symposium on Sulphur for Korean
Agriculture, Seoul, South Korea (May 1988).
-------
air
NH-
*- CENTRIFUGE
DRYER
air
particulates
(NH4)2S04
Figure 1
Wet Ammonia FGD With
Crystalline Byproduct
-------
flue gas from
boiler ESP
AMMONIA
STORAGE
TANK
air
ammonium
sulphate
Figure 2
Wet Ammonia FGD With
Compacted Byproduct
-------
VAPOUR COMPRESSOR
EVAPORATOR
AMMONIUM
SULPHATE SOLUTION CONDENSATE
COLLECTOR
FERTILIZER
ADOmVE
STORAGE
TO
AMMONIUM
SULPHATE
STEAM CONOeNSATE FERTILIZER
SILOS
Figure 3
European Granulation Process
-------
PRODUiT RECYCLE
RECYCLED PRODUCT
ASPIRATION D'AIR
AIR OUTLET
AIR DE FlUIDISATION
FLUIDIZATION AIR
SOLUTION
SOLUTION
Figure 4
European Granulator
-------
OPERATING EXPERIENCE
OF THE SIMPLIFIED FGD SYSTEMS
Kimihito YAMAZAKI
International Activities Department
Electric Power Development Co., Ltd.
15-1, Ginza 6-chome, Chuo-ku,
Tokyo, 104 JAPAN
Abstract
hi 1992, Japan proposed "the Green Aid Plan" as a comprehensive assistance program concerning
mergy and environmental technologies for developing countries. As part of this plan, Japan's
Ministry of International Trade and Industry (MITI) proposed to China a project to demonstrate
(te simplified Flue Gas Desulfurization (FGD) systems and has entrusted the execution to EPDC.
This project was organized as a cooperative project of Japan and China, aimed at demonstrating
the feasibility of reducing desulfurtzation cost by 30% in return for a lower but acceptable SO?
removal efficiency. EPDC procured Chinese equipment as much as possible and entrusted the
O&M works to Chinese personnel to facilitate the diffusion of the technology. The following 2
systems were installed:
• Simplified semi-dry FGD system was installed at No. 4 unit (210 MW)of Huang-dao power
plant in Oingdao, Shandong Province, for small power plant.
» Simplified wet FGD system was installed at No. 12 unit (300 MW)ofTaiyuanNo.lpower
plant, Shanxi Province, for bigger power plant.
Various problems which were not expected at the planning stage occurred during the
demonstration. By solving them one by one we reached the stable operation.
This paper describes the operating experience.
Simplified semi-dry FGD
Simplified wet FGD
-------
Simplified Semi-dry FGD System
1. System Outline
The engineering, procurement and construction of the simplified semi-dry FGD system were
performed by Mitsubishi Heavy Industries, Ltd. Table 1 shows the Design Condition.
Item
Table 1
Design Condition
Specification
Gas flow rate
SO2 removal efficiency
Gas temp.
Dust concentration
300,000 Nm3/h
70%atCa/S1.4
Inlet: 145°C
Inlet: 600mg/Nm3
(1 00 MW equivalent)
(Inlet SO2: 2,000 ppm)
Outlet : 65°C
ESP Outlet: 300 mg/Nm3
Fig. 1 shows the Flowsheet. The design condition of Spray dryer is as follows; gas velocity: 2.3m/s,
gas retention time : 10s, atomized slurry volume : Max 18 t/h, atomized slurry size: 70nm, CaO
purity of quicklime: 70%.
FGD Fan
Ash Pit
Fig. 1
Flowsheet of Semi-dry FGD
-------
Fig.2 shows the spray dryer and rotary atomizer. The spray dryer was made as slim as possible (the
diameter is 8.6m) by taking the gas flow pattern into consideration through Computational Fluid
dynamics and model tests. It reduced the space for spray dryer installation.
Directly coupled with high speed motor with inverter control, the innovated rotary atomizer reduces
electricity consumption by 20% and the weight by half. The rotational speed of the rotary atomizer
is set at 8,000 rpm.
i Slim type spray dry scrubber
Flue gas outlet
Fig.2
Spray Dryer and Rotary Atomizer
-------
2. Emissions Control Performance
The flue gas characteristics such as SO2. dust, temp, at Huang-dao No. 4 unit fluctuated widely.
Inlet SO2 was 1,200 ppm on the average. Fig. 3 shows the variability of Inlet SO2 concentration.
Inlet SO2 (ppm)
2000
1000
[July 13th, 19951
_, Time
15=00 18:00 20=00 22:00 24:00
Fig. 3
Variability on Inlet SO2 Concentration of Huang-dao
This system uses the absorbent of Lively Intensified Lime-Ash Compound (LILAC) which is made
from quicklime, fly ash and by-product through a hot water curing process to enlarge the absorption
surface. This absorbent was developed by Hokkaido Electric power Co., Ltd. and Mitsubishi
Heavy Industries, Ltd. Fig. 4 shows the SO2 Removal Efficiency. By using LILAC absorbent,
SO2 removal efficiency was 30% higher than normal slaked Lime. Also ESP shows good
efficiency by lower dust resistivity due to lower gas temp. Dust concentration at ESP outlet was
less than 70 mg/m'N.
LILAC
Normal absorbent
Inlet SO2 1,200 ppm
Approach temp. 15°C
1.2
1.4
1.6
1.8
Ca/S Ratio
Fig. 4
SO2 Removal Efficiency of Spray Dryer
-------
3. Main Problems
• Scaling
Because Inlet gas temp, increased up to max 175°C and SO2 concentration fluctuated more than
500 ppm a day, the scaling problem of spray dryer was caused by non evaporating particles.
We could solve this problem basically by changing to 250,000 Nm3/h of gas to be treated and
increasing the approach temp, to 18°C.
• Abrasion
Because the quicklime purchased at Huang-dao is calcined in the old-fashioned kiln with coal,
it includes a lot of impurities such as fly ash and uncalcined limestones. These impurities
caused the severe abrasion of the rotary atomizer disk.
We could solve this problem by installing hydro-cyclones, which reduced the particle size in the
absorbent slurry to below 100 um. The life of the disc became more than 6,000 hours by
applying zirconia as a liner material, forming a thick calcium protect layer, and manufacturing
the nozzles in tungsten carbide. Fig. 5 shows the Rotary Atomizer Disk.
Fig. 5
Rotary Atomizer Disk
-------
4. Cost Evaluation
The capital cost was approximately 1.7 billion JP¥ (Japanese yen) for the full turn key contract,
including the costs of construction supervisors and the modification for connection with the existing
facilities. The cost of the commercial plant will be much lower because this demonstration plant
includes many extra devices such as advanced monitoring and control systems required for the
evaluation.
The annual O&M cost in!996 was approximately 280 million JP¥, which includes living expenses
for the three Japanese supervisors. The annual O&M cost without all of Japanese support could be
approximately 140 million JP¥. The annual employment cost of the 30 Chinese operators was
approximately 14 million JP¥, and the cost of quicklime was 322 RMB.¥ (Chinese yuan) a ton.
The average utilities consumption at a gas volume 250,000 Nnr/h were 2 t/h of quicklime, 15t/h of
water and 800 kWh of electricity. The electricity and water is provided free by Chinese.
5. Future
The demonstration will be completed in March, 1998 and all the Japanese supervisors will leave
Huang-dao for Japan. After that, the Chinese will make the best use of the demonstration plant on
the basis of the knowledge and know-how acquired through the demonstration.
It remains however understood that EPDC continues to be ready and willing to cooperate with the
Chinese on the follow up to the project, if required. The final evaluation of the project will be
completed by the year 2000.
-------
Simplified Wet FGD System
1. System Outline
The engineering, procurement and construction of simplified wet FGD system were performed by
Hitachi, Ltd. and Babcock-Hitachi K.K. Table 2 shows the Design Condition.
Item
Table 2
Design Condition of Simplified Wet FGD
Specification
Gas flow rate
SO2 removal efficiency
Gas temp.
Dust concentration
Mist concentration
600,000 NnvVh (200 MW equivalent)
80% (Inlet S02: 2,000 ppm)
Inlet: 140°C (max. 170°C ) Outlet: 47°C
Inlet: 500 mg/m'N Outlet: 100 mg/m3N
Inlet: Outlet: 150 mg/m3N
Fig. 6 shows the Flowsheet. Limestone is crushed to below 5 mm size by a hammer crusher, and
pulverized to 100 mesh-pass 95% by a dry tube mill. Power consumption can be reduced to half
by making the pulverized size coarser by adopting 325 to 100 mesh-pass 95%.
f) BeltFiltei{_ j
Waste water
Fig.6
Flowsheet of Simplified Wet FGD
-------
Fig. 7 shows the Absorber. All chemical reactions take place in the single integrated absorber.
Gypsum is withdrawn from the absorber and sent to dewatering. The primary dewatering is
accomplished by hydrocyclones, followed by secondary dewatering in a belt filter.
The compact absorber (Height 12m x Length 17m) was realized by high gas velocity (10 m/s) and
minimized flue gas ducts by adopting the horizontal flow.
L/G and PH are set at 15 ^/Nm3 and 5.7. The spray nozzles are arranged in 3 rows. The first row
is sprayed co-currently, while the second and third rows are sprayed Counter-current to minimize the
amount of mist entrained downstream. As the stack inlet temp, is kept at 90°C by mixing with
untreated gas, a gas reheater was not provided.
flue gas outlei
Flue gas inlet
Fig. 7
Absorber of Simplified Wet FGD
-------
2. Emissions Control Performance
The flue gas characteristics at Taiyuan No. 12 unit fluctuated widely. Inlet SO2 was 1,100 ppm on
the average. Fig. 8 shows the variability of Inlet SO2 concentration.
Inlet S02 ppm
2000 [•
1000
12th, 1997]
10:0
0
12=0
0
14=0
0
Fig. 8
16=0
0
18=0
0
Time
Variability of Inlet SO2 Concentration of Taiyuan
Fig. 9 shows the SO2 removal efficiency, which was 84% at inlet SO2 concentration of 1,100 ppm.
This SO2 removal efficiency corresponds to 81% in the design condition of 2,000 ppm. SO2
removal efficiency could reach 90% by operating three recycle pumps (equivalent L/G=19). Two-
stage chevron type mist eliminators were installed. The mist concentration of FGD outlet was
121mg/Nm3 for FGD inlet 31 lg/Nm3
S02 Removal Efficiency (%)
90-
80.-
70 -
10
Measured
Corrected
Design point
(Inlet SO2 1,100 ppm)
' L/G ratio
15
Fig. 9
20
SO2 Removal Efficiency of Simplified Wet FGD
-------
3. Main Problem
It is known for wet FGD process that limestone dissolution is prevented and PH in the absorber is
lowered due to the increase of aluminum's dissolution from flyash. Similar phenomenon has been
frequently observed at Taiyuan.
4. Cost Evaluation
The capital cost was approximately 2.1 billion JP¥ for the full turn key contract. The cost of the
commercial plant will be much lower because this demonstration plant includes many extra devices.
The annual O&M cost in 1996 was approximately 240 million JP¥, which includes living expenses
for the four Japanese supervisors. The annual O&M cost without all of Japanese support could be
approximately 90 million JP¥. The cost of limestone was 57 RMB.¥ a ton. The average utilities
consumption were 3 t/h of limestone, 50t/h of water and 1,800 kWh of electricity.
5. Future
This demonstration will continue until March, 1999. Further improvement of SO2 removal
efficiency and water consumption will be attempted. The Chinese are building the gypsum factory
to make use of the by-product gypsum. The factory will start the operation in August, 1997.
If we design the FGD plant by the conventional concept for the flue gas property as we experienced,
the increase of the cost will be inevitable. But "the granular limestone FGD system" originated by
Hitachi will break through the barrier. Fig. 10 shows the Flowsheet.
Hydrocy clones
Neutralization Tank
Fig. 10
Flowsheet of Granular Limestone FGD System
10
-------
Limestone crushed to below 5 mm is fluidized by the suction force of the recycle pump in a
neutralization tank. And gypsum produced on the surface of limestone particle is separated by
making use of collisions of particles with each other. It produces much pure gypsum than using
pulverized limestone. Elimination of milling system is one of the advantages of this system.
In addition, it can prevent the masking phenomena caused by the fluoride or alumna.
Demonstration of this system will start in 1998 at Taiyuan No.l power plant. (Patent of this
system has been worldwide applied by Hitachi).
Conclusion
Various problems occurred during the demonstration. The operating ratio of both FGD systems
had decreased because of problems such as unstable water supply and big change in flue gas
properties. By solving them one by one, we achieved stable operation. A lot of experience and
knowledge have been obtained through the demonstration.
As further cost reduction and system improvement are necessary for spreading these simplified FGD
systems, we are planning to improve these systems through the demonstrations. We hope that this
operating experience will be useful for you. Thank you.
References
1. M. Furukawa, K. Tomimatsu and S. Nozawa, "Development of Simplified FGD System", World
Energy Council 16th Congress in 1995, Tokyo.
2. H. Kuroda, S. Nozawa and H. Kaku, "Development of Economical compact Type Wet-limestone
FGD Process", EPRI SO2 Control Symposium in 1995, Miami.
11
-------
Clear Liquor Scrubbing with Anhydrite Production
O. W. Hargrove, Jr.
T. R. Carey
P. S. Lowell
Radian International, LLC
P.O. Box 201088
Austin, TX 78720-1088
F. B. Meserole
Meserole Consulting
8719RidgehillDr.
Austin, TX 78759
R. G. Rhudy
Electric Power Research Institute
P. O. Box 10412
Palo Alto, CA 94303-0813
Thomas J. Feeley, III
Federal Energy Technology Center
Department of Energy
P.O. Box 10940
Pittsburgh, PA 15236
Abstract
The Department of Energy and the Electric Power Research Institute are funding this project to de-
velop the Clear Liquor Scrubbing (CLS) process. The CLS process is an advanced FGD process
designed to remove greater than 99% of the SO, from flue gas, to avoid problems associated with
many existing systems, and to produce a reusable byproduct—anhydrous calcium sulfate (anhy-
drite). Substantial cost reduction and greater byproduct marketing flexibility are projected for this
process. An initial task to confirm the feasibility of the anhydrite process has been completed suc-
cessfully. Further development is currently underway at EPRI's Environmental Control Technol-
ogy Center (ECTC) wet FGD pilot unit. This paper presents the bench-scale results and the initial
findings from the pilot work.
Introduction
This project is funded by the U.S. Department of Energy's Federal Energy Technology Center
(DOE/FETC) under a cost-sharing PRDA with Radian International. The Electric Power Research
Institute (EPRI) is providing co-funding and technical oversight. The project is part of FETC's
Advanced Power Systems Program, whose mission is to accelerate the commercialization of af-
fordable, high-efficiency, low emission, coal-fueled electric generating technologies.
A process concept—Clear Liquor Scrubbing (CLS) with Anhydrite Production—was proposed
and accepted by FETC as a Phase I project in its Mega-PRDA program. The project integrates three
process operations—chloride control upstream of the flue gas desulfurization (FGD) system, a
clear liquor process for enhanced SO, removal performance, and production of anhydrite
(anhydrous calcium sulfate) rather than calcium sulfite or gypsum (calcium sulfate dihydrate). The
CLS-Anhydrite process is an advanced FGD process designed to remove greater than 99% of the
-------
SO, from flue gas, to avoid reliability problems associated with many existing FGD systems, and
to produce a reusable byproduct—anhydrous calcium sulfate (anhydrite). Substantial cost reduc-
tion and greater byproduct marketing flexibility are projected for this process.
The CLS-Anhydrite process fulfills the objectives of the PRDA: an advanced flue gas desulfuriza-
tion (FGD) process that has decreased capital and operating costs, higher SO, removal efficiency,
and better byproduct solids quality than existing, commercially available technology. The CLS
process (which uses a scrubbing liquid with no solids) will be used to accomplish this objective
rather than a slurry liquor process (which contains solids). This CLS project is focused on three
research areas:
• Development of a clear liquor scrubbing process that uses a clear solution to remove SO,
from flue gas and can be operated under inhibited-oxidation conditions;
• Development of an anhydrite process that converts precipitated calcium sulfite to anhydrous
calcium sulfate (anhydrite); and
• Development of an alkali/humidification process to remove HCI from flue gas upstream of
the FGD system.
The anhydrite process also can be retrofit into existing FGD systems to produce a valuable by-
product as an alternative to gypsum. This fits well into another of FETC's PRDA objectives of de-
veloping an advanced byproduct recovery subsystem capable of transforming SO, into a useable
byproduct or high-volume valuable commodities of interest.
Each of the three processes proposed in this program can benefit the power industry individually
and are at different stages of development. Program plans were to develop each of the processes at
the appropriate scale in Phase I. Since the anhydrite process was the least developed, the initial
plans were to perform bench-scale tests in the laboratory followed by further development at the
4-MW pilot scale. The chloride removal and clear liquor scrubbing processes had been tested pre-
viously by EPRI on the 0.4-MW mini-pilot scale, and tests were planned the 4-MW scale for fur-
ther development. Thus, all three processes were to be tested on the pilot unit at the ECTC during
Phase I but will not be operated in an integrated, steady-state manner. This was to be included in
Phase II.
However, the majority of the pilot work in Phase I was devoted to the anhydrite production proc-
ess because of the small scale of development of the anhydrite process before this project, progress
in producing reactive anhydrite from solution, and potential interest in an alternative byproduct by
the utility industry. This paper describes the proposed processes, outlines the test approach, and
preliminary Phase I test results.
Process Description
Figure 1 illustrates the basic flow diagram for the CLS process. This figure includes incorporation
of the alkali/humidification process for HCI reduction and the anhydrite production process. How-
ever, each of the three processes can be applied independently.
Clear Liquor Scrubbing Process
The basic CLS concept is to recirculate clear liquor that contains a sufficient liquid-phase alkalinity
to achieve the desired SO2 removal efficiency without the need for solid-phase alkalinity. The liq-
uor then flows to a limestone reactor and solid-liquid separator that precipitates a calcium-sulfur
solid and returns clear liquor to the scrubber. The CLS process can be operated as an inhibited-
oxidation system (calcium sulfite production) or as a forced-oxidation system (gypsum produc-
tion). Either of these two products can then be converted to anhydrite; however, the economics of
the anhydrite process are more favorable if calcium sulfite is produced in the CLS system.
-------
Limestone
Na,CO,
Calcium Sulfate _^_
Anhydrite Product
Figure 1
Clear Liquor Scrubbing with Anhydrite Production
To be successful, the CLS process must generate solids that are easily dewatered at high rates so
that the high volume of liquor passing through the regeneration system does not require large or
expensive tanks and equipment. Other key aspects for the success of the process include low con-
sumption rates of the buffer, low L/G in the scrubber, and control of the chemistry to achieve good
utilization and low scaling potential. The CLS process had been developed during prior EPRI test-
ing using the 0.4-MW mini-pilot system at EPRTs Environmental Control Technology Center
(ECTC).
Based on the previous studies of the CLS process, several issues require further investigation. The
first need is to verify the findings observed on the mini-pilot system on the pilot system that is 10
times larger. Full-scale design criteria need to be developed and refined at the 4-MW pilot-scale
before a full-scale demonstration is attempted. Variables to be tested at the larger scale include resi-
dence time, pH, suspended solids concentration, recycle ratio, and degree of agitation. In addition,
the potential of combining the CLS process with the anhydrite process needs to be investigated.
The high TDS concentration necessary for the anhydrite process may have a negative effect on the
calcium sulfite solids properties in the CLS process.
-------
Anhydrite Process
As shown in Figure 1, a process that produces anhydrous calcium sulfate (anhydrite) can be added
to the CLS process. The anhydrite process could potentially be added to any conventional FGD
system. Potential advantages of crystallizing anhydrite rather than gypsum from solution include
the following: energy savings over thermally dehydrating gypsum, producing a material of lesser
specific volume than gypsum, potentially displacing gypsum in some commercial applications, and
producing a material for different commercial applications than gypsum. These potential advan-
tages are somewhat speculative because the anhydrite market is currently small relative to that of
gypsum in the U.S. However, if a relatively cheap source of anhydrite is developed, the use of
anhydrite could increase in several different applications that are discussed later.
In a solution saturated with calcium and sulfate, anhydrite, rather than gypsum, is the thermody-
namically stable form above about 100°F. However, typical forced-oxidation FGD systems oper-
ating at 120°F-140°F produce only gypsum. The gypsum product results because the kinetics of
gypsum precipitation are very fast relative to the anhydrite precipitation kinetics. Figure 2 shows
the solubility of gypsum and anhydrite as a function of temperature and ionic strength. Both solids
become more soluble at lower temperatures, but only gypsum solubility is affected by ionic
strength. This property is key to the anhydrite process.
Gypsum - low Gypsum - high
Anhydrite - • • ,
J ionic strength ionic strength
Increasing Temperature
Figure 2
Calcium Sulfate Solubility Data
In the proposed process, calcium sulfite produced in the CLS process is converted to anhydrite in a
secondary reaction tank by oxidation of sulfite to sulfate and subsequent precipitation of anhydrite.
As discussed above, the kinetics of anhydrite precipitation are relatively slow at typical FGD oper-
ating conditions. To increase the precipitation rate of anhydrite, the process must be operated at an
elevated temperature and at a high ionic strength. As shown in Figure 2, gypsum solubility in-
creases with ionic strength while anhydrite solubility does not. Based on information in Figure 2,
conditions can be chosen that provide a driving force for anhydrite precipitation while the solution
remains subsaturated with respect to gypsum (i.e., gypsum will not be formed). Points A and B in
Figure 2 represent such points at lower and higher ionic strengths, respectively. The driving force
for anhydrite precipitation is represented by the distance between either Point A or B and the solid
-------
line. By increasing the solution ionic strength, the driving force for anhydrite precipitation can be
increased without forming gypsum solids.
Previously conducted batch tests determined the precipitation rate of anhydrite at 140°Ffrom a high
ionic strength solution. Based on these data, an anhydrite precipitation rate was calculated from the
slope of a fitted exponential function. This rate was sufficiently high to warrant further investiga-
tion to determine how the rate might be further increased. The success of the anhydrite process de-
pends on finding a set of operating conditions under which the anhydrite precipitation rate is suffi-
ciently fast to yield an economically sized reaction tank and which minimizes gypsum production.
Alkali/Humidiflcation Process
Maintaining a low chloride concentration in a FGD system has several potential benefits. Corrosion
should be minimized at low concentrations that may allow materials of construction to be cheaper.
In limestone systems, the limestone utilization will generally be higher. Finally, the scaling poten-
tial is reduced in the CLS process.
EPRI has examined a number of chloride control options including prescrubbing, tail-end separa-
tion or concentration, and injection of scrubber blowdown into the flue gas. The first two options
are relatively expensive and result in some waste water streams. The last option has been tested
with limited success, but chloride removal is limited by off-gassing of HC1 in the process of drying
the liquid.
The proposed chloride removal process involves injection of an alkaline slurry upstream of an ESP
or other particulate removal device. HC1 in the flue gas reacts with the alkali to form a solid which
is dried by the flue gas and collected in the particulate removal device. EPRI has developed a simi-
lar process to react flue gas SO3 with alkali reagent to mitigate opacity problems and has completed
considerable work on the effect of injection on SO2 removal.
Bench-Scale Testing
Approach
Development of the anhydrite process began with a series of batch and flow-through reactor tests
in the laboratory. The objective of these tests was to provide kinetic data over a range of operating
conditions. Batch test variables included solution ionic strength, temperature, and initial liquid-
phase calcium-to-sulfate ratio. These tests were conducted by generating the desired initial solution
composition with reagent grade chemicals and seeding the batch reactor with anhydrite. The anhy-
drite seed material was generated in the lab by thermally converting gypsum in a high ionic strength
solution at 80-90°C. Complete conversion of the gypsum to anhydrite was confirmed using thermal
gravimetric analysis (TGA) and Infrared (IR) spectroscopy. During each batch test, the liquid cal-
cium and sulfate concentrations were measured with time by analyzing liquid samples with either
atomic absorption spectroscopy (AA) or ion chromatography (1C). Changes in the calcium and/or
sulfate concentration with time yield the anhydrite precipitation rate. Solids collected at the end of
each test were analyzed using TGA and IR to confirm that no gypsum was generated.
Based on the batch test results, laboratory flow-through reactor tests were conducted at the most
promising conditions. Variables included temperature, ionic strength, residence time, solids load-
ing, and seed crystal size. The objective of these tests was to provide determine the best conditions
for initial pilot-testing. Liquid flow-through tests generally provide more realistic kinetic data than
batch tests since the data is generated at a constant liquid composition (i.e., constant anhydrite rela-
tive saturation).
-------
Figure 3 illustrates the configuration used during the flow-through tests. The desired solution
composition was established, and a known charge of anhydrite seed material was added to the re-
actor at the beginning of each test. Throughout each test, liquid was filtered from the reaction ves-
sel while '"fresh" solution, saturated with respect to gypsum, was added to the reaction vessel. In
this way, these experiments approached steady-state with respect to the solution chemistry. Since
the charge of seed material was large relative to the mass of anhydrite precipitated during most
tests, the change in solid surface area was small, and changes in the soluble species concentration
was also small. The rate of reaction was determined by a liquid-phase material balance between the
inlet and outlet streams. The weight percent solids was measured at the beginning and end of each
test to confirm the liquid-phase measurements, although the error in the solid-phase balance is
much larger given the relatively small change in suspended solids concentration.
Solution from M"e"°
Pump
gypsum saturator ^/~*\
(dissolved spcaes ^ 1
—Na". Mg2*. Cl —
Sampling
Bottle
Figure 3
Bench-Scale Flow-Through CrystaUizer
Bench Results
This section presents the results of batch and flow-through laboratory experiments. In both batch
and flow-through tests, anhydrite precipitation rates were measured that are equivalent to gypsum
precipitation rates at typical FGD conditions (0.06 g/hr-g seed at 125°F). This indicates that tanks
of an economically attractive size may be used in the production of anhydrite at the commercial
scale, so long as another process not simulated in the laboratory (such as calcium sulfite dissolu-
tion) does not become rate limiting.
A total of 13 valid batch reactor tests were completed. The precipitation rate continuously changes
throughout a batch experiment and can be calculated by measuring the change in solution composi-
tion over time. By comparing the anhydrite precipitation rate 10 minutes into each run, the effects
of several important variables were determined. Although the amount of solids precipitated was
relatively small compared to the initial seed inventory, IR andTGA analyses indicated that gypsum
was not formed during the batch experiments.
The effect of temperature on the anhydrite precipitation rate in an NaCl solution is shown in Fig-
ure 4. These results show the expected increase in reaction rate as the temperature increases over a
25°C range. If the rate constant were to exhibit an Arrhenius-type relationship with temperature (k
A -~E*/R ), a plot of the log of the precipitation rate versus 1/T (in °K) should yield a straight line.
-------
Figure 4 shows this relationship. The precipitation rate increases by slightly more than a factor of
two for each 10°C temperature rise over the temperature range.
According to theoretical calculations (see Figure 2), the anhydrite precipitation rate should be
greater at higher ionic strength (higher concentration of dissolved solids). This effect was observed
in the batch tests as shown in Figure 5. There was more than a 20-fold increase in the precipitation
rate as the NaCl concentration increased from by a factor of 6. Clearly, for anhydrite production to
be successful, higher ionic strength solutions have to be used.
T, + 12°C
i/r, LTK ^
Figure 4
Effect of Temperature on Batch Precipitation Rate
{
I
NaCl Concratration
Figure 5
Effect Solution Strength on Anhydrite Precipitation Rate (NaCl Solutions)
Several liquid flow-through, batch solids runs were conducted to confirm rates measured in the
batch experiments. The batch rates are determined by differentiating concentration data obtained
from a solution that is continually changing in composition. The flow-through tests approach a
steady-state solution composition more similar to a continuous full-scale process, and, for this rea-
-------
son, the flow-through results are considered to be more representative. A total of 12 valid flow-
through experiments were completed.
The initial runs were conducted such that the composition of the inlet stream (calculated after as-
sumed mixing) was slightly subsaturated with respect to gypsum. Different anhydrite seed loading
and different flow rates were tested at different temperatures to develop an understanding of the
steady-state precipitation rates. The pH was also decreased substantially in one run and iron added
to the reactor in another run to determine if either had a positive effect on the precipitation rate.
The anhydrite precipitation rates measured in the first several tests were much lower than the initial
rates measured in the batch experiments (based on comparison of batch rates at similar solution
composition and temperature). Changes in the feed rate to the reactor, calcium concentration in the
feed solution, and amount of anhydrite in the initial charge to the reactor were made. After these
changes were made, anhydrite precipitation rates of 0.03 to 0.04 g/g^-hr were achieved. These
anhydrite precipitation rates in the flow-through reactor compare favorably with gypsum precipita-
tion rates of about 0.06 i/g^-hr at typical FGD forced oxidation conditions.
The effect of pH and iron (a known modifier of crystal surface reactions) addition were examined
in an attempt to increase the anhydrite formation rate. The effect of pH was negligible. The iron
addition decreased the anhydrite precipitation rate compared to runs without iron that were con-
ducted under otherwise similar conditions. Apparently iron affects the anhydrite surface reaction
but in a deleterious manner.
The final two runs were conducted at two different temperatures with substantially higher feed
stream calcium concentration to determine if the precipitation rate could be further increased. The
experiment conducted at the higher temperature showed that a rate of 0.11 glg^-bf was achieved.
Analysis by IR indicated that no gypsum was formed. At a lower temperature, a rate of 0.8 g/gs^-
hr was measured, but the IR analysis indicated presence of substantial amounts of gypsum at the
end of the run.
Pilot Testing
Approach
Figure 6 shows the equipment configuration used to test the anhydrite process at the ECTC. The
major objective of the pilot testing was to demonstrate that anhydrite could be produced from FGD-
generated calcium sulfite slurry. Once it had been demonstrated that anhydrite production was fea-
sible, important variable interactions were to be explored. Several modifications were made to the
original test plan because the pilot anhydrite reactor was an existing pilot reaction vessel, not a
newly constructed one. Because of some limitations, test conditions had to be modified to achieve
the desired reaction temperature. However, a good understanding of several important variable in-
teractions was developed during the Phase I testing.
A number of techniques were employed to determine the phase of calcium sulfate produced in the
pilot unit including:
• % Bound Water (measure solid weight loss at about 180°C) - provides estimate of anhydrite
wt.% assuming solids are mixture of only anhydrite and gypsum;
• X-Ray Diffraction (XRD) - indicates the presence of gypsum, hemihydrate, and/or anhydrite;
and
• Visual Microscope and Scanning Electron Microscope (SEM) - indicate changes in particle size
and shape.
-------
The percentage of bound water and the visual microscope were available on site and were the pri-
mary means of characterizing the solids to provide operating feedback. In using the bound water as
an indication, it was assumed that the solids were either anhydrite or gypsum. For 100% gypsum
solids, the bound water should be about 21% by weight. Anhydrite should have no bound water.
Visual observations of the centrifuge product were also valuable once experience had been gained.
Lime or Limestone for
HC1 Control
from
NYSEC's
Kinligh
Spray
Dryer
t
ESP
Return to Kintigh
Station
Anhydrite
Product
PSD
jQ Lirnei
Figure 6
ECTC Flow Configuration for Anhydrite Test Block.
The XRD and SEM were used in Radian's Austin laboratories to confirm on-site results.
Pilot Results
The initial operating conditions specified did not produce anhydrite at the ECTC pilot unit. Many of
the problems were associated with converting an existing reaction tank to the anhydrite reaction
tank. To achieve the shorter residence times specified required the operating slurry level to be rela-
tively low which increased the air rate required to achieve complete sulfite oxidation. Since one of
the major heat losses associated with this process is the heat of vaporization of water vapor leaving
with the spent oxidation air, the higher air rate prevented the target temperature from being reached.
There were other heat losses as well, including inadequate insulation around the tank and lines to
the pH measurement and sampling locations. Much of the first month was spent understanding and
resolving the temperature limitations of the pilot anhydrite reactor.
-------
Once the temperature issues were resolved satisfactorily, anhydrite was produced with varying de-
grees of success for the remainder of the pilot testing. Figure 7 shows how the percentage anhy-
drite changed as a function of time during the first 3 months of testing. For purposes of simplify-
ing the presentation, Figure 7 considers the solids produced to be either gypsum or anhydrite. Ac-
tually, there were periods in which other hydrate phases was also observed, but the predominant
phases were gypsum and anhydrite.
Period 1 was the first attempt to make anhydrite during which the temperature issues had to be re-
solved. The target temperature was not achieved until late January. During January, Figure 7
shows that some anhydrite was formed, but only a maximum of about 50%. Near the end of Janu-
ary the temperature dropped in the reactor and the anhydrite converted back to gypsum. Figure 8
shows the SEM results for the gypsum formed at the end of Period 1. An of the XRD scan
showed the sample to be all gypsum.
Wed 1/22 Sat 2/31 Tue 2/11 Fn 2J21 Won 3/03 Thu 3/13 Sun 3/23 Wed 4/02 Sat 4/12
|^Penod1>|
-------
Figure 8
SEM of Gypsum in Period 1
Figure 9
SEM of Period 2 Solids
Following the drop in anhydrite content at the end of Period 2. it was postulated that anhydrite
would form initially small crystals which would subsequently grow in size until the surface area
was insufficient to maintain the gypsum relative saturation below 1. Once the gypsum saturation
exceeds 1 by any significant amount, gypsum will begin to form rapidly because of its faster pre-
cipitation kinetics. It was therefore decided to control the anhydrite surface area through external
means. The initial method of surface area control began on February 19 and concluded on
March 11.
Figure 7 shows that the anhydrite content increased rapidly following initiation of anhydrite surface
area control. The XRD of a sample taken on February 21 showed about equal amounts of hemi-
hydrate and anhydrite with no gypsum. Over the next week, the solids gradually changed until they
were mainly anhydrite with small amounts of hemi-hydrate and gypsum. The centrifuge product
hardened upon cooling similar to the previous operation when anhydrite and hemi-hydrate were
present.
11
-------
Over a weekend, the characteristics of the centrifuge product changed dramatically. On March 3,
these solids were hardening rapidly upon cooling and the solids were forming one large rock in the
collection dumpsters used at the ECTC. When the dumpsters were emptied into the trucks for ulti-
mate disposal, the solids were discharged in a single piece that did not break apart when colliding
with the truck bottom as had been observed previously. The bound water measurement and the
XRD anaJyses showed the solids to be 80 to 90% anhydrite with the remainder gypsum. These
characteristics were observed until the end of the period when method of surface area control had
to be changed. Figure 10 shows the SEM results from March 7 (end of Period 3). The XRD
shows a majority of the solids to be anhydrite with small amount of gypsum. The SEM shows a
change in crystal size and shape with smaller crystals that appear to be "stacked platelets" under the
visual microscope.
Materials and equipment limitations necessitated a change in the method used for anhydrite surface
area control on March 11. Over the next two days, the anhydrite content decreased from 85% to
65% on March 14. The centrifuge solids appeared to be drier and did not set as quickly or as hard.
Experimentation with methods of surface area control continued throughout March. Figure 7
shows a gradual decrease in the anhydrite concentration through the latter half of March. The only
exception was on March 17 when samples taken 17 showed 82 to85%> anhydrite, and the centri-
fuge product had returned to its more reactive properties. The surface area control method used at
this time could not be maintained for extended periods of time with the equipment on site. The an-
hydrite content dropped to as low as 33% during this period. Figure 11 shows the SEM of solids
produced toward the end of Period 4. The XRD results indicated a large amount of gypsum with
some anhydrite. The SEM shows some very large gypsum crystals as well as some of the finer
anhvdrite crystals.
Figure 10
SEM of Solids from Period 3
A successful method of control anhydrite surface area was re-established in late March, and the
anhydrite concentration increased to 90% by mid-April. Properties of the anhydrite product re-
turned to those observed during Period 3.
Application
Prior to beginning the pilot anhydrite test program at the ECTC, a preliminary market survey to
identify the most promising uses of anhydrite was conducted. Perhaps the most attractive alterna-
tives include the self-leveling floor and aggregate markets.
12
-------
Figure 11. SEM of Solids from Period 4
Self-Leveling Floors
The use of anhydrite in Europe for self-leveling floor screeds is successful and growing. Some of
the earliest use has been in Germany and the Netherlands'. There are some significant differences
between the production of anhydrite by one of leading European processes and our FGD solution
precipitation method. The European method heats FGD byproduct gypsum to 800°C by a series of
cyclones that contact hot air (produced through combustion of natural gas, oil, or coal) with the
gypsum in a countercurrent manner. During the process, the gypsum is first dried, then converted
to hemihydrate, and ultimately to anhydrite as the solids reach 800°C. The product crystals have a
limited internal pore volume, probably as a result effusion at high temperature. The product anhy-
drite is pulverized to obtain the desirable fluid properties.
The FGD anhydrite solution precipitation process should be much more thermally efficient than the
Dutch process. This is major advantage since fuel would be expected to be a major operating cost
in the Dutch process. On the other hand, the anhydrite produced in the Dutch process may be more
reactive and therefore suitable for the self-leveling floor application. However, FGD precipitated
anhydrite should have reactive characteristics similar to synthetic byproduct anhydrite from the hy-
drofluoric acid manufacture. This material has been successfully employed in the application of
self-leveling floors. The Canadians have examined a process to produce self-leveling floors from
natural anhydrite produced in Nova Scotia." Application of FGD anhydrite in self-leveling floors is
one goal of Phase II of the PRDA project.
Synthetic Aggregate
The insoluble calcium sulfate anhydrite (CaSO4) solids produced as a byproduct from wet lime or
limestone FGD systems will, upon rehydration, "set up" to form a coherent, hard mass. This
property, which is not characteristic of other byproduct options such as calcium sulfate dihydrate
(gypsum), may yield potential market opportunities for the anhydrite byproduct option. In survey-
ing the potential market for anhydrite as a synthetic aggregate, an existing commercial operation
was identified in which more than 500,000 tons per year of anhydrite waste from hydrofluoric acid
production is being converted to aggregate and sold as a substitute for crushed stone. This anhy-
drite product is being used by a number of municipal and county agencies as well as private com-
panies for construction fill and road base material. Assuming that this material has similar proper-
ties to those of FGD anhydrite, there appears to be a large potential market for this material in or
near urban areas where conventional crushed-stone aggregates are in short supply.
13
-------
There does not appear to be any potential for using byproduct anhydrite to produce aggregate for
use in concrete. In this use, sulfate is known to react with concrete, causing undesirable expansion
and cracking.
Future Activities
Currently a final report on this project is being prepared and a plan to optimize the anhydrite proc-
ess and test its use in potential markets is being developed.
Acknowledgments
This research is being sponsored by the U.S. Department of Energy's Federal Energy Technology
Center (Pittsburgh) under contract DE-AC22-95PC95253. The authors thank Dr. Perry Bergman,
FETC's original Contracting Officer's Technical Representative, for his interest and support during
the initial stages of the Phase I effort. Carl Richardson and Ron Skarupa also made key contribu-
tions to this work as did Parsons Power (ECTC operations contractor), Mr. J. H. Wilhelm, and
ORTECH Corporation.
The work presented in this paper is partly the result of research carried out at EPRJ's Environ-
mental Control Technology Center (ECTC) located near Barker, NY. We wish to acknowledge the
support of the ECTC cosponsors: New York State Electric and Gas, Empire State Electric Energy
Research Corporation, Electric Power Development Corporation, and the U.S. DOE.
References
' Kappe, J., "High Quality Anhydrite from Flue Gas Desulphuization Gypsum," Environment and
Technology, November 1991.
- ''Anhydrite Self-Leveling Flooring Screed: Final Report," Microlog (233440-1-9044-01-SQ.)
14
-------
HIGH VELOCITY WET SCRUBBING OF SO2 AND NOt
Bruce W. Lani Ronald W. Breault
Manyam Babu Tecogen Division
Dravo Lime Company Thermo Power Corporation
3600 Neville Road 45 First Avenue
Pittsburgh, PA 15225 Waltham, MA 02254
Abstract
Recent proposals by the EPA and regional authorities to further reduce SO2, NOX, and
fine particulate emissions will require fossil fuel combustors, including those utilizing
low sulfur fuels, to implement more stringent air pollution control measures. To address
these new emission limits, Dravo Lime Company (DLC) in conjunction with Tecogen is
developing a two stage wet scrubbing process which operates at gas velocities up to 25
ft/sec for the removal of SO2 and NOX. This process utilizes DLC's ThioClear® or
Thiosorbic magnesium-enhanced lime technologies for SO2 and HNOs capture with
Tecogen's TecoLytic™ corona reactor for converting NO to HNOs. The by-products of
this process are wallboard quality gypsum, magnesium hydroxide, an excellent reagent
for water treatment, and a commercial grade calcium nitrate solution. This paper reports
on the process concepts and preliminary bench scale evaluations in preparation of testing
to be conducted at DLC's 4.5 MW wet scrubber pilot plant located at the Miami Fort
Station of the Cincinnati Gas and Electric Company. This project is a cooperative effort
of the Ohio Coal Development Office, Cincinnati Gas and Electric Company, Tecogen,
and DLC.
Introduction
The 1990 Clean Air Act Amendments have made the control of SO2 and NOX emissions a
prominent national issue. These emissions are leading precursors to acid rain and fine
particulates. Furthermore, NOX is identified as a strong contributor to photochemical
smog. Presently, utilities are addressing SO2 reductions by either switching to lower
sulfur containing fuels or utilizing conventional wet scrubbing technologies. With
regards to NOX emissions, the reductions have been achieved by combustion zone
modifications.
As a result of existing and proposed regulations, SO2 and NOX emissions will be
restricted further to the point where fuel switching and combustion modifications by
themselves will not be sufficient to achieve the proposed emission limits. Therefore
advanced control processes will be required to limit the SO2 and NOX emissions. Such a
process is the ThioClear / TecoLytic process which is described in this paper.
-------
Process Technology
The ThioClear / TecoLytic process is the integration of three technologies for the
efficient and cost effective treatment of combustion gases including those resulting from
the utilization of high sulfur fuels. The conceptual commercial scale process shown in
Figure 1 will incorporate a two-stage high velocity horizontal scrubber operating at a
nominal gas velocity of 25 ft/sec. The first stage will utilize DLC's ThioClear process to
remove up to 98% of the SO2 and generate salable by-products, gypsum and magnesium
hydroxide. The second stage will incorporate Tecogen's TecoLytic reactor module
which will convert insoluble NO to extremely soluble HNO3 and a scrubbing section
which will capture the HNOs. To achieve cost effective NOX reductions to the 0.1 to 0.15
Ib/MMBtu level, the TecoLytic process will work in conjunction with combustion
modifications. Low NOX burners, air and/or fuel staging will be utilized to achieve NOX
levels to 0.3 Ib/MMBtu. The TecoLytic process will further reduce these emissions by 50
to 66%. The captured HNOs will be neutralized with lime to produce the salable by-
product, calcium nitrate. Preliminary economic analysis indicates that this process would
be cost competitive with a similar scenario utilizing SCR as the NOX removal system.
Figure 1. Conceptual ThioClear* /TecoLytic™ Process
The Horizontal Scrubber
There are a number of advantages to utilizing a horizontal scrubber operating at much
higher flue gas velocities over the current design standard of 10 ft/sec for the capture of
-------
flue gas emissions. For greenfield and retrofit applications, there would be smaller
and/or fewer scrubber modules required to scrub the total flue gas from the combustor,
greatly reducing the initial capital investment. The utilization of horizontal flow
demisters which are an integral part of the horizontal scrubber are more efficient at
eliminating carryover than vertical flow demisters which are limited to gas velocities
below 20 ft/sec. Additional advantages of the horizontal scrubber include reduced height
requirements when compared to vertical scrubbers and minimal impact in connecting to
existing ducting. These advantages translate into greater flexibility in materials of
construction and lower power consumption by pumps for meeting hydraulic head
requirements and by fans due to lower inlet and outlet pressure losses.
Aside from the reduction in capital and operating costs, improvements in scrubbing
efficiency can be realized by operating at higher velocities and/or utilizing mass transfer
devices. The scrubbing liquor associated with a magnesium-enhanced lime FGD system,
such as the ThioClear process, is highly alkaline due to the presence of magnesium
sulfite. It rapidly neutralizes the captured SO2 making the absorption process controlled
by the rate at which SC>2 transfers from the gas phase to the liquid phase. To take
advantage of this alkalinity, a mass transfer enhancement device can be installed and/or
the scrubber operated at a higher gas velocity.
The relationship between the mass transfer coefficient and SC>2 removal is defined by the
following equation:
where: NTU 862 removal as number of transfer units (dimensionless)
H height of scrubber (ft)
KB overall mass transfer coefficient (— - )
8 ft'-hr-atm
a mass transfer surface area per unit volume (ft2 / ft3)
G flue gas flowrate (lb~mole)
ft -hr
P total system pressure (atm)
Therefore, the benefits of utilizing mass transfer devices and operating at higher
velocities will have a positive impact on SO2 removal provided that the product of the
enhancements to the mass transfer coefficient (minimizing film resistance by increased
turbulence of high velocity operation) and the surface area (increase of interfacial mass
transfer area resulting from the mass transfer device) is greater than the detrimental off-
set due to the increased gas velocity. The reliance on the mass transfer characteristics of
the spray droplets alone are diminished and the amount of liquor required to treat the
same volume of gas can be reduced.
Utilization of the high velocity horizontal scrubber in the crossflow mode of gas/liquid
contact is not new to the FGD industry. However, performance and reliability of these
-------
earlier scrubbers were adversely effected as the length of the vessel was reduced.
Nonuniform gas flow distributions were created in the spray zone from top entry spray
nozzles that compressed the gas toward the scrubber floor and reduced the effectiveness
for SC>2 removal. The combination of the skewed velocity profile and spray headers
installed immediately in front of the demisters contributed to excessive carryover exiting
the mist eliminators.
The horizontal scrubber installed at the Miami Fort pilot plant was designed to minimize
the above drawbacks of the prior commercial designs. The cross sectional dimension of
the scrubber is 51" x 25" Each of the three available recycle pumps, capable of moving
175 gpm, supply two spray headers, each equipped with two Bete TF40 full cone spiral
style nozzles. These spray headers are located within the cross sectional area of the
scrubber to minimize gas maldistribution and are oriented counter-current to the gas flow
to enhance mass transfer. Taking advantage of the nonscaling tendencies of the
magnesium-enhanced lime FGD process and minimal suspended solids in the recycle
liquor, multiple layers of Kimre Kon-Tane 37/97 tower packing can be installed
downstream of the spray zone. The packing further reduces gas maldistribution,
enhances mass transfer, and acts as a predemister. The demister section consists of two
stages of horizontal impingement separators with three wash headers on a time sequence.
Prior testing of magnesium-enhanced lime FGD processes in the pilot plant horizontal
absorber has demonstrated many of the advantages described above. Figure 2 is
representative of the resulting velocity profile between the packing and the first stage of
the demisters. The plug-like flow profile insures that the SC>2 removal capabilities of this
scrubber and the performance of the mist eliminator will not be degraded by velocity
variations within the scrubber. The results of the parametric SC>2 removal studies are
detailed in Figure 3. As can be seen, increasing the velocity of the flue gas from 20 ft/sec
to 25 ft/sec permitted a decrease in the L/G to achieve the same SC>2 removal.
Figure 2. Horizontal Scrubber Velocity Profile
VELOCITY (fps)
15
-------
Figure 3. Horizontal Scrubber-SO2 Removal
98
86
63
20
30
LJG
40
50
60
One must be cautioned concerning the direct interpretation of pilot scale scrubber
performance to commercial size facilities. In general, the SO2 removal achieved at the
pilot plant is lower than that of a utility FGD scrubber at similar operating conditions.
The droplets exiting the spray nozzles in the scrubbers utilized for these studies impact
the walls of the scrubbers within 1 to 1.5 feet from the nozzle. The liquor lost to the
walls tends to have minimal beneficial impact on SO2 removal. This is not the case in the
utility scrubbers where greater utilization of the liquor is achieved due to minimal spray
contacting the walls. Prior testing at the Miami Fort pilot plant has shown that up to 50%
of the liquor entering the scrubber does not contribute to SO2 removal when duplicating
operating conditions of commercial facilities.
Due to the prior successful demonstration of high removal efficiencies of SC>2 in a
horizontal absorber utilizing a magnesium-enhanced lime FGD process, this effort will
not be repeated for this project at the Miami Fort pilot plant. The existing horizontal
absorber will be retrofitted with the TecoLytic reactor module and removal studies will
focus on the capture of the oxidized NO by-products (NO2 and HNO3) resulting from the
corona device. The same concepts that have been effectively demonstrated in enhancing
SO2 removal will be utilized for the removal of the NO by-products.
The ThioCleat® Process
To improve the economics of the conventional magnesium-enhanced lime (Thiosorbic)
wet scrubbing FGD process that produces a by-product that requires disposal by
landfilling, DLC developed the ThioClear process. ThioClear utilizes a magnesium
based absorber liquor which contains less than 1% suspended solids to capture SO2.
Magnesium-enhanced lime regenerates an oxidized bleed stream of scrubbing liquor to
produce magnesium hydroxide and gypsum as by-products.
-------
A detailed description of the process is shown in Figure 4. The ThioClear process takes
advantage of the catalytic effect of magnesium to increase the alkalinity of the scrubbing
liquor. Magnesium hydroxide (Mg(OH)2) is added to the recycle tank where the pH is
maintained in the range of 6.0-6.5. The hydroxide reacts with dissolved SC>2 in either the
recycle tank or the absorber to form the soluble alkaline salt, magnesium sulfite
(MgSOs). The sulfite rapidly neutralizes additional absorbed 862 by forming magnesium
bisulfite (Mg(HSO3)2). This capability provides the scrubbing liquor additional capacity
in absorbing SC>2 compared to systems where magnesium sulfite is not present.
Figure 4. ThioClear® Process
so.
/ S02 + H20 * HSOi* H-
A — MgSO3 + H- + HSOJ ^ Mg(HSOJ2
\ Mg(OH)2 + S02 ^ MgS03 + H2O
/ Mg(OH)2+S02 ^ MgS03 + H20
B \ Mg(OH)2 + Mg(HSOj)2 * 2MgSO3
/ MgSOj + % O2 ^ MgSO4
C ^— Mg(HSO3)2 + O2 ^ MgSO4 + H2SO4
• Mg(OH)2 + H2SO4 * MgSO4 + 2H2O
D — MgSO4 + Ca(OH)2 ^ Mg(OH)2 + CaSO4- 2H2O
MgSO,
Mg(OH)2
Unlike typical FGD operations, the blowdown from the scrubber to the oxidizer in the
ThioClear process is a slipstream of low pH liquor from the gas/liquid contact zone. The
control of this effluent flowrate and pH is important to the ThioClear process.
Modulating the blowdown flowrate provides control of the scrubber alkalinity via the
liquor magnesium concentration and thereby provides excellent scrubber performance
during plant start-up and boiler load changes. The lower pH associated with the
blowdown improves oxidation.
The scrubber effluent feeds low pH liquor in the range of 5.5-6.0 to the oxidizer.
Oxidation of the liquid phase magnesium sulfites and bisulfites produces magnesium
sulfate (MgSO^). Due to formation of sulfuric acid (H2SO4) from the oxidation of the
-------
bisulfites, magnesium hydroxide additions are required to maintain the pH of the oxidizer
between 4.5-5.5.
A second stream entering the oxidizer is the gypsum-rich hydroclone underflow.
Recycling the process gypsum to the oxidizer before final filtering accomplishes multiple
tasks. The recycle stream provides sufficient crystal surface area to eliminate scaling
potential that can result from calcium sulfate formation during oxidation. Additionally,
the magnesium hydroxide impurities in the gypsum stream are effectively dissolved in
the low pH regime. Therefore, the amount of process magnesium hydroxide required for
pH control is decreased and the recovery of magnesium is maximized. Thus, minimal
solid phase magnesium compounds will leave the process with the gypsum after filtering.
The gypsum slurry exiting the oxidizer is fed to a continuously operating filter.
Utilization of cake washes aid in the recovery of the magnesium sulfate liquor and
improve the purity of the gypsum cake. The gypsum cake is deposited on a lay down
area while the magnesium sulfate filtrate is returned to the process.
The recovered filtrate is returned to the regeneration tank. Slaked magnesium-enhanced
lime that is composed of+90% calcium oxide (CaO) and 4 to 8% magnesium oxide
(MgO), is added to the regeneration tank to maintain the pH between 10.0-10.8. The
resulting mixture precipitates the sulfates as gypsum (CaSO^BbO) and regenerates the
magnesium species to magnesium hydroxide.
Taking advantage of their size differential, separation of the magnesium hydroxide and
gypsum crystals from the regeneration tank slurry is achieved by the use of two banks of
hydroclones in series. The magnesium hydroxide-rich overflow from the first bank of
hydroclones is delivered to a thickener for settling and thickening. The thickened
magnesium hydroxide is used for pH control in the absorber and the oxidizer. The excess
production is available as the salable by-product. The hydroclone underflow is diluted
with the clear liquor leaving the thickener. The resulting slurry is fed to the second bank
of hydroclones for a second stage of physical separation. The liquor from the overflow of
the second hydroclone returns to the recycle tank to maintain liquid level. The gypsum-
rich underflow stream is recycled to the oxidizer for eventual filtration as described
previously.
Two of the more prominent features of the ThioClear process are its efficiency in
removing SOj and the purity of the gypsum produced. Comparisons of limestone forced
oxidation spray tower pilot plant data to that generated from the ThioClear process in the
vertical spray and tray towers of the Miami Fort pilot plant are shown in Figure 5. Since a
15 ft/sec condition was not tested with the ThioClear process, an interpolation of the 14
and 16 ft/sec data was conducted to provide a trend line for the comparisons. As evident
from the graph, the highly alkaline scrubbing liquor associated with magnesium-
enhanced lime and the resulting approach to gas-film limited mass transfer characteristics
enable exceptionally high SO2 removals with minimal L/G when compared to the
limestone scrubber operating at similar velocities.
-------
Figure 5. Scrubber Reagent Effect on Performance
LSFO Spray @10fps
LSFOSpray@15fps
M g-Um e Spray @ 10f ps
M g-Ume Spray @ 15f ps
M g-Ume Tray @10fps
M g-Ume Tray @ 15fps
o™
100 110 120 130 140
For the ThioClear process to be feasible, sufficient quantities of magnesium hydroxide
need to be produced and recovered to operate the process and provide a by-product
stream. In ThioClear, this is accomplished by recycling the gypsum stream from the
hydroclones to the oxidizer and continuous filtration. The resulting increase in the
recovery of magnesium has resulted in a very pure snow white gypsum filter cake. The
purity of the gypsum and magnesium hydroxide by-products is shown in Figure 6.
Figure 6. ThioClear® By-Product Purity
98.20
• Mag Hydroxide
Gypsum '
20%
0% -1
Gypsum
Magnesium Hydroxide
-------
The TecoLytic™ Process
The TecoLytic technology is based on using a cold, reactive plasma to generate chemical
reactions that result in the destruction of the NOX molecule, either into nitrogen and
oxygen or other specie forms which can be more easily removed from the flue gas and
sold as a by-product. In operation, a high electric field is produced to enable free
electrons to be accelerated to an energy level sufficient to dissociate some gas molecules
producing free radicals. The chemically reactive species then initiate the NOX reactions,
much in the same way as conventional catalysts, except over a wide temperature range
(less than 200°F to over 1200°F). An important technology feature is that the flue gas
contains the reactants and therefore the use of a reducing agent is not required.
The TecoLytic corona reactor has a unique geometry from which its advantages derive
relative to conventional corona reactors. Conventional corona reactors, as used on a large
scale in ozone production, are generally coaxial annular tubes with a small gap thickness,
typically 1/8" to 3/8" The length can be up to several feet. The gas flows axially
through the annular gap. This geometry has several disadvantages. A large fraction of
the reactor frontal area is lost and not used (i.e., space between tubes, electrode space),
just from the geometry. Also, the corona forms only near the central electrode where the
electric field is the greatest. As a result, only a small area near the electrode surface is
filled with the corona that limits the effectiveness on the bulk gas. The combination of
these two effects results in a very small fraction (-15%) of the reactor volume being
effective for chemical reactions. Additionally each tube's gas is not mixed with that of
the other tubes. Therefore electrode failures result in straight bleed through of that tube's
unreacted gases.
Figure 7 illustrates the TecoLytic corona reactor. Positive and negative electrodes are
alternately spaced so that each electrode is surrounded by four opposite polarity
electrodes. This geometry results in a unique corona comprised of an intense corona
surrounding each electrode and a corona "sheet" between each positive and negative
electrode. The gas flow is perpendicular to the electrodes and the entire reactor cross
section is open to gas flow. The treated gas is subjected to turbulent gas mixing between
the electrodes and must flow through multiple corona regions. Hence, there is no
opportunity for the gas to bypass the chemically active regions even if multiple electrodes
would fail. Also, since each electrode is alternating in polarity during operation,
particulate buildup is avoided as would normally occur with DC systems. The chemical
reactions, all of which are likely to occur to an appreciable extent within the reactor, are
detailed in Table 1.
-------
Figure 7. TecoLytic™ Corona Reactor
Dielectric
Ground
Gas How
Table 1. Corona Reactor Reactions
Reactant(s)
Produces)
Number
N2 + e
NO + e
NO" + NO
NO"
N + O
N + N
NO + O
2NO2
O2 + e
O2 + O
2NO + O3
3NO2 + H2O
4HNO3
H20 + 03
N02 + OH
HNO3 + H2O0)
2N + e
NO"
NO2 + N + e
N + O + e
NO
N2
NO2
N2 + 2O2
2O + e
03
2NO2 + 02
HNO3 + NO
2H2O + 2N2 + 5O2
2OH + O2
HNO3
HNO3(«q)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Bench Scale Results
The concepts behind S02 scrubbing in the horizontal absorber and the ThioClear process
have been successfully demonstrated at the pilot scale. Current bench scale efforts have
focused on the TecoLytic reactor and the tiOx scrubber. The testing of the reactor was
-------
conducted at Tecogen's test laboratory in Waltham, Massachusetts. Scrubber evaluations
have been conducted at DLC Research Center's 5 kW scrubber facility in Pittsburgh,
Pennsylvania.
Tecogen has been involved in developing corona initiated reactions for pollution control
for the past seven years. During that time, work has focused on developing this
technology for NOX control. These NOX removal efforts utilizing the corona reactor have
focused on reactor design and optimizing energy utilization.
The design of the corona reactor is critical for many reasons. Decreasing the spacing
between the electrodes improves the gas mixing through turbulence and enables lower
power usage at the expense of a greater number of electrodes and higher pressure drop.
Table 2 details the influence of electrode geometry on power density and reactor pressure
drop. Utilizing this data with the design criteria of maintaining a pressure drop of ~2
inches of HzO and minimizing electrode wear by operating at low power densities has
resulted in the pilot scale reactor design criteria of a power density of <40 W/ft and a
pressure drop of 2.3 inches of HbO at a gas velocity of 25 ft/sec. The pilot plant reactor
will have 1600 electrodes which are 0.25 inches in diameter on staggered 0.75 inch
spacing which translates into a reactor depth of ~9.0 inches.
Table 2. Effect of Electrode Geometry on Power Density
Electrode
Diameter
(in.)
.25
.375
Electrode
Spacing
(in.)
.75
.75
.75
1.00
1.00
1.25
1.25
1.60
Corona
Initiation
Voltage O-P
(KV)
9
9
9
12
12
12
12
17
Operating
Voltage O-P
(KV)
17
20
30
30
40
30
40
40
Power
Density
(Watts/Ft)
30
65
200
130
400
40
250
125
Pressure
Drop
(in. H2O)
2.3
1.3
.5
.8
.3
2.1
.4
.8
In addition to evaluating the reactor design parameters, evaluation of various power
supplies were conducted. The results of this study are shown in Table 3. Increasing the
frequency of the power supplied to the reactor improves the stability of the corona.
However, the potential degradation of the electrodes operating at the higher power
densities associated with the higher frequencies has resulted in a frequency of 1000 Hz
being selected for the pilot plant operations. Utilization of a pulsed DC system was not
effective due to the lack of generating a corona.
-------
Table 3. Power Supply Evaluation Status
Supply Type Frequency Mai. Voltage Measured Corona
Zero-to Peak Power Density Characteristic
(Hz) (Volts) (Watts/Ft)
Variable A.C. 100-2,000 40,000
Fixed A.C.
Fixed A.C.
Pulsed D.C.
5,000
40,000
40,000
25,000
25,000
40,000
100 Increasing stability
and smoothness with
frequency.
150 Very smooth corona
300 Very smooth corona
No corona
NOX conversion versus specific power is shown in Figure 8. Specific power is defined as
the power input to the reactor (W) divided by the product of the flue gas flowrate (scfrn)
and the inlet NOX concentration (ppm). It is the same dimensional quantity as
eV/molecule but in a fashion more easily used for design purposes. Inherent in this
specific power definition is the NOX concentration's effect on conversion. That effect is
the power required to achieve a fixed NOX conversion level and is directly proportional to
the inlet NOX concentration. Simply, if the inlet NOX level is reduced by 50%, the power
required for a given conversion level is also halved. Because the kinetics are so fast,
except for very short residence times (ms), it has been found that input power per unit
mass flow is most critical.
Figure 8. TecoLytic Reactor Operating Data
60 -
50 •
40 -
20 -
10 -
I/
/
/
/
/
/
/
/ ^>
^
*2 that would generate
gypsum and calcium carbonate solid contaminants, respectively. The resulting calcium
-------
nitrate solution by-product generated from these studies compares favorably with the
specifications of fertilizer grade calcium nitrate solution shown in Table 4.
Table 4. High Grade Calcium Nitrate Fertilizer Specifications
Analysis Acceptable Range Bench Scale By-Product
Calcium (%)
Nitrate Nitrogen (%)
PH
Specific Gravity
Suspended Solids
9.8-11.4
9.3 - 12.3
4-6
1.39-1.46
<0.8
10.49
9.44
6.41
1.4
0.5
Summary
Multiple synergisms exist in combining the ThioClear and TecoLytic processes in a
horizontal absorber for the removal of SO2 and NOX. The high §62 removals of the
ThioClear process insures minimal contamination of the nitrate by-product generated in
the second stage and therefore provides greater flexibility in fuel selection. Complete
humidification of the flue gas is accomplished in the first stage that promotes the
formation of the hydroxyl radical required for nitric acid formation in the corona reactor.
Operation at high gas velocities enhances both SO2 removal in the first stage and gas
mixing for higher NO reductions in the reactor module. Additional advantages of this
process for emissions control include:
• Flue gas reheat not required.
• Avoids ammonia handling, storage, and associated fouling and flyash contamination
problems.
• Flexibility of NOX control.
• Zero discharge process with salable by-products.
• Capability to eliminate other air pollutants such as air toxics and fine particulates.
The concepts behind the ThioClear / TecoLytic process have been successfully
demonstrated on either the bench or pilot scales. The integration of these technologies
provides a promising advanced control process for SO2 and NOX emissions. Planned
pilot plant operation within the next year will be devoted to optimizing the integration of
the technologies and gathering pertinent design and cost data.
Acknowledgements
The authors of this paper would like to thank the Ohio Coal Development Office and The
Cincinnati Gas and Electric Company for their support.
-------
References
1. Breault, R_, and Lani, B.W., Combined High Velocity SO? and NO* Scrubbing
Process, presented at the 22nd International Ti
and Fuel Systems, Clearwater, Florida, 1997.
Process, presented at the 22nd International Technical Conference on Coal Utilization
2. Lani, B.W., College, J., and Babu, M, Results of ThioClear Testing: Magnesium-
Enhanced Lime FGD with High SO? Removals and Salable Bv-Products. presented at
the 1995 SO2 Control Symposium, Miami, Florida, March 1995.
3. Lani, B.W., Babu, M., Phase II: The Age of High Velocity Scrubbing, to be presented
at the EPRJ - DOE - EPA Combined Utility Air Pollutant Control Symposium,
Washington, D.C., August 1997.
-------
SIMULTANEOUS SOi AND NO2 REMOVAL BY ALKALINE SOLIDS
Christopher H. Nelli
Gary T. Rochelle
Department of Chemical Engineering
University of Texas at Austin
Austin, Texas 78712-1062
Abstract
At conditions typical of a bag filter exposed to a coal-fired flue gas which has been adiabatically
cooled with water, calcium hydroxide and calcium silicate solids were exposed to a dilute,
humidified gas stream of nitrogen dioxide (NC>2) and sulfur dioxide (SOi) in a packed-bed reactor.
A prior study found that NO2 reacted readily with surface water of alkaline and non-alkaline solids
to produce nitrate, nitrite, and nitric oxide (NO). With SOi present in the gas stream, NC>2 also
reacted with S(IV), a product of SOi removal, on the exterior of an alkaline solid. The oxidation
of S(IV) to S(VT) by oxygen reduced the availability of S(TV) and lowered removal of NC>2. In
addition, S(IV) reacted with nitrous acid to produce sulfur-nitrogen compounds on the surface.
Subsequent acidification of the sorbent by the removal of NC>2 and SC>2 facilitated the production
of NO. A reactor model based on empirical rate expressions predicted rates of SO2 removal, NO2
removal, and NO production by calcium silicate solids. Rate expressions from the reactor model
were inserted into a second program, which predicted the removal of SO2 and NOX by a
continuous process such as the collection of alkaline solids in a baghouse. The continuous process
model, depending upon inlet conditions, predicted 30-40% removal for NOX and 50-90% removal
for SO2- These results are relevant to dry scrubbing technology for combined SO2 and NOX
removal that first oxidizes NO to NO2 by the addition of methanol into the flue duct.
Introduction
The Acid Rain Program (Title IV) of the 1990 Clean Air Act Amendments requires the electric
power industry to significantly reduce sulfur dioxide (SO2) and nitrogen oxides (NOX) emissions
from fossil-fueled boilers. For older, existing power plants and/or those that operate on a seasonal
basis, dry scrubbing technology, which has a lower capital cost compared to conventional
technologies Like limestone slurry scrubbing or lime spray drying, may be the least-cost option for
SO2 control.
To meet the new NOX compliance requirements, the application of low-NOx burner technology to
existing boilers should be sufficient in most cases1. However, any additional removal of NOX
beyond the current level of reductions set forth in Title IV will likely require a post-combustion
process. A dry scrubbing process that simultaneously removes both SO2 and NOX will not only
satisfy the current SO2 requirements but also provide additional NOX removal at considerable
operating and cost efficiencies.
For simultaneous removal of SO2 and NOX, nitric oxide (NO) is first oxidized to nitrogen dioxide
the addition of methanol or other hydrocarbons into the flue duct at an optimum
-------
temperature2-3. NO2, along with SC>2, is then removed by contact with an injected alkaline
material. For systems with baghouses, most of the contact between flue gas and injected sorbent
will occur at the bag filter, where cycle times can range from a few minutes to hours. Since the
flue gas will be adiabatically cooled with water, the relative humidity at the bagfilter is expected to
be 10-60% at temperatures of 60-90 °C.
A significant amount of previous work4-5-6 has been conducted regarding SC>2 removal by
calcium-based materials at the conditions listed above at both bench and pilot scale. On the other
hand, prior to this study, little work had been found regarding NO2 removal by alkaline materials
at similar conditions. Because of the lack of prior research and the implications toward combined
SO2/NOX removal via NO oxidation, the focus of this research was to measure the NO2 and SO2
reactivities of calcium hydroxide and calcium silicate solids when exposed to a synthetic flue gas at
bagfilter conditions.
Chemistry
It can be assumed that a gas-liquid interface exists on the surface of the sorbent when it is in
contact with a gas stream of 50% relative humidity. Multilayer coverage of water on the surface,
due to adsorption and water-filled capillary pores, provides enough surface water to absorb gas
species such as NO2 and SO2. In a previous study by Nelli and Rochelle7, absorbed species were
found to react as they would in an aqueous solution. For example, NO2 was found to react on
inert surfaces with surface water in the following manner:
N02 (g) <-> N02 (1) (1)
2NO2 (1) + H2O (1) <--> HNO2 (1) + HNO3 (1) (2)
3HNO2 (1) <--> HNO3 (1) + 2NO (1) + H2O (1) (3)
NO (1) <--> NO (g) (4)
Reactions 1-4, added together, produced the following overall reaction stoichiometry that was
observed experimentally:
3N02 (g) + H20 (1) <--> 2HN03 (1) + NO (g) (5)
In addition, removal of NO2 by surface water was second order in NO2 concentration.
The kinetics and stoichiometry of the surface mechanism found by Nelli and Rochelle7 were
similar with the observations made by previous researchers studying NO2 removal by aqueous
solutions. Takeuchi et al.8 and Shen and Rochelle9, both studying the absorption of NO2 into
aqueous systems at low NO2 concentrations, found NO2 hydrolysis to be second order in NO2
concentration. Takeuchi et al.8, using a stirred-cell reactor as a gas-liquid contact device,
determined the forward pseudo-second order rate constant of reaction 2 to be 6.20* 104
mole/L/atm2/sec at 25°C (their reported rate constant was adjusted to match the reaction
stoichiometry shown in reaction 2). Nelli and Rochelle7, studying the NO2-H2O reaction on the
surface of sand particles, found the rate constant to be roughly four times faster at 48% relative
humidity than that determined by Takeuchi et al.8 However, at higher relative humidity, the rate
constant was observed to approach the value determined by Takeuchi et al.8 in bulk aqueous
solution.
Nelli and Rochelle7 also studied the reaction of NO2 with surface water on alkaline solids such as
hydrated lime and calcium silicate. Unlike inert surfaces, the alkaline materials produced little NO
-------
at initial reaction times. The available alkalinity neutralized a large portion of the nitrous acid
(HNO2) produced on the surface by the removal of NC>2. By keeping nitrous acid dissociated, NO
production was reduced by the inhibition of reaction 3.
The presence of SO2 in the gas phase provides an additional route for NO2 removal. The
hydrolysis of SO2 in an alkaline solution supplies the necessary S(IV) concentration to react with
NO2. Nash10 proposed that such a reaction in an aqueous solution was likely to involve electron
transfer leading to a chain mechanism of free radicals. Littlejohn et al.11, on the other hand,
proposed that a transient intermediate was necessary to form the reaction products of nitrite ion and
sulfite radical. In either case, the overall reaction of NC>2 in solution with sulfite proceeds in the
following manner1':
NO2 + SO3= -> NO2- + SO3- • (6)
803-- + SO3 > SO3= + SO3 (7)
SO3 + H2O --> SO4= + 2H+ (8)
The following overall reaction stoichiometry is obtained when reactions 6-8 are added:
2NO2 + SO3= + H2O -> 2NO2- + SO4= + 2H+ (9)
NC>2 reacts with bisulfite in aqueous solution like it reacts with sulfite. The overall stoichiometry is
as follows:
2NO2 + HS03- + H20 -> 2N02- + SO4= + 3H+ (10)
The NO2-S(TV) reactions are first order with respect to both NO2 and S(TV). Table 1 lists the rate
constants found by Takeuchi et al.8 and Shen and Rochelle9 using a stirred-cell apparatus as a gas-
liquid contacting device.
Table 1
NO2-S(IV) reaction rate constants in aqueous solution.
T(°C)
25
55
NO2-SO3=
6.6x1 0s L/mole/s
11.2xl05L/mole/s
NO2-HSO3-
l.SxlO* L/mole/s
2.8X104 L/mole/s
Researcher
Takeuchi et al.8
Shen and Rochelle9
It is known that O2, in the presence of NO2, autoxidizes S(IV) to S(VI)9-11. Free radicals
produced by NO2 reacting with sulfite and bisulfite enhance sulfite oxidation by the following
chain reaction:
NO2 + SO3=->NO2- + SO3-' (11)
SO3-- +O2->SO5-- (12)
SO5-« + SO3= -> SO4= + SO4-« (13)
SO4-- + SO3= -> SO4= + SO3-- (14)
SO3-- + S03-- --> S206= (15)
-------
Thus, for every mole of NCb absorbed, several moles of sulfite (or bisulfite) can be consumed if
oxygen is present. As a result, the presence of oxygen inhibits the NC>2-S(IV) reaction by
reducing S(IV) concentration. Takeuchi et al.12 observed that the absorption rate of NO2 into
sodium sulfite solutions was reduced by 40% when air, rather than nitrogen, was used as a
diluent.
Finally, there is an important side reaction between nitrous acid and bisulfite whose product,
hydroxylamine disulfonate (HADS), is a precursor to numerous sulfur-nitrogen compounds13'14:
HNO2 + 2HSO3---> HON(SO3)2= + H2O (16)
The production of HADS, relevant in aqueous systems at a pH range of 3-8, is important because
it reduces NO production by removing nitrous acid from solution. Otherwise, at low pH, nitrous
acid is converted to NO by reaction 3. The kinetics of this initial reaction, at low to moderate
acidities, are believed to follow the rate expression shown below13:
d[HDS] = 3.7xl012 (L2/mole2/sec) exp (-6100 / T(K)) [H+] [NO21 [HSO3-] (17)
Experimental Method
Sorbent reactivity was measured in a packed-bed reactor system (Figure 1). Solid reagents,
dispersed in sand to prevent channeling and agglomeration, were placed within a cylindrical,
Pyrex® reactor (3.5 cm in diameter and 19.5 cm in height) and supported by a coarse glass frit (2
mm in thickness). A water bath equipped with a PK) controller regulated the temperature of the
submerged reactor.
Water fed from a syringe pump to a helical Pyrex® tube within a furnace supplied humidity to an
inert feed stream comprised of N2 and air (a source of O2). The combination of this humidified
feed stream with commercially supplied gases of 1% NO2 in N2 and 0.5% SO2 in N2 provided a
synthetic flue gas for the reactor apparatus. Mass flow controllers regulated gas flows. A bypass
line around the reactor allowed the synthetic flue gas to stabilize before beginning an experiment.
To reduce gas concentrations within the ranges of the gas analyzers and to prevent condensation
downstream of the reactor, house air diluted the reactor outlet stream by a factor of 50-70. Air
flow was regulated by a gate valve and known approximately by a high flow air rotameter. A
Thermo Electron chemiluminescent NO/NO2/NOX analyzer with a molybdenum converter and a
Thermo Electron pulsed fluorescent SO2 analyzer sampled a portion of the diluted stream. The
balance of the stream was scrubbed with 13 wt% NaOH solution.
In a typical experiment, sorbent mixed with sand was placed inside the reactor. A nitrogen stream
containing a known relative humidity preconditioned the reactor contents for 18 minutes. Flue gas
bypassing the reactor was then synthesized with the same relative humidity as the preconditioning
stream, allowed to reach steady-state, and afterwards sent into the reactor. Gas phase material
balances around the reactor gave removal/production rates of NO2, NO, and SO2.
The primary reagent of interest was calcium silicate or AD VACATE (ADVAnced siliCATE).
Calcium sih'cate is comprised of varying amounts of calcium hydroxide reacted with fly ash in a
heated, aqueous slurry15. The reaction between the silica in the fly ash and the calcium hydroxide
produces a calcium silicate material with a high surface area and porosity. The material for this
study was prepared by Johnson16 by slurrying 1 part hydrated lime and 3 parts Shawnee fly ash at
90 °C for 12 hours. Solids were made ready by filtering the slurry, drying the filter cake overnight
-------
at 100 °C, and sieving the solids through 80 mesh sieve. Additional details of the synthesis of this
material can be found in Johnson's thesis16.
Hydrated lime, produced by the Mississippi Lime Company, was the secondary reagent of interest.
This material has been used by previous researchers for AD VACATE production15'17. Unlike
AD VACATE, hydrated lime has little internal surface area and porosity.
As mentioned previously, the purpose of the sand was to prevent channeling and agglomeration of
the solid reagents within the reactor. This dispersant, noted in this study simply as sand, was
purchased from Aldrich Chemical Company and labeled as white quartz with a size of 50-70 mesh.
Table 2 lists the respective BET surface areas of the reagents including sand. Experiments
typically lasted 20 to 60 minutes.
Table 2
BET surface areas of solid reagents.
Material Surface Area (m2/g)
Sand 0.57
Hydrated Lime 8.76
Fly Ash AD VAC ATE 49.9
Results
Hydrated Lime
Figure 2 shows the rate of NC"2 removal obtained by a mixture of sand and hydrated lime in the
reactor with a humidified gas stream of NC>2 and N2. hi some experiments, SC>2 and 02 were
added to the gas stream to observe their effect on NC>2 removal. The experiment without SC>2 or
O2 represented the base case experiment in Figure 2.
Without SC>2, a mixture of sand and hydrated lime produced a measurable rate of NO2 removal by
the NO2-H2O reaction on the surface of the hydrated lime surface. Removal declined only slightly
throughout the course of the experiment because the hydrated lime neutralized the nitrous and nitric
acids accumulating on the surface. The addition of 5% ©2 to the gas stream did not alter the
removal rate appreciably.
SQz was added to the synthesized feed stream to test the existence of the NO2-S(TV) reaction. As
shown in Figure 2, the addition of SC>2 without C>2 greatly enhanced the ability of hydrated lime to
remove NO2- Hydrated lime with S(IV) on its surface from the removal of SC>2, either in the form
of SOs= or HSOs' or a combination thereof, pushed the NC-2 rate of removal close to the maximum
rate possible, or the inlet feed rate of NC>2.
To test for S(IV) oxidation, C"2 was added to the feed stream along with SC>2. The addition of 62
was expected to oxidize S(IV) on the surface to sulfate or S(VI). The loss of S(IV) should have
had an adverse effect on the NC>2 rate of removal. Indeed, as Figure 2 shows, the presence of C>2
lowered the rate of NO2 removal, the degree to which related to the oxygen content in the feed
stream.
In the experiments with SC>2 present, NC>2 removal increased slightly with time. It is hypothesized
that since SC>2 removal was high, a steep concentration gradient of 862 in the bed was created,
-------
which resulted in most of the S(TV) at early times being deposited at the front end. As additional
S(TV) was deposited over time towards the back end of the bed, the rate of NC>2 removal increased.
Figure 2 shows a sharp decrease in the NO2 removal rate at approximately 20 minutes for the
experiments with SC>2 present. The decline in NOa removal rate is hypothesized to be the result of
product layer formation on the solids that prevented hydroxide or other alkaline agents from
reaching the surface to neutralize the acidic reaction products. The acidification of the outer surface
of hydrated rime was expected to shut down the NO2-H2O and NO2-S(IV) reactions either through
the build-up of nitric and nitrous acid or the loss of sulfite and bisulfite by protonation to sulfurous
acid.
Fly Ash ADVACATE
Typical rates of SC>2 removal, NC>2 removal, and NO production by fly ash ADVACATE exposed
to synthetic flue gas in the sandbed reactor are shown in Figure 3. Rates of removal and
production are plotted against sorbent conversion where conversion is defined as the cumulative
moles of SC>2 and NOX removed normalized by the divalent alkalinity of the sorbent material.
Alkalinity of fly ash ADVACATE was determined by an acid-base titration of unreacted sorbent
dissolved in acid. Conversion assumed a stoichiometry of 1.0 mole of divalent alkalinity per mole
SC>2 removed and 0.5 mole divalent alkalinity per mole NOX removed.
Figure 3 shows a decreasing removal rate of SC»2 and NC>2 as a function of sorbent conversion. It
is hypothesized by the authors that the formation of a product layer on the solids likely resulted in
an acidification of its outer surface, thus inhibiting the NOj-H^O and NO2-S(TV) reactions and
facilitating NO production. Nonetheless, the SO2 and NO2 rate of removal was sufficient to
achieve a conversion close to 100%. Unlike hydrated Lime which obtained conversion only in the
range of 20-30%, fly ash ADVACATE in most cases achieved conversion of approximately 100%.
In addition to the removal of SO2 and NO2, a significant amount of NO was produced. For the
experiment shown in Figure 3, only 61% of the total NO2 that was removed from the gas phase
was retained on the surface. The remaining 39% converted to NO and returned to the gas stream.
The decrease in NO production at high conversion was the result of a decrease in NO2 removal.
Though not shown in this paper, chemistry similar to that seen with hydrated lime, such as the
negative effect of 02 on NOa removal, was observed to occur with fly ash ADVACATE18. The
main difference between the two sorbents was the much higher conversion achieved by fly ash
ADVACATE over hydrated lime. The higher surface area of fly ash ADVACATE over hydrated
lime is believed responsible for creating a thinner product layer on the alkaline surface and thus
increasing the availability of alkalinity underneath it.
Figure 4 shows the results of ion chromatography (1C) analysis of spent solids from the reactor.
All of the contents from the reactor were dissolved in 100 ml of 10-2 M HC1 solution and agitated
for 15 minutes. A portion of the solution was sampled, diluted by a factor of 10, and injected into
the 1C. Standard solutions of nitrite, nitrate, sulfate, hydroxylamine disulfonate (HADS), and
amine disulfonate (ADS) were prepared for calibration purposes. Hydrogen peroxide (^C>2) was
added to some of the sample solution to oxidize any S(TV) to S(VT). S(VT) species, such as
sulfate, were easier to detect than S(IV) species.
The bar graph shows the results of three experiments where exposure of fly ash ADVACATE with
synthesized flue gas varied from 7 to 30 minutes. After 30 minutes of reaction, the end products
of nitrogen were nitrate (a product of the NO2-water reaction), HADS (the first stable product of
the nitrous acid-S(IV) reaction), and ADS (a product derived from HADS). Note the increasing
proportion of ADS to HADS as a function of exposure time. Nitrite or nitrous acid was not
detected hi any experiment, indicating the ability of the solids to produce NO from nitrous acid. It
is apparent that if nitrous acid is not converted to HADS and ADS in the course of its lifetime on
the sorbent surface, then the only alternative pathway is NO production.
-------
Analysis of spent AD VACATE solids showed good closure of the gas phase material balances
used to determine removal/production of SC>2, NC>2, and NO. From all the experiments studied,
(96 +/- 10)% of the nitrogen and (98 +/- 17)% of the sulfur were recovered in the solid phase. A
confidence level of 95% was used in the statistical calculations.
Mathematical Model of Sandbed Reactor
Figure 3 showing NO2 removal, NO production, and SO2 removal as a function of sorbent
conversion gives little indication of the removal that fly ash AD VACATE would provide on a
continuous basis in an industrial process such as a baghouse. This section attempts to model the
experimental results from the packed-bed reactor by developing empirical and semi-empirical rate
expressions for NO2 and SO2 removal and NO production for a reactor loaded with fly ash
AD VACATE. Rate expressions from the reactor model will then be inserted into a second
modeling program designed to predict SO2 and NOX removal by a continuous process. The
purpose of the model is not to validate the hypotheses put forward by the authors, but instead to
merely reproduce the data for predictive purposes.
The model attempted to predict the rates shown in Figures 5-7. The model used five rate
expressions to account for SO2 removal, NO2 removal, NO production, and sulfur-nitrogen
production. These rate expressions were based on the five reactions shown below.
S02 + H20 <-> HSO3- + H+ (18)
2NO2 + H2O <-> NO2- + NO3- + 2H+ (19)
2N02 + HS03- + H2O -> 2N02- + SO4= + 3H+ (20)
HNO2 + 2HS03- -> HON(S03)2= + H2O (21)
HNO2 <-> 2/3NO + V3HN03 + 1/3H20 (22)
Rates, in conjunction with the known stoichiometry of the reactions, were used to calculate
accumulations of total nitrate, nitrite, S(FV), S(VT), and sulfur-nitrogen. These values of acid
accumulation were then used to calculate local sorbent conversion. The forms of the equations,
along with the numerical values of the adjustable parameters, are listed below:
(23)
(24)
(26)
(27)
™.ve,s.o.= *„,.„. .S^r"'1*'5'^ <28>
-------
ab , „ ju i / 1.0 - conversion ^ ,OQ,
—r- where a = ki PSO2 and b = k2 ( rnnve,,lon— ) (29)
conversion
ki = 893 mole/Vi/sec/atm
k2 = 0.15mole/Vi/sec
= k3 PN022 - k4 [N03-]T [N02-]T
k3 = 4.05xl05 mole/Vi/sec/atm2
k4= 0.0010 Vi/mole/sec
conversion
/ , converson \
r3 = ks PN02 n exp (- k6 pi/ )
k5 = 2000.0 1/atm
ke = 0.0663 atm1^
r4 = ky [N02-]Total (32)
k7 = 0.001 I/sec
r5 = k8 [N02-]T Pso2 (33)
kg = 8.88 1/sec/atm
Rates concerning SO2 removal, NO2 removal, and NO production were linked to removal from (or
production to) the gas phase by the following gas phase mass balances:
-'I^
-2(r2 + r3)[^ (35)
2/3r5^] (36)
where v is the gas velocity,
x is the length of the sandbed reactor,
Q is the gas concentration of SO2, NO2, or NO,
r; is the removal or production rate of i (mole/Vi/sec),
Vi is the volume of water on the surface of the sorbent,
and Vr is the volume of the reactor.
The amount of water on the surface of the sand was calculated by assuming a water molecule has
six nearest neighbors, the diameter is equal to its collision diameter, and the surface area and
number of monolayers is approximated by BET theory. The overall rates of SO2 removal, NO2
-------
removal, and NO production were calculated by multiplying the differences in inlet and outlet
concentrations of SO2, NC>2, and NO by the gas flow rate, G. Rates were calculated at 200 steps
along the length of the reactor and after every time step (1 sec). The results of the model are
shown in Figures 5-7.
Mathematical Prediction of a Continuous Process
With rate expressions quantified from the reactor model, a second modeling program was used to
predict the removal of NOX and SO2 by a continuous process such as the collection of fly ash
ADVAGATE in a baghouse. Though the same rate expressions were used in both programs, the
continuous process model simulated a vastly different physical system regarding contact time
between the sorbent and the gas stream. Thus, the difference between the two models is physical
and not chemical.
In a baghouse, solids are collected continuously on bagfilters and periodically cleaned by a
reversed air stream or a pulsed jet. The time it takes to clean all the bagfilters, regardless if only a
portion of the bagfilters are cleaned or all are done together, is called the cycle time. The
continuous process model, therefore, has to account for an intermittent, growing bed of collected
solids that has the freshest sorbent on the exterior of the bed and the most spent sorbent at the
interior of the bed. In this arrangement, the freshest sorbent receives the highest gas concentration
of SO2 and NO2 and the most spent sorbent receives the lowest gas concentration of SO2 and
NO2. This arrangement is exactly opposite of the case for the fixed-bed reactor.
A key variable in a continuous dry scrubbing process is the stoichiometric ratio, the ratio of the
molar feed rate of alkalinity (where alkalinity is a function of the type of sorbent material used) and
the feed rate of acid gas. A stoichiometry of 1 mole divalent alkalinity per mole SO2 and 0.5 mole
divalent alkalinity per mole NOX was used to calculate the stoichiometric ratio. The second key
variable in a continuous process, in this case a baghouse, is cycle time. Contact between flue gas
and injected solids prior to the baghouse was assumed to be negligible.
Figure 8 shows the time average removal of NO2 and NOX as a function of stoichiometric feed
ratio and baghouse cycle time. Tune average removal is simply the integral of the instantaneous
removal and represents actual removal obtained by a baghouse for a given cycle time. The
decrease in the removal of NOX as a function of cycle time is the result of NO production.
Otherwise, NOX removal would taper off asymptotically like NO2 removal.
As shown in Figure 8, low removal (30-40%) of NOX was obtained at cycle times greater than 10
minutes. Modest improvement in NOX removal was achieved when the stoichiometric feed ratio
was increased from 1 to 1.5. A stoichiometric ratio of 1.5 is probably the maximum ratio feasible
since larger ratios would add a significant operating cost that would weaken the economic
advantage of a low capital cost process such as the AD VACATE process. Finally, the removal of
NO2 is presented to show the potential NOX removal if NO production was reduced to zero. Even
at zero NO production, only approximately 60% removal of NOX would be obtained. These
results show that removal of NOX is limited by both NO production and the rate of NO2 removal,
particularly at long times or high solids conversion.
Figure 9 shows SO2 removal by fly ash AD VACATE as predicted by the continuous process
model for stoichiometric feed ratios of 1 and 1.5 and inlet NO2 gas concentrations of 0 and 200
ppm. Depending upon process conditions, moderate to high removal (50-90%) of SO2 was
obtained. SO2 removal was significantly enhanced by the addition of NO2-
Figures 8-9 show predictions of removal based on a flue gas with no O2. To account for a flue gas
with O2 (in this case, 5%), multiply the NO2 and NOX removals shown in Figure 9 by 0.85 (©2
has no effect on SOi removal). This correction was based on the experimental observation that a
flue gas stream of 225 ppm NO2/1000 ppm SO2/5% O2 was equivalent to a flue gas stream of 225
-------
ppm NO2/500 ppm SO2/0% O2 in terms of NO2 and NOX removal by fly ash AD VACATE. By
substituting 500 ppm SO2 for 1000 ppm SO2 in the continuous process model and noting the
decrease in NO2 and NOX removal, the addition of 5% O2 decreased the removal of both NO2 and
NOX by approximately 15%.
Conclusions
NO2 reacted readily with surface water of alkaline solids like calcium hydroxide and calcium
silicate. The addition of SO2 to the gas phase enhanced the rate of NO2 removal, whereas the
addition of O2 lowered the rate. Sulfur-nitrogen compounds were found to be produced on the
surface. Acidification of the sorbent by the removal of NO2 and SO2 facilitated the production of
NO. Hydrated Lime typically obtained a conversion in the range of 20-30%, whereas fly ash
AD VACATE in most cases achieved conversion of approximately 100%. Rate expressions from
the reactor model were used to predict the continuous removal of SO2 and NOX in a process like
the collection of solids in a baghouse. The continuous process model, depending upon inlet
conditions, predicted 30-40% removal for NOX and 50-90% removal for SO2.
Acknowledgment
This work was supported by the Texas Advanced Technology Program (grant no. ATP 3658064).
References
1. S.C. Wood. Chemical Engineering Progress. Jan., p. 32 (1994).
2. M. Hori, N. Matsunaga, P.C. Malte, and N.M. Marinov, Twenty-Fourth Symposium
(International) on Combustion/The Combustion Institute, Sydney, Australia, p. 909 (1992).
3. V.M. Zamansky, L. Ho, R.M. Maly, and W.R. Seeker, Third International Conference on
Combustion Technologies for a Clean Environment, Lisbon, Portugal, Vol. 2 (1995).
4. W. Jozewicz and G.T. Rochelle. Environmental Progress. Vol. 5, No. 2, p. 219 (1986).
5. P. Chu and G.T. Rochelle. JAPCA. Vol. 39, p. 175 (1989).
6. W. Jozewicz, C. Jorgensen, J.C.S. Chang, C.B. Sedman, and T.G. Brna. JAPCA. Vol. 38,
p. 796 (1988).
7. C.H. Nelli and G.T. Rochelle. Ind. Eng. Chem. Res. Vol. 35, p. 999 (1996).
8. H. Takeuchi, M. Ando, and N. JCizawa. Ind. Eng. Chem., Process Des. Dev. Vol. 16, p.
303 (1977).
9. C.H. Shen and G.T. Rochelle, 7995 SO2 Control Symposium, Miami, FL, Vol. 3 (1995).
10. T. Nash. Atm. Emir. Vol. 13, p. 1149 (1979).
11. D. Littlejohn, Y. Wang, and S. Chang. Environ. Sci. Tech. Vol. 27, No. 10, p. 2162 (1993).
12. H. Takeuchi, K. Takahashi, and N. Kizawa. Ind. Eng. Chem., Process Des. Dev. Vol. 16,
p. 486 (1977).
13. S.G. Chang, D. Littlejohn, and N.H. Lin. ACS Symp. Ser. Vol. 188, p. 127 (1982).
14. J.B. Jarvis, P.A. Nassos, and D.A. Stewart, 1985 EPA/EPRI Symposium on Flue Gas
Desulfurization, Cincinnati, OH, Vol. 1, p. 261 (1985).
15. K.K. ICind and G.T. Rochelle. /. Air & Waste Manage. Assoc. Vol. 44, p. 869 (1994).
16. H.L. Johnson. M.S. Thesis. The University of Texas at Austin (1992).
17. L.F. Arthur and G.T. Rochelle, 7995 SO2 Control Symposium, Miami, FL,Vol. 2 (1995).
18. C.H. Nelli. Ph.D. Dissertation. The University of Texas at Austin (1997).
-------
PUMP
H2O
Air, N2
REACTOR IN
WATER BATH
FURNACE
SO2/N2, NO2/N2
Air
Calibration Gases
HOOD .
1.5 LPM
SO2 ANALYZER
0-50 PPM
NOX ANALYZER
0-10 PPM
SCRUBBER
Figure 1
Schematic of the sandbed reactor system.
-------
CD
IT
Figure 2
The effect of SC>2 and C>2 on NCb removal by hydrated lime. Experimental results only are shown.
A reactor loading of 20 g sand and 0.400 g hydrated lime was used for all experiments in addition
to the following inlet gas conditions: 70 °C, 60% relative humidity, and 225 ppm NOa. Open
points corresponded to experiments without SC>2 and closed points corresponded to experiments
with SC>2 (910 ppm inlet concentration). Rates are compared with the inlet feed rate of NC>2.
0.0 0.2 0.4 0.6 0.8 1.0
Conversion (mole rem./divalent alk.)
Figure 3
Hue gas cleaning by fly ash AD VACATE. Experimental results only are shown. Inlet feed rates
of SO2 and NO2 were l.OlxlQ-6 and 2.31xlQ-7 mole/sec, respectively.
-------
E
70°C
60% RH
225 ppm N02
1000 ppm SO»
- 11.4 m2 Sand
: 12.5 m2 ADV
7 15 30
Time of Exposure in Reactor (min)
Figure 4
Speciation of nitrogen reaction products on fly ash AD VACATE. Total divalent alkalinity of fly
ash AD VACATE loaded into reactor was 0.675 rnmole.
9.0 10-
I
«S 3.0 1 0 =,
O
CO
70-C
60% RH
20g Sand
O.ZSOg ADV
225 ppm NO,
0.0 0.2 0.4 0.6 0.8 1.0
Conversion (mole rem./divalent alk.)
FigureS
SO2 rate of removal by fly ash AD VACATE. Curves predicted by equations 23-36. Conversion
assumed a stoichiometry of 1 mole divalent alkalinity per mole SC>2 removed and 0.5 mole divalent
alkalinity per mole NOX removed.
-------
° 3.0 1 0"
E 2.0 1 0' 7
O
z
70°C
60% RH
20g Sand
- 0.250g ADV
225 ppm NO2
970 ppm SO,
533 ppm SO2
274 ppm SO,
-Model
0.0 0.2 0.4 0.6 0.8 1.0
Conversion (mole rem./divalent alk.)
Figure 6
NC>2 rate of removal by fly ash AD VACATE. Curves predicted by equations 23-36. Conversion
assumed a stoichiometry of 1 mole divalent alkalinity per mole SC>2 removed and 0.5 mole divalent
alkalinity per mole NOX removed.
_ 1-210
2 removed and 0.5 mole divalent
alkalinity per mole NOX removed.
-------
20 30 40
Cycle Time (min)
Figure 8
Predicted removal of NC>2 and NOX by fly ash AD VACATE as a function of stoichiometric feed
ratio and baghouse cycle time. Base conditions were 1000 ppm SOi, 200 ppm NC>2, and 0% C"2
for a gas flow of 70 °C, 60% relative humidity, and a 1:1 stoichiometric feed of alkalinity to acid
gas. To account for a flue gas with C>2 (in this case, 5%), multiply the NC>2 and NOX removals
shown above by 0.85.
:_ so -
o
2 by fly ash AD VACATE as a function of stoichiometric feed ratio, NC>2
gas concentration, and baghouse cycle tune. Base conditions were 1000 ppm SC>2, 200 ppm NC>2,
and 0% O2 for a gas flow of 70 °C, 60% relative humidity, and a 1:1 stoichiometric feed of
alkalinity to acid gas. Oxygen has no effect on SC>2 removal.
-------
Wednesday, August 27; 8:30 a.m.
Parallel Session C:
Continuous Emission Monitors
-------
THE ON-LINE REAL-TIME MONITORING OF AMMONIA, NOx and
SO2 IN FLUE GAS USING A UV-PDA ANALYZER
Richard D. Driver, Israel M. Stein, and Gariy S. Zaslavskiy
NovaChem BV
One Gateway Center, Suite 415
Newton, Massachusetts 02158
Abstract
An on-line real-time photodiode-array (PDA) analyzer, optimized for the UV, has been developed
to monitor the injection of ammonia in Selective-Catalytic-Reduction (SCR)
Selective-Non-Catalytic-Reduction (SNCR) DeNOx installations. The analyzer, coupled to the stack
via a high temperature sampling system, uses a sensitive chemometric algorithm to measure ammonia
slip in the flue gas in the 0-50 ppm range in the presence of varying levels of SO2 and NOx The unit
was tested on a 120 Megawatt cyclone coal fired burning generator. The measurement of ammonia
obtained on the NovaChem analyzer during operation of an aqueous ammonia SNCR showed
comparable results to the wet chemistry laboratory method. Studies were obtained at the end of a
four week trial and the analyzer was in continuous operation for over four months.
Ammonia Monitoring and DeNOx Pollution Prevention Systems
The reduction efficiencies realized in DeNOx installations vary widely depending on the nature and
operation of an installation and limited data is available the optimization of such systems. Expensive
DeNOx systems are being installed, commissioned and operated in many coal fired power stations'
but limited analytical data during installation and operation may lead to disappointing levels of NOx
reductions. Much research work goes on in this area2"5 but surprisingly little on-line data is available.
There are many variables, such as injector positioning, combustion dynamics, burner operation and
the choice of measurement point, which must be considered in the optimization of a DeNOx system
and any of these can affect the actual efficiency of a final installation. What is required in the
installation, testing, evaluation and day to day operation of DeNOx systems is the availability of high
quality real-time CEM data of the ammonia slip, together with correlated SO2 and NOx species
concentration data.
It is, therefore, advantageous to control the injection of ammonia or urea in Selective Catalytic
Reduction System ("SCR") or Selective Non Catalytic Reduction System ("SNCR") DeNOx
installations in order to optimize NOx reduction, to reduce ammonia slip and to provide combustion
air preheater protection. Using ammonia the chemical nature of the DeNOx process for complete NO
reduction is described by the following reaction:
4NH3 + 4NO + O2- 4N2 + 6H20
-------
To control the injection rate in coal fired power stations, ammonia slip (0-50 ppm) is measured,
simrltaneously with NOx levels (400 -1200 ppm) in the flue gas. In regard to potential combustion
air preheater fouling, it is also desirable to measure S02 (700 -1500 ppm) in the flue gas. Ammonia
is expensive and a toxic material giving both economic and environmental reasons for optimizing
DeNOx systems based on its use.
Beta Testing the NovaChem IPM-Mark II 350DN Analyzer on Unit #1
A study of the NovaChem IPM 350DN Process Diode-Array Flue Gas Analyzer for DeNOx
applications was conducted at the Public Service of New Hampshire from in the summer and fall of
1996 . The NovaChem flue gas monitor, which is designed to simultaneously measure sulfur dioxide
(0 - 2000 ppm), oxides of nitrogen (0 - 2000 ppm), and ammonia (0 - 50 ppm), was beta-site tested
on Unit #1, a 120 Megawatt, coal burning generator (Figure 1). Merrimack Station Unit #1 is a
Babcock and Wilcox cyclone fired unit with an aqueous ammonia SNCR installed for control of NOx
emission Unit #1 is the smaller and older of the two power units at the Merrimack station.
Figure 1 The IPM-Mark H 350DN Installed on Unit #1
The IPM-350DN analyzer was installed on a steel platform on the Unit #1 stack, at a point where flue
gases have a temperature of approximately 450°C.
The Detection of Flue Gases by Ultraviolet Spectroscopy
The UV offers certain major advantages in the detection of flue gas components at the ppm level
when compared to other analytical techniques. The species of interest, namely NH3, S02, NO and
N02 all have distinct spectral signatures in the 180-320 nm spectral region (Figure 2)6"10. The spectra
were obtained on the NovaChem IPM-Mark II 350DN UV flue gas analyzer The ammonia (NH3)
gas phase spectrum is restricted to the 180-220 nm spectral region consisting of 10 equally spaced
vibrational lines on 4 nm intervals with the absorbance maximum at approximately 191 nm The S02
-------
gas phase spectrum spans the spectral region from 180-350 nm. The spectrum consists of two broad
electronic features one centered at 195 nm of width 35 nm and a second centered at 291 nm with a
width of approximately 40 nm. The 195 nm feature is a factor of 10 stronger than the higher
180 200
Absorbance / Nanometers
File#1 :DSKELE1A
SINGLE file Split from
MULTIFILE
280 300
Overlay Y-Zoom
CURSOR Res=None
Figure 2: Calibration Spectra; NH3 (58 ppm), SO2 (981 ppm) and NOx (767 ppm)
wavelength feature and exhibits undulating vibrational peaks, approximately 25 in all on 1.5 nm
centers. The major NOx component of flue gas spectra is NO with a much weaker contribution from
NO. The NO2 gas phase spectrum consists of a slowly varying absorbance rising to a maximum at
180 nm and slowly decreasing out to 260 nm. The NO gas phase spectrum consists of 8 distinct and
unequally spaced features between 180 and 240 nm, with the two strongest features centered at 185
nm with a 3 nm spacing.
With the use of a photo-diode array spectrophotometer and application specific software, the
technique offers almost instantaneous detection capability. The parallel spectral acquisition capability
of the photo-diode array spectrophotometer allows concentration information to be produced in a
matter of seconds with high measurement accuracy. The UV is completely transparent to the
presence of water vapor (H2O) which can make the task of calibration and measurement difficult in
other spectral regions. Of even greater importance, the details or spectral signature of a UV spectrum
is very little affected by changes in sample temperature which makes the task of calibration and
sample prediction straightforward. Finally, using UV spectroscopic techniques, the flue gas may be
measured with a high temperature flow cell, without the need for sample preparation or
manipulation. It is only necessary that the sample be maintained above a level of 300° C to prevent
destruction of the ammonia.
The measured flue gas spectrum is a superposition of the individual spectral signatures of the three
spectra components, weighted by the relative densities of the species. It is most convenient to carry
out all calculations in absorbance space where the relationships are linear with respect to species
-------
concentration. Absorbance is linearly related to the measured optical transmission, defined as the
ratio of the transmitted radiation to that of the incident radiation of an absorbing sample, by the
following equation:
- 10--4
'o
The net absorbance A (to base 10), which is in general wavelength dependant, may be related to the
individual concentration ct of each species k through the molar extinction coefficient ek by the sum:
where L is the cell pathlength. This equation represents the absorbance relationship at a single
measurement wavelength and the complete spectrum represents a series of linear matrix equations
for the different measured wavelengths.
Present Wet Chemistry Method of Ammonia Slip Measurement
The conventional method of ammonia slip measurement uses wet chemistry according to the NIOSH
6701 procedure based on ion chromatography. Moisture content of the sample, which can interfere
with the 6701 Method, was determined by EPA Test Method 4. The 6701 Method is laborious and
requires at least 30-60 minutes of sampling at the low ppm levels (0-10 ppm) which are expected to
be encountered in real DeNOx installations. The method also gives delayed reporting of 3-4 days
between the data collection and the resulting reporting and the numbers and comparison with system
parameters has a strong presumption of steady state. Using the NIOSH 6701 procedure it is
extremely difficult to carry out the type of measurements over a long time period, required to verify
and control the DeNOx system.
The IPM-Mark II 350DN Analyzer
The NovaChem IPM-350DN Flue gas analyzer consists of an IPM-Mark II photodiode array
spectrograph bench and fiber-optic coupled transmission cell and flue gas sampling system contained
in a large NEMA 4 purged enclosure (Figure 1). The system also incorporates a flue gas flanged
sampling probe which is placed at a suitable point on the power station stack to extract the sample
gases. The purged analyzer enclosure is placed directly next to the probe flange and a heated trace
line connects the probe to the analyzer.
The optical path, shown in Figure 3, consists of the light source, UV transmitting fiber optic cables,
sample cell and a photodiode array spectrograph assembly combining a holographic grating
spectrograph with a UV optimized photodiode array. The light source assembly, spectrograph bench,
together with power supply modules, an electronic module, including a powerful computer for data
-------
acquisition and analysis, and an output display module, are contained in a single, smaller NEMA4
enclosure. Each component module is designed to be easily installed or removed from the system,
without disturbing the operation of the other modules. A software module, including a sophisticated
chemometric prediction engine, allows the system to operate as a full multi-component flue gas
analyzer.
The light source is a super quiet Xenon flash tube capable of a lifetime of greater than 10,000 hours.
The UV light is directly transmitted to the sample cell. The sample cell is a transmission type cell
Diode Arcjy
Figure 3: Schematic Diagram of IPM-Mark n Optical System
where a collimated light beam interacts with the flowing gases. After passing through the sample,
the light is collected by a return fiber, which transmits the light to the entrance slit of the
spectrograph. The fully solid-state design of the analyzer, including the fully solid-state optical bench
contains no moving parts and requires the minimum of system maintenance. The optical system is
configured with high mechanical tolerance pre-aligned optical modules for ease of manufacture and
routine system maintenance. The optical fibers allow the added advantage that the heated
transmission cell may be easily thermally isolated from the analyzer assembly.
The spectrograph separates the light beam into its component wavelengths and reflects the light onto
a linear array composed of 512 UV enhanced photodiodes; effectively 512 separate detectors, each
targeting a different portion of the spectrum, which monitor the sample simultaneously. In contrast
to conventional filter based instruments which measure only one or two discrete wavelengths, the
diode array spectrophotometer obtains a complete UV spectrum of the sample in under 1 second.
The analog signal generated by each diode is a function of the amount of light at that specific
wavelength. This analog signal is digitized and then mathematically manipulated to produce the
measured transmission spectrum and the resulting component concentration values of the flue gas
sample. The individual detector photodiodes, placed on 0.25 nm diode centers, span the spectral
-------
region 180-308 run. The spectrograph assembly has an effective cubic spline interpolated spectral
half width of 0.6 nm (full-width-at-half-maximum). The spectral region covered by the
spectrophotometer and the resulting spectral resolution permits the simultaneous measurement of
NH3, SO2, NOx in real time.
The software module consists of data acquisition algorithms, spectral preprocessing algorithms and
an advanced multivariate chemometric prediction engine capable of turning the raw measured data
values of the light intensity into a meaningful and reproducible UV spectra of the flue gas. The
analyzer may be used to collect raw calibration spectra, or with the addition of a suitable PLS (Partial
Least Square) or PCR (Principal Component Regression) calibration files, may be used on-line in
real-time and be turned into a full flue gas analyzer. Factory calibrations may be carried out on the
analyzer with suitable gas or liquid sample mixtures. The calibration spectra of known concentrations
of mixtures are manipulated to give the required calibration prediction matrix required by the
prediction engine. NovaChem Chemometric Prediction Engine (NCPE), consist of the PLS/PCR
calibration file and real-time measurements of transmission spectra. During process analysis a
proprietary using the chemometric calibration files, operates on the spectra of unknown samples in
the flow-cell and generates chemical concentration values which may be outputted to the analyzer
screen, 4-20 mA outputs and an RS232 or RS485 interface.
The NovaChem IPM 350 DeNOx analyzer, allows a measurement as frequently as every three
seconds, allows immediate transmission to the control room for process optimization, and in
comparison to conventional wet chemistries permits a quantum leap in data measurement.
The Sampling System and Flue Gas Probe
The sampling system is fabricated in accordance with an number of important design criteria. First,
the sampling system must be capable of removing all particulate matter from the flue gas while
maintaining sample integrity. Additionally the sample had to be sampled at a constant and
controllable volumetric flow rate to the flow cell, while maintaining all sample parameters during
transport to and during measurement in the flow cell. Finally, the sampling system has to have the
flexibility of allowing the introduction of various gas compositions into the flow cell, under controlled
pressure and temperature conditions for calibration purposes.
In designing the sampling system, two critical problems were addressed. One issue is the formation
of ammonium bisulfate at temperatures below 280°C and the second was the catalytic reaction of
ammonia on metallic surfaces. Sample integrity may be maintained only if the same amount of
ammonia that enters the probe reaches the flow cell. Great care was taken with the design and
construction of the sample heated lines and sampling system in order to keep the temperature of the
sample line and the sample cell above the 280°C at which ammonium bisulfate would be formed.
PLC controllers control the heating of the probe, delivery line and sample cell, in order to keep
sample temperature at 330°C. A careful choice of inert construction materials for the probe and
sampling system ensure the prevention of catalytic reactions, which again will decrease the ammonia
levels in the analyzer.
-------
The flue gas probe, a seven foot probe with an internal ceramic filter, is mounted on a suitable flange
and inserted into the duct (Figure 4). The probe with three sampling points at 5, 6 and 7 feet into the
flue provided a sample of the stack composition. The flange is connected to the external sampling
side of the probe. The external section of the probe assembly is heated to 330°C, while the sample
in the flue typically ranges from 350°C to 450°C. The probe filters which may coat with flyash are
automatically cleaned by a "blow back" process every thirty minutes
The sample is aspirated through the flow cell by means of the air driven eductor (aspirator) at a rate
of approximately 3 liters per minute. The sampling system is designed to maintain the sample at
constant pressure and at temperature of330°C After measurement, the sample is discharged to air
A low temperature valve is used to facilitate the introduction of calibration gases The sample line
between probe and flow cell is held at 330°C to prevent ammonia from reacting with other
Figure 4: The Flue Gas Probe
compounds. High temperature UV transmitting fiberoptic cables allows the flow cell to be housed
within a high temperature cabinet to maintain sample integrity An optimized temperature control
system maintains temperature stability of better than 1°C.
The modular construction of the probe and sampling system are designed for simplified maintenance
with minimum down-time. The flow cell can easily be removed for cleaning and replacement. The
pre-aligned nature of the flow cell optics, coupled to the optical fiber cables, require no optical
alignment after cell re-assembly. The high signal-to-noise requirements of the IPM 350DN requires
the flow cell to have high optical efficiency to UV radiation Great care is taken in the design and
construction of the optical components of the flow cell in order to guarantee optical efficiency
-------
Chemometrics and System Calibration
TheUV (or IR) spectra of species such asNH3, SO2 and NOx are distinct but highly overlapping,
and require the use of advanced chemometric, multivariate pattern recognition techniques to extract
the concentration information data from the highly overlapping UV spectra of the major species".
The PCR (Principal Component Regression) chemometric technique is used to form a robust analyzer
calibration. The PCR technique involves the generation of a prediction matrix from measured
calibration spectra of mixtures of the species of interest The calibration matrix may then be used to
analyze the spectrum of the flue gas to produce the measured component concentrations
Calibration involves the systematic collection of spectral data within the analyzer at known species
concentrations and the generation of the calibration prediction matrix The complete measurement
Figure 5 Calibration Spectra of Various NH3, SO2, NOx Mixtures
space must be mapped out in the calibration cycle The IPM-Mark II 350DN system was calibrated
with individual and derived mixtures of certified SO2, NOx, and NH3 gases (Figure 5). The figure
shows a data set of 50 calibration spectra of various known mixtures of NH3, SO2 and NOx plotted
as a three dimensional grid with the z-axis being the spectrum number from 1 to 50.
The chemometric calibration process is carried out with a proprietary NovaChem software package
which takes the spectral calibration files and the known concentration information and generates a
maximum likelyhood calibration file, which is used by the NovaChem Chemometric Prediction
Engine. The chemometric process involves the generation of Cross-Validation information on the
self consistency of the data set. Simply, some spectra are removed from the set, a calibration matrix
is produced with the remaining data and the spectra removed from the data set are then operated on
with the prediction matrix and the prediction numbers compared to the known concentrations of the
samples. The comparison between the predicted numbers and the known concentrations are the so
-------
,7"
"Z
Figure 6: Cross Validation; NH3
Figure 7: Cross Validation- SO2
called Cross-Validation plots. The more linear the plots, the better the calibration. These are shown
forNfB (Figure 6) and SO2 (Figure 7) for the calibration data set plotted in Figure 5. A similar plot
is available for NOx. The high linearity of the data confirms the quality of the calibration data set and
of the independently determined concentration values for each spectrum. Spectral data noise and/or
inaccuracies in the concentration values will lead to higher data scatter about this line.
Wet Chemistry Study
In order to evaluate the results for the measurement of ammonia obtained on the NovaChem IPM
350DN Ammonia Analyzer and to allow calibration based upon wet chemistry measurement, a
comparison was made with ammonia measurement using NIOSH Method 6701, according to the
following procedure:
Seven wet chemistry sample collection periods were employed varying in length from 35 to 80
minutes with the length of each period primarily dependent on the requirement for obtaining at least
20 cubic feet of sample gas for the wet chemistry procedure.
CEM Services, Inc. of Norton, MA conducted the wet chemistry according to the attached NIOSH
6701 procedure. ESA of Chelmsford, MA performed the lab analysis. The measurement values for
NH3, S02, and NOx for the IPM 350DN for all sampling periods were recorded every 30 seconds
on a computer disk with time and date labels. The measurement values from the NovaChem IPM
350DN Analyzer were witnessed by and reported to Public Service of New Hampshire staff at the
time of measurement and before the results of the independent ESA lab studies were known.
The first four sampling periods involved the simultaneous collection by both measurement techniques
of stack gases using a split sample line. The high vacuum pressures required by the wet chemistry
techniques (10-20 inches of mercury) required the NovaChem IPM 350DN to operate at vacuum
pressure, which is not the usual operating condition.
-------
The method of splitting samples during period 1 to 4 with high and varying vacuum pressure created
by the wet chemistry system proved unsatisfactory for obtaining continuous results from the analyzer
forNHS, SO2 andNOx over the entire measurement period. Therefore, in periods 5 and 6, separate
measurements for each technique were obtained with sampling by the NovaChem analyzer for a
period of at least five minutes before and at least 5 minutes after each 30-40 minute wet chemistry
sampling period. It was assumed that the plant operation would remain constant during the entire
period.
In the period 7, a wet chemistry sample was obtained from an analyzed cylinder of Ammonia (53 ppm
NIC balance N2) with the analyzed value blinded to the CEM Services staff. Analyzed NH3 gas (53
ppm) and SO2 gas (981 ppm) were sequentially injected into the sample port of the IPM 350DN
analyzer and measurements were obtained.
Ammonia injection rates for the SNCR system were varied from 0 to 2 GPM during the testing.
Injection rates were to be held constant during each of the sampling periods. The operation of the
ammonia injection was controlled by PSNH personnel and was blinded to the NovaChem and CEM
Services staff.
Results with the NovaChem Analyzer
The results obtained from the comparison study are shown in Table I below.
Table I
Comparison of NovaChem IPM 350DN Analyzer and Wet Chemistry Results
Run#
1
2
3
4
5
6
7 (Calibration Gas)
Run Time
09:13-09:58
10:11-10:56
11:18-12:28
12:45-14:09
14:30-15:21
16:04-16:58
17:15-17:45
WET
CHEMISTRY
0
0
0
0
1
1
49
NovaChem
350DN
1
1
1
1
3
3
49
The wet chemistry studies confirmed the results of many previous studies with the NovaChem
analyzer on the Unit #1 generator showing that there was essentially no ammonia slip at this point
in the flue with the ammonia set at 2 GPM. The results of a 20 hour monitoring period is shown in
Figure 8. The ammonia level is indicated by the axis on the PJIS of the figure. The level of SO2 and
10
-------
NOx are given by the LHS axis. The ammonia slip showed a drift of approximately +/-1.5 ppm about
zero which represented the accuracy of the analyzer during the study period. Recent improvements
in chemometric calibration and optical stability have brought this accuracy level down to below +/-
1 ppm over a 24 hour time period. The vertical spikes on the figure represent the positions of the
blowback periods every 30 minutes.
During a four month trial period the system was operating continuously better than 99% of the time
and analyzer stability remained at the level seen above. The probe required cleaning once because of
the high vacuum pressure required during wet testing. The flow cell and light source required no
adjustment or cleaning during the trial period. No fouling of the windows of the flow cell was noted.
8-15-96 -8-16-96
1200
1000-
1720.3518:33:3519^6:3520:59:332212:332325:33 0:38:34 151:34 3:0434 41734 5:3035 6'4335 7:5635 90933 10.22:3311:35:33124634
8/15/96 8/15/96 8/15/96 8/15/96 8/15/96 8/15/96 8/16/96 8/16/96 8/16/96 8/16/96 8/16/96 8/1ff96 8/1&96 8/16/96 8/16/96 8/16/96 8/16/96
Figure 8: Operation of the IPM-Mark II350DN Analyzer Over a 20 hour Period
Additional studies were conducted by spiking the system at the probe with small sample of 20%
ammonia for one second. The result of the spiking is shown in the Figure 9. The spiking experiment
was repeated twice in quick succession. It is seen that the IPM 350DN responds within 15 seconds
Figure 9: On-Line Data Showing Ammonia Spiking and Blowback Periods
11
-------
and returns to normal values within 3 minutes during both spiking episodes. No perceptible drop in
S02 level is seen during the spiking period. Note that the range of SO2 and NOx levels shown on
this araph is from 500 to 1100 ppm. An extremely small drop in NOx is perceptible during the first
spiking episode but not during the second. Blowback periods are shown even- 30 minutes.
The concentration measurements of Figure 9 show three periods in which the ammonia level on unit
1 was set to 3 gpm (gallons-per-minute), decreased to 0 pgm and then increased again to 3 gpm. In
both cases in which the level of ammonia was increased, the NOx level initially dropped from
approximately 700 ppm to 600 ppm but after a few minutes rose to a level not significantly below the
inital value showing a net reduction of no more than 5% in the NOx level. During this time period
no significant change is seen in the ammonia measurement level, which is essentially reading zero.
Conclusions
The ammonia measurements obtained on the PM-350DN analyzer consistently 1-2 ppm higher than
the wet chemistry studies.
The IPM350DN analyzer and the wet chemistry measurement gave identical readings of calibrated
ammonia sample in nitrogen.
During simultaneous IPM-350DN and wet chemistry measurements of the flue gas neither analyzer
showed any appreciable ammonia slip in six separate readings.
Spiking the IPM-350DN analyzer with ammonia in the presence of S02 and NOx produced the
expected ammonia reading increase and demonstrated the robustness of the chemometric prediction
engine for the measurement of ammonia in the presence of the other two species.
Upon ammonia injection, although the NOx levels in the flue were shown to decrease initially, after
just a few minutes the levels increased to approximately 95% of the original value. This pattern
repeated itself a number of times. This increase suggests conversion of ammonia to NOx at the high
temperature levels at the point of injection, and may explain the absence of ammonia and the poor
efficiency of denoxificatioin.
The NovaChem analyzer has the accuracy, stability and robustness required in a flue gas ammonia
analyzer for coal fired power stations. The PSNH Unit #1 power plant was producing negligible
ammonia slip when measured by the NovaChem or wet chemistry technique.
Acknowledgments
The authors wish to acknowledge the help of Mr. James Philbrick, Mr. Roger Deshaies and the
engineering staff of the Public Service of New Hampshire, Merrimack Station for their help in the
installation and testing of the NovaChem IPM-Mark II 350DN Analyzer.
12
-------
References
1. "Collaborative Decision Making to Reduce Emissions from Centralia Power Plant", J.
Peterson, PNWIS-A&WMA Annual Conference, Seattle, December 11-13, 1996.
2. "The NOxSO Process: Simultaneous Removal of SO2 and NOx from Flue Gas", J.T. Yeh,
C.J. Drummond, J.L. Haslbeck and L.G. Neal, Presented at 1987 Meeting of AIChE,
Houston, Texas, March 29-April 2, 1987.
3. "Evaluation of the NOxSO combined NOx/SO2 Flue Gas Treatment Process", H.P. Tseng,
J.L. Haslbeck and L.G. Neal, DOE Report; DE-AC22-FE6-0148, September 30, 1983.
4. "A Mathematical Model for Flue Gas Conditioning by Ammonia and Sulfur Trioxide", C.
Chang, H. Bai and C. Lu, 88th Annual Meeting A&WM Conference, San Antonio, Texas,
(1995).
5. "Combustion", I. Glassman, Academic Press, New York, (1977).
6. "The Middle Ultraviolet in Science and Technology", A.E.S. Green, Editor, John Wiley &
Sons, Inc., New York, 1966.
7. "The Identification of Molecular Spectra", Fourth Edition, R.W.B. Pearse & A.G. Gaydon,
Chapman and Hall, London, (1976).
8. "Molecular Spectra and Molecular Structure; Electronic Spectra and Electronic Structure of
Polyatomic Molecules", Volume III, G. Herzberg, Krieger Publishing Company, Malabar,
Florida, (1991).
9. "Ultraviolet Absorption of S02: Dissociation Energies of SO2 and SO", P. Warneck, P.P.
Marmo & J.O. Sullivan, J. Chem. Phys., 40, 1132-1136, (1964).
10. "Spectres dabsorption dans le proche ultraviolet et visible des composes minoritaires
atmospheriques N02 et SO2 entr 200 et 300 K", E. Hicks et. al., J. Chem Phys., 76, 693,
(1979)
11. D.Haaland & E. V. Thomas, Anal. Chem., 60, 1193, (1988).
13
-------
In situ Analyzer Using a Near Infrared Diode Laser
for Process Control and Environmental Monitoring
Joel Brand, Kenneth Groff, Garth Monlux, and Patrick Zmarzly
Monitor Labs, Inc.
76 Inverness Drive East
Englewood, CO 80112
Abstract
An ammonia monitor designed for in situ smoke stack or exhaust duct applications is
discussed. This instrument is useful in a wide variety of ammonia monitoring and
process control applications, particularly ammonia-based NOX control technologies,
including selective catalytic reduction (SCR) and selective non-catalytic reduction
(SNCR). The sensor technology utilized in this instrument is second harmonic
spectroscopy using a near infrared diode laser. A probe comprised of a diffusion cell
with protected multi-pass optical measurement cavity provides the optical interaction
with the sample. Other components of the system include signal processing electronics
and an embedded computer platform. The in situ design eliminates sample handling
issues, associated with extractive analysis of ammonia, such as sample line adsorption
and heated sample trains and cells. Data collected during field trials from both SCR and
SNCR applications demonstrate the feasibility and robust operation of this instrument in
traditionally problematic operating environments.
Introduction
Accurate, reliable, real-time ammonia monitoring has been a difficult process control
problem for those using ammonia in a variety of applications. In particular, the use of
ammonia and ammonia generative compounds in combustion NOX reduction technologies
such as selective catalytic reduction (SCR) and selective non-catalytic reduction (SNCR)
have shown great promise, however the measurement of ammonia is very problematic in
these environments. Too much ammonia results in the following problems':
1. Contamination of fly ash. What would normally be a salable by-product for use as a
filler material in wallboard, paving or other construction applications now becomes a
hazardous waste when laden with ammonia. Removal from the plant can be
problematic if truck drivers refuse to haul a load of ash that is too odoriferous.
2. Excess reagent cost. Any excess ammonia emissions beyond what reacts in the
process represents additional cost to the plant operator.
-------
3. Fouling of equipment. With an overabundance of ammonia in flue gas, ammonium
salts can form causing severe fouling in air preheaters and catalyst beds. This can
cause such pressure drop and loss of efficiency that it may necessitate a plant
shutdown for cleaning; while the cleaning process itself may not be too difficult an
unplanned shutdown can be very costly.
4. Regulatory issues. In some areas, particularly in California, there are regulatory
limits on ammonia slip from the stack.
In developing an ammonia analyzer, there are issues to be addressed which are common
to all stack monitoring equipment, and those which are specific to ammonia. The main
problem with ammonia is that it is very "sticky"; the unbonded electron pair on the
nitrogen atom in ammonia allows it to interact weakly with many substances. This means
that it will adsorb and desorb on many surfaces in any sample handling system leading to
inaccurate measurement of ammonia, especially under conditions of changing load.
Another problem with ammonia is that it can be reactive and form salts that can be
dropped out in a traditional extractive system.
The general problems common to any stack measurements are: a hot, wet sample,
typically with acid dewpoint problems and heavy particulate loading; severe vibration
resulting from fans and baghouses; exposure to adverse weather conditions including
temperature extremes; and the survivability in an industrial environment where the
instrument may suffer from unstable power and mechanical shocks from falling items or
unplanned collisions with other equipment. There are many strategies that have been
used in the past to monitor ammonia in this environment, including ultraviolet dispersive
in situ, extractive with conversion and chemiluminescence detection, and NO, difference
monitoring methods.
Ultraviolet dispersive methods, typically in situ, have proven themselves for monitoring
NO and SO2. This technique can also be applied to measure ammonia, but the overlap
with the electronic transitions in SC>2 proves to be very problematic. This approach can
be acceptable under the rare process control applications in the absence of SC>2.
One of the most common techniques used to measure ammonia is to extract the sample
and run it through any variety of traditional analyzers, with a chemiluminescence NOX
analyzer and thermal ammonia converter being the most common. In this configuration,
the stream is switched back and forth between an oven that converts all of the ammonia to
NO and another stream that measures just the NOX in the sample. The ammonia
concentration is then just the difference in NOX readings between these two channels.
While this approach gives an accurate measurement of the ammonia concentration at the
location of the analyzer and oven, it is not necessarily very reflective of the concentration
in the stack or duct because of the sample handling issues mentioned above.
Another technique is to monitor inlet and outlet NOX from a control device, e.g. SCR.
Unless all of the other process variables are well-characterized, the true ammonia
-------
concentration may not be inferred to within a few ppm of ammonia, which is generally
the accuracy required to keep a tight rein on the excess ammonia problems listed above.
There are a number of other novel techniques which have been applied to this problem,
for example ion mobility spectroscopy or Fourier Transform Infrared Spectroscopy
(FTIR) While some of these are excellent choices for open path or laboratory work, they
are not well-suited to real-time stack measurements. This paper describes the
development of tunable diode laser based instrument operating in the near infrared that
makes the measurement entirely in situ, avoiding all of the sample handling issues
associated with ammonia. It utilizes the solid state aspects of the laser diode and makes
for an instrument with no moving parts (other than the solenoid valves for calibration
gases). Other advantages of this technique are the sharp wavelength specificity, high
sensitivity when using harmonic spectroscopy techniques, and small and rugged
packaging.
Theory of Operation
The present instrument uses second harmonic spectroscopy2, which is closely related to
the more familiar second derivative spectroscopy3 which has been used in stack
instruments for quite some time4. In fact, the so-called second derivative technique is
really an approximation to the true second harmonic, and in practical applications the two
techniques are generally indistinguishable.
Consider the absorption of light through a gas sample. The intensity I(v) transmitted
through a pathlength / and density p is given by Beer's law:
7(v) = v-wM, (1)
where Io is the incident intensity, S, is the linestrength of the absorbing Line, and /(v) is
the normalized lineshape function. For weak absorption lines, the exponential may be
approximated by the first term in the Taylor series. In this approximation, the fractional
absorption is given by the exponent.
In the pressure broadened regime, which is valid above a few tens of Ton, /(v) can be
well represented by a Lorentzian:
where v is the frequency, VQ is the linecenter, and F is the half-width half maximum
(HWHM).
Now consider what happens if the laser frequency is modulated back and forth across the
absorption line. In practice this is accomplished by modulating the laser current about a
DC value that is fixed to correspond to the line center. An absorption dip will be seen at
twice the frequency of the laser modulation: once on the first half of the cycle as the laser
frequency is increasing and again on the last half of the cycle as the laser frequency is
-------
decreasing. Heuristically, we would expect to see a component of the return signal at
twice the modulation frequency. Mathematically, if the laser is modulated at an angular
frequency eo with an amplitude a about it's center frequency, vc, the resulting expression
for the frequency will be:
v(;) = vc +acos(
-------
Figure 1 shows a schematic diagram of the instrument The probe which houses the
optical cavity and interacts with the stack gas is mounted upon a weldment and inserted
into the stack. The enclosure which houses all of the electronics and remaining optics
mounts to a flange on the weldment so that the whole instrument may be installed as a
one piece unit.
Following the signals downstream through the instrument, the light source is a distributed
feedback (DFB) communications diode laser operating hi the l.SjJ. band. It is temperature
controlled and current controlled. Temperature control ensures stability of the laser and
provides a coarse tuning of the wavelength. The current is modulated and the DC value
is controlled via a feedback loop from a captive gas sample, denoted the reference cell.
This current control provides precise wavelength control and modulation. The detector
on this cell monitors the third harmonic from the captive ammonia sample and keeps the
laser line-locked precisely to the absorption line of interest; the reference cell is in no way
associated with the calibration of the instrument. The light from the diode is collimated,
sent through an optical isolator to reduce unwanted external cavity feedback, and split
into two beams: one goes through the reference cell described above and the other is
coupled into a high temperature fiber.
The fiber launches the beam towards the multipass Herriott cell where it interacts with the
stack gas. One beam reflects off a lens in front of the Herriott Cell and back to a detector
in the instrument enclosure while another beam traverses the measurement cavity and
falls upon another detector in the enclosure. Any spurious fluctuations in the second
harmonic resulting from laser modulation or fiber coupling may then be accounted for by
subtracting the normalized signals from these two beams leaving just that contribution
from the measurement cell. The signals from these two detectors are filtered and
analyzed with lock-in amplifiers to extract the second harmonics on an analog signal
processing board. The other analog control and measurement functions including the
third harmonic lock in for the reference cell are located on this board. The sensor board
scales signals from and drives temperature and pressure sensors, and controls valve
switching for calibration.
The measurement cavity which houses the Herriott cell is a diffusion cell formed by a
tube with sintered metal filters welded into the surface to prevent particulates from
entering the cavity. The filter area and porosity have been engineered to deliver a rapid
response time in the stack while not requiring an excessive amount of calibration gases.
Calibration gases are injected into the cavity after passing through a heat exchanger; the
instrument calibration is thus performed with gases at the stack conditions, hi this
manner the effects of temperature and pressure correction to the concentration
measurements are minimized since the deviations from calibration conditions are
generally small.
The instrument is run by two microprocessors utilizing a LonWorks™ based network
technology, one mounted on the analog board and the other on the sensor board. The first
-------
micro reads in fast signals from the lock in amplifiers, performs startup scanning and
linelocking function, controls the laser during normal operation, and calculates raw
concentration. The second micro reads in the temperature and pressure, performs
averaging, calibration and compensation functions. The final concentration is then sent
out over the LonWorks™ bus to a plant control system, data acquisition system, or
display and analog output module. The use of the Lon Works™ network allows for easy
remote access to variables and troubleshooting as well as a proven industrial interface
that is resistant to noise and high voltage transients.
Results
Various prototypes that have evolved to the present instrument have been tested over the
last several years. In January of 1996, an early prototype was installed near the entrance
to the stack at the Public Service of Colorado Arapahoe Station. This unit is a coal fired
boiler with an SNCR. Stack conditions were approximately 10 % water and 135 °C.
Figure 2 shows data from this prototype. Although it was quite noisy, the data showed
rough agreement with other instruments being tested. Most of all, it showed excellent
survivability and proof of principle was demonstrated. Lessons learned from this field
trial allowed us to redesign much of the electronics and optics to reduce the lower
detectable limit (LDL).
The second and third field trials were conducted at the GE cogeneration station in
Kingsburg, California. This is a gas fired turbine, with water concentration nearly 30%
with the use of steam injection and a stack temperature of 165 °C. On first attempt,
dramatic water interference from an unrecorded hot band was observed. This was not
observed at the first trial since the LDL was nearly 5 pptn and both the temperature and
water concentration were much lower. This necessitated finding a laser that could access
an interference free line. Figure 3 shows laboratory data taken after the new laser was
installed and many additional design modifications were employed to boost performance.
Figures 4 and 5 show data from this field trial showing low drift and noise and favorable
comparison to their extractive system, although close comparisons might not be too
meaningful.
Summary and Future Work
This instrument design has proven to be a rugged, reliable unit with excellent data
availability and good accuracy. Noise and drift numbers in the field are well below a one
ppm LDL, an approximate threshold for the desired accuracy of a process control
ammonia monitor. The optical alignment and mirrors survived well in this environment
and calibration drift was negligible over a four month trial.
The final test of this design will be to install it in a high SO2, high temperature
environment as might be encountered immediately downstream from an SCR in a coal
fired application. Freedom from interference from hot bands, survivability of the mirrors,
-------
and general survivability of the instrument at high temperature will be the major hurdles
to be overcome in this environment.
Acknowledgments
The authors wish to acknowledge the assistance of many collaborators at various stages
in the research, design, and testing of these instruments. Colleagues at Monitor Labs, Inc.
include Dr. Greg Fetzer, Mr. Scott Harrison, Mr. Benjamin Halsted, Mr. Richard Hooper,
and Ms. Xia Zhong; Colleagues at Spectral Sciences, Inc. include Dr. Neil Goldstein, Dr.
Jamine Lee, and Dr. Steve Richtsmeier; Dr. Pam York from David Samoff Research
Center; Mr. Terry Hunt from PSCO-Arapahoe; and Mr. Mel Murphy from GE-
Kingsburg.
References
1. See, for example, Leslie L. Sloss, et. al. Nitrogen Oxides Control Technology Fact
Book, Park Ridge, New Jersey: Noyes Data Corporation, 1992, p 94-95.
2. H. Wahlquist, "Modulation Broadening of Unsaturated Lorentzian Lines", J. Chem.
Phys. vol. 35, 5, pp. 1708-1710 (1961)
3. J. Reid and D. Labrie, "Second Harmonic Detection with Tunable Diode Lasers-
Comparison of Experiment and Theory", Appl. Phys. B. 26, pp. 203-210 (1981)
4. For example, the Monitor Labs 8100 series of in situ UV analyzers.
-------
Calibration gas
injection tube and
beat exchanger
Detectors
Stack Range
Calibration
Valves and
connections for
gases
Electronics Boards
(including LonWorks™
hardware)
Sintered Metal
Filters
Herriott Cell
Weldment
High
Temperature
Optical Fiber
Laser Diode Module,
(including reference
cell and optics)
LonWorks™
Network
Connection
Figure 1. Schematic of the TDL instrument design. See text for description of
components
-------
50
45 h
40
35
30
e- 25
20
15
10
5
0
Q-
Q.
30
60
Time (Hours)
90
120
Figure 2. Comparison of first prototype data to wet chemistry measurements taken from
2/23/96 to 2/28/96 at the Public Service Company of Colorado Arapahoe Power Station
in Denver, Colorado. This prototype was installed from February to April, 1996. The
flat lines at 30 and 75 hours are periods when the instrument was down for maintenance.
-------
25
20
"E 15
0.
_CL
3?
10
5
!
-r
1
J
^
.
r
.
I t i
t
-
-
15 30 45 60 75 90
Time (Hours)
Figure 3. Laboratory measurements of the second generation instrument showing 80 ppb
zero drift and 210 ppb span drift.
-------
30
25-
20
10
50
100
Time (Hours)
150
200
Figure 4. Comparison of prototype (solid line) to measurements made with a Monitor
Labs dilution extractive system (circles). These data were taken at Kingsburg
degeneration Facility in Kingsburg, CA.
-------
25-
20-
15-
I. f(
s /rv
i" i; i
10
5-
0 —
V* ^yUMX'
J w
20 40
60 80
Time (Hours)
100
120 140
Figure 5. Prototype data from Kingsburg field trial showing daily zero and span checks.
Dotted lines are expected zero and span levels.
-------
LOW MAINTENANCE, HIGH PERFORMANCE GEMS FOR NOx, CO, AND O2
MONITORING
Paul F. Schubert, Thomas Dadisman, and William Ryder
Monitor Labs, Inc.
76 Inverness Drive East, Englewood, CO 80112
Abstract
An innovative new sensor based CEMS has been developed for the measurement of NOx,
CO, and Oj. This new highly integrated system is designed with a NEMA 4X type
housing to allow it to be installed in harsh environments without the traditional analyzer's
need for a temperature controlled shelter. Field experience has shown that the sensors
themselvei and the supporting systems require little maintenance. The rugged design and
low maintenance requirements result in a substantial reduction in initial installation cost
as well as reduced operating cost compared to traditional CEMS. The relative accuracy,
drift, response time, and thermal stability of the sensor based analyzer when mounted in
an outdoor environment is comparable to the performance of a traditional CEM system
installed in a controlled environment. The system has been certified in natural gas fired
applications.
Introduction
Industry continues to confront increasing requirements for monitoring emissions while
competitive and economic forces drive them toward cost and staff reductions. These
forces increase the desirability of new, simplified approaches to meeting the regulatory
monitoring requirements. The traditional continuous emissions monitoring (CEM)
systems measure oxides of nitrogen (NOx), carbon monoxide (CO), and oxygen (02)
using three independent instruments. Each instrument utilizes a different sensing
technolgy to achieve acceptable sensitivity and specificity. These systems are typically
expensive to install and maintain, and require substantial technician level support. As a
result, considerable resources have been applied to create cost-effective monitoring
solutions. Generally, sensor based approaches have been chosen because they provide
the opportunity to replace costly instrumentation with low cost sensors1.
Sensor based approaches have typically been associated with predictive or parametric
methods of emissions monitoring. This approach creates a model correlating data
collected such as flow, temperature, pressure, torque, etc. under various source operating
conditions with emissions data collected under the same conditions using traditional
analyzers2"3 While some success has been achieved using these methods, direct
-------
measurement of desired exhaust gases provides greater confidence than can be achieved
using modeling approaches. The recent development of a solid state catalytic NOx
sensor4 allowed development of an analyzer combining the advantages of sensors with
the direct measurement capability of traditional continuous emissions monitors.
The new sensor based analyzer, the CEMcat™ continuous emissions monitor, utilizes a
single, compact, sensor module containing the three sensors for NOx, CO and O:
measurement. The development of this analyzer was funded primarily by the Gas
Research Institute5 (Chicago, IL). The objectives of the program were develop an
analyzer with the following features.
• performance equivalent to traditional CEM systems for natural gas, refinery off-gas,
and light fuel oil applications;
• ability to install the system outside without the need for a separate shelter; and
• very low maintenance requirements, allowing installation in remote sites such as
unstaffed natural gas transmission stations.
The system was tested in prototype form extensively in the natural gas transmission
industry to monitor emissions from gas turbine engines and large gas-fired reciprocating
engines6. In these applications, the CEMcat analyzer demonstrated its capability to meet
40CFR60 relative accuracy requirements7"8 Further development of the system allows it
to meet the requirements of 40CFR75 for gas fired applications where fuel flow
measurements can be used to calculate emission rates. The system is also being applied
for SCR system process control.
Catalytic Sensors
The analyzer uses spark plug sized catalytic sensors to measure NOx and CO. Each
sensor has two probes, an active and a reference, exposed to the sample stream. At the
tip of each probe is a 100 fi precision resistance temperature detector (RTD). The active
catalytic probe is coated with a thin layer of catalytic material which is selective for the
gas of interest. The reference probe is coated with an inert ceramic material (Figure 1)
Both probes are placed within a sample chamber maintained at 300 °C (572 °F). This
temperature is selected for optimum catalyst performance, to equilibrate the exhaust gas
sample temperature with the sensor temperature, and to minimize effects of temperature
variation in the exhaust stack. The system controls the temperature of the sample
chamber to within about 0.1°C. Critical flow orifices ensure constant flow across the
probes (Figure 2).
As the gas flows across the probes, the target molecules react on the catalyst surface and
give offheat (an exothermic reaction). The heat changes the resistance in the RTD of the
catalytic probe, while the reference probe remains at the ambient temperature of the
sample chamber. The resistance difference between the two probes is measured using a
bridge circuit, and is directly proportional to the concentration of the target molecules in
the sample gas stream. The sensors are designed to give a resistance difference of
-------
approximately 1.2 mO per ppm of the target gas (NOx or CO). The sensors have a rapid
response time, giving a 90% of full scale reading in less than 30 seconds.
Catalytic coating
100n RTD's
Reference coating
Stainless steel sensor body
Figure 1
Catalytic sensor design
Sensor
Stainless Steel
Block at 300° C
Sample
Gas
Outlet
Figure 2
Schematic of a catalytic sensor in the heated sensor module
Unlike pellistor or catalytic bead sensors9, the catalyst coatings for the NOx and CO
catalysts are high surface, high porosity material. This results in a large number of active
sites for catalytic reactions throughout the coating. This ensures a long life and
resistance to poisoning for the sensors. In addition, the numerous active sites in the
porous structure prevent the system from becoming saturated. This distinguishes them
electrochemical cells, which are prone to saturation. It is this same property which gives
these sensors an exceptionally large dynamic range. They have been used successfully
for measuring normal emissions in the 20 ppm range up through emissions in the
-------
percentage range, and are linear over that entire range. Further advances in electronics
and system control are expected to allow extension beyond this range.
The NOx Sensor
The NOx sensor is a true NOx sensor, measuring both NO and NO2 simultaneously.
Conversion of NO2 into NO (which is required for chemiluminescence based analyzers)
is unnecessary for catalytic monitors. This sensor uses a vanadium based catalyst ,
similar to the catalyst used in selective catalytic reduction (SCR) units, to reduce NO and
NO: as shown in equations (1) and (2).
2 NO + 2 NH3 + 1/2 O2 ->• 2 N2 + 3 H2O AH°6oo K = -97.1 kcal/mol (1)
2 NO2 + 2 NH3 -> 2 N2 + 3 H2O + 1/2 O2 AH°60oK = -83.1 kcal/mol (2)
For these reactions, trace levels of ammonia are introduced into the sensor chamber as a
co-reagent. The ammonia is typically added as an ammonia/oxygen blend (about 1.5%
ammonia in air for a 500 ppm NOx source). The use of the ammonia/air mixture ensures
sufficient oxygen to complete the NO reaction in applications where the normal oxygen
content of the gas stream is low, and eliminates any effects of oxygen concentration on
NOx sensor response. In addition, the use of this diluted ammonia supply avoids safety
and handling concerns surrounding anhydrous ammonia.
Noise for the sensor system is due primarily to electronics noise, temperature
variability within the sensor module, and turbulence within the sensor chamber. The
system has been designed to minimize these effects. Figure 3 shows the zero noise for
the sensor in a 0 to 1000 ppm system over a 10 hour period. As can be seen in the figure,
the noise level is quite low. Similar results are obtained for noise at the span values for
the system (530 ppm for NOx, and 440 ppm for CO), as shown in Figure 4.
ppm
-10
ft^r^lWV1**
0123456789 10
Hours
Figure 3
NOx sensor zero noise
-------
Q.
D.
600
500
400
300
200
100
0.
c
NO
CO
) 1 2 3 *•
Hours
Figure 4
NOx and CO sensor span noise
The CO Sensor
The CO sensor uses a precious metal based catalyst to oxidize CO to CO2 as shown in
equations.
CO +1/2 O2 -> CO2
AH°6oo K = -67.6 kcal/mql (3)
Oxygen already present in the exhaust gas serves as the required co-reagent for this
oxidation reaction. The sensor has a large dynamic range, and has been successfully used
to measure emissions levels from the 0-60 ppm range up to the 0-8000 ppm range in the
field.
Experience in combustion control applications has shown that these precious metal based
sensors also respond to unburned hydrocarbons in the exhaust gases. For example,
methane can oxidize over the catalyst as shown in equation (4).
CH, + 2
CO2 + 2 H2O
(4)
However, the catalyst is designed to minimize the sensor's response to such
hydrocarbons at the 300 °C operating temperature. As a result, the CO sensor's response
to CO is 150 times greater than its response to methane when operating at 300 °C. In
most applications, methane represents over 90% of the hydrocarbons present in the
effluent stream. This low sensitivity to unburned hydrocarbons allows the sensor based
continuous emissions monitor to meet regulatory requirements in most applications.
-------
The Oxygen Sensor
The new continuous emissions monitor uses a high performance electro-catalytic
zirconium oxide (ZrOj) based oxygen sensor. These sensors are well characterized, and
have been extensively used in emissions monitoring and combustion control
applications10 Although the oxygen sensor is housed in the same 300 °C (572 °F) sensor
module as the two catalytic sensors, it has its own integral heater which operates at about
850 °C (1550 °F). The high temperature allows the zirconium oxide to operate as a
conduit for oxygen between the higher concentration reference side and the lower
concentration sample side. The concentration of oxygen is then determined from the
electromotive force (emf) produced by this process. This sensor generates a millivolt
output that is a logarithmic function of the net oxygen concentration in the sample gas.
An instrument amplifier conditions the sensor signal which is then digitized and
linearized in software.
Analyzer Configuration and Operation
The basic design and operating parameters of the CEMcat analyzer are
summarized in Table 1 The integrated analyzer can be considered as consisting of three
major systems in addition to the instrument housing. These major systems are the sample
conditioning module, the electronics module and user interface, and the sensor module.
The system is enclosed in a NEMA 4X type enclosure, which allows it to be mounted
outdoors and exposed to the weather The sample conditioning module includes the
probe which penetrates the stack, the short heated sample line, the water removal system,
the sample pump, and a paniculate filter. The electronics module includes the signal
conditioning, microprocessor, I/O and communications circuitry, the power supply, and
operator interface (keypad and display). The microprocessor controls system operation
(display, keypad, I/O), temperature of the manifold and sensor module, and features like
automatic calibration and blow-back. The sensor module consists of the heated sensor
block containing the the NOx, CO, and O2 sensors, and the eductor, which is used to
draw conditioned sample gas through the sensors.
The sampling system pulls sample from the stack through the probe and sample line at a
rate of about 3 slpm (0.10 scfin). The sample is passed through a two stage thermo-
electric chiller to cool the sample and decrease the moisture content of the sample gas to
less than 1%. Sample is brought under vacuum through the first stage of drying and
maintained at an elevated pressure through the second stage in order to ensure a low
sample dewpoint. The condensate formed is continuously moved from the sample using
a peristaltic pump. Once the sample is conditioned and filtered, a secondary eductor pulls
a portion of the sample gas from the main stream at a rate of 1.8 slpm (0.06 scfm). This
secondary stream is then split into three equal portions and drawn separately through
each sensor using an orifice to control flow. The vacuum generated by the secondary
eductor must be sufficient to cause the flow through each orifice to be critical. The
specificity of the three sensors can be improved by correcting interferences to the
-------
measurement of one sensor from the reading of the others. This is accomplished by real-
time algorithms built into the CEMcat's multiplexing operating software. The gas flow
path for the analyzer is shown in Figure 5.
Table 1
CEMcat Analyzer Design and Operating Parameters
Physical Design
Design and Dimensions
Materials and Protection Ratings
Gas Inlet Conditions
Temperature
Pressure
Flow Rates
Operating Conditions
Temperature
Humidity
Power Supply
Input voltage
Power consumption
Dimensions 76 cm (30") x 122 cm (48") x 34 cm (13.25")
Weight 118 kg (262 Ib)
Housing of rust and acid resistant steel (1.4571;316ti per
SAE/AISI)
Polycarbonate widow for viewing display
Standard NEMA4X/TP 56 type
Sample conditioning provided for 600 C (1112 F) inlet
temperature.
Inlet, Atmospheric to -5.08 mm water (-2 inches of water) at
4°C.
Outlet. Atmospheric
Inlet to instrument 3 liters per minute of stack gas sample
removed
Inlet to individual sensors: approximately 0.6 liters per
minute per sensor (3)
Drive air to operate eductor: <1 liter/sec at 5143 hPa (<1 scfin
at60psig)
-30°C to +50°C (-20°to +120T) ambient
up to 100% relative humidity
105-120-132 VAC @ 60 Hz? or 195-230-264 VAC @ 50 Hz
1200 watts for integrated package
82 watts/meter (25 watts/foot) for heated sample line
300 watts for standard heated probe
-------
Stack
Sample
3 Ipm
Exhaust
Exhaust
Instrument Air
6 Ipm
Figure 5
Gas flow path through the analyzer
Analyzer Installation
With its pre-engineered design, the CEMcat analyzer is has simple installation. It can be
installed like any other extractive CEM system inside a temperature conditioned shelter.
However, it has the added advantage of being able to be mounted outside, and exposed to
the weather. This gives the user the flexibility to choose the most cost effective and
convenient location for the system Typically, it has been mounted at the base of the
exhaust stack, inside a shelter if one is available, or outside if one is not. Installation at
the base of the stack allows a short a vertical heated sample line to be run to the heated
sample probe. This configuration minimizes the amount of heated sample line required.
Figure 6 shows the basic installation mode for the CEMcat analyzer.
4
Probe
> Stack
Wall
^— fc.
W
' Emission
Row
Pipe
v Flange
j^BFL, Heated sample
*fa*^ t£ line 3' to 400'
(1m to 120 m )
•^r
.1 CEMcat Unit
"4
^Communication Line
y To Control Room
Figure 6
Typical CEMcat analyzer installation configuration
-------
The system requires power, instrument quality air, and calibration gases at the installation
site for operation. Installation for the system is typically less than one day, assuming that
utilities are already available. This is a significant advantage over traditional CEM
systems, which must be installed inside shelters or control rooms.
The analyzer can be configured to communicate via analog outputs, serial output, or both.
The analyzer has an IEEE 485/422 serial communication port, and uses a Modbus
communication prototcol. The system is also available with four analog output channels,
which can be used to connect the system to a data acquisition system or distributive
control system. There are also a number of digital inputs and outputs that can be used for
valve actuation, alarms, and other user defined functions.
Analyzer Maintenance
The analyzer was designed to minimize the maintenance required, so that it could be
located at installations where there is little or no operator presence. In pipeline
applications this has included remote, unstaffed transmission stations. In the boiler
installation, staffing constraints require that the system operate for long periods of time
with no attention. The most frequent periodic maintenance requirement is to check the
inlet particle filter on a monthly basis. As with any extractive system, contamination of
the filter can result in degraded performance including slower response time, drift, and
other problems. In the clean fuel applications particulate build-up on the filter is rarely
an issue.
As with any continuous emissions monitoring system, the calibration gases need to be
replaced at regular intervals. In most cases this is every three to four months. These
calibration gases are CO/O2 span gas, and NO span gas, and a CO/NO/low C>2 zero gas.
The ammonia in air used to operate the NOx sensor is replaced on the same schedule as
the calibration gases.
There are a number of other preventative maintenance items which should be done on an
annual basis. This includes checking and replacing as needed the tubing for the
peristaltic pump, which drains the water from the thermoelectric chiller. The probe
should also be checked and cleaned annually. The sensors can be replaced at annual
intervals corresponding to relative accuracy testing of the system as part of a preventative
maintenance program, or can be replaced on an as needed basis. The system diagnostics
provide guidance for the replacement of the sensors by monitoring the sensor response to
the span gas, and determining when the sensor span is insufficient to maintain the
performance characteristics of the system. Individual sensors can then be replaced, or the
entire sensor module housing all three sensors, the heater elements, and the critical flow
orifices can be exchanged.
-------
Analyzer Performance and Certification
The CEMcat analyzer has recently been tested by third party evaluation teams on boiler
and reciprocating engine applications. In both cases, the systems were tested for
compliance with 40CFR60 requirements. The boiler is a subpart DB boiler in a food
processing plant, and the engine is a Clark TLAD-6 two stroke reciprocating natural gas
fired engine in gas transmission service. Each of these series of tests is described in
detail below.
Boiler Performance Testing
The analyzer was installed on a subpart DB boiler in a food processing plant in the U. S.,
and certified under 40CFR60 regulations. Certification consisted of a relative accuracy
test, and a seven day drift test. The relative accuracy was determined by having a third
party source tester conduct a Relative Accuracy Test (Appendix B). The source tester's
sample probe was located approximately one foot above the sensor based analyzer's
probe on the boiler's exhaust stack. Both probes extended to the approximate center of
the stack. The source testers measured NO and NOs using reference method 7E.
Reference method 3A was used for oxygen, and moisture was determined
gravimetrically.
The source tester's measurements were compared against the CEMcat analyzer's total
NOx and oxygen. The relative accuracy test consisted of 9 separate side by side
comparisons of the continuous emissions monitoring system against the reference
method. Each trial lasted a minimum of 21 minutes. Before and after each trial the
source tester re-checked their instrument calibration. The CEMcat system was not
adjusted during the test.
The relative accuracy test produces nine separate mean differences (d,), and one overall
average mean difference (d), for each species under comparison. The final calculation of
relative accuracy is as follows:
Relative Accuracy = (|d[ + |CC|) / Reference Method Mean * 100 (5)
The value CC, or confidence coefficient, is a statistical means to compensate for
variations in the data spread between the CEMS and the source tester. Specifically, it is
the product of the student t for a 95% confidence interval (t=2.306 for 9 tests), and the
standard deviation (Sd) of the mean difference between the CEMS and the reference
method (d), divided by the square root of the number of trials (n), i.e.
^F ¥
CC = (t,K)*~r (6)
V77
Typical emission ranges for the boiler were about 50 to 120 ppm for NOx, and about 2%
to 3% for 02. The average emission levels obtained during the tests, and the overall
relative accuracies are shown in the Table. The U.S. regulations (40 CFR part 60) require
-------
a relative accuracy of 20% for NOx, and either 20% for oxygen or 1% absolute difference
for oxygen. The data show that the analyzer is capable of meeting these accuracy
requirements.
The seven day drift test requires the variation in the daily zero and daily span on the test
gases be less than 2.5% of the full scale in any day during the seven day period. Full
scale for the subpart DB boiler is 500 ppm for NOx, and 21% for O2. The results of the
drift test for NOx are shown in Figure 7. Throughout the entire period, the zero and span
drifts were less than the 2.5% of span limit set by the regulations. There was no
appreciable drift observed for the oxygen sensor. Successful completion of the relative
accuracy and drift test resulted in certification of the system.
Table 2
Relative Accuracy Tests on Subpart DB Boiler
Gas
NOx
O2
Reference
Method
71 ppm
2.52%
Sensor
Measurement
67 ppm
2.49%
Relative
Accuracy (%)
7.0
6.4
-.--/—-=Ss-\ v V 1
-Zao Drift
1 Span Deft
Figure 7
Seven day drift test on subpart DB boiler in a food processing plant
Reciprocating Engine Testing
The testing protocol used for the reciprocating engine is essentially the same as for the
same as used for the subpart DB boiler. However, unlike the boiler, which was operated
-------
at essentially constant conditions, the engine operation was varied to allow determination
of how well the CEMcat analyzer could track emissions under different load conditions.
Nine thirty minute long runs were performed, with the engine adjustments performed
between each run. The reference method analyzers were checked for drift between every
two runs. During the testing, the NOx levels varied from about 50 ppm to 150 ppm, the
CO varied from about 130 ppm to 300 ppm, and the oxygen level ranged from about
14.5% to 16.5 percent. The exhaust flow rate varied from about 226,000 to 400,000 1pm
(8000 to 14,000 dscfrn). The results of the performance testing on the unit are
summarized in Table 3. As can be seen from the data, the system performed very well,
with all sensors giving relative accuracy results of less than 6%. In addition, all of the
precision results during the testing fell close to or well below 5%.
Table 3
CEMcat Performance Testing on Clark TLAD-6 Reciprocating Engine
Load Speed NOx CO O2
HP RPM PPM ppm %
Minimum 1734 239 53.54 140.6 15.62
Maximum 2742 302 145.4 235.7 1641
Average 2159 287 83.95 193.7 15.94
Correlation Coefficient 1 00 0.998 0.983
Average Error -2.33%-l .63%-0.40%
Relative Accuracy 3.14% 5.74% 0.69%
Conclusions
The analyzer has demonstrated excellent performance in boiler, turbine, and
reciprocating engine applications, meeting the requirements of 40CFR60. The excellent
relative accuracy results also indicate that the system will meet the more stringent
requirements of 40CFR75, in applications appropriate to the CEMcat analyzer design.
The success of the testing also confirms that a "clean sheet of paper" approach to
continuous emissions monitoring design and development can create a more cost
effective solution to monitoring. By using the sensor based approach, the instrument
accomplishes the low maintenance objective sought by parametric or predictive
monitoring while simultaneously accomplishing direct measurement of the compliance
gases which is the strength of the gas analyzers.
Acknowledgments
We wish to acknowledge the Gas Research Institute and Southern California Gas
Company for financial support of the emissions monitor, and the U.S. Department of
Energy for support in developing the NOx sensor. In addition, the authors would like to
thank David Sheridan, Mike Cooper, Andrew Banchieri, and Dina Rostrup-Nielsen, all
-------
formerly of Advanced Sensor Devices, and Stephen Devita, formerly of Monitor Labs for
their contribution to the design and development of the CEMcat monitor and the NOx
sensor.
References
1. Schubert, Paul F. National Environmental Journal, January/February 1996, 102-4.
2. Hung, W. S. Y., ASME Paper No. 91-GT-306, June 1991.
3. Hung, W. S. Y., AWMA Paper 94-TA29A.-3, June 1994.
4. Dalla Betta, R. A.; Sheridan, D. R., U. S. Patent 5314 828, assigned to Catalytica,
Inc., May 1994.
5. Schubert, Paul F.; Sheridan, David R.; Banchieri; Andrew J., AWMA Paper 95-
MP16A.03, June 1995.
6. Schubert, Paul F.; Sheridan, David R.; Cooper, Michael D.; Banchieri; Andrew J.
Mechanical Engineering, in press.
1. EPA, "Specifications and Test Procedures for SO2 and NOx Continuous Emission
Monitoring Systems in Stationary Sources," Code of Federal Regulations. Chapter 1,
Title 40, Part 60, Appendix B, Performance Specification 2, 1992, pp. 1108-1115.
8. EPA, "Specifications and Test Procedures for Continuous Emission Rate Monitoring
Systems in Stationary Sources," Code of Federal Regulations. Chapter 1, Title 40,
Part 60, Appendix B, Performance Specification 6, 1992, pp. 1218-1219.
9. W. Gopel, J. Hesse, J.N. Zemel, Sensors : A Comprehensive Survey, Vol. 2, VCH
Publishers Inc., New York, NY, 1991, p. 539-540.
10. Jahnke, James A. Continuous Emissions Monitoring; VanNostrand Reinhold, New
York, 1993, 108-109.
-------
Study of stack flow test and ultrasonic monitor fundamentals
Yiu-Lam Albert Lati
Houston Lighting and Power Company
12301 Kurland, Houston, Texas 77034
Abstract
Based on the W. A. Parish Unit 5 and 6 results, the reference method (KM) flow value is 12 to
15% higher than the three dimensional traverse (3D) value. One benefit of the 3D flow traverse is
a three dimensional map of flue gas flow at the plane of traverse. This velocity profile is essential
in understanding the reasons as well as proposing postulations for the phenomena responsible for
the differences between the two methods. The KM profile is flat compared to 3D profile. This
profile difference and other known phenomena, such as S-type pitot flow coefficient bias under
non-axial flow condition from the literatures and normal velocity vector correction can be used to
quantitatively reconcile the difference in flow value. After the KM value is adjusted then there is
reasonably good agreement between the two methods considering the differences in pitot tip,
instrumentation, time and space, calculation and test technique. Thus, further investigation is
required to verify the fundamental reasons and postulations for the differences between the two
traverse methods as described in this study. When the ultrasonic flow monitor (FM) calculation
and parameters are checked, there are certain inaccuracies, in particular the path length and the
transducer vertical off-set. Since this technique uses a calculated average stack temperature for
standard conditions correction, the path length inaccuracy is translated into temperature
inaccuracy; this in turn translates into error in temperature corrected volumetric flow in standard
cubic feet per hour (SCFH). Another correction is the stack pressure which is a measured value,
again any inaccuracy is translated into error in pressure corrected volumetric flow in SCFH. This
study emphasizes on the fundamentals of the physics and geometry of the stack flow monitors in a
circular stack, and the importance and sensitivity of the key ultrasonic flow monitor parameters to
the final temperature and pressure corrected volumetric flow. Thus, in view of these probable and
potential errors and inaccuracies, these flow monitor parameters should be thoroughly checked
for accuracy, then corrections to these parameters and other pertinent adjustments to flow based
on known phenomena, such as path velocity bias, should be implemented. After implementation
the calculated stack volumetric flow could be more accurate and representative than the reference
method due to the known problems being identified within the industry as well as in this report.
Introduction
With the introduction of stack flow monitors in the Houston Lighting and Power Company
(HL&P) solid fuel generating stations there is a need to understand the principles applied to the
ultrasonic flow monitor (FM) as well as stack flow tests with reference method (KM)1 and three
-------
dimensional probe (3D) method. These monitors have upstream and downstream sensors
mounted across the stack liner at one diameter only. The geometry and the physics are essential
to the accuracy of the calculation of the actual flue gas velocity, actual volumetric flow rate, gas
temperature and the correction to standard conditions which is the reported CEMS value. The
monitor is subjected to Relative Accuracy Test Audit (RATA) for the relative accuracy and bias.
The RM is defined in the federal register1 and relies on the S-type pitot tube as the inferential
velocity sensor which develops a differential pressure signal proportional to the point velocity.
This method produces a one dimensional velocity profile with respect to the point velocity.
Industry wide there are concerns regarding the accuracy of the reference method2. HL&P has
performed stack flow tests with 3D probe, a United Sensors DAT type probe tip at W. A. Parish
Unit 5 and 6 (WAP5 and WAP6). In order to obtain accurate measurement HL&P calibrated test
grade instrument plus an automated computer data acquisition system is used. This 3D method is
a single null method with the probe always nulled or balanced in the yaw angle. This method is
capable of resolving the test point velocity into the yaw and pitch vector. This study addresses
the results obtained by the two methods as well as the impact of velocity profile on the stack FM.
Stack Flow Traverse and Sonic Path
The impact of velocity profile on circular stack is briefly mentioned below for clarity; detailed
treatment can be found in other references3. The FM averages the stack flow linearly between the
two sensor tubes (typically 6 inches into the flow stream), thus the measured FM velocity is an
average path velocity (Up) as defined mathematically in Equation 1.
fu(r)-dr
R
(1)
As for the stack flow traverse with either the RM or 3D method, the test is conducted at the
centroid of equal area for the stack1. Thus, the flow traverse yields an average or mean velocity
for the cross sectional area at traverse plane. This better represents the true average velocity of
flue gas flow normal or perpendicular to the stack cross sectional plane. Mathematically the mean
velocity (Um) is defined in Equation 2.
_ fu(r)-dA
U = I- (2)
'mean i V /
A
The velocity profile function, u(r) is purely empirical and can only be derived from curve fitting of
test points. Some in the industry have used Prandtl's one-seventh power law to emulate this
velocity profile. HL&P has derived a velocity profile function based on actual test data using a
statistical fit program4. This empirical function is flatter and has a better fit than Prandtl's power
law.
-------
The velocity function is the quotient of two quadratic equations as follows:
TZ , ., , , (a+c*r+e*r2)
Velocity, u(r) ± '-
(l+b*r+d*r2)
where r is radius in inches, and a, b, c, d, and e are coefficients.
Equation 1 and equation 2 are equal only for the special case of a constant velocity profile,
otherwise the average path velocity is biased high compared to the mean velocity. Another
potential bias is that the flow monitor only measures the flue gas flow between the outlet planes of
the upstream and downstream extension tube, the average path velocity ignores the annulus from
the stack liner to the interface of extension tube and flue gas. With a known velocity profile (Ur),
one can calculate the positive bias due to the exclusion of this annulus. Using a mathematical
program5 one can numerically integrate the inner core, for example the WAP6 bias in path
velocity is 3.3%. Likewise, the total bias due to the actual sonic meter path compared to the
theoretical full cross sectional (area integration) volumetric average velocity is 7.6%. These
observations and postulations are purely based on known phenomena and mathematics, the
magnitude is dependent on actual test velocity function/profile for the unit. These are
independent of errors and inaccuracies of the parameters in the ultrasonic flow monitor.
Ultrasonic flow monitor
There are many sonic stack flow monitor parameters as shown in Figure 1 which are essential to
the accuracy of the velocity calculation as well as the volumetric flow. Also, there are parameters
important to the correction to standard gas conditions6 or SCFH.
Key parameters6 are:
D dimension - stack diameter
L dimension - actual path length, B-C
H dimension - transmitter off-set, B-C
Upstream pulse travel time in flue gas, tt
Downstream pulse travel time in flue gas, tj
Transit time between transmitter and end of extension tube, t^ or tta and distance A-B and
C-D
Gas constant, R' (derived from R, k and mole percent of gas constituents)
Sonic velocity, Cs
Stack gas temperature, T
Stack static pressure, P
while the fundamental equations are:
-------
Flue gas velocity, Up
2H
I '2
.Some velocity in gas, Cs —*
2 i
Flue gas temperature, T = R'(Cs)2
This study covers the key parameters and their effect on velocity, volumetric flow and corrected
SCFH. Specific unit examples are used to illustrate the fundamentals as well as to fortify the
postulations and underlying principles. Due to the potential + and - deviations/errors in these
parameters, the net effect on corrected flow to standard conditions can be + or -
L (pt B to pt C) is the
actual path traversed
by sound pulse
through flue gas.
Pomt A and Point D represent
the position of the transducers.
dt = distance between transducer
planes into gas stream
D = slack diameter
Elevation (pt A to pt D)
' is not the same as H or
flue gas flow
Note on key dimensions H and D:
Here L*2 does not equal to H*2 plus 0*2,
L"2 equals to H^2 plus dT2
Figure 1 Ultrasonic flow monitor dimensions and nomenclature
The key dimensions are obtained from HL&P drawings7, while flow monitor parameters are
obtained from the unit monitor printouts of the real time parameters and protected variables.
-------
Spreadsheets are used to effectively simulate the calculation in the flow monitor and to conduct
sensitivity analyses, the algorithm follows the line by line variables stated in the manual6
Results
Stack Flow Traverse
HL&P conducted stack flow traverses for WAP5 and WAP6 concurrently with reference method
tests conducted by the contractor at the same test ports. Due to the additional instrumentation
the 3D method requires approximately twice the time for a 3 by 4 port traverse compared to RM,
therefore, the values from the 3D and RM methods are not exactly the same time-wise and
spatial-wise. Nevertheless, based on the stack flow traverse the reference method results are
consistently higher than the 3D results. WAP 6 results from three flow rates show that the
average RM value is 14.5% higher than the average 3D values. Also the WAP 5 results show that
the average RM value is 12% higher than the 3D value. There are many studies *•3 addressing the
causes for bias from the S-type pilot tube. The current 12 to 14.5% high bias agrees with the
literature, part of the bias can be reconciled by the differences between the pilot tips and
calibration constant or flow coefficient213 as discussed below.
WAP6 - 690MW
WAP6 - 550MW
WAP6 - 320MW
WAP5 - 690MW
Wet flue gas flow, KSCFH
3D
100,954
87,365
60,932
102,567
Reference method
117,709
102,768
73,261
114,979
One bias is the effect of non-axial (yaw and pitch angle) component on S-type pilot calibration
coefficient Based on 3D resulls the flow is fully developed and uniform, the veclor is mainly
axial wilh zero yaw angle and small positive (+2 to +8 degrees) pitch angle. Based on RM resulls
Ihere is a 3 lo 4 degree angle of flow, Ihis is the yaw angle since the S-type can not resolve pitch
angle. From the literatures *3 a +5 lo +10 degree pilch can decrease Ihe S-lype pilot coefficient
by approximalely 2 to 5%, this amount varies with different sludies2. This would ihen
increase/bias Ihe velocity or flow by Ihe same amount Similarly Ihe effect of yaw angle has Ihe
same effecl and magnitude on pitol coefficient2. There is a discrepancy between 3D and RM in
lhal Ihere is no apparenl yaw angle wilh 3D while Ihere is 3 lo 4 degrees yaw wilh Ihe RM.
Another issue is Ihe need lo accounl for Ihe normal velocity vector by multiplying the raw velocity
by Ihe cosine of Ihe pilch angle. Based on literature review and lesl resulls Ihe combined effect of
pitch and yaw angle on pilot flow coefficienl and normal veclor could increase the measured
velocity by 6 to 10% over Ihe irue normal velocity crossing Ihe plane of iraverse. Therefore,
when the above known phenomena and corrections are applied then the adjusted RM value is
-------
closer to the 3D value, the adjusted RM value is still higher than the 3D value. The above is
based on the assumption of small yaw and pitch angle. The actual pitch angle with the reference
method is not known since it is not measured.
Under test conditions the pitch angle could be artificially introduced by the RM test probe if the
probe is not lined up horizontally2 A droop into the gas flow generates a positive pitch angle.
Likewise for a long probe traversing the furthest point from the port in a large diameter stack
could droop into the gas flow due to the gravity. The latter may not be controlled by the
operator. With the 3D probe tip the pitch angle can be determined through calibration curve.
Since the reference method can not resolve the pitch component, therefore, the RM is susceptible
to introducing positive bias in velocity. This is the result of 1) positive bias by positive pitch and
2) the normal component of the test point velocity is always less than the measured velocity since
cosine of + or - small angle is always less than unity. This potential bias can not be obtained or
determined with the current RM.
Thus far the known and potential bias resulting from S-type probe can account for some or most
of the bias depending on the actual S-type probe orientation and test technique, however, there is
one component of the 3D traverse results which could account for an additional 4 to 5%. This is
the velocity profile difference between the RM and the 3D method. The HL&P 3D profile in
figure 2 shows the smooth profile from zero (artificially added point for non-slip condition at the
stationary interface) to incrementally higher flow toward the centerline. However, the RM shows
a relatively flat velocity profile9 as shown in Figure 3. One could postulate certain phenomena: 1)
the RM is insensitive to near wall effect, 2) the S-type probe has inherent high bias at/near stack
liner, and 3) the manometer reading bias. The net effect is an inflated velocity profile by RM
when compared to 3D. This is also illustrated in Figure 4 for WAP58 The above is supported by
reference method results from other units, generally for all units the RM velocity profile is
relatively flat10- n
One issue related to the reference method is the limited number of equal area test points used for
the average stack velocity. In other words the method does not yield information on velocity near
the wall. Based on the WAP6 3D velocity fit and numerical area integration method,
mathematically one can demonstrate that the equal area test method has a + 3.3% bias in flow
when compared to true flow or area integration method. This applies to all equal area methods
with limited test points or limited knowledge of the near wall flow condition. The magnitude of
this bias can only be determined with accurate velocity profile.
Ultrasonic Flow Monitor Fundamentals
Based on recent data from six flow monitors, drawings and the knowledge gained from the stack
flow traverse, there are parameters with errors and deviations from the true value. Since certain
parameters rely on the accuracy of two or more variables, this does cause problems in that one
can not easily solve one equation with two or three unknowns. The latter can be estimated or
approximated with certain prior knowledge, known phenomena and postulations, this would
reduce the overall uncertainty or deviation. Nevertheless the key findings and postulations for the
system monitors are described below.
-------
Dimensions and geometry. There are significant errors in the L dimension and H dimension.
The L and H dimensions do not necessarily deviate in the same direction. One can easily verify
the derived values based on Pythagorean relation with the stored dimensions and stack drawings.
Figure 5 shows the sensitivity of L and H on velocity. The net effect on actual velocity is +8 to
+9% for three units and -4 to -5% for three units. One can speculate the reasons: a) errors in
input, b) wrong dimensions were supplied or calculated, and c) adjustments to the L dimension
were made during startup temperature 'calibration'. The latter is discussed in the next section.
For three units there are significant errors in the D dimension which lead to stack cross sectional
area error resulting in volumetric flow error. These units have tapered stack, the stored value in
the monitor is assumed to be derived from HL&P supplied information. In this study the mean
stack diameter between the upstream and downstream extension tube is used as the true cross
sectional area for the mean sonic path, this value is used for the volumetric flow error calculation.
For two units, one can speculate that the upstream (larger) diameter is being used, this results in a
larger area and a +4.8% bias in calculated volumetric flow. For one unit there is a negative 2.3%
bias, this is probably due to precision error in selecting the diameter.
The following illustrates the general sensitivity of key monitor parameters on other calculated
variables:
Parameter
L
L
H
D
Cs
T
P
dtort^
change
T
T
T
t
T
T
T
r
Affected variable
U,p (squared)
and Cs (linear)
Cs (linear)
U,p (linear)
Volume (linear,
velocity is not affected)
T (square root)
Corrected volume (linear)
Corrected volume (linear)
U,p and Cs (squared)
change
T
t
J
T
t
1
T
t
Sonic velocity, stack temperature and stack pressure. The L dimension directly
affects the calculated sonic velocity which is used hi the temperature equation. There is a stack
temperature 'calibration' procedure6 where the L variable is changed so as to match the test
temperature. By changing L, one effectively changes the calculated path velocity also. Thus the
real reason for the temperature error should be thoroughly investigated. The most common
reason to change the L variable is due to the inaccuracy of the L dimension and/or the location of
the transducers inside the extension tube, the A-B and C-D distances. One has to be absolutely
-------
sure that there are errors in these dimensions and the test temperature measurement is accurate,
otherwise one can introduce another unknown into the temperature equation as explained below.
Based on the real time parameters, one can infer that four out of the six units have been adjusted
for the above reason.
One should follow up after the temperature 'calibration' by checking the reasonableness of the
new L with geometry, plus verify H and A-B and C-D dimensions through gas velocity and sonic
velocity calculation and convergence. The reason is that these distances can introduce error in
the temperature calculation through an assumed transit time. Without adjusting the transit time,
one inadvertently 'changes' both the upstream and downstream travel time, t, and t2, since the
transit time remains the same in the program. This leads to path velocity error even though the
indicated temperature has been 'calibrated' and 'corrected'. One approach is to measure the L
dimension and the A-B and C-D dimensions, when there is reason to change A-B and C-D
dimensions, then change the transit time in the program to reflect the increased or decreased
transit time due to the change in distance. The above is the correct way to 'calibrate' the
temperature based on the fundamental equations because when the dimensions are accurate the
physics and mathematics should produce the correct temperature. One caution in changing the
transit time is that the program may only allow multiples of preset transit time. Further
calculation shows that at high load this preset transit time (times 2 tubes) is a significant
percentage, 7.45% compared to the average travel time of the sonic pulse. This coded and
quantized number should be adjusted to accommodate the actual distance without introducing
artificial (least count) error in the flue gas path velocity and sonic velocity calculation.
The coded 'gas constant'/coefficient, R' used in the temperature calculation is basically a
constant. A derived 'gas constant' should be a function of temperature, specific heat ratio and
mass/mole fraction of flue gas constituents. Thus, the R' has a slight positive error in the order of
1%, this in turn translates into +1% error in stack temperature. Figure 6 shows the effect of stack
temperature on flow due to temperature correction.
The stack pressure is derived from elevation adjustment to barometric pressure, the barometric
pressure is measured by a pressure transducer. Upon examination of the real time values for the
six units, there is a significant spread in stack pressure value compared to test values. The units
with low stack pressure have pressure correction bias of-2 to -8% while the units with high stack
pressure have bias of+1 to +2%, see Figure 6 for the effect of stack pressure on flow correction.
Thus the pressure transmitter should be periodically calibrated and the reasonableness of the
adjustment to elevation should be verified.
Conclusion and discussion
Based on the WAP 5 and 6 results the reference method yields higher flow than the 3D method.
This observation is consistent with literature. The probable causes for the higher velocity are 1)
pitch and yaw angle effect on flow coefficient, 2) cosine of pitch and yaw angle, and 3) flat
velocity profile (higher velocity near wall). The physics related to these phenomena require
further investigation. The true differences between the reference method and the 3D method can
-------
only be defined through thorough understanding of the individual and combined effect of these
phenomena. The actual velocity profile and velocity vector is essential to the understanding of
these phenomena as well as for determining bias due to limited equal area points with reference
method vs. true area integration. The profile is also important in understanding the potential bias
from the sonic flow monitor. The reason is that the flow monitor only measures an average path
velocity by traversing sonically across the stack at one diameter at a 45 degree angle, thus flow
monitor does not represent area integration except for the special case of an extremely flat
velocity profile.
Based on the real time and stored parameters from the six flow monitors, stack drawings and the
knowledge gained from the stack flow traverse, there are apparent errors and deviations in key
parameters. The top contributing parameters are L, H and D for actual velocity and volumetric
flow, and stack pressure for pressure correction to standard conditions. Based on the single
parameter effects, there is a range of single effect values from -5% error to +9% error. Thus,
when the errors are in one direction then the final error could be very significant. Therefore, the
key parameters and dimensions should be verified in order to have confidence and certainty in the
final corrected volumetric flow calculated by the flow monitor.
In summary the aforementioned results, explanation and postulations on the error and deviation
could serve as a guide in investigating, verifying and correcting these key parameters. Figure 7
summarizes the type of biases mentioned in the results section. After the flow monitors have been
thoroughly checked and the parameters are implemented, then the path velocity adjustment based
on actual velocity profile should be implemented. The resultant volumetric flow should be
accurate for emission and heat rate calculation. After implementation the calculated stack
volumetric flow could be more accurate and representative than the reference method due to the
known problems identified within the industry as well as in this report. Finally, this flow value
together with the CO2 value can be used to back calculate the fuel heating value and fuel quality12
which can be useful for units burning fuel with variable quality.
Acknowledgment
Author appreciates the efforts of Performance Analysis personnel who conducted these stack
traverses.
References {Any mention of brand name does not represent endorsement.}
1. Code of Federal Regulations, applicable sections on Stack gas velocity and volumetric
flew rate (Type-Spilot tube).
2. Muzio, L. I, Martz, T. D., McRanie, R. D. and Norfleet, S. K., Flue Gas Flow Rate
Measurement Errors, Electric Power Research Institute (EPRI) Report, TR-106698, June
1996.
3. Baker, S. S., Ettema, R J. And Martz, T. D., Guidelines for Flue GasFlowRate
Monitoring, EPRI Report, TR-104527, June 1995.
4. TableCurve, a curve fitting program by Jandel Scientific.
-------
5. MathcadPLUS 6.0, a mathematics program by MathSoft, Inc.
6. Monitor Manual - Ultra/low Model 100, United Sciences, Inc.
7. HL&P - As-built drawings of the CEMS ultrasonic meters.
8. W. A. Parish Unit 5 stack CEMSRATA report, METCO, October 1996.
9. W.A. Parish Unit 6 stack CEMS RA TA report, METCO, October 1996.
10. W.A. Parish Unit 7 stack CEMS RATA report, METCO, October 1996.
11. W.A. Parish Unit 8 stack CEMS RA TA report, METCO, October 1996.
12. Munukutia, S. and Khodabakhsh, F., Enhancement of boiler performance evaluation
methods using CEMS data, American Society of Mechanical Engineers Paper No. PWR-
Vol. 29, 1995 Joint Power Generation Conference, 1995.
-------
W. A. Parish Unit 6 - Oct 1996
3D probe stack flow comparison
144 1*2
Indie* From One Wall Of Stack
* 3D-690-avg-2diam.
• Ref method, avg@690MW
- 3D-S50-2diam-avg
' Ref method, avg@S501fW
"— 3£>-320-2diam-avg
B - Ref method, avg@325MW
Figure 2. WAP 6 - average traverse velocity - 3D vs. Reference Method
W. A. Parish Unit 6 - Oct 96@690 MW
METCO - S-type - normalized velocity
144 192
Inches From One Wall Of Stack
r4-B*D-Vjiotmld ••
! • r4-AiC-Vjioniild
-• r7-B4D-Vjii>nnld
Figure 3. WAP 6 - Example of reference method traverse data points from three runs.
-------
W. A. Parish Unit 5 - Oct 1996
3D probe stack flow comparison
at 1*4 192
Inch** From On* Wmll or Stic*
3D-«vg@690MW -X- Ref method, »vg@690MW — «- - R«f method, «vg@550MW -B) - Rcf method, avg@320MW |
Figure 4. WAP 5 - average traverse velocity - 3D vs. Reference Method.
Solid fuel - stack flow sensitivity
to dimensional error
-20%--
-24K M I I i I I I I I I I I I I I I I I I I I I I I I I I I I | I | I | I | I | I | I | I | I |
-12% -10% -8% -6% -4% -2% 0% 2% 4% 6% 8% 10% 12%
Percent error in dimension
L-error on Vel/Flow""°"~ H-error on Vel/Flov
Figure 5. Ultrasonic flow meter - Sensitivity of L and H dimension on calculated velocity.
-------
Solid fuel - stack flow sensitivity
to stack T and P error
-10 -'« -»« -14 -»2 -'<" -•-«-« -2 0 Z 4 8 • 10 12 14 1» 11 20
* temp effect on std flow calculation
* stack P effect on std flow calculation
Figure 6. Sensitivity of Corrected volumetric flow to Stack temperature and Stack pressure
20%
0%
Example - W. A. Parish Unit 5
Flow test BIAS and Monitor BIAS
Row test
Monitor
Jill core flow/wall effect biu E±J 3D vs. S-type
II] monitor errors fUJi path vs. area integration
stack flo\*|
Figure 7 Sources of Flow test bias and Flow monitor bias
-------
OVERVIEW OF DEVELOPING TECHNOLOGIES FOR
CONTINUOUS EMISSION MONITORING
James A. Jahnke, Ph.D.
Source Technology Associates
P.O. Box 12609
Research Triangle Park, NC 27709
Abstract
Continuous emission monitoring (CEM) systems have evolved considerably since
their early applications in the 1970s. However, the increasingly competitive nature
of the CEM field has driven a search for new monitoring platforms and the
application of new analytical technologies. In particularly, developments in sensor
and microsensor technology appear promising for reducing both the size and costs
of continuous monitoring systems.
History
Continuous emission monitoring (CEM) systems have undergone continual
development since their first application to electric power stations in the 1970s.
Extractive systems using existing process analyzers, were among the first to be
installed. In-siru systems, monitoring the flue gas without extraction, followed as
an attempt to avoid the problems associated with gas extraction and conditioning.
The next wave of development, resulting from a combination of the European
designed dilution probe with American developed ambient air luminescence
analyzers, exhibited considerable success in the Part 75 Acid Rain CEM program.
As "platforms" for gas analysis, CEM systems have improved since the 1970s due
to both the requirements of the regulatory agencies and the demands of users.
Analyzers have also exhibited considerable development since their 1970
applications. With the advent of digital electronics and microprocessor systems,
analyzers have become the most reliable part of the CEM system. In the late 1980s,
microprocessor controlled analyzers with interactive menu systems, were
introduced by a number of manufacturers. Today's analyzers can output digital
signals, perform internal diagnostic functions, perform internal corrections and
provide data backup.
-------
Analytical techniques used in CEM system analyzers have remained much the same
since the 1970s. The electro-optical methods of differential absorption spectroscopy,
gas filter correlation spectroscopy, and luminescence are used widely.
Electrocatalytic, paramagnetic, and electrochemical techniques are also applied
routinely. However, with the increasing concerns and requirements for the
measurement of "air toxics," newer methods are making the transition from
traditional laboratory application to field use. Ion-mobility spectrometry, the use
of diode lasers, photoacoustic spectrometry, and the field application of Fourier
Transform Infrared Spectrometry (FUR) and mass spectrometry are now being
incorporated in CEM systems for air toxics monitoring.
In the background however, other analytical approaches are being considered for
a wider range of monitoring applications. Principal among these are the use of
"sensors/' either small, miniature, or micro devices that respond to varying
concentrations of flue gas components. The first of these sensors used in CEM
applications was the electrocatalytic zirconium oxide sensor for monitoring flue
gas oxygen. Catalytic sensors for monitoring NOx and CO are also available
commercially today. Metal oxide sensors, surface acoustic wave (SAW) devices,
and fiber optic techniques show potential for environmental monitoring
application.
But further in the future, one can expect the development of "micro-sensors" for
environmental monitoring applications. Spurred by research programs of the
Department of Defense and the Department of Energy, numerous programs are
underway to miniaturize sensing systems by using the techniques of
microelectronics. This field of "microelectromechanical systems" (MEMS) offers
the potential of miniaturizing spectrometer systems, designing arrays of multiple
sensors, and incorporating sensors into integrated circuits for the development of
miniature analytical systems. However, whether the potential of microsensing
will be realized in the continuous emission monitoring systems will depend more
on market forces than on the potential of the technology itself.
Continuous Monitoring Approaches
Today, there are basically six different approaches to monitoring flue gas emissions
on a continuous basis (Figure 1).
-------
Approaches to Continuous Monitoring
Parameter
Predictive
— Ex-situ
Close-coupled
Figure 1
Approaches to Monitoring Flue Gas Emissions
The most traditional technique has been to extract a sample of the flue gas from the
stack and analyze it on either a wet or dry basis. In-situ systems avoid problems of
corrosion and non-representativeness often associated with extractive systems by
monitoring the gas in the stack, without extraction. Here, analysis is performed by
interaction of light with the gas constituent or gas interaction with some type of
sensor inserted into the stack.
Remote sensing conceptually avoids problems of both extractive and in-situ
methods by monitoring emissions at the stack gas exit. An active laser beam or
absorption of UV solar radiation by the flue gas are typical analytical approaches in
this technique. However, accuracy problems associated with plume dispersion have
limited this method for regulatory applications.
Two inferential methods that are beginning to see wider application are parameter
monitoring and predictive monitoring. In both of these methods, "sensors" are
used to provide data that can either be used as surrogate for emissions (parameter
monitoring) or to predict the emissions from some type of model. The predictive
method has adopted the acronym, "predictive emissions monitoring" (PEM).
Sensors used in either of these methods are not typically gas sensors, but sensors
that monitor process operations such as temperature, pressure, fuel and gas flow
rates.
-------
Approaches to Analytical Instrumentation
In this increasingly competitive field of emissions monitoring, CEM system
integrators and instrument manufacturers are seeking to develop lower-cost
systems. As one example, to monitor multiple gases, traditional approaches that
"gang'' analyzers are being re-examined. This has led to a number of analytical
options, which are categorized in Figure 2.
Figure 2
Analytical Approaches to Multi-gas Analysis.
The traditional techniques are those that have been used in monitoring systems
since the 1970s. The developing methods listed are not strictly "new," but are being
applied more frequently today in order to gain competitive advantage. The
developing methods seek to gain this advantage by reducing: 1) the number of
system components, and 2) the cost of the system components. Although one
would expect that reduced component costs should lead to reduced system costs, this
has not always been the case. CEM system prices are based on what the market will
bear and subsequently have little relationship to component costs. However, the
increase in the number of CEM systems vendors and integrators because of the
utility Acid Rain CEM market and the market's subsequent collapse have led to
increased competitiveness and increased interest in the "developing" strategies
-------
listed in Figure 2. As a result, monitoring systems are available today (1997) in the
price range of $50,000, compared to $150,000 or more, demanded several years ago.
New Sensor Technology for CEM Applications
Strictly defined, a "sensor" is a device that receives a signal or stimulus and
responds with an electrical signal. In general, it responds to a nonelectrical value
(such as a concentration of gas molecules) and converts it into and electrical value.
In this definition, the optical/detector assembly of a differential absorption system,
a flame ionization detector of a total hydrocarbon analyzer, or an electrochemical
cell of a combustion analyzer could all be characterized as sensors. The typical
pollutant analyzer, however, is not thought of as a sensor, but as an integrated
assembly of component parts.
In a narrower, more typical view, a "sensor'' is a single device, or a tightly integrated
assembly whose components are, in general, not interchanged or repaired. It is a
device that is thrown-away upon failure, is easily replaced, and is relatively
inexpensive. Sensors and the smaller microsensors lend themselves for application
in the internal multi-gas modular analyzers due to their small size and
intechangeability.
Rapid advances are being made in the design of miniature and micro gas sensors for
both domestic and military applications. As sensor-based technology develops, it is
anticipated that CEM systems will adapt the technology when cost savings can be
realized without sacrificing monitoring accuracy and precision.
Several approaches are being taken in sensor development programs. These can be
characterized into two groups, 1) the development of miniature sensors (centimeter-
scale) for proof of concept, and 2) the further miniaturization to a micro-level
(10"6 cm) scale, where miniaturization is accomplished by applying microelectronic
and micro-machining techniques developed in the semiconductor electronics
industry. The versatility of these techniques has led to the new field of
"microelectromechanical systems" or MEMS, which extends the planar, two-
dimensional techniques of microelectronics into the third dimension. By using
a variety of deposition, lithographic, and etching techniques, valves, cells, free-
standing structures, etc. can be "micro-machined."
Figure 3 illustrates approaches being taken in new sensor development:
-------
Micro-machined
Devices
Analytical Approaches
Figure 3
Approaches in the Development of Sensor Technology for Monitoring Gases
Two primary forces drive the development of miniature and micro chemical gas
sensors. One is the development of low cost commercial products that can be used
to monitor gases (such as CO and CO2) in the home, workplace, and automobiles.
The other is the interest in developing low cost sensors to monitor chemical warfare
agents. Private sector research programs are funded primarily by companies that
have a commercial stake in these areas. Military sensor development is funded by
the government to national research laboratories, defense contractors, and
university laboratories.
Most of the developing techniques are targeted for monitoring gases present in
ambient air. Although low cost miniature gas sensors and microsensors may
become available commercially in the near future, they still face the formidable
challenge of monitoring specific analytes in a complex and varying flue gas in CEM
applications. Sample delivery systems used in conjunction with sensor-based CEM
systems will be as crucial to system performance as they are to current CEM system
designs. Since, economies associated with the mini- and micro-scale can be realized
in system design it is expected that developments in miniaturization should lead to
lower cost CEM systems.
-------
Outlook
A trend can be seen from discrete analyzers to multi-gas modular systems and
multi-gas analytical systems, particularly in the case of monitoring air toxics.
Coincident with the multi-gas measurement approach is the decreasing size of the
analytical instrumentation from bread-box sized analyzers to mini-integrated
systems, to microsensors.
A driver in many of these developments has been an attempt to reduce equipment
costs in face of an extremely competitive market. Another driver is the continuing
search for a proprietary product, one that is uniquely competitive, but in addition,
provides operational advantages over other systems.
Although there are many analytical methods (such as SAW, optical fiber,
microcalorimetric, and other methods) that show potential in emissions
monitoring applications, those chosen for evaluation and /or further development
must meet a large number of hurdles before they can fit into the CEMS market.
Its a large step from the laboratory to the stack installation, where measuring a
corrosive gas in an instrument-hostile ambient atmosphere can be extremely
difficult. The factors that any new analytical technique (whether it is macro- or
micro- size) must address are the following (after Clifford, 1996):
• Sensitivity
• Susceptibility to interferences
• Susceptibility to poisons
• Response time
• Linear, nonlinear, or integrating response
• Temperature and humidity dependence
• Reversibility, recoverability
• Memory and hysteresis effects
• Stability, decomposition
• Lifetime
• Sensor reproducibility
• Sensor comparability
Unfortunately, many of chemical microsensors are weak in many of the areas noted
above. Interferences, poisoning, slow response times, recovery times, reversibility,
may not be a problem for proof of concept studies in a laboratory but frequently
become issues that are difficult to resolve on a practical level.
Most of the current research and development work being done in microfabrication
is focused on future large volume sales to either residential, automotive, or military
markets. Its expected that CEM system technology should benefit from microsensor
developments in these other application areas.
-------
It should not be assumed that microsensors or micro-analytical systems will
necessarily reduce total CEM system costs. Even if a sensor can be manufactured for
$5 or $20, it is merely one component in a system that must meet both operational
challenges and regulatory requirements. The analytical system costs are
approximately 30 to 50% of the total system costs, where data acquisition and shelter
costs can be comparable to the analyzer costs. The one area that has not realized cost
reductions with technological development is the CEM data acquisition system.
The plethora of regulatory reporting requirements, (State, Federal, and local) has
hindered any standardization here, resulting in the necessity for custom
programming in most applications. Users have also contributed to this problem
due to their requests for customized features for specific plant applications.
Ancillary costs also contribute to a system's total cost. Such costs may include:
the gas calibration system, technical specification preparation and bid evaluation,
installation, platforms, elevators, certification, QA plan preparation, and
engineering overhead. These ancillary costs are in general, unavoidable.
With the onset of micro-analytical systems, our current perceptions of what a CEM
system is and does may have to change. The traditional extractive system with
heated, air conditioned shelters protecting extractive, analytical, and data acquisition
subsystems may no longer hold. The path for future CEM development is clear:
smaller, less costly, better. Its not inconceivable that the functions of each of the
CEM subsystems could be incorporated unto a single chip. The timeline, however
is less clear. While many promising technologies are being investigated many
hurdles to commercialization remain. One thing is certain, we are on the verge of
a revolution in environmental sensor technology that will challenge our existing
monitoring concepts. The future of CEM technology will not be merely an
extension and refinement of the past, but will be a major shift in adaptation to
new technologies.
References
Sze, S.M. 1994. Semiconduct Sensors. John Wiley & Sons, Inc., New York, NY.
Clifford, P.K. and Dorman, P. 1996. Test Protocols for Residential Carbon Monoxide
Alarms - Phase I (draft). Gas Research Institute - Environment and Safety Research
Group. GRI-96/0055.
-------
Development and Field Testing of a Continuous Real-Time, Speciating Mercury
Analyzer
Ms. Sharon Sjostrom
Dr. Daryl L. Roberts
Mr. Gary Anderson
Mr. Frank Sagan
Mr. Justin Smith
ADA Technologies, Inc.
304 Inverness Way South, Suite 365
Englewood, CO 80112
Abstract
The mercury concentration in utility flue gas from coal-fired boilers is in the range of 0.1 to 1 part per
billion. Wet chemical and solid sorbent measurement methods are available that can provide reasonable
time-average measurements. However, a method that responds in real-time to mercury perturbations in
the flue gas stream, can differentiate between elemental mercury and mercury compounds, is not affected
by varying levels of SC>2, and that operates at typical stack temperatures to avoid sampling problems,
would be a useful alternative to manual methods.
ADA Technologies has been developing such a continuous mercury analyzer for the past few years. The
analyzer is based on absorption of ultraviolet light emitted by a mercury lamp. The lamp itself resides in
a permanent magnetic field, creating two wavelengths of source light that are polarized 90° with respect
to each other (Zeeman Effect). Mercury extinguishes only one of these wavelengths, and common
interferents such as SOi extinguish both, allowing for on-line cancellation of the interferent species. The
system consists of two sample cells each with sensitive photodetectors, a mercury species converter, and
a calibration system. Elemental mercury is detected in the first sample cell. Sample gas exiting the first
cell is passed through a "converter" to change speciated mercury compounds to elemental mercury that
is then measured in a second cell (total mercury). Signals from both detectors are continually processed.
The analyzer is calibrated regularly with known concentrations of mercury, and the operation is verified
periodically by comparing to a manual mercury measurement method.
Extensive laboratory evaluation and field testing has recently been undertaken to optimize the
performance of the analyzer. New developments in the calibration technique, sampling system,
measurement, and signal processing have resulted in a system capable of measuring mercury
concentrations on a slipstream from a coal-fired power plant in real time (0.1 to 0.8 ppb mercury). The
system is configured to automatically calibrate on a regular basis. We have found that we can also
measure the SO2 concentration because SO2 absorbs UV light at the wavelength of the mercury lamp
emission. With the Zeeman splitting of the light source, the SOi concentration does not affect the
mercury signal except when the SO? varies more than 1000 ppm between calibrations. Descriptions of
these recent developments and resulting field test data are included in this paper.
-------
Introduction
Continuous monitoring of mercury emissions from combustion sources may be required by regulatory
agencies in the near future. In addition, the effectiveness of research efforts aimed at mercury control
technologies can be greatly enhanced with a continuous mercury monitor. For these reasons, ADA
Technologies has been developing a continuous monitor for mercury found in flue gases and other
combustion gases. The monitor has now undergone four months of testing at a coal-fired pilot facility
and six months of field testing at a coal-fired power plant where the mercury concentrations have been in
the range of 0.1 ppb to 0.8 ppb. The analyzer has also been configured to monitor SC>2 while recording
total and elemental mercury concentrations. This paper describes the fundamentals behind the operation
of the analyzer, the techniques used for calibration, non-ideal factors that influence operation, and an
example of data collected from a coal-combustion flue gas stream.
Mercury Analyzer Fundamentals
Elemental mercury both emits and absorbs UV light in a very narrow wavelength range centered around
2537 A. Therefore a mercury lamp is the ideal source of light to pass into a sample cell that contains an
unknown concentration of mercury. When elemental mercury vapor is in the light path between the
lamp and a light detector, the mercury vapor absorbs light in the wavelength emitted by the lamp. The
extinction of light gives a direct measure of the concentration of mercury through the application of
Beer's law as has been practiced for "cold vapor atomic absorption" analyses for decades. The difficulty
in applying this technique to flue gases is the presence of species, such as SO2, that also absorb UV light
in the same region. Since SO2 is often present in concentrations that are one million times that of
mercury, the extinction of light by SO2 swamps that caused by the mercury, rendering conventional UV
absorption spectroscopy useless.
ADA's approach to cancelling the influence of such interferents in real time is based on the Zeeman
effect on the UV emission spectrum of the mercury lamp (Figure 1). The mercury lamp shown in the
figure emits light at 2537 A with the subpeaks of several isotopes closely grouped. If the lamp is placed
in a magnetic field, the spins of outer shell electrons will align, causing the emission profile to change.
The new profile adds two "winglets" spaced a fraction of an angstrom on either side of the main
emission grouping and orthogonally polarized to the main emission grouping.
The absorption profile of mercury is superimposed over the emission spectrum in Figure 1 (100% on this
figure means 100% transmission and zero absorption). Elemental mercury vapor absorbs the n
wavelength (the main emission grouping) but does not significantly absorb the CT+ or a" wavelengths (the
"winglet" emissions). In contrast, SO2 absorbs the n and CT wavelengths almost equally. Application of
Beer's law to both wavelengths (jr and CT) yields two equations that are solved simultaneously for the
mercury concentration and the SO2 concentration. Because the molar absorptivity of elemental mercury
for the n wavelength exceeds that of SO2 by a factor of about 1,000, the mercury concentration can be
determined even though the SO2 is present at concentrations one million times that of mercury.
To exploit this principle, it is necessary to introduce alternatively the n and then the a wavelength to the
sample cells. The "switch" that chooses which wavelength enters the sample cell is a half-order
waveplate rotating at approximately five rotations per second. The light coming through this waveplate
follows the equation I = Iacos229 + I^sin^G , where 6 is the angle of rotation of the waveplate. In
one rotation of the waveplate, the light entering the sample cell has four peaks and four valleys
corresponding to the intensity of either Tt-light or a-light emitted by the mercury lamp.
-------
Figure 2 presents the photodetector response with zero gas, mercury span gas, and SO2 present in the
measurement cell. The photodetector response follows the sinusoidal equation of the waveplate. It does
not matter if Ia is greater than or less than 1^, but for illustration purposes, Figure 2 shows a situation
where Ia is greater than 1^. The first third of the figure represents the response with zero gas in the
measurement cell. When mercury is in the measurement cell, the intensity of the n-light (main emission
grouping) reaching the photodetector is decreased, but the intensity of the a-light reaching the
photodetector is essentially unchanged. Thus, the difference between the maximum and minimum
intensities is greater, and the average light intensity at the photodetector is slightly lower. The extinction
of the slight is related by Beer's law to the concentration of mercury. For the purposes of discussion,
the difference between maximum and minimum light intensities can be considered as the "mercury
signal strength." In the presence of 862, both jt and o polarizations are attenuated, decreasing the
mercury signal strength and the average light intensity.
Figure 3 is a graph of mercury signal strength versus average light intensity for concentrations of SO2
from zero to 4500 ppm measured in the field during an SC>2 calibration. The mercury signal strength is
seen to be a linear function of average light intensity for this broad band absorber. For the calibration
shown in the figure, a least squares linear fit was applied to the data. The r2 for this data set was 0.9991.
Because the correlation is very strong in the range expected at coal-fired utilities, the analyzer can report
the SC>2 concentration simultaneously with the concentrations of elemental and total mercury.
Figure 4 is a graph of mercury signal strength plotted against average light intensity for varying
concentrations of mercury. Four concentrations of mercury were passed through the measurement cell
from the calibration system with and without SO2. The lines representing each mercury concentration
have nearly identical slopes and are spaced linearly with mercury concentration. Again, the r2 from a
least squares linear fit is quite high. Therefore, when sampling conditions typical of coal-combustion
flue-gas streams (0-5000 ppm SC>2 and less than 20ug/m3 Hg), simple linear interpolation can be used to
calculate SC>2 and mercury concentrations.
Figures 3 and 4 define an algorithm for this analyzer to calculate mercury concentration. This
calculation is graphically summarized in Figure 5. A three point calibration (zero gas, SO2 span gas,
mercury span gas) is used to define the slope of the SO2 line and the spacing of the mercury lines. These
points are identified in Figure 5 and represent the points referred to in Figure 2. The SC>2 concentration is
calculated by determining the distance the sample point is from the zero SO2 line. This line is defined
by A^ and Aspan on Figure 5. The mercury concentration is proportional to the distance the sampling
point is from the zero mercury line. This line is defined by A^K, and Aso2 on Figure 5
The calibration and data manipulation procedure described above has been used in the laboratory and
one analyzer operating at a coal-fired power plant. The graph in Figure 6 shows a trace of the mercury
measurements and SO2 measurements reported by the analyzer during a field calibration. SO2
concentrations from zero to 1500 ppm were introduced into the system in the absence of mercury to
verify that the mercury concentration reported by the analyzer was not affected by changes in SO2. Total
and elemental mercury traces on the plot (scale on left) do not change significantly as a result of changes
in the SO2 concentration. Mercury span gas was introduced into the analyzer at 7.08 hours on Figure 6
while maintaining the SO2 concentration. The SO2 concentration reported by the analyzer was not
affected. These data suggest two key points: 1) the Zeeman effect permits successful cancellation of
S02 during mercury measurements, and 2) since SO2 absorbs light in the wavelength region being
monitored, the analyzer can report both the SC>2 concentration and the mercury concentration in real
time.
-------
The absorption profile of elemental mercury is shown in Figure 1. Other forms of vapor-phase mercury
do not absorb light in the same region. To obtain total mercury measurements, all mercury in the sample
gas is passed through a region which converts all forms of mercury to elemental mercury. Calibrations
have been conducted using various concentrations of HgCb as the non-elemental species. When HgCl2
is introduced into the analyzer, insignificant mercury is reported as elemental mercury and the expected
calibration concentration is reported as total mercury by the analyzer. Other forms of mercury have not
been introduced to the analyzer during total mercury calibration checks; however, thermochemical
decomposition of non-elemental forms of mercury is guaranteed at the conditions of the converter.
Noise, temperature and other non-ideal effects
The theory behind the analyzer operation is well known. However, when this theory is applied to an
instrument designed to measure mercury concentrations less than 1 ppb, engineering details become
tremendously important. The effect of changes in the temperature of the environment seems to be the
most critical in the prototype ADA instrument.
The analyzer at the coal-fired power plant was placed in a "temperature-controlled trailer" in anticipation
that this would stabilize the signals, reduce the noise and improve resolution of the instrument. Figure 7
shows changes in the mercury signal strength, average light intensity, detector temperature and mercury
lamp temperature over a continuous 40 hour period. The sawtooth pattern in the detector temperature
profile from hours 25 to 35 and 45 to 60 reflect temperature fluctuations caused by a room air
conditioner cycling on and off during the day. The air conditioner was off at night. The signal strength
and average light profiles follow the temperature profile. These fluctuations represent 10 to 100 ug/Nm3
swings in the calculated mercury concentration. These data were collected while sampling actual flue
gas from a coal-fired utility and the true fluctuations in the actual mercury concentration were probably
less than 3 |ig/Nm3.
In an attempt to minimize the temperature fluctuations, the air conditioner is off during analyzer
operation. In addition, the lamp temperature and the detector temperature controlled independently.
This decreases the frequency of room temperature cycling to 24 hours and holds two critical components
of the system relatively stable. Diurnal effects are still noted and these are believed to be related to
changes in temperature.
Periodic calibrations are employed to minimize the drift of the reported mercury measurements and
mathematical corrections for temperature drift over time are used. Following each calibration the
average light and mercury signal strength are compared with the previous zero values. The real-time
measurements are then corrected to the slope of the drift curve. This is not ideal because the slope of the
drift line over the previous drift period is used to predict the subsequent drift slope. However, the
technique gives a reasonable representation if instrument temperatures are not changing quickly.
Fortunately, the mercury algorithm is simple and can be applied to data after testing if needed, allowing
reasonable post-test temperature drift corrections.
Attempts have been made to calculate the mercury signal strength and average light intensity as a
function of temperature to mathematically correct for temperature in real-time. However, it has been
difficult to select which temperature to use in the calculations. When the detector temperature is held
constant and the lamp temperature is allowed to drift, the signals drift. This would indicate a change in
light intensity from the lamp. If the lamp temperature is held constant and the detector is allowed to
drift, the signals still indicate some drift. This would imply a change in the responsivity of the detector.
When attempts are made to hold both constant, slight fluctuations in temperature remain due to null
bands in the temperature controllers and a temperature effect is still noted. Figure 8 presents three plots
-------
including short term drift of the average light intensity and mercury signal strength for hours 37-47 from
Figure 7, and the corresponding trace of lamp temperature versus time for this period. Three time
periods are highlighted: a constant lamp temperature segment from hour 37 to 38.2 (light triangles), and
increasing temperature period from hour 38.2 to hour 42.3 (gray squares), and a decreasing temperature
period from hour 42.3 to hour 45.5 (black diamonds). The average light and mercury signals in the top
two graphs show a hysteresis associated with increasing or decreasing temperature. This characteristic
of the analyzer needs to be examined further before a real-time mathematical correction can be
developed.
Analyzer Sampling and Calibration System
A critical component in assuring that an analyzer is capable of providing meaningful mercury
measurements is to confirm that a representative gas sample is passed through the analyzer. Achieveing
reasonable quantitative certainty on this issue with sub-ppb levels of mercury is, however, an
engineering challenge, and studies are continuing to determine the optimal material and gas flow rates
required in a transport assembly to bring flue gas from the duct to the analyzer. A system was fabricated
with a special sampling filter, valve boxes and controls to allow automatic calibration through the
sampling lines and sampling filter and to allow automatic switching between inlet and outlet sampling
lines.
A sketch of the stack sampling filter is shown in Figure 9. An oversized buttonhook nozzle turned out
of the flow (non-isokinetic sampling) is used for one program to minimize the particulate included in the
flue gas sample. This is followed by a ceramic or glass filter held at 300 °F to remove any remaining
particulate at a temperature where little adsorption of vapor-phase mercury is expected. It is possible to
calibrate through the filter to determine if any mercury is being adsorbed by particulate on the filter or by
the filter material itself, or to calibrate without passing through the filter. A second option would have
been to install a cyclone followed by a filter. When sampling in high ash flue gas streams, removing the
flyash from the gas stream without affecting the mercury measurement may not be possible with
conventional methods. Some flyash may adsorb or desorb mercury if the temperature of the ash is
altered or if the flue gas is passed through the filter. Other ashes may change the speciation of the
mercury passing though the filter. If a total mercury measurement is adequate, the filter temperature can
be elevated until all mercury is desorbed from the ash. It is quite possible that the fractional mercury
speciation may not be maintained if this technique is used.
The sampling lines used for this program are 1/4" diameter heated Teflon® tubes. Teflon® is used for
mercury permeation tubes and it would follow that Teflon® may absorb mercury during sampling.
However, it is hoped the Teflon® equilibrates after some period time of constant flow and that the
sampling lines adequately transport the flue gas to the analyzer without changing the mercury
concentration in the gas. Two additional 1/4" Teflon® lines are heat-traced with the sample lines to
transport calibration gas from the analyzer to the sampling filter. One line is exposed only to zero gas,
and the second line is exposed to mercury span gas. Calibrations through the sampling hot-lines are
referred to as "overboard" calibrations ("onboard" for a calibration directly to the analyzer). The
confidence in the ability of the hot-line to transport mercury laden flue gas is enhanced when an
overboard calibration closely matches an onboard calibration, which is the case for the tests at the
current field site with elemental mercury calibrations.
Non-elemental mercury may be more likely to "stick" to the Teflon® lines. Early sampling data (see
Table 1) indicates that the flue gas from the test location contains mostly elemental mercury, thus
calibrations at this location were conducted only with elemental mercury. Teflon® may pose problems
for other locations with higher fractions of non-elemental mercury.
-------
Analyzer Operation at a Coal-Fired Pilot Facility
An upgraded version of the analyzer was used for four months at a pilot coal combustion facility
(Consol, Inc.; Library, PA) under a DOE-sponsored project aimed at assessing the performance of a
novel mercury control technology (Roberts, et al., 1997). The analyzer was equipped with auto-zero and
auto-span features to help overcome some of the problems associated with drift. In Figure 10, we show
a typical sequence of zeroing, spanning, measuring the inlet to the control device (the native flue gas),
and the outlet from the control device. The span concentration is 16 (jg/m3. Adjusting for zero drift, this
figure shows that the total mercury in the flue gas is approximately seven ug/m3, that 3.5 ug/m3 of this
total is elemental mercury, and that all of the mercury is removed by the control device. This type of
output typifies how the speciating analyzer can be used to enhance the speed at which mercury process
development can be performed.
Discussion and Conclusions
A real-time continuous mercury analyzer capable measuring elemental and total total vapor-phase
mercury measurements is of interest both to those responsible for developing control technologies and
those responsible for setting regulations on mercury emissions. Our studies conducted in the past six
months indicate that the Zeeman effect and a mercury lamp can be used as the basis of an analyzer
capable of measuring sub-ppb concentrations of both total and elemental mercury in real-time while
monitoring the SO2 concentration in the sample gas stream. Unlike some mercury analyzers, this
analyzer accepts the gas at stack temperatures, minimizing sampling problems, and does not require
chemical reagents or sorbents that may be expended or fouled during routine operation.
The Zeeman analyzer has the potential to provide accurate mercury and SC>2 data. However, it is a
complex instrument, and there are several engineering design issues that must still be addressed to
improve the accuracy of the instrument in the 0.1 to 1 ppb mercury range. Several general observations
can be made based on field testing with the analyzer to-date:
1) The analyzer is capable of measuring < 1 ug/Nm3 mercury in the presence of SOi.
2) Concentrations of SO2 from 0 to 4500 ppm, the highest concentration tested, do not significantly
affect the mercury concentration reported by the analyzer.
3) The analyzer can be calibrated to measure and accurately report SO2 in real-time.
4) Several sampling concerns must be addressed in further investigations, such as removing paniculate
in high ash flue gas streams and choosing materials for sample transport lines.
References
Roberts, D. L., R. M. Stewart, T. E. Broderick, "Capturing and Recycling Part-per-Billion Levels of
Mercury Found in Flue Gases," presented at EPRI-DOE-EPA Combined Utility Air Pollutant Control
Symposium, Washington, DC, August 25-29, 1997
Acknowledgments
The authors would like to thank Mr. Tom Brown of DOE and Dr. Ramsay Chang of EPRI for their
support and encouragement during the analyzer development process.
-------
Figure 1. Zeeman splitting: Emission profile of natural mercury and mercury in a magnetic field.
2537-A Emission Profile
Mercury Lamp
Magnet
cr
Figure 2. Ideal response of photodetector.
o
o.
CO
0>
cr
o
"o
Y
Zero Gas
A zero
Hg Span Gas
A span
Zero Gas with S02
AS02 I
Aso,
Time
-------
Figure 3. Linear response of analyzer to SC>2 concentrations from 0 to 4500 ppm.
-j
IF 1.5 -
o
-i
-------
Figure 5. Graphical summary of mercury concentration calculation.
Average Light Intensity
Figure 6. Mercury measurements with varying 862 concentrations.
Elemental Hg
Total Hg
SO2
Cfl
6.94 6.96 6.98
7.02 7.04 7.06
Time (hrs)
7.12 7.14
-------
Figure 7. Temperature drift effects on mercury analyzer signals.
40
Time (hours)
10
-------
Figure 8. Short term temperature drift of mercury signal strength and average light intensity
*~ **
1650 1700
Average Light (m v)
1800 1BSO
350
Hg Signal {m v)
37 36
40 41 42 43 44 45
Tim e (hours)
11
-------
Figure 9. Analyzer stack sampling filter.
Gas Flow
Oversize Buttonhook Nozzle
Temperature Controlled
Filter Box
, Ceramic Filter
Filter calibration
(ca! gas
passes through filter)
Sample
Figure 10. Mercury concentrations during typical cycle of zeroing, spanning, and inlet and outlet
sampling on ADA's mercury control device (Unit 2; 6/7/97) at Consol, Inc., Library, PA.
10.59
10.64
10.79
Time of Day (hours)
12
-------
Evaluation of Particulate Matter and Total Mercury Continuous Emission Monitors for Compliance
Monitoring at Hazardous Waste Combustion Facilities
Franklin M. Stevens, Jr.
Steven P. Schliesser
Energy and Environmental Research Corporation
1001 Aviation Parkway, Suite 100
Morrisville, NC 27560
H. Scott Rauenzahn
US Environmental Protection Agency
Office of Solid Waste
401 M Street, SW
Washington, DC 20460
ABSTRACT
New rules governing the emission of particulate and Hazardous Air Pollutants (HAPs) from hazardous
waste burning facilities were proposed by the EPA in 1996. These rules are scheduled to be
promulgated in 1998. These rules cover hazardous waste incinerators, cement kilns, and light-weight
aggregate kilns. Continuous monitoring of emissions for compliance is desired where practical in
support of the new regulations. Activities aimed at incorporating the use of Continuous Emissions
Monitoring Systems (CEMS) in the new rules have included assessments of the state-of-the art of CEMS
technology, the drafting of proposed CEMS performance specifications, and field evaluations of a
number of CEMS. The field evaluations have taken place in two (2) programs at two (2) different
facilities. The first CEMS program, carried out from July 1996 through May of 1997 at Dupont's
Experimental station in Wilmington Delaware, evaluated six PM CEMS based on four different
principles of operation. Overall the PM monitors demonstrated that they could meet the criteria of the
Performance Specification 11. They also demonstrated that they could operate with a high degree of
reliability. The second CEMS program, also carried out from July 1996 through May of 1997 at the
Holnam's cement manufacturing facility in Holly Hill, South Carolina, was designed to evaluate three
Hg CEMS. All three (3) Hg CEMS failed to meet the proposed relative accuracy specification were
adversely impacted by the interference gases and had demonstrated marginal data availability.
-------
INTRODUCTION
New rules governing the emission of particulate and Hazardous Air Pollutants (HAPs) from hazardous
waste burning facilities were proposed by the EPA in 1996. These rules are scheduled to be
promulgated in 1998. These rules cover hazardous waste incinerators, cement kilns, and light-weight
aggregate kilns. Continuous monitoring of emissions for compliance is desired where practical in
support of the new regulations. Activities aimed at incorporating the use of Continuous Emissions
Monitoring Systems (CEMS) in the new rules have included assessments of the state-of-the art of CEMS
technology [1], the drafting of CEMS performance specifications, and field evaluations of a number of
CEMS. The field evaluations have taken place in two programs at two different facilities. In the first
program, six (6) particulate matter (PM) CEMS were evaluated at a commercial hazardous waste
incinerator with a low temperature, wet air pollution control system. In the second program, three total
mercury CEMS were evaluated at a cement kiln burning hazardous waste. The results of these
evaluation programs will be summarized in this paper.
Under the proposed rules, CEMS for PM and total mercury could be required in addition to the suite of
CEMS that are required under the current rules. The draft performance specification for PM CEMS is
similar to the new ISO 10155 [2], and is based on calibration of the CEMS response using manual
gravimetric reference method measurements (Method 5, modified for measurements at low loadings).
The performance specification contains minimum requirements for the quality of the calibration relation
derived from these measurements, which are as follows. The correlation coefficient must be greater than
or equal to 0.90, the confidence interval (95%) of the correlation at the emission limit will be within
20% of the emission limit value, and the tolerance interval (bounding with 95% confidence 75% of all
possible measurements) of the correlation at the emission limit will be within 35% of the emission limit
value. Additionally, calibration and zero drift are required to be no more than 2% of the emission limit
value. Since the correlation depends on the type and size distribution of the particulate, which in turn
can be affected by changes in the fuel and facility operation [3,4], the correlation must be developed on a
site specific basis and is valid only over the range of operating conditions and PM loadings used to
develop the calibration relation. It is mainly these factors and the precision of the reference method
measurements which determine the quality of the calibration.
The draft performance specification for total mercury CEMS contains the requirement that the CEMS
measure elemental, speciated, and PM bound mercury. Performance is assessed by comparison to
manual reference method measurements (relative accuracy) and calibration checks using sources of
elemental mercury and mercuric chloride (calibration error). The form of these requirements is similar
to those found in Performance Specification 2 of 40 CFR Part 60, Appendix B. Relative accuracy must
be less than 20% of the mean of the reference method test data or 10% of the emission limit, whichever
is greater, and calibration error, assessed at three different levels, must be less than 15% of the reference
level. Calibration drift, assessed using a standard for elemental mercury, must be less than 10%. Zero
drift must be less than 5%. The specification also requires an interference test on the monitors for a
variety of compounds normally found in flue gas (HC1, S02, CO, C02, 02, NO2, C12, Moisture).
The first CEMS program, carried out from July 1996 through May of 1997 at Dupont's Experimental
station in Wilmington Delaware, evaluated PM CEMS based on four different principles of operation:
Extractive light-scattering (Sigrist KTNR), insitu light-scattering (Durag DR-300, ESC PSA), beta
2
-------
particle transmissivity (Emissions SA Beta 5M, Verewa F-904-KD) and'an acoustical energy monitor
(Jonas Consulting's Acoustic Energy PM monitor). The measurement location was at the exit of a wet
electrostatic precipatator installed down stream of a wet venturi scrubber. The site presented stack
conditions that represented a reasonably worst case scenario with PM loadings ranging from 5 to 100
mg/dscm at 7% O2 and saturated conditions. All six monitors were configured on the stack in the same
relative area. The test results for five of the monitors meet all the requirements of the performance
specification suggesting that with proper handling optical and beta gauge PM monitors can be calibrated
at a facility of this type in conformance with the current performance specification.
The second CEMS program, also carried out from July 1996 through May of 1997 at the Holnam's
cement manufacturing facility in Holly Hill, South Carolina, was designed to evaluate three Hg CEMS.
The monitors used identical measurement technologies (ultraviolet (UV) photometer at 253.7 r|m)
however, each monitor uses a slightly different approach to preparing the sample for analysis. All three
analyzers were located on the transfer duct leading to the stack from the exit of the ESP. The location
did not meet all EPA criteria for a sampling location. Manual method measurements of mercury
concentrations at the location yielded similar to the results from the stack. The test results from the nine
month demonstration did not meet the requirements of the performance specifications.
DESCRIPTION OF THE TEST SITES AND CEMS
Test Sites
The site selected for the PM CEMS demonstration is the incinerator at the Dupont Experimental Station
in Wilmington, Delaware. A Nichols Monohearth incinerator is used as the primary combustion
chamber. Waste is fed to this combustion chamber using three separate means: 1) a ram feeder for solid
waste, 2) a cylindrical chute for batched waste material, and 3) a Trane Thermal liquid waste and No. 2
fuel oil burner. The primary combustor exhausts to a secondary combustion chamber (afterburner)
where No. 2 fuel oil is fed using a Trane Thermal burner. This afterburner chamber discharges to a
spray dryer where the elevated temperature exhaust gases are used to dry the scrubber liquid to remove
dissolved and suspended solids previously collected by the wet scrubber system. Some PM is removed
by the spray dryer; recycling the scrubber water back into the gas stream serves as another source of PM
as does the waste feed streams. The exhaust gas from the spray dryer discharges to a cyclone where
additional PM is removed from the gas stream. The exhaust gas from the cyclone discharges to a reverse
jet gas cooler/condenser, which reduces the gas temperature to the dew point. The reverse jet gas
cooler/condenser discharges into a variable throat venturi scrubber which is used to remove PM and acid
gases. The venturi discharges into a spray absorber where a soda ash neutralized scrubbing solution is
used to absorb acid gases. The gas is subcooled in the absorber to the dew point by the use of the
cooling tower water spray before exhausting through a chevron-type mist eliminator. After this, the gas
is further treated by a set of electrodynamic Venturis (EDVs), which is used to remove fine PM and the
metals that condense as a result of the gas subcooling. The gas then passes through a set of centrifugal
droplet separators, it is then drawn through the induced draft fan and a series of steam heat coils, and it is
exhausted out the stack.
-------
The site selected for the Hg CEMS demonstration is a long wet cement kiln co-firing hazardous waste
owned and operated by Holnam, Inc. in Holly Hill North Carolina. Holnam operates two (2) wet
process kilns at the facility. The larger Kiln #2 is 18'6" in diameter and 580' long, with a design
capacity of 2,100 tons per day of clinker. The main raw materials in the Portland cement manufacturing
process are limestone (providing calcium) and clay (providing silica, alumina, and iron). Limestone and
clay are obtained from the on-site quarries. The facility also obtains other raw materials, such as fly ash,
to supplement on-site raw materials and obtain the correct raw mix to manufacture Portland cement.
The raw materials are finely ground, mixed with water to form a slurry, and fed to the kilns at an
approximate solids content of 65%.
The hot end of each kiln is equipped with a multi-fuel burner, which can be fired with coal, petroleum
coke, waste carbon, shredded tires, fuel oil, and natural gas, with coal being the primary fuel for both
kilns. The burner can also be fired with supplemental solid and liquid hazardous waste fuel. The rated
capacity of the burners are 300 million Btu/hr and 600 million Btu/hr for Kiln #1 and Kim #2,
respectively. The remaining parts of this section provide a more detailed description of the operation of
Kilns 1 and 2 at Holnam's Holly Hill facility.
Slurry is fed into the kiln from the higher (cold) end, while the fuel is introduced at the lower (hot) end.
The material is heated as it travels down the kiln. As the material progresses through the kilns, it
undergoes physical and chemical changes. First, the material loses water as it is heated and passed
through the chain section. Next, calcium carbonate in the slurry is calcinated into calcium oxide (lime),
fuses at high temperature with silicates, iron, and aluminum to produce an intermediate product of
Portland cement, called "clinker." The production of clinker requires that the solid material be heated to
approximately 2,650 to 2,700°F, while the gaseous material reaches a temperature in excess of 3,000°F.
The clinker produced from this process is conveyed through a grate cooler and is cooled by air from
forced draft fans to about 175°F. The cooled clinker is then stored for subsequent grinding, during
which approximately 5% gypsum is added to produce Portland cement, the final product. The exit
gases from the kiln are passed though specially designed electrostatic precipitators (ESPs) which were
supplied for Holnam at this facility.
Particulate Matter CEMS
Light-Scattering CEMS. The light-scattering technologies can be configured as either in-situ or
extractive systems. The three monitors infer participate concentration in the stack by measuring the
amount of light scattered by the particulars in either the forward or backward direction. Various types of
light sources (halogen, infrared, and incandescent) are being used to generate a beam with a known
wavelength. A light sensor or photometer appropriately positioned in either the forward or backward
direction measures the scattered light. Each CEMS is designed with an air-purge system to minimize
PM buildup on the optics. Each monitor adjusts and compensates the detector's signal for interferences,
such as stray light and PM accumulation on its optics. Also, each CEMS has an automatic zero and
calibration check performed daily. The instruments' responses are proportional to the "dry" PM
concentration for a given set of PM characteristics (composition, density, size distribution, index of
refraction) and provide detection levels near 0.5 to 1.0 mg/m3. Each individual instrument undergoes a
factory calibration to ensure the same response for a given set of PM conditions, so a monitor can be
-------
replaced with an identical model without the need for re-calibration. However, since the instrument
response is dependent on PM characteristics, a site-specific calibration is generally required to ensure or
adjust instrument response. These CEMS produce nearly continuous output. Each of the three CEMS
are installed on more than 100 stacks worldwide.
Beta Gauges. Each of the two (2) beta instruments uses a heated sampling line to obtain and deliver an
isokinetic or a close-to-isokinectic sample which is collected on a filter roll. The sampling flowrate and
duration is programmable or adjustable, though the optimal sampling duration depends on PM loading.
After the sampling period is completed, some form of probe purge is performed to entrain any PM
deposit onto the filter. Analysis of the filters begins with determining the beta transmission through
each blank filter is before sampling begins. After a batch sample is collected over the sampling period,
an automatic filter indexing mechanism moves the loaded filter position to a location between the
carbon-14 beta source and a detector. Analysis takes about 2 minutes. The difference between the two
analyses is representative of the PM collected on the filter. Thus, the response of the instrument is
relatively independent of the PM characteristics. These CEMS produce results concurrent with the
sampling period. Each beta gage CEMS are installed on more than 100 stacks worldwide.
Acoustic Energy. In this technique shock waves caused by the impact of particles with a probe inserted
into the gas flow are used to measure particle loading. The device counts the number of impacts and
measures the energy of each impact. This information, coupled with knowledge of the gas velocity,
allows calculation of the particle mass and thus concentration. Since the probe inherently distorts the
localized flow pattern, changes in flow velocity or particle size distribution will, in principle, alter the
instrument's response. Since the instrument response is dependent on PM characteristics, a site-specific
calibration is expected to be required to ensure or adjust instrument response. This CEMS produces very
frequent signals on a nearly continuous basis. This vendor has not yet presented any evidence that this
technology is used for a PM air emission application.
Total Mercury CEMS
The total mercury CEMS tested were made by Perkin Elmer (Mercem), Monitor Labs.Inc (Verewa HM-
1400) and EcoChem Technologies, Inc (Seefelder's Hg-Mat 2) . The Verewa HM-1400 continuous total
mercury analyzer was provided and supported by Monitor Labs. This monitor is designed to measure
total mercury: elemental mercury, mercury compounds, and particulate matter (PM) bound mercury.
Analyzer output is on a continuous weight/volume basis (ug/dscm) as mercury. A gas sample is
extracted non-isokinetically at a constant sample rate of two (2) 1/min from the stack through a resistance
heated, polytetrafluoroethylene (PTFE) lined probe. The sample is transported to the analyzer through a
PTFE sample line heated to 120°C. In the analyzer, the sample gas first passes through an oven where it
is heated to 800°C. This vaporizes any mercury present in the particulate and partially dissociates
mercury compounds. The sample is then mixed with hydrochloric acid (HC1) at 70°C to transform all
mercury compounds to HgCl2. This solution then reacts with sodium borohydride (NaBH4) at 10°C to
reduce all the mercury to the elemental form. A gas-liquid separator at 2°C strips the gas phase mercury
vapors out of solution. The mercury vapor is then detected in an ultraviolet (UV) photometer at 253.7
T)m. The photometer uses a double-beam configuration. Mercury sample flows through one (1) beam, is
then scrubbed of mercury and passes through the second beam. This technique eliminates most
potential interferences. The signal difference is due to any interfering substances detected by the
-------
photometer, minus the mercury fraction. However, its effectiveness depends on the selectivity of the
scrubbing. Finally, the sample gas is dried and the volume flow rate measured so that the instrument
output can be reported on a dry basis. The reagents used in the analyzer are stored outside the CEMS
analyzer enclosure and pumped continuously into the reactors. Replenishment of the reagents and
removal of the waste solutions is required every two (2) weeks.
The Seefelder Messtechnik Hg-Mat 2 continuous total mercury analyzer was provided and supported by
EcoChem Technologies. This monitor is designed to measure total mercury: elemental mercury,
mercury compounds, and PM bound mercury. Analyzer output is on a continuous weight/volume basis
(ug/dscm) as mercury. A gas sample is extracted non-isokinetically at a constant sample rate of 1.5
1/rnin from the stack through a heated stainless steel probe and transported to the analyzer through a
PTFE sample line heated to 200°C. In the analyzer, the sample gas passes through two (2) reactors
where it is cooled and all speciated mercury is converted to the elemental form. PM bound mercury is
also desorbed in the reactors. The elemental mercury is carried in the vapor phase (separated from the
liquid by a demister) to a UV photometer operating at 253.7 r|m. Finally, the sample gas is dried and the
volume flow rate measured so that the instrument output can be reported on a dry basis. The reagent
used in the analyzer is stored outside the CEMS analyzer enclosure and pumped continuously into the
reactors. Replenishment of the reagents and removal of the waste solutions is required every four (4)
weeks.
The Perkin-Elmer MERCEM continuous total mercury analyzer was provided and supported by
Wheelabrator Clean Air Systems. This monitor is designed to measure total mercury: elemental and
mercury compounds. Analyzer output is on a weight/volume basis (ug/dscm) as mercury. A gas sample
is extracted non-isokinetically at a constant sample rate of about 17 1/min from the stack through a
heated platinum probe equipped with two (2) heated scintered metal filters. The sample is transported to
the analyzer through a PTFE sample line heated to 185 °C. At the analyzer, a small portion of the
sample flow (about 0.5 1/min) enters a reactor in which mercury compounds are reduced to elemental
mercury by a tin (II) chloride (SnCl2) solution.
The sample gas containing vapor phase elemental mercury is separated from the reaction chamber liquid
and enters a thermo-electric cooler, where it is cooled and dried to a dew point of 5 °C. The dried sample
then enters an amalgamation unit during a continually cycling batch operation. Mercury vapor is
collected on a cool gold/platinum trap. At the end of the batch collection time, the trap is purged with
instrument air, and the photometer baseline is determined. Then the trap is heated to 750°C to thermally
desorb the mercury, which is released into the nitrogen carrier gas flow and measured in the photometer.
The sensitivity of the instrument can be varied by changing the collection time, which is about 15
seconds for the 0-100 ug/dscm measuring range. Since the flow through the photometer consists of only
nitrogen and elemental mercury, optical interferences are eliminated. Sample flow rate is measured so
that the mercury concentration can be reported on a dry basis. The reagents used in the analyzer are
stored in the CEMS enclosure and added continuously into the reactor with a controlled reflux rate.
Replenishment of reagents and removal of waste solutions is required every three (3) months. The
Perkin-Elmer unit uses an internal chiller to maintain the required instrument temperature.
-------
EXPERIMENTAL PROCEDURE
PM Demonstration
The overall scope of the PM CEMS demonstration included prescreening measurements for PM, HC1,
and particle size distribution; development and laboratory testing of a Modified Method 5 for low PM
loading measurements; and field demonstration of the PM CEMS. The elements of the endurance test
included:
Two calibration relation tests;
Monthly RCAs (comparison to reference method measurements);
Monthly ACAs;
Continuous recording of CEMS data for nine months;
Documentation of daily calibration and zero checks;
Documentation of all performed maintenance/adjustments; and
Documentation of all periods of data non-availability. The test which establishes the
correlation between the CEMS outputs and the reference method is called the calibration
relation. Subsequent tests to determine whether that calibration is still valid are referred
to as response calibration audits (RCAs).
Two calibrations were performed under a similarly wide variety of facility operating conditions. The
first calibration was performed in September and October 1996. These tests established the initial
calibration. As experience was gained, data quality improved, and due to suggestions from the public, a
second calibration relation test was conducted in April 1997 to evaluate the data quality produced from
the PM CEMS and the reference method. In addition, six (6) RCAs were performed from November
1996 to March 1997 and again in May 1997 to evaluate the stability of the respective PM CEMS
calibration relations over time. The RCAs were conducted under normal operating conditions, and no
attempt was made to modify process conditions. The RCA test serve a two-fold purpose: 1) to
determine the acceptability of the RCA data relative to the two calibration relation test results, and 2) as
additional supporting data for each PM CEMS into forming an overall or cumulative database which
consisted of all acceptable data. The evaluation protocols used were those found in the proposed PS 11
for PM CEMS and Appendix to Subpart EEE, the data quality objectives. All tests consisted of
comparing CEMS outputs to concurrently run, paired modified Method 5 (M5) measurements as the
reference method.
Hg Demonstration
Draft Performance Specification 12 (DPS 12) sets the standards which the monitors must meet. It
requires the use of calibration standards of a known concentration of mercury presented to the analyzer
section of the CEMS to determine the calibration drift or calibration response of the analyzer. Examples
of calibration standards include stock mercury solutions of known concentration, filters with a known
composition, or permeation tubes with certified mass permeation rates at a given temperature.
However, no specific calibration techniques or auxiliary procedures are stated for the precision of these
techniques, nor for the accuracy of the standards.
-------
The MERCEM and Hg-Mat 2 CEMS use permeation tubes to introduce the calibration gas into the
analyzer portion Hg CEMS on a daily basis. However, the perm tube systems are not NIST-certified,
and the calibration gas produced has no reference value traceability. Thus, the calibration check
conducted may not be precise or repeatable, and gives no measure of the true bias of the procedure. The
procedure was designed only as a general check on analyzer operation. The Verewa CEMS challenges
the analyzer by introducing a standard solution into the reaction chamber. However, this procedure must
be conducted manually; it is not used for a daily auto-calibration. The Verewa analyzer does not
complete a calibration procedure on a continuous daily basis.
Calibration error (CE) and zero drift (ZD) were evaluated once daily over a 7-day period. The Hg
CEMS outputs were recorded and flagged during these tests to allow calculation of the drift. Zero
calibrations were completed automatically by all three (3) CEMS. ZD was determined by introducing a
zero gas into the analyzer portion Hg CEMS. High level calibrations were conducted automatically by
the permeation tube technique for the MERCEM and Hg-Mat 2 CEMS, and by manual calibration with
mercury solutions for the Verewa CEMS.
The CE tests were performed by challenging the Hg CEMS with concentrations of elemental mercury
(Hg°) and mercuric chloride (HgCl2) generated using permeation tubes. The CE was carried out
separately for both Hg° and HgCL>. The calibration system introduced a quantity of gas phase Hg° or
HgCl, in nitrogen to the sampling system of the Hg CEMS, upstream of all filters, scrubbers, and other
gas conditioning. In each case, the target CE test concentrations were 0%, 40-60%, and 80 - 120% of
the emission limit (50 ^ig/dscm). The challenge was conducted three times non-consecutively at each
level. The reference gases were delivered simultaneously to all three (3) CEMS sample systems. The
gas concentrations were verified by a simultaneous measurement during each Hg CEMS challenge using
the Method 101A based verification system.
The interference testing was conducted following the CE test at the high concentration level for both
Hg° and HgCk As with the CE testing, calibration gases were generated with a permeation tube device,
and the concentration verified with a modified M101A verification train. After the CEMS response to
each high level calibration gas was recorded, the interference test gases were substituted for the
nitrogen dilution gas flow used in the CE test. The response of each Hg CEMS was recorded and
compared with that from the Hg° and HgCL, individually to calculate the interference as described in the
performance specification. Each interference test gas is introduced singly.
The relative accuracy test audits (RATAs) were conducted by comparing simultaneous Hg CEMS and
reference method measurements. The reference methods used in this demonstration was the proposed
Draft Method 101B (M101B) train for speciating mercury and a standard Method 29 (M29) sampling
tram. Initially, the reference method measurements were conducted at two (2) locations: the standard
stack sampling location, and on the transfer duct co-located with the Hg CEMS. The stack
measurements utilized both M101B and M29 trains during the first test series with a full traverse of the
stack. The transfer duct measurements were made at a fixed point with a single M101B sampling train.
Nine (9) runs were made at both locations. Relative accuracy (RA) was calculated according to the draft
performance specification for each individual method and location. The initial RA testing was conducted
during the first week of the performance test period. RATAs were being conducted every four (4)
-------
weeks after.
After the first two RAT As the testing was placed on hold for the months of November and December of
1996 to allow the vendors to reconfigure their monitors. The flue gas environment was presenting
problems to the gas conditioning portions of the monitors causing instrument failure.
RESULTS AND ANALYSIS
Participate Matter CEMS
Over the 9-month program a total of two calibration relation tests and six response calibration audits
(RCAs) were performed. Calibration test results produced by paired Method 5 (M5) trains ranged from
5 to 100 mg/dscm (0.002 to 0.044 gr/dscf) @ 7 % O, across a range of gas temperatures from 290 to 320
°F and moisture levels from 15 to 30 %. There were 85 out of 102 runs in which the paired M5 trains
produced acceptable results within established QA criteria for relative standard deviations < 30 %. Most
of the weight gain was associated with the filter catch, as there was generally 5 to 25 mg weight gain on
the filters and 0.5 to 4 mg weight gain from the front-half probe rinses.
One-minute average readings from each of the five PM CEMS were recorded on a data logger and
complied for the corresponding M5 sampling periods for each test condition in consideration of their
respective response times. Since five monitors were undergoing calibration testing collectively, the
policy was to proceed with testing even though a monitor(s) was not operating properly or undergoing
self-calibration. Review of field notes and CEMS data for each condition was made to determine
whether the CEMS data produced was valid or not. On balance, the PM CEMS produced acceptable
data for about 95 % of the test conditions.
Individual pairs of M5/PM CEMS results for each run in the two calibration relation tests were evaluated
according to EPA draft Performance Specification 11 and ISO 10155. This evaluation included the
procedures and statistical equations for calculating the correlation coefficient, confidence interval, and
tolerance interval. Briefly, these involve performing a regression analysis on the correlations between
paired set of CEMS and M5 data. Depending on the measuring conditions experienced by the individual
monitor, the PM CEMS/M5 correlations are based on either actual in-stack (mg/actual cubic meter
(acm) or dry standard PM concentration units (mg/dscm) from the M5 results. This is done because the
three light-scattering monitors and the ESA Beta monitor perform their analysis under actual in-stack
conditions (thus producing PM concentration data proportional to that reference), whereas the Verewa
measures the sample gas volume under dry conditions and then calculates PM concentration to a (dry)
standard reference temperature. A linear calibration relation is then calculated by performing a linear
least squares regression. The CEMS data are taken as the x values and the reference method data as the
y values. The calibration relation, which gives the predicted PM concentration, y', based on the CEMS
response x, is given by:
y' = 3*x + b
where :
a = slope of the linear regression line, and
b = y intercept.
-------
Following this, the 95 % confidence interval for the regression relation is computed, as is a tolerance
interval bounding 75 % of the population of the paired data with 95 % confidence; both intervals are
calculated at the proposed emission limit level (69 mg/dscm @ 7 % O2). The equations provided in draft
PS 11 were put on a spreadsheet, while values for tf, vf, and u,,' were automatically inserted from Table I
in draft PS 11. In essence, the confidence interval give the 95 % confidence on the uncertainty of the
PM concentrations calculated from the CEMS response using the regression relation. The tolerance
interval bounds the region within which one would expect continued paired data sets to fall, based on the
measurement pairs used to perform the calibration. Subsequent measurements during RCA tests
comparing CEMS responses to reference method data are considered consistent with the current
calibration relation if at least 75 % of them fall within a tolerance level of either 25 % or 35 %.
Calibration Relation Test Results
Table 1 presents the results using the acceptable data for each calibration relation test. For the first
calibration relation tests, only the ESA Beta monitor produced data meeting the draft PS 11 criteria for
the correlation coefficient (> 0.90), confidence interval (< 20 %), and tolerance interval (< 35 %).
However, for the second calibration relation tests, all five monitors produced data not only meeting and
exceeding all the draft PS 11 criteria but also the more strict ISO criteria for the correlation coefficient
(> 0.95), confidence interval ( < 10 %), and tolerance interval (< 25 %) with one exception. That single
exception was the Sigrist light-scattering monitor not meeting the ISO confidence interval criterion.
The results of the draft PS 11 statistical calculations from both second calibration relation tests for two
CEMS are included in the graphical illustrations in Figures 1 and 2 for the ESA and Durag monitors,
respectively. Review of the comparison between the two calibration relation regression lines and their
corresponding linear equations in Figures 1 and 2 give insight into two important aspects of these two
calibration relation tests : 1) reproducibiliry of CEMS performance in terms of the equations defining
their linear relationship with M5 measured results, and 2) reproducibility in data quality associated with
the M5 results produced in the early and then the later stages of the program. For the first point on
CEMS performance, stability over time and relative insensitivity to changes hi PM properties would be
reflected by the level of agreement between the equations defining each CEMS/M5 relationship for the
two sets of tune periods/calibration test operating conditions. Comparison of the slopes of the linear
equations are the key measure in this comparison, since all of the y-intercepts in the equations had
nominal values of 5 (mg/acm or mg/dscm) or less. Below is a comparison of the linear equations' slopes
between the two calibrations for each CEMS, showing agreement within 11 % for all five monitors.
This agreement between the equations defining their relation with M5 serves to reflect acceptability of
subsequently applying the calibration relation test results for each monitor over time and variations in
waste feedstreams, PM properties, and gas conditions.
CEMS First Calibration Slope Second Calibration Slope % Difference
ESA 0.83 0.77 7.2
Verewa 1.30 1.42 8.4
Durag 0.36 0.37 2.7
ESC 0.40 0.45 11
Sigrist 0.29 0.31 6.5
10
-------
Regarding the second aspect on M5 results, duplication in the data quality from the M5 results between
the two calibration tests would be reflected, given the same/similar reproducibility in GEMS
performance, with same/similar correlation coefficients, confidence intervals, and tolerance intervals.
Although a high degree of correlation and reproducibility is afforded in the second calibration, this is not
the case for the first one. This and other information available, but not reported in this paper,
collectively show an improvement in data quality between the M5 results produced in the first and
second calibration.
PM CEMS Monthly RCA Test Results
The discussion focuses on the results produced during six calibration tests conducted monthly from
November through March and then in May in relation to RCA criteria specified in draft PS 11. As a key
part of this national demonstration program, the monthly test results form another basis on which the
acceptability of subsequently applying the calibration relation test results are evaluated to see if CEMS
performance vary over time. Due to the non-reproducibility of waste feedstreams, the monthly tests also
furnish a measure of evaluating CEMS reproducibility in terms of variations produced in PM properties
and flue gas conditions. The overall set of monthly test data are assembled collectively to evaluate if
they meet the RCA draft PS 11 criteria.
The RCA results are compared to the calibration relation as the basis of this evaluation. If 75% of the
RCA data fall within a certain tolerance intervals, then continued use of the most recent calibration
relation is considered acceptable to monitor PM emissions until the next RCA is performed. If the 75%
threshold is not met, then a new calibration relation must be developed to monitor PM during
subsequent operations.
The following explanation, along with the accompanying tables and figures, illustrates the (revised) draft
PS 11 procedure to evaluate acceptability of the subsequent RCA data relative to the calibration
relation. First, a figure is produced showing the calibration relation regression line based on the
calibration relation data and the tolerance intervals set at either 25 % or 35 % from the PM standard (69
mg/dscm @ 7 % 02). The tolerance interval boundary lines apply the same slope as the regression line
and intersect the emission standard at values +/- 25 % and +/-35 % of the proposed PM limit. Second,
the values from the paired sets of data from the RCA tests are plotted and overlaid onto the above figure.
Third, the number of points inside and outside the tolerance intervals are separately counted to determine
if at least 75 % of the RCA points fall within the tolerance interval boundaries.
The RCA evaluations are illustrated in Figures 3 and 4 for the Verewa's February, March, and May
monthly calibration tests and for the ESC's February, March, and May monthly calibration tests,
respectively. Results from the RCA evaluations for all the monthly tests for these same two monitors
relative to the 25 % and 35 % tolerance intervals are presented in Table 2. Included in Table 2 are the
number of data points for each month within and outside of each respective calibration relations' range
of PM concentrations. The data within the calibration range of PM concentrations represents valid data
applicable for the RCA evaluation; data outside the range of PM concentrations are shown but excluded
from evaluation. Each monitor produced data meeting the RCA criterion for each monthly data set
relative to each calibration with few exceptions. On the collective set of monthly data, the monitors
11
-------
produced valid data with the following percentages relative to each calibration within the respective
tolerance intervals (T I):
First Calibration Second Calibration
100 100
76 86
85 93
92 94
92 94
These results clearly demonstrate that each monitor produced acceptable data beyond the 75 %
RCA criterion relative to each calibration within each of the 25 % and 35 % tolerance intervals. This is
another firm indication of the reproducibility of these CEMS for continuous monitoring of PM
emissions and compliance.
Total Mercury CEMS
This program was designed to determine whether Hg CEMS are available that can meet the draft
performance specifications. During the performance test the Hg CEMS were challenged with gas phase
mercury and mercuric chloride (HgQ2) as part of the calibration error testing and RAT As were carried
out at two (2) different facility operating conditions, one with normal chlorine levels derived from the
co-firing of hazardous waste, and the other with lower chlorine levels when no hazardous waste is
burned. The initial RATA was a part of the performance test, and was performed at the normal chlorine
conditions. One of the subsequent monthly RATAs were performed at the "low" chlorine condition. All
other RATAs were run at normal chlorine levels. The program was performed over a nine month period.
All the data has not been analyzed to date. Table 3 presents the results of the calibration error and
RATAs performed through February 1997.
All three (3) Hg CEMS failed to meet the proposed interference response specification for total
(absolute) interference of 10% of the high-level reference value. The Perkin-Elmer MERCEM cannot
measure dry SO2, N02, or HC1 without first humidifying the gas. Since the CEMS optical bench
measures mercury on a dry basis, it must be assumed that these gases are removed along with water
vapor (scrubbing) in the reaction chamber and/or gas conditioner (dryer). Interference response may
therefore vary with moisture content under actual operating conditions. The Seefelder Hg MAT-2
exhibited extreme interference response to both SO2 (500 ppm) and N02 (250 ppm). H2O proved too
difficult to handle as an interference test gas with the currently available equipment. No data is available
for H2O interference response.
All three (3) Hg CEMS failed to meet the proposed relative accuracy specification based on 20% of the
average reference value. The characteristics of the kiln gas matrix had detrimental effects on CEMS
performance. Condensible components in the stack gas effluent caused zero drift problems with the
Verewa HM-1400, and reaction chamber clogging with the Perkin- Elmer MERCEM. The gold trap
component of the Perkin-Elmer MERCEM also showed poor collection efficiency after two (2) months
and was replaced. The monitors had relative accuracy values from 24 to 74%.
12
-------
The program also evaluated the overall reliability of the monitors, operational problems exhibited by the
individual CEMS analyzers during and after the initial RATA test have necessitated that all three (3)
CEMS be modified or upgraded to minimize the effects of the sample gas matrix. In general, these
problems were associated with the content of the stack gas matrix generated by the cement kiln, and its
effect on sample transport, zero drift, and interference response. The analyzer systems evaluated for this
program were designed to be used on a combustion gas sample that has been scrubbed of acid gas and
has low non-condensible particulate concentrations, neither condition of which exists at the host facility.
After the first RATA modifications were made to enhance the performance of the CEMS, and increase
long-term availability of the CEMS. Since the systems were modified the performance of the Perkin-
Elmer MERCEM and the Hg-Mat 2 CEMS have improved. The Verewa HM-1400 went offline in
March of 1997 and needed almost continuous maintenance prior to that to keep it on line for any period
of time. The preliminary data availability figures for the monitors show the Verewa HM-1400 to be
available approximately 40% of the time while the other two monitors were online an average of 88% of
the time.
SUMMARY AND CONCLUSIONS
Overall the PM monitors demonstrated that they could meet the criteria of the Performance Specification
11. They also demonstrated that they could operate with a high degree of reliability under reasonable
worst case (as-found, normal day-to-day) HWC facility operations. All five PM CEMS produced
accurate/precise data meeting Draft PS 11 and ISO 10155 acceptance criteria for correlation coefficient
and confidence/tolerance intervals during the second calibration relation testing which occurred after
seven months of operation. All five PM CEMS produced accurate/precise/stable data meeting Draft PS
11 acceptance criteria for either 25 % or 35 % tolerance interval requirements in the RCA tests and
four of the five PM CEMS produced reliable data available from 85 % to 99 % of the time.
The mercury monitors did not fair as well during the testing program. All the data has not been analyzed
however, the general trend in the data suggests that the monitors will not meet the performance
specifications. The frequency for required maintenance of the monitors during the program was on
almost a weekly basis which makes it unlikely that the monitors could be economically used to
determine compliance.
ACKNOWLEDGMENTS
This work was carried out under EPA contract 68-D2-0164.
REFERENCES
1. Draft Technical Support Document for HWC MACT Standards. Volume IV: Compliance with the
Proposed MACT Standards. U.S. Environmental Protection Agency, Washington, DC, 1995.
2. International Standards Organization, "Stationary Source Emissions - Automated Monitoring of Mass
Concentrations of Particles - Performance Characteristics, Test Procedures and Specifications," ISO
10155, available from ANSI.
13
-------
3. A.W. Gnyp, S.J.W. Price, C.C. St. Pierre, and D.S. Smith, "Long Term Field Evaluation of
Continuous Particulate Monitors," in Proceedings: Advances in Particle Sampling and Measurement.
EPA 600/7-79-065, U.S. Environmental Protection Agency, Research Triangle Park, 1979, pp. 122-168.
4. Jahnke, J.A., "Transmissometer Systems - Operation and Maintenance, An Advanced Course," APTI
Course SI:476A, EPA 160/2-84-004 (1984).
14
-------
TABLE 1. CALIBRATION RELATION STATISTICS
GEM
ESA
VEREWA
Durag
ESC
SIGRIST
Draft PS 1 1
ISO 10155
Initial Calibration
Second Calibration
Initial Calibration
Second Calibration
Initial Calibration
Second Calibration
Initial Calibration
Second Calibration
Initial Calibration
Second Calibration
Correlation
Coefficient
>0.90
>0.95
0.919
0.985
0.897
0.961
0.760
0.972
0.767
0.978
0.758
0.951
Confidence
Interval
< 20 %
< 10%
14.3%
7.1%
13.6%
9.8%
21.6%
8.7%
21.2%
7.8%
21.7%
12.0%
Tolerance
Interval
< 40 %
< 25 %
24.9%
11.1%
22.6%
16.7%
42.4%
13.4%
40.9%
12.1%
42.5%
17.0%
-------
TABLE 2. FRACTION / PERCENT OF RCA WITHIN TOLERANCE LEVEL
Calibration 1
CEM
ESA
VEREWA
Durag
ESC
SIGRIST
Feb. RCA
March RCA
May RCA
Cumulative Data
Feb. RCA
March RCA
May RCA
Cumulative Data
Nov. RCA
Dec RCA
Jan. RCA
Feb. RCA
March RCA
May RCA
Cumulative Data
Nov. RCA
Dec RCA
Jan. RCA
Feb. RCA
March RCA
May RCA
Cumulative Data
Nov. RCA
Dec RCA
Jan. RCA
Feb. RCA
March RCA
May RCA
Cumulative Data
Data Deve oped
> 75 %
7/7
4/7
6/6
17/20
7/7
9/9
3/6
19/22
4/4
5/5
4/4
7/7
5/7
6/6
31/33
5/7
5/5
6/6
7/7
9/9
7/7
39/41
6/7
5/5
8/8
7/7
8/8
8/8
42/43
100%
57%
100%
85%
100%
100%
50%
86%
100%
100%
100%
100%
71%
100%
94%
71%
100%
100%
100%
100%
100%
95%
86%
100%
100%
100%
100%
100%
98%
25%
> 75%
7/7
7/7
6/6
20/20
7/7
9/9
6/6
22/22
4/4
5/5
4/4
7/7
3/7
6/6
29/33
4/7
5/5
6/6
7/7
9/9
7/7
38/41
4/7
5/5
8/8
7/7
8/8
8/8
40/43
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
43%
100%
88%
57%
100%
100%
100%
100%
100%
93%
57%
100%
100%
100%
100%
100%
93%
35%
> 75%
7/7
7/7
6/6
20/20
7/7
9/9
6/6
22/22
4/4
5/5
4/4
7/7
4/7
6/6
30/33
5/7
5/5
6/6
7/7
9/9
7/7
39/41
5/7
5/5
8/8
7/7
8/8
8/8
41/43
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
1 00%
57%
100%
91%
71%
100%
100%
100%
100%
100%
95%
71%
100%
100%
100%
100%
100%
95%
Out
ofCal.
Range
4
2
5
11
4
3
3
10
1
4
4
5
6
20
2
4
3
5
14
-
4
4
2
10
-------
TABLE 2 (CONT). FRACTION / PERCENT OF RCA WITHIN TOLERANCE LEVEL
Calibration 2
CEM
ESA
VEREWA
Durag
ESC
SIGRIST
Feb. RCA
March RCA
May RCA
Cumulative Data
Feb. RCA
March RCA
May RCA
Cumulative Data
Nov. RCA
Dec RCA
Jan. RCA
Feb. RCA
March RCA
May RCA
Cumulative Data
Nov. RCA
Dec RCA
Jan. RCA
Feb. RCA
March RCA
May RCA
Cumulative Data
Nov. RCA
Dec RCA
Jan. RCA
Feb. RCA
March RCA
May RCA
Cumulative Data
Data Developed
> 75 %
7/8
1/7
9/9
17/24
5/10
9/9
8/8
22/27
1/4
2/5
5/5
7/7
3/5
4/8
22/34
0/7
5/5
6/6
6/7
7/8
5/7
29/40
1/7
2/5
5/6
7/7
8/8
7/8
30/41
88%
14%
100%
71%
50%
100%
100%
81%
25%
40%
100%
100%
60%
50%
65%
0%
100%
100%
86%
88%
71%
73%
14%
40%
83%
100%
100%
88%
73%
25%
> 75%
8/8
7/7
9/9
24/24
10/10
9/9
8/8
27/27
4/4
5/5
5/5
7/7
4/5
8/8
33/34
3/7
5/5
6/6
7/7
8/8
7/7
36/40
4/7
5/5
6/6
7/7
8/8
8/8
38/41
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
80%
100%
97%
43%
100%
100%
100%
100%
100%
90%
57%
100%
100%
100%
100%
100%
93%
35%
> 75%
8/8
7/7
9/9
24/24
10/10
9/9
8/8
27/27
4/4
5/5
5/5
7/7
4/5
8/8
33/34
4/7
515
6/6
7/7
8/8
111
37/40
4/7
5/5
6/6
7/7
8/8
8/8
38/41
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
100%
80%
100%
97%
57%
100%
100%
100%
100%
100%
93%
57%
100%
100%
100%
100%
100%
93%
Out
ofCal.
Range
3
2
2
7
1
3
1
I
1
-
3
4
7
4
19
-
-
2
4
4
5
15
-
2
4
4
2
12
-------
Table 3 : Summary of Ha CEMS Demonstration Tests
CEMS
Model
Test Dale
Proposed Perf Level
Less (hail slntcd value
/crewa HM-
1400
8/96
12/96
01/97
02/97
03/97
03/97
04/97
05/97
ALL
Hg
llgCI,
Hg
HgCI,
Hg
1-lgCl,
Hg
HgCI,
Hg
HgCl,
Hg
HgCI,
Hg
HgCI,
Hg
HgCl,
Hg
MgCl!
Cosls
Purch.
n/a
Maint.
(Ann.)
n/a
Perlormance Achieved
RA
±20% of
mean
n/a
n/a
-364
B
247.6
P
±10%
of sld
n/a
n/a
11.3
B
29.6
r
Off
Off
Off
Off
CE
±15% of
Rcf Cone
a(3 levels
n/a
n/a
P
P
P
P
22.7
-460
-60.1
7.6
-89.7
Off
Off
....
Off
Off
Off
Off
CD
± 10%
of
sld
n/a
n/a
NP
n/a
NP
n/a
NP
n/a
NP
n/a
....
Off
n/a
Off
n/a
NP
n/a
ZD
±5%
or
sld
n/a
n/a
TDD
n/a
TBD
n/a
TBD
n/a
Off
n/a
....
Off
n/a
Off
n/a
n/a
Data
Avail.
n/a
n/a
862
67.6
90.1
II 4
0
0
400
Observed Interference Levels (§4.7)
CO CO, O, SO, NO, 11,0 IICI Cl, TOTAI
n/a n/a n/a n/a 'n/a n/a n/a n/a +10% of
sld
-33 -71 -4.7 -5.4 -3.3 NP 17.2 -2.6 -93
1.6 02 91 -4.9 NP NP 8.1 -10. 1 4 1
NP
-3.3 -7.1 -4.7 -54 -33 NP 17.2 -2.6 -9.3
1.6 0.2 9.1 -4.9 NP NP 8.1 -10.1 4.1
-------
Table 3 : Summarv of Hg CBMS Demonstration Tests
CEMS
Model
Proposed PC
.ess than sla
MERCEM
Sec. Ilg-
MAT2
Test Date
f. Level
cd value
=
8/96
12/96
01/97
02/97
ALL
8/96
12/96
01/97
02/97
ALL
Hg
HgCI,
Hg
HgCI,
Hg
HgCI,
Hg
HgCI,
Hg
llRCI,
Hg
HgCI,
Hg
HgCI,
Hg
Hgcl,
Hg
HgCI,
Hg
HECI
Costs
Purch.
u/a
Maint.
(Ann.)
n/a
Performance Achieved
RA
+20% of
mean
n/a
n/a
63.1 B
71.4 nrf
P
n/a
n/a
-24.2
B
716.4
P
+ 10%
ofsld
n/a
n/a
19.6 B
8.7 nrf
P
n/a
n/a
-7.5
B
87.0
P
CE
+ 15% of
RefConc.
at 3 levels
n/a
n/a
NP
NP
P
P
37.8
13.2
-19.9
41.2
-20.8
n/a
n/a
NP
NP
P
P
-1.5
-61.3
-68.8
0.6
-84.3
CD
+ 10%
of
Sid
n/a
n/a
TBD
n/a
TBD
n/a
TBD
n/a
n/a
n/a
n/a
TBD
n/a
TBD
n/a
TBD
n/a
n/a
ZD
+ 5%
of
std
n/a
n/a
TBD
n/a
TBD
n/a
TBD
n/a
n/a
n/a
n/a
TBD
n/a
TBD
n/a
TBD
n/a
n/a
Data
Avail,
n/a
n/a
94.4
89.9
96.8
89.9
n/a
95.7
87.6
96.0
87.0
Observed Interference Levels (§4.7)
CO
n/a
3.9
0.8
C0;
n/a
O.I
-11.5
o,
n/a
•31.8
-9.3
SO,
n/a
-2.8
-3.7
NO,
'ii/a
0.6
NP
11,0
n/a
NP
NP
1ICI
n/a
0.7
-2.0
Cl,
n/a
-0.8
-1.2
TOTAL
+ 10% of
std.
32.3
-27.0
NP
3.9
0.8
0.2
-3.3
0.1
-11.5
1.8
-4.9
31.8
-9.3
1 8
4.0
-2.8
-3.7
58.3
0.7
0.6
NP
-92.6
NP
NP
NP
NP
NP
0.7
-2
-50.4
5.2
-0.8
-1 2
3.0
3.3
32.3
-27
-78.0
3.7
NP
0.2
-3.3
1.8
-4.9
1.8
4
58.3
0.7
-92.6
NP
NP
NP
-50.4
5.2
3
3.3
-78
3.7
NP: Sequence not performed.
DNA: Dala not yel available from analytical lab
B: Baseline performance: Accuracy without response factor applied
P:Tcst postponed
TBD: To be determined
*: Data questionable, CEM malfunction
Off: CEM Offline
nrf: no response factor applied
-------
30.00
25.00 ---
20.00
-2 15.00 —
c
o
U
T3
O
.a
5.00 —-
0.00
0.
-5.00 J-
B Cal 2
• Cal 1
Poly, (cal 1 y_pred + lol)
. „ ... Poly, (cal 2 y_pred + tol)
Poly, (cal 2 y_pred + ci)
- -Linear (Cal 2)
Poly, (cal 1 y_pred + ci)
Poly, (cal 2 y_pred - ci)
Linear (Cal 1)
- Poly, (cal 2 yjred - tol )
Poly, (cal 1 yjred - ci )
Poly, (cal 1 y_pred - tol )
3000
CEM'SPM Response
Figure 1. ESA First and Second Calibration Relation
-------
35.00 -,
Cal 2
Cal 1
Poly, (cal 1
Poly, (cal 1
•--Poly, (cal 2
-—-Poly, (cal 2
- - - Linear (Cal
Linear (Cal
Poly, (cal 2
— Poly, (cal 2
Poly, (cal 1
Poly, (cal 1
y_pred + tol)
y_pred +ci)
y_pred + tol)
y_pred + ci)
1)
2)
y_pred - ci)
yjpred - tol)
yjred - ci)
y_pred - tol)
.00
-5.00 I—
CEM'S PM Response
Figure 2. Durag First and Second Calibration Relation
-------
70 000
60.000
1 50.000
VI
•o
a 40.000
a
| 30.000 -
a
« 20.000 -
o
O
"> 10.000
•a
o
A
"S o ooo
0
-10 000
-°0 000
,•'"'
•' ,. -^
j'
- - x-;5*'
)0 /lO
,./
,x
J/X =>
•^-//'
jjJ s/^ ^^
Ss' ,
^^ X _, • '
k~*^'' /
. ^'^^^'^
..,//*•' /
OP^"^ 20
.
S^/^
^/^ y=c
£^\raB' ,-
t - __ »
LI , -
' ' ' _X^
^€^
S^s<
00 30
' u
1.4207x-2.51
y=1.2988x-f
—
00 40
"D "
)3
.312
00 50
A RCA-feb
a RCA-mar
x RCA-may
Linear (Cal 1)
Linear (cal 2 y pred +35%)
- — Linear (cal 2 yjred +25%)
• Linear (Cal 2)
- Linear (cal 2 y pred -25%)
Linear (cal 2 y__pred -35%)
CEM'S PM Response
Figure 3. Verewa RCA Evaluation : February, March, and May
-------
,4000-
ation (mg/ACM
Con
A RCA-feb
n RCA-mar
x RCA-may
Linear (cal 1 yjpred +35%)
Linear (cal 2 yjpred +35%)
Linear (cal 1 yjpred +25%)
Linear (cal 2 yjpred +25%)
Linear (Cal 1)
Linear (Cal 2)
Linear (cal 1 yjpred -25%)
Linear (cal 2 yjpred -25%)
Linear (cal 1 yjjred -35%)
— -Linear (cal 2 yjpred -35%)
00
CEM'S PM Response
Figure 4. ESC RCA Evaluation : February, March, and May
-------
Development of a Continuous Emissions Monitor for HCI and CI2
Andrew D. Sappey, Ph.D.
Richard J. Schlager
Michael D. Durham, Ph.D.
Douglas W. Jackson
Francis J. Sagan
ADA Technologies, Inc.
304 Inverness Way South, Suite 365
Englewood, CO 80401
Abstract
Emission of HCI and C12 from a variety of sources is currently regulated under the
Clean Air Act. However, there is no commercially available continuous emissions
monitor that can measure HCI and C12 simultaneously. During a Phase I SBIR program
funded by the Department of Energy, we demonstrated a new technique, Flame
Chemiluminescence Emission (FCE), that is able to measure HCI and C12 simultaneously
at sub-ppm concentrations. The FCE technique is simple and the hardware required to
perform the measurement is inexpensive and robust facilitating its use for the
development of an HC1/C12 CEM. In addition, the technique is intrinsically sensitive to
all chlorinated compounds which will allow its use for the detection of chlorinated
solvents and other hazardous species. During Phase I, we demonstrated a detection limit
below 250 ppb for both HCI and C12; improvements hi design to be implemented during
Phase II will lower the detection limit below 100 ppb. Interference from common flue gas
constituents amounted to no more than 1 ppm for realistic SO2, NO2, and NO
concentrations. A carbon dioxide concentration of 15% (150,000 ppm) was also tested
for interference and showed a minor negative bias equivalent to approximately 1.5 ppm
of HCI. Oxygen (210,000 ppm) and H2O (30,000 ppm) did not exhibit any spectral
interference; however oxygen did exhibit predictable quenching behavior. During Phase
II, a calibration technique will be devised to correct for these slight interferences.
-------
Introduction
HC1 and C12 are among the 189 hazardous air pollutants (HAPs) identified
by the Clean Air Act Amendments (CAAA). Sources of HC1 that will be
regulated under the CAAA include hazardous waste incinerators, coal-fired
boilers and municipal solid waste incinerators (MSWs). HC1 produces negative
environmental effects such as acid rain and negative health effects caused by the
inhalation and subsequent attack of tissues. Consequently, there is a significant
need to measure and control HC1. Continuous monitors that are currently
available for measuring HC1 are either very expensive (e.g. FTIR instruments), are
sensitive to the amount of water vapor in the sample, (e.g. gas filter correlation
technique), or simply are not accurate.
Chlorine, C12, is a strong oxidant and photochemically active. Free Cl
atoms produced via photolysis can catalytically destroy ozone. The amount of C12
emanating from sources is currently a matter of some controversy. However,
recent results from cement kiln hazardous waste combustors1 indicate that the HC1
concentration is approximately a factor of 100 greater than the C12 concentration;
whereas for hazardous waste incinerators, the HC1 concentration is only a factor
of 10 greater than the C12 concentration. For LWA kilns, the ratio is 1000:1. For
other sources such as coal-fired power plants, data of this caliber is not yet
available. Although the relative concentrations of HC1 and C12 being emitted from
a particular source may not be known, concern over greenhouse gases and ozone
depletion may cause C12 from these sources to be regulated under the CAAA as
well.
ADA has successfully developed a new technique for detecting gas-
phase, chlorinated species that can be applied to the simultaneous
measurement of HC1 and C12 in flue gas. This technique, which we refer to as
Flame Chemiluminescence Emission (FCE), converts the chlorine in the sample
into a compound that emits brightly in a hydrogen or hydrocarbon flame
excitation source. A CEM based upon FCE will provide a measure of total
chlorine in a sample regardless of chemical form including HC1, C12, and any
chlorinated organics. HC1 and C12 can be measured independently by
incorporating a switching sample interface in the instrument which will provide a
scrubbed, HC1 - or C12 - free sample to the analyzer or one that converts HC1 to
Cl, or C12 to HC1. The difference between the total chlorine and
scrubbed/converter measurements provides an HC1 and C12 concentration, since
the organic chlorine fraction resulting from a properly operating incinerator is
known to be small. However, Phase I experiments focused on demonstrating the
required detection sensitivity for total chlorine content (HC1 and C12) without
speciation and in the presence of typical flue gas constiuents. During Phase I, we
demonstrated a detection sensitivity of less than 250 ppb for both HCl and Cip
well below the range necessary for envisioned applications. For comparison, a
controlled HCl emission source produces from 10 - 100 ppm HCl.
-------
Continuous data from an HC1/CL, CEM will provide many benefits to users. First,
continuous data are necessary to define and isolate causes of pollution, and therefore
develop cost-effective solutions to pollution problems. Second, continuous data are
necessary to provide performance assurance to the public, regulators, and workers that
processes are operating safely. And third, continuous data can be used to avoid capital
costs by optimizing a system to avoid extra pollution control equipment when complying
with air regulations. Without continuous data, facilities must rely solely on up-front
waste characterization and periodic effluent sampling to estimate emissions and therefore
compliance.
Advantages of the FCE technique are due to the following characteristics:
•FCE is specific to Cl; it does not suffer from interferences caused by other flue
gas constituents such as SO2, CO2, NO2, and NO.
•FCE can be instituted as a simultaneous monitor of both HC1 and C12.
•FCE can be implemented with inexpensive components resulting in a cost-
effective instrument.
Competitive Technologies
Unlike HC1, C12 can not be detected using infrared absorption, so FTIR
and the gas filter correlation technique are not options for simultaneous
measurement of HC1 and C12.2 There exists a commercially available ion mobility
spectrometer that is capable of detecting HC1 and C12 with two separate
instruments and appropriate flue gas pre-conditioning. However, the ion mobility
spectrometer has never been field tested on flue gas for the detection of HC1 and
C12. In contrast, the ADA Technologies HC1/C12 CEM is being developed and
will be field-tested at a variety of sites. In addition, ion mobility spectrometers
traditionally use a radioactive ionization source, typically Ni63 foil, a beta emitter
with a half life of 93 years. It has been shown that the Ni63 can react with acid
gases to form mobile, radioactive Ni compounds that can be released into the
environment. Research on alternative atmospheric pressure ionization sources is
progressing, but none are commercially available, to our knowledge. Ion mobility
spectrometers have poor resolution for low molecular weight species such as HC1
and C12. Higher resolution requires a long drift path which in turn requires a
large, bulky instrument. Finally, an HC1/C12 ion mobility spectrometer will cost
approximately $30,000 without the sample preconditioning system which costs
approximately $10 k. We believe that the FCE HC1/C12 CEM will be at least a
factor of two cheaper.
-------
Details of the FCE Analyzer and Experimental Results
An FCE CEM provides a measure of total chlorine in a sample
regardless of chemical form including HC1, C12, and any chlorinated
organics. We have demonstrated in the Phase I program sensitivity to the 250
ppb range, well below regulated levels of HC1 (25 ppm). The Phase I experiments
also demonstrated detection of C12 by FCE for the first time. FCE consists of a
simple process by which all of the chlorine in a sample is converted to a highly
emissive chlorine-containing molecule. This molecule is subsequently excited to
emit light in a hydrogen flame. Emission from the flame is collected, color
filtered, and transmitted to a detector as described below.
A two-lens light collection system was employed consisting of 3-inch
focal length, 2-inch diameter lenses placed 3 inches from the flame and detector,
respectively. Thus the f # of the collection system was 1.5. In the collimated
section between the two lenses was placed a filter wheel consisting of two pairs of
identical 2-inch diameter interference filters. The filters were located in an
alternating configuration so that, as the wheel turned, 540 run light was sampled
by the detector followed by a background measurement at 400 nm. With the two
sets of filters, the signal versus background measurement was made twice for each
rotation. A diode pickup on the wheel produced a synchronous signal for the
custom-built, lock-in detection circuitry. The detector was a Hamamatsu mini-
PMT, model H5784-03 which combines a high voltage power supply, the PMT,
and a current-to-voltage preamp in one small package. The detector signal was
sent to a 100-gain voltage amplifier and on to an Analog Devices AD630
demodulator chip. The sync signal from the wheel allows the chip to produce a
DC voltage equal to 1/2 of the difference between the on- and off-resonance
signal. This voltage was sampled at a 1 Hz rate by a 4-channel 12-bit Blue Earth
microcontroller. The microcontroller performed a ten-sample moving average
and saved the resultant values to a file. The data was analyzed and graphed using
a simple spreadsheet program.
The entire apparatus was enclosed in a blackened, light-tight enclosure to
prevent spurious signals from reaching the detector. Fumes from the flame were
vented through a flexible duct. The ambient temperature inside the box was
monitored during each experiment.
Initial testing focused on determining detection limits for HC1 and C12.
Calibrated flows of HC1 and C12 were produced by flow dilution of standard HC1
(1031 ppm) and C12 (100 ppm) mixtures in nitrogen. The total flow was
maintained at a constant 2.0 slpm by reducing the nitrogen flow as the HC1 or C12
flow increased. In this way, concentrations ranging from 0 ppm to 50 ppm of HC1
and C12 in nitrogen were produced. The mass flow controllers were calibrated
using a bubble flow meter prior to use in Phase I testing.
-------
During a routine test, the torch burned with a rich mixture of hydrogen/oxygen at
a total flow rate of 1.91 slpm with 1.55 slpm of the total flow being H2. These
flows were chosen as a compromise between increased signal and increased flame
flicker noise at low or no oxygen flow and low signal and low noise for
stoichiometric flames. After a brief warmu-up time, HC1 or C12 was introduced
and the emission signal was monitored. Various concentrations of HC1 and C12
were added and the response of the detector was recorded as a function of time.
Representative data are given in Figure 1 for HC1 and Figure 2 for C12. One point
of interest is that the amplitude of the noise decreases for lower signal levels.
This is readily seen in both the HC1 and C12 data. Although not understood at this
point, this fact allows extremely low levels of HC1 and C12 to be detected with
relative ease.
HCL TEST 2/28/97A
o
a.
co
0.5 . .
25.1 PPM HCL 12 7PPM HCL 7 7PP(* HCL 5.1PPM HCL 2.1PPM HCL
1000
TIME (SEC)
SYS10SAMPAVG(V)
Figure 1: Detector response versus HC1 concentration.
-------
CL2 TEST 3/05/97B
Figure 2: Detector response versus C12 concentrations. The detector gain is the
same as in Fig. 1. Note that the noise decreases as the concentration decreases
allowing measurement of very low concentrations.
Calibration plots resulting from this data are given in Figures 3 and 4. A
linear fit to the data is shown which will be adequate for most applications.
Figure 5 shows a comparison of the instrumental response to HC1 and C12. As can
be seen immediately from the data, the response to C12 is greater by a factor of
approximately four. This difference in response can be easily explained using a
simple model. A quadratic fit to the data is shown hi the figure for comparison to
the linear fits in Figures 3 and 4. The coefficients for the quadratic fit are given in
Table 1. Note that the linear terms for the quadratic fit differ by almost exactly a
factor of four as seen for the linear fits in Figures 3 and 4.
-------
HCL TEST 3/05/97A
^Z
x"
r
PMTGAJN
X
1 9
x
X
SYS=.045(H
R2=.997
3 5 10 15 2
HCL CONCENTRATION (PPM)
X
L)-.049
!
3 25 30
g| HCL DATA [--j HCL FIT
Figure 3: Calibration curve for HC1 using the data from Fig. 1. A linear fit to the
data is shown for comparison. The slope of the fit is 0.045 volt/ppm.
CL2 TEST 3/05/97B
YS=.166(CL2}-.
CL2 CONCENTRATION (PPM)
. CL2 DATA r-i CL2 FIT
Figure 4: Calibration curve for C12 using the data from Fig.2. A luiear fit to the
data is shown for comparison. The slope of the fit is 0.166 volt/ppm.
-------
HCL V. HCLx-. CL3y CL2.V
CONCENTRATION (PPM)
Figure 5: Graph showing the data for HC1 and C12 on the same scale. The
system response to C12 is greater than for HC1 by a factor of four although the
shape of the curves is very similar. Also shown is a quadratic fit to the data. The
data points are indistinguishable from the fit.
Coefficients
A
B
C
HCI
3.73 x 1CT1
0.035
-0.004
C12
0.001
0.136
-0.094
Table 1: Coefficients for the quadratic fit to the data shown in Figure 5. The
equation used for the fit is y = Ax2 + Bx + C
The slight non-linearity of the calibration curves and the fact that equal
concentrations of HCI and Cl, produce signals that differ by a factor of four is
easily explained and results from the process used to convert the chlorine in the
sample to the species excited in the flame. We have developed a simple model of
the conversion process which predicts the functional form of the curves
accurately.
-------
Interference Experiments
Another set of experiments was completed to determine the system
response to various constituents of flue gas with a potential ability to interfere
with the HC1/C12 measurement. The constituents studied were NO, NO2, CO2,
SO2, H2O, and O2. None of these species, except for NO2, has an electronic
transition at either the background wavelength (400 nm) or the signal wavelength
(540 nm); however, there are several other mechanisms by which interference
may occur. For instance, any of these species may react hi the converter to form a
species that does emit at either the background or signal wavelength.
Alternatively, the species may interfere by quenching emission from the excited
state.
Each of the six potential interferents was tested individually by adding a
known and realistic quantity to the torch with and without 10 ppm of HC1 present
and observing the system response. The concentrations studied were: 100 ppm of
NO, 25 ppm of NO2, 15% CO2 (150,000 ppm), 500 ppm of SO2, 3 % H2O (30,000
ppm), and 21 % O2 (210,000 ppm). Figure 6 shows that there was no interference
from 100 ppm NO both with and without 48 ppm of HC1. (The sudden drop in
signal at 1900 seconds is unrelated to the NO concentration.) The NO2 data are
not shown, but 25 ppm of NO2 exhibited no interference.
HCLTEST2/26/97C
I
f^^t^
Figure 6: System response to 100 ppm of NO with and without 48 ppm of HC1.
No interference is observed. The slight dip in signal at about 1900 seconds is
unrelated to the NO concentration.
Figure 7 shows the system response to 500 ppm of SO2 with and without
the presence of 48 ppm of HC1. There is a slight positive interference observed
when the HC1 is present as well as the SO2. The origin of the interference is not
definitely known; however, apparently, SO2 and HC1 react to form a species that
-------
emits more strongly at 540 run than at 400 nm. In any case, the effect is quite
small amounting to an equivalent HC1 concentration of about 2 ppm. In addition,
the SO, concentration is very high for real flue gas systems. Due to SO2
regulations, 100 ppm is a more likely value. Finally, we have produced a 100%
(500 ppm) variation in SO2 concentration during this test. In a realistic CEM
application, the SO2 concentration will not vary more than approximately + 50
ppm.
HCL TEST 2/26/97B
5QOPPM J 02 ON
f/
1500 3000
TIME (SEC)
Figure 7: System response to 500 ppm of SO2 with and without 48 ppm of HCL
There is a slight positive SO2 interference amounting to about 1.5 ppm of HC1;
however, it is only present when the HC1 is on. The origin of the interference is
not known.
Figure 8 shows the system response to 15% CO2 with and without 5.1 ppm
of HC1 present. The CO2 shows a small, but definite negative bias amounting to
an equivalent HC1 concentration of about -1.2 ppm. Interestingly, the bias is
present with or without the HC1 which suggests that HC1 is not involved in the
mechanism that produces the effect. Apparently, CO2 reacts in the chlorine
converter to produce an unknown species that emits more strongly at 400 nm than
at 540 nm. In any case, the CO2 interference will not be significant in real
applications. We have tested + 100% fluctuations in CO2 concentrations;
whereas, in realistic situations, the CO2 concentration is expected to fluctuate by
perhaps + 10%.
10
-------
HCL TEST 2/28/97B
Figure S: System response to 15 % CO2 (150,000 ppm) with and without 5.1
ppm of HC1. A negative interference is present that appears to be due to the
formation of an unknown chemical species that emits in the flame more strongly
at 400 nm than at 540 nm. The interference is slight in any case, amounting to an
equivalent HC1 concentration of only -1.2 ppm of HC1.
H2O and O2 interferences were tested in a slightly different way. For O2,
two kinds of tests were performed. First, the flame stoichiometry was altered by
changing the amount of O2 added to observe the effect on signal levels and flame
stability. Basically, it was found that more oxygen reduced the signal but
increased flame stability. In the other type of experiment 21 % oxygen in
nitrogen was added to a 10 ppm mixture of HC1 in nitrogen and added through the
sample injector to observe the effect on signal level. Oxygen added in this way
was also observed to decrease the signal level. In this case, the interference
appears to be excited state quenching; O2 is widely known to be extremely
effective as an excited state quenching agent. The effect is pronounced but not
prohibitive. Adding 21% oxygen to the flow reduced the signal by only a factor
of two. In contrast, flue gas nominally contains only 5 10 % residual O2. As
long as the 02 concentration remains relatively constant which is a goal in most
combustion processes, the effect of O2 will be completely mitigated. As an added
note, it may be possible to use the residual 5-10% oxygen hi the flue gas as the
oxidizer for the hydrogen fuel. In this way the amount of O2 that will need to be
added to the flame will be minimized.
Water did not influence the signal when added as a component of the
fuel/oxidizer mix at 3% concentration; neither did the water resulting from the
combustion process in the flame. Therefore, water is not a spectral interference.
-------
In conclusion, none of the species tested, NO, NOy COa SOy Oy or H2O, will
have a significant effect on analyzer performance in realistic situations.
Sample Transport Issues
Sample transport issues are common to all extractive CEMs. In order to
test potential problems with HC1, we added water to the HC1/N2 mixture at 3%
concentration, and the signal was completely extinguished. The cause of this
effect can not be quenching like that observed for O2 since water has no effect
when added through the torch. Instead, the effect probably involves sample
transport, although this has yet to be definitively demonstrated. In any case,
membrane dryers are known to remove water with nearly 100 % efficiency and to
leave both the HC1 and C12 in the sample flow, untouched. Therefore, whatever
the effect of water might be, it can be removed from the sample without
compromising the HC1 and C12 measurement. Sample conditioning of this type is
common in CEMs and is not considered an impediment to the technological or
commercial potential of the FCE technique.
Other Observations
Glowing particles from the torch are occasionally emitted in the flame.
Most often these particles glow with a bright incandescent color causing a
dramatic instantaneous increase in signal. Occasionally, the particles are
sufficiently hot to emit more strongly at 400 nm than at 540 nm causing a
negative excursion. These instances are far less prevalent, but they do occur. We
eliminated the effect of these "sparks" by implementing a routine in the data
processing software that ignores samples having large positive or negative
excursions from the previous average values. This data treatment worked quite
effectively (Figure 9). The figure shows HC1 data repeatedly introduced at a
concentration of 33 ppm. The data shown have not been averaged for this test
which accounts for the high frequency noise that is not present in the data
presented earlier. The software routine ignores any samples with a positive or
negative excursion greater than 25% of the last average value. As seen in the
data, the large one-sample excursions are removed effectively by the routine.
12
-------
HCL TEST 2/26/97A
PMT GAIN=1.9
33.8PI 'M HCL
33.BPPM HCL
1000 1500
TIME (SEC)
UNFILTERED
FILTERED
Figure 9: Comparison of filtered versus vmfiltered data. Notice the spikes due to
the sparks are largely removed from the filtered data. No averaging is used in
either set of data
References
1. Draft Technical Support Document for HWC MACT Standards Volume II: HWC
Emissions Data Base, U.S EPA pages 2-18 - 2-20 (February 1996).
2. Draft Technical Support Document for HWC MACT Standards Volume IV:
Compliance with Proposed MACT Standards (February 1996).
13
-------
NOTES
-------
NOTES
-------
NOTES
-------
NOTES
-------
NOTES
-------
NOTES
-------
NOTES
-------
NOTES
-------
NOTES
-------
NOTES
------- |