EPRI
Electric Power
Research Institute
August 1997
TR-108683-V3
EPRI-DOE-EPA Combined Utility
Air Pollutant Control Symposium
The Mega Symposium
Particulates and Air Toxics
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
Particulates and Air Toxics
August 25-29, 1997
Washington Hilton & Towers Hotel
Washington, DC
Conference Chairpersons
George Offen, EPRI
Lawrence Ruth, U.S. DOE
David Lachapelle, U.S. EPA
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
High 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
SO2 control Continuous emission monitoring
-------
CONTENTS
Thursday, August 28; 8:00 a.m.; 1:00 p.m.
PARALLEL SESSION A: Particulate Control - New Technologies
Advanced Fine Particulate Control and Implications Toward Solving Multiple
Pollutant Control Problems in Coal-Fired Boilers
Impact of Coal Characteristics and Boiler Firing Conditions on ESP
Performance
Outstanding Improvement of SOx and Particulate Removal Efficiency in Flue
Gas Treatment
Recent Combustrolฎ FACT Flue Gas Conditioning Experience
Cost-Effective Retrofit Technology for Enhanced ESP Performance on
Low-Sulfur Coal
Summary of Wet ESP Operation at NSP's Sherco Station
Testing of a Combined Dry and Wet Electrostatic Precipitator for Control of
Fine Particle Emissions from a Coal-Fired Boiler
Electrostatically Enhanced Core Separator System
Development of the Laminar-Flow Fine-Particle Agglomerator
Advanced Hybrid Particulate Collector, a New Concept for Air Toxics and
Fine-Particle Control
Thursday, August 28; 8:00 a.m.
PARALLEL SESSION B: Air Toxics Control- Mercury Capture by Sorbents:
Lab Scale Research
Combined Mercury and Sulfur Oxides Control Using Calcium-Based Sorbents
Performance of Activated Carbon for Mercury Control in Utility Flue Gas Using
Sorbent Injection
Fixed-Bed Control of Mercury; Role of Acid Gases and a Comparison Between
Carbon-Based, Calcium-Based, and Coal Fly Ash Sorbents
Capturing and Recycling Part Per Billion Levels of Mercury Found in Flue Gases
Novel In Situ Generated Sorbent Methodology and UV Irradiation for Capture of
Mercury in Combustion Environments
A Circulating Fluid Bed Fine Particulate and Mercury Control Concept
-------
Thursday, August 28; 1:00 p.m.
PARALLEL SESSION B: Air Toxics Control-Mercury Capture by Sorbents:
Pilot Scale Studies
Demonstration of Dry Carbon-Based Sorbent Injection for Mercury Control in
Utility ESPs and Baghouses
Mercury Control in Utility ESPs and Baghouses through Dry Carbon-Based
Sorbent Injection: Pilot-Scale Demonstration
Novel Vapor Phase Mercury Sorbents
Thursday, August 28; 3:00 p.m.
PARALLEL SESSION A: High-Gas-to-Cloth Ratio Baghouses
Alabama Power Company E.G. Gaston 272 MW Electric Steam Plant-Unit
No.3 Enhanced COHPAC I Installation
Predicting COHPAC Performance
Performance Response of a COHPAC I Baghouse During Operation with
Normal and Artificial Changes in Inlet Fly Ash Concentration and During
Injection of Sorbents for Control of Air Toxics Emissions
Operation and Performance of COHPAC atTU Electric's Big Brown Station
Thursday, August 28 3:00 p.m.
PARALLEL SESSION B: Air Toxics Control-Mercury Capture by FGD
Factors Affecting Control of Mercury by Wet FGD
Improved Mercury Control in Wet Scrubbing Through Modified Speciation
Mercury Absorption in Aqueous Hypochlorite
Mercury Emissions Control in FGD Systems
-------
Friday, August 29; 8:00 a.m.
PARALLEL SESSION A: Particulate Control-Engineering Studies
Particulate Control for Year 2000 and Beyond for Power Plants
Baghouse Evaluation and Optimization at the Bonanza Power Plant
The Use of Treatment Time and Emissions Instead of SCA and Efficiency for
Sizing Electrostatic Precipitators
Computational Fluid Dynamics Applications for Power Plants
Friday, August 29; 8:00 a.m.
PARALLEL SESSION B: Air Toxics Control-General Topics
Trace Metals Removal From Residual Fuel Oils
The Use of Coal Cleaning for Trace Element Removal
Prediction of Trace Element Partitioning in Utility Boilers
Summary of Key Air Toxic Results From the Center For Air Toxic Metals
(CATM)
Field Validation of Sampling Procedures for the Speciation of Mercury in Flue
Gas
-------
Thursday, August 28; 8:00 a.m.; 1:00 p.m.
Parallel Session A:
Particulate Control - New Technologies
-------
ADVANCED FINE PARTICULATE CONTROL AND IMPLICATIONS TOWARD
SOLVING MULTIPLE POLLUTANT CONTROL PROBLEMS IN COAL-FIRED
BOILERS
John P. Gooch
Southern Research Institute
P.O. Box 55305
Birmingham, AL 35255
and
Charles B. Sedman
National Risk Management Research Laboratory
Air Pollution Prevention and Control Division
Research Triangle Park, NC 27711
Abstract
Recent advances in fine particulate matter (PM) control, including wet electrostatic precipitation
(ESP) and combined ESP/fabric filtration, have demonstrated advanced control of fine particles.
These and other improvements in fine particle control offer intriguing routes to solving other
emissions problems facing the utility industry. These include sulfates, nitrates, and metals, both
solid and vapor which contribute to plume opacity and the formation of secondary particles in the
atmosphere. Current commercial PM control systems are compared. In addition, future advances
in fine PM control are discussed, along with their integration with other gas cleaning technologies,
to alleviate stack opacity problems and reduce secondary particle formation.
Introduction
The potential impact of anticipated new national ambient air quality regulations (NAAQS), the
announced intent of EPA to issue more comprehensive air emission regulations for combustion
sources, and the long range impacts of the Clean Air Act Amendments of 1990 are compelling
reasons to reassess the current and future compliance strategies of electric steam generators. For
these reasons, suppliers of technology should similarly reassess their near-term markets and
potential research and development opportunities.
-------
First and foremost aref the NAAQS revisions to include new limits for particles less than 2.5
micrometers in diameter (PMZS). Many issues of source apportionment and compliance schedules
need to be resolved, and the unraveling could continue for the next decade. Of the controllable
sources of PMZ5, combustion sources are an important industrial sector, and issues center on
costs of control. For controlling primary fine particle emissions, smaller combustion sources are
dominated by fabric filter baghouses, while larger sources are a mix of baghouses and ESPs;
research efforts to improve performance are in progress.1
Secondary particles (i.e., particles that are formed in the atmosphere from gaseous emissions),
may become a primary target of the implementation plans; and would cause further emissions
limits on fine particle precursorssulfur oxides (SOJ, nitrogen oxides (NOJ. Systems designed
to control both primary and secondary fine particles are also being developed and refined.2-3
Revision of the oxidant (ozone) NAAQS would similarly define additional reductions in NOX
emissions from combustors and should factor in compliance and technology strategies.
Mercury emissions from major stationary sources are a continuing concern; the strategy toward
utility boilers is as yet undefined, as regulators and researchers struggle to define the current
capabilities of conventional control technologies toward mercury from coal combustion, and
determine a cost-effective course of action. A possible approach is to include incentives for
mercury control in a future comprehensive regulatory action, as defined in EPA's Clean Air
Power Initiative (CAPI).4 Similar considerations for smaller combustion sources may also be
forthcoming under EPA's Industrial Combustion Coordinated Rulemaking (ICCR), currently
ongoing.5 Both efforts attempt to consider the total environmental costs and benefits of various
compliance strategies, before defining emissions limits.
Visibility impairment from combustion sources has prompted litigation and enforcement actions,
and will also influence future implementation plans and compliance strategies. Near-stack opacity
violations due to combinations of sulfur trioxide, nitrogen dioxide, and fine particles are not
uncommon, and may actually be exacerbated by the use of wet scrubbers for sulfur dioxide
(SOJcontrol and ammonia-based NOX controls. Remote visibility impairment in pristine areas has
been directly linked to SO2 emissions from utility coal combustors.6
Implications For Control Strategies
A logical goal for resolving the plethora of SO/NO/PM^-related issues is to develop a more
comprehensive approach to control, as opposed to the serial controls now in vogue, which include
combustion modifications and ammonia injection, with and without a fixed-bed catalyst, followed
by particle control, followed by SOX control, with perhaps activated carbon somewhere in the
scheme for additional mercury removal.
With all of these additional concerns, and possible new reductions in many emissions categories,
the ideal of a multi-pollutant control system, with one sorbent and one collection box should be
revisited. Refinements of existing serial controls should similarly be considered because of the
large investments already made in these technologies.. Since major efforts are underway to
-------
improve, and perhaps redefine, fine particle control technology, the development of a one system,
one sorbent approach is a logical focal point. Fine particle controls have traditionally focused on
the physical capture of particles, and have generally ignored condensible vapors and gases, as well
as non-condensible gases, which are subject to emissions regulations. Systems which are capable
of operation at lower flue gas temperatures have an inherent advantage in collection of
condensible gases, as well as non-condensible but reactive gases. These systems can further be
modified to serve both as gas absorbers and particle colllectors; fabric filters or wet ESPs have
substantial potential to play this dual role.
The ability to convert gases to particles needs refinement. Gas cooling without expensive
absorbers/humidifiers has been limited to modest water injection systems, primarily for
improvement of dust collection. Better multi-pollutant sorbents which are capable of both acid
gas and metal vapor sorption, and could include catalytic activity for oxidation/absorption of nitric
oxide, need development.7 Combustion modifications which either preferentially produce more
controllable forms of pollutants, e.g. larger particles and/or more oxidized species such as
nitrogen dioxide (NO2) and mercuric chloride (HgCl2), should be investigated further. If technical
progress is made in all the above, in conjunction with advanced fine particle control systems, then
the goal of one sorbent, one vessel, becomes more achievable.
In the following sections, results are presented from fine PM measurements of two commercial
and two pilot-scale emission control systems. Results are discussed in the context of anticipated
new PM regulations, development needs, and opportunities for multi-pollutant control.
Performance Measurements and Comparisons
State-ofthe-art Commercial Systems
The U. S. Department of Energy (DOE) has sponsored a project to perform a "Comprehensive
Assessment of Air Toxic Emissions" from coal-fired electric generating stations. Southern
Research Institute performed sampling, chemical analyses, and mass balances on two large power
stations with modern emission control systems for SOX and PM. The primary focus of these
measurement programs was the quantification of hazardous air pollutant (HAP) emission rates;
results have been previously reported.8'9'10 The following re-examines the particulate
measurements obtained at these two power stations with an emphasis on the fine particulate
fraction and the performance of the control systems on such emissions.
A dry flue gas desulfurization (FGD) system with three spray dryer modules (Joy/Niro design) in
parallel, followed by two baghouses in parallel, is operating on the Springerville Generating
Station Unit 2 of Tucson Electric Power Company. The Unit was operated at its maximum
capacity of 422 megawatts (MWe) gross electrical output during the test series, which was
performed in June 1993. Coal burned at the plant is subbituminous and is obtained from the Lee
Ranch Mine in New Mexico. Average sulfur and ash contents for this coal are 0.7 and 19%,
respectively. The boiler for Unit 2 is a Combustion Engineering (CE) corner-fired design with
-------
balanced draft and overfire air for reduction of NOX. Figure 1 illustrates the arrangement of the
spray dryer-baghouse system at Springerville Unit 2.
The second plant, the Bailly Station of Northern Indiana Public Service, is the site of a Clean
Coal Technology Project to demonstrate the Pure Air Advanced Desulfurization Process. Units 7
and 8 of the Bailly Station (175 and 345 MWe gross capacity, respectively), are Babcock &
Wilcox (B&W) cyclone boilers, each equipped with an ESP.. Flue gas from both units is
combined to form a single stream which is treated by the wet scrubber. Coal burned at both units
is an Illinois/Indiana high sulfur bituminous coal, with sulfur contents ranging from 2.5 to 4.5%.
Figure 2 illustrates the arrangement of the boiler and emission control system for Bailly
Generating Station Units 7 and 8. Average electrical output for the units during the test program
in September, 1993, was 511 MWe.
Table 1 presents a summary of PM concentration data obtained at the Springerville and Bailly
test sites. The spray dryer/baghouse system was extremely effective at PM removal, with overall
collection efficiency, based on inlet fly ash concentrations, of 99.9%. If the sorbent injected in the
spray dryers were considered in the efficiency calculation across the baghouse, the efficiency
would exceed the 99.90% value given in the table.
At the Bailly site, the interpretation of PM measurements in the stack was complicated by the
formation of sulfuric acid mist in the scrubber system and by varying amounts of ammonia
injection to control acid mist formation and plume appearance. However, the analytical work
performed suggested that about 75% of the stack PM emissions at Bailly consisted of sulfuric acid
aerosol, with the remaining 25% fly ash from the boilers. If the estimated fraction of condensed
acid is subtracted from the stack mass concentration, the fly ash concentration drops to 13.6
mg/DNm3, and the apparent fly ash collection efficiency across the system increases to 99.71%.
Note that the term "apparent" is used for the efficiency results because, for both types of systems,
calcium-containing sorbents are introduced downstream from the inlet measurement locations.
Although both systems are characterized by relatively high overall mass collection efficiencies, the
particle size dependent performance of the systems differs markedly Figure 3 illustrates PM
collection efficiency of the baghouse, along with the overall (spray dryer/baghouse) system at
Springerville, calculated using the ratio of the incremental mass concentrations (dM/dlogD) across
the baghouse and across the entire system. The overall system efficiency is lower because of the
high concentration of solids injected in the spray dryer, thus increasing the particulate
concentration at the baghouse inlet. The baghouse, however, exhibits the characteristically high
values of collection efficiency in the sub-two micrometer range, with no fractional efficiency value
lower than 99%.
Figure 4 illustrates the penetration ratios (defined as (100-efficiency)/100) across the Bailly Unit 8
ESP and across the scrubber downstream of the two precipitators. The Unit 8 ESP, which
treated about 65% of the flue gas entering the scrubber, was performing well, with an overall
mass efficiency of 99.82%, and an outlet mass concentration of only 9 mg/DNm3 The Unit 7
ESPwas not performing as well (outlet =70 mg/DNm3) due to fields out of service. Although the
-------
calculated fractional efficiencies for the ESP exhibit considerable scatter in the 1 to 2 micrometer
diameter range, the niinimum collection efficiency in this size band is about 96%. In contrast, the
penetration ratios across the scrubber in this size band are greater than 1.0, reflecting the
generation of fine particles across the scrubber.
Nominal ammonia injection rates on Units 7and 8 were 15 ppm by volume, but the ammonia
supply was depleted during the test series, and no ammonia was injected to the ESPs on the last
day of stack sampling. If the ammonia formed ammonium sulfate particles upstream of the ESPs,
little or no sulfuric acid vapor would be expected to be available to condense in the scrubber.
However, if ammonia were present in excess, it is possible that ammonium sulfite particulate
could form hi the scrubber.
Analytical data on stack solids suggest that insufficient calcium is present in the stack PM for the
solids consist of calcium compounds, but ammonia determinations were not performed. As a
result of the depletion of the ammonia supply, sulfuric acid aerosol is the most probable source of
the PM produced in the scrubber.
In view of the recently announced fine PM regulations, it is of interest to examine control system
performance on a cumulative mass basis as a function of particle size. Figure 5 presents
cumulative mass concentrations measured in the stack at Springerville and Bailly in terms of
mg/DNm3 corrected to 3% O2. The graph illustrates that the cumulative mass at Bailly below 2.5
micrometers diameter exceeds the value measured at Springerville by a wide margin (15 vs 0.7).
This difference results from both the particle generation in the scrubber and from the high fine
particle collection efficiencies achieved by the baghouse at Springerville. Coincidentally, the total
mass concentrations measured by the impactor sampling systems in the stack at Bailly and
Springerville were both approximately 18 mg/DNm3. (Since impactor systems are constrained to
sample at a rate isokinetic to the average stack velocity, these concentrations may not compare
well with EPA Method 17 or Method 5 total mass concentrations.)
Also shown on Figure 5 is a hypothetical cumulative concentration curve for Bailly which would
be expected if a small wet ESP were installed downstream from the scrubber. These calculated
results were based on pilot-scale experiments performed at Southern Research which are being
presented at this conference in another paper.11 This hypothetical configuration would reduce the
cumulative mass below 2.5 micrometers from 15 to 2.5 mg/DNm3, which is a reduction of
approximately 83%. As a result of acid mist formation during the cooling process, the pilot-scale
wet ESP was collecting sulfuric acid aerosol during the experiments when fractional efficiencies
were measured, so the results should be applicable to the circumstances at the Bailly scrubber
exit.
Additional insight into overall system performance can be obtained by examining cumulative mass
efficiency as a function of particle size. Figure 6 presents the results obtained by calculating
cumulative mass efficiencies from the cumulative mass distributions at the control system inlet and
outlet sampling locations. Again, it is appropriate to use the term "apparent" cumulative mass
efficiency since calcium-containing sorbent PM is introduced downstream of the inlet sampling
-------
locations at both plants. Note that the overall system sub-2.5 micrometer collection efficiency at
BailJy is about 90%, even though particle generation occurs in the scrubber. This results from the
relatively high collection efficiencies obtained by the ESPs and the feet that the mass of the fine
PM formed in the scrubber is relatively small compared with the mass in this size range exiting the
boiler.
Secondary Particulate Issues
The preceding discussion has dealt exclusively with primary PM. However, ambient fine PM will
be a mixture of particles which form from atmospheric reactions as well as primary particles
exiting combustion sources. In general, the mass of SOX and NOX in power plant flue gas is much
greater than the paniculate mass exiting from high efficiency ESPs and fabric filters. Since fine
particle concentrations are available for these well-controlled combustion sources, it is of interest
to compare these values with potential secondary ambient concentrations resulting from SOX and
NOX on a stack-equivalent basis.
Table 2 compares PM2 5 measurement results performed with impactors on the indicated sources
with: 1) mass concentrations of sulfetes resulting from the combustion of a Phase I compliance
coal, and 2) the mass of nitrates corresponding to 0.45 Ib NO^/106 Btu (194 ng/j). A recent
report dealing with this subject12 speculates that about 10% of SOX and 5% of NOX are converted
to sulfate and nitrate fine PM. Ammonium sulfate and ammonium nitrate were used here as a
basis of calculation. Note that, even with only 5 and 10 % PM formation from the precursors, the
sum of the nitrates and sulfates would be over a factor of 25 times greater than the primary fine
PM emissions from Bailly on a stack equivalent basis.
This comparison emphasizes the importance of an integrated approach in considering additions to
an emission control system which may be needed as a result of the forthcoming fine PM
regulations. Another point which should be made is that, although the sulfuric acid condensation
which has been observed in scrubbers treating flue gas from high sulfur coals can cause significant
plume opacity problems, the acid mist is a small fraction of the mass of SO, and the potential
secondary sulfate PM which is removed by the scrubber. The SO2 removal efficiencies at
Springerville (0.7% S coal) and Bailly (2.5-4.5% S coal) were 60% and 93%, respectively.
Metals Removal Efficiencies
Since the major goal of the DOE sampling projects at Springerville and Bailly was to quantify
emissions of HAPs, it is of interest to compare efficiencies of metals removal for these two
systems. The metals of interest and the efficiencies of removals across the two systems are
presented in Table 3. Note that the efficiencies shown for Bailly were calculated from the
-------
fractional penetrations across the Unit 8 ESP and the scrubber. This should be more
representative of a system of this type (high efficiency ESP + wet scrubber) because the Unit 7
ESP performance was low due to fields out of service. Antimony and selenium removals are not
reported for Bailly due to analytical problems which are discussed in detail in the report.12
In general, a high efficiency of particle removal will equate to a high efficiency of trace element
removal because most of the trace elements will be associated with the paniculate phase. The
exceptions are mercury, boron, and selenium, which have significant volatilities at stack gas
temperatures.
Since selenium is normally ranked with mercury as a volatile element, the high capture efficiency
found at Sprmgerville was not expected. A possible explanation is that the selenium was present
as an acidic selenium oxide, which was in the vapor state and susceptible to capture in the spray
dryer. However, the data obtained at the spray dryer inlet indicated the percentage of selenium in
the vapor state ranged from 0.7 to 23% prior to the addition of the alkaline sorbent in the spray
dryer.
Boron is most likely to occur in the vapor state as boric acid; therefore, reaction with lime and
removal to the solid phase is expected to occur.
At Springerville, the more reliable mercury data were obtained from carbon trap sampling. These
data indicated spray dryer inlet concentrations in the range of 7.21-7.88 micrograms/Nm3.
Corresponding values at the stack were in the range of 4.92-6.02 micrograms/Nm3, for an
apparent removal of 22%.
At Bailly, the carbon trap method indicated an average mercury removal of-5% across the Unit 8
ESP, +8% across the Unit 7 ESP, and +53% across the scrubber. Since there is no reason to
expect significant removal of mercury across the ESP, the overall system removal is about 50%,
based on the carbon trap data.
Redinger and Evans13 have reviewed recent efforts concerning mercury speciation and emissions
control in FGD systems, and has concluded that 50% removal is representative of a minimum
average baseline for existing scrubbers. Higher values have been achieved in both wet and dry
FGD systems.
The different values observed for the efficiency of mercury removal at Springerville and Bailly
may be caused by differences in the relative amounts of elemental vs oxidized mercury which exist
in the flue gas at various points in the flue gas cleaning system. Work is currently underway to
provide more reliable data on mercury speciation and to improve the fraction of mercury captured
in FGD systems.13
-------
Navel Systems
Although both the Springerville and Bailly emission control systems provide effective reduction of
PM and SO2 emissions, both types of systems have certain inherent disadvantages. The wet
scrubber has the problem of acid mist formation which requires injection of ammonia to improve
plume opacity. Addition of other basic reactants upstream of the ESP in ESP/scrubber systems
and humidification have also been evaluated as potential solutions to the plume opacity problem
The spray dryerftag house system provides excellent particulate and trace element control (with
the exception of mercury), but the reagent costs generally make such systems uneconomical for
high sulfur coals.15
Dry scrubbing systems, when combined with a properly designed fabric filter system, generally
provide better overall particulate and trace element removal than ESP/wet scrubber combinations.
The absorptive and particulate collection properties of calcium-containing filter cake preclude the
formation of acid mist in the stack. As a result, significant development efforts are underway to
improve the economics and performance of the dry scrubbing/filtration processes.
An advanced dry collection system which includes boiler injection of limestone, on-site
preparation of lime for use in a dry scrubber, and a fabric filter operating at relatively low
temperature is being developed by B&W as a part of DOE's Combustion 2000 program.16 Asea
Brown Boveri (ABB), also as a part of the Combustion 2000 Project, is developing another dry
process in which the process of gas cooling and SO2 removal is integrated into the fabric filter.2
Still another advanced dry process which has been employed in Europe is the Reflux Circulating
Fluid Bed Scrubber.17 This system reportedly has the capability for removing SO2 at an efficiency
up to 97% with competitive system cost. A key feature of the system is the ability to use
relatively low cost forms of calcium oxide instead of the more expensive hydrated lime that was
used in previous applications of circulating fluid bed scrubbers.
Cumulative mass efficiency data as a function of aerodynamic particle diameter have been
calculated for pilot-scale test performed by Southern Research on a Compact Hybrid Particle
Collector (COHPAC) II pilot-scale test unit located at Alabama Power's Miller Electric
Generating Plant near Birmingham, Alabama.18 This demonstration unit consisted of a two-field
ESP operated at about 188 ft2/! 000 acfin, followed by a pulse-jet baghouse section which was
operated at an air-to-cloth (A/C) ratio ranging from 10.4 to 11.2 ft/min(3.0 to 3.4 m/tnin), and
with average tubesheet pressure drops ranging from 4.6 to 6.6 niches H2O (1.2 to. 1.6 kPa).
These calculated cumulative efficiencies are presented in Figure 7. The graph illustrates that the
ESP/fabric filter combination is extremely effective in controlling PMZ5, with a cumulative mass
efficiency greater than 99.5%
Also in Figure 7 are cumulative efficiencies calculated from one of the wet ESP experiments.11
These efficiencies are considerably lower than the results from Springerville and the COHPAC II
-------
tests, but it should be noted that the wet ESP is a small unit designed to be used as a "polishing
device" to be placed behind an ESP/wet scrubber combination or a dry ESP which is not meeting
current emission requirements. Referring to Figure 5, the wet ESP would increase the overall
PMi5 collection efficiency at Bailly from 90 to 98.3%, based on the fractional efficiencies
measured during the pilot-scale test program.
Development Needs
For retrofit of existing systems, the most difficult situation involves systems with little or no gas
absorption opportunities, i.e., a low-sulfur coal application with ESP control. A COHPAC or
Electrically Stimulated Fabric Filter (ESFF) addition with sorbent injection upstream should
dramatically improve multi-pollutant control capability. Figure 8 illustrates these concepts.
For systems having a low-sulfur coal, fabric filter controlled system, sorbent injection upstream of
the fabric filter (Figure 9 ) is a simple improvement. For a more robust multi-pollutant system,
conversion of the baghouse entry from bottom to top-side with diffusion baffles to establish
uniform gas flow, enhances the likelihood that sorbent particles will be utilized more efficiently, as
shown in Figure 9b . The relocation of the baghouse entry then frees up additional vertical space,
allowing the inclusion of a vertical fluid bed reactor for sorbent injection, gas/solids mixing, and
gas cooling, shown in Figure 9c . This concept, along with more advanced multi-pollutant
sorbents, represents an opportunity for high performance, multi-pollutant control. This system
could be added downstream of an ESP, at a higher cost, and ,therefore, would be more attractive
to industrial applications..
For existing wet scrubbers with either ESP or baghouse upstream (Figure 10), improvements in
condensible fine PM and mercury control will require further study. Sorbents injected upsteam of
the ESP- scrubber combination will have limited residence time to remove sulfur trioxide (SO3) or
mercury vapor. Sorbents injected upstream of the fabric filter/scrubber system will have more
solids residence time, but at less favorable gas temperatures for SO3/mercury vapor sorption.
Sorbents injected in the more favorable conditions downstream of the scrubber would require an
additional solids removal step. Integration of wet ESP downstream of the scrubber could solve
the plume opacity problem, but not enhance mercury or acid gas capture significantly.
Spray dryer/fabric filter systems are few in number, but potentially represent the easiest
conversion, accomplished by changing to a multi-pollutant sorbent from soda ash or slaked lime.
Opportunities for Development
The new PMZ5 regulations will cause utilities operating in an increasingly deregulated
environment to search for the most economical approach to meet any new emission limitations
which may be imposed on specific plants. The wide variations in local conditions and system
configurations should present opportunities to develop site-specific applications to minimize
-------
costs, even though the basic processes for SO^ NO^, and PM control are considered to be
relatively mature technologies. There is also a development opportunity caused by the need for
economically feasible sorbents which can effectively capture mercury in both wet and dry FGD
systems.
The opportunities for development of multi-pollutant control technology are probably as many as
there are available pilot facilities which have coal-firing capability and modem particle control
technology. The discussion here is limited to those facilities and sponsors which have ongoing
fine PM, acid gas, and mercury control programs readily adaptable to the concepts
aforementioned.
Several pilot-scale efforts are underway to develop improved, low-cost technologies for fine
particle collection which can be used in an integrated control system. These include the
Electrostatically Enhanced Core Separator (LSR Technologies) and an advanced hybrid
paniculate collector which combines features of an ESP and a baghouse.19
B & W, with significant support from DOE's-Federal Energy Technology Center (FETC) and the
Ohio Coal Development Office, has facilities capable of investigating many of the above concepts.
Their ongoing mercury/ air toxics efforts are evidence of their capabilities, and their facilities offer
a range of existing control technologies for evaluation and upgrade.13
Southern Research Institute facilities and programs, with support from thr Tennessee Valley
Authority (TVA), Southern Company Services (SCS), and EPA, designed to evaluate advanced
sorbents in conjunction with advanced fine PM controls, such as the modified fabric filter with
fluid bed solids mixing described previously, as well as ESFFs and wet ESPs.
EPA/APPCD has Cooperative Research Agreements with the Institute of Clean Air Companies
and Hosokawa Mikropul Environmental Systems to develop multi-pollutant sorbents and
integrate with advanced fine PM controls. Progress in these areas is the subject of a separate
paper at this meeting.20
Field pilot facilities remaining from Clean Coal Technology Demonstrations and EPRI field
demonstrations are still available. The COFfPAC system at SCS's Miller Station is an example.
It is an expressed desire to jointly plan these and other future programs among EPA, DOE, and
users/vendors, such that the best systems are developed and supported from bench to commercial
demonstration. The authors by this paper invite the participation of all potential users of multi-
pollutant technology to join in this effort. In this manner, the most cost-effective and
flexible 11111 systems are likely to emerge.
Conclusions
1. The rapid development of new fine PM technologies provides new opportunities for an
integrated, multi-pollutant control system.
10
-------
2. Development of new sorbents with acid gas and metal vapor sorption capacities, and perhaps
with enhanced catalytic activity for NOX control, is vital.
3. Augmentation of PM technologies for better gas/solids contact and lower gas temperatures is
necessary for satisfactory sorbent utilization and optimum cost-effectiveness.
References
1. W. Marchant, et al., "Advances in Fine Particle Control Technology," presented at the
Ukraine-United States Technology Meeting, Kiev, Ukraine (September 10-11, 1996).
2. J. W. Regan, "ABB's LEBS Technologies: Practical Solutions for Controlling Air Emissions,"
presented at the 22nd International conference on Coal Utilization and Fuel Systems,
Clearwater, Florida (March 17-19, 1997).
3. U. S. Patent 5,601,791, "Electrostatic Precipitator for the Collection of Multiple Pollutants,"
(February 11, 1997).
4. "EPA's Clean Air Power Initiative," Office of Air and Radiation, U.S. Environmental
Protection Agency, Washington, D.C. (Revision 2, June 1997).
5. "Industrial Combustion Coordinated Rulemaking, Organizational Structure and Process,"
Industrial Combustion Coordinated Rulemaking Federal Advisory Committee, (Revision 2,
June 1997).
6. Modeling the Incremental Impacts of Sulfur Dioxide Emissions from the Hoyden and Craig
Power Plants on Mount Zirkel Wilderness Air Quality and Visibility. Denver, CO, U.S.
Environmental Protection Agency Region 8, November 7, 1996.
7. C.B. Sedman and W.P. Linak, "An Investigation of Multi-Pollutant Sorbents and Sorption
Mechanisms", Internal Grant Proposal, Research Triangle Park, NC U.S. Environmental
Protection Agency (May 27, 1997).
8 Springerville Generating Station Unit No. 2. Birmingham, AL, Southern Research
Institute, June 21, 1994. SRI-ENV-94-476-7960.
9. Bailly Station Units 7 &8 andAFGD ICCT Project. Birmingham, AL, Southern Research
Institute, October 1994. SRI-ENV-94-827-7960.
10 G.F. Weber, et al., "A Summary of Utility Trace Element Emissions Data from the DOE Air
Toxics Study," presented at the EPRI/DOE Conference International Conference on
Managing Hazardous and Particulate Air Pollutants, Toronto, Ontario, Canada (August 1995)
11
-------
11. L.S. Monroe, et al. "Testing of a Combined Dry and Wet Electrostatic Precipitator for
Control of Fine Paniculate Emissions from a Coal-Fired Boiler," presented at the EPRI-DOE-
EPA Combined Utility Air Pollutant Symposium, Washington, D.C. (August 1997).
12. Review of the National Ambient Air Quality Standards for Paniculate Matter: Policy
Assessment of Scientific and Technical Information. Washington, D.C.:U.S. Environmental
Agency, July 1996. EPA-452/R-96-013 (NTIS PB97-115406).
13. K.E. Redinger and A.P. Evans. "Mercury Speciation and Emissions Control in FGD
Systems," presented at the 22ฐ(J International Technical Conference on Coal Utilization and
Fuel Systems, Clearwater, Florida (March 1997).
14. J.R. Peterson, A.F. Jones, and F.B. Meserole. "SO3 Removal from Flue gas by Sorbent
Injection-EPRI HSTC Phase E Tests," in Proceedings: 1993 SO2 Control Symposium, Vol. 3,
EPA-600/R-95-015c, (NTIS PB95-179248), (February 1995).
15. R.J. Keeth, P.A. Ireland, and P.T. Radcliffe. "Economic Evaluations of 28 FGD Processes,"
in
Proceedings: 1991 SO2 Control Symposium, Vol. 1, EPA-600/R-93-064a (NTIS PB93-
196095), (April 1993)
16. D.A. Madden and W.F. Musiol. "Enhanced Limestone Injection Dry Scrubbing (E-LIDS)
Development as Part of B&W's Combustion 2000 LESS," presented at The 22nd
International Conference on Coal Utilization and Fuel Systems, Clearwater, Florida. (March
1997)
17. R.E. Graf, et al. "Commercial Operating Experience with Advanced-Design, Circulating Fluid
Bed Scrubbing," in Proceedings: 1993 SO2 Control Symposium, Vol. 2, EPA-600/R-95-
015b, (NTIS PB95-179230),(February 1995).
18. W.A. Harrison, et al., "Pilot Scale Demonstration of a Compact Hybrid Particulate Collector
(COHPAC) for Control of Trace Emissions and Fine Particles From Coal-Fired Power Plants,"
presented at the 22nd International Conference on Coal Utilization and Fuel Systems,
Clearwater, FL. (March 1997).
19. T.J. Feeley and L.J. Ruth. "The U.S. Department of Energy's Advanced Environmental
Control Technology Program," presented at the 22nd International Conference on Coal
Utilization and Fuel Systems, Clearwater, FL. (March 1997).
20. B. Ghorishi and C. Sedman, "Combined Mercury and Sulfur Oxides Control Using
Calcium-Based Sorbents," presented at EPRI-DOE-EPA Combined Utility Air Pollutant
Symposium, Washington, D.C. (August 1997).
12
-------
Table 1
Springerville and Bailly Participate Data (g/DNm3)
and Apparent Mass Collection Efficiency
(Reference Conditions: 298.15ฐK, 3% 02
Springerville Bypass A Bailly Unit 8 Inlet
Inlet 21.83 4.631
Stack 0.0211 0.0543 (75% sulfate)
Stack w/o sulfate NA 0.0136
Efficiency, % 99.90 98.83
Efficiency w/o sulfate, % NA 99.71
NA Not Available
Table 2
Primary and Secondary PM2S Concentrations
Stack Equivalent Basis
Source Concentration
(mg/DNm3)
Bailly 15
Springerville 0.7
High Eff. Dry ESP 1
SO2 as (NH4)2S04 3,230
(from 0.6% S coal)
10% of above 323
NOxasNH4NO3 1,270
(00.1 lbNOx/106Btu)
5% of above 64
13
-------
Table 3
Trace Element Removal Efficiencies, %
Element
Antimony
Arsenic
Barium
Beryllium
Boron
Cadmium
Cobalt
Chromium
Copper
Lead
Manganese
Mercury
Molybdenum
Nickel
Selenium
Vanadium
Bailly Unit 8 +
Scrubber
NA
99.34
99.97
99.99
91.56
99.74
99.92
100.0
99.89
99.94
99.94
50.
99.69
99.84
NA
99.96
Springerville SDA +
Baghouse
99.3
99.9
99.95
>99.96
90.5
99.99
>99.91
99.99
99.91
99.4
99.8
22.0
98.1
>99.94
>99.96
99.96
NA Not Available
14
-------
Air Heaters
Manifold [
Dry FGD Modules
Fabric Filters
Stack
Figure 1. DuctAn'angementatSpringerville
15
-------
ESP
vvvv
Legend
Sampling Location
ESP
vwv
ESP
ww
FGD
To Slack
BoilerS
Figure 2. Process Flow Diagram and Gas Sampling Locations for Baily Generating Station Units 7 & 8
-------
99.99
99.95
99.90
U
jj= 99.50
T5
ง 99.00
ts
u
s
U- 98.00
95.00
90.00
0.10
ฎ"
' Baghouse ., ' '
-<4
Cf
System
-i' ini| i r-r-
1.00 10.00
Aerodynamic Diameter, Microns
100.00
Figure 3. Fractional Collection Efficiency of Baghouse Alone and Overall
Baghouse-SDA System vs. Aerodynamic Particle Diameter
17
-------
1 .OOE+2
1.00E+1 -=
o 1 .OOE+0
3 1.00E-1 -=
O
1 -OOE-2
1.00E-3 -=
1 .OOE-4
Scrubber
Unit 8 ESP
i i i i i n n i i i rrrTTi i i i i i i 11
0.10 1.00 10.00 100.00
Aerodynamic Diameter, Microns
Figure 4. Ratio of Outlet to Inlet Differential Mass Concentrations vs
Aerodynamic Particle Diameter
18
-------
100.00
10.00
o
E
3
O
1.00
0.10
Impactor Total Mass
Bailly & Springerville = 18
"
Springerville
i i i i 111
4 5 ซ 7 69
1.00
4 56789
10.00
56789
100.00
Aerodynamic Diameter, Microns
Figure 5. Cumulative Mass Concentrations vs Aerodynamic Particle Diameter
19
-------
99.90
-
_
cT 99.50
c
"o
fc: 99.00
111
v>
^ 98.00
0)
^
eo
3
E 95.00
o I
_
_
90.00
1 1 1 1 1 I I 1 i | | | | i i i 1 I i i I 1 1 l_L
1
_ ,
i
i x
i
ซ i
l / Springerville _, - "
ix x**
~~ ~ ' ป
" l ^
.
1 *
1
l /
f
l /
l 0'
i /
/ Bailly
1 ซ
1 , '
X
, _ 1
- - ' * 1
1
1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
2 34S6789 2 3456789 2 345^789
1.00 10.00 10C
).00
Aerodynamic Diameter, Microns
Figure 6. Apparent Cumulative Mass Efficiency vs Aerodynamic Diameter
20
-------
99.95 -Jj
99.90 -^
-
o
_| 99.50
_o ~
1" 99.00
ซo
^ 98.00
3
"5
| 95.00
O
90.00
on nn
i i i i i i 1 1 i i i i i i i i 1 i i i i i i i i
1
1 <
1
1 ป' *
1 ป '
*~T~ "* COHPAC II
1
1
I
r'***
1 '
/f WESP
1
x
X |
- *
1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
2 3 456789
1.00
2 3456789
10.00
2 3 456789
100.00
Aerodynamic Diameter, Microns
Figure 7. Cumulative Mass Efficiency vs Aerodynamic Particle Diameter
for COHPAC II and WESP Pilot Plants
21
-------
Sorbent
I
I
Flue Gas
ESP
To Stack
PULSE JET
I FABRIC
1 FILTER
a. Pulse Jet Fabric Filter Addition Downstream
Sorbent
I
{
Flue Gas
PULSE JET
ESP I FABRIC '
FILTER |
1 I
,. k.
b. Pulse Jet Fabric Filter
Last ESP Field
Replaces
Figure 8. Sorbent Injection Retrofit Options for Existing ESPs
-------
ro
CO
Sortent
I
Flue Gas
To Stack
a. Sorbent Injection Plus Pulse Jet Fabric Filter
Diffusion
Baffle
To Stack
Sortoent
b. Sorbent Injection Plus Top Load Pulse Jet Fabric Filter
Water +
Sorbent
b. Fluidized Bed Sorbent Injection / Gas Cooling Plus
Top Load Pulse Jet Fabric Filter
Flue Gas
Figure 9, Sorbent Injection Options With Existing Pulse Jet Fabric Filter
-------
0
ro
Flue Gas
ESP
or
Fabric Filter
I PULSE JET "-*
FABRIC I
1 FILTER !
| OR
WET ESP
To Stack
Figure 10. FGD Retrofit With Optional (1, 2, or 3) Secondary Sorbent Injection Points
-------
Impact of Coal Characteristics and Boiler Conditions On ESP Performance
Srivats Srinivasachar
ABB Combustion Engineering, Inc., Windsor, CT, USA
and
Kjell Porle
ABB Flakt Industri AB, Vaxjo, Sweden
Current drivers in electrostatic precipitator (ESP) development include the tightening of
emission standards for particulate and cost reduction. Today's standards for particulate
emission are in the 20-50 mg/Nm3 range, with future standards likely below 10 mg/Nm3.
The current trend towards firing low sulfur coals for SO2 emission reduction poses additional
challenge. Low sulfur coals typically generate high resistivity and difficult-to-capture ash.
This paper describes a detailed study of fly ash properties for different coals, the influence of
combustion conditions, and their impact on ESP collection efficiency. In the laboratory-
scale portion of the study, the generation of fly ash in a full-scale boiler was simulated by
burning pulverized coal in a drop-tube furnace. Extensive analysis of ash particles showed
how a more detailed consideration of coal and mineral properties could explain the
significantly different ash size and composition distribution obtained from similar-looking
fuels.
In the pilot-scale study, tests were performed on a movable ESP integrated with a 1
pulverized coal-fired combustor. Data are presented for a comprehensive test campaign with
a low-sulfur U.S. coal (Powder River Basin sub-bituminous with 0.4% sulfur) generating
moderately high resistivity ash. ESP performance was obtained as a function of boiler
operating conditions including varying coal grind and flame temperature. The effect of flue
gas temperature was also studied in this campaign.
Introduction
The electrostatic precipitator (ESP) is a reliable and efficient particulate control device with
low operating costs and maintenance. It has also been proved to be able to achieve the
required emission levels for most major applications. Over the years, environmental
legislation have demanded increasingly lower dust emissions. This has been achieved by
ESPs using a combination of improvements in the control of the high voltage supply and
increased collecting area. Today's emission standards for coal-fired boilers are in the range
of 20-50 mg/m3 NTP. Future standards are expected to be much stricter, most likely below
10 mg/m3 NTP.1 Additional regulations for fine particulate emissions (PM2 5) are also being
considered.
The need for inherent improvement to ESPs has to be considered keeping in perspective the
current trend towards the use of low sulfur coals. Switching to low sulfur coals is the
dominant approach for SO2 emission reduction in the utility industry.2 Low sulfur coals may
-------
generate high resistivity ash, which causes an undesirable phenomenon called "back corona.
Higher participate emissions occur if there is back corona in the ESP.3
In the Unites States, utilities are switching to low sulfur coals mainly from the sub-
bituminous class mined in the Powder River Basin (PRB). These coals have lower heating
values than the bituminous, higher sulfur coals they are replacing. Switching to PRB fuels
increases the total flue gas flow for a given operating load and lowers the heat extraction in
the boiler. The decreased heat removal increases the temperature of the flue gas entering the
ESP. For most units, which have an ESP after the air heater (cold-side ESPs), the increased
flue gas temperature further exacerbates the problem of high resistivity ash. Particulate
emissions can increase by a factor often when a utility burning a medium- or high-sulfur
coal switches to a low-sulfur coal.
Can the ESP meet these new requirements cost effectively? What factors need to be
considered to extend design rules applicable for current ESPs to those for ultra-low
emissions and for high resistivity ashes? Most likely, one will need to predict more
accurately the detailed characteristics of the ash entering the ESP, and design the ESP with
flexibility to achieve the emission limits (total and fine particulate) for a wide range of
operating conditions.
The overall motivation of this paper is to examine how the characteristics of the ash formed
from combustion are related to the coal, mineral properties and boiler operating conditions.
Furthermore, the paper intends to link some of these ash and boiler characteristics to ESP
performance.
Fuel Selection
Coals spanning a wide range of properties were selected for this study. Bulk fuel analyses
are provided in Table 1. Coals A and B were Polish coals. Coal C and D were Indian coals.
Coal E was from Australia, Coal F from South Africa and Coal G from Colombia. Coal H
was a sub-bituminous coal from the Powder River Basin (PRB) in the United States.
Crushed coal samples were obtained from the field, and these were ground to a fineness
ranging from 65 percent to 90 percent smaller than 74 urn.
Table 1. Bulk Coal and Ash Analysis For Coals In This Study
Parameter
(on dry basis)
Ash (%)
Volatile matter(%)
Total Sulfur (%)
Coal
A
24.0
31
0.9
Coal
B
13.1
33
0.8
Coal
C
40.2
28
0.4
Coal
D
51.6
24
0.5
Coal
E
12.2
31
0.6
Coal
F
17.1
21
0.4
Coal
G
10.2
36
0.8
Coal
H
7.7
43.5
0.4
-------
Component
SiO2
A1203
Fe203
CaO
MgO
Na20
K2O
TiO2
P205
SO3
Coal A
51.1
26.4
6.9
3.2
2.5
0.72
2.8
1.1
0.4
3.8
CoalB
46.8
25.3
9.0
4.7
2.9
1.13
2.3
1.0
0.8
5.2
CoalC
62.6
27.1
5.1
0.8
0.5
0.11
0.9
1.7
0.3
0.4
CoalD
66.0
23.8
5.5
0.7
0.5
0.08
1.0
1.4
0.3
0.3
CoalE
72.6
20.4
4.0
0.3
0.2
0.08
0.6
0.8
0.2
0.3
CoalF
44.9
31.0
3.7
7.6
1.8
0.34
1.09
1.6
3.7
5.5
CoaIG
56.0
19.8
7.8
4.1
2.2
0.6
2.2
0.9
0.2
2.0
CoalH
35.3
19.0
4.8
20.3
3.8
1.4
0.6
1.5
0.9
11.5
Methods
Bench-Scale Facility [Drop-Tube Furnace System]
A drop-tube reactor was used for generating the ash samples. The facility was electrically
heated to achieve temperatures up to 1550ฐC. Coal combustion temperature was modulated
by changing either the gas temperature or the oxygen concentration in the reactant gases. A
pulverized coal feeder provided a uniform feed rate ranging from 0.05 to 2 g/min. Gas flows
ranged from 20 to 30 1/m, which provided a residence time of up to 5 seconds in the reactor.
A detailed description of the drop-tube system has been provided in the literature.6
The drop-tube reactor was mated to a cooling and sampling section. Isokinetic sampling of
the flue gases and the particulate was performed with a collection probe. The collection
probe was connected to several particle characterization systems to measure either the mass
or the number size distributions. These methods are discussed in slightly more detail below.
Pilot-Scale Test Facility
Pilot-scale testing was conducted in a 1 MWft (3.5 MBtu/hr) integrated test facility. This
facility consisted of a combustor, a cooling loop for the furnace gases and a newly
constructed mobile pilot ESP (Figure 1). Pulverized coal was fired up through a single
swirled burner into a refractory-lined furnace. The combustor, which has an extensive
operational history, simulates the time-temperature-oxygen concentration profile of a field
unit. This ensured that the fly ash-vapor phase species partitioning that occurs in the radiant
zone of a field unit would be replicated in the pilot combustor. The ash-laden flue gases
were cooled by a series of water-cooled heat exchangers and water-cooled ducts. The final
temperature control was automatically performed by an air-cooled heat exchanger located
just before an induced draft fan. Stable and accurate control of the flue gas temperature at
the ESP inlet, to within +/- 2ฐC, was achieved during the tests.
The range of coal fineness tested in the pilot unit ranged from a coarse grind of 65% smaller
than 74 u.m to a fine grind of 90% smaller than 74 urn. Combustion or flame temperature
was varied by changing the degree of preheat provided to the secondary combustion air into
-------
the burner. In this manner, flame temperature could be varied without changing the firing
(coal feed) rate.
The pilot ESP was of the wire-plate design with spiral discharge electrodes. Operation of the
pilot ESP at collection efficiencies similar to a field unit is critical for correct scaling of the
performance data. The pilot ESP had three electrical and mechanical fields. At a nominal
gas flow of about 0.6 m3/s , the maximum specific collection area was about 80 m2/(m /s)
[400 ft2/kacfm]. The high SCA allowed operation at very high collection efficiencies even
with the most difficult (high resistivity) ash. The pilot facility was operated continuously for
three weeks. Measurements were conducted after a reasonable steady state was achieved in
the ESP for a given operating condition. Steady state was assumed when flat operating
voltages and emission levels (opacity) were obtained.
Measurement Methods
On-line opacity monitoring was the principal method used for measuring ESP performance
on the pilot facility. The opacity meter (Sick Optik-Electronik, OMD-41, Optical Density
Monitor) was located at the ESP outlet, straddled across a 2 m length of the flue gas duct.
The advanced self-correcting feature, which compensated for any fouling of the optical
elements, allowed accurate long-term opacity measurements. Particulate loading was
gravimetrically measured at the ESP inlet and outlet using EPA Method 5/29.
Mass size distributions of the ash were obtained with a Berner-type low pressure impactor
(BLPI).7'8 BLPI was used both in the pilot and drop-tube furnace tests. BLPI resolves the
ash sample into 11 size classes between 0.01 and 25 urn. A cyclone with a Stokes cut
diameter of approximately 9 |j.m (reference density = 2.45 g/cm3) was used as a pre-cutter to
prevent overloading of the upper BLPI stages at the ESP inlet and in the drop-tube furnace.
In the aforementioned configuration, the BLPI provided size classification between 0.01 urn
to 4 u.m; particles between 4 to 10 um were collected partially in the cyclone with the
remainder in the upper stages of the BLPI. At the outlet of the pilot-scale ESP, the impactor
was used without the cyclone, and the impactor data represented the entire ash sample.
Ash number distribution measurements for the ultra-fine particles were determined with a
Differential Mobility Analyzer (DMA), which separates particles based on their electrical
mobility.9 The DMA is used in conjunction with a condensation nucleus counter (CNC) to
count the classified particles. DMA measurements covered the particle size range from 0.02
to 0.6 |j.m. Flue gas was sampled through a cyclone pre-cutter having a Stokes cut diameter
of about 4 u.m, diluted with clean, dry heated air using an ejector-based dilutor, and
transported via a neutralizer, before admission to the DMA.
A cyclone with a cut size of less than 1 urn was used to collect the ash from the flue gas for
computer controlled scanning electron microscopy (CCSEM) analysis. CCSEM analysis
measures the particles individually, and determines for each particle the composition and the
area. Such data for about 2000 particles were used to statistically calculate volume-based
size distributions.
-------
Results and Discussion
Coal Mineralogy and Ash Characteristics
The mineralogy of the pulverized coal samples was determined with the CCSEM technique
which classifies the minerals as a function of both composition and size. Table 2 provides
the compositional classification for the different coals. The iron content for the all the coals
was relatively low (less than 10 percent); ash compositions from these coals were therefore
dominated by components such as silica, aluminosilicates and calcium and potassium
aluminosilicates.
Quartz (SiO2) content was highest for Coal E. For this coal, the minerals were also very fine,
with 76 percent of the minerals smaller than 4.6 urn. The content of modifier ions found in
minerals such as illite, calcite, pyrite and mixed aluminosilicates, was also the lowest for this
coal. This composition suggests that a high resistivity ash will be formed from this coal. In
contrast, Coal A and Coal B were expected to produce ash of low to moderate resistivity.
Coal H had about 25 percent of the ash present as calcium, organically bound to the coal
matrix. During combustion the calcium coalesced with the discrete minerals such as quartz
and kaolinite to form calcium silicate and calcium aluminosilicate particles.
Ash composition, particularly its distribution with size, is expected to strongly affect the
performance of the precipitator. CCSEM analysis of the total ash samples showed that coal
mineral constituents such as quartz, kaolinite and illite were depleted in the finer ash sizes
during this process, while new compositions such as iron, calcium and other mixed
aluminosilicates were formed (Figure 2). These mixed compositions are formed from
coalescence of the discrete minerals that are included within the coal particles.
For coals containing elements that lower resistivity (for example, Na, K and Fe) the
enrichment of mixed compositions in the fine size range would contribute to lowering
resistivity for the fine ash relative to the bulk. Other parameters such as the structure of the
dust cake also affect resistivity, and the actual resistivity of the dust cake formed with the
fine ash could indeed be higher, as determined experimentally in some cases.
On the other hand, the amount of the low resistivity ash particles formed is very small for
coals such as Coal E, due to the low concentration of elements such as sodium, potassium
and iron. Back corona in the later fields of the precipitator would be aggravated for such
coals because of the predominance of non-conductive "fine" particles formed.
The different ash compositions formed from the combustion of the coal particles have
different propensity for adsorbing gas species such as SO3 and H2O. Adsorption of SO3 and
H2O impacts both the dust cake resistivity, via a surface conduction mechanism, as well as
the cohesivity of the collected ash layer. Knowledge of the ash compositions is the first step.
Further work needs to be conducted to determine how the adsorption characteristics of the
different compositional categories vary.
-------
Table 2. Mineral Analysis (CCSEM) For Coals Used In This Study
(Percent Concentration)
COALS :
MINERAL CATEGORY:
Quartz(Si)
Kaolinite (Si-Al)
Montmorillonite (Si-Al-Ca-Fe)
Illite (Si-AI-K)
Calcite (Ca)+Dolomite (Ca-Mg)
Pyrite (FeS2)
Others (Mixed)
A
14
23
8
28
8
6
13
B
10
28
10
29
6
6
11
C
22
48
9
9
-
4
8
D
26
46
9
7
-
3
9
E
52
29
6
4
-
3
6
F
7
59
6
8
6
2
12
G
31
23
2
18
8
7
11
H
31
20
5
2
35
Correlation Of Ash Characteristics To Field Performance
Current sizing methodologies for the ESP assume some form of size distribution and use the
ash loading as one of the parameters to characterize the ash feed to the ESP. In conjunction
with experience obtained with the performance of operating field units and with other
parameters such as ash resistivity, the average migration velocity is then calculated and used
to determine the collection area for a given collection efficiency. In fact, different ash size
distributions can result from the combustion of coals with very similar bulk ash
composition.6 One of the elements of this study was to examine if there was any correlation
between the ultra-fine (< 0.5 u.m) portion of the ash and precipitator performance.
The ultra-fine ash size distribution (number concentration) for the various coals, as measured
by the DMA, are plotted in Figure 3. All data, except for Coal H, were obtained in the drop-
tube furnace.6 Ash in this size range is formed mainly by vaporization in the boiler and
subsequent condensation in the downstream heat exchangers. The largest size for the
agglomerates (and fraction of total ash) was obtained for Coal E. The high silica content and
its presence as very fine inclusions in the coal was responsible for the high levels of
vaporization. Data for Coal H showed a smaller number concentration compared to the other
coals. This is related to the scale at which the different experiments were conducted. In the
pilot unit with Coal H, a longer residence time and a more gradual cooling profile allows
more agglomeration of the ultra-fine ash resulting in a lower number concentration. The
slightly lower number concentrations in the pilot unit are similar to previous data obtained in
field units.6
Particles in the ultra-fine size range can strongly influence the total surface area of the ash,
hence the condensation characteristics of SO3 and H2O. The surface area distribution as a
function of the ash size, determines the surface density of adsorbing species, such as SO3 and
H2O, on the various particles. For example, if the "ultra-fine" (< 0.5 urn) number
concentration is relatively high, the adsorbing species will be depleted in the "fine" (0.5-4
urn) ash fraction. The "fine" ash fraction dominates the total mass emission from the ESP
(Figure 6, 7 and 9).
-------
ESP performance for the various coals was therefore examined in relation to the surface area
of the ash resulting from these coals. Migration velocities were determined from field data
and normalized to a common ESP operation condition. As shown in Figure 4, there was a
positive correlation between the migration velocity and the ash surface area (corrected for
sulfur content). In particular, Coal E had the lowest migration velocity and highest ash
surface area. It is important to note that for quite similar fuels, based on traditional analysis,
the migration velocity varies over a factor of 5.
Effect Of Combustion Conditions On Ash Characteristics and ESP
Performance
A range of combustion temperatures, coal fineness and ESP inlet temperatures can occur in
the field units boiler design and operation. Typically, the ESP designer is only provided the
bulk coal ash analysis and the ESP inlet boundary conditions. A significant departure from
design performance can therefore be expected in cases, where an atypical boiler design or
operation scheme changes the characteristics of the ash and the flue gas into the ESP.
Effect of Coal Fineness. The effect of coal fineness on the ash characteristics was
examined in both the drop-tube reactor and in the pilot-scale furnace. The ash loading in the
0.1 to 4 micron size range for Coal A, as measured with the Berner impactor, is shown in
Figure 5 for the coal fineness range expected during full-scale operation. The size
distribution data, generated in the drop-tube furnace, showed little difference for the various
coal fineness levels. Coal A, which was a moderately swelling bituminous coal, was
expected to form a cenospheric char upon coal pyrolysis. These chars exhibit a higher
degree of char fragmentation compared to coals that form solid char structures.10 Different
levels of coal grinding was therefore expected to have limited impact on the ash size
distribution for these types of coals.
Slightly different data were obtained in the pilot-scale tests with Coal H. This coal did not
swell upon pyrolysis and formed a solid char. Mineral coalescence was expected to
dominate the characteristics of the ash formed from combustion of this coal. Ash size
distributions measured at the pilot ESP inlet are shown in Figure 6. There was little change
in ash loading for particles smaller than 2 urn, suggesting little impact of coal grinding on the
formation of particles in this size range. The finer coal grind, however, generated a larger
proportion of particles in the 4-20 urn range at the expense of even larger ash particles. For
all tests, a portion of the large ash was collected in the flue gas ducts leading up to the ESP.
The ash loading to the ESP with the fine grind coal was consequently higher than the coarse
grind coal (Figure 6).
Ash loading measured at the pilot ESP outlet are also shown in Figure 6 for the two different
levels of coal fineness. The measured data showed no effect of coal grind on the ESP outlet
emissions. This suggests that, in this case, ash composition was the more important variable
rather than ash fineness, and that grinding the coal finer did not change the individual ash
particle compositions significantly.
Coal fineness could still have an impact on ESP performance in an indirect fashion. A finer
coal feed would ignite earlier and generate higher local combustion temperatures in the
-------
boiler. This can affect the ash formation process, particularly the vaporization of minerals
and the quantity of the ultra-fine ash fraction. ESP performance, as was discussed earlier,
can, in some cases, be quite sensitive to the amount of ultra-fine ash.
Effect of Combustion Temperature. Drop-tube experiments, reported earlier,6
confirmed the strong sensitivity of the fraction of ultra-fine ash generated with respect to
combustion temperature. In these experiments, particle combustion temperature was varied
by changing the oxygen content in the reactant gases. The number size distributions for Coal
E showed that the median size varied from about 40 nm at 5% O2 to 200nm at 20% O2,
indicating a 25-fold increase in surface area. Average coal particle temperatures during
combustion were estimated to be about 1510ฐC at 5% O2, 1590ฐC at 10% O2 and 1790ฐC at
20% 02 for a gas temperature of 1375ฐ C. Increased char particle temperatures lead to
increased ash vaporization. The larger quantity of vaporized ash subsequently condenses
and coagulates to form larger sizes in the ultra-fine range at the higher O2 condition. It is
worth noting that for Coal E, which inherently generates a high "ultra-fine" ash loading
(Figure 3), the ESP performance in various field units was found to be strongly affected by
the boiler operation.
The effect of combustion temperature was studied in the pilot-scale tests by changing the
degree of preheat of the combustion air. Ash size distributions, measured at the ESP inlet at
two combustion temperatures, 1500ฐC and 1625ฐC, are shown in Figure 7. There was about
4 percent O2 in the flue gas for these tests. The pilot-scale data confirmed the bench-scale
data; a higher flame temperature increases the amount of ultra-fine ash (< 0.5 u,m).
For larger ash particles, the higher combustion temperature and the increased burner swirl
(due to higher volumetric flow through the swirler at elevated air preheat temperature)
increased their deposition on the furnace walls. This translated to a lower ESP inlet loading
of ash particles greater than 1 u,m for the high flame temperature case (Figure 7). Also shown
in Figure 7 is the fractional penetration of particles through the ESP for the two cases. ESP
performance was largely unaffected, in spite of the altered loading at the ESP inlet. ESP
operating voltages also did not change to any appreciable extent when flame temperature
was increased.
The minimal impact on the ESP operation can be understood in light of the amount of the
ultra-fine ash formed from the combustion of this coal. The ultra-fine size spectrum for Coal
H is plotted in Figure 3. In comparison to other coals, the mean size and the amount of the
ultra-fine ash agglomerates from Coal H was one of the lowest. It was therefore not
expected to present any competitive surface for SO3 adsorption or deplete the SO3 inventory
available for the larger particles (0.5-4 urn) that constitute the bulk of ESP emissions.
Effect Of ESP Inlet Flue Gas Temperature. Flue gas temperature is a critical
parameter affecting ESP performance. Flue gas temperature changes when the heat
extraction in the boiler is altered. For example, when switching from a bituminous coal to a
sub-bituminous coal, the decreased heat extraction due to the lower heating value of the sub-
bituminous coal can increase the temperature of flue gas entering the ESP by as much as
40ฐC. For the test discussed here with the PRB coal, the initial ESP inlet temperature was
-------
137ฐC. It was increased in one step to 170ฐC (Figure 8). Only a portion of the ESP was
energized during these tests to magnify the differences in outlet emissions.
The extinction plot (Figure 8) shows that there was a significant increase in emissions as a
result of increasing flue gas temperature. Extinction, which is proportional to outlet mass
emissions, increased from 0.07 at 137ฐC to 0.15 at 170ฐC. Clear stack extinction was about
0.03 during these tests. Emissions were 23 mg/Nm3 at 137ฐC and 60 mg/Nm3 at 170ฐC.
The voltage-current (V-I) data from the ESP indicated back corona during operation at the
high temperature. Operating voltage in the rear field (C-field) decreased significantly as the
flue gas temperature was increased from 137ฐC to 170ฐC. The increase in back corona, as
detected by the decrease in the operating voltage, was responsible for the increased
emissions at the higher flue gas temperature.
The size distribution of the ash leaving the ESP at the two flue gas temperatures is shown in
Figure 9. Emission of both ultra-fine (< 0.5 um) and fine ash (0.5-4 um) increased at the
higher flue gas temperature. With potential PM2 5 rules, it is important to measure and
predict not only the total emissions, but also the size distribution of the emitted ash particles.
Summary
This study examined the impact of coal properties and boiler operating conditions on
characteristics of ash formed from combustion and, consequently, on the ESP performance.
Ash from seven coals ,which spanned a range of properties, was generated in an bench-scale,
drop-tube furnace system and characterized for size using various advanced analytical
techniques. Emphasis was placed on two size ranges: less than 0.5 urn and 0.5 to 4 um.
Detailed ESP performance investigations were conducted in an integrated pilot-scale
combustor-ESP facility on an eighth coal.
The experimental data showed that the amount of the ultra-fine ash varied widely for the
different coals. The ultra-fine ash fraction, which is predominantly formed via a
vaporization-condensation process, was exponentially dependent on temperature. Similar
behavior was observed in both the bench- and pilot-scale experiments. For the specific coal
investigated in the pilot facility, ESP performance was not affected by changing the flame
temperature, as the amount of ultra-fine ash was relatively small. However, competition for
SO3 can become an issue when the surface area of the ultra-fine ash becomes comparable to
that of the remainder of the ash. ESP performance can be severely impacted in such cases,
particularly if the ash composition is such that it forms a high resistivity dust cake.
Coal fineness and the temperature of the flue gases leaving the air heater (ESP inlet) were
two other parameters investigated. The ash size distribution was not altered for the case in
which the char from coal devolatilization had a propensity to fragment during combustion.
A non-swelling coal was used in the pilot tests, and changing coal fineness increased the
quantity of ash particles larger than 2 um at the ESP inlet. In spite of this, ESP performance
was unchanged, suggesting the importance of ash composition over ash size in affecting ESP
operation. ESP performance degraded with increasing flue gas temperature, due to increased
back corona. Emissions increased over the entire size spectrum at the elevated temperature.
-------
These results show that consideration of the detailed impact of the fuel and boiler operation
is important for accurately sizing ESPs, and also suggests methods to improve ESP
operation.
Acknowledgments
The authors thank Christer Mauritzson, Mats Thimanson and Inga-Lill Samuelsson of ABB
Flakt Industri, Vaxjo, Sweden, Karl-Heinz Schmidle of ABB Corporate Resarch Centre in
Dattwil, Switzerland and Ben Pease, Gary Tessier, Stan Bohdanowicz, Greg Burns, Rudy
Jackobek, Dana Raymond and other support staff at ABB Combustion Engineering, Inc.,
Windsor, Connecticut, who supported the drop-tube furnace and pilot ESP testing. The
authors also acknowledge Microbeam Technologies Inc. for the CCSEM analysis. The
authors thank Esko I. Kauppinen, Terttaliisa M. Lind and Bertram Schleicher of VTT
Chemical Technology, Espoo, Finland for the Differential Mobility Analyzer data. The
authors acknowledge the support of Thomas D. Brown, project manager from Federal
Energy Technology Center, U.S. Department of Energy. This work was partially funded by
the U.S. Department of Energy under Contract Number DE-AC22-95PC95259.
References
1. Porle, K., Klippel, N., Riccius, O., Kauppinen, E.I. and Lind, T., "Full-Scale ESP
Performance After P.C. Boilers Firing Low Sulfur Coals," presented at EPRI/DOE
International Conference on Managing Hazardous and Particulate Air Pollutants,
Toronto, Canada, August 1995.
2. Platte, J. B. (1991) "Scrub vs. Trade: Enemies or Allies?", The 1991 SO2 Control
Symposium, Washington, DC, Dec 4-8.
3. Soud, H.N., (1995) "Developments in Particulate Control for Coal Combustion", IEA
CR/78 (Apr 1995), IEA Coal Research, London, UK.
4. Guiffre, J. T. (1980) "Precipitator Upgrading and Fuel Control Program for Particulate
Compliance at Pennsylvania Power and Light Company", Second Symposium on the
Transfer and Utilization of Particulate Control Technology, EPA 600/9-80-039a (Sept.
1995).
5. Johnson, S.A., Helble, J.J., Srinivasachar, S., "Evaluating Differences in Slagging
Behavior of Similar Coals," Coal Blending and Switching of Low-Sulfur Western
Coals. Bryers, R.W., and Harding, N.S., Editors, ASME, New York, 1994.
6. Srinivasachar, S., Bradbum, K., Mukherjee, T., Porle, K., Samuelsson, I.L.,
Kauppinen, E.I., Lind, T.M., and Ylatolo, S., " Coal Ash Characterization and
Precipitator Performance," Powergen95, Anaheim, CA December 1995.
7. Kauppinen, E.I. and Pakkanen, T.P., Environ. Sci. Technol, 24, 1811, 1990.
8. Hillamo, R.E., and Kauppinen, E.I. Aerosol Sci. Technol. 14, 33-47, 1992.
9. Joutsensaari, J, Kauppinen, E.I., Jokiniemi, J.K. and Helble, J.J. The Impact of Ash
Deposition on Coal Fired Plants. Proceedings of the Engineering Foundation
Conference, Williamson, J. and Wigley, F. (Eds.) 613-624, 1994.
10. Kang, S.G., Helble, J.J., Sarofim, A.F. and Beer, J.M., Twenty-Second Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh, PA, p.231, 1988.
10
-------
ABB Integrated Pilot ESP Testing Facility
Figure 1
0 0.5 1 1.5 2
Relative Enrichment In Ash Less Than 4.6 microns/Total Ash
Figure 2
Composition of Fine Ash Relative to Bulk Ash for Various Coals
II
-------
1.00E+09
1.00E+05
0.01
0.1
Dp, microns
Figure 3
Ultrafine Ash Loading for Various Coals - Drop Tube Furnace,
15% O2,1450 ฐF - Coal H - Pilot Test 3% O2,1620 ฐC
0.8
ฃ 0.6
0.4
0.2
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Migration Velocity (Relative)
Figure 4
Relationship Between Ash Properties and Migration Velocity
12
-------
100000
10000
1000
100
10
0.01
Size (microns)
10
100
Figure 5
Distribution of Ash Loading in 0.3-4 jim is a Weak Function of Coal Grind:
Coal A - Drop Tube Furnace Data
(Impactor Only; Cyclone Data not Included)
10000
1000
CO
8f
i J
i!
2 Q
,2 "S
S o
a ;o 100
II
II
a
co
-o-Outlet - 65% < 75 microns
-o- Outlet - 90% < 75 microns
-ป- Inlet - 65% < 75 micron
-m- Inlet - 90% < 75 micron
0.010
0.100 1.000
Particle Size, microns
10.000
100.000
Figure 6
Effect of Coal Grind on Size Distribution of Ash Entering and Exiting ESP
(Coal H) - Pilot Testing
13
-------
10
0.1
1
! ^-"
i
/
m^'-L.
[*~~~
:
j
-j
.^x.
Jf -
^Q
x
^
T^^
" -=1^^
.-'--.
!
I
ii-
a-
*
ป
I
X i
a^\ ,.'|
J-\XT~ "
: "^k
1 1
!
-1
L
k
\^?
V
! I
^
i ! I
*
_ Low Temperature - Penetration
_ High Temperature - Penetration
. Low Temperature - Inlet Loading
_ High Temperature - Inlet Loading
I
- 1000.0
100.0
10.0
1.0
0.01
0.1 1
Particle Size (microns)
10
Figure 7
Effect of Combustion (Flame) Temperature on Inlet Loading and Penetration
- Coal H, Pilot-Scale Testing
0.4
0.3
20:00 22:00 0:00 2:00 4:00 6:00 8:00 10:00
Figure 8
Effect of Flue Gas Temperature at ESP Inlet on Extinction and ESP Operating
Voltage - Pilot Testing, Coal H
14
-------
so
70 .
60 -
ซ
1 ,0
O)
o 40 -
"3
0
=5 30
a
20 .
10 -
n
I I
I
I
_o- Rue Gas Temp. = 170 C |
-- Flue Gas Temp. = 137 C
|
0.01
I
!
i
i
I
ii
M
TP=
i
-^
/
/
/
/
T
I
/
~M
/
/ |
/ <
>
/
-~^
/ I ""
' ', '
i ,
;
\
\
I
\
\
\
\
*i
i
I
i
i
i
i
!
\
V
^ K
I
i
i
(=
M
;
0.1 1 10
I
i
1C
Particle Size (microns)
Figure 9
Effect of Flue Gas Temperature at ESP Inlet on Emissions
15
-------
OUTSTANDING IMPROVEMENT OF SOx AND PARTICULATE
REMOVAL EFFICIENCY IN FLUE GAS TREATMENT
Shinichiro KOTAKE, Takeo SfflNODA
ShintaroHONJO, Toru TAKASfflNA
Mitsubishi Heavy Industires Ltd.
2-5-1 Marunouchi, Chiyoda-ku, Tokyo 100, JAPAN
Abstract
In order to respond to electric power demand in Japan, new fossil power plants are being
required, but securing places for these plants is becoming increasingly difficult, while on the
other hand emission regulations corresponding to flue gas treatment facilities of thermal
power stations are also getting to be more and more stringent year by year.
MHO, to respond to such market needs, has developed a high performance flue gas treatment
system which is being promoted for application in actual plants. Two representative
examples of such application are introduced in this paper.
1. Koa Oil Osaka Refinery
The flue gas treatment facility for the refinery's vacuum residue (VR) fired 145 MW output
boiler required the outlet SC>2 cone, to be 4 ppm which translates to a desulfurization
efficiency 99.8%, and outlet particulate cone, below 1.3 mg/m3N for which there are no
prevailing examples so far. However, such extraordinarily difficult requirement have been
successfully coped with a design comprising combination of co-current and counter-current
Double-Contact-Flow Scrubber type absorber (to be called DCFS hereinafter) and Wet
Electrostatic Precipitator (ESP). The basic design of this super high performance FGD unit
is now completed and fabrication/erection of components are currently in progress.
2. Tohoku Electric Power Co. Haramachi Power Station
In Haramachi power station's FGD facility for 1000 MW No. 1 coal-fired unit, a nonleak
GGH heat extractor (raw gas cooler) was located upstream of the dry ESP which is the first
such instance in the world. With this configuration, a highly compact system with high
dedusting performance as compared to that of the conventional system could be realized.
The confirmation tests conducted in June 1997 prior to start of the commercial operation had
proved the supeior particulate removal performance of this new system.
Introduction
Fossil fuels like coal, petroleum etc., are steadily being adopted as energy source by thermal
power stations, petrochemical complexes and so forth for operation of their very large-size
new facilities which are under construction at various locations in recognition of their
handling ease, stability of supply etc. However, from the view point of environmental
protection, as the regulation concerning emission of particulates, sulfur dioxide, nitrogen
-------
dioxide etc., are becoming increasingly stringent, provision of environmental protection
facilities exhibiting higher performance has become indispensable inconjunction with the
construction of above mentioned new power stations, petrochemical complexes etc.
Particularly in case of Japan, for construction of new petrochemical complexes which are
located in industrial zones near to the human habitations, such emissions close to zero are
being demanded. On the other hand, it is a fact that simultaneous with demands to scale up
the facility size with higher performance, demands are being made to reduce both the
equipment investment cost and operation cost. Mitsubishi Heavy Industries Ltd., (MHI) in
an effort to respond to such mutually contradictory needs, started developing an absorber
called the Double-Contact-How Scrubber (DCFS) from the year 1987 with superior SOx
absorption and dedusting performance than the conventional ones. Together with the
adoption of DCFS for improved desulfurization performance, MHI succeeded in developing a
system where paniculate concentration at stack outlet could be achieved below 5 mg/m N by
locating the heat extractor of nonleak GGH upstream of the dry ESP (low temperature ESP).
The system is further featured for its low equipment investment cost and operation cost as
compared to that of the conventional systems.
This paper, while introducing the flue gas desulfurization facility adopted with the aforesaid
DCFS type super high performance absorber which could achieve a desulfurization
percentage of 99.8% in Koa Oil Refinery's FGD unit, also reports on the outline of design and
performance test results of Tohoku Electric's Haramachi FGD Unit 1 adopted with a nonleak
GGH heat extractor and a dry ESP.
Table 1. Flue Gas Treatment System - Design Values
Power station
Output
Fuel
Treated gas volume
Inlet SO2 Cone.
Outlet SO2 Cone.
Desulfurization
percentage
Inlet particulate cone.
Outlet particulate cone.
Outlet gas temp.
Koa Oil Osaka
149 MW
VR, CO
674,000 mJN/h* .
2,170 ppm"
5.4 ppm
99.8%
1 .3 mg/mJN
90 ฐC
Tohoku Electric
Unit 1
1000MW
Coal
3,61 0,000 mJN/h
690 ppm
69 ppm
90%
22 g/m3N
25 mg/mJN
103 ฐC
* Simultaneous treatment of VR combusted gas and CO combusted gas
** VR combustion only
FGD facility for Koa Oil Osaka Refinery
Koa Oil Osaka Refinery is located at Japan's Hanshin industrial region together with
numerous other chemical plants where emission of environmental pollutants from each plant
is strictly regulated. To comply with such regulations, Koa Oil has been mixing the vacuum
residue of oil refining with light oil and shipping it out as heavy fuel oil.
However, with the deregulation of electric power market resulting from revision of Japanese
electric utility law in 1995, as the sales of electric power from other industries to power
-------
utilities became a possibility and the utilities started to invite tenders from the independent
power producers (IPP) in every region of the country, Koa decided to participate in this
business and build a new power generating unit utilizing the vacuum residue at their Osaka
refinery for efficient energy use.
For construction of this power generating unit however, Koa Oil had to comply not only with
the above mentioned stringent emission regulations of the region but also with even stricter
public demand concerning the emissions. As a result, contract requirement for the FGD
facility called for compliance with 99.8% desulfurization efficiency and 1.3 mg/m3N
particulate concentration at stack inlet. To cope with such hard demand MHI proposed their
developed combination design of DCFS and wet ESP which Koa accepted to adopt. The
FGD plant is now under construction and scheduled to be started up in July 1998.
Development of Super High Performance Double-Contact-Flow Scrubber (DCFS)
The DCFS shown in Fig. 1 is suitable for application where high particulate removal needs to
be considered in particular. In DCFS, with liquid absorbent uniformly spouted up through
nozzles located at absorber lower part, and further with mutual collision between the liquid
drops, the gas/liquid contact is made effective to achieve high desulfurization and dedusting
efficiency under low pressure loss condition.
Gas Outlet
Absorber Misi Eliminator
Absorber Recirculabon Pump
'ARS
Oxidation Air Blower
Fig. 1 Sectional Diagram of DCFS
After completion of its fundamental tests with 15,000 m3N/h pilot facility at MHI Hiroshima
R&D Center, the DCFS was subjected to joint Verification tests with Chubu Electric Power
Co. and Chugoku Electric Power Co. using actual coal flue gas in 15,000 m3N/h and 300.000
m3N/h facilities respectively. Thereafter, counter-current DCFS was adopted for the 136
MW heavy oil-fired No. 2 unit of Kashima-minami Joint power station in 1993 followed by
co-current DCFS adopted for coal-fired 175 MW No. 1 unit of Chugoku Electric's
Shimonoseki power station in 1994. Further, in 1995 the counter-current DCFS was also
adopted for treating 10 X 104 m3N/h flue gas of Weifang Chemical factory in China, in 1997
for 250 MW Mikuni thermal power station of Fukui Joint Thermal Power, and 149 MW
generating unit of Sumitomo Osaka Cement factory. In DCFS, for ensuring long-term
-------
stable operation, the nozzle for spouting up the recirculated absorbent liquid is made from
abrasion resistive ceramic (Fig. 2) or urethane rubber with the header pipe made of 6% Mo
steel or 18-8 stainless steel or FRP depending on the properties of the recirculated liquid.
Since commencement of the commercial operation, the FGD facilities are maintaining their
design desulfurization and dedusting performances without any problems like abrasion,
plugging and so forth under the design operation conditions.
Fig.2 Liquid Column Nozzle
Design of Super High Performance Absorber
Desulfunzation performance is affected by various FGD design factors like absorbent liquid
recirculation flow rate. FGD system inlet SOi concentration, absorber dimension etc. MHI
on the basis of above mentioned tests and actual plant operation experience clarified the
effect of these conditions on the desulfurization performance, and successfully developed a
simulated performance evaluation system. Compiled data for the co- and counter-current
DCFS are shown respectively in Table 2 and 3.
Table 2 Co-current DCFS
Plant
Inlet SO2 (ppm)
Section Area (m2)
Ug (m/s)
L1 (m-Vm^r.)
HL (m)
Cold model
2005000
1
2.5-10
112440
2.0-13
A plant
140-580
21
4.3-5.3
172-283
4.412
B plant
450-795
44
4.14.9
231-271
8-11
KOA
2142
54
4.2
238
8.13
Table 3 Counter-current DCFS
Plant
Inlet SO2 (ppm)
Section Area (m2)
Ug (m/s)
L1 (m3/m^n)
HL (m)
Cold model
200-5000
1
2.510
112.2440
2.0-13
C plant
1510-1650
42
2.1-3.9
172-197
4.6-6.1
KOA
182
74.52
3.0
169
4.16
Remarks)
A : Demonstration plant
B, C : Actual plant
Ug : Gas velocity
L : Recirculation flow rate
HL : Height of spouted up liquid
-------
Example of desulfurization performance curve corresponding to liquid recirculation flow rate
in co- and counter-current DCFS are shown respectively in Fig. 3 and 4. As can be seen
from these figures, there is good agreement between the estimated performance curve and the
actual performance curve for each case to allow sufficient evaluation of both the co- and
counter-current DCFS. However, to achieve above 99% desulfurization efficiency with only
the co-current or the counter-current DCFS, the recirculation flow rate becomes so large as to
make the design impractical due to big increase in equipment and operation costs.
Desulfurization
Efficiency
Desulfurization
Efficiency
(%) 70
160 180 200 220 240 260 280 300 320
Recirculation Row Volume L' (m3/m2)
Fig.3 Co-Current Absorber Estimated
Desulfurization Performance
SO 100 150 200
Recirculation Flow Volume L' (m3/m2)
Hg.4 Counter-Current Absorber Estimated
Desulfurization Performance
250
Thereupon, system design optimization with the combination of the co- and counter-current
DCFS was made. The result is shown in Table 4. Based on this result, the twin-tower co-
and counter-current DCFS was adopted -because of its advantage in respect of equipment cost
etc.
Table 4 Result of Study made with Absorber Combination
Case No.
Item -
Design condition
Calculation Result
Absorber spec.
Flue gas vol. (mJN/h)
Inlet S02 (ppm(d))
Absorber type
Design desulfurization effect. (%)
Outlet SO2 (ppm(d))
Theoretical liquid column height (m)
Power consumption
Gas side pressure loss (mmAq)
Absorber equip, investment cost
Absorber MTO (Material Take Off)
1
445,620
2,790
3 towers
Counter- + Co- + Counter Current
99.87
569/137/3.7
3.7/4.0/3.7
100
180
100
100
2
445,620
2,790
2 towers
Co + Counter Current
99.87
372/3.7
8.8/8.4
156
137
68
60
Power consumption, equipment cost, and MTO (Material Take Off) are taken as 100 for Case 1
-------
On the basis of above results, desulfurization performance evaluation test was conducted with
MHl's co- and counter-current DCFS test facility (Fig. 5). In Table 5, design conditions as
well as hot model test conditions for Koa Oil's FGD unit are shown.
Fig. 5 Co-Counter Current DCFS Test Facility
Table 5 Design Conditions of Actual Plant and Hot Model Test Conditions
Item
1
2.
3.
Common Conditions
Gas flow rate
Inlet SO2
Outlet SO2
Overall desulfurization effect
Co-current section condition
Gas flow rate
Liquid column height
Co-current section desutfurization effect
Counter-current section conditions
Gas flow rate
Liquid column height
(m3N(w)/h)
(ppm(d))
(ppm(d))
(%)
(m/s)
(m)
(%)
(m/s)
(m)
Counter-current section desulfurization effect (%)
Actual Plant
Design
Condition
674.000
2,170
5.4
99.8
4.1
8.8
89.3
3.0
8.2
99.0
Hot Model
Condition
12,100
2,200
4.0
9.0
4.0
4~7
-
Result of test made with liquid column height fixed for co-current DCFS and varied for
counter-current DCFS is shown in Fig.6.
-------
10
2
CD
o
(3.7ppm)
Design figure
3456789
Counter-current liquid column height [m]
Hg 6 Co/Counter Current DCFS Test Result
In the Hot Model test, an extremely high desulfurization efficiency was obtained which
sufficiently met the design condition of 3.7 ppm. The absorber specification for Koa Oil
Osaka Refinery is shown in Table 6.
Table 6
No. of Unit
Dims. Width (co- / counter-current)
Depth
Height (co- / counter-current)
Nozzle Material
Header Material
3.9m
13
18.9m
1
5.4m
.9m
20.9m
Urethane Rubber
Type 304L
FGD Facility for Tohoku Electric Power Co. Haramachi Power Station
In Japan, since the start of coal utilization in utility thermal power stations in 1975, the FGD
system process flow shown in Fig. 1A for maintaining particulate and SOx emissions from
power stations at 100 mg/m3N and below 100 ppm respectively has been in general use.
Since then there is a trend to impose increasingly stricter particulate emission value for each
plant site, and in keeping with this trend, the flue gas treatment system got changed from the
conventional system shown in Fig. 7A to the one shown in Fig. 7B for high particulate
removal. However, as the overall cost of this system got largely inflated starting with that
for the wet ESP while requiring large installation space at the same time, the need for a
compact system with higher performance but of low cost rose. To respond to this need, a
system was jointly developed with Chubu Electric Power Co. in which by installing the
nonleak GGH upstream of the dry ESP, performance equivalent to that of the one of Fig. 7B
could be achieved without the provision of wet ESP. This system shown in Fig. 7D has
been adopted for FGD facility of Tohoku Electric Power Co.'s 1000 MW No. 1 unit.
-------
A
B
C
s
D
Conventional system
Dry ESP _
LjuDgstrom
type GGH
-
Grid-bed
Absorber
Ljungstrom
- typeGGH
STACK
Conventional System for nigh dedusting performance
Dry ESP -
NonJeak
GGH
-
Grid-bed
Absorber
We! ESP -
Nonleak
GGH
- STACK
-------
Dispersion
Quantity
Swing
Dispersion
Continuous
Dispersion
Time
Fig. 10 Steel Shot Dispersion Cycle
Heat transfer tubes are of fin type carbon steel tube. For prevention of tube abrasion by the
dispersion of steel shots, fin protector is provided at the upper part of the tube bundle which
while absorbing energy of the dispersed steel shots evenly drops them on the tube bundle.
For prevention of tube erosion further, suppression of gas pressure-drop rise by suppressing
gas velocity within the bundle was considered. For separation of ash and steel shots, rotary
separator is provided which enables almost complete prevention of ash adhesion on the steel
shots at separator outlet.
Dry ESP
In the gas temperature region of 90ฐ to 100ฐC, the electrical resistance of particulates is lower
than that of the 130ฐ to 140 ฐC region to allow prevention of back corona discharge and
achieve high particulate collection efficiency under stable ESP charged condition. However,
with drop in electrical resistivity of ash, the electrical attraction force towards the electrodes
drops, in addition, with fouling of electrodes, particularly if there is ash corpulence in the
discharge wire, the corona current distribution becomes disproportioned to cause weak
adhesion of ash at spots having low current flow and thus to ash reentrainment in gas. To
prevent such reentrainment, off-flow damper rapping system, and modified electrodes for
uniform current distribution have been adopted in this dry ESP. Even though the collection
electrode and the discharge electrode are made of such general use material like SPH and
SPC, but they are found to be corrosion-free without any dust adhesion or solidification on
them. In order to avoid hopper plugging by ash which was expected because of reduced
fluidity of ash caused by temperature lowering of ash for ash separation, such means as
enhanced steam heating, infrared painting of hopper inside/outside, enhanced aeration as well
as ash handling system linked with ESP damper opening/closing and rapping were adopted.
Absorber
Co-current grid-packed absorber has been adopted where the grids are packed in lattice
formation. With this arrangement, by making the absorbent liquid drip from the top part of
these packings, the gas-liquid contact surface was enlarged to effectively absorb SO2.
Moreover, due to the co-current flow of gas and liquid the pressure loss was comparatively
low. In addition, lower absorber pump power consumption than that of the absorber with
pressurized spray nozzles was achieved because of the adoption of low backpressure spray
nozzles.
-------
Absorber Ancillaries
As oxidation equipment, Air Rotary Sparger (ARS) has been adopted. With ARS
improved air-liquid contact efficiency is achieved with fine air bubbles created in the water
current by discharging air through its agitating arms. This resulted in highly effective
oxidation with reduced amount of air. Further, due to mixed current of air bubbles and
water, adequate mixing of solids and liquid takes place to prevent settling of solids in the
tank. This enables extraction of high grade gypsum. In consideration of the corrosive and
highly abrasive nature of the absorbent liquid, 254SMO has been adopted for absorber
sections in contact with the liquid.
Six horizontal vane-pitch controlled absorber pumps have been adopted. In this type of
pump, impeller vane pitch is continuously changed in accordance with boiler load and FGD
load signals to regulate the pump flow rate and thus control the pump power consumption to
its minimum limit. For prolonged operation, the pump impeller is made of 2 phase stainless
steel with tungsten carbide surface coating.
Ca utilization ratio is maintained at a fixed value while keeping the absorber outlet SO2
concentration below the set value to control desulfurization by detecting unreacted CaCO3
percentage in absorber recirculation liquid with a CaCO3 analyzer. This enables the
recirculation flow rate to be kept at minimum and thus contribute towards lowering of the
operation cost.
Performance Test Results
Trial operation of Tohoku Electee's Haramachi FGD Unit 1 which started in November 1996
was successfully carried out while confirming various performances without any trouble.
The operation of GGH which was placed for the first time under high dust loading was good
together with that of the operation of SSCS. Although about six months have passed since
the start of the trial operation, but no trouble has been experienced with dust adhesion or
SSCS malfunction.
Results of performance test made in May 1997 are shown in Table 7. As regards the test
conditions, except the inlet SO2 concentration which was slightly lower than the design value,
the remaining were almost same as that of the design ones. Under these conditions, it was
confirmed that operation is possible with higher desulfurization and dedusting as well as
lower power consumption performances than the estimated ones.
Particularly, with the adoption of very low temperature ESP, together with the achievement of
high ESP dedusting performance, a very high dedusting performance was achieved in the
absorber as well. This is because of the fact that particulate size at ESP outlet is larger than
that in the conventional system.
-------
Table 7 Performance Test Result at 1OOOMW Output (1997/5/22)
Unit
Actually measured
value
FGD Plant Inlet
Gas Flow Rate
Gas Flow Rate
Flue Gas Temp.
SOx
SOx
O2
H20
m3N/h(w)
m3N/h(d)
"C
ppm(d)
m3N/h
vol%(w)
vol%
3071000
2782300
100
621
1728
4.2
9.4
FGD Plant Outlet
Gas flow rate
Gas flow rate
Flue gas temp.
SOx
SOx
O2
H2O
Desulfurization percentage
m3N/h(w)
m3N/h(d)
'C
ppm(d)
m3N/h
vol%(w)
vol%
%
3072000
2749400
104
42
115
4.6
10.5
93.2
Dedusting Performance
FGD plant inlet
Stack inlet
mg/m3N(d)
mg/m3N(d)
22
0.6
Utility Consumption
Power consumption
FGD plant make-up water
Limestone consumption
1 4K steam flow rate
MW
t/h
t/h
t/h
8.7
107.4
7.0
35.6
Conclusion
1) With the combination of a co-/counter-current DCFS and a wet ESP it became possible
to offer a high efficiency flue gas desulfurization system which can achieve absorber outlet
SO2 concentration below 4ppm (desulfurization efficiency 99.8%) and paniculate
concentration below 1.3 mg/m3N.
2) From the trial operation of Tohoku Electric's coal-fired Haramachi 1000 MW FGD Unit
1, the performance of the compact system having superb dedusting capability adopted with
nonleak GGH heat extractor placed under high dust loading and a dry ESP operating under a
temperature between 90ฐ ~ 100ฐC was proved to be stable and higher than the estimated one.
3) The Mitsubishi High Efficiency Flue Gas Desulfurization System which comprises the
system of item 2 above added with a DCFS has exceedingly high desulfurization and
dedusting capacity has been adopted for several FGD units which are presently under
construction or design. These are, Chugoku Electric's 1000MW Misumi power station,
Electric Power Development Company's 1050 MW Tachibanawan thermal power station,
and Shikoku Electric's 700 MW Tachibanawan power station.
We are confident, this Mitsubishi High Efficiency Rue Gas Desulfurization system would
become the model system for coping with rigorous conditions which are imposed on large
size coal-fired power plants for environmental protection.
-------
Bibliography
1. K. Iwashita, T. Takashina, S. Okino, Y. Endo
"Commercial Application of New Type Scrubber"
1995 SO2 Control Symposium
2. S. Kotake, K. Iwashita, Y. Tsuchiya, T. Higashi
Development of High Efficiency Wet Limestone Gypsum Flue Gas Desulfurization
System"
1993 SO2 Control Symposium
3. H. Owaki
"Development of High Efficiency Dust Collecting System in High efficiency Flue
Gas Cleaning System"
Special document for EERE Member R-9112, 1992. 1
4. S. Kotake, K. Iwashita, Y. Tsuchiya, H. Kataoka
"Development of High Efficiency Flue Gas Desulfurization System"
ICOPE-97
-------
RECENT COMBUSTROLฎ FACT
FLUE GAS CONDITIONING EXPERIENCE
Robert R. Crynack, Ph.D.
Wheelabrator Air Pollution Control
441 Smithfield Street
Pittsburgh, PA 15222-2292
Abstract
The Combustrol* FACT (Fly Ash Conditioning and Treatment) technology is a simple, cost
effective system to improve marginally performing electrostatic precipitators when collecting high
resistivity fly ash. This technology uses non-hazardous, water based conditioners to reduce fly
ash resistivity and improve precipitator efficiency. This paper describes the technology and
presents three case histories of recent demonstration installations. All three demonstrations were
on units that were load limited due to high paniculate emissions and opacity when burning lower
sulfur coals than originally designed. All three demonstrations showed that the FACT technology
provides an economically attractive alternative to purchasing replacement power.
Introduction and Technology Development
The FACT technology has been permanently installed on seventeen boiler units. Numerous test
installations have also proven its capability. The FACT technology has been successfully
demonstrated on units as small as 15 megawatts (mw) and as large as 720 mw. The largest
permanent installation currently using the FACT technology is 350 mw. The technology is not
new, but has been slow to be accepted because of the poor reputation of past suppliers of
proprietary chemical conditioners.
The Calgon Corporation has been a leader in the water treatment chemical business for many
years. In the early 1980s they were approached by their utility customers to provide flue gas
conditioning chemicals. Calgon put their R&D, chemical, and process expertise to the task. In
1983 they demonstrated their first system and in 1984 they patented the process.
In August 1994 Wheelabrator Air Pollution Control became the exclusive world wide licensee for
Calgon's Combustrol* FACT technology. This license agreement brought together the chemical
expertise of Calgon and the precipitator experience of Wheelabrator. This technology has been
developed over a fifteen year period and is a proven technology to improve the performance of
marginally performing precipitators.
-------
FACT Conditioners
The proprietary conditioners are not "off the shelf chemicals or simple combinations of chemicals
that are blanketed in secrecy. The FACT conditioners are a carefully conceived and accurately
formulated combination of chemicals to reduce fly ash resistivity and enhance particulate cohesion
to improve precipitator performance. Unlike many past and present suppliers of proprietary
conditioners, Wheelabrator and Calgon are very open about the chemistry of the conditioners.
There are several conditioner formulations, but the one most effective on low sulfur, western
coals is FACT-5000. FACT-5000 is a water based conditioner composed of 50% proprietary
polymer, 25% ammonium nitrate, and 25% sodium nitrate. Ammonia and sodium have
traditionally been used as conditioning chemicals, but the polymer is unique.
The polymer in FACT-5000 is comprised of three monomers that are polymerized to form the
proprietary patented conditioner. These three monomers are: (1) acrylamide, (2) 2-acrylamido 2-
methlypropane sulfonic acid, and (3) acrylic acid. The resulting polymer has three functional
groups that carry negative charges through the deposited ash layer. These charge carriers
enhance the ability of the ash to conduct the particulate charge to the grounded plate. The
polymer combined with the ammonium and sodium salts produces an effective fly ash
conditioner.
As with all chemicals, concern arises about the toxicity and hazards of this conditioner. The
FACT-5000 conditioner has no flammability or reactivity hazards. There is no inhalation health
hazard that requires respiratory protection during use or handling. Its low health hazard
classification identifies that prolonged skin contact may cause minor skin irritation and eye contact
may cause temporary eye irritation. These effects can be eliminated by use of common personal
protection equipment. There are no transportation restrictions or special handling requirements.
FACT Feed Equipment
The FACT feed equipment is simple and reliable, resulting in minimal capital, operating, and
maintenance costs. The feed system consists of three assemblies (1) the Feed Skid, (2) the
Air/Liquid Panel, and (3) the Injection Lances. See Figure 1 for a schematic diagram of the
system. See Figures 2, 3, and 4 for pictures of the Feed Skid, Air/Liquid Panel, and Injection
Lance, respectively.
The chemical feed pump on the Feed Skid provides a precise amount of the concentrated
conditioner at a controlled rate. The water booster pump provides the dilution water to dilute the
concentrated conditioner to the required concentration. The diluted conditioner is then fed to the
Air/Liquid Panel where the flow to each Injection Lance is monitored. The compressed air flow
to each Injection Lance is also controlled at the Air/Liquid Panel. The prescribed air and liquid
flows are then fed to the Injection Lances, where two fluid (air and liquid) nozzles atomize the
diluted conditioner.
-------
Demonstration #1
Early in 1996, a demonstration unit was installed and tested at a Michigan power plant on a 29
mw boiler unit. The coal being burned was a Pennsylvania bituminous with 1.2% sulfur, 7.5%
moisture, and 8% ash. The precipitator is a 1973 Research-Cottrell weighted wire unit. The two
field unit provides an SCA of 167 for treatment of the 110,000 acfin of gas at a temperature of
300 ฐF. The boiler was load limited because of opacity concerns. The plant is near a summer
resort area and has to be very careful not to exceed the six minute average 20% opacity limit.
Table #1 is a brief summary of the results of the demonstration project.
Within 2 hours of the start of conditioning of the dirty precipitator, power levels started increasing
and opacity started decreasing. Within 24 hours the boiler was able to reach its full rating of 29
mw with only 16% opacity. Within 48 hours the opacity was further reduced to 10% at full load.
After one month of conditioning, operation stabilized and full load was achievable with an average
opacity of 10%. The corona current in the first field stabilized at 62 ma, a 4.4 times increase over
pre-conditioning levels. Similarly, the second field corona current increase by a factor of 6. The
best operation in the three month demonstration project achieved 8% opacity at full load with
nearly a six (6)fold increase in inlet field corona current and a twelve (12) fold increase in outlet
field current.
TABLE #1
Demonstration #1
Summary of Results
Operation
No conditioning
After 2 hours
After 24 hours
After 48 hours
After 1 month
Best of 3 months
MW
24
24
29
29
29
29
Opacity
18%
15%
16%
10%
10%
8%
KV1
35
37
37
37
37
38
MAI
14
22
30
52
62
82
KV2
34
35
35
35
36
37
MA2
25
54
60
150
150
300
This was a very successful demonstration. Full boiler load could be achieved at all times with
opacity averaging 10% compared to the derated level of 24 mw at close to 20% opacity.
Conditioner usage varied due to fuel variations, but the cost of the conditioner was approximately
9 to 16 cents per megawatt-hour of generation.
-------
Although this was an economically attractive solution to achieving full load capability at
significantly reduced opacity, the plant decided to pursue other options. The plant tested the
burning of a mixture of petroleum coke and Powder River Basin (PRB) coal. The cost of the
PRB coal and petroleum coke was significantly less than the eastern coal being burned. The
precipitator performance was expected to be worse and the FACT system was temporarily kept
on site. With burning about 10-20% petroleum coke, the performance of the precipitator was
very good, achieving at least as good of results as with the conditioning. It is hypothesized that
the high vanadium content of the petroleum coke produced significant SO3 to provide natural
conditioning of the fly ash. The plant continues to operate with these blended fuels.
Demonstration #2
This demonstration project was conducted in the Fall of 1996 at a mid-western municipal utility.
The boiler is a 1968 vintage, tangentially fired unit, producing 360,000 pounds per hour of steam
flow. The two field precipitator was reworked in 1990 to make three fields by converting the 9
foot inlet field into two 4.5 foot fields. A third power supply was added. Barbed wires replaced
the round weighted wires in the first two fields. The collecting area of 21,840 square feet and gas
volume of 167,000 acfm gives an SCA of 131.
The coal burned at this plant is a Wyoming Powder River Basin (PRB) coal from the
Wyodak/Anderson Seam of the Rochelle Mine. This coal has a heating value of 8800 BTU/lb,
moisture of 27%, ash of 4-5%, and sulfur of 0.2-0.3%. Based on the mineral analysis of a typical
sample, the resistivity at an operating temperature of 390 ฐF is 1 x 10 n ohm-cm.
This is a peaking unit rated at 34 mw and is used to meet peak summer air conditioning demand,
which is about 2000 hours/year. The opacity limitation is 40%, based on six minute averages, but
the plant likes to maintain less than 30% average opacity. This unit's output was typically derated
by 20% to maintain acceptable opacity.
The results of this demonstration project are summarized in Table #2. With just two hours of
conditioning, the dirty precipitator showed improvements in operation. The precipitator voltage
increased to near the transformer/rectifier rating of 45 KV in all three fields. This also
significantly reduced the spark rate. The corona voltage and current increases resulted in a 60%
increase in precipitator power (KW). After 24 hours of conditioning, the full load rating of 34
mw was achievable at an opacity of 25-30%, compared to the same opacity at only 28 mw
without conditioning. The precipitator power increased by 2.5 times the pre-conditioning level,
even though boiler load increased 21%.
The amount of conditioner feed varies widely with this summer peaking unit. When in operation
and requiring the full 34 mw production, the cost of conditioner is approximately $4.75 per hour.
This is equivalent to $0.14 per mw-hour or .014 cents per kw-hr of generation capacity. Lesser
or no conditioning is required at lower loads.
-------
TABLE #2
Demonstration #2
Summary of Results
OPERATION
No conditioning
After 2 hours
After 24 hours
MW
28
28
34
AVGKV
38
44
45
AVGMA
30
50
77
AVG SPK
45
2
2
TOTKW
5.6
9.1
13.5
This system is a cost effective way of getting full load production. The municipal utility
purchased a permanent system that started up in June 1997. Results similar to those described
above have been realized with the permanent system.
Demonstration #3
The third demonstration project was conducted in Spring 1997 at a large mid-western utility.
This 340 mw unit has a tangentially fired dry bottom boiler. The four field, sixteen bus section
electrostatic precipitator provides an SCA of 116 for the 1,230,000 acfin gas volume. The plant
has no coal pile and is referred to as a "just in time" facility. The 6 minute average stack opacity
limit is 30%.
The plant burns three coals - (1) Decker, (2) Rochelle, and (3) Jacob's Ranch. Decker is a low
sulfur, western coal from Montana, and is well known for its high sodium content that makes its
ash "very collectable" The Decker coal burned at this plant has a sodium content of 2-2.5% in
the ash. The Rochelle coal is similar to that described above for Demonstration #2. The Jacob's
Ranch coal is similar to the Rochelle coal, but the sodium content of the ash is only 1.5% and the
ash content is 6%, making it the most difficult of the three coal ashes to collect.
The test program was done entirely by the utility, with only check out and start up support from
WAPC. The demonstration unit was initially installed on only one-half of the precipitator. Power
levels and opacity improved adequately to justify completing the installation on the entire unit.
Tests were conducted for all three coals. Table #3 is a summary of the results. All of the data
provided in Table #3 is from a draft report from the utility.
-------
TABLE #3
Demonstration #3
Summary of Results
Decker
Max. Load (mw) w/o Conditioning 330
Max. Load (mw) with Conditioning N/A
Opacity (%) 18
EP Power (KW) with Conditioning 3 8 1
Conditioner Feed (%) 0
E^belle
270
325
25
385
70
Jacob's Ranch
240
290
27
312
100
With the Decker coal only 330 mw was achievable due to boiler limitations, even though the unit
is rated at 340 mw. No conditioning was needed to achieve 18% opacity. When burning the
Rochelle coal, the precipitator power levels with conditioning were similar to those for the Decker
coal. A load of 325 mw was achievable for a 25% opacity with a conditioner feed rate of 70%,
compared to only 270 mw without conditioning. Higher conditioner feed rates were not tried to
obtain higher load or lower opacity. Although full load was not achieved, an additional 55 mw of
load, or 20% increase, was realized.
When burning the Jacob's Ranch coal, a load of 290 mw was reached at an opacity of 27% with
conditioning, compared to only 240 mw without conditioning. Power levels were about 80% of
those with the other coals even with conditioning at the maximum rate of this demonstration
system. This was an increase of 50 mw, or 20%, over unconditioned maximum achievable loads.
For the Rochelle coal, the cost of the conditioner to gain the additional 55 mw of load would be
approximately $1.37 per mw-hour of additional load or $0.23 per mw-hour of achievable load.
For the Jacob's Ranch coal, the conditioning cost to gain the additional 50 mw load would be
$2.10 per mw-hour of additional load or $0.36 per mw-hour of achievable load. This compares
favorably to the utility's estimated cost of $8-10 per mw-hour to purchase the additional power.
At the writing of this paper, the demonstration unit is still installed on this unit. With the high
demand for the summer cooling load, the demonstration unit provides the utility with an
economical alternative to buying that incremental power that the FACT conditioning system
permits. However, the performance of the demonstration did not meet all of the utility's
expectations. Their goal was to achieve full load with both PRB coals and with a wider opacity
safety margin. Conditioner usage was higher than expected. The high 7.5 feet per second face
velocity through the precipitators with the resulting low treatment time of only 3.2 seconds,
makes this a very difficult application to achieve higher precipitator performance than obtained
here when burning PRB coals .
-------
Summary
The three recent demonstration projects presented in this paper show that the Combustrol* FACT
flue gas conditioning technology can improve the performance of marginally performing
precipitators. In two cases, a 20% increase in load was achieved, allowing the boiler to reach full
load. In the third case 50-55 mw of additional load was achieved with conditioning and was a
cost effective option to purchasing replacement power.
The simplicity and low capital cost of the FACT technology make this a viable and cost effective
way for utilities to reduce emissions and opacity. Production can be increased to regain that lost
by the derating needed to maintain compliance. A low cost demonstration such as those reported
above, is an effective way for any utility to evaluate the merits of the system in their own plants
Compressed
Air
or Steam
COMBUSTROL
FACT
Conditioner
Lances
Diluted
Water
Compressed
Air
or Steam
Lances
Figure 1
FACT Feed System
-------
Figure 2
FACT Feed Skid
-------
Figure 3
FACT Air/Liquid Panel
-------
Figure 4
FACT Injection Lance
-------
COST-EFFECTIVE RETROFIT TECHNOLOGY FOR
ENHANCED ESP PERFORMANCE ON LOW-SULFUR COAL
Michael D. Durham, Ph.D.
C. Jean Bustard,
Kenneth E. Baldrey
Cameron E. Martin
ADA Environmental Solutions, LLC
7931 S. Broadway, #349
Littleton, CO 80122
(303)792-5615
Abstract
Electrostatic precipitators (ESPs) serve as the primary particle control devices for a majority of
coal-fired power generating units in the United States. ESPs are used to collect particulate matter
that range in size from less than one micrometer in diameter to several hundred micrometers.
Utilities are considering a variety of options to respond to Title IV of the 1990 Clean Air Act
Amendments. Many of these options could result in changes to the ash that will be detrimental to
the performance of the ESP causing Increased emissions of fine particles and higher opacity. For
example, a switch to low-sulfur coal significantly increases particle resistivity while low-NOx
burners increase the carbon content of ashes. Both of these changes could result in derating of
the boiler to comply with emissions standards.
ADA Environmental Solutions has developed a chemical additive that is designed to improve the
operation of ESPs to bring these systems into compliance without the need for expensive capital
modifications. The additives provide advantages over competing technologies in terms of low
capital cost, easy to handle chemicals, and relatively non-toxic chemicals, hi addition, the new
additive is insensitive to ash chemistry which will allow the utility complete flexibility to select
the most economical coal. Finally, this technology is effective in applications where
conventional SOs flue gas conditioning does not work such as hot-side ESPs and cold-side ESPs
that operate above 375ฐF. This paper will present the results of several full-scale demonstrations
that have been sponsored by DOE, EPRI, and private utilities. These will include ESPs
collecting flyash from various low-sulfur coals including Powder River Basin coals, washed
eastern coals, and blends.
-------
Introduction
ADA Environmental Solutions has commercialized a proprietary, flue gas conditioning agent to
provide utilities and industries with a cost-effective means of complying with environmental
regulations on particulate emissions and opacity. The flue gas conditioning system decreases
particle resistivity which will improve the performance of electrostatic precipitators (ESPs) with
resistivity related performance problems. ESPs serve as the primary air pollution control device
for the majority of coal-fired utility boilers in the Eastern and Midwestern United States.
Particle resistivity is the dominant parameter affecting the performance of an ESP. When the
resistivity is too high, the ESP must be increased in size by a factor of 2 to 3 resulting in
proportionally increased capital and operating costs. High resistivity is produced by low-rank
and low-sulfur coals and by many other thermal and industrial processes. Decreasing resistivity
and increasing cohesivity of the particles are the most effective means to increase collection
efficiency and decrease outlet emissions from ESPs. Resistivity modification can improve
electrical conditions, while increasing the cohesiveness of the particles can reduce emissions
from reentrainrnent. Both of these properties can be modified through flue gas conditioning
(FGC).
Development of a cost-effective technology to decrease outlet emissions, and associated
particulate air toxics, is driven by a comprehensive plan in Title III of the Clean Air Act
Amendments of 1990 to achieve significant reductions in emissions of hazardous air pollutants
from major sources. The mandate to reduce air toxics will require upgrading particulate control
equipment such as electrostatic precipitators and fabric filters, since many of the toxic metals
exist as fine particles. Concentration of trace metals in the fine particle fraction may force
utilities to achieve very high overall particle collection efficiencies. This presents a challenge to
users of existing particulate control equipment to identify modifications or upgrades which
increase the collection efficiency of fine particles.
A related development is that reducing 862 and NOx emissions from utility boilers has the
potential to substantially increase particulate emissions from existing ESPs. This occurs because
switching to low-sulfur coals as an SC>2 control strategy exacerbates ESP performance problems
associated with high-resistivity flyash. In addition, several utilities that have installed low-NOx
burners have experienced increased particulate emissions following combustion modification.
This can be caused by several different changes to the gas stream that impact ESP performance
such as: increased back-end temperatures which increases gas velocity and particle resistivity;
increased LOI resulting in a greater number of carbon particles which are difficult to capture in
an ESP; and increased particle loading due to a shift in the flyash/bottom ash ratio.
This paper presents results from laboratory, pilot-scale, and full-scale test programs
demonstrating the capabilities of this new technology. Test programs were conducted on both
hot-side and cold-side ESPs burning a variety of low-sulfur coals. Data showing reduced particle
resistivity, increased ESP power, and decreased opacity will be presented for different
applications.
-------
Flue Gas Conditioning with SOS
For cold-side ESPs collecting high-resistivity dusts, increasing the SO3 concentration in the flue
gas will result in a reduction of particle resistivity. Conditioning of the ash by injecting small
amounts of S03 into the flue gas is a well proven technique for improving the performance of the
ESP. In most commercial SO3 conditioning system, SO2 is transported to the plant where a
catalyst converts it to SO3 before being injected into the duct. Newer configurations, such as
EPRICON, use a catalyst to oxidize native SCh in the flue gas to generate SO3.
Although SO3 conditioning is an effective technique for resistivity reduction, it does have several
disadvantages. By increasing the SO3 concentration, the potential for corrosion due to H2SO4 is
increased. The mechanism by which SO3 conditioning lowers resistivity is by increasing the acid
dew point to a level closer to the ESP flue gas temperature to enhance condensation of acid onto
the surface of the particles. One of the benefits of switching to a low-sulfur coal is the reduced
potential for corrosion because both the acid dew point temperature is lower and the
concentration of H2SC>4 is reduced; SO3 conditioning negates this benefit.
SO3 flue gas conditioning systems are rather expensive. The hardware required to convert SC>2 to
SO3 and control flow rates requires a capital investment of approximately $2,500,000 for a 500
MW plant, hi addition, the cost of the SC>2 and the energy to maintain the converter at 800ฐF
involves a significant operating expense. Since SO3 and NH3 are both toxic, these systems create
worker health and safety concerns. Leakage of the ammonia tanks could lead to exposure of the
public.
There are many situations where SO3 conditioning is not very effective. As temperatures
increase above approximately 350ฐF, very high concentrations need to be injected to obtain
decreasingly smaller benefits. When the ash is high in alkali materials, such as calcium or
magnesium, the SO3 reacts with the flyash which renders it inactive. With some of the more
acidic flyashes, ammonia has to be added so that the ash will respond to conditioning. This ash
when collected is difficult to dispose of because of the odor produced by the ammonia.
New and Emerging Regulations
There are several new and emerging regulations that could impact decisions on flue gas
conditioning systems. On Earth Day this year, President Clinton announced that coal-fired
electric utilities would have to start reporting the release of toxic chemicals from their facilities.
These Toxic Release Inventories (TRI) must be file annually by each plant to report the amount
of each toxic chemical that was emitted by the plant as a gas, liquid, or solid during the previous
year. Therefore, any ammonia injected to condition the flue gas will have to be reported as it will
exit the plant in either the gas phase or as contained in the ash.
EPA is also considering new regulations to require point sources to measure and report
condensable particulate matter as part of the total participate emissions. Although, SO3 is
injected in the vapor form, upon cooling, it reacts with the water vapor in the flue gas and
-------
condenses as a sulfuric acid mist. This would be collected on a filter system designed to measure
condensable particulate. This could result in a significant increase in particulate matter. The
complete condensation of 15 ppm of sulfuric acid vapor results in an acid mist mass loading of
approximately 0.025 gr/dscf, or a mass emission rate of about 0.05 Ib/MMBtu. This would
exceed the latest New Source Performance Standards by nearly a factor of two.
Several states have recently enacted regulations requiring utilities to include any opacity
excursions that occur after the unit is offline as exceedances. These "off-line" emissions occur
after the unit has stopped feeding coal to the boiler but continues to induce air flow through the
unit to accelerate cooling. Without the moisture being supplied to the gas by the hydrogen in the
fuel the resistivity of the ash on the plates increases dramatically leading to near instantaneous
reductions in power. This occurs even with SOj conditioning and results iin high opacity levels.
Laboratory Results with ADA-23 ESP Conditioning
Based upon the fact that ADA-23 matched performance of SO3/NH3 in a fabric filter, tests were
conducted to measure its impact in an ESP.1 ADA-23 was immediately seen to be successful as
a resistivity modifier in laboratory trials. Redispersed flyash was obtained from a utility, which
fires a Texas lignite coal. Sulfur content of the lignite coal is approximately 0.6% and the flyash
is low in alkali metals, sodium and potassium. It has long been observed that the combination of
low-sodium and low-sulfur content is associated with high-resistivity conditions at cold-side ESP
temperatures for a variety of coals.
Two parameters were varied to determine the impact of additive ADA-23 on flyash resistivity.
Tests were run over an extended flue gas temperature range of 200ฐF to 450ฐF. The
concentration of additive was tested over a range comparable to that for SO-j conditioning
systems. The additive demonstrated dramatic reductions in resistivity of the lignite flyash over
the entire temperature range tested. The effect of concentration of the additive on resistivity at a
constant temperature of 310ฐF is presented in Figure 1, where the impact is shown to range from
about two to over four orders of magnitude reduction in flyash resistivity. It should be noted that
the liquid injection rate was identical for all test cases. For tests without additive, water was
injected at the same liquid feed rate.
Similar tests were run on flyashes from a variety of different low-sulfur coals including several
Powder River Basin coals. In all cases, the resistivity could be decreased to the optimal 1010
ohm-cm range. Typical test results from a Powder River Basin ash are plotted as a function of
flue gas temperature hi Figure 2. Significant reductions in flyash resistivity were measured for
all temperatures in the test matrix.
While the results from these tests were excellent and it appeared that ADA-23 would provide an
alternative to SOs/NHs conditioning, a new technical challenge came to mind. Would this
additive work at hot-side ESP temperatures (greater than 600ฐF)? The problems unique to hot-
side ESPs have been virtually without solution for the past 20 years. Options to meet particulate
removal requirements include a cold-side retrofit or installation of a baghouse. To test whether
-------
ADA-23 could be used to condition hot-side ESPs, laboratory resistivity tests were conducted.
Initial tests were first run at 700ฐF with an ash sample from a hot-side ESP at a plant firing a
western subbituminous coal. This ESP has historically exhibited severe performance problems
associated with a low-sodium content hi the coal. Resistivity of fresh ash at hot-side operating
temperatures has typically been measured in the 1010 -1011 ohm-cm range; it is expected that the
resistivity increases significantly over time as the ash ages on the ESP plates. When the additive
and ash were injected at 700ฐF in a 10% moisture, air environment, the resistivity was reduced by
up to two orders of magnitude. The amount of the resistivity reduction can be controlled via the
concentration of the additive. This test demonstrated that conditioning could be effective at hot-
side temperatures.
Pilot-Scale Results with ESP Conditioning
The dramatic laboratory results led to several additional field tests on flue gas slipstreams at
different sites and with different coals.2 Two performance tests with small-scale ESPs were
conducted for ADA-23. In the first, a 300 acfm slipstream of the offgas from a trial burn of a
Powder River Basin coal at the Southern Research Institute (SRI) Coal Combustion Facility
(CCF) was conditioned in a large spray contact chamber. In the second test, a 5,000 acfm flue
gas slipstream taken downstream of the air preheater at a Colorado power plant was conditioned
by additive spray injection in an 18 inch duct.2 In all of these tests, the conditioning produced
decreased resistivity, increased power, and improved ESP performance.
Additional pilot-scale tests were conducted on a 1 MW cold-side ESPs. Tests were conducted at
380ฐF which was representative of peak summertime operating conditions for many utilities.
This is a difficult application because conventional SOa flue gas conditioning is ineffective at this
temperature. Tests were run for two different coals a washed medium-sulfur Alabama coal
and a low-sulfur Powder River Basin coal. The results were identical for both ashes. Figure 3
shows typical test results. At baseline conditions, the high-resistivity ash severely limits that
power available for particle collection by the ESP. However, with the injection of the
conditioning agent, the power increases by greater than a factor often. When the injection was
halted, the ESP returned to poor operating conditions.
Full-Scale Demonstration Test Description
Central & South West Services (CSWS), a utility holding company that has several hot-side
ESPs, became interested in the potential benefits that this flue gas conditioning agent could
provide.3 Hot-side ESP performance degrades over time because of sodium depletion of the ash
layer. When this condition becomes severe, opacity increases above compliance limits and the
ESP must be brought off-line for washing, which results in lost generation and power sales.
ADA-23 showed the potential to control resistivity and eliminate undesirable conditions such as
back-corona and sparking at reduced voltage due to advanced sodium depletion. The value
associated with this potential improvement was significant and funding mechanisms were
-------
immediately pursued. The final agreement was a "Tailored Collaboration" contract between
EPRI, CSWS, and ADA Technologies.
The test was organized into four phases, with the continuation of each phase dependent on
positive results from the previous phase. A brief description of each phase follows:
Phase I: Evaluate the effect of ADA-23 on host-site ash resistivity in the laboratory.
Phase II: Evaluate the effect of ADA-23 on ash resistivity when a slipstream sample is
extracted from the host site hot flue gas duct and subsequently conditioned.
Phase III: Design, fabricate and install an additive injection system and conduct a short-term,
two week demonstration on one half of an existing hot-side ESP.
Phase IV: Conduct a long-term demonstration of flue gas conditioning with ADA-23 to
optimize injection parameters and document performance.
Results from the first three phases were reported in an earlier paper.4 The following section
describes results from Phase IV.
Phase IV Results
The objective of Phase IV was to evaluate the long term operation of the full-scale additive
system and its impact on the performance and collection efficiency of a hot-side ESP. Phase TV
also allowed for a more thorough evaluation of the injection system and the opportunity to make
the appropriate mechanical improvements.
The demonstration was conducted on a 550 MW boiler firing a low-sulfur western coal. Figure 4
shows the configuration of the ducting and ESP. The duct splits into two parallel systems just
downstream of the economizer with separate ESPs. Each ESP treated a flue gas flow of
approximately 1,600,000 acfm at 750ฐF. This arrangement allowed a direct comparison between
one ESP that was conditioned and the other ESP with no conditioning. Each ESP had an opacity
monitor downstream to monitor particulate emissions.
Phase IV began in June 1996 and continued until October. The injection system was similar to
that used in Phase HI. Improvements were made to the injection lances, a storage tank was
installed, and the injection skid was upgraded for remote operation. Inspection during this test
showed that build-up on the injection lances experienced in Phase III was substantially improved.
This was important to overall system reliability. Since the lances are easy to remove, inspections
were made every two weeks. Recent design improvements will reduce this even further.
Problems with the additive skid control program that became evident in Phase ITI were
eliminated and changes were made to increase versatility of the skid.
-------
Results during this longer-term test showed that performance improvements documented in
Phase HI were repeated. The effect of flue gas conditioning with ADA-23 is evident in both
increased ESP power levels and decreased outlet opacity. Figure 5 shows the impact of
conditioning during a two-week period in September. With the injection system turned off,
September 14, the power levels from both the treated and untreated side were nearly identical.
However at the optimum feed rate, September 16-25, the power levels for the treated side
increased by 100 kW over levels from the untreated side. When the injection system was turned
off, the power levels began to decrease to the untreated levels.
Figure 6 shows the opacity levels during this same tune frame. With the injection system off,
September 14, the opacity on the treated side was typically higher than the untreated side.
However, during injection at the optimal level, the opacity on the conditioned side falls well
below that on the unconditioned side. When the injection system was turned off on September
25, the opacity on the treated side returned to a level greater than the untreated side. This
demonstrates that the increase in power produces improved ESP performance resulting in
decreased emissions.
Figure 7 shows the impact of the conditioning agent on a much shorter time frame. When load is
reduced at night, the flue gas temperature drops, producing increased particle resistivity and
increased opacity. For many hot-side ESPs, low-load conditions represent the most difficult
period of operation. For example, on this ESP the hourly duct opacity on the unconditioned side
increased to greater than 35% at reduced load. However on the treated side, the conditioning
agent effectively reduced the particle resistivity producing improved ESP operating conditions
which reduced opacity to below 5%. This dramatic decrease in emissions demonstrates the
capabilities of this new technology.
Cold-Side ESP Results
Utilities with cold-side ESPs are actively pursuing methods to maintain ESP performance while
firing lower-sulfur, compliance coal. Pilot testing with ADA-23 showed that this additive
effectively reduces flyash resistivity when low-sulfur coal is fired, even at temperatures where
SC>3 is ineffective. Two utilities that are currently assessing Powder River Basin (PRB) coals at
their stations evaluated ADA-23 during PRB test burns. These demonstrations were completed
in March 1997.
The first demonstration was on a nominal 250 MW unit where high flue gas temperatures have
historically caused less than ideal ESP performance. The goal of the plant was to be able to burn
a blended coal consisting of at least 80% PRB coal. In the past, any attempt to burn more than a
60% PRB coal required significant derating of the unit to avoid opacity exceedances. During the
demonstration program, the plant was able to achieve its goal by burning an 80% PRB blend and
without exceeding opacity limits.
The second demonstration was cofunded by EPRI. At this site ADA-23 was compared directly
to SOs and SOs/NHs conditioning. Testing was completed on two different coals, including
-------
100% PRB. The results from this test showed that conditioning with ADA-23 was comparable
to SOs in maintaining ESP design operating performance on both coals. A new injection skid
designed for commercial use and modified injection lances were evaluated during this test. Both
systems proved to be reliable and addressed design concerns encountered during the first
demonstration at Central & South West Service's station.
A third cold-side demonstration is currently underway on a 400 MW plant. This demonstration
will provide a side by side comparison with SOs/NHs which is being injected on a parallel unit
burning the same coal.
Ash Characteristics
One property that was important in the development of these new additives was ash
characteristics after conditioning. It was a goal that the application of the additive would not
detrimentally impact ash chemical characteristics or odor. Based on tests that ADA
Environmental Solutions has conducted to date, additive conditioning with ADA-23 at effective
concentrations does not alter flyash chemical and handling characteristics. During a trial at a
Colorado power plant firing a low-sulfur Powder River Basin coal, samples of ash from a pilot
particulate control device were taken during a baseline period with no conditioning and during
additive injection. The samples were submitted to an independent testing laboratory for
elemental analysis and for TCLP analysis for leachable metals. The TCLP results indicate that
the conditioned flyash is not enriched in any of the leachable metals, including the volatile metals
mercury and selenium. Conditioning with ADA-23 should produce a non-hazardous waste
which will continue to be suitable for commercial sale or conventional landfilling.
Conclusions
The ADA-23 additive promises to be an effective conditioning agent for ESPs and baghouses to
improve removal of fine particles from coal-fired combustion flue gas streams. This new flue
gas conditioning agent offers a solution to the time-dependent deterioration of hot-side ESPs and
reduces resistivity for cold-side ESPs, including those operating at temperatures greater than
375ฐF.
The advantages associated with this new additive technology include:
Increased particle cohesion and hence better collectability and easy cleaning;
Reduced particle resistivity;
Temperature and sodium effects are not factors controlling performance;
Provides a practical, cost-effective solution to hot-side ESP operating problems;
-------
The agent is insensitive to ash particle chemistry which allows users a wider range of options
in purchasing coal;
The injection system is simple and low capital cost;
Conditioned ash remains suitable for commercial use;
Ash handling is not impaired at conditioning levels required for resistivity control.
Several additional demonstrations on both hot-side and cold-side ESPs will occur in the next six
months. These demonstrations will provide the opportunity to evaluate different coals,
temperatures, alternative injection scenarios, and long term operations. Recent data shows that in
some cases, conditioning does not need to be continuous.
References
1. CJ. Bustard, K.E. Baldrey, T.G. Ebner, S.M. Sjostrom, R.H. Slye, "Demonstration of Novel
Additives for Improved Fabric Filter Performance," presented at EPRI/DOE International
Conference on Managing Hazardous and Particulate Air Pollutants, Toronto, Canada,
(August 1995).
2. K.E. Baldrey, J.R. Butz, "Flue Gas Conditioning for Improved Particle Collection in
Electrical Precipitators, Roll-up Monthly Status Report, March through November, 1994,"
ADA Technologies, DOE Contract No. DE-AC22-91PC90364 (December, 1994).
3. N.N. Dharmarajan, B.W. Moore, R.L. Chang, R.L. Glover, M.D. Durham, K.E. Baldrey, C.J.
Bustard, C.E. Martin, "Cost-Effective Alternative Solutions for Hot-Side ESPs," presented at
POWER-GEN '96 International Conference, Orlando, FL (December, 1996).
4. M.D. Durham, K.E. Baldrey, C.J. Bustard, C.E. Martin, "Flue Gas Conditioning for
Improved Electrostatic Precipitator and Fabric Filter Performance," presented at Air & Waste
Management Association's 89th Annual Meeting, Nashville, TN (June 1996).
-------
FIGURES
1E+13
1E+12
V 1E+11
1E+10
5; 1E+09
1E+08
0.0
0.1
Relative Additive Concentration
1.0
Figure \
The Effect of Additive ADA-23 on Flyash Resistivity
as a Function of Concentration for a Texas Lignite at 300ฐF
E 1E+12
ฃ 1E+11
& 1E+10
f 1 E+09
.2 1E+08
(A
S 1E+07
1
Baseline
^ Additive
A A
A
300 350 400 450 500 550 600 650 700
Temperature (ฐF)
Figure 2
Flyash Precipitation Rate as a Function of Temperature for a Powder River Basin Coal
-------
1400
10
20 25 30
Time (Hours)
35
40
45
Figure 3
The Effect of Additive ADA-23 on ESP Power Levels
Tests Conducted on a IMW(e) Pilot ESP Downstream of a Boiler Firing PRB Coal
A Side
BSide
Figure 4
Ductwork Configuration During Central and South West Services Demonstration
-------
300
- ป- Treated
-Untreated
Figure 5
Effect of Conditioning on ESP Power During Full-Scale Demonstration
- ซ- Treated
- Untreated
Figure 6
Effect of Conditioning on Opacity During Full-Scale Demonstration
-------
Untreated
Figure 7
Effect of Conditioning on Hourly Opacity
-------
Summary of Wet Esp Operation at NSP's Sherco Station
R.M.Henningsgaard
Northern States Power Co.
Sherburne County Generating Plant
1399 Industrial Blvd.
Becker Mn. 55308
S.D.Lynch
Southern Environmental, Inc
6536 West Nine Mile Road
Pensacola, Florida 32526
R.F.AItman
Electric Power Research Institute
516 Franklin Building
Chattanooga, Tennessee
Abstract
Northern States Power Company's Sherbume County Generating Plant air quality control system
consists of 12 individual scrubber modules on each of its first two units. These are rod venturi
scrubber modules designed to remove both flyash and SO2 from the incoming flue gas. Although
the venturi scrubbers are capable of removing 99% of the incoming flyash, the remaining 1%
consists of submicron paniculate that causes excessive opacity in the stack. After reviewing the
available control options, it was determined that wet esp's had the most potential for removing
this submicron particulate. NSP contracted with Southern Environmental Inc to design and
retrofit a wet esp within the spray tower of one of Sherco's existing scrubber modules. The wet
esp scrubber module became operational in May 1995. Based on over two years of successful
operation, NSP has decided to retrofit 15 more of Sherco's scrubber modules with wet esp's. This
report reviews the results of Sherco's wet esp testing.
Introduction
This report summarizes the results of testing of a wet electro-static precipitator (esp) that was
retrofitted within one of Northern States Power Company's (NSP) Sherbume County Generating
Plant (Sherco) Units 1 and 2 scrubber modules. Sherco's wet esp is the single largest wet esp
designed to handle coal fired flue gas in the United States. Sherco's Unit 1 has been in operation
since 1976. Unit 2 came on line in 1977. Both Sherco units 1 & 2 are capable of loads up to 760
-------
MW. Both units have identical Combustion Engineering tangential pulverized coal fired boilers.
The permitted fuel source is low sulfur Powder River Basin coal
Figure 1 illustrates how Sherco's scrubber modules are laid out within the plant. There are 12
scrubber modules on each unit, 10 of which are required for full load operation. Consequently,
each module is rated for 75 MW, leaving two modules on each unit available for maintenance or
cleaning as required. Sherco has no upstream particulate control device. The scrubbers are
designed to remove both the flyash as well as the SC>2 coming from the boilers.
Figure 1:
NSP Sherco Unit 1 & 2 Combustion Air/Flue Gas Path
The Sherco's scrubber module design is illustrated in figure 2. Each module handles 660-830
klbs/hr of flue gas at 275-325ฐF. The flue gas is initially contacted within the venturi, consisting of
125 closely spaced venturi rods. The venturi sprays adiabatically cool the flue gas to 120- 125ฐF,
removing 99% of the incoming flyash and 50% of the incoming SCv The venturi liquid-to-gas
ratio is approximately 5-10 gal/kacfin. The flue gas then turns the comer into the spray tower,
where approximately 50% of the remaining SO2 is removed at a hquid-to-gas ratio of 8-14
gal/kacfin. Sherco's scrubbers were initially designed to add limestone for SO2 removal; However,
the Powder River Basin coal that Sherco currently uses has sufficient calcium in its ash to provide
all the required SOa removal Two sets of demisters, single pass flat demisters and double pass
chevron demisters, remove entrained slurry droplets from the flue gas stream. The flue gas then
passes through a reheater which heats the flue gas to 160ฐF before leaving the module. The
calcium sulfite shrrry in the module's reaction tank is force oxidized to essentially 100% calcium
suifate. A vertical mixer is used to keep the shiny solids in suspension. Slurry is routinely bled
from the reaction tank to maintain the shrrry solids level at approximately 10%.
-------
X-^Vl HoSBwTMT
ป*ป f-0 ) iSaM
~ L V_/ Votta
Figure 2
NSP Sherco Standard Design Module
Sherco's venturi scrabbers remove paniculate relatively well Their emissions average about 0.07
Ibs/Mbtu. They also do a good job of removing SCb, emitting about 0.20 Ibs/Mbtu. However,
stack opacity has been a problem Minnesota law requires coal fired units of Sherco's vintage to
maintain stack opacity below 20%. Sherco routinely runs between 30 and 40% opacity. NSP has
operated Sherco under a variance since its startup, allowing it to exceed its 20% opacity limit.
However, this variance is due to expire by the end of the year 2001.
The causes of Sherco's high opacity has been studied extensively. Two important factors are
involved. First, the flue gas from both units are emitted from a common stack The diameter of
the stack (32.5 ft) is an abnormally large optical path length. Second, the particles that are not
removed by the venturi are primarily submicron in diameter. The average diameter of the emitted
particles is 1/2 micron, the size range known to have the greatest impact on opacity. Particle
sizing at the Met to the scrubbers shows the mass mean diameter to be about 10 microns.
However, there is sufficient submicron particles to account for the number emitted downstream.
As result, Sherco's boilers generate a sufficient quantity of submicron particles that pass through
the venturi scrubbers to cause the high stack opacity. It has been estimated that reducing Sherco's
particulate emissions to approximately 0.03 Ibs/Mbtu would reduce its opacity to 20%. This is
illustrated in figure 3, which shows a composite of all of Sherco's stack tests with their
corresponding stack opacity readings.
-------
S. 30-
O
Variance Limit
State Limit
NSPS
Operating Permit Limit
0.02 0.04
Particulate Emissions (Ibs/Mbtu)
0.12 0.14 0.16
Figure 3
NSP Sherco Paiticulate Emissions vs Corresponding Stack Opacity
Initial Design Concept
NSP in conjunction with Black & Veatch studied the various options available for bringing Sherco
back into compliance with the 20% opacity limit. It was ultimately decided that retrofitting wet
esp's within the existing scrubber modules was the option with the most potential The wet esp
could replace both the flat and chevron demisters, since the wet esp is itself a highly efficient
demister. Additional space is provided by relocating the spray tower headers to the crossover duct
between the venturi and the spray tower. The residence time lost by relocating these sprays could
be compensated for by increasing the flow of the spray pumps, which are currently underutilized
on Sherco's scrubber modules. In 1991, NSP decided to test this concept by performing a
slipstream test. Upon the successful conclusion of the slipstream testing, NSP decided to continue
testing the wet esp concept at full scale on one of Sherco's scrubber modules. In 1993, NSP
entered into an agreement with EPRI to cofund a one module wet esp test using EPRI's Tailored
Collaboration funding mechanism. Dr. Ralph Airman, the project's primary EPRI contact, became
a member of the design team which drew up the wet esp bid specification and evaluated the
designs from the prospective wet esp vendors.
The successful bidder for the module wet esp design was Southern Environmental Inc. (SEI).
Figure 4 illustrates how SEI's wet esp is situated within Sherco's scrubber module. The module
has two wet esp fields. Each field has 600, 5 foot long rectangular collector tubes. An electrode is
centered within each tube with a 4" clearance to ground. SEI's design supports their electrodes
solely from the top, eliminating the requirement for insulator support compartments in the
relatively wet and dirty space below the wet esp. Wash sprays are situated above and below each
-------
Run PUT
Figure 4
SEI's Wet Esp Scrubber Module Design
field. The fog sprays located below each field point up and are designed to operate continuously
while the module is in service. The flush sprays located above each field point down and are
designed to operate only when the module is taken out of service for a flush. The out of service
flush is scheduled once each day while the unit is down on load (usually at night). A catch basin
dernister is situated directly below the lower field wet esp. The top side of the catch basin
captures water cascading from the wet esp wash sprays above, diverting it to the surge tank,
thereby segregating it from the slurry in the reaction tank below. The bottom side of the catch
basin, acting as a two pass demister, removes the shiny droplets entrained in the flue gas coming
from the crossover duct, preventing slurry droplets from severely loading the lower field wet esp.
The catch basin demister also helps distribute the flue gas across the wet esp by creating a 1/2
inwc pressure drop across it. Wet esp wash water removed by the catch basin is recirculated from
the surge tank by the return pump to a treatment tank situated above the scrubber inlet ducts.
Wet Esp Module Operation
During the course of the wet esp module testing, the module's venturi was set in three different
configurations:
I 077100T11MPW
-------
High Venturi Dp.... This is the venturi dp that is normal for the other modules, ranging
between 13 and 21 inwc depending upon the number of modules in service and the unit load.
Low Venturi Dp This is the minimum venturi dp that was deemed to be feasible, ranging
between 1 and 4 inwc depending upon the number of modules in service and the unit load.
Intermediate Venturi Pp....This was a venturi dp approximately halfway between the high and
low venturi dp setting, ranging between 8 and 13 inwc depending upon the number of
modules in service and the unit load.
The main question to be addressed was how long the wet esp would remain clean for each of the
venturi dp settings. The answer to this question will be discussed in the following sections.
High Venturi Dp Operation/Continuous Wash Mode (May 18 - Sep 26, 1995)
The wet esp module was initially setup in the high venturi dp mode. Upon startup of the module it
was immediately evident that the sprays designed to wash the top of the bottom field could not
have flow high enough to provide full coverage without reducing the lower field power output to
unacceptably low levels. Consequently, it was necessary to compromise by accepting only partial
spray coverage of the top of the lower field. The sprays designed to wash the bottom of the lower
field were acceptable.
Figure 5 shows the wet esp operation for the high venturi dp phase after startup. The graphs
illustrate the secondary kV for each wet esp field as well as the resulting opacity in the duct
downstream of the module. The opacity monitor path length is corrected to the stack diameter so
the opacity reading would be equivalent to the stack opacity if all the modules had wet esp's.
Based on ability of the wet esp to maintain voltage levels that were sufficiently high to keep
opacity low throughout the test period, the wet esp module operation at high venturi dp was
considered a success.
5/1/95 5/16/85 5O1/85 6/15/85 6/30/85 7/15/85 7/30/85 8/H/9S 8/29/95 9/13/95
Figure 5
High Venturi Dp/Continuous Wash Operation
-------
Low Venturi Dp Operation (Sep 27, 1995 - Mar 6, 1996)
The next phase of the test involved configuring the wet esp module for low venturi dp operation.
Figure 6 illustrates the results of this phase. As shown on the graphs, the power output of both
fields quickly deteriorated within short periods of time. Because the venturi was less efficient at
the low dp's, more ash was allowed to penetrate to the wet esp. While the continuous wash, of the
bottom of the lower wet esp field was able to keep the portions of the collector tube reached by
the sprays clean, the wet esp prevented the water from rising more than 6 to 9 inches above the
bottom of the collector tubes. As a result, the ash simply built up beyond this point to limit the
output of the lower field. As the lower field power output dropped, ash loading to the upper field
increased, bringing down its power output. Ultimately, the power output became too low to
sustain opacity at acceptable levels. The module then had to be taken down to have the wet esp
cleaned.
By March 6, after three unsuccessful attempts at miming at low venturi dp, it was decided that the
low venturi dp mode of operation would not be feasible with SEFs wet esp scrubber module
design.
-Upper tv D Lmmf tV QOpadt)r \
10/12/95 10/27/95 11/11/95 11/26/95 12/11/95 12/26/95 1/10/96 1/25/96
Figure 6
Low Venturi Dp Operation
Intermediate Venturi Dp Operation (Mar 8 - Sep 29, 1996)
Having established the success of high venturi dp operation and the failure of low venturi dp
operation, the question now was....how low was too low? The next test phase sought to answer
this question. The venturi dp was set up at an intermediate level approximately halfway between
the high and the low venturi. The results of the intermediate dp testing is shown in figure 7.
The wet esp started this phase of the test with continuous wash sprays on the bottom of the lower
field wet esp. By May of 1996, the lower field power output had dropped to unacceptable levels.
Two separate attempts were then made to clean the wet esp with nitric acid. Neither attempt
could successfully remove the majority of the buildup from the lower field. Consequently, on May
tBPM H1QปTซBOTJ68&(ป
-------
24th the wet esp was set up to run without wash sprays while the module was in service, making
the lower field rely solely on the daily out of service flush as does the upper field. The lower field
power output promptly rebounded. Consequently, the test was able to continue.
After switching to the dry mode of operation, the wet esp was able to run in the intermediate
venturi dp mode for 4 additional months before it required cleaning, making the total time
between cleanings interval 9 months. However, recall that this phase of the test did not start with
clean collector tubes on either field and since the previous mode was the low venturi dp
operation, the equivalent time since the previous cleaning could be considered to be much longer.
It is estimated that the lower field would likely have gone perhaps as long as 12 months had it
started in this mode with completely clean collector tubes. NSP considers this time between
cleanings interval to be more than adequate. Consequently, the intermediate venturi dp mode of
operation was considered a success.
o
I
8,
&
ง
s
3/7/96 3/22/96 4/6/96 4/21/96 5/6/96 5/21/96 6/5/96 6/20/96 7/5/B6 7/20/96 B/4/96 0/19/96 9/3/96 9/16/96
Figure?
Intermediate Venturi Dp Operation
Based on the success of the wet esp operation at high and intermediate venturi dp's, NSP decided
to install wet esp's within 15 additional scrubber modules at Sherco. A full conversion of all 24 of
Sherco's scrubber modules was deemed unnecessary if operation of wet esp modules at high
venturi dp's could reduce the opacity through those modules to 5%. It was estimated that 7-8 wet
esp modules in conjunction with 2-3 nonwet esp modules on each unit would be sufficient to
reduce stack opacity far enough below 20% to provide an adequate margin of safety.
High Venturi Dp/Dry Mode Operation (Oct 28, 1996 - Apr 4, 1997)
Given NSP's decision to proceed with a partial wet esp conversion, it was decided to refocus the
test efforts again on high venturi dp along with the dry mode of operation based on the dry mode
operation's potential to substantially simplify the final wet esp design. Therefore, after cleaning
the lower field wet esp (no cleaning was done on the upper field at that time), the wet esp module
was reconfigured to operate in this mode. Figure 8 illustrates the operation during this test phase.
-------
As shown, the wet esp module performed very well throughout this test period. By April 4th
Sherco's unit 2 came down for an extended 2 month outage, effectively idling the wet esp for the
duration of the outage. By that time the lower field had operated over 5 months since it was last
manually cleaned and voltage was still in the upper 20 kV range. This is less than one third of the
total reduction in kV that the lower field was deemed to be capable of before cleaning would be
required. Consequently, if it is assumed that the voltage reduction would be relatively constant
over time, a between cleanings interval in excess of 15 months can be inferred. At the same time
the upper field had operated almost 15 months since its last manual cleaning while maintaining its
relatively high power output. Based on these results, It is clear that operating the wet esp module
at high venturi dp while in the dry mode of operation is a highly successful mode of operation for
Sherco's wet esp scrubber module.
I O Upper kV QImnr tV OCorOpnc
10/1/86 10/10/96 10131/96 11/15/95 11OO/96 12/15/96 12/30/96 1/14/87 1/29/97 2/13/97 2/2S/97 3/15/97 3/30/97 4/H/97 *29/97
Figure 8
High Venturi Dp/Dry Mode Operation
Wet Esp Cleaning
The time interval between wet esp cleanings is only one part of the equation for determining the
success of the wet esp concept at Sherco. Another consideration was how quickly the module
could be cleaned and at what cost. The existing scrubber module reheaters routinely require
cleaning every 3 to 4 months on average. When taken out of service for cleaning, about 4 to 5
days is required to clean the module from top to bottom. Consequently, NSP has cleaning crews
which routinely go through a module on each unit every week This leaves one other spare
module available to be worked on as required, leaving the other ten modules on each unit
available for ftdl load operation.
Given the ability of SEI's wet esp module to operate 9 to 18 months between cleanings while
running at intermediate and high venturi dp's, a maximum of 3 to 6 weeks of cleaning time
intervals would result in a labor hours equivalent to the current situation. On Sherco's wet esp
module, high pressure washing of the reheater has not been required. Consequently, a significant
portion of the work required on the other modules has been eliminated. The main area to be
cleaned on Sherco's wet esp module is the lower field collector tubes. The upper field has been
11 PW HiQ*TAlECRUBeSIปซฃTEH^Bซ1SnCOWrwnปOPWT DOC
-------
capable of running much longer. In fact it has not had to be cleaned yet while operating hi the
intermediate or high venturi dp mode of operation.
The ash deposits that occurred on the wet esp collector tubes tended to be relatively soft in nature
compared to the hard calcium sulfate scale that generally forms on the reheaters and demister
surfaces of the other modules. It is possible to remove these deposits with a putty knife or any
other tool having a steel blade. Analysis has shown that the primary constituent of the wet esp
deposits are calcium sulfate. Therefore, it may be the case that the scale inhibitor chemical is
having the desired affect upon the calcium sulfate by preventing it from forming into the hard
scale deposits that are typical on Sherco's other scrubber modules.
While several manual cleaning methods have been tried, the method which had the most success
has been to use a high pressure washer which Sherco refers to as a hydrolaser. The hydrolaser is
typically used to clean reheaters on the other scrubber modules. It is capable of pressures up to
15,000 psi. A six foot long wand was used with a nozzle which sends out a cone having a
diameter large enough to cover the perimeter of the tube with sufficient pressure to remove the
buildup on a single pass. Thus, a tube can be cleaned by simply lowering the nozzle into the tube
from above. A second pass might be required to completely clean deposits on the other side of the
electrode. A side benefit to this method is that the electrode is cleaned at the same time. Using
this method, approximately 5 minutes was required to clean one of the 600 tubes on a field. As a
result, approximately 1 week was required to clean a single field with crews working around the
clock.
Wet Esp Scrubber Module Performance
The discussion so far has focused primarily on how the wet esp has performed over tune with
regards to collector tube buildup. The following sections will focus on other aspects of the wet
esp performance such as voltage vs current curves, opacity reduction capability, particulate
removal, air toxics removal, and SC>2 removal.
Voltage (kV)
Figure 9
Upper and Lower field Wet Esp Voltage vs Current
10
-------
Voltage vs Current Curves
Figure 9 illustrates the relationship between current and voltage for both upper and lower wet esp
fields. These curves are based on actual numbers taken on June 26, 1997. Note that the Upper
field's secondary current limit is 1650 mA and the lower field's secondary current limit is 1200
mA.
Opacity Reduction
Based on the results of the long term testing, figure 10 illustrates the expected relationship of
venturi dp to opacity for the wet esp module design. The clean operation on the graph refers to
operation that occurs initially after startup with collector tubes that are completely clean. This is
the optimum performance curve. The steady state curve is what the performance settles into over
time as the collector tube accumulates calcium sulfate scale. Generally, the operation reaches
steady state relatively quickly (after a couple weeks), then degrades more slowly as the calcium
sulfate scale accumulates on the tube. Of course the operation degrades quicker with lower
venturi dp.
5 6 7 8 9 10 11 12 13 14 15 16 17 18 18 20 21
Figure 10
Venturi Dp vs Opacity Relationship
Particulate Removal
Figure 11 shows how opacity has related to particulate emissions for the various tests conducted
on the wet esp module. The line indicates the expected opacity to particulate emissions
relationship. As shown the results were somewhat erratic, but were generally within the expected
range. Note that some of the testing was conducted while operating with heavily loaded collector
tubes at low venturi dp, representing a wide range of operation.
11
-------
Vt,
SRI
Pace
1 0.02 0.03 O.M 0.05 0.(
Participate Emissions (Ibs/Mbtu)
Figure 11
Wet Esp Paniculate Emissions vs Outlet Duct Opacity
Table 1 estimates the particulate removal efficiencies along with the associated opacities while
operating with a clean wet esp for each of the venturi dp settings. Note that even at the low
venturi dp settings, the venturi removed the vast majority of the incoming ash.
Table 1
Venturi Dp/Particulate Removal Efficiency Estimates for Clean Wet Esp Operation
"Clean Wet Wet Esp
Single Single
Operational Mode
High Venturi DP
Intermediate Venturi Dp
Low Venturi Dp
Venturi
Dp Range
(inv*:)
13-21
8-13
1-4
Venturi
Particulate
Removal
Efficiency
99
98
97
Esp
Particulate
Removal
Efficiency
90
90
90
Module
Particulate
Removal
Efficiency
99.9
99.8
99.7
Venturi
Emission
Rate
(Ibs/Mbtu)
0.07
0.14
021
Wet Esp
Emission
Rate
(Ibs/Mbtu)
0.007
0.014
0.021
Module
Venturi
Emission
Rate
(Ibs/Hr)
50
100
150
Module
Wet Esp
Emission
Rate
(Ibs/Hr)
5
10
15
Wet Esp
Outlet
Opacity
3%
6%
10%
* Wet esp partoJate removal efficiency ffops off as bulcLp accLfnUates onlte cdlector tubes.
Air Toxics Removal
In an effort to characterize the impact of the wet ESP design on emissions of various air toxic
substances, Southern Research Institute (SRI) was contracted to quantify these emissions. In
order to isolate the effects of the wet ESP module from Sherco's other modules, the testing was
conducted downstream of the wet esp module as well as downstream of another Unit 2 scrubber
module in the far south row (210). Testing was also conducted at the inlet to the wet esp module
hi order to characterize the flue gas entering the scrubber modules so that the efficiency of both
designs could be estimated. The results of the testing are summarized in table 2.
12
-------
As shown, the wet ESP module had additional removal capability beyond what the existing
scrubber modules were capable offer all the air toxic substances listed except for elemental and
divalent mercury. The additional removal capability is attributed to the tendency of many air toxic
substances to concentrate on the submicron particles that the wet esp is designed to remove.
Neither module was capable of removing elemental mercury. Both modules removed about the
same amount of divalent mercury. Note that elemental mercury was by far the most prevalent
form of mercury.
Table 2
Wet Esp Module Air Toxics Emissions Test Results
Substance
As
Ba
Be
Cd
Ca
Cr
Co
Cu
Pb
Mn
Mo
Ni
Se
Ti
V
Elemental Hg
Divalent Hg
Inlet
Concentration
(Wj/Nm3)
85.8
10000
4Z1
0
188000
228
129
800
162
2440
113
193
0
31400
1030
7.05
0.57
Wet ESP Module Outlet
Concentration Removal
((ig/Nm3) Efficiency
0 100.0%
296 97.0%
0.152 99.6%
0
4150 97.8%
1.63 99.3%
0 100.0%
9.73 98.8%
3.61 97.8%
10.94 99.6%
6.78 94.0%
0 100.0%
0
81.8 99.7%
9.86 99.0%
8.12 0.0%
0.13 77.2%
Standard Module Outlet
Concentration Removal
(jig/Mm3) Efficiency
5.37 93.7%
646 93.5%
0.482 98.9%
0
8950 95.2%
4.98 97.8%
0.814 99.4%
18.6 97.7%
9.25 94.3%
21.6 99.1%
9.08 920%
214 98.9%
21.8 0.0%
258 99.2%
34.2 96.7%
8.40 0.0%
0.11 80.7%
SO2 Emissions
Figure 12 shows how the wet esp scrubber module's SO2 emissions compare to the standard
design scrubber module average. Although the wet esp module was started in May 1995,
problems with its SO2 monitor resulted in invalid SO2 readings until Jury 1995. The numbers
shown are monthly averages.
As can be seen, operating the wet ESP module at high venturi dp (May to September 1995 and
November 1996 to April 1997), resulted in SO2 emissions which were slightly below the standard
design module average. While operating the wet ESP module at low venturi dp (September 1995
to March 1996), its SO2 emissions were higher than the standard design module average.
Operating the wet ESP module at intermediate venturi dp (March 1996 to November 1996)
resulted in SO2 emissions that were generally comparable to the standard design module average.
13
-------
S E
-Standard ModLriซ SO2 Emitdon* (I
-Wet E*p ModUe S02 EmlMlon* (Ib/Mblu)
Figure 12
Wet Esp Module SC>2 Emissions
Higher SC>2 emissions while running at low venturi dp is most likely due to a reduction of mixing
of the flue gas with the slurry within the venturi Another likely factor is that the low venturi dp
will not shear the slurry droplets to as small as they would be sheared when operating at higher
venturi dp. The flue gas flow in the latter high venturi dp operating period was intentionally higher
than the flow during the earlier period. Consequently, the SC>2 removal capability is not as
efficient as would be expected when operating at a reduced liquid-to-gas ratio.
Future Testing
The main priority of the test program at this time is to continue to establish the interval of time
between cleanings for the current operation; high venturi dp and dry mode operation. As of April
1997, the lower field wet esp was still averaging higher than 25 kV. Thus, there has been very
little degradation in the lower field since it was started up in this mode in October 1996. The
current plans are to operate the module in this mode until the lower field dips below 25 kV,
allowing an approximation of the time between cleanings interval The assumption is that when
this occurs, the field will be approximately one third of the way through its time between
cleanings interval Since this mode of operation is the simplest possible design, long intervals
between cleanings will make NSP less inclined to further alter the design. While NSP has decided
to retrofit wet esp's within 15 additional Sherco scrubber modules, there is still some further
enhancements that may be feasible. Further testing is required to determine the feasibility of these
enhancements. The following sections describe the proposed testing hi detail
Intermittent Lower Field Cleaning With Upper Fog Spray Nozzle Headers
Tests conducted while operating an entire level of sprays showed it was not possible to
simultaneously achieve full spray coverage and still have sufficient voltage on the fields located
below the spray level However, tests have shown that operation of only a single upper fog spray
nozzle header, having 17 nozzles, can provide full spray coverage while attaining sufficient
voltage on the lower field. Consequently, if individual control valves are installed on each of the
nine upper fog spray nozzle headers, it would be possible to operate these sprays one header at a
tune while the module is in service. The resulting spray flow would not upset the reaction tank
water balance, making it unnecessary to recirculate the wet esp wash water.
14
-------
After the time between cleanings interval of the high venturi dp/dry mode operation is established,
this will be the next concept to be tested. The plans are to clean a portion of the lower field when
the module is taken down to install the nozzle header valves. This will: allow a baseline area to be
monitored. Success would be gauged by the amount buildup observed on the cleaned section of
tubes after the module is returned to service in this mode. The initial testing of this concept will be
at high venturi dp. It is possible that this concept may also make low venturi dp operation feasible.
Scale Inhibitor Testing
To date all of the testing has been conducted using the same scale inhibitor chemical added at the
same concentration to the wet esp wash water. The current use rate of this chemical costs about
$2500 per module per month. When all the aforementioned testing has been completed, this will
be the next area to be studied. The initial phase of this test will be to completely remove the scale
inhibitor from the module. The direction of further testing will depend upon the results. Other
options include optimizing the concentration of the current chemical and testing alternative scale
inhibitor chemicals. This testing is likely to be long term and will involve other scrubber modules
as they are retrofitted with wet esp's.
Acknowledgments
The authors would like to acknowledge the efforts of all those that helped make this project a
success including Dr. Virgil Marple & Dr. Dale Lundgren of Wet Inc. for then- efforts on the
slipstream test, all past and present employees of Southern Environmental for their efforts at
bringing the full scale wet esp module to fruition, all the engineers at Black & Veatch who
worked on various aspects of the project over the years and all the members of the NSP design
team. The authors would also like to acknowledge those who initiated earlier wet esp efforts in
this application; specifically, Kal Maki of Combustion Engineering, who originally proposed the
wet esp scrubber module concept for Sherco back in 1977 and Evan Bakke of Peabody, who
proposed a similar concept at Minnesota Power's Clay Boswell Station at approximately the same
time.
I 07/10071 11 PW
15
-------
TESTING OF A COMBINED DRY AND WET ELECTROSTATIC
PRECIPITATOR FOR CONTROL OF FINE PARTICLE EMISSIONS FROM A
COAL-FIRED BOILER
Larry S. Monroe
Kenneth M. Gushing
Southern Research Institute
P. O. Box 55305
Birmingham, AL 35255-5305
Wallis A. Harrison
Southern Company Services, Inc.
P. 0. Box 2625
14N8195
Birmingham, AL 35202
Ralph Altaian
Electric Power Research Institute
513 Franklin Building
Chattanooga, TN 37411
Abstract
A marginal dry electrostatic precipitator could be augmented by substituting a wet ESP
section in the last field of this dry ESP. Since the wet ESP section would eliminate
rapping reentrainment, would improve the capture of fine particles, and would remove
most of any sulfiiric acid aerosol present, this combination could allow enhanced control
of fine particle emissions without requiring additional space. A pilot-scale wet ESP was
testing in this configuration at the Southern Company Services and Southern Research
Institute Combustion Research Facility, supported by EPRI, TV A, and Southern Compnay
Services. The effectiveness of this wet ESP polishing unit in removing fly ash particles,
acid aerosol, mercury, and trace metals is presented. Issues in scaling this technology to
full scale use are also discussed.
-------
Introduction
There are two types of electrostatic precipitators (ESPs) used in the electric utility
industry: hot-side ESPs, which operate at 600+ฐF (315+ฐC), and cold-side ESPs, which
operate around 300ฐF (150ฐC). These ESPs use grounded plates to collect the fly ash,
where the ash is held to the plate by the electrostatic forces imposed in the ESP.
Periodically, the plates are rapped using mechanical acceleration to remove the collected
layer of ash. There are two problems that hinder the performance of these ESPs. First,
when a plate is rapped, some of the ash is released into the flue gas stream and escapes.
The final field in an ESP has nothing downstream to collect these rapping losses, and they
are emitted out of the stack as paniculate emissions. Second, the flow of flue gas across
the collected layer of ash on the collection plate may erode some of the collected
paniculate and result in particle escape from the collection zone of that field. In addition
to these two effects, a dry ESP relies on electric field charging and interparticle cohesive
forces to hold the collected ash onto the plate. The collected layer conducts the current
through it to the collection plate, and acts as a current-limiting resistor (particularly if the
electrical resistivity of the dust is greater than about 5 x 1010 ohm-cm). By limiting the
collection current in the ESP, this collected ash layer limits the amount of power that can
be applied to the collection process, and therefore limits the collection of fine fly ash in the
ESP.
In other industries, particularly ore smelters and plastics manufacturers, wet electrostatic
precipitators (WESP) are often utilized. In a WESP, water flows down the collection
plate, either overflowing a weir or being electrostatically collected from a spray. This
water serves to sweep the collected particles off the plate, ensuring that they are not
reentrained into the flue gas or that the dust layer does not limit the current applied to the
ESP. Normally, WESPs are operated near the flue gas saturation point.
In principle, a wet ESP field, discharging saturated flue gas, could be installed in the outlet
field of a typical dry ESP casing at a coal-fired electric utility plant. However, there are
two difficulties in this approach: plume buoyancy, where the density of saturated flue gas
would cause the plume to sink from the stack outlet, and accelerated stack corrosion, due
to the condensation of water from the saturated gas. These problems can be solved
through either reheating the saturated flue gas or converting the plant to wet stack
operation, both rather costly choices. However, it might be possible to operate a wet ESP
under conditions where the outlet flue gas is above the saturation point, thus avoiding the
problems mentioned above. By limiting the contact between the flue gas and the water, it
is likely that the outlet flue gas may be kept 50-70ฐF (28-39ฐC) above the saturation point.
Other operational problems may be caused by not saturating the flue gas in the WESP,
such as dry spots on the plates collecting dry ash deposits or uneven flue gas temperature
distributions (and therefore uneven velocities) in the collection zones. Representatives of
the Electric Power Research Institute (EPRI) and Southern Company Services proposed
that a WESP, installed as a simulation of the last field of a dry ESP, be tested in the pilot-
scale combustion facility located at Southern Research Institute. A major goal of this
program was to demonstrate successful operation of a WESP at temperatures well above
-------
the saturation point, and then to document the fine particulate collection and reduction of
sulfur oxide concentrations during these operating conditions.
WESP Design
For this test, instead of designing the wet ESP section to be housed in the final field of a
dry ESP, the WESP was designed as a stand-alone unit to be installed downstream of an
existing dry ESP. This allowed greater flexibility in the test program. However, the
WESP was designed to a size appropriate for installation inside the final field of a dry
ESP. By controlling the power input to the dry ESP, the inlet loading to the WESP unit
could be maintained at different levels. By separating the two units, it was possible to
sample the flue gas between the two types of ESPs for particulate mass concentration
(mass loading), particle size, and mercury concentration and speciation. Additionally, the
space between the two units allowed the introduction of activated carbon at the inlet of
the WESP fields to evaluate the use of this additive in mercury capture.
Southern Environmental, Inc., of Pensacola, Florida, designed and constructed the wet
ESP unit, which was configured as a wet-wall adaptation of a conventional dry wall ESP.
This stand-alone unit had a single flow channel, with a set of discharge electrodes located
on the centerline and the collection plates along each side. The WESP was divided into
two fields with an overall length of about 9 feet (2.74 m), with the front field twice the
length of the second field. The first field was 5 ft (1.5 m) high by 5 ft 5 in. (1.65 m) long,
while the second field was 5 ft (1.5 m) high and 2 ft 8 in. (0.8 m) long, for a total plate
collection area of 80 ft2 (7.43 m2). The average flue gas flow rate at the WESP inlet for
this test program was 2,300 acfm (1.08 m7s). The average flue gas velocity ranged from
6.5 ft/s (2.0 m/s) to 8 ft/s (2.4 m/s), while the specific collecting area (SCA) ranged from
33.7 (1.84) to 37.2 (2.03) ft2/1000 acfm (x 10"3 m2/m3/s) during the tests. The spacing
between the two collection plates was 12 inches (30.5 cm), with 6 inch (15.3 cm) spacing
between the discharge electrode and each of the collection plates. The collection plates
were constructed of flat mild steel. The discharge electrodes were a barbed-type design.
The inlet field had four vertical discharge electrodes, spaced 16 inches (40.6 cm) apart,
and the exit field had two discharge electrodes, also spaced 16 inches (40.6 cm) apart.
The water was sprayed into the WESP through a series of nozzles. The first set of five
nozzles was located in the inlet transition to the WESP. Also, two pairs of nozzles (a
high- and low-flow set) were located on the centerline above the discharge electrodes,
with the spray directed down at a 45ฐ angle from the horizontal. By choosing either the
low-flow or high-flow set of nozzles, good fluid atomization could be achieved at any
flow. The nozzles for the second field were similar, with a set of three nozzles located on
a pipe header placed vertically on the centerline between the two fields and directed
horizontally downstream. There were also two pairs of nozzles located above the
discharge electrodes of the second field, just as in the inlet field. Each of the four sets of
nozzles was equipped with a flow-control valve and a rotameter-type flow indicator.
-------
All of the nozzles were the FullJet style manufactured by Spraying Systems Company.
These nozzles provide a wide-angle, full-cone spray and produce a relatively coarse
droplet size The water supply pressure to the nozzles was typically in the range of 10 to
15 psig (69 to 103 kPa). At these supply pressures the nozzles produced droplets ranging
in diameter (median volume diameter) from 1,000 to 1,800 micrometers.
Test Matrix
The major portion of the WESP testing consisted of a set of conditions, where the inlet
paniculate mass concentration and the amount of water supplied to the WESP were
varied. Generally, the performance goal for this test program was to simulate the
conditions at the outlet of the second field of a typical three-field ESP with a second-field
outlet mass loading of 0.7 Vo/lQ6 Btu (303 ng/J). The desired WESP outlet mass emission
rate was 0.03 lb/10s Btu (13 ng/J) which would meet new source particulate emission
limits. Table 1 summarizes the major goals on this test program, namely, to evaluate the
ability of the WESP to capture particulate, acid gases, and trace metals. As shown in
Table 1, there were initially six test conditions to study particulate capture, followed by
five other tests to evaluate acid gas and trace metals capture. The inlet conditions to the
WESP were stable, with the inlet temperature generally around 285 to 290ฐF (140 to
143ฐC). The flue gas flow rates, oxygen levels, and water content were also very stable
during the test program.
The coal chosen for the testing was a local Alabama coal, Shoal Creek, mined by
Drummond Coal Company. This coal is a compliance coal for sulfur, high in heating
content but low in volatile content and classified as a medium volatile bituminous coal. A
typical analysis for Shoal Creek coal would be: 6.9% moisture, 8.6% ash, 60.2% fixed
carbon, 24.3% volatiles, and 0.72% sulfur. The natural mercury content of this coal
averaged about 20 ppb (by weight). To make the ideal coal for these tests, the level of
mercury was increased fivefold to around 100 ppb.1
Discussion of Results
Particulate Matter Capture
The WESP was effective at collecting fly ash. Tables 2, 3, and 4 summarize the test data
for each measurement method. A comparison of the emissions as measured by Method
29, Method 17, and the cascade impactors at the WESP outlet is presented in Figure 1. In
this plot, the results from the cascade impactors and Method 17 were of the same
magnitude (except for Tests 11 and 12), while Method 29 emissions are significantly
higher. It is not unusual to find that a cascade impactor, with a much smaller sampling
volume, produces, in many cases, lower total mass concentrations than a measurement
method designed to accurately determine mass (Method 17). During testing, the cascade
impactors and the Method 17 in-duct filter holders were inside the duct and were heat-
traced with their temperatures controlled to 300ฐF (149ฐC). On the other hand, the
-------
Method 29 test protocol used a heated probe connected to an external filter contained in
an oven, with both the probe and filter controlled to 250ฐF (121 ฐC).
Of the three methods used to measure the WESP collection efficiencies, two agreed,
namely the EPA Method 17 results and the cascade impactor results. However, the EPA
Method 29 results show consistently lower efficiencies than the other two methods.
Looking at the particle loadings, the Method 29 measurements of the outlet of the WESP
are 2 to 3 times higher than the Method 17 or cascade impactors. Often, the mass
balances on these systems are checked by also looking at elements in the fly ash that are
expected to be distributed equally in all sizes of particles. These elements, commonly
referred to as major metals, are aluminum, calcium, iron, magnesium, and titanium.
Therefore, the capture efficiency of each of these metals can be used as an internal
consistency check with the overall measurements of total mass-based efficiency. The
major metals analysis will not be sensitive to any species that condenses in the sampling
train, and therefore is a good method to obtain the collection efficiencies for the solid
particles only.
In this testing effort, the capture efficiencies of the major metals for the Method 29
measurements agree with the capture efficiencies of the cascade impactors and the Method
17 results, as can be seen in Figure 1. The check on Method 17 collection efficiencies by
the major metals is also shown in Figure 1. For two of the three Method 17
measurements, there is very little difference in the total mass collection efficiency and that
calculated from the average of the major metals collection efficiencies. The third Method
17 major metals collection efficiency (Test 6) is lower than the total, but if iron is excluded
from the average, then the major metals agree with the total mass collections efficiency
(98.8% for major metals without iron versus 99.1% for total mass). Therefore, if iron is
excluded from the Test 6 average for Method 17, the collection efficiencies as measured
by cascade impactors, Method 17, Method 29 major metals, and Method 17 major metals
all agree. Therefore, the agreement of these methods for collection efficiencies gives
confidence in the measurements of the collection of the solid phase particles through the
WESP.
Since the outlet loadings of the Method 29 measurements are 2 to 3 times higher than
those measured at the same time by the cascade impactors, it appears that some species
may be condensing in the Method 29 WESP outlet measurement train that is not
condensing in the higher temperature filters of the impactors or Method 17 trains. Since
both the Method 17 and the impactors are heated to 300ฐF versus 250ฐF for the Method
29, there is the possibility that the amount of water in the sulfuric acid droplets may be
much higher in the Method 29 measurement. In any case, more study would be needed to
identify the species that is condensing in the Method 29, but for the solid phase particles, it
is clear that the three methods agree in the measured collection efficiencies.
The correlation of WESP capture efficiencies with WESP outlet temperature, as presented
in Figure 2, shows that there is a general decrease in fly ash capture efficiency with
increasing temperature. The dashed line is the linear regression fit to all of the data. This
-------
effect is probably due to the simple relationship between gas temperature and precipitator
SCA. (As the temperature increases, the volume of flue gas increases and decreases the
SCA which decreases capture efficiency). Similarly, the increase in WESP capture
efficiency with higher values of inlet mass loading, Figure 3, is probably due to the average
inlet particle size increasing as the particle loading increases. Again, the dashed line in this
figure is the linear regression fit to the data. Since the dry ESP is collecting fly ash ahead
of the WESP, the increase in the WESP inlet loading is caused by allowing more ash to
escape the dry ESP. This increment in ash loading is most likely composed of larger ash
particles which will be collected with a high efficiency, thus increasing the overall (mass-
weighted) capture efficiency.
Sulfuric Acid Aerosol Capture
The capture of sulfuric acid aerosol is a principal advantage of a WESP unit. Indeed, the
observations in this work showed that the capture of SO3, as H2SC>4, by the WESP was
generally between 60 and 70 percent. Table 5 presents the SO2 and SOs test data. The
capture efficiency for each species is shown graphically in Figure 4. There appears to be
no sensitivity of the capture efficiency to temperature or inlet fly ash loading.
Interestingly, the addition of 5 times the natural SOs present (Test No. 8) also resulted in a
capture efficiency of 68%. This result implies that regardless of the original concentration
of SOs in the inlet flue gas, the resulting aerosols had similar particle size distributions and
were all collected within the WESP with efficiencies in the range of 65 to 70%.
The use of sodium hydroxide as an additive to the WESP feed water (Test No. 9) and the
addition of activated carbon to the flue gas (Test No. 10) did not significantly change the
collection efficiency of the sulfuric acid aerosol.
Capture efficiencies of SC>2 were typically between 10 and 25%. Higher capture
efficiencies were seen during Test Numbers 5, 7, 8 and 9. Tests 5 through 8 were for
conditions where the most water was used in the WESP sprays, therefore there is probably
a connection between water volume and capture of SO2. These tests also demonstrated
that capture of sulfur dioxide was not appreciably affected by the injection of sodium
hydroxide (Test No. 9).
Mercury Capture
The mercury capture results in the WESP testing were puzzling. Although the doping of
coal for the desired level of mercury was highly successful, it was expected that some of
the oxidized mercury would be captured, while the elemental mercury would pass through
the WESP This behavioral assumption was based purely on the sensitivity of collection of
each species to temperature.
-------
In general, there was little or no mercury found in the solid phase in the inlet or outlet.
The blank train analysis confirmed that there was no contamination in the operation of the
M29 trains. There appeared to be no detectable mercury in the reagents used.
The input values of mercury from the coal compared reasonably well with the total
mercury measured at the WESP inlet. The concentrations of oxidized and elemental
mercury emitted from the WESP were comparable. The total mercury capture of the
WESP was around 30%, while the oxidized mercury capture was 50 to 60%, and the
elemental mercury capture was -20 to -55%. A negative capture efficiency means that
there is more of that particular species leaving than entering the WESP. Figure 5 shows
that for every test, there was more elemental mercury leaving than entered the WESP.
Either oxidized mercury was converted to elemental mercury during its passage through
the WESP or this result was an artifact of the sampling method. Dr. Ed Dismukes
suggested the possibility that the oxidized mercury was captured in a liquid phase in the
WESP, where the liquid also captures SO2 and SOs. The sulfite ion can be oxidized to
sulfate by the reduction of Hg2* to the elemental form. The resulting elemental mercury
would then be volatile at WESP temperatures, and could escape. At room temperature,
this conversion is strongly favored thermodynamically.
Another possible explanation for the higher concentration of elemental mercury at the
WESP outlet compared to the WESP inlet is a known problem with the Method 29
sampling protocol Several studies have been performed to better understand the apparent
problems with Method 29.2>0'4'5 These studies have shown that the hydrogen peroxide that
is used in the first set of impingers following the filter will, in the presence of sulfur
dioxide, collect elemental mercury that should pass unaffected to the potassium
permanganate impinger. For a wet scrubber or spray dryer collecting a significant fraction
of the sulfur dioxide this would result in a higher than actual indicated collection efficiency
of oxidized mercury and would produce a higher apparent concentration of elemental
mercury measured at the outlet compared to the scrubber inlet value.
In contrast to a wet scrubber or spray dryer, the WESP was marginally efficient at
collecting S02 from the flue gases. Figure 4 shows that the SO2 collection efficiency
ranged from 10 to 25%. As compared to a wet scrubber or spray dryer with a 90%+
collection efficiency of S02, the WESP should have had only a modest affect on Method
29 results because of the relatively small difference in the concentrations of SO2 passing
through the inlet and outlet Method 29 trains. In this case, there should not have been the
grossly different inlet and outlet speciation results observed in these tests. The conclusion
then is that either another chemical mechanism is distorting the Method 29 measurements
or there is an actual change in mercury speciation across the WESP. Clearly, further work
on Method 29 chemistry and the behavior of mercury in this environment is needed to
allow this question to the settled definitively.
-------
The additives, NaOH and activated carbon, seemed to have no effect on the behavior of
mercury through the WESP. In Figure 5, the emissions for Test Nos. 9 and 10 are the
same as the previous test, within the apparent experimental scatter.
Trace Metals Capture
As observed in the DOE Air Toxics Testing Program, the control of fine particles seems
to determine the control of most trace metals. With the exceptions of mercury and
arsenic, other trace metals are expected to be in the solid phase. It was expected that the
WESP would be efficient in removing these trace metals in the solid phase, given its
superior collection efficiency. Indeed the testing in this program did confirm the ability of
the WESP to remove these trace metals. The capture efficiencies in the WESP for Cd, Pb,
and Sb were very high, exceeding 75 percent. Unfortunately, the analyses for arsenic and
selenium were not useable due to poor spike recovery, so no information was generated
on the effectiveness of the WESP in removing these elements.
Conclusions
The testing of a pilot-scale wet ESP unit installed downstream of a conventional cold-side
ESP was successful. After correcting initial problems with the WESP installation, namely
sulfuric acid condensation and high-voltage tracking on the insulators, high thermal losses,
and excessive sparking from the water flows off the bottom of the plates; the unit
performed very well through an extended testing period of over 250 hours. The WESP
was operated with the outlet flue gas at 10 to 50ฐF (5 to 28ฐC) above the saturation point
without noticeable problems for test periods of about 24 hours.
Overall collection efficiency of the WESP ranged from 90 to 99 % among the twelve
individual tests. WESP average inlet paniculate mass concentrations were slightly higher
overall than desired for simulating the outlet emissions from the second field of a dry ESP
(1.00 lb/106 Btu (433 ng/J) actual versus a 0.7 lb/106 Btu (303 ng/J) goal), while the
WESP average outlet mass emission rates were also slightly higher than desired (0.065
lb/106 Btu (28 ng/J) actual versus a 0.03 lb/106 Btu (13 ng/J) goal), consistent with the
higher inlet mass concentration. These outlet emissions reflect the raw numbers for the
Method 29 trains. If the major metals collection efficiencies are used to estimate the
WESP emissions, the outlet emissions would average 0.04 lb/106 Btu (18 ng/J).
Therefore, it appears that if the inlet loading to the WESP were controlled down to the
desired input of 0.7 lb/106 Btu (303 ng/J), then the outlet emissions of solid particles, as
determined by the major metals, would be below the goal of 0.03 lb/106 Btu (13 ng/J).
The average mass collection efficiency for the twelve tests, using the Method 17 results
and the Method 29 major metals analysis, was 95.0%.
The WESP was effective in collecting SO3 that condensed to sulfuric acid aerosol in the
device. During normal operations, the WESP removed 67% of the 3.5 ppm sulfuric acid
(measured @ 3% O2) contained in the flue gas entering the device. When SOs was
injected to the flue gas, the WESP still collected 68% of the sulfuric acid, despite a five-
-------
fold increase in the inlet concentration. Obviously, the sulfuric acid is nucleating and
growing to a size distribution which is collected with an overall 67% fractional efficiency.
The WESP was able to collect approximately 30% of the total mercury entering the unit.
The test data indicate an apparent collection efficiency of 50% for oxidized mercury
(probably HgCb) and -30% for elemental mercury. A negative collection efficiency would
mean that more elemental mercury is exiting the device than is measured at the inlet. It
has also been hypothesized that this could be due to either a reduction of dissolved Hg2*
by SC>2 oxidation to SOs in the WESP or simply an artifact of Method 29. More work is
needed to fully understand this result.
Trace metals, with the exception of arsenic and mercury, are typically in the solid form at
the conditions exiting the wet ESP. Therefore, a good collection efficiency of fly ash
particles would translate into a good collection of trace metals. Indeed, it was observed
that lead, antimony, and cadmium were collected at an efficiency of 75 to 95%.
Unfortunately, the analyses for arsenic and selenium were not useable due to poor spike
recovery, so no information was generated on the effectiveness of the WESP in removing
these elements.
Testing the WESP unit at reduced power showed that the performance of this unit was
not substantially changed as the current density was decreased from between 75-120
uA/ft2 (0.81 to 1.29 mA/m2) to 38-45 uA/ft2 (0.41 to 0.48 mA/m2). Therefore, it is
expected that a larger scale installation would give similar performance to that measured in
this testing.
A final goal of the WESP testing program was to provide data to evaluate the possibility
of using a wet ESP polishing field as a retrofit on the outlet stage of a marginal dry
precipitator. The collection efficiency for this configuration was estimated by assuming
that the inlet particle loading had not changed from the previous test and by comparing
that value to the Method 17 measurements made at the WESP outlet. The collection
efficiency of the WESP was estimated to be 93.5%, very similar to the other test results.
Acknowledgements
The testing of the WESP was supported by the Electric Power Research Institute, the
Tennessee Valley Authority, and the U. S. Environmental Protection Agency. The
Southern Company, through the contribution of the usage fee for the facility and the time
of employees, also supported the WESP testing. The following individuals were directly
responsible for the successful testing of the wet ESP: Mr. David Salladay of TV A,
Messrs. Sam Rakes and Charlie Sedman of EPA, and Mr. Roy Clarkson of SCS.
The staff of the facility was responsible for the operation of the facility and the WESP:
Mr. Wim Marchant, supervisor, Mr. Sam O'Neal, Mr. Bill Page, Mr. Ken O'Neal, and Mr.
David Smith. Other employees of Southern Research were also involved in sampling and
analytical efforts in this project: Mr. Marvin Steele, Mr. Dave W. Smith, and Ms.
-------
Wynema Kimbrough. Messrs. John Carter and Brian Reid of SeaTec, Inc., were also
responsible for operation and sampling efforts during the testing.
TTL, Inc , of Tuscaloosa, AL, was used to analyze the Method 29 samples taken during
the testing for trace metals. Dr. Edward Dismukes, now retired from Southern Research,
was very helpful in designing the doping procedure for adding mercury to coal and in
understanding the mercury capture effects of the WESP. Discussions with Mr. John Lytle
of TV A, Dr. John Gooch, and Dr. Grady Nichols of Southern Research were invaluable in
performing the work successfully.
References
1. Finkleman, Robert B. "Mode of Occurrance of Toxic Elements in Coal: Level of
Confidence" In Proceedings' Trace Element Transformations in Coal-Fired
Power Systems, Electric Power Research Institute, April 19-22, 1993, Scottsdale,
AZ
2. Bush, P. V., E. B. Dismukes, and W. K. Fowler. "Characterizing Mercury
Emissions from a Coal-Fired Power Plant Utilizing a Venturi Wet FGD System."
In: Proceedings of the Eleventh Annual Coal Preparation. Utilization, and
Environmental Control Contractors Conference. DOE/PETC, Pittsburgh, PA, July
12-14, 1995, pp 105-12.
3. Bush, P. V., E. B. Dismukes, J D. McCain, and W. K. Fowler. "Sampling and
Analytical Challenges for Air Toxics Assessments." In: Proceedings of the
EPRI/DOE International Conference on Managing Hazardous and Particulate Air
Pollutants: Book 2. EPRI, Toronto, Ontario, Canada, August 16, 1995.
4. Laudel, D. L., M. K. Heidt, T. D. Brown, B. R. Nott, and E. P. Prestbo
"Mercury Speciation: A Comparison between EPA Method 29 and Other
Sampling Methods." In: Proceedings of the 89th Annual Meeting of the Air and
Waste Management Association: Paper 96-WA64A.04. AWMA, Nashville, TN,
June 23-28, 1996.
5. Khosah, R. P., T. J. McManus, A. J. Bochan, J. L. Clements, and R. O. Agbede.
"A Comparative Evaluation of EPA Method 29 and the Ontario-Hydro Method
for Capture and Analysis of Mercury Species by ATS and Three Other Research
Laboratories." In: Proceedings of the 89th Annual Meeting of the Air and Waste
Management Association: Paper 96-WA64A.03. AWMA, Nashville, TN, July
23-28, 1996.
-------
Table 1. Schedule and Description of WESP Tests.
Test#
1 - ' >
1
2
3
4
5
6
7
8
9
10
11
12
Date (1995)
, ' / ',:-, , ,,-, - ',
Julv18&19
July 20
Julv21
Julv22
Julv 23
Julv 23
Julv 24
Julv 25
Julv 26
July 27
Julv 27 & 28
Julv 28
Time
day
day
day
dav
day
night
day
dav
day
day
night
evenina
Particulate
Concentration
' >','-, i ' ^ -; ~ '
med
med
med
low
low
hiah
med
med
med
med
med
very hiah
Approach
Temperature
' \ -i "\ -'-',!
med
med
hiah
high
low
low
low
low
med
med
med
Test Description
" " ^ ^ "" " " ' ^ ^ "" "*O ^ " * *" "x " ^ . V " "^ % \ ^ ^ ^ O 5 * l" " * v v
WESP Startup Tests
Goals: 0.4 lbs/106 Btu (173 na/J) and 50ฐF f28ฐC) approach
Goals: 0.4 lbs/106 Btu /173 na/J) and 75ฐF (42ฐC) approach
Goals: 0.1 lbs/106 Btu (43 na/J) and 75ฐF (42ฐC) approach
Goals: 0.4 lbs/106 Btu (173 na/J) and 15ฐF (8ฐC) approach
Goals: 0.6 lbs/106 Btu (260 na/J) and 15ฐF (8ฐC) approach
Goals: 0.4 lbs/106 Btu (173 na/J) and 15ฐF (8ฐC) approach
Space charae test: SOi & NH^ injection, S0\ effect on Ha
NaOH injection test for Ha and SOi capture
Activated carbon injection for Ha and SOi capture
Low Power Test: Reduce current to 50 pA/ft (0.54 mA/m2)
Sinale Vessel Simulation: First field drv & wet second field
TABLE 2. WESP Average Particulate Mass Concentrations and Collection
Test#
'-? ' '"
1
2
3
4
5
6
7
8
9
10
11
12
Date (1995)
ฐ ",-"S ' ~! ,
July 18 & 19
July 20
July 21
Julv 22
Julv 23
Julv 23
Julv 24
July 25
Julv 26
Julv 27
Julv 27 & 28
Julv 28
Jest Description
WESP Startup
Matrix Test A
Matrix Test B
Matrix Test C
Matrix Test D
Matrix Test E
Matrix Test F
SO, & NK-, Injection
NaOH Injection
Activated Carbon Injection
Low Power Case
COHPAC II Simulation
Type Trains
In - Out
j:,4*^ ;:;*
M17-M17
M29 - M29
M29 - M29
M29 - M29
M29 - M29
M17-M17
M29 - M29
M29 - None
M29 - M29
M29 - M29
M17-M17
Est. -M17
Average Loading, lbs/106 Btu (na/J)
Inlet
-"y--^"- v v-v,
0.79(341)
1.09(471)
0.79 (343)
0.28(122)
0.47 (203)
1.62(700)
1.24(538)
1.39(601)
0.81 (351)
1.30(564)
1.27(552)
1.27(552)**
Outlet
- ;;>--VT'"-' -:\ ;' V
0.022 (9.7)
0.080 (34.7)
0.077 (33.5)
0.029(12.4)
0.033(14.4)
0.015(6.4)
0.11 (46.9)
0.085 (36.9)
0.12(54.1)
0.056(24.1)
0.082 (35.7)
Collection
Efficiency, %
\*;~ " ~v; ;" X
97
93
90
90
93
99
91
89
90
96
94
*: The inlet loading for Test #12 is assumed to be the same as for Test #11.
-------
TABLE 3. Average WESP Major Metal Collection Efficiencies.
Test#
1
2
3
4
5
6
7
9
10
11
Type Train
M17
M29
M29
M29
M29
M17
M29
M29
M29
M17
Collection Efficiencies
Aluminum,
%
98.0
96.4
96.0
88.6
93.6
99.6
97.2
96.8
97.3
98.1
Calcium,
%
93.2
93.8
92.0
99.5
92.4
96.4
89.8
95.3
95.9
100.0
Iron,
%
95.9
95.5
95.1
83.0
94.1
84.2
96.3
95.6
96.0
88.5
Magnesium,
%
100.0
96.1
94.2
89.3
94.8
100.0
97.0
96.1
96.6
100.0
Titanium,
%
96.7
95.9
95.8
92.9
95.4
99.1
97.5
96.2
96.7
97.5
Average MM
Efficiency, %
; ^ %"^" w" s \ " "" " "
96.8
95.6
94.6
90.7
94.1
95.9
95.6
96.0
96.5
96.8
Italicized numbers represent efficiencies calculated by eliminating data from one or two of the three individual runs.
TABLE 4. WESP Cascade Impactor Test Results.
Test#
- \ %\ vs !" "-$ซ
2
3
4
5
7
9
11
Inlet Impactors
Concentration,
gr/dscf (g/dsm3)
"~ " ^^ , \ v ^ - ^ v ^
0.14(0.32)
0.17(0.39)
0.055(0,13)
0.050(0.11)
0.17(0.39)
0.13(0.29)
0.16(0.36)
Emission Rate,
lb/106Btu(ng/J)
0.49(214)
0.63(271)
0.21 (90)
0.19(83)
0.68 (294)
0.47 (203)
0.62 (268)
Outlet Impactors
Concentration,
gr/dscf (mg/dsm3)
,\ - !.' ' /; - > *
0.0063(14.4)
0.0065(14.9)
0.0029 (6.6)
0.0021 (4.7)
0.0030 (6.9)
0.0030 (6.9)
0.0049(11.3)
Emission Rate,
lb/106Btu(ng/J)
:,V;, .. ;.-.!
0.027(11.6)
0.028 (12.2)
0.012(5.35)
0.0094 (4.05)
0.014(5.99)
0.012(5.16)
0.023 (9.77)
Collection
Efficiency, %
*"- ',ป* "
94.6
95.5
94.1
95.1
98.0
97.5
96.4
-------
TABLE 5. WESP SO2 and SO3 Concentrations and Collection Efficiencies.
Test#
, '" ff ,
1
2
3
4
5
6
7
8
9
10
11
12
Test Description
WESP Startup
Matrix Test A
Matrix Test B
Matrix Test C
Matrix Test D
Matrix Test E
Matrix Test F
S03 & NH3 Injection
NaOH Injection
Activated Carbon
Low Power Case
Dry Field-Wet Field
WESP SO,, Concentrations @ 3% 02 dry
Inlet S03, ppm
12.4
7.2
9.1
8.3
9.2
6.9
59.3
10.3
10.6
10.6
10.6
Inlet S02, ppm
645
511
530
510
510
504
513
502
500
500
500
Outlet S03, ppm
3.3
3.1
3.7
3.5
3.5
2.4
19.2
3.5
3.3
3.5
3.4
Outlet S02, ppm
V ' ' ; s ;NJT,S!"''
I 550
415
466
451
389
385
382
380
415
425
416
WESP Efficiency, %
S03
, -x ^t"^':
73
57
60
58
61
65
68
66
69
67
68
S02
: X^'kr;?';-^:^
* ซ&stckงซ%SJ^-
15
19
12
12
24
24
25
24
17
15
17
-------
100
a
"a
E
LU
O
O
5
a
3
-------
ss
5"
o
0 Q^-
ฃ 95
UJ
C
ts
"5
u
m 90
3
e
ซ
(
0.
*
5
""
J - " *
^ ' " ^
c
200 400
. ! ji . iii i i
(
JS "^
0 '
600
0 0.4 0.8 1.2 1.
WESP Inlet Loading
1
J
800
6 2.
KI17-MM
O M17 Total
^ M2S^aS
0
Figure 3. WESP Collection As a Function of Inlet Fly Ash Loading.
100
90
>; ao
U
I 70
UJ 60
O
O 40
U
CO
O 30
O 20
CM
O
U) 10
<
c
5
<
L
I
/
) C
1
E
>
' C
1 E
(
>
I
1
)
I
<
I
> fl
> <
3 C
1
[
^
' C
3 [
) <
-
-
3 1
O
D SO2
56789
Test Number
10 11 12
Figure 4. WESP SO2 and SO3 Collection Efficiency.
-------
>
O on
5
n .0
LU ,n
c 10
O
u
O
o 10
ฐ 20
=
O
fin
r
<
I
2
1
[
>
(
L.
1 C
1 <
y
[
J
}
\
A
I
>
t
^
I
*
J
4
2
^
i
i
.
Totai
A Elemental
0123456789 10 11 12
Test Number
Figure 5. WESP Collection Efficiency for Mercury Species.
-------
ELECTROSTATICALLY
ENHANCED CORE SEPARATOR SYSTEM
Bruce H. Easom
Leo A. Smolensky
S. Ronald Wysk
LSR Technologies, Inc.
989 Main Street
Acton, MA 01450
Ralph F. Altaian
Electric Power Research Institute
516 Franklin Building
Chattanooga, TN 37411
Wallis A. Harrison
Robert R. Hardman
Southern Company Services
P.O. Box 2625
Birmingham, AL 35202
Abstract
The combustion of coal in Fossil Energy Power Systems releases trace amounts of
chemical elements identified by the Clean Air Act Amendments of 1990 as hazardous air
pollutants (HAPs). LSR Technologies, Inc. is working to reduce the emissions of these
HAPs by developing a novel, high efficiency particle collection system known as the
Electrostatically Enhanced Core Separator (EECS) system. The concept involves placing a
high efficiency particle separator downstream of a conventional ESP that strips the particles
from the incoming flow and returns them, along with a small amount of bleed flow, back to
the inlet of the ESP. The main component of the system is the separator, called the EECS.
The design is based on the mechanical Core Separator that was developed by LSR as a high
efficiency centrifugal separator. Enhancing the Core Separator by adding an electrical field
improves the separation efficiency of particles in the sub-micron range which is the range
where centrifugal separation becomes ineffective. In the combined system, the centrifugal
forces operating on the particles are augmented by electrostatic forces so the EECS has high
separation efficiency for particles of all sizes. Field tests have shown than an EECS
operating downstream of an underperforming ESP can reduce the paniculate emission rate
to below 12.9 ng/J (0.03 Ib^million Btu) even for ESPs with emission rates as high as 260
12.9 ng/J (0.6 Ibjmilh'on Btu). The projected capital cost of a 250 MWe EECS retrofit
system is about $25/kW.
Introduction
The Clean Air Act Amendments of 1990 have identified 189 chemical elements and
compounds as hazardous air pollutants (HAPs). The U.S. Environmental Protection
-------
Agency (EPA) has been given the task of evaluating the health risks of these so-called air
toxics and determining their acceptable emission rates. The U.S. Department of Energy
(DOE), in cooperation with private industry, is sponsoring research into developing new
air toxic emission control technologies.
Coal typically contains trace amounts of HAPs, the species and amounts vary with coal
type and source. Combustion of coal in Fossil Energy Power Systems releases these
HAPs into the combustion gas where they flow to the stack by passing through the gas
cleanup system. Except for mercury and selenium, the HAPs exist primarily in the solid
phase temperatures typical of flue gases so they can be removed by paniculate cleanup
systems. In fact, these solid phase HAPs are removed with approximately the same
efficiency as the other paniculate matter. Table 1 shows the removal efficiency of 15 HAPs
by an electrostatic precipitator having an efficiency of approximately 99 percent.
These results suggest that very high removal efficiencies of paniculate HAPs can be
achieved by using a very high efficiency paniculate collector. Epidemiological studies'
suggest that there may be health benefits from reducing concentrations of paniculate matter
in the ambient air. This paper will discuss the development of a very high efficiency
paniculate collector called the Electrostatically Enhanced Core Separator (EECS) system
that may have the potential to significantly reduce the emission rates of both particulates and
paniculate air toxics from Fossil Energy Power Systems.
Table 1
Hazardous Air Pollutant Removal Efficiency of an Electrostatic Precipitator2
Removal
Element Symbol Efficiency
[percent]
Removal
Element Symbol Efficiency
[percent]
Antimony
Arsenic
Barium
Beryllium
Cadmium
Chromium
Cobalt
Sb
As
Ba
Be
Cd
Cr
Co
81.00
99.10
99.80
97.40
99.20
99.20
99.30
Copper
Manganese
Mercury
Molybdenum
Nickel
Phosphorus
Vanadium
Cu
Mn
Hg
Mo
Ni
P
V
99.60
99.60
<20.00
96.00
98.20
98.00
99.50
Approach
The Electrostatically Enhanced Core Separator (EECS) system is derived from LSR's Core
Separator technology that was developed as a mechanical particle collector. The mechanical
system consists of a cyclone, recirculation fan and a particle separator called the Core
Separator. The Core Separator is a cylindrical centrifugal separator. The gas with
entrained panicles enters the cylinder through a tangential slot. The tangential inlet induces
a spinning motion to the gas that causes the entrained panicles to migrate toward the outer
wall. The clean gas is removed axially from the center of the cylinder and the particles,
along with some bleed flow, are extracted through a second tangential slot. The bleed flow
with the entrained particles is directed to the cyclone then through the recirculation fan back
to the Core Separator inlet In this configuration, the system collection efficiency is a
function of both the Core Separator efficiency and the cyclone efficiency as shown in
-------
Equation 1. The important feature of Equation 1 is that the system efficiency is primarily
governed by the efficiency of the Core Separator. For example, if the Core Separator
efficiency is 99.9 percent and the cyclone efficiency is only 30 percent, the system
efficiency is still 99.67 percent.
Where
TJ^j Core Separator system collection efficiency
77cs Core Separator separation efficiency
I] cyc cyclone collection efficiency
What is happening physically is that, while the process flow passes through the separator
once and leaves the system, the particles keep recirculating through the cyclone until they
are eventually collected. To particles, this system appears to a large number of cyclones in
series. That is why the system efficiency can be significantly higher than the efficiency of
the cyclone alone.
Due to the nature of turbulent two-phase flows, it is much easier to design a high efficiency
separator than a high efficiency collector. Cyclones suffer from trying to satisfy two
conflicting requirements. They need high inlet velocities to bring the entrained particles to
the walls and they need low velocities to ensure that the separated particles flow down the
walls into the hopper without being reentrained into the swirling flow. In the Core
Separator we do not collect the particles so we can use high velocities to achieve very high
separation efficiencies without having to worry about reentrainment.
We employ basically the same idea in the EECS system as in the Core Separator system
except that the components are configured in a collector-first orientation, a particle
precharger is added upstream of the separator, and electrostatic field is applied to the Core
Separator to improve its performance with sub-micron sized particles. The efficiency of the
collector-first orientation is given by Equation 2.
(2)
Where
77 EECS system collection efficiency
T]EECS EECS separation efficiency
77 Efficiency of particle collector
col
In the collector-first arrangement, if the EECS has an efficiency of 99.9 percent and the
collector has and efficiency of 30 percent then the system efficiency is still 99.77 percent
-------
Because the EECS system requires a separate collector, it is ideally suited for retrofit
applications where the existing electrostatic precipitator (ESP) can be used as the collector.
Since the EECS system performance is relatively independent of the collector performance,
almost any ESP, no matter how poorly it operates, can be successfully retrofitted by the
addition of an EECS. In fact, it may be possible to remove the last section of the ESP,
place the EECS inside the existing housing and still achieve very high particle collection
efficiencies with no increase in the size of the ESP footprint.
What physically happens in the EECS system is that particles that make it through the ESP
on their first pass are separated from the exhaust gas and returned to the ESP inlet for a
second pass. For these particles, the ESP appears to have an SCA twice what the original
ESP had. If any particles make it through on the second pass, they are directed back to the
inlet for a third. If the EECS had a separation efficiency of 100 percent, no particles could
escape with the exhaust gas and the system efficiency would be 100 percent as shown by
Equation 2. To recirculating particles, the system appears to have infinite SCA which is
what the Andersen-Deutch equation indicates is required for 100 percent collection
efficiency.
Of the three components that make up the EECS system, the EECS, the ESP and the
recirculation fan, the focus of this Phase I research program has been on the development
of a high efficiency EECS that is compact, has low capital cost and operates with low
pressure drop. The next section describes that development process.
Project Description
The task of developing the EECS involved taking the basic Core Separator geometry,
adding an axial discharge electrode and determining which flow regimes produced the best
performance. This was done in two steps. Some geometry changes were made in the
second EECS directed toward improving the electrical characteristics and reducing
manufacturing costs in a full scale system.
Laboratory tests were conducted using a simulated exhaust stream created by disbursing fly
ash from Public Service of Colorado's Comanche Station into the exhaust gas from a
natural gas burner. Comanche Station bums pulverized western subbituminous coal in a
subcritical boiler and the fly ash is collected in a baghouse and shipped to LSR in 55 gallon
drums. A measured amount of ash is disbursed through an air ejector into the EECS inlet
duct and the concentration of dust was measured in the EECS clean flow outlet using a
simplified EPA Method 5 sampling procedure. The calculated efficiency was then
determined using Equation 3.
Tfe^l-Q,^^, T/m. (3)
Where
WEECS EECS separation efficiency
Coul Mass concentration of paniculate matter in the EECS clean flow outlet as
determined by EPA Method 5
V Average gas velocity in the EECS clean flow outlet duct
-------
Cross-sectional area of the EECS clean flow outlet duct
Test duration
Mass of paniculate matter introduced at the EECS inlet
In addition to the efficiency, the EECS gas inlet velocity and bleed flow were measured so
that the EECS efficiency could be plotted against these two flow variables. Also, the
particle size distribution was measured at the inlet and clean flow outlet order to determine
the EECS partial separation efficiency.
While the first EECS tested had a design flow rate of 950 m3/hr (560 acfm) with a length to
diameter ratio of about 2.5 to 1, the second unit, shown in Figure 3 and Figure 4, had a
design flow rate of 2850 m3/hr (1680 acfm) with a length to diameter ratio of about 7.5 to
1. All the interior edges of the second EECS were rounded or fitted with corona shields to
allow higher field strengths before the onset of sparking. Also, the longer length means
that, for a given gas flow rate, fewer EECS subassemblies would be required thereby
reducing the capital cost Tests of this geometry showed the cost savings could be
achieved without paying a performance penalty.
After completion of the laboratory tests, the EECS unit was shipped to Alabama and
installed at Alabama Power Company's James H. Miller, Jr. Electric Generating Plant in
Quinton, Alabama. The unit was tested in an exhaust gas slipstream from Unit No. 3
burning a sub-bituminous coal from the Powder River Basin. The EECS was installed
downstream of the COHPAC n unit3 which was operated with its filter bags removed. By
reducing the output voltage of COHPAC unit's high voltage power supplies, we were able
to control the EECS paniculate inlet loading and thereby simulate the inlet loading coming
from a range of under-performing ESPs.
The final Phase I task was to produce an updated cost estimate of an EECS retrofit based
on a 250 MWe plant. The cost estimate was performed by Sargent and Lundy, LLC as part
of a subcontract. LSR provided them with a conceptual design based on the results of the
EECS testing.
Results
Laboratory Tests
Most of the EECS testing was done at ambient temperature. Due to the simplicity of
conducting ambient temperature tests, it made the process of identifying the most
interesting operating regimes much quicker. Once the ambient temperature performance
map was created, a series of high temperature tests was conducted to determine the effects
of elevated temperature.
The results showed an efficiency high of over 99 percent. They also show a trend of
decreasing efficiency with increasing inlet velocity, a trend typical of conventional ESPs.
In the mechanical Core Separator the efficiency increased with increasing inlet velocity due
to the higher centripetal forces generated so it appears that in the EECS the electrical forces
are dominant in the separation process. The EECS is indeed performing as an electrostatic
separator, not a mechanical one.
-------
Using (3, it is possible to estimate the efficiency of a complete EECS system by assuming
an efficiency for the ESP collector. For an ESP having a collection efficiency of 97
percent, the estimated system efficiency varies from 99.84 to 99.97 percent. It should be
pointed out that these are only rough estimates of the system efficiency. A more systematic
approach is to use an assumed system inlet size distribution, a partial collection efficiency
for the ESP and the partial separation efficiency for the EECS. Even this more involved
procedure still provides only an approximate result because it does not include the
performance coupling effect due to particle agglomeration and attrition.
Field Test Results
Six EECS efficiency tests were conducted at Alabama Power Company's Plant Miller. The
tests were conducted with two EPA Method 5 gas sampling trains to measure the EECS
inlet and EECS clean flow outlet paniculate concentrations. The EECS was operated at a
nominal specific collecting area of 19.7 m2/(m3/s) (100 ftVkacfm) The first test was
conducted with no electric power applied to the EECS to get a baseline performance of the
mechanical separator. The measured efficiency was 44.5 percent. The results of the final
four tests with the precharger operating normally are shown it Figure 1 in S.I. units and in
Figure 2 in English units. The efficiencies of these runs are between 86 and 96 percent.
The data show that the EECS is capable of reducing the emissions from under-performing
ESPs having outlet loadings as high as 280 ng/J (0.65 Ibm/million Btu) down to below
13 ng/J (0.03 Ibm/million Btu). All runs had a pressure drop of less than 100 pa (0.4 in
we) and during the approximately two weeks of operation the prechargers were cleaned
once with a compressed air lance while the EECS itself was not cleaned at all.
Application
The most immediate application of the EECS system will be to provide a retrofit system for
coal-fired power plants currently operating under-performing ESPs. According to a cost
study performed by Sargent and Lundy, LLC in 1994, the capital cost for a retrofit EECS
is about $8 million for a 250 MW plant including a 40 percent process contingency.4 This
is approximately the same capital cost as EPRI's COHPAC concept where a high air-to-
cloth ratio baghouse is placed downstream of an under-performing ESP. Toward the end
of this project Sargent and Lundy, LLC will re-evaluate the EECS costs taking into account
the design modifications and operating experience gained from the field tests conducted at
Alabama Power Company's Miller Steam Plant.
The second application of the EECS system is to retrofit ESPs that currently meet
paniculate emission requirements but may not meet the species-specific air toxic emission
regulations that are expected from EPA in the near future. The development of EECS
concept will proceed with the immediate needs of the first group in mind and will be ready
for the second group if the if the EPA air toxic emissions rules should require it.
The final task of this Phase I project was an economic analysis performed by Sargent and
Lundy, LLC. In 1994 Sargent and Lundy performed an economic analysis showing the
EECS to have a capital cost of about $8 million for a 250 MWe plant. This cost estimate
included a 40 percent process contingency due to its immature stage of development at the
time of the analysis. The new analysis showed a decrease in the capital cost to
$6,25 million due primarily to the fact that, with a pressure drop of only 100 pa (0.4 in
we) the forced draft fans will not need to be modified. An additional reduction is due to
lower contingencies now that the technology has been demonstrated in the field.
-------
18
16-
14 -
g 12-1
o>
c
13
3
o> 10
I
D.
O
I
UJ
8
= 11.6%,
= 283 nAfcm*2
in = 1300C,
8 = 13.8%,
J = 283 nA/cmA2
= 1270C,
8=16.0%,
J = 330 nA/cm"2
50 100 150 200 250 300 350 400
EECS Normalized Inlet Paniculate Loading [ng/J]
Figure 1: EECS Field Test Results, S.I. Units
-------
0.040
0.035 -
3
ffl
0.030 -
0.025 -
T3
eo
o
_i
ฃ 0.020
J5
3
u
ea
a.
5 0.015
O
S
0.010
W
u
111
LU
0.005 -
0.000
in = 265 ฐF,
6 = 13.8%,
J = 283 nA/cm"2
Tin = 271ฐF,
= 11.8%,
= 313nA/cm"2
Tin = 261 ฐF,
B = 16.0%,
J = 330 nA/cmA2
1.6%,
= 283 nA/cmA2
0.0 0.2 0.4 0.6 0.8
EECS Normalized Inlet Paniculate Loading [Ibjmillion Btu]
1.0
Figure 2: EECS Field Test Results, English Units
-------
Future Activities
In Phase II, a 34000 m3/hr (20000 acfm) EECS module will be constructed and tested for
particle separation efficiency and for separation efficiency of 16 HAPs. Activated carbon
sorbent will be injection upstream of the EECS to measure the separation efficiency of gas
phase air toxics such as mercury. The second phase will also include testing of the first
wet EECS element As in Phase I, the final task of phase n will be a detailed economic
assessment of the EECS system to be performed by Sargent and Lundy.
Acknowledgment
This work has been supported by the U.S. Department of Energy under contract number
DE-AC22-95PC95261, by the Electric Power Research Institute and by Southern
Company Services, Inc. During this period of performance our FETC Contracting
Officer's Representatives (CORs) for this project have been Dr. Perry Bergman and Mr.
Thomas Feeley.
References
1 D.W. Dockery, C.A. Pope HI, X. Xu, J.D. Spengler, J.H. Ware, M.E. Fay, E.G.
Ferris, F.E. Speizer. "An Association Between Air Pollution and Mortality in Six
U.S. Cities." New England Journal of Medicine. Vol. 329, p. 1753 (9 Dec 1993)
2 G.B Nichols, Letter to the editor, Air and Waste, Air and Waste Management
Association, Vol 44, p. 1028 (Aug 1994)
3 W.A. Harrison, K.M. Gushing, R.L. Chang. "Pilot Scale Demonstration of a
Compact Hybrid Paniculate Collector (COHPAC) for Control of Trace Emissions
and Fine Particulates from Coal-Fired Power Plants." presented at the EPRI/DOE
International Conference on Managing Hazardous and Paniculate Air Pollutants.
Toronto, Ontario, (Aug 1995)
4 R.P. Gaikwad, D.G. Sloat, R.F. Altman, R.L. Chang. "Engineering Evaluation of
Novel Fine Paniculate Control Technologies." presented at the EPRI/DOE
International Conference on Managing Hazardous and Particulate Air Pollutants.
Toronto, Ontario, (Aug 1995)
-------
Figure 3: EECS Showing Clean Row Outlet
-------
Figure 4: EECS Showing Precharger and Bleed Flow Outlet
-------
Figure 5: EECS at Plant Miller
-------
DEVELOPMENT OF THE LAMINAR-FLOW FINE-PARTICLE
AGGLOMERATOR
Paul Feldman
Kevin Mills
Environmental Elements Corporation
3700 Koppers Street
Baltimore, MD 21227
Abstract
This paper presents the current status of the commercial development of a new technology to
more efficiently control fine particulate emissions. The technology is based on an invention by
Environmental Elements Corporation (EEC) which utilizes laminar flow to promote contact of
fine submicron particles with larger particles to form agglomerates prior to their removal in a
conventional particulate control device, such as an electrostatic precipitator. As agglomerates
the particles are easily captured in the control device, whereas a substantial amount would pass
through if allowed to remain as fine particles. EEC has developed the laminar-flow agglomerator
technology through the laboratory proof-of-concept stage, which was funded by a DOE SBIR
grant, to pilot-scale and full-scale demonstrations. Preparations are now underway for two
additional demonstrations.
The Need For Fine Particulate Control
There is little question that standards for the control of particulate emissions are becoming more
and more stringent, especially for fine particles in the micron and submicron size range. The
driving forces for tightening of fine particulate emission standards in the United States are PM 2.5
regulations, visibility concerns, stack opacity requirements, and the need for control of air toxics
emissions. High-efficiency control of the submicron fraction of particulate emissions is necessary
to meet each of these needs.
The air toxics resulting from coal combustion tend to concentrate in the fine particulate fraction
of the emissions. Heavy metals and heavy organics which were volatilized at the high
temperatures of combustion, tend, during cooling, to nucleate as very fine particles, or to
condense on existing flyash particles. This condensation takes place mainly on submicron
particles because of their great number and total high surface area compared to larger particles.
Stack opacity and long-range visibility are also largely determined by the fine particulate fraction
of the flyash because the light extinction coefficient reaches a maximum near the wavelength of
light, which is between 0.1 to 1 micron.
-------
Proposed PM 2.5 regulations focus specifically on the reduction of fine particulate matter in the
atmosphere. Although a primary goal is the reduction of secondary particulate matter, Le.
particles formed by condensation or reaction of gaseous species such as sulfur and nitrogen oxides
after emission from the stack, the pre-stack control of primary fine particulate matter will also be
important in satisfying the PM 2.5 requirements.
Approaches To Effective Fine Particulate Control
Typical particulate emissions, e.g. from a coal-fired boiler, are composed of a range of particle
sizes from submicron to tens of microns. Particles larger than a few microns in diameter are
relatively easy to remove from the gas stream by inertial and electric forces. Very fine particles,
Le. those less than 0.05 microns, move freely by diffusion and are removed by contact with other
particles or surfaces. The most difficult particles to remove from the gas stream are those in the
size range 0.1 to 2 microns because they are not affected strongly by either inertia or diffusion. It
is unfortunate that particles in this size range affect plume opacity, visibility and health much more
adversely than particles outside this range. It is further unfortunate that for a typical coal-fired
boiler, the median diameter of particulate emissions on a number basis is 0.27 microns, the most
difficult size for control purposes.
The need for fine particulate control is obvious, but the challenge remains to accomplish the
control effectively. It is possible to increase the efficiency of capture of particles in the 0.1 to 2
micron size range by maximizing the effectiveness of diffusion or inertia, but, in past practice,
economic considerations have limited this approach. For example, to increase the effectiveness of
diffusive capture, it is necessary to provide greater surface area and/or more time for diffusion to
occur, but this means a significant increase hi equipment size. Inertial forces can be increased by
increasing the relative velocity of the particle to the collecting surface, but this can be achieved
only at the expense of large increases in pressure drop and power input to the collecting device.
Although the economic control of particles in the 0.1 to 2 micron range is a difficult problem,
there are several approaches which can potentially lead to its solution. The more promising of
these are particle growth by agglomeration, particle growth by condensation, wet electrostatic
precipitation, and electrostatic enhancement of filters. This paper focuses on the growth of fine
particles by agglomeration and specifically on the laminar flow agglomerator which is now under
development by EEC.
Particle Growth by Agglomeration
In the agglomeration process the individual fine particles become part of larger, more easily
collected particles by cohering to themselves or attaching to larger particles. It would be ideal if
particles could be agglomerated while they are suspended in the flue gas, but normal utility
particulate concentrations are too low to allow for frequent enough particle-particle contact to
promote significant agglomeration. Attempts have been made to enhance the suspended particle
agglomeration rate electrostatically or by applying sonic energy, but without practical success.
The authors believe there are more promising approaches to promote fine particle agglomeration,
Le. agglomeration in very high particulate concentrations, or agglomeration on a contacting
surface in a laminar flow section of an electrostatic precipitator. EEC has initiated programs to
evaluate both of these approaches.
-------
High-Concentration Agglomeration. This approach is to increase the paniculate
concentration in the gas phase to the point where the probability of contact by paniculate
diffusion of the fines is high enough to allow sufficient contact in a short residence time.
Concentration of several hundred grains per cubic foot is required to achieve a satisfactory rate
of contact, and this can be achieved in a circulating fluid bed using foreign bed material or using
recirculated flyash. In fact, there is ample experience with this concept in circulating dry
scrubbers used for FGD applications where lime is injected in the fluid bed to achieve the
necessary high paniculate concentration. Typically the electrostatic precipitator following the bed
is extremely efficient in collecting the agglomerated particles from the bed in this application.
EEC completed a Phase I U.S. Department of Energy SBIR program to verify the concept, and is
now starting a Phase n pilot program to demonstrate the effectiveness of the concept on a
slipstream at a major U.S. power plant.
Laminar Flow Agglomeration. A conventional electrostatic precipitator runs in highly
turbulent flow, and its performance is modeled by the Deutsch model which assumes complete
mixing of all particles between the collecting plates. Electrical forces are operative only across
the laminar boundary layer. This model leads to the familiar exponential relationship between
collection efficiency and the product of the electrical migration velocity of the particles and the
specific collecting area (SCA) of the precipitator. The exponential nature of the Deutsch equation
means that the investment of SCA yields decreasing returns in efficiency at high collection
efficiency levels; ie. 100% collection efficiency is approached only asymptotically in turbulent
flow and cannot be reached no matter how large the precipitator.
The advantage of laminar flow becomes apparent when comparing the laminar flow model to the
Deutsch model Figure 1 is useful in visualizing this comparison. In laminar flow the flow
streamlines are parallel and in the direction of flow; there is no force causing particles near the
collecting surface to be thrown back into the central flow region as there is in the Deutsch model
Therefore the electrical forces tending to move the particles toward the collecting surface are
effective across the entire flow cross section, not just across the laminar sublayer. This model
yields an equation which shows collection efficiency directly proportional to the product of
migration velocity and specific collecting area (SCA), ie. 100% collection efficiency is possible in
laminar flow with a relatively small SCA requirement.
Figure 2 illustrates the striking difference in paniculate collection capabilities between turbulent
and laminar precipitation collecting typical utility flyash. The comparison in electrostatic
precipitator size from laminar to turbulent is dramatic: at 99% collection efficiency a turbulent
precipitator is more than twice the size of an equivalent laminar precipitator, and at 99.99%
efficiency the turbulent precipitator is more than five times larger than its laminar counterpart.
Despite these advantages, a practical industrial-scale laminar-flow electrostatic precipitator has
yet to be put in operation. The reason is that a reduction in Reynolds Number by almost an order
of magnitude from conventional practice is necessary to reach laminar flow: this is not feasible if
the electrostatic precipitator is to operate as a conventional single-stage particle collector, e.g. it is
felt that the effective flow dimension cannot practically be reduced enough to approach laminar
flow, and, because of flow distribution and aspect ratio considerations, freedom to approach
laminar flow by lowering gas velocity is quite limited. An additional upsetting factor is the corona
wind which is induced by the corona discharge current in the precipitator. By itself the corona
wind is enough to promote turbulence.
-------
TURBULENT: DEUTSCH (COMPLETE MIXING) MODEL:
PARTICULATE REMOVAL
; 1 - EXP(-wUvS) RATE OF COLLECTION PROPORTIONAL TO (1-ri)
LAMINAR:
r
) = wL/vS
PARTICULATE REMOVAL
RATE OF COLLECTION CONSTANT
FIGURE 1: COMPARISON OF
TURBULENT AND LAMINAR FLOW PRECIPITATION
-------
I
a
ซ 500
400-1
300-1
to 200-1
99 99.5 99.9 99.95
COLLECTION EFFICIENCY (%)
99.99
TURBULENT
LAMINAR
FIGURE 2: PRECIPITATOR SIZES
LAMINAR VS. TURBULENT FLOW
-------
Rather than discouraging further consideration of laminar flow, these observations should instead
be regarded as establishing important criteria for taking advantage of true laminar flow in an
electrostatic precipitator-type device. I.e. the device should ideally not be of the typical parallel
plate design, and it should not produce corona discharge in the particle collecting zone. A design
which meets these criteria is a two-stage, tubular precipitator in which the particle collecting area
is comprised of many parallel small-diameter tubes sized to give laminar Reynolds Number. The
two-stage provision means that there is no corona discharge in the collecting section, but a non-
discharging electric field must be applied to provide the necessary electrical force for particle
collection, e.g. by non-discharging rods along the axis of each tube. Particles must be charged, of
course, ahead of the collecting section in a separate charging section in which a copious corona
discharge is provided for rapid particle charging in a compact charging zone. A problem with this
design, however, is that, by eliminating corona discharge in the collecting section, there is no
holding force to keep the collected particles on the collecting surface, and reentrainment of these
particles is likely. Reentraimnent in this type design can be quite severe, even in laminar flow, and
can destroy the usefulness of the device as a high-efficiency electrostatic precipitator. However, if
this "collecting section" is regarded as a means of assuring that all of the particles, including fines,
are able to come together on a surface, they can then form agglomerates which will be more easily
collected in a downstream conventional precipitating section. The net result is for the two-stage
device to act as an agglomerator for a downstream collector, with overall fine paniculate
collecting efficiency being much higher than if the entire device were of conventional electrostatic
precipitator design.
Although the theoretically ideal design for the agglomerator is the tubular design, it was found, as
discussed below, that closely spaced parallel plates with horizontal flow can also be used
effectively as a laminar-flow agglomerator. This arrangement is, of course, more compatible with
conventional electrostatic precipitators and can find practical use as a modification to
conventional precipitators.
Development Of The Laminar Flow Agglomerator
Laboratory Tests in Tubular Geometry
The laminar flow agglomerator concept was first tested in the EEC laboratory under a DOE SBIR
grant using a tubular electrostatic precipitator geometry. The lab system used in this program was
designed to treat up to 100 cfin of gas and consisted of a gas preparation section, the laminar flow
vessel, a tubular precipitator to represent conventional precipitation, and associated
instrumentation and equipment.
Ambient air was used as the carrier gas and entered into the gas preparation section which
consisting of a heater, humidification section, and a cooler to adjust to the desired operating
temperature. Particulate was fed into the system via a screw feeder connected to an air powered
eductor to insure good particle distribution. The particulate used in the program was a mixture of
approximately 75% silica-magnesia mixture and 25% sodium sulfate. Its characteristics were
similar to coal flyash in that its particle size distribution was a mass median diameter of 10
microns with a geometric standard deviation of 3, and its electrical behavior in the precipitator
was as moderately high resistivity at the 250ฐF operation.
-------
Following the gas preparation section the gas entered the laminar flow device which consisted of
21-1.5" tubes, each 5 feet long. Negative polarity electrodes were suspended axially along the
center of each tube, the first foot of which was of small diameter to allow corona discharge for
particle charging. For the two-stage tests the remainder of the center electrode was 1/4" rod to
provide a strong, non-discharging field for particle collection. For the single-stage tests, the
discharge geometry was continued for the whole length of the center electrode. The length of the
center electrode was varied as a means of testing the importance of collecting section length. The
laminar flow device was designed to operate in either dry or wet operation. For wet operation
water was introduced at the top of the tubes and flowed over tube extensions into each tube.
Flow was downward through the device, and a hopper was situated below the tubes for
particulate removal.
Downstream of the laminar flow device was located an 8"-diameter, 5 foot long tubular
electrostatic precipitator fitted with a central negative polarity standard 0.109" discharge
electrode. The purpose of this device was to represent conventional precipitation. It too was
designed to operate both wet and dry.
The most important measurements made during the test program were particle size and
concentration. Particle size was routinely measured in an Insitec PCSV-UFP particle counter
through probes located at the inlets and outlets of the laminar device or the 8" tube. The Insitec
is a single particle counter that measures the light scattered from individual particles as they pass
through a focused laser beam. Particle concentration was also indicated by the Insitec with
checks by occasional filter catches. Loading and particle size checks were also made using a
University of Washington cascade imp actor.
The primary variables for the test program were the type of operation, ie. single stage or two
stage and wet or dry; the lengths of the discharging and non-discharging sections; gas velocity;
and temperature. The resultant variables of Reynolds number and residence time were controlled
through these primary variables. Particulate characteristics were not variables in this program, the
single dust type which was described above being used in all runs. The following table shows the
range of conditions used in the test program. In this table DE is the discharge electrode length,
NDE is the length of the non-discharging electrode, Re is Reynolds Number, TEMP is gas
temperature, VEL is gas velocity through the device, and RES is residence time of the gas stream
in the device.
TEST CONDITIONS
DE
ft
5
5
1
1
1-5
NDE
ft
0
0
2-4
2
0
Re
8000
17000
1160-3063
1561-2961
1160-2034
TEMP
F
160-250
100
160-250
100
250
VEL
ft/sec
2.56-5.72
5.08
2.34-6.62
2.20-4.20
2.50-4.39
RES
sec
0.87-1.95
0.98
0.45-2.14
.71-1.61
0.40-2.00
MODE
8" tube esp - dry
8" tube eso - wet
Two-staqe - drv
Two -staae - wet
Sinale-staae - dry
-------
The procedure used for each run was to first run the fan without dust feed for several minutes and
then to bring the gas to desired levels of temperature and humidity by adjusting the heater, cooler,
and steam injection. When the temperatures throughout the system stabilized and flow was set to
desired velocity, the dust feeder was turned on. After a few minutes, power was applied to the
test precipitator and the run timer was started. Data was then taken after 5 minutes of operation
and then periodically for the remainder of the run.
The test program verified the agglomeration concept by demonstrating the growth of the particle
size distribution after passage through the agglomerator. To determine the extent of
agglomeration taking place, the characteristics of the outlet particle size distributions were
examined in comparison to the inlet distribution and to computer-predicted outlet distributions
assuming no agglomeration. The outlet size characteristics used were the d(10) and d(50) values.
The quantity d(10) represents the particle diameter in microns below which 10% of the total
particulate mass resides. It is a good indicator of the fine particulate population of the size
distribution. For the inlet size distribution used in this program, Le. 10 micron mass median
diameter (MMD) & standard deviation 3, d(10) is 2.4 microns. In conventional precipitation, as
collection efficiency increases, d(10) and the MMD decrease, reflecting the loss of the more easily
captured larger particles.
Figure 3 plots d(10) versus mass penetration for all of the two-stage, small-diameter tube, dry
operation data points as well as for the 8" tube, dry operation data points. The solid line on the
figure represents the theoretical prediction of d(10) as a function of penetration. The broken lines
represent the regression fits for the 1.5" tubes and for the 8" tube data. The figure shows that, for
the 8" tube, the d(10) values and the theory agree very well, indicating that the emissions from the
8" tube represent the particle size distribution expected of particles that were not collected in the
tube. The data for the 1.5" tube, however, lie significantly above the theoretical and 8" lines,
meaning that the outlet emissions are less fine than would be expected with no agglomeration
taking place; Le. the size distribution of the emitted particles reflect the fact that fine particle
agglomeration had occurred in the tube. Electrostatic precipitator performance projections based
on the particle growth data indicate that replacement of a single electrostatic precipitator field
with an agglomerating field can be equivalent to the addition of two or more conventional fields
to the preciprtator. Economic advantages are significant because of the avoidance of the need for
space, structure, ducting, ash handling provisions, etc. for an extra precipitator field.
Laboratory Tests in Parallel Plate Geometry
After the successful tubular pilot tests, it was decided to test the concept hi a parallel plate type
geometry, even though achievement of laminar flow is more difficult because of the necessarily
high vertical plate dimension. The horizontal-flow, parallel plate arrangement is much more
suitable to practical use than the vertical tubular design. Encouragement for the viability of
parallel-plate laminar operation was gotten after testing a plexiglass model of parallel plate
configurations. Smoke was injected into air flow at various velocities between plexiglass plates
spaced at various distances. Visual observation of the smoke patterns at the various velocities
and plate spacings showed that very well behaved, virtual laminar flow was achievable with
reasonable plate spacings and velocities.
-------
V)
z
o
DC
O
Q
0.0001
0.001
0.01
MASS PENETRATION
1.5" TUBES
+ 8" TUBE
THEORY
REGRESSION FITS
FIGURE 3: ILLUSTRATION OF
D(10) GROWTH IN LAMINAR AGGLOMERATOR
-------
In order to explore the paniculate collection capabilities of a precipitator fitted with a parallel-
plate laminar agglomerator field, a single-gas-passage, three-field pilot electrostatic precipitator
was erected in the EEC laboratory with 5 ft plate height. The inlet field was 3 ft long and the
second and third fields were 5 ft long. The first and third fields were conventional precipitator
fields with 9 inch plate spacing and rigid discharge electrodes. In the agglomerator mode of
operation, the second field was configured as a non-discharging plate-to plate field without
discharge electrodes, and with variable plate spacings. Alternate plates were grounded or at high
voltage in the agglomerator field.
The pilot unit was operated at ambient temperature with gas velocities ranging from 3 to 6 ft/sec.
Ash was injected at the rate of 1 grain per cubic foot, with particle size distribution characterized
by a mass median diameter of 10 microns and geometric standard deviation of 3. A series of runs
were made with the second field at different plate-to-plate spacings. For comparison to
conventional performance, runs were made with the second field fitted with conventional
electrostatic precipitator geometry similar to the first and third fields.
Figure 4 shows the results of the runs by plotting on log-log paper the negative log of penetration
versus a parameter, K3. The K3 parameter is, in effect, the product of the electrostatic
precipitator SCA and the effective particle migration velocity based on the mass median diameter
of the inlet particle size distribution. Theoretically a plot of the negative log of penetration vs. K3
should yield a straight line on log-log paper. For reference, the right hand y-axis shows actual
penetration values as fractions of the inlet particulate concentration. Three sets of data are
plotted: (1) the base data with all three fields in conventional precipitator configuration, (2) data
for closest plate spacing in the agglomerator field (labeled as AGGLOMERATOR 1), and (3)
data for somewhat wider agglomerator plate spacing (labeled as AGGLOMERATOR 2). It can
be seen from the figure that the data points for each mode of testing align very well Most
importantly it can be seen that overall electrostatic precipitator performance increases
significantly, and the data for the closest plate spacing, AGGLOMERATOR 1, is best, as should
be expected.
The data were analyzed with the assumption that essentially all particles contacting the plates in
the agglomerator field were reentrained, and that essentially all particulate collection and removal
occurred only in the first and third fields. It is important to note that h was possible to achieve
virtually complete particulate collection in some of the agglomerator runs, especially with the
configuration using the closest agglomerator plate spacing. Completion of this test program left
no doubt that the laminar flow agglomeration concept is truly a promising means for high-
efficiency control of fine particles.
An example of the effectiveness of the parallel-plate agglomerator performance can be found by
comparing the data sets for a K3 value of 1: While base performance with conventional
precq>itator geometry yielded a penetration of 2.5%, the wider spaced agglomerator allowed a
penetration of only 0.09% and the closer spaced agglomerator allowed only 0.0045% penetration.
Analyzing the data in a more pragmatic way, it was found that replacement of the center field of a
conventional ESP with a parallel-plate agglomerator field resulted in total precipitator particle
collection efficiency equivalent to a conventional precipitator with more than twice the SCA of
the original precipitator.
10
-------
100-
g
I
tH 10-
01
Q.
0.1
^
ei-
j>r
X
X
- 1E-3
1E-2
X BASE DATA
1
K3
CD AGGLOMERATOR 1 AGGLOMERATOR 2
10
FIGURE 4: EFFECT OF AGGLOMERATION
ON PARTICULATE PENETRATION
-------
Full-Scale Field Demonstration
Encouraged by the results of the parallel plate pilot program, EEC took advantage of an
opportunity to install a laminar agglomerator field in the second field of a full-scale three-field
industrial electrostatic precipitator. The electrostatic precipitator follows a coal-fired industrial
boiler, and, in fact, is one of three parallel, similar boiler/electrostatic precipitator units at the site.
Thus, the installation of the agglomerator field in one of these units allowed direct comparison of
the performance of the agglomerator electrostatic precipitator with two similar ESP's of standard
configuration.
The agglomerator field was installed with specially designed 24-foot high plates, spaced closely
together to effect laminar or near laminar flow in the field. The plates were alternately at high
voltage or ground. No discharge electrodes were used in the agglomerating field. The
agglomerator precipitator was operated as needed, generally five days per week, for five months,
during which time its stack opacity was always lower than the other two parallel precipitators.
Typically, the opacity comparison was 6% for the agglomerator precipitator vs. 20% for the
standard ESP's when collecting high-resistivity ash, and 3% vs. 10% when collecting moderate-
resistivity ash.
A test program was planned to collect quantitative emissions data to better characterize the
performance of the agglomerator. However, a fuel upset in the boiler serving the agglomerator
precipitator caused a fire which passed through the electrostatic precipitator and resulted in plate
warp age and twisting in the first field and similar but less severe damage in the agglomerator field.
The test program could not be carried out. However, it is important to note that the
agglomerator field, despite some plate distortions, was able to continue operation in an
electrically- and mechanically-acceptable manner.
Laminar Agglomerator Programs in Progress
At the present time EEC is installing a pilot-scale 3-field agglomerator precipitator on a slip-
stream from a recovery boiler to test the applicability of the laminar agglomerator to the collection
of saltcake. Saltcake from a recovery boiler is a very fine particulate and can be somewhat
"sticky" One objective of the test program is to determine if the saltcake "stickiness" causes
buildup problems which may be difficult to overcome in the close plate spacing of the
agglomerator. A factor mitigating against buildups is the non-discharging nature of the
agglomerating field. High field strength with no current introduces a strong electrical
reentrainment component which will tend to keep the plates clean. The cohesive nature of the
saltcake can then be a positive factor in building strongly bonded agglomerates for downstream
collection. The test program will be completed later hi 1997.
As part of EEC affiance with Wisconsin Electric Power Company (WEPCO), EEC is preparing to
install an agglomerator field in a precipitator at WEPCO's Presque Isle Station. The experience
gained in the full-scale industrial boiler installation was an important step in the planning, design
and preparation for this first full-scale utility agglomerator demonstration. The agglomerator field
will be retrofit in the second of three fields in the Presque Isle precipitator. Installation will be in
early 1998.
12
-------
Economics Of Laminar Flow Agglomeration
The primary advantage of laminar flow agglomeration is that an electrostatic precipitator fitted
with an agglomerating field can achieve particle collection efficiencies equivalent to a much larger
conventional precipitator. Laboratory tests showed the agglomerator precipitator to be equivalent
to a conventional precipitator of more than twice its size. This is perhaps more than should be
expected when scaled up to full-scale application, but it does give confidence that the
agglomerator modification to a precipitator is likely to be equivalent to a substantial increase in
the precipitator's SCA A more quantitative definition of agglomerator equivalence will be
developed following full-scale demonstrations of the technology.
On a retrofit basis, the agglomerator modification merely replaces the hardware in a center
precipitator field or portion of a field, and does not require additional plant space as addition of
fields would. For a new precipitator the same advantage applies in that the agglomerator
precipitator can be much smaller than conventional
The economic comparison of interest here is the cost of converting a field or portion of a field of
an existing precipitator to close-spaced plates versus the cost of adding new fields to the
precipitator. For the case of a precipitator upgrade, adding new fields to the precipitator adds
costs for the precipitator casing and internals, hoppers and ash handling for the new fields, as well
as costs for additional ductwork to reroute exhaust from the newly added fields to the stack and
additional support steel and foundations for the newly added fields.
For the agglomerator modification to the precipitator, the major costs are for the new plates to
replace the original collecting plates and discharge electrodes, and reinforcement of existing
support steel and foundations. Very importantly from a cost standpoint, no additional hoppers or
ash handling systems or ductwork rerouting are necessary with this modification because all of the
change is restricted within the original precipitator.
The following table shows comparative costs for an agglomerator modification vs. addition of
fields to a precipitator.
COMPARATIVE COSTS: AGGLOMERATOR VS. UPGRADE
Internals & Casing
Ash Handling
Ductwork Mod.
Foundation/Structures
Total Installed Cost
Agglomerator, $K
1,483
150
1,633
Upgrade, $K
2,450
1,440
400
340
4,630
The costs are for a precipitator collecting flyash from a 300 MW boiler and requiring an increase
of 85% of its SCA from 267 to 495. Assuming the agglomerator modification can achieve the
same level of performance as the increased SCA, it is clear from this table that the agglomerator
option is by far the most economical means for solving this particular plant emissions problem,
total installed cost being nearly one third of that for the conventional precipitatpr add-on to
achieve the same result. The reason for such a large difference is the fact that additional ash
handling systems and ductwork modifications are not needed with the agglomerator option. A
13
-------
coincident advantage for the agglomerator option is the avoidance of the need for plant space for
the preciphator addition, which in many cases can be quite difficult. For most of the existing
boiler population, therefore, an electrostatic preciphator performance upgrade strategy that
confines the upgrade within the existing electrostatic precipitator casing is most attractive.
Acknowledgment
Environmental Elements Corporation is appreciative of DOE SBBR. Grant No. DE-FG05-
94ER81760 under which the tubular precipitator portion of the work was funded.
14
-------
ADVANCED HYBRID PARTICULATE COLLECTOR, A NEW
CONCEPT FOR AIR TOXICS AND FINE-PARTICLE CONTROL
S.J. Miller
G.L. Schelkoph
G.E. Dunham
Energy & Environmental Research Center
15 North 23rd Street
Grand Forks, ND 58203
K. Walker
W.L. Gore & Associates, Inc.
101 Lewisville Road
Elkton, MD 21922-1100
H. Krigmont
Allied Environmental Technologies Company
One Pacific Plaza
7755 Central Avenue, Suite 1118
Huntington Beach, CA 92647
Abstract
A new concept in particulate control, called an advanced hybrid particulate collector (AHPC), is
being developed under funding from DOE. The AHPC combines the best features of electrostatic
precipitators (ESPs) and baghouses in a manner that has not been done before. The AHPC
concept combines fabric filtration and electrostatic precipitation in the same box, providing
major synergism between the two methods, both in the particulate collection step and in transfer
of dust to the hopper. The AHPC provides ultrahigh collection efficiency, overcoming the
problem of excessive fine-particle emissions with conventional ESPs, and solves the problem of
reentrainment and recollection of dust in conventional baghouses.
Phase I of the development effort consisted of design, construction, and testing of a 200-acfm
working AHPC model. Results from both 8- and 100-hour tests are presented.
Introduction
The primary technologies for state-of-the-art particulate control are fabric filters (baghouses) and
electrostatic precipitators (ESPs). However, each of these has limitations that prevent it from
achieving ultrahigh collection of fine particulate matter. A major limitation of ESPs is that the
fractional penetration of 0.1- to 1.0-um particles is typically at least an order of magnitude
greater than for 10-um particles, so a situation exists where the particles that are of greatest
1
-------
health concern are collected with the lowest efficiency. Fabric filters are currently considered to
be the best available control technology for fine particles, but they also have weaknesses that
limit their application. Emissions are dependent on ash properties and typically increase if the
air-to-cloth (A/C) ratio is increased. In addition, many fabrics cannot withstand the rigors of
high-S03 flue gases, which are typical for bituminous fuels. Fabric filters may also have
problems with bag cleanability and high pressure drop, which has resulted in conservatively
designed, large, costly baghouses.
The objective of the advanced hybrid particulate collector (AHPC) is to. overcome the
deficiencies of ESPs and fabric filters and achieve >99.99% particulate collection efficiency for
all particle sizes from 0.01 to 50 um, be applicable for use with all U.S. coals, and be cost-
competitive with existing technologies.
Approach
State-of-the-art ESPs can provide 99.9% total mass particulate control, but collection efficiency
for 0.1- to 1.0-um particles is significantly lower. Current fabric filters can achieve 99.9%
collection efficiency on large coal-fired boilers, and when advanced fabrics are employed or
when flue gas conditioning is used, fabric filters can achieve 99.99% collection efficiency with
no significant deterioration in performance for sizes from 0.1 to 1.0 um. Fabric filters cannot
routinely achieve that level of control for all coals within economic constraints, and studies have
shown that collection efficiency is likely to deteriorate significantly when the face velocity is
increased. u An approach to make fabric filters more economical is to employ smaller baghouses
that operate at much higher A/C ratios. The challenge is to increase the A/C ratio for economic
benefits and to achieve ultrahigh collection efficiency at the same time. To achieve high
collection efficiency, the pores in the filter media must be effectively bridged (assuming they are
larger than the average particle size). With conventional fabrics at low A/C ratios, the residual
dust cake serves as part of the collection media, but at high A/C ratios, only a very light residual
dust cake is acceptable, so the cake cannot be relied on to help achieve high collection efficiency.
The solution is to employ a sophisticated fabric that can ensure ultrahigh collection efficiency
and endure frequent high-energy cleaning. In addition, the fabric should be reliable under the
most severe chemical environment likely to be encountered (such as high SO3). Such a fabric is
already commercially available but is not widely applied to coal-fired boilers because of its
higher cost compared to conventional fabrics. The fabric is GORE-TEXฎ membrane on GORE-
TEXฎ felt which can achieve very high collection efficiencies at high A/C ratios. GORE-TEXฎ
membrane filter bags consist of a microporous, expanded polytetrafiuoroethylene (PTFE)
membrane laminated to a felted or fabric backing material. Consequently, even fine,
nonagglomerating particles do not penetrate the filter, resulting in significant improvements in
filtration efficiency, especially for submicron particles. This fabric is also rugged enough to hold
up under rigorous cleaning, and the all-PTFE construction alleviates concern over chemical
attack under the most severe chemical environments. Although GORE-TEXฎ membrane filter
media is more expensive than conventional fabrics, the much smaller surface area required for
the AHPC makes the use of GORE-TEXฎ membrane filter media economical.
Successful operation of fabric filters at high A/C ratios is a challenge because of the increasing
difficulty in controlling pressure drop. The size of fabric filters and bag-cleaning frequency are
-------
determined by pressure drop. For viscous flow, pressure drop across a fabric filter is dependent
on three components:
AP = KfV + K2WRV + K2 C V2177000 (1)
where:
AP = differential pressure across baghouse rube sheet (in. of water)
Kf = fabric resistance coefficient (in. of water-min/ft)
V = face velocity or A/C ratio (ft/min)
K2 = specific dust cake resistance coefficient (in. of water-ft-min/lb)
WR = residual dust cake weight (Ib/ft2)
C = dust loading (grains/acf)
t = filtration time between bag cleaning (min)
The first term in Eq. 1 accounts for the pressure drop across the fabric. For conventional fabrics,
the pore size is quite large and the corresponding fabric permeability is high, so the pressure drop
across the fabric alone is negligible. To achieve better collection efficiency, the pore size can be
significantly reduced, without making the fabric resistance a significant contributor to pressure
drop. The GORE-TEXฎ fabric allows for this optimization by providing a microfine pore
structure while maintaining a sufficient fabric permeability to permit operation at high A/C
ratios. The second term in Eq. 1 accounts for the pressure drop contribution from the permanent
residual dust cake that exists on the surface of the fabric. For operation at high A/C ratios, the
bag cleaning must be sufficient to maintain a very light residual dust cake and ensure that the
pressure drop contribution from this term is not unreasonable (e.g., up to 50% of the total). The
third term in Eq. 1 accounts for the pressure drop contribution from the dust accumulated on the
bags since the last bag cleaning. K2 is determined primarily by the fly ash particle-size
distribution and the porosity of the dust cake. Typical K2 values for pulverized coal (pc)-fired fly
ash range from about 3 to 15 in. of water-ft-min/lb, but may, in extreme cases, cover a wider
range. Of interest is the maximum A/C ratio at which a baghouse can be expected to operate
reliably for the range of K2 values likely to be encountered. All three terms in Eq. 1 may require
increased bag-cleaning frequency with increased A/C ratio, but the third term dictates the
minimum bag-cleaning interval. From Eq. 1, with a face velocity of 2 ft/min, a dust loading of
3 grains/acf, and a AP increase of 4 in. of water, the required bag-cleaning frequency is greater
than 100 min when K2 is less than 23 in. of water-ft-min/lb. In a reverse-gas utility baghouse,
cleaning takes place off-line and may require several minutes per compartment and more than an
hour to clean all of the compartments. This is one reason why most reverse-gas baghouses are
conservatively designed for a face velocity of 2 ft/min. To ensure that adequate cleaning time is
available when K2 is not known demands a conservative approach. On the other hand, if K2 were
known to be less than 6 in. of water-ft-min/lb, Eq. 1 implies that a face velocity of 4 ft/min could
be employed. However, to date, reverse-gas baghouses have not been designed much above face
velocities of 2 ft/min because an effective method of controlling K2 has not existed and excessive
residual dust cake weight is frequently encountered.
Pulse-jet baghouses have the potential to operate at much higher face velocities because bags can
be cleaned more often and adequate pulse energy can usually prevent excessive residual dust
-------
cake buildup. Assuming that bag life is acceptable and that low particulate emissions can be
maintained through the use of advanced filter materials, face velocities much greater than
4 ft/min should be possible. Assuming 10 min is the minimum cleaning cycle time for a pulse-jet
baghouse, a face velocity of 4 ft/min is adequate to handle a dust with a K2 greater than 30 in. of
water-ft-min/lb (see Figure 1). If K, is less than 15 in. of water-ft-min/lb, the face velocity can be
increased to 8 ft/min. For many dusts, this might be possible with conventional systems.
Doubling face velocity again to 16 ft/min implies that K2 would have to be less than 4 in. of
water-ft-min/lb. This is lower than most typical K2 values; however, through the use of flue gas
conditioning, it may be possible. Increasing the face velocity beyond 16 ft/min appears to be
stretching the theoretical limit for a full dust loading of 3 grains/scf. However, if the actual dust
loading that reached the fabric were reduced by a factor of 10, the allowable K2 would increase
by a factor of 10, while keeping the cleaning interval at 10 min. If a process could collect 90% of
the dust before it reached the bags, a K2 of up to 40 would be allowable at an A/C ratio of
12 ft/min and a 10-min bag-cleaning interval (see Figure 2). The K2 for almost all coal fly ash
dusts is likely to be less than 40, even allowing for some size fractionating between the
precollected dust and the dust that reaches the bags. Therefore, a theoretical basis exists to
operate a fabric filter at a reduced dust loading and high A/C ratio with a reasonable bag-cleaning
frequency.
1000n
- 100-
_o>
ฃ
O
o>
_
ง 10 -J
O
Dust Loading = 3 grains/ft'
DP Increase = 4 in. H O
10 15 20 25
K, in. of water-ft-min/lb
30
Figure 1
Cleaning Cycle Requirements for Full Dust Loading
-------
1000
C
E
100
CD
c
03
jO)
o
I Dust Loading = 0.3 grains/ft3
AP Increase = 4 in. H2O
10
0 5 10 15 20 25
K2, in. of water-ft-min/lb
Figure 2
Cleaning Cycle Requirements for Reduced Dust Loading
The preceding analysis is valid as long as the dust can be effectively removed from the bags and
transferred to the hopper without significant redispersion and re-collection. With pulse-jet
cleaning, heavy residual dust cakes are not typically a problem because of the fairly high cleaning
energy that can be employed. However, the high cleaning energy can lead to significant
redispersion of the dust and subsequent re-collection on the bags. The combination of a very
high-energy pulse and a very light dust cake tends to make the problem of redispersion much
worse. The barrier that limits operation at high A/C ratios is not so much the dislodging of dust
from the bags as it is transferring the dislodged dust to the hopper. Therefore, any improvement
that facilitates transfer of the dislodged dust to the hopper without re-collection on the bags will
greatly enhance operation at higher A/C ratios. The AHPC achieves enhanced bag cleaning by
employing electrostatic effects to precollect a significant portion of the dust and to facilitate
moving the dust from the bags to the hopper.
While very large ESPs are required to achieve >99% collection of the fine particles, a small ESP
can remove 90% to 95% of the dust. Including rapping puffs, 90% to 95% collection efficiency
can be achieved with full-scale precipitators with a specific collection area (SCA) of less than
100 ft2 of collection area/1000 acfrn. In the AHPC concept, the goal is to employ only enough
ESP plate area to remove approximately 90% of the dust. Similarly, the cloth area should be held
to a minimum to keep the cost reasonable. If the fabric is operated at an A/C ratio of 12 ft/min
and the SCA of the ESP is 83, the filtration collection area will be the same as the plate
collection area. An SCA of 83 should be sufficient to easily remove at least 90% of the dust (note
that an alternative definition of SCA is simply the inverse of A/C ratio multiplied by 1000). A
baghouse operating at an A/C ratio of 2 ft/min has the same collection area as an ESP with an
-------
SCA of 500. Both of these are typical of the size of collectors employed for new power plants.
Therefore, an AHPC operating at an A/C ratio of 12 ft/min and an SCA of 83 would offer an
83% reduction in fabric area over a conventional baghouse operating at 2 ft/min and an 83%
reduction in plate area over a conventional ESP with an SCA of 500. The combined collection
area in the AHPC would be 67% lower than either the conventional baghouse or the ESP. These
key features of the AHPC are shown in Figure 3.
Scope of Work and Results
Phase I of the development effort included design and construction of a 200-acfm working
AHPC model. The first experimental tests were cold-flow tests with air for the purpose of
Ultraclean Flue Gas
GORE-TEX" Filter Bags
A/C-12
GORE-TEX' bags
provide ultrahigh
collection efficiency.
SCA-90
Greater than 90%
collection
Agglomerated Ash
Falls to Hopper
EERC 11223SM.CDR
Figure 3
Key Features of AHPC
-------
adjusting the bag-cleaning parameters to achieve the best interaction between the ESP and
filtration zones. Reentrained dust (fly ash) was injected into the carrier air upstream of the
AHPC, operating at an A/C ratio of 12 ft/min.
After successful completions of the cold-flow tests, 8-hour verification tests where the AHPC
was required to collect fly ash from real flue gas produced from coal combustion were
completed. The fractional collection efficiency and system pressure drop were determined as a
function of the main independent variables, coal type and A/C ratio. These tests were followed
by 100-hr tests to evaluate the longer-term operability of the AHPC over multiple cleaning
cycles. Extensive inlet and outlet particulate measurements were completed to thoroughly
document the performance of the AHPC as a function of time.
The objectives of the 8-hr tests on coal were to evaluate:
The AHPC under real flue gas conditions firing Absaloka subbituminous coal.
AHPC operability with the ESP alone and all bags removed
On-line versus off-line cleaning
A/C, 3.66 and 4.8 m/min (12 and 16 ft/min)
Operational variables for the Absaloka-fired tests (conducted on the EERC particulate test
combustor [PTC]) are found in Table 1. With the ESP on and bags removed, the particulate
emissions ranged between 0.2840 to 0.3728 g/m3 (0.1240 to 0.1628 gr/scf). The AHPC efficiency
with the ESP on without bags was about 95%. This result was encouraging because 90%-95%
efficiency for the ESP was the basis for the concept.
Aerodynamic particle sizer (APS) respirable mass data, presented in Figure 4, show the AHPC
outlet particulate emission when the ESP was on and when the ESP was off. With the ESP on,
the APS showed an average outlet emission of about 220 mg/m3. The average respirable mass
emission when the ESP was off was approximately 1300 mg/m3. The ESP by itself achieved 83%
collection efficiency of respirable mass compared to 95% total mass. However, this result is
encouraging because it shows that the ESP removes a substantial portion (83%) of the fine-
particle mass.
Table 1
Test Parameters for Absaloka Coal, PTC Test
PTC-A8-574 PTC-AB-575 ... FTC-AB-576... PTC-AB-577 PTC-AB-57S
Air/Cloth, m/min >iA
Inlet Temp., ฐC 149 ,
On-Line and Off-Line Cleaning NA
No. of Bags in Use $
3.7
149
Off
4
3.7
149
oป
4
4.8
149
Off
3
4.X
149
On
3
In Test PTC-AB-576, the cleaning mode of the AHPC was set in the on-line mode. The dP
versus time graph is shown in Figure 5. The change in dP before and after cleaning was at 1.1 to
1.2 kPa (4.5 to 5.0 in. W.C.), with pulse intervals averaging around 70 min (see Figure 6), similar
to the results with off-line cleaning in Test PTC-AB-575.
Since Runs 575 and 576 both demonstrated excellent pressure drop control and at least a 70-min
bag-cleaning interval, the decision was made to increase the A/C ratio from 3.7 m/min
(12 ft/min) to 4.9 m/min (16 ft/min). For the first test (AB-577) at 4.9 m/min (16 ft/min),
cleaning was off-line and for the next test (AB-578) cleaning was on-line. Similar to the off-line
-------
IE
D)
co"
CO
CO
CD
-Q
cfl
"5.
CO
CD
cc
1000 -
900
800
700
600
500
400
300
200
PTC-AB-574
I ESP Off
f %
H *^ *^
: | ESP Only j
10 11 12
Time, hr
Figure 4
APS Data for Test PTC-AB-574
13
14
2.5
2.0
CglC GST 14004. COB
1.0
0.5
o.o_''-
0 50 100 150 200 250 300 350 400 450 500 550
Time, min
Figure 5
Pressure Drop as a Function of Time for Test PTV-AB-576 with On-Line Cleaning
-------
100
80
4-ซ
c
O.
60
40
20
EERCSS1386S.CDR
I
I
I
50 100 150 200 250 300 350 400 450 500 550
Time, min
Figure 6
Pulse Interval as a Function of Time for Test PTC-AB-576 with On-line Cleaning
cleaning mode, changes in the dP before and after the cleaning cycle ranged from 0.6 to 0.5 kPa
(2.5 to 2.0 in. W.C.) at steady state, shown in Figure 7. Pulse intervals were about 10-12 min,
shown in Figure 8.
The 4.8 m/min (16 ft/min) tests were successful because the pressure drop was readily controlled
with a pulse-cleaning interval of 10-12 min. However, this is a smaller interval than would be
anticipated based on the sole effect of increasing A/C ratio. The theoretical increase in dP as a
function of time is proportional to the square of the face velocity, so at 4.8 m/min (16 ft/min), the
theoretical cleaning interval should be 0.56 times the interval at 3.7 m/min (12 ft/min). Based on
the 60-70-min pulse interval observed at 3.7 m/min (12 ft/min), the theoretical interval is about
35 min compared to the 10-12-min interval observed. This indicates additional nonideal effects
occurred. This may have been caused by the nonideal geometry of the AHPC for the higher A/C
ratio tests.
Figure 9 plots the inlet and outlet particulate particle-size distributions for Test PTC-AB-573.
The inlet concentration is a combination of data from the scanning mobility particle sizer
(SMPS), multicyclone sampling, and Coulter counter analysis of the first cyclone catch.
While the multicyclone collects all of the inlet ash, it divides it into only six fractions: the five
cyclones and a backup filter. The cut point of the first cyclone is typically about 7 urn (depending
on temperature and sampling flow rate), and the cut point of the last cyclone is typically about
0.7 um. To obtain more information about the entire particle-size distribution requires further
resolution of the first cyclone catch (which usually contains about 80% of the total mass) and the
particles collected on the backup filter. A convenient method to expand the resolution of the first
cyclone catch is to perform a Coulter counter analysis on it and to combine the data. Since no
instrument is available to accurately determine the distribution of the multicyclone filter catch,
the submicron distribution is most conveniently determined by direct sampling of the flue gas
-------
EEPCGS13981.CDR
0 50 100 150 200 250 300 350 400 450 500
Time, min
Figure 7
Pressure Drop as a Function of Time for Test PTC-AB-578 with On-Line Cleaning at 16 ft/min
50
40
D.
20
10
0
EEfiC GSt3867.CDFI
0 50 100 150 200 250 300 350 400 450 500
Time, min
Figure 8
Pulse Interval as a Function of Time for Test PTC-AB-578 with On-Line Cleaning at 16 ft/min
10
-------
10000n
1000-
100-
10-
1-
0.1-
0.01-
0.001 -
0.0001 -
0.00001 n
0.000001
EEKO BSI3SSS.CDK
0.01 0.1 1 10
Particle Aerodynamic Diameter,
100
Figure 9
Inlet and Outlet Particle-Size Distributions for Test PTC-AB-573
with the SMPS. At the outlet, the dust concentration is too low to use mutlicyclones or the
Coulter counter, so the distribution is obtained by combining data from the SMPS and an APS. A
number of assumptions and conversions were made to combine the data into single particle-size
distributions. First, the particle sizes for the SMPS and Coulter data were converted to
aerodynamic diameters, which is a function of the geometric particle diameter and the particle
density. It was assumed that the SMPS diameter was roughly equal to the geometric diameter,
and the density for all particles was 2.5 g/cm3. Once the particle size was represented in terms of
aerodynamic diameter, the mass concentration (mg/m3) for each channel (size) was calculated.
The Coulter counter data represent the first cyclone catch of the multicyclone, which is the upper
end of the particle-size distribution. The mass concentration was calculated based on the
percentage of total mass in each channel, the total mass in the cyclone catch, and the volume of
gas sampled. The SMPS mass concentration was calculated from the particle geometric diameter,
assumed density, and the number concentration for each channel. The mass distribution for the
multicyclone sample was calculated directly from the data. Once the mass concentration was
calculated, it was plotted (on a dM/d log [Dp] basis) as a function of particle size. By comparing
the inlet and outlet concentrations for a given particle size, the fractional efficiency for the entire
range from 0.02 to 50 urn is obtained. From Figure 9, the collection efficiency was greater than
99.99% over the entire particle-size range.
The objectives of the 100-hr tests on coal were to:
Determine operability of AHPC for an extended period at steady-state conditions.
Determine the fate of seven trace metals using the AHPC system.
Evaluate the effect of carbon injection on trace elements and system operability.
The AHPC was operated at an A/C of 3.7 m/min (12 ft/min) with on-line cleaning. The
temperature of the AHPC was maintained at 147ฐ C (300ฐF) throughout the baseline test. EPA
Method 29 multimetals sampling train was used to determine trace metal concentrations for As,
11
-------
Cd, Cr, Hg, Ni, Pb, and Se as well as the particulate loading of the gas stream. Sampling of the
inlet and outlet flue gas from the AHPC was performed simultaneously. Time duration for
Method 29 inlet sampling was 1 hour while the time duration for the outlet sampling was 4 hr. In
addition, 24-hr outlet dust-loading samples were taken.
The purpose of the sorbent injection test was to evaluate the AHPC performance while injecting
activated carbons for mercury control. The AHPC operating temperature was lowered from
149ฐC (300ฐF) to 135ฐC (275ฐF) to improve the absorption characteristics of the sorbents. Two
sorbents were used for this test. A lignite activated char (LAC) and an iodated impregnated
activated char (I AC) were mixed in a ratio of 4:1 LAC to I AC, respectively. The sorbent addition
rate was adjusted to achieve a sorbent-to-Hg ratio of 3000:1.
Figures 10 and 11 show the pulse interval for the last day of the 100-hr tests. The interval was
quite steady at about 25 to 35 min at the end of the tests. This shows that in extended operation
the AHPC pressure drop was well controlled and that the injection of a carbon-based sorbent did
not adversely affect the pressure drop.
The 24-hr dust-loading results showed that the total mass collection efficiency was greater than
99.999%. Even after sampling for 24 hr, the dust-loading filters looked clean. Respirable mass
data also show that an ultrahigh collection efficiency was achieved. Figure 12 is an example of
the data showing slight spikes that occur after bag cleaning, but most of the data are around
0.001 mg/m3. When the sampling included the short-term cleaning spikes (see Figure 12), the
integrated average values were still all below 0.01 mg/m3. Since the inlet respirable mass was
typically about 1000 mg/m3, this corresponds to a fine-particle collection efficiency greater than
99.999%.
The trace element measurement results are shown in Table 2. At the outlet, only mercury,
chromium, and selenium were above detection limits. The chromium, possibly, was the result of
50
c
"E
"cc
o> 25
_c
(U
tn
CL
EgffC GST4353.CDfl
3900 4100 4300 4500 4700 4900 5100 5300 5500 5700 5900
4000 4200 4400 4600 4800 5000 5200 5400 5600 5800 6000
Time, min
Figure 10
Pulse Interval for Last Day of 100-hr Baseline Test PTC-AB-585
12
-------
50
c
"E
"Jo
ฎ 25
CD
EfflC GS14354.CDR
4900 5000 5100 5200 5300 5400 5500 5600 5700 5800 5900
Time, min
Figure 11
Pulse Interval for Last Day of 100-hr Sorbent Injection Test PTC-AB-586
1 -i
fEHCGST4340.CDR
ra
JE
^5
ID
0.1 -
0.01 -
JD
|
"a.
0> 0.001 -:
DC
0.0001
PTC-AB-585-3
A/C Ratio = 12 ft/min
Integrated
Averages
9 10 11 12 13 14 15 16 17 18
Time, hr
Figure 12
Respirable Mass Emission for 100-hr Test
13
-------
TABLE 2
Trace Element Data
PTC-AB-585
Baseline
PTC-AB-586
Sorbent Injection
Av. Inlet
Av. Outlet
% Removal
Av. Inlet
Av. Outlet
% Removal
Hg,
ug/m3
5.5
3.6
34.5
6.8
2.7
60.3
As,
ug/m3
83
<1.0
>98.8
99
<1.0
>99
Cd, Cr,
ug/m3 ug/m3
1.7
<0.08
>95.3
1.8
<0.08
>95.6
463
0.57
99.9
534
2.2
99.6
Pb,
ug/m3
319
<0.50
>99.8
291
<0.50
>99.8
Ni,
ug/m3
223
<2.0
>99.1
229
<2.0
>99.1
Se,
Ug/m3
32
16
50
25
3.8
84.8
contamination, since stainless steel components were used in the construction of the AHPC. The
injection of sorbent resulted in a modest increase in mercury collection efficiency from 35% for
the baseline case to 60% with sorbent. The Absaloka coal is known to produce primarily
elemental mercury, which has proven in other work to be more difficult to capture with sorbents.
The fine-particle data indicate a collection efficiency greater than 99.999%, but the individual
trace element data for the nonvolatile elements indicate collection efficiencies primarily in the
range from 99% to 99.9%. Even though the actual efficiency may have been orders of magnitude
higher, the detection limits with Method 29 do not allow accurate measurement beyond about
99.9% collection efficiency.
Summary
These tests provided the following results:
Particulate collection efficiencies greater than 99.99% for all particle sizes from 0.01 to
50 um were achieved.
Pressure drop was well controlled, steady, and not adversely affected by injection of
carbon. Time interval between bag-cleaning cycles ranged from 25 to 35 min at the end
of the 100-hr tests.
Emissions of arsenic, cadmium, lead, and nickel were below detection limits. Mercury
and selenium were detected hi measurable quantities in vapor form at the outlet.
Chromium was detected at the outlet, but may have been the result of contamination.
No increased participate emissions were noted during mercury sorbent injection.
Plans for Phase II are to scale up the AHPC to the 5000-10,000-acfm pilot size and test on a
slipstream of a full-scale boiler.
Specific anticipated benefits of this approach are as follows:
Solves the problem of excessive fine-particle emissions with conventional ESPs.
Provides ultrahigh collection efficiency, even at high A/C ratios.
14
-------
Solves the problem of reentrainment and re-collection of dust in conventional pulse-jet
baghouses caused by the close bag spacing and the effect of cleaning one row of bags at
a time.
Solves the problem of chemical attack on bags, allowing application to all U.S. coals.
Requires significantly less total collection area than conventional ESPs or baghouses.
Is suitable for new installations or as a retrofit replacement technology.
Acknowledgment
The EERC wishes to acknowledge that the work described in this paper was funded by the U.S.
Department of Energy, Federal Energy Technology Center (FETC). Thank you to the entire
project team for their respective contributions-Dr. Perry Bergman at FETC for serving as the
DOE COR, Mr. Ken Walker at W.L. Gore and Associates as a technical and financial partner,
and Dr. Henry Krigmont at Allied Environmental Technologies Company for modeling and
technical review.
References
1. Dennis, R. et al. "Filtration Model for Coal Fly Ash with Glass Fabrics," EPA-600/7-77-084,
Aug. 1977.
2. Leith, D.; Rudnick, S.N.; First, M.W. "High-Velocity, High Efficiency Aerosol Filtration,"
EPA-600/2-76-020, Jan. 1976.
3. Oglesby, S.; Nichols, G.B. Electrostatic Precipitation; Marcel Dekker, Inc.: New York, 1978.
15
-------
Thursday, August 28; 8:00 a.m.
Parallel Session B:
Air Toxics Control - Mercury Capture by Sorbents: Lab Scale Research
-------
COMBINED MERCURY AND SULFUR OXIDES CONTROL USING
CALCIUM-BASED SORBENTS
S. Behrooz Ghorishi
Acurex Environmental Corporation
4915 Prospectus Drive
Durham, NC 27713
and
Charles B. Sedraan
National Risk Management Research Laboratory
Air Pollution Prevention and Control Division
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Abstract
The capture of elemental mercury (Hgฐ) and mercuric chloride (HgCl2) by three types of calcium
(Ca)-based sorbents was examined in this bench-scale study under conditions prevalent in coal-fired
utilities. Ca-based sorbent performances were compared to that of an activated carbon. Mercury
capture of about 40% (nearly half that of the activated carbon) was achieved by two of the Ca-based
sorbents. The presence of sulfur dioxide (SOj) in the simulated coal combustion flue gas enhanced
the capture of Hgฐ from about 10 to 40%. Increasing the temperature in the range of 65-100ฐC also
caused an increase in the Hgฐ capture by the two Ca-based sorbents. Mercuric chloride (HgCl2)
capture exhibited a totally different pattern. The presence of SO2 inhibited the HgCI2 capture by Ca-
based sorbents from about 25 to less than 10%. Increasing the temperature in the studied range also
caused a decrease in HgCl2 capture. Upon further pilot-scale confirmations, the results obtained in
this bench-scale study can be used to design and manufacture more cost-effective mercury sorbents
to replace conventional sorbents already in use in mercury control.
-------
Introduction
Title m of the 1990 Clean Air Act Amendments (CAAA) requires the U.S. Environmental Protection
Agency (EPA) to submit a study on 189 hazardous air pollutants (HAPs). This study would include
emissions and a risk (to public health) assessment of the 189 HAPs. Among these compounds,
mercury has drawn special attention due to its increased levels in the environment and the well
documented food chain transport and bioaccumulation of this specie and its compounds such as
methyl mercury.1>2 An EPA report to Congress cites the largest emitters of mercury as coal-fired
utilities, medical waste incinerators (MWIs), municipal waste combustors (MWCs), chlor-alkali
plants, copper and lead smelters, and cement manufacturers.3 These sources are estimated to account
for over 90% of all anthropogenic mercury emissions. Utility boilers account for nearly 25% of the
total anthropogenic emissions, of which more than 90% are attributed to coal-fired utility boilers.
Mercury, a trace constituent of coal4, is readily volatilized during coal combustion.5 Mercury is
the most volatile trace metal, and major portions of it can pass through existing paniculate matter
(PM) control devices.5 A sorbent reacting with this metallic species can effectively convert the
vapor to a sorbed liquid or solid phase, facilitating its removal with sorbent particles in a PM
control device. Mercury control processes which use adsorption on dry sorbents do not pose the
problem of the treatment and stabilization of a waste liquid stream and, therefore, seem very
attractive for coal combustors.
Several methods of controlling mercury emissions are in either commercial use or development for
MWCs and MWIs.6 Dry sorbent injection (DSI) of activated carbon, followed by fabric filtration (FF)
has shown consistently high (>90%) mercury removal in MWC applications. Spray drying (SD)
followed by FF, and wet scrubbing (WS) have both been successfully applied for acid gas control,
and have been found to remove substantial (60-90%) amounts of mercury in MWCs. However, all
three technologies have been less successful in removing mercury from coal-fired flue gases.7
There are primarily three reasons suspected for the observed differences in mercury capture between
-------
MWC and coal-fired cases: (a) the differences in the mercury components (species) present in the two
flue gases, (b) mercury species concentrations, and (c) composition of the two flue gases. On account
of the larger concentrations of hydrogen chloride (HC1) present in a MWC flue gas, mercury is
thought to exist primarily as mercuric chloride (HgClj).8 Recent pilot plant studies on coal-fired flue
gas indicate that for some Ohio coals, a considerable portion of mercury vapor may be HgCl2.
However, the same study indicated that elemental mercury (Hgฐ) vapor concentration may actually
increase across a wet limestone scrubber, presumably due to the reduction of HgCl2 vapor entering
the scrubber.9 The lower concentration of HC1 in a coal-fired flue gas is believed responsible for a
portion of the mercury to exist as Hgฐ.
Another difference in the two types of flue gases is their total mercury concentrations. The total
mercury concentration in a MWC flue gas is typically several orders of magnitude higher than the
mercury concentration in a coal-fired flue gas. The typical mercury concentration observed in coal
combustion flue gas (2-3 ppb)10 was simulated throughout this study. The third difference between
MWC and coal-fired systems is the composition of the flue gases. Sulfur dioxide (SOj) is present at
higher concentration in coal combustion flue gases and is believed to influence the capture of mercury
by different sorbents and emission control devices. The effect of SO2 on mercury capture was
investigated in this study.
Pilot-scale studies have shown that, to achieve high removals of mercury in coal-fired power plants,
activated carbon to mercury (by weight) ratios of around 3000/1 were required.11'12 At an activated
carbon cost of $1.125 /kg, the material cost would be approximately $500,000 per year for a 500
MW power plant. Chang et al.11 arrived at an annual cost of $100,000 to $1 million for mercury
control in a 500 MW power plant. A recent study by Chang and Offen7 estimates that removing 50%
of the mercury emitted in flue gas by U.S. power plants could range from $1 billion to $10 billion per
year. Therefore, bench-scale efforts to study process parameters and sorbent types for mercury
control hi coal-fired flue gas are needed to develop effective and economic mercury capture
technology. In addition, improvement of mercury control using existing technologies for SO2 and fine
PM control would appear to be prudent. Therefore this study focuses on improving the existing SO2
-------
control sorbents for a potentially combined mercury and SO2 control.
Bench-scale results from laboratories at the Air Pollution Prevention and Control Division (APPCD)
of the U.S. Environmental Protection Agency (EPA) showed that Ca-based sorbents were effective
in controlling HgCl2 under MWC operating conditions (in the absence of SOj)13. It was found that
calcium oxide (CaO) and calcium hydroxide [Ca^Hy were effective in capturing HgCl2 at 100ฐC.
At 140ฐC, however, the Ca-based sorbents were found to be less efficient in capturing HgCl2. Also,
during the Hgฐ capture experiments, only activated carbons exhibited significant capture at both 100
and 140ฐC in the absence of SO2. Pilot-scale tests showed that injection of Ca-based sorbents into
a furnace reduced total mercury emissions at the outlet of the furnace.14 Stouffer et al. have shown
that, in an air toxics control pilot plant, high system HgCl2 removal can be achieved with the injection
of hydrated lime as the sorbent.15 At 93ฐC, removals of HgCl2 from the gas were about 55 and 85%
at Ca/Hg weight ratios of 5,000 and 100,000, respectively. The corresponding Hgฐ removals ranged
only from 10 to 20%, even at Ca/Hg weight ratios as high as 300,000.
Considering the above observations, a potential method of cost reduction in controlling mercury
emissions in coal-fired utilities (low mercury concentration) would be to utilize the cheaper Ca-based
sorbents. This paper reports results of experiments to study Hgฐ and HgCl2 capture by several Ca-
based sorbents and their performance compared with a lignite-coal-based activated carbon
(DARCOฎ FGD, Norit Americas Inc.). Hgฐ and HgCl2 concentrations were roughly 2 to 3 ppb in
a simulated flue gas in order to replicate conditions (as close as possible) prevalent in a coal-fired flue
gas.10 Among the Ca-based sorbents evaluated in this study were reagent grade hydrated lime
(calcium hydroxide), a mixture of fly ash and hydrated lime (advanced silicate -- Advacate), and a
modified Advacate. More details on the sorbents tested in this study are given in the next section.
Capture of Hgฐ and HgCl2 by these sorbents was studied as a function of system temperature and SO2
present in the simulated flue gas.
-------
Sorbents
The capture of Hgฐ and HgQ2 by five types of sorbents was studied in this investigation. Of the five,
the three Ca-based sorbents were hydrated lime, Advacate, and a modified Advacate. Hgฐ and HgCl2
capture by Clinch River Fly Ash (CRFA) and an activated carbon (FGD) was also measured for
comparison. Preparations of Advacate and modified Advacate are discussed below.
Preparation of Advacate and Modified Advacate
Advacate was prepared in a pressure hydrator at 150ฐC by mixing a 3/1 ratio (by weight) of CRFA
to hydrated lime. The modified Advacate was prepared by addition of a chemical agent during this
process. The entire reaction time in the pressure hydrator was 1 h. After preparation of the sorbents,
they were placed in a vacuum oven at 165ฐC for 24 h before use. Several batches of Advacate and
modified Advacate were prepared, and their physical characteristics were studied using nitrogen (N2)
sorption. Very similar physical characteristics were obtained for different batches of each, indicating
their reproducibility. The following subsection describes, in detail, the structural properties and
chemical compositions of the studied sorbents.
Structural Properties/Chemical Compositions of the Sorbents
Information about the internal pore structure (total and incremental volume and surface area) of the
three Ca-based sorbents was determined by a Micromeritics ASAP 2600 using N2
adsorption/desorption with a Bmnauer-Emmett-Teller (BET) method. BET analyses on the three
Ca-based sorbents (hydrated lime, Advacate, and modified Advacate) obtained from N2 sorption are
shown in Figure 1. Of the three, hydrated lime had the lowest internal pore volume and surface area.
Advacate, containing only 25% of hydrated lime by weight, had a higher surface area than hydrated
lime. Modified Advacate had the highest surface area among the three. A bimodal pore size
distribution was seen for the three Ca-based sorbents with most of the pore diameters being
-------
approximately 15 to 50 nm. The total pore volume of modified Advacate was over five times that for
hydrated lime and three times the pore volume of Advacate. The internal pore structure of the studied
activated carbon (FGD) is shown in Figure 2. Bimodal pore size distribution was not observed in the
case of FGD. Unlike Ca-based sorbents, pores with diameters less than 5 nm were the significant
contributors to the total pore area and volume in FGD, causing the average pore diameter in FGD
to be considerably lower than that of Ca-based sorbents. Structural properties and chemical
composition of the studied sorbents are summarized as:
Hydrated lime: The Hydrated lime used in this research was reagent grade (Sigma Inc.)
containing 97.6% Ca(OH)2 and 1.8% calcium carbonate (CaCO3). This material has a total surface
area of 13.0 m2/g and an average pore diameter of 33.4 nm.
Clinch River Fly Ash (CRFA): This material is a fly ash obtained from the Clinch River
Virginia power plant. This power plant uses a local bituminous coal. The mineral content of this fly
ash is: 5.2% CaO, 51.6% SiO2, 24.7% A1203, 0.5% Na2O, 1.8% MgO, 3.3% K2O, 7.8% Fe20, and
1.4% TiO2. CRFA has a total surface area of 2.3 m2/g and an average pore diameter of 8.1 nm.
Advacate. A reaction product of a 3/1 mixture of CRFA and hydrated lime. As prepared for
this study, it had a total surface area of 30.9 mVg and an average pore diameter of 21.2 nm.
Modified Advacate: Advacate prepared with an additional chemical agent. Modified Advacate
for this study had a total surface area of 91.4 m2/g and an average pore diameter of 22.2 nm.
FGD. A trademark for an activated carbon known as "DARCOฎ FGD" manufactured by
Norit Americas Inc. FGD is a lignite-coal-based activated carbon manufactured specifically for the
removal of heavy metals. It has a total surface area of 575 m2/g and an average pore diameter of 3.2
nm. More information about physical characteristics and mercury capture performance of FGD can
be found elsewhere.16 Table 1 summarizes the physical properties of the studied sorbents.
Experimental Apparatus and Procedures
Figure 3 is a schematic of the experimental apparatus used to study capture of Hgฐ and HgCl2. Pure
HgCl2 powder in a diffusion vial was the source of HgCl2 vapor, and pure Hgฐ liquid in a permeation
tube was the source of Hgฐ vapor. The relative concentration of HgCL, or Hgฐ vapor in the gas
-------
stream was varied by adjusting the water bath temperature. The generated HgCl2 or Hgฐ vapor was
carried into the main system by a nitrogen (N^) stream where it was mixed with water vapor (H2O),
air, sulfur dioxide (SOj), and carbon dioxide (CO2) in the manifold. The composition of the simulated
flue gas and the total system flow rate was kept constant throughout these studies as follows: 2-3 ppb
HgCl2 or Hgฐ, 5% HjO, 7% Q,, 10% CQ, 173 ppm SQ,, and balance of nitrogen, total system flow:
300 cmVmin
A 3-way valve placed before the manifold (Figure 3) diverted the Hgฐ or HgCl2 in the N2 stream away
from the manifold when desired. The first 3-way valve placed after the manifold was used to direct
flow to or away from the fixed-bed reactor. The sorbent to be tested (approximately 0.1 g) was
placed in the constant temperature reactor. A furnace kept at 850ฐC was added downstream of the
reactor to convert any oxidized mercury vapor to Hgฐ. According to thermodynamic predictions, the
only Hg specie at this temperature is Hgฐ.13 The presence of the furnace enabled detection of non-
adsorbed HgCl2 as Hgฐ by the on-line ultraviolet (UV) Hgฐ analyzer, thus providing actual,
continuous Hgฐ or HgCl2 capture data by the packed bed of sorbent. Prior to the mercury analyzer,
an ice bath served as a water trap. Quality control experiments had previously indicated no loss of
Hgฐ or SO2 in the water trap.
It should be noted that the Hgฐ research apparatus is made of Teflon The HgCl2 apparatus is made
of quartz and avoids the use of Teflon, which is known to adsorb HgQ2.
The UV Hgฐ analyzer responded to SO2 concentrations as well as to Hgฐ. For instance, a gas stream
consisting of 173 ppm SO2 and 3 ppb Hgฐ produced a SO2/Hgฐ signal ratio of 8/18. Contributions
from S02 were accounted for by placing a SO2 analyzer (UV) on-line downstream of the Hgฐ
analyzer. The SO2 analyzer was incapable of responding to mercury in the concentration range used
in this study. By subtracting the SO2 signal measured by the SO2 analyzer from the total response of
the mercury analyzer, the mercury concentration was obtained.
In addition to sorbents, other parameters studied in this investigation were packed bed temperature
-------
and the presence of SO2. The two studied temperatures were 100 and 65ฐC. These temperatures are
typically observed in the air pollution control system of coal-fired utility boilers. The effect of SO2
was studied by performing two sets of packed bed experiments in the absence of SO2 and in the
presence of 173 ppm SO2. This relatively low level of SO2 was selected to minimize UV interference
with Hgฐ detection, and to be consistent with the previous activated carbon sorption experiments.
.For the same reasons, UV monitor interference and consistency with activated carbon data, the
experiments reported here have been limited to simulated flue gas with no nitrogen oxides, HC1, and
only 5% moisture. The effects of temperature and SO2 on Hgฐ and HgCl2 capture by different
sorbents were studied independently. In each test, the packed bed was exposed to the simulated flue
gas for 30 minutes during which the exit concentration of mercury was continuously monitored. The
percent removal of Hgฐ or HgCl2 was obtained according to: percent removal= 100 ^mercury-
mercuryout)/mercuryjn. It should be noted that each set of parameters was run in duplicate. If the
duplicates did not meet the precision goal (the data quality indicator) of ฑ10%, the parameters were
tested a third time.
Results and Discussions
Capture of Elemental Mercury
Figure 4 shows the effect of SO2 on Hgฐ capture performance of the Ca-based sorbents as compared
to the activated carbon (FGD) and CRFA at 100ฐC. Removals presented in Figure 4 (and Figure 7)
are obtained by averaging the removal data acquired during the exposure period (30 minutes). Of the
five, FGD showed the highest capture of Hgฐ during the 30 minutes of exposure (constant during this
period). Both CRFA and hydrated lime exhibited insignificant capture of Hgฐ (approximately 5%).
Among the Ca-based sorbents, it is seen that Hgฐ capture increases as the total surface area and
cumulative pore volume increases (Figure 1 and Table 1). The presence of SO2 significantly increased
the capture of Ca-based sorbents, especially Advacate and modified Advacate. The insignificant Hgฐ
capture by the Ca-based sorbents in the absence of SO2 indicated the lack of any interaction (physical
or chemical) between the Hgฐ and the Ca-based sorbents. The enhancement effect of SO2 at 100ฐC
-------
may indicate that the reaction of SO2 and sorbents created active sulfur (S) sites for the adsorption
of Hgฐ, possibly through formation of Hg-S bonds (chemisorption). Conversely, the lack of
significant improvement in Hgฐ capture for hydrated lime with SO2 present (Figure 4) indicates the
need for a fine pore structure as well as SO2. If indeed, the major Hgฐ capture mechanism is
chenrisorption by SO2-generated active sites, then decreasing the system temperature should decrease
the overall rate of "active site generation and chemisorption" leading to a decrease in Hgฐ capture.
The effect of temperature (65 vs 100ฐC) on Hgฐ capture by Advacate and modified Advacate in the
presence of SO2 is illustrated in Figure 5. This figure shows Hgฐ capture throughout the 30 minutes
of exposure. The observed higher captures at higher temperature support the chemisorption theory
of Hgฐ capture by the Ca-based sorbents in the presence of SO2.
The capture of SO2 by the five sorbents during the Hgฐ tests was also monitored at 100ฐC (Figure 6).
All three Ca-based sorbents showed higher captures of SO2 than activated carbon (FGD), which was
expected because of their alkaline nature. After approximately 10 minutes of exposure to the
simulated flue gas, the SO2 reaction rate (change of percent removal with time) showed diminishing
removal with increasing time. One explanation is that the reaction of SO2 with Ca-based sorbents
may lead to pore mouth closure, thus blocking the access of SO2 to the interior of the Ca-based
sorbents. This would occur within the first 10 minutes of exposure of sorbent to flue gas. One may
speculate that since all three Ca-based sorbents had the same average pore diameter (20-30 nm), they
should exhibit the same monotonically decreasing SO2 capture pattern.
In summary, Hgฐ can be captured by previously reacted mixtures of fly ash and hydrated lime
(Advacate and modified Advacate) when SO2 is present in the flue gas. Based on this observation,
one may conclude that, in terms of Hgฐ control, the optimum region for injection of Ca-based
sorbents is upstream of SO2 control systems in which a higher concentration of SO2 is present, and
flue gas temperatures are higher. In this way, both SO2 and Hgฐ emissions may be controlled for
approximately the cost of SO2 control by sorbent injection alone. Modifying sorbents to increase the
total surface area and fine pore structure increases Hgฐ uptake in the presence of SO2 for the sorbents
-------
studied.
Capture of Mercuric Chloride
Figure 7 depicts the effect of SO2 on HgCl2 capture performance of the three Ca-based sorbents as
compared to the activated carbon (FGD) and Clinch River Fly Ash (CRFA) at 100ฐC. Similar to Hgฐ,
FGD captured the highest fraction of incoming HgCl2 (constant removal during the exposure period),
with the three Ca-based sorbents and CRFA showing from 10 to 20% HgCL, capture in the absence
of SO2. Unlike the Hgฐ case, the presence of SO2 inhibited the HgCl2 capture by Advacate and
modified Advacate, indicating that perhaps HgCl, is not attracted to the sites preferred by Hgฐ, and
has affinity for SO2 capture sites. The SO2 inhibition effect may also confirm the earlier conclusion
that the presence of SO2 caused a blockage of pores in Advacate and modified Advacate, and
therefore limited the access of HgCl2 to the interior structure of the sorbents. One may also attribute
the S02 inhibition effect to the competition of SO2 with HgCl2 (both acid gases) for the alkaline sites
located inside the pores or on the external surface of the sorbent.
Figure 8 illustrates the effect of temperature on HgCl2 capture by Advacate and modified Advacate
in the absence of SO2 (optimum condition), throughout the 30 minutes of exposure. Unlike the Hgฐ
case, decreasing the temperature caused an increase in HgCl2 capture by these sorbents. The effect
of temperature may be explained by a physisorption mechanism through which the HgCl2 molecules
are adsorbed by the sites.
An interesting observation can be made by comparing Figure 8 (effect of temperature on HgCl2
capture) to Figure 5 (effect of temperature on Hgฐ capture). Unlike Hgฐ, HgCl2 capture increased
(with time) at the lowest studied temperature (65ฐC). The reason may be outlined as follows.
At lower temperatures, water vapor present in the simulated flue gas may condense on the surface
of Advacate and modified Advacate. It should be noted that the homogeneous dew point of 5% water
vapor in air is below 65ฐC, but that the actual dew point above hygroscopic solids (such as calcium
10
-------
silicates) can be significantly higher, favoring condensation of water vapor at the solid surface. If
water vapor molecules were to condense on the sorbent sites, they could readily dissolve the
incoming HgCl2 molecules but not the insoluble Hgฐ molecules. As the time of exposure progresses,
an increasing number of water vapor molecules condense, thus the capture percentage of HgCl2
increases. This dissolution effect, yet to be proven for these sorbents, may be very important in
practical situations where the concentration of water vapor is likely higher than for these bench-scale
simulations.
Conclusion
The capture of elemental mercury (Hgฐ) and mercuric chloride (HgClj), the mercury species
identified in coal flue gas, by three types of calcium-based sorbents differing in their internal structure,
was examined in a packed-bed, bench-scale study under simulated flue gas conditions for coal-fired
utilities. The results obtained were compared with Hgฐ and HgCl2 capture by an activated carbon
(FGD) under identical conditions. Tests were conducted with and without SO2 to evaluate the effect
of SO2 on Hgฐ and HgCl2 control by each of the sorbents.
The Ca-based sorbents showed insignificant removal of Hgฐ in the absence of SO2. However, in the
presence of SO2, Hgฐ capture was enhanced for the three Ca-based sorbents. It was postulated that
the reaction of hydrated lime with SO2 would result in pore mouth closure as evidenced by the sharp
drop in the SO2 removal rate after the initial 10 minutes of exposure. Despite the loss of internal
surface area, the relatively high uptake of Hgฐ observed for these sorbents in the presence of SO2,
suggests that Hgฐ and SO2 do not compete for the same active sites, and the sites for Hgฐ capture are
influenced positively by the presence of SO2. Moreover, the capture of Hgฐ in the presence of SO2
increased with sorbent surface area and internal pore structure.
Conversely, the three Ca-based sorbents showed decreased removal of HgCl2 in the presence of SO2.
In the absence of SO2, roughly 25% of the incoming HgCl2 was captured. The alkaline sites in the
Ca-based sorbents were postulated to be instrumental hi the capture of acidic HgCl2. SO2 not only
11
-------
competed for these alkaline sites but also, as mentioned, likely closed pores with subsequent
reduction in accessabiliry of the interior of the Ca-based sorbent particles to the HgCl2 molecules.
It was hypothesized that the capture of Hgฐ in the presence of SO2 may occur through a
chemisorption mechanism, while the nature of the adsorption of HgCl2 molecules may be explained
through a physisorption mechanism. The effect of temperature studies further confirmed this
hypothesis. Increasing the system temperature caused an increase in Hgฐ uptake by the sorbents in
the presence of SO2. However, the increase in temperature resulted in a significant decrease in the
HgCl2 uptake in the absence or presence of SO2. Increased sorbent surface area and internal pore
structure had no observable effect on HgCl2 capture in the presence of SO2.
With the relatively large quantities of Ca needed for SO2 control at coal-fired boilers, the above
results suggest that Ca-based sorbents, modified by reaction with fly ash, can be used to control total
mercury emissions and SO2 cost-effectively. The most effective Ca-based sorbents are those with
significant surface area (for SO2 and HgCl2 capture) and pore volume (for Hgฐ capture).
Sorbents injected upstream of a fabric filter should perform as indicated by the fixed-bed reactor
simulation in this study. Confirmation of these results on a 50 cfm (0.024 m/s) pilot plant is
anticipated later this year.
Acknowledgements
The authors would like to acknowledge the logistical support of Richard E. Valentine (EPA/APPCD),
experimental assistance from Lisa Adams and Hamid Bakhteyar and sorbent development assistance
from Wojciech Jozewicz and Carl Singer (Acurex Environmental Corporation).
References
1 D.G. Langley. "Mercury Methylation in an Aquatic Environment," J. Water Pollut. Control
Fed., 45: 44-51, (1973).
12
-------
2. G. Westoo. "Methyl Mercury a Percentage of Total Mercury in Flesh and Viscera of Salmon
and Sea Trout of Various Ages," Science, 181: 567-568, (1973).
3. U.S. Environmental Protection Agency, Mercury Study Report to Congress, Book 1 of 2,
External Review Draft, EPA/600/P-94/002a (NTIS PB95-167334), Environmental Criteria
and Assessment Office, Cincinnati, OH, January 1995.
4. C.E. Billings; A.M. Sacco; W.R. Matson; R.M. Griffin; W.R Coniglio; and R.A.Harley.
"Mercury Balance on a Large Pulverized Coal-fired Furnace," J. Air Pollut. Contr. Assoc.,
23:9, 773, (1973).
5. D.H. Klein; A.W.Andren; J.A. Carter, J.F. Emery; C. Feldman;W. Fulkerson; W.S. Lyon,
J.C. Ogle; Y. Talmi; R.I. Van Hook, and N.Bohon. "Pathways of 37 Trace Elements Through
Coal-Fired Power Plant," Environ. Sci. & TechnoL, 9:10, 973, (1975).
6. T.G. Bma; and J.D. Kilgroe. "The Impact of Particulate Emissions Control on the Control
of Other MWC Air Emissions," J. Air & Waste Mgt. Assoc., 40(9): 1324 (1990).
7. R. Chang and G.R. Offen. "Mercury Emission Control Technologies: An EPRI Synopsis,"
Power Engineering^ November 1995.
8. B. Hall; O. Lindqvist, and E.Ljungstrom. "Mercury Chemistry in Simulated Flue Gases
Related to Waste Incineration Conditions," Environ. Sci. & TechnoL, 24: 108, (1990).
9. Redinger, K.E. Babcock & Wilcox to William Maxwell, Letter,U.S. EPA, dated Nov. 7,
1996, attachment p.5.
10. D.L. Laudal; M.K. Heidt; T.D. Brown, B.R. Nott, and E.P. Prestbo. "Mercury Speciation:
A Comparison Between EPA Method 29 and Other Sampling Methods," in Proceedings of
the 89th Air & Waste Management Association Annual Meeting, 96-WP64A.04, A&WMA,
Nashville, TN, (1996).
11. R. Chang; C.J. Bustard; G.Schott; T. Hunt; H.Noble, and J.Cooper. "Pilot-Scale Evaluation
of Carbon Compound Additives for the Removal of Trace Metals at Coal-Fired Utility Power
Plants," paper presented at the Second International Conference on Managing Hazardous Air
Pollutants, Washington, D.C., July 13-15, 1993.
12. S.J. Miller.; D.L. Laudal; R. Chang, and P.D. Bergman. "Laboratory Scale Investigation of
Sorbents for Mercury Control," presented at the AWMA Annual Meeting, Paper no. 142,
13
-------
Cincinnati, OH, June 20-24, 1994.
13. S.V. Krishnan; B.K. Gullet, and W. Jozewicz. "Mercury Control by Injection of Activated
Carbon and Calcium-Based Sorbents," paper presented at Solid Waste Management: Thermal
Treatment & Waste-to-Energy Technologies, U.S. EPA/AEERL & AWMA, Washington,
D.C., April 18-21, 1995.
14. B.K. Gullett; and K. Raghunathan. "The Effect of Sorbent Injection Technologies on
Emissions of Coal-Based, Metallic Air Toxics," In Proceedings of SO2 Control Symposium,
Vol. 3, p. 63-1, EPA-600/R-95-015c (NTIS PB95-179248), February 1995.
15. M.R. Stouffer; W.A Rosenhoover, and P.P. Burke. "Investigation of Flue Gas Mercury
Measurement and Control for Coal-Fired Sources," Paper 96-WP64B.06 presented at the
89th Annual Air and Waste Management Association Meeting, Nashville, TN, June 23-28,
1996.
16. S.V. Krishnan; B.K. Gullett, and W. Jozewicz. "Sorption of Elemental Mercury by Activated
Carbons," Environ. Sci. & Tech., 28:8, 1506-1512 (1994).
Table 1. Total Surface Area and Average Pore Diameter of Sorbents
Sorbent Total Surface Area (m2/g) Average Pore Diameter (nm)
HydratedLime 13.0 33.4
Advacate 30.9 21.2
Modified Advacate 91.4 22.2
Clinch River Fly Ash (CRFA) 2.3 8.1
Activated Carbon (FGD) 575 3.2
14
-------
10 20 50
Pore Diameter (nm)
10 20 50
Pore Diameter (nm)
10 20 50
Pore Diameter (nm)
10 20 SO
Pore Diameter (nm)
200
Figure 1. Internal Pore Structure Characteristics of Lime, Advacate, and Modified Advacate
Obtained from Nitrogen Sorption
15
-------
I 003
ฃ
5 10 20 an 100 200
Pore Diameter (nm)
Pore Diameter (nm)
5 10 20 50 100
Pore Diameter (nm)
Figure 2. Internal Pore Structure Characteristics of DARCO FGD Obtained from Nitrogen Sorption
16
-------
Mercury Source
Packed Bed Reactor
N.
Carbon
Trap
3
Water Bath'
N2
I
(D -
-Way Valve
Heater
-^-
Water
Manifold
i
On
i
/
-off Valve
i j i ,
kir
5S ! By-Pass
~i ; O
^>
/" -
) l LJ ^^
1 i 3-Way Valves
j PO
^u^
o/-\
oU
Data Acquisition
Ionic Mercury
Reduction Furnace
* .
Trap
^
S02
A n^l\/"7pr
Mercury
/-\rialV/.cl
^
^
Figure 3. Schematic of the Bench-Scale Packed Bed Reactor
17
-------
D No SO2 D 173 ppm SO2
FGD
M. Advacate Advacate Lime
Sorbent
CRFA
Figure 4. Effect of Sulfur Dioxide on Elemental Mercury Capture by the Sorbents at 100ฐC
18
-------
10 15 20
Time (min)
25
30
Adv @ 65 C
Adv(S> 100"C
mod. Adv @ 65 C
-*-
mod. Adv (a) 100ฐC
Figure 5. Effect of Temperature on Elemental Mercury Capture by Calcium-Based
Sorbents (Advacate and Modified Advacate) in the Presence of 173 ppra Sulfur Dioxide
19
-------
15
Time (min)
30
Figure 6. Capture Percentage of Incoming Sulfur Dioxide by the Sorbents at 100ฐC
20
-------
100
80
60
o
Z 40
O
o
20
Dl73ppmS02DNoSO2
FGD M. Ad vacate Ad vacate
Sorbent
Lime
CRFA
Figure 7. Effect of Sulfur Dioxide on Mercuric Chloride Capture by the Sorbents at 100 C
21
-------
100
20
10 15 20
Time (min)
25
30
Adv @ 65ฐC
Advfffi !00ฐC
mod. Adv @ 65ฐC
mod. Adv @ 100ฐC
Figure 8. Effect of Temperature on Mercuric Chloride Capture by the Calcium-Based
Sorbents (Advacate and Modified Advacate) in the Absence of Sulfur Dioxide
22
-------
Performance of Activated Carbon for Mercury Control in Utility
Flue Gas Using Sorbent Injection
T.R. Carey, O.W. Hargrove, Jr., C.F. Richardson
Radian International LLC
8501 N. Mopac Blvd.
Austin, TX 78759
R. Chang
Electric Power Research Institute
3412HillviewAve.
Palo Alto, CA 94303
F.B. Meserole
Meserole Consulting
8719 Ridgehill Drive
Austin, TX 78759
Abstract
Mercury continues to be considered for possible regulation in the electric power industry under
Title III of the Clean Air Act Amendments of 1990. EPRI is conducting research to investigate
mercury removal using sorbents. This paper describes the results of parametric bench-scale tests
investigating the removal of mercuric chloride and elemental mercury by activated carbon and
fly ash. Results indicate that the adsorption capacity of carbon is dependent on many factors,
including the type of mercury being adsorbed, flue gas composition, and adsorption temperature.
The adsorption capacity of fly ash appears to be related to carbon content; however, the data are
not always consistent. These results provide insight into potential adsorption mechanisms and
suggest that the removal of mercury involves both physical and chemical mechanisms.
Understanding these effects is important since the performance of a given sorbent could vary
significantly from site-to-site depending on coal- or gas-matrix composition. These data are
being used to develop a theoretical model for predicting mercury removal by sorbents at different
conditions.
Introduction
The Clean Air Act Amendments of 1990 listed 189 substances as hazardous air pollutants. Thirty
seven of these substances have been detected in power plant emissions with eleven being trace
metal species. Mercury is the trace metal species of greatest concern because of perceived risks
from its environmental release1. Most of the trace metal species are efficiently removed in
-------
properly operated paniculate removal systems. Mercury, however, is present mainly in the vapor
form and is not captured effectively by existing particulate removal systems.
Injection of activated carbon or other sorbents upstream of a particulate control device is one
potential method for controlling mercury emissions from utility boilers. Carbon-based and other
sorbents have been developed for control of mercury emissions from municipal- and hazardous-
waste incinerators. Existing data from the incinerators provide some insight into mercury control,
but these data cannot be used directly for coal-fired utilities because mercury concentrations,
species, and process conditions differ greatly2. For example, municipal solid waste (MSW)
mercury concentrations (200 to 1000 ug/m3) are one to two orders of magnitude higher than for
coal combustion sources (5 to 20 ng/m3). The relatively high mercury concentrations in MSW
systems and other differences in process conditions generally result in relatively low carbon
injection rates. Tests have shown that the carbon to mercury ratio in MSW incinerators is more
than an order of magnitude lower than that necessary to achieve similar mercury removal in coal
combustors.
Several bench-3"5, pilot-6"9, and full-scale101" tests have examined the influence of carbon type,
carbon structure, carbon surface chemistry, injection method (dry or wet), amount of carbon
injected, and flue gas temperature on mercury removal. Inconsistencies in the data suggest that a
wide variety of factors may influence the mercury removal obtained when injecting sorbents into
flue gas upstream of an electrostatic precipitator (ESP) or a baghouse. These factors potentially
include the mercury species being removed (oxidized vs. elemental), the flue gas composition,
process conditions (e.g., temperature), sorbent characteristics, and the presence of other active
surfaces (e.g., fly ash). Many of the earlier tests did not consider or measure all of these
variables. Thus, data have been difficult to compare.
To develop a better understanding of how the above parameters affect mercury adsorption and
sorbent effectiveness, EPRI has developed a systematic approach for evaluating potential
sorbents and developing a theoretical model to predict performance. This approach and the
theoretical model have been discussed in more detail previously12' '3. Laboratory tests are used to
characterize the physical and chemical properties of a given sorbent such as size, shape, surface
area, porosity, and chemical composition. Bench-scale, fixed-bed tests are then conducted to
determine the equilibrium adsorption capacity and breakthrough characteristics. These tests are
conducted under simulated flue gas conditions, and results using different sorbents can be
compared to provide relative indicators of performance. The sorbent properties and adsorption
equilibrium data are then used in the theoretical model that incorporates mass transfer
considerations to predict mercury removal during sorbent injection. This model will be refined as
bench-scale and pilot-scale injection test data are collected. Tests upstream of both electrostatic
precipitators and baghouses are needed to account for differences in residence time and physical
configuration.
This paper discusses recent bench-scale, fixed-bed adsorption results using both elemental
mercury and mercuric chloride under simulated flue gas conditions. A lignite-based activated
carbon obtained from Norit Americas (commercial name Darco FGD carbon) has been tested
under a wide variety of mercury concentrations and flue gas conditions. Several fly ashes with
-------
different carbon contents have also been tested in the lab. This work is significantly different
than the work of others since the majority of previous work has been done in nitrogen. Before
discussing the mercury adsorption results, the bench-scale test equipment and procedures are
described.
Test Methods and Procedures
In order to model the performance of a given sorbent, the equilibrium adsorption capacity and
characteristics of the sorbent must be known. If a particular sorbent has a low equilibrium
capacity for mercury, the mercury removal rate will be adversely affected since equilibrium will
be rapidly approached. Sorbent equilibrium data can be generated by conducting fixed-bed,
adsorption breakthrough tests. The bench-scale studies presented in this paper have focused on
evaluating the breakthrough characteristics of Norit Americas' Darco FGD carbon and different
fly ashes under a wide variety of test conditions.
Figure 1 shows a schematic diagram of the bench-scale, mercury sorbent test apparatus. Two
identical test units have been constructed. One unit is used to test elemental mercury adsorption
while the other is used to test mercuric chloride adsorption. In both systems, a simulated flue gas
is prepared by mixing heated, nitrogen gas streams containing SO2, HC1, CO2, O2, and water.
The gas composition can be varied by appropriately adjusting the various gas rates. Mercury is
injected into the gas by contacting nitrogen carrier gas with either recrystallized mercuric
chloride solids or with an elemental mercury diffusion tube (VICI Metronics) in a mercury
saturation vessel. The mercury concentration is controlled by the temperature of the mercury
saturator and the nitrogen flow rate .through the saturator. All gas mixing, water saturation, and
mercury injection occur within a closed, temperature-controlled box designed to prevent water
condensation which can affect the behavior of mercury and the gas concentrations in the flow
lines.
The reaction gas flows at about 1 standard L/min through heated Teflon lines (225ฐF) to a
temperature-controlled column (0.5-inch ID) containing the sorbent to be studied. The sorbent is
mixed in a sand diluent prior to being packed in the reaction column. In the column, the gas is
heated to the reaction temperature before contacting the sorbent by passing it across a bed of
pyrex spheres designed to enhance heat exchange. The column temperature is controlled using an
internally mounted thermocouple shielded from the gas with a glass sheath. The reaction gas
flows downward through the column to minimize the chance of selective flow or channeling
through the bed. The bed material is supported by a fritted glass disk and packed with quartz
wool. "Blank" tests have verified that the sand diluent and quartz wool do not adsorb mercury.
The linear gas velocity through the empty column is approximately 36 fVmin at 275ฐF.
During each test, the sorbent/sand mixture is equilibrated at the desired adsorption temperature
for at least one hour before contacting gas. During this time, the Inlet gas bypasses the sorbent
column and passes to the analytical system to determine the inlet mercury concentration. The
analytical system (described below) consists of a gold amalgamation unit and a cold-vapor
atomic absorption (CVAA) spectrophotometer. After the inlet mercury concentration is
established, the adsorption test is initiated by diverting the reaction gas through the sorbent
-------
column. The amount of mercury exiting the column is measured on a semi-continuous basis until
100% of the inlet mercury is detected at the outlet (100% breakthrough). The reaction gas
continually passes through the sorbent column during each test by sending the effluent gas to a
waste scrubber during the analysis step.
During normal operation of a test, the effluent gas from the fixed-bed column flows through
heated lines to an impinger containing SnCl2 solution which reduces any oxidized mercury
compounds to elemental mercury. After passing through the SnCL, impinger, the gas flows
through a buffer solution (Na2CO3) to remove the SO2 and HC1 from the gas, thus protecting the
downstream, analytical gold surface. Gas exiting the impinger solutions flows through a gold
amalgamation column housed in a tubular furnace where the mercury in the gas is adsorbed
(<100ฐC). After adsorbing mercury onto the gold for a fix period of tune (typically 6 minutes),
the mercury concentrated on the gold is thermally desorbed (>750ฐC) in nitrogen and sent as a
concentrated mercury stream to a cold-vapor atomic absorption (CVAA) spectrophotometer for
analysis. Therefore, the total effluent mercury concentration is measured semi-continuously with
a six-minute sample tune followed by a six-minute analytical period.
When testing elemental mercury adsorption, the effluent mercury can be fully or partially
oxidized due to reactions between the inlet elemental mercury, sorbent, and flue gas components.
The percentage of inlet elemental mercury oxidized across the sorbent is determined by replacing
the SnQ2 impinger with an impinger containing tris(hydroxymethyl)aminomethane (Tris)
solution. The Tris solution has been shown in other EPRI studies to capture oxidized mercury
while allowing elemental mercury to pass through without being altered14. Therefore, with the
Tris impinger in place, any effluent .oxidized mercury is captured by the Tris impinger, and only
elemental mercury is detected by the downstream analytical system. The difference between the
total measured effluent mercury (SnCl2 impinger) and the effluent elemental mercury represents
the amount of elemental mercury oxidized across the sorbent bed.
Using the automated apparatus described above, mercury adsorption breakthrough curves were
determined at various operating conditions. The percent breakthrough is determined as a function
of time by normalizing the measured mercury concentration at the outlet of the sorbent bed to the
inlet mercury concentration. The capacity of the sorbent to adsorb mercury (|ig Hg/g sorbent) is
determined by summing the total mercury adsorbed to the sorbent through a given time. The
initial breakthrough capacity is defined as the capacity of the sorbent at the time when mercury is
first detected at the outlet. The 100% breakthrough (equilibrium) capacity is the capacity at the
tune when the outlet mercury concentration is first equal to the inlet concentration. The time to
initial breakthrough and the time to complete equilibrium (100% breakthrough) have to be
determined in order to calculate the respective capacities. These times are determined based on
the slope of the breakthrough curve. A linear regression of the breakthrough curve is determined
over the range of 20% to 80% breakthrough. The initial breakthrough time and 100%
breakthrough time are then determined based on this linear fit. To calculate the respective
capacities, the amount of mercury adsorbed from the start of the test to the respective time is
determined.
-------
Mercury Adsorption Test Results
Although the bench-scale test units are capable of testing nearly any sorbent type, results in this
paper focus on Darco FGD carbon and several different fly ashes. FGD carbon is lignite-based
and was obtained from Norit Americas Incorporated. It is a commercial product primarily sold as
a sorbent for heavy metal species in incinerator flue gas streams. Table 1 summarizes the FGD
carbon physical and chemical properties measured in the lab. The chemical composition was
obtained using energy dispersive x-ray analysis (EDX). The particle size distribution was
determined using both a scanning electron microscope method and Microtrac analysis (laser
scattering method). Typical product properties obtained from Norit Americas datasheet are
included for comparison. The analytical methods used to make these measurements were not
specified.
In addition to FGD carbon, several different fly ashes have been tested hi the lab. Fly ash results
presented in this paper were obtained using four different samples collected from three different
full-scale facilities. Two of the fly ash samples are from the same site. The fly ash samples differ
in their loss-on-ignition (LOI) content, varying from 0% to 82%. Analyses from two samples
show that the fly ash carbon content is only slightly lower than the measured LOI.
Table 2 summarizes the baseline conditions used during both the elemental mercury and
mercuric chloride FGD carbon and fly ash tests. FGD carbon has also been tested over a range of
conditions shown in Table 2. The baseline mercury concentrations are reported as a range
because these concentrations can not be precisely controlled from test-to-test due to the nature of
using a saturation vessel. FGD carbon is tested at baseline conditions on a regular basis to
determine variability in the measurement methods and to monitor unexpected changes in the
results.
Bench-scale results presented in this paper include the effect of the following variables on the
FGD carbon elemental mercury and mercuric chloride adsorption capacities:
Mercury concentration,
SO2 concentration,
HC1 concentration, and
Mercury oxidation (elemental mercury only).
The effect of temperature on FGD carbon mercury adsorption has been reported previously15. In
addition to the FGD carbon parametric results, the elemental mercury and mercuric chloride
adsorption capacities of the fly ashes are compared to those of FGD carbon.
At each set of test conditions, the initial breakthrough capacity and the 100% breakthrough
capacity were determined. Changes in these capacities at different conditions provide
information about how these variables might affect mercury removal. Changes hi the initial
breakthrough capacity suggest that the adsorption rate has been affected while changes hi the
-------
100% breakthrough capacity suggest that the equilibrium characteristics have been affected.
These results can be combined with the theoretical model to eventually predict changes in
mercury removal.
Effect of Mercury Concentration
Figure 2 illustrates the effect of inlet mercuric chloride concentration and elemental mercury
concentration on the FGD carbon equilibrium adsorption capacities at baseline conditions (see
Table 2). The inlet HgCl2 concentrations and HgCl2 adsorption capacities are reported based on
the mass of mercury rather than the mass of HgCl2 so that the comparison is on an equal molar
basis. The data in Figure 2 show that the FGD carbon adsorption capacities for mercuric chloride
and elemental mercury are equal at baseline conditions when the HgCl2 inlet concentration and
capacity are expressed in this way. For both mercury types, the equilibrium adsorption capacity
increases as the inlet mercury concentration increases. This effect of concentration is consistent
with a physical adsorption mechanism. Data presented in a previous paper showed that the
capacity also increases as the temperature decreases for both types of mercury15, also suggesting
a physical adsorption mechanism.
FGD carbon has appreciable capacity for mercuric chloride and elemental mercury even at
concentrations as low as 5 ug/Nm3. This result is important since both types of mercury are
typically present at 5-20 ug/Nm3 in coal-fired utility flue gases. Preliminary modeling work
suggests that a capacity of 500 ug/g should be sufficient to remove mercury from flue gas using
carbon. The measured capacities at about 10 ug/Nm3 varied from 200 to 1000 ug/g.
Although the equilibrium adsorption capacity for elemental mercury is equal to that for mercuric
chloride at baseline conditions, elemental mercury is not adsorbed as efficiently as mercuric
chloride. When testing mercuric chloride, total adsorption (i.e., 100% removal) is generally
achieved until the carbon is saturated and breakthrough occurs. With elemental mercury,
however, typically about 15% breakthrough is observed from the start of the test when the carbon
is "fresh" until rapid breakthrough occurs as the carbon becomes saturated. These data suggest
that although FGD carbon has a similar equilibrium capacity for elemental mercury and mercuric
chloride, the initial adsorption rate for mercuric chloride is higher, and therefore, this carbon may
be more effective at removing mercuric chloride than elemental mercury.
Effect of Flue Gas Composition.
Figure 3 illustrates the effect of SO2 on the FGD carbon capacity for elemental mercury and
mercuric chloride at baseline conditions. The inlet mercury concentration was 60-70 ug Hg/Nm3
during the elemental mercury tests and 30-45 ug Hg/Nm3 during the HgCl2 tests. The FGD
carbon adsorption capacity for both mercury types increases as the SO2 concentration decreases
from about 500 ppm to 0 ppm. The capacity for elemental mercury increased dramatically when
SO2 was removed entirely from the gas. Above about 500 ppm SO2, the capacity for both
mercury types does not change significantly as the SO2 concentration increases. Since most full-
scale utilities have at least 200 ppm SO2 in their flue gas, the effect of SO2 on capacity may not
be practically important.
-------
Figure 4 illustrates the effect of HC1 on the FGD carbon capacity for elemental mercury and
mercuric chloride at baseline conditions. The inlet mercury concentration was 53-61 |ig Hg/Nm3
during the elemental mercury tests and 28-33 ng Hg/Nm3 during the HgCl2 tests. As the HC1
concentration increases from 0 ppm to about 50 ppm, the mercuric chloride capacity increases,
with appreciable adsorption at each HC1 concentration. However, the effect of HC1 on the
elemental mercury capacity is more dramatic. Increasing the HC1 concentration from 0 ppm to 50
ppm increases the FGD carbon elemental mercury capacity from 0 ng/g to about 2500 |ig/g.
Above about 50 ppm HC1, neither the elemental mercury or HgCl2 capacity changes
significantly. The ability of FGD carbon to adsorb elemental mercury is clearly affected by the
HC1 concentration at the gas conditions tested. At low HC1 concentrations, essentially no
elemental mercury is adsorbed, and the adsorption capacity for elemental mercury is less than
that of mercuric chloride. At high HC1 concentrations, the elemental mercury capacity is about
equal to the mercuric chloride adsorption capacity (at the same inlet mercury concentration).
Results shown in Figures 3 and 4 clearly indicate that flue gas composition affects carbon
performance and that performing adsorption tests under realistic operating conditions is
important. The effect of flue gas composition on adsorption capacity suggests that the mercury
adsorption mechanism is not purely physical. Interactions between mercury and flue gas
components on the carbon surface may be important. Other researchers have generally done
bench-scale, mercury adsorption tests in nitrogen3"5. The results discussed above indicate that
tests conducted in nitrogen will probably produce different results than tests conducted in
simulated flue gas.
Effect of Mercury Oxidation on Elemental Mercury Adsorption
During most of the elemental mercury adsorption tests, the percentage of inlet elemental mercury
oxidized across the sorbent bed was determined at 100% breakthrough. The percent oxidation
was determined using the Tris solution method described earlier. The inlet elemental mercury
stream was also checked with Tris solution to verify that the inlet mercury was elemental
mercury and not partially oxidized before contacting the sorbent. These measurements showed
that the inlet mercury was generally less than 5% oxidized.
Figure 5 illustrates the FGD carbon, elemental mercury adsorption capacity at 275ฐF as a
function of mercury oxidation across the carbon. Data are shown for baseline tests, variable SO2
tests at baseline conditions, and variable HC1 tests at baseline conditions. The data indicate that
as oxidation of elemental mercury across the carbon increases, the adsorption capacity Increases.
These data follow a relatively tight trend considering the wide range of test conditions included.
Sorbents other than FGD carbon, including fly ash, have also been tested. In general, if a sorbent
adsorbs a significant amount of elemental mercury, oxidation across the sorbent is high.
However, high oxidation does not necessarily mean that adsorption occurs (i.e., some sorbents
oxidize mercury without adsorbing it). All of these data suggest that elemental mercury
adsorption, and probably mercuric chloride adsorption, occurs through a complex mechanism.
Figure 5 shows that oxidation (capacity) has varied over a wide range while operating at constant
-------
baseline conditions. The cause of this variation in oxidation is not known, but subtle differences
in adsorption conditions and sorbent surface conditions may be important.
Mercury Adsorption Using Fly Ash
Several fly ash samples have been tested for their ability to adsorb elemental mercury and
mercuric chloride. Figure 6 compares the equilibrium adsorption capacities for various fly ash
samples to those of FGD carbon. The percent LOI of the fly ash samples varied from 0% to 82%.
The 17% and 82% LOI samples were collected from different locations at the same facility. The
82% LOI sample was manually ground using a mortar and pestle due to its coarse particle size.
In general, Figure 6 suggests that the elemental mercury and mercuric chloride adsorption
capacities increase as the fly ash percent LOI increases. Since LOI is essentially equal to carbon
content, the data suggest that mercury removal (capacity) increases as the carbon content of fly
ash increases. However, with only a limited number of samples, more tests are required to fully
characterize the effect of LOI on mercury adsorption and distinguish differences in capacity. For
example, the differences in adsorption capacity between the 18% and 82% LOI fly ashes are not
as large as might be expected based only on the difference in LOI. This observation suggests that
factors other than LOI may be important.
Field data also suggest that factors other than LOI are Important. The 0% LOI fly ash, which
showed no adsorption capacity in the bench-scale test, has been reported to remove significant
levels of mercury in pilot-scale studies6. The differences in mercury adsorption observed between
the lab and field tests again suggest.that other factors are important to fly ash adsorption. One
possibility is the NOX concentration; the lab tests were conducted with no NOX in the gas.
However, lab results using carbon and no NOX have generally correlated with carbon
performance in pilot studies. The inconsistent fly ash results indicate further study is needed.
As shown in Figure 6, the fly ash adsorption capacities are significantly lower than those of FGD
carbon. Despite these low capacities, modeling predictions suggest that the measured capacities
are sufficient to result in measurable mercury removal in particulate removal devices. The low
capacities are offset by the large mass of fly ash in the gas. Again, further study is required to
verify these model predictions.
Conclusions
The adsorption of elemental mercury and mercuric chloride by Norit Americas' Darco FGD
carbon has been studied over a wide range of conditions. The adsorption capacities for both types
of mercury increased as the temperature decreased and as the inlet mercury concentration
increased. These data are consistent with a physical adsorption mechanism. However, the
adsorption capacities were also affected by flue gas composition which suggests the mechanism
is not purely physical. The FGD carbon adsorption capacity for both types of mercury increased
as the SO2 concentration decreased. Similarly, the FGD carbon adsorption capacity for both types
of mercury increased as the HC1 concentration increased; however, elemental mercury adsorption
was affected more dramatically by HC1 and SO2 than mercuric chloride adsorption. The
-------
elemental mercury adsorption capacity increased from 0 ng/g at 0 ppm HC1 to 2500 |j.g/g at 50
ppm HC1. These results illustrate the importance of testing sorbents under realistic flue gas
conditions. In addition, these results indicate that, contrary to earlier assumptions3'5, removing
elemental mercury from flue gas using activated carbon is, in general, not much more difficult
than removing oxidized mercury. Oxidation of elemental mercury appears to be important to
mercury adsorption, with high levels of oxidation needed to achieve high adsorption capacities.
The carbon content of fly ash appears to affect the ability of fly ash to adsorb mercury; however,
further study is required to understand the factors affecting mercury adsorption by fly ash.
Acknowledgments
The authors acknowledge the Electric Power Research Institute for funding this work.
References
1. "Mercury in the Environment - A Research Update," EPRI Report TR-107695, December,
1996.
2. R. Chang and G. Offen. "Mercury Emission Control Technologies: An EPRI Synopsis."
Power Engineering. Vol. 5, November, p. 51 (1995).
3. C.D. Livengood, H.S. Huang, and J.M. Wu. Proceedings of the EPRI/DOE International
Conference on Managing Hazardous and Particulate Air Pollutants, EPRI TR-105749,
February, 1997.
4. S.V. Krishnan, B.K. Gullett, and W. Jozewicz. "Mercury Control in Municipal Waste
Combustors and Coal-fired Utilities." Environmental Progress. Vol 16, No. 1, p. 47 (1997).
5. R.D. Vidic and J.B. McLaughlin. "Uptake of Elemental Mercury Vapors by Activated
Carbons." J. Air & Waste.Manage. Assoc. Vol. 46, March, p. 241 (1996).
6. S. Sjostrom, J. Smith, T. Hunt, R. Chang, and T. Brown, "Demonstratrion of Dry Carbon-
Based Sorbent Injection for Mercury Control in Utility ESPs and Baghouses," Paper 97-
WA72A.07, presented at the 90th Annual Meeting & Exhibition of the Air & Waste
Management Association, Toronto, Canada (June 8-13, 1997).
7. C.L. Senior, E.B. Lawrence III, G.P. Huffman, F.E. Huggins, N. Shah, A. Sarofim, I. Olmez,
and T. Zeng, "A Fundamental Study of Mercury Partitioning in Coal Fired Power Plant Flue
Gas," Paper 97-WP72B.08, presented at the 90th Annual Meeting & Exhibition of the Air &
Waste Management Association, Toronto, Canada (June 8-13,1997).
8. J.T. Maskew, W.A. Rosenhoover, M.R. Stouffer, et. al. Preprint ACS Fuel Division. Vol 40,
No. 4, p. 843(1995).
9. G.E. Dunham, SJ. Miller, R. Chang, and P. Bergman, Paper 96-WP64B.03, presented at the
89th Annual Meeting & Exhibition of the Air & Waste Management Association, Nashville,
TN (June 24-27, 1996).
10. K. Felsvang, R. Gleiser, G. Juip, and K.K. Nielsen. Fuel Proc. Technol. Vol. 39, p. 417
(1994).
11. D.M. White, W.E. Kelly, M.J. Stucky, et. al. U.S. EPA, EPA/600/SR-93/181, January, 1994.
12. R. Chang, F. Meserole, T. Carey, O.W. Hargrove, C. Richardson, M. Rostam-Abadi, and S.
Chen, "Utility Flue Gas Mercury Control via Sorbent Injection," Paper No. ES96-42,
-------
presented at the Air & Waste Management Association Specialty Conference, Clearwater, FL
(February 28-March 1, 1996).
13. S. Chen, M. Rostam-Abadi, and R. Chang, In Proceedings of the American Chemical
Society, New Orleans, LA, March 23-28, 1996.
14. T.R. Carey, O.W. Hargrove, Jr., D.M. Seeger, C.F. Richardson, R.G. Rhudy, and F.B.
Meserole, "Effect of Mercury Speciation on Removal Across Wet FGD Processes," Paper
No. A405, presented at the AIChE Spring National Meeting, Session 47, New Orleans, LA
(February 25-29, 1996).
15. T.R. Carey, O.W. Hargrove, Jr., C.F. Richardson, R. Chang, and F.B. Meserole, "Factors
Affecting Mercury Control in Utility Flue Gas Using Sorbent Injection," Paper 97-
WA72A.05, presented at the 90th Annual Meeting & Exhibition of the Air & Waste
Management Association, Toronto, Canada (June 8-13, 1997).
10
-------
Table 1
Properties of Darco FGD Powdered Activated Carbon
Carbon Property
Lab Data Norit Americas' Data
General Bulk density (g/mL)
Properties: Surface area (mVg)
Molasses decolorizing efficiency
Iodine number
Particle % passing 325 mesh
Size: Avg size from SEM analysis (urn)
Avg size from Microtrac analysis (|J.m)
Pore Size micro, <20 A
Distribution meso, 20-50 A
(mg/g): macro, 50-150,000 A
Chemical Oxygen
Composition Carbon
(wt%): Silicon
Calcium
Iron
Aluminum
Sulfur
Magnesium
Table 2
-
-
-
94
15
14
-
28
22
14
13
7.4
7.1
3.7
2.9
0.51
600
90
600
95 minimum
-
-
0.18
0.25
1.06
-
-
-
-
-
1.8
Bench-Scale Test Conditions
Parameter Baseline Value
Gas Rate (L/min at 75ฐF) 1 .0
Gas Composition:
HgCl2 (ug Hg/Nm3) 25-45
Hgฐ(ngHg/Nm3) 40-80
SO2(ppmd) 1600
HC1 (ppmd) 50
02 (%) 6
CO2 (%) 12
H20 (%) 7
Adsorption Temperature (ฐF) 275
Range Tested
~
5-105
5-130
0-3000
0-100
-
-
225-400
-------
Temperature-controlled
Enclosure .
Waste Gas
Scrubber
N2 Purge
(p) Pressure guage (A) Manual Valves 00 Electronically Controlled Valves
Figure 1
Bench-Scale, Fixed-Bed Mercury Adsorption Apparatus
-------
ฃ
O
cd
Cu
IJ
B- ซ
i
20 40 60 80 100 120
Inlet HgCl2 or Hgฐ Concentration (ng Hg/Nm3)
140
Figure 2
Effect of Inlet Hgฐ and HgCl2 Concentration on the FGD Carbon Adsorption Capacity at 275ฐF
II
16000 n
14000 .
12000 .
10000 .
8000.
6000.
4000 -
2000 .
0.
: Hg(0)
i A HgC12
,
A A
A A
0 500 1000 1500 2000 2500 3000
Inlet SO2 Concentration (ppm)
Figure3
Effect of SO2 Concentration on the FGD Carbon Hgฐ and HgCl2 Adsorption Capacities at 275ฐF
13
-------
"o
D.
II
8- ซ
o u
^ Jp3
ฃ
'3
3500
3000 -
2500 .
2000 .
1500 .
1000 .
500 '
0 ,
i
;
:-
9
~f
':
: " ปA' Hg(0)
^ - HgCI2
L^^ ^.i.., , , , , i , i , i 1 i i ' i 1 ' ' i '
20 40 60 80
Inlet HCI Concentration (ppm)
100
Figure 4
Effect of Inlet HCI on the FGD Carbon Hgฐ and HgCl2 Adsorption Capacities at 275ฐF
I?
"o
s.
7000 ^_(
| S
i
a Variable HCI, 1600 ppm SO2
20 40 60 80
Oxidation of Effluent Mercury (%)
100
Figure 5
Effect of Mercury Oxidation on the FGD Carbon Hgฐ Adsorption Capacity at 275 ฐF
14
-------
2500
n HgC12 Adsorption Data
Hg(0) Adsorption Data
-------
Fixed-Bed Control of Mercury; Role of Acid Gases and
i Comparison Between Carbon-Based, Calcium-Based, and Coal Fly Ash Sorbents
Behrooz Ghorishi
Acurex Environmental Corporation
4915 Prospectus Drive
Durham, NC 27713
Brian K. Gullett
U.S. Environmental Protection Agency
Air Pollution Prevention and Control Division (MD-65)
Research Triangle Park, NC 27711
Abstract
The effects of hydrogen chloride (HC1, 50 pprn) and sulfur dioxide (SO2, 1000 ppm) on the
capture of both elemental mercury (Hgฐ, 40 ppb) by a thermally activated carbon (FGD), and
mercuric chloride (HgCl2, 73 ppb) by calcium hydroxide [Ca(OH)2] were examined in a fixed-bed
system for application to waste- and coal-fired combustors. Presence of both acid gases inhibited
the adsorption of HgCl2 by Ca(OH)2, possibly through depletion of available alkaline sites. The
inhibition effect of SO2 was more drastic than that of HC1. Capture of Hgฐ by FGD, however, was
improved in the presence of these acid gases. It appears that the reactions of SO2 and HC1 with
FGD create sulfur (S) and chlorine (Cl) sites. These sites may enhance the Hgฐ capture through
formation of S-Hg and Cl-Hg bonds in the solid phase (a chemisorption mechanism). The Cl sites
were more active in capturing Hgฐthan the S sites. Comparisons between a carbon-based sorbent
(FGD), Ca(OH)2, and coal combustion residues (fly ash, bottom ash, and scrubber sludge)
revealed that non-carbon-based sorbents with relatively high calcium (Ca) content can effectively
remove HgCl2 from a simulated flue gas. HgCl2 removals of up to half of the removal shown by
FGD were observed with some of the non-carbon-based sorbents.
Introduction
Title III of the 1990 Clean Air Act Amendments (CAAA) requires the U.S. Environmental
Protection Agency (EPA) to submit a study on 189 hazardous air pollutants (HAPs) from
industrial sources. This study will include an emission and risk (to public health) assessment of the
HAPs. Among the 189 HAPs, mercury species (hereafter, mercury) have drawn special attention
due to their increased levels in the environment and well-documented food chain transport and
bioaccumulation.1>2
-------
Mercury, present in hazardous/municipal wastes and in coal3, is readily volatilized during
combustion and incineration processes.4 Mercury is the most volatile among various trace metallic
species, and major portions of it can pass through existing paniculate control devices.4 A sorbent
that can react with mercury can effectively shift this metal from the vapor phase to the particulate
(sorbent) phase, facilitating its removal. Mercury control processes which use adsorption on dry
sorbents do not require treatment and stabilization of the waste liquid stream and, therefore, seem
very attractive for coal combustors and hazardous/municipal waste incinerators. The need to
develop mercury control technologies and the attractive features of adsorption processes on dry
sorbents led researchers to focus their efforts on the evaluation of the adsorption kinetics and
sorbent capacity of many different solid sorbents. Past research has identified two different classes
of sorbents to be effective in mercury removal: activated carbons and calcium-based sorbents.5"8
An important scientific issue that needs to be addressed is the chemical form in which mercury is
released from combustion systems. Previous investigations conducted in EPA laboratories6"8 have
indicated that mercury control strategies are dependent upon the type of mercury species that
exists in the coal/waste combustion flue gases. These studies have shown the relative ease of
controlling oxidized mercury (specifically HgCl2) as opposed to Hgฐ. Hall et al.9 showed that, in a
simulated municipal waste combustor (MWC), flue gas mercury is mainly found as HgCl2. They
postulated that HgCl2 is the most favorable mercury species due to the relatively high HC1
concentration in MWCs. On the other hand, Hgฐ is believed to be the prevailing form of mercury
in emissions from coal combustion processes.10 However, recent pilot-scale coal combustion test
results have indicated that combustion of certain types of coal (Blacksville, a bituminous coal
from the Pittsburgh No 8 seam) can lead to a flue gas mercury species profile dominated by
oxidized mercury (most probably HgCl2)." Research in this area is on-going to determine the
conditions that favor formation of oxidized mercury in coal combustion processes.
The presence of both Hgฐ and HgCl2 in combustion flue gases justifies a comprehensive research
study on the adsorption of both these species by solid sorbents. As mentioned, preliminary
investigations found activated carbons to be efficient Hgฐ sorbents,6'8 and calcium-based
compounds to be effective HgCl2 sorbents.7'8 Combustion residues (fly ash, bottom ash, and
scrubber sludge) produced from burning coal may have some intrinsic properties capable of
adsorption of mercury species. As part of the investigation presented here, the mercury sorption
properties of different residue samples generated during combustion of Illinois coals were
examined.
Previous activated carbon studies6'8 focused on determining the effects of temperature (60-140ฐC)
and Hgฐ concentration [8-40 ppb (65-327 ug/dscm)] on the adsorption of Hgฐ on two kinds of
thermally activated carbons (FGD and PC-100, Norit Americas Inc.) in a fixed-bed reactor
system. It was concluded that surface area and temperature have a strong effect on the Hgฐ
adsorption capacity of the activated carbons. PC-100, with twice the specific surface area of
FGD, captured about 4 times the amount of Hgฐ captured by FGD. The adsorption capacity of
both activated carbons increased by an order of magnitude when the fixed-bed temperature was
reduced from 140 to 60ฐC. Hgฐ concentration had minimal effect on the activated carbons'
-------
adsorption capacities. This study reports on the continuation of those investigations6-8 with
emphasis on the effects of two acid gases, SO2 and HC1, on the adsorption of Hgฐ by the most
commonly used activated carbon (FGD). These acid gases are believed to be the important
coal/waste combustion flue gas components influencing the capture of mercury species by solid
sorbents. The effects of SO2 and HC1 were studied at a concentration of 1000 ppm (2.6 g/dscm)
and 50 ppm (0.07 g/dscm), respectively.
Previous investigations7'8 with solid sorbents illustrated the relative ease of removal of HgCl2 as
compared to Hgฐ. It was determined that Ca(OH)2 can be effective in adsorption of HgCl2 under
fixed-bed conditions. Those investigations were focused on determining the effects of bed
temperature (60-140ฐC) and HgCl2 concentration [11-73 ppb (122-809 ug/dscm)] on theHgC!2
capture by Ca(OH)2. It was concluded that gas-phase HgCl2 concentration has a strong effect on
its adsorption; a 6.6-fold increase in HgCl2 concentration (from 11 to 73 ppb) resulted in a 12-fold
increase in HgCl2 uptake. Fixed-bed temperature had minimal effect on this adsorption. The
investigation presented here reports on the continuation of the previous studies;7'8 it focuses on
deducing the effects of SO2 (1000 ppm) and HC1 (50 ppm) on HgCl2 uptake by Ca(OH)2.
In summary, the scope of the present work was to study the mechanisms of Hgฐ uptake by FGD
and HgCl2 uptake by Ca(OH)2 under more realistic conditions; i.e., in the presence of acid gases.
Moreover an attempt was made to evaluate the potential usage of coal combustion residues as
mercury species sorbents. All the experiments were performed in a laboratory-scale, fixed-bed
apparatus, in which inlet simulated flue gases were contacted with the sorbent under study. The
fixed-bed results presented here will be used to design experimental test programs to be
conducted in a 50 cfm (0.024 mVs) coal/waste combustion pilot plant. The future pilot-plant tests
will investigate the potential effectiveness of these sorbents (carbon-based, lime-based, coal
combustion residues) for in-flight capture of mercury species from actual flue gas.
Experimental Procedure
A schematic of the experimental apparatus used to study the capture of Hgฐ and HgCl2 is shown
in Figure 1. A detailed description of this system has been presented elsewhere.6"8 The important
feature of this system is the ionic mercury reduction furnace. This furnace (kept at 850ฐC) is
placed downstream of the fixed-bed reactor to convert any oxidized mercury (Hg++, as in HgCy
to Hgฐ. According to thermodynamic predictions, the only mercury species that exists at this
temperature is Hgฐ.12 The presence of the furnace enabled detection of non-adsorbed HgCl2 as
Hgฐ by the on-line ultraviolet (UV) Hgฐ analyzer, thus providing actual, continuous Hgฐ or HgCl2
capture data by the fixed-bed of sorbent.
In each test, the fixed-bed of sorbent was exposed to the mercury-laden flue gas. During the
exposure period, the exit concentration of mercury was continuously monitored. The percent
removal of Hgฐ or HgCl2 at any time was obtained as:
instantaneous removal, % = 100*[(Hg)b-(Hg)0J/(Hg)in
-------
Selected experiments conducted during this test program were run in duplicates and indicated a
variation of ฑ10% in the experimental results.
The presence of HC1 (50 ppm) in the simulated flue gas created a major interference in the on-line
mercury analysis system (Figure 1). Repeated tests confirmed that HC1 reacts with Hgฐ in the gas
phase and at the outlet of the furnace during the natural cool-down period, thus converting
portions of Hgฐ back to HgCl^ which is not detectable by the on-line Hgฐ analyzer. This
interference undermined the use of the furnace as an ionic mercury thermal converter. An off-line
method was adopted to study the effect of HC1 on Hgฐ and HgCl2 adsorption. In this method, the
sorbents were exposed to the mercury-laden flue gas for a period of 24 hours and subsequently
the amount of mercury adsorbed on the sorbent was determined using an X-ray fluorescence
(XRF) technique (Siemens SRS 303 XRF analyzer). Selected tests in the off-line test program
were run in duplicates and indicated a variation of ฑ13% in the experimental results.
Sorbents
The role of acid gases on the capture of Hgฐ was studied using a thermally activated carbon:
FGD. FGD, manufactured by Norit Americas Inc., is a lignite-coal-based activated carbon. The
effects of SO2 and HC1 on HgCl2 uptake were investigated using reagent grade Ca(OH)2 (Sigma
Inc.) containing 97.6% Ca(OFf)2 and 1.8% calcium carbonate (CaCO3). A detailed description of
the physical properties of these two sorbents can be found elsewhere.6"8
The mercury adsorption properties of four different combustion residue samples were evaluated in
this study (Table 1). These samples were obtained from four Illinois power plants. Power plant #1
is in southern Illinois and burns Herrin (No. 6) coal which is mined in southern Illinois. The coal is
combusted in pulverized coal combustion (PCC) and cyclone boilers. Power plant #2 is near the
Illinois-Indiana border. It bums Springfield (No. 5) and Herrin (No. 6) coals that are combusted in
PCC boilers. This plant has two wet limestone scrubber units: one utilizes an inhibited oxygen
cycle and the other employs a forced-oxidation cycle. Power plant #3 is in central Illinois and
bums Herrin (No. 6) and Springfield (No. 5) coals which are mined in southern and central
Illinois, respectively. This plant uses fluidized bed combustion technology. Power plant #4 is in
central Illinois and bums Springfield (No. 5) coal. The coal is combusted in one PCC boiler and
two cyclone boilers. A wet limestone scrubber utilizing a forced-oxidation cycle has been installed
on the PCC unit.
Elemental compositions of these residue samples were obtained using the XRF technique. These
results, in comparison to the elemental composition of FGD (carbon not included), are shown in
Table 2.
-------
Experimental Results and Discussions
Comparison Between Carbon-Based Sorbents, Lime-Based Sorbents, and Coal Combustion
Residues
Coal combustion residue evaluations [in comparison to FGD and Ca(OH)2] were conducted in the
fixed-bed reactor setup (Figure 1) under the following conditions: 10% carbon dioxide (CO^, 7%
oxygen (OJ, 5% water vapor, 173 ppm SO2, and the balance nitrogen (NJ. Total system flowrate
of 300 cirrVmin, bed temperature of 100ฐC, and 0.1 g of sorbent were used throughout this study.
HC1 was not included in the simulated flue gas due to its interferences with the ionic mercury
reduction furnace and the on-line Hgฐ analyzer. A low SO2 concentration (173 ppm) was chosen
to minimize the interferences in the on-line Hgฐ analyzer.6"8
HgCl2 sorption capabilities of the residue samples were evaluated by introducing 73 ppb HgCl2 in
the simulated flue gas prior to the fixed-bed reactor. HgCl2 removals by the four residue samples
in comparison to FGD and Ca(OH)2 are shown in Figure 2. As expected, FGD exhibited the
highest removal under these conditions (about 90-80%), maintained for at least 200 minutes.
Ca(OH)2 also showed considerable removal: it removed 68% of incoming HgCl2 for about the
first 20 minutes of exposure followed by a decreasing removal pattern to about 20% after 200
minutes of exposure. The residue sample 2PPS exhibited a relatively good HgCl2 capture
capability: initially it removed about 50% of incoming HgCl2 with a slow decrease (note the log
scale of the time axis) to about 20% after 200 minutes of exposure. This sorbent is a scrubber
sludge sample from a power plant that uses a limestone wet scrubber. One may relate the HgCL,
removal activity of this sample to the presence of Ca compounds in this sludge (see Table 2). The
other scrubber sludge sample (4PPIS) also exhibited considerable HgCl2 removal (Figure 2). The
good HgCl2 capture performance of Ca(OH)2 and scrubber sludge samples may indicate a
correlation between Ca content and HgCL, capture.
Hgฐ capture capabilities of the residue samples were evaluated by introducing 40 ppb Hgฐ in the
simulated flue gas. Of the four residue samples (Table 1), only two exhibited measurable Hgฐ
removals: 3PF and 2PPS. These results in comparison to FGD and Ca(OH)2 are shown in Figure
3. Results indicate the degree of difficulty in Hgฐ adsorption (as compared to HgClj). Activated
carbon, as expected, was the most efficient sorbent for removal of Hgฐ. Sample 3PF exhibited the
next best removal: a 13% initial removal and exhaustion after 20 minutes of exposure to the Hgฐ-
laden flue gas. XKF analysis of this residue sample (Table 2) revealed that, unlike other samples,
3PF contains bromine (Br) at a low concentration of 8.5 ppm. As indicated in the next section, the
presence of Cl sites in activated carbon may enhance the capture of Hgฐ through formation of Cl-
Hg bonds in the solid phase. In this case, Br (another halogen element) might have been
instrumental in adsorption of Hgฐ through formation of Br-Hg bonds. Sample 3PF also showed
considerable HgCl2 removal (Figure 2) probably due to the presence of Ca (Table 2).
-------
Role of Acid Gases on HgC^ Uptake by Ca(OH)2
Acid gases are believed to be important flue gas components affecting the adsorption reactions of
mercury species. Results presented in the previous section and other investigations7'8 showed
Ca(OH)2 to be an efficient HgCl2 sorbent. Thus, the effect of acid gases on HgCl2 capture was
studied using this sorbent in the fixed-bed reactor. As mentioned, due to the interference of HC1 in
the on-line mercury analyzer, an off-line method using XRF was adopted. In the fixed-bed
reactor, 1 g of Ca(OH)2 was exposed to different flue gases (total flow of 300 crrrVmin) at a bed
temperature of 100ฐ C for a total exposure time of 24 hours. Five runs at different flue gas
conditions were performed to deduce the effect of HC1 in the presence and absence of SO2.
Results are shown in Table 3.
Run #la was the blank run in which the bed was exposed to a stream of N2. As expected, the
XRF analysis of the exposed sorbent revealed no solid-phase mercury. In run #2a (the baseline
run), the sorbent [Ca(OH)2] was exposed to a flow of N2 and 73 ppb HgCl2. Subsequent XRF
analysis of the exposed sorbent (Table 3) revealed a solid-phase mercury concentration of 6.03
mg mercury/g Ca(OH)2 the highest solid-phase mercury concentration observed during this test
program. In run #3a, the sorbent was exposed to a flow of 73 ppb HgCl2, 50 ppm HC1 and N2.
The solid-phase mercury concentration in the exposed sorbent dropped to 3.09 mg mercury/g
Ca(OH)2. The presence of HC1 inhibited the HgCU capture by a factor of 2. This inhibition may be
due to a competition for the available alkaline sites. Run #4a was designed to deduce the effect of
SO2 (1000 ppm) on the capture of HgCl2 by Ca(OH)2. The presence of SO2 inhibited the HgCl2
capture by a factor of 5. It appears that SO2 at this concentration is a more influential inhibitor
than HC1 at a concentration of 50 ppm. Finally in run #5a, the sorbent was exposed to both SO2
and HC1 simultaneously, resulting in the capture of 1.78 mg of mercury. Considering the
experimental variability of ฑ13%, the results indicate the prevailing inhibition effect of SO2 on the
capture of HgCl2 and.the fact that the inhibition effects of these acid gases are not additive.
In summary, the results indicate that 50 ppm HC1 in the flue gas can decrease the HgCl2 sorption
capability of Ca(OH)2 by half. The inhibition effect of 1000 ppm SO2 is even more drastic (factor
of 5), and essentially controls the uptake of HgCl2.
Role of Acid Gases on Hgฐ Uptake by Activated Carbon, FGD
Results presented in the previous section and other investigations6'8 showed FGD to be an
efficient Hgฐ sorbent. Thus, the effects of acid gases on Hgฐ uptake were studied using FGD in the
fixed-bed reactor. As mentioned, due to HC1 interferences in the on-line system, the off-line
method using XRF technique was adopted. In the fixed-bed reactor (Figure 1), five experiments
were conducted under different flue gas conditions with 1 g of FGD (100ฐC, flowrate of 300
cirrVmin, 24h). Results are shown in Table 4.
Run #lb (the blank run) revealed an absence of any detectable mercury (detection limit of 0.02
mg mercury/g sorbent). In run #2b (the "baseline" run), the FGD was exposed to a flow of 40 ppb
-------
Hgฐ in N2. Subsequent XRF analysis revealed a solid-phase mercury concentration of 0.187 mg
mercury/g FGD-- the lowest solid-phase concentration observed during the Hgฐ test program (see
Table 4). Unlike HgCl2 capture by Ca(OH)2, FGD was least active toward adsorption of Hgฐ
when no acid gas compounds were present. In run #3b, FGD was exposed to a flow of N2, 40 ppb
Hgฐ, and 50 ppm HC1. Solid-phase mercury concentration on exposed FGD increased to 0.874
mg mercury/g FGD. The presence of HC1 enhanced the Hgฐ capture capability of FGD by a factor
of about 5. It appears that HC1 reacted with FGD under these conditions and created Cl sites;
these sites were then instrumental in capturing Hgฐ through formation of Cl-Hg bonds in the solid
phase. Run #4b was designed to deduce the effect of SO2 on the capture of Hgฐ by FGD through
exposing it to a flow of N2, 40 ppb Hgฐ, and 1000 ppm SO2. The presence of SO2 enhanced the
Hgฐ capture by a factor of about 3. Similar to HC1, one can conclude that the reaction of S02 and
FGD may create active S sites which are effective in capturing Hgฐ through formation of solid-
phase S-Hg bonds. Despite a lower gas-phase concentration, HC1 had a more pronounced
enhancement effect on Hgฐ capture, suggesting that the Cl sites may be more active than the S
sites. Finally in run #5b (Table 4), the sorbent (FGD) was exposed to both SO2 and HC1
simultaneously. Under these conditions, the enhancement effect was the highest at a factor of
about 6, indicating that both types of active sites were instrumental (but not additive) in capturing
Hgฐ simultaneously.
It was hypothesized that exposing FGD to the acid gases, HC1 and SO2, creates active Cl and S
sites. Analysis of these elements in the exposed sorbent could potentially provide some evidence
for this hypothesis. The concentrations of Cl and S in the exposed FGD were determined using
the XRF technique; results are summarized in Table 5. The concentration of Cl in FGD exposed
to N2 was 0.226% wt. This is the background amount of Cl present in activated carbon (sample
#1). The Cl concentration in FGD exposed to HC1 (50 ppm) and Hgฐ (40 ppb) for 24 hours
increased to 0.547%, a 2.4-fold increase in Cl concentration. The increased concentration (or
number) of Cl atoms might then have been instrumental in capturing Hgฐ (see Table 4) through
formation of Cl-Hg bonds. The presence of SO2 (1000 ppm) in the flue gas decreased the Cl
uptake by FGD (sample #3); however, it created S sites (sample #5) which were hypothesized to
be active in capturing Hgฐ. Concentration of S in FGD exposed to N2 was 0.856%. This is the
background amount of S present in activated carbon (sample #4). The S concentration in FGD
exposed to SO2 and Hgฐ for 24 hours increased to 1.819%, a 2.1-fold increase in S concentration
(sample #5). Presence of HC1 apparently did not have any effect on the S uptake by FGD (sample
#6). In summary, it appears that the presence of acid gases in the simulated flue gas creates active
sites that are instrumental in capturing Hgฐ. Based on these results, one may conclude that the
optimum region for the control of Hgฐ by injection of activated carbon, FGD, is upstream of the
acid gas removal system.
Conclusion
The presence of acid gases (HC1 and SO2) can drastically affect the sorption behavior of gas-
phase mercury species by activated carbons and Ca-based sorbents. The effects of HC1 (50 ppm)
and SO2 (1000 ppm) on the adsorption of HgCl2 by Ca(OH)2 were studied in a bench-scale, fixed-
-------
bed reactor. The presence of both acid gases decreased the HgCl2 capture capability of Ca(OH)2.
The inhibition effect of 1000 ppm SO2 was more drastic than 50 ppm HC1, and essentially
controlled the uptake of HgCl2. It was hypothesized that the inhibition effect is due to a
competition between these acid gases and HgCl2 for the available alkaline sites. The presence of
acid gases, however, had a positive effect on the capture of Hgฐ by a lignite-coal-based activated
carbon (FGD). It appears that the presence of these acid gases in the flue gas creates active S and
Cl sites, which are instrumental in capturing Hgฐ, possibly through formation of S-Hg and Cl-Hg
bonds in the solid phase (chemisorption). Based on these results, one may conclude that the
optimum region for the control of Hgฐ by injection of activated carbon is upstream of the acid gas
removal system.
A comparison between activated carbon, Ca(OH)2, and coal combustion residues (fly ash, bottom
ash, scrubber sludge) indicated that non-carbon-based sorbents with relatively high Ca contents
can be effective HgCl2 sorbents. These Ca-based sorbents exhibited HgCl2 removals as high as
half the removal shown by one type of activated carbon (FGD). Hgฐ sorption studies, however,
showed the superior efficiency of the carbon-based sorbent as compared to the other studied
sorbents.
The sorbents and flue gas conditions studied in this investigation will be used to conduct a
comprehensive test program in a 50 cfm (0.024 m3/g) coal/waste combustion pilot plant. Future
pilot-plant tests will investigate the effectiveness of these sorbents for in-flight (short residence
time) capture of mercury species under more realistic conditions.
Acknowledgments
This work was cosponsored by the Illinois Department of Commerce and Community Affairs
through the Illinois Coal Development Board and the Illinois Clean Coal Institute (ICCI) (Project
Managers: Ken Ho and Ronald Carty). This work and use thereof are subject to the disclaimer
statement of the ICCI, available upon request. The authors acknowledge the logistical support of
Richard Valentine (EPA/APPCD), experimental assistance from Lisa Adams and Hamid
Bakhteyar, and technical assistance from Wojciech lozewicz (Acurex Environmental
Corporation).
References
1. D.G. Langley, "Mercury Methylation in an Aquatic Environment," J. Water Pollut. Contr.
Fed., 45: 44(1973).
2. G. Westoo, "Methyl Mercury a Percentage of Total Mercury in Flesh and Viscera of
Salmon and Sea Trout of Various Ages," Science, 181: 567(1973).
3. C.E. Billings, A.M. Sacco, W.R. Matson, R.M. Griffin, W.R. Coniglio, and R.A. Harley,
"Mercury Balance on a Large Pulverized Coal-fired Furnace," J. Air Pollut Contr Assoc
23(9): 773(1973).
-------
4. D.H. Klein, AW. Andren, J.A Carter, J.F. Emery, C. Feldman, W. Fulkerson, W.S. Lyon,
J.C. Ogle, Y. Talmi, R.I. Van Hook, andN. Bolton, "Pathways of 37 Trace Elements
Through Coal-fired Power Plant," Environ. Sci. & Technol., 9(10): 973(1975).
5. R. Meij, "A Sampling Method Based on Activated Carbon for Gaseous Mercury in
Combustion Flue Gases," Water, Air, and Soil Pollut., 56: 117(1991).
6. S.V. Krishnan, B.K. Gullett, and W. Jozewicz, "Sorption of Elemental Mercury by
Activated Carbons," Environ. Sci. & Technol., 28: 1506(1994).
7. S.V. Krishnan, H. Bakhteyar, and C.B. Sedman, " Mercury Sorption Mechanisms and
Control by Calcium-based Sorbents," in Proceedings of the 89th Air & Waste
Management Association Annual Meeting, 96-WP64B.05, Nashville, 1996.
8. S.B. Ghorishi, and B.K. Gullett, "An Experimental Study on Mercury Sorption by
Activated Carbons and Calcium Hydroxide," Presented at the Fifth Annual North
American Waste-To-Energy Conference, Research Triangle Park, NC, April 1997.
9. B. Hall, O. Lindqvist, and E. Ljungstrom, "Mercury Chemistry in Simulated Flue Gases
Related to Waste Incineration Conditions," Environ. Sci. & Technol., 24: 108(1990).
10. M.S. Devito, P.R.Tunati, R.J. Carlson, and N. Bloom, "Sampling and Analysis of Mercury
in Combustion Flue Gas," in Proceedings of the EPRI's Second International Conference
on Managing Hazardous Waste Air Pollutants, Washington, DC, 1993.
11. D.L. Laudal, M.K. Heidt, T.D. Brown, B.R. Nott, and E.P. Prestbo, "Mercury Speciation:
A Comparison between EPA Method 29 and Other Sampling Methods," in Proceedings of
the 89th Air & Waste Management Association Annual Meeting, 96-WP64A.04,
Nashville, 1996.
12. S.V. Krishnan, B.K. Gullett, and W. Jozewicz, "Mercury Control by Injection of
Activated Carbon and Calcium-based Sorbents," in Proceedings of Solid Waste
Management Thermal Treatment & Waste-To-Energy Technologies, USEPA/AEERL and
A&WMA, Washington, DC, 1995.
Table 1
Illinois coal combustion residue designations
Sample ID #
1PPB
2PPS
3PF
4PPIS
Power Plant #
1
2
3
4
Sample Type
Bottom Ash (PCC boiler)
Scrubber Sludge (inhibited oxygen cycle)
Fly Ash
Scrubber Sludge (forced-oxidation cycle)
-------
Table 2
Elemental composition of Illinois coal combustion residues (% wt)
Element
Na
Mg
Al
Si
S
K
Ca
Ti
Mn
Fe
Cu
Br
Sr
Sample Identification #
2PPS
0.043
0.544
0.174
1.190
18.4
0.070
24.9
0.024
0.010
0.151
0.001
0.000
0.025
4PPIS
0.025
0.094
0.047
0.412
18.4
0.019
24.2
0.004
0.006
0.057
0.001
0.000
0.009
1PPB
0.548
0.536
7.34
27.9
0.248
1.25
4.2
0.289
0.056
7.3
0.005
0.000
0.014
3PF
0.457
3.54
4.35
12.8
6.4
1.01
19.5
0.114
0.043
4.71
0.005
0.001
0.031
Activated Carbon (FGD)
0.182
0.569
1.72
3.57
0.856
0.069
1.82
0.095
0.019
0.910
0.001
0.000
0.057
10
-------
Table 3
Effect of HC1 (50 ppm) on the HgCl2 (73 ppb) capture by Ca(OH)2 in the absence and
presence of SO2 (1000 ppm)
Run#
la
2a
3a
4a
5a
Flue Gas Component
N2 (blank)
N2 + HgCl2 (baseline)
N2 + HgCl2 + HCl
N2 + HgCl2 + SO2
N2 + HgCl2 + HC1 + S02
Solid-phase Mercury Concentration
mg mercury/g Ca(OH)2
0
6.03
3.09
1.28
1.78
Table 4
Effect of HCI (50 ppm) on the Hgฐ (40 ppb) capture by activated carbon, FGD, in the
absence and presence of SO2 (1000 ppm)
Run#
Ib
2b
3b
4b
5b
Flue Gas Component
N2 (blank)
N2 + Hgฐ (baseline)
N2 + Hgฐ + HCI
N2 + Hgฐ + SO2
N2 + Hgฐ + HCI + SO2
Solid-phase Mercury Concentration
mg mercury/g FGD
0
0.187
0.874
0.549
1.072
11
-------
Table 5
Concentration (% wt) of chlorine (Cl) and sulfur (S) in simulated-flue-gas-exposed
activated carbon, FGD
Sample #
1
2
3
4
5
6
Flue Gas Component
N2 (blank)
N2 + Hgฐ +HC1
N2 + Hgฐ + HC1 + SO2
N2 (blank)
N2 + Hgฐ + SO2
N2 + Hgฐ + HC1 + SO2
Element
Cl
Cl
Cl
S
S
S
Concentration (% wt)
0.226
0.547
0.355
0.856
1.819
1.733
12
-------
Mercury Generation
System
N.
Carbon Trap
Manifold
Carbon Trap
On-off Valves
Q-
Reactor
By-Pass
-e
3-Way Valves
HCI
_ CD
S> i=
-------
AC (FGD) lime 2PPS 4PPIS 1PPB 3PF
100
10 20 50
Time (min)
100 200
Figure 2
HgCl, removal by four coal combustion residues (see Tables 1 and 2) in comparison to
activated carbon, FGD, and Ca(OH)2
14
-------
60
50
CD
O 40
E
0)
a:
30
0
CO
ฃ
CD
E
_0)
LU
20
10
AC (FGD) lime 3PF 2PPS
5 10 20
Time (min)
100 200
Figure 3
Hgฐ removal by two coal combustion residues (see Tables 1 and 2) in comparison to
activated carbon, FGD, and Ca(OH),
15
-------
CAPTURING AND RECYCLING PART PER BILLION LEVELS OF MERCURY
FOUND IN FLUE GASES
Dr. Daryl L. Roberts
Ms. Robin M. Stewart
Mr. Thomas E. Broderick
ADA Technologies, Inc.
304 Inverness Way South, Suite 365
Englewood, CO 80112
Abstract
ADA Technologies is developing a sorbent-based process that removes and recovers mercury
found in flue gases made by the combustion of coal. Coal-fired power plants are a prime
candidate for regulations on the emissions of mercury since they constitute 20% to 40% of
man-made emissions of mercury to the atmosphere. Mercury is receiving significant
attention as an air toxic compound since health effects are documented and since the Clean
Air Act Amendments of 1990 dictate that the Environmental Protection Agency report to
Congress on a prudent regulatory approach for mercury. EPA is currently under a court order
to promulgate regulations for public comment by January 14, 1998, although some delay is
possible.
ADA's process concept involves the uptake of the mercury on a sorbent that contains a noble
metal, thermal regeneration of the sorbent, and the recovery of liquid, elemental mercury for
commercial distillation and re-use. Multiple sorbent beds insure the continued removal of
mercury from the flue gas when one or more sorbent beds are being regenerated. Because of
the attributes of the system, we have adopted the name "Mercu-RE" to describe the process.
The mercury is removed from the biosphere, eliminating the eventual re-release of the
mercury via leaching or volatilization from a solid or liquid waste.
There are several ways to configure the noble metal sorbent, such as a packed bed, a
monolith, or on a filter bag. In laboratory work with synthetic flue gases, we have
-------
demonstrated the regenerability of the sorbent through 56 cycles of uptake and regeneration.
The sorbents have removed both elemental mercury and oxidized mercury in laboratory
testing and in testing at a coal-fired pilot combustion facility.
We installed a skid designed to handle 20 ACFM of flue gas on a slip stream of a coal-fired
pilot combustion facility operated by Consol, Inc. (Library, PA). Between late January and
late June, 1997, the skid-mounted unit treated flue gas from four different coals during
approximately 700 hours of run time. Smooth operation and reliable regeneration were
difficult to achieve. However, when the beds are properly regenerated, the unit removed
essentially 100% of the mercury found in the flue gas. According to ADA's speciating
mercury analyzer and according to wet chemistry methods of mercury analysis, more than
50% of the mercury in these flue gases has typically consisted of oxidized forms of mercury.
These pilot tests constitute the first demonstration of a process that simultaneously removes
elemental and oxidized mercury quantitatively from realistic coal flue gases.
Introduction
ADA is developing a novel, regenerable mercury capture technology that involves a highly
efficient, regenerable sorbent. The main attributes of this patented process (Durham, et al.,
1995) are highly efficient mercury removal, mercury recovery, sorbent regeneration, and
sorbent re-use, and as a consequence, ADA has adopted the name "Mercu-RE" to describe
the process. The Mercu-RE process has the following advantages:
Mercury removal efficiencies exceeding 95 % regardless of the chemical form of the
mercury compared with 25% to 75% efficiency of alternative technologies,
A substantial reduction in the cost of mercury control compared with alternative
approaches,
Elimination of mercury-contaminated solid or liquid wastes, and
Removal of mercury from the biosystem.
Figure 1 contrasts the fate of mercury in the Mercu-RE process with the fate of mercury in an
uncontrolled waste combustor and in a system using state-of-the-art carbon injection for
mercury control. The end product of the Mercu-RE process is liquid, elemental mercury,
which is suitable for recycle and re-use and is thereby not available to be distributed into the
biosystem. Further, no secondary wastes are made. In contrast, state-of-the-art carbon
injection technology produces a mercury-contaminated carbon with approximately 300 times
the mercury concentration of the original fuel, in many cases mixed with fly ash. Although
this mercury-contaminated fly ash/carbon mix may pass the Toxic Characterization Leach
Procedure (TCLP), the mercury on this carbon is susceptible to eventual leaching and
volatilization, introducing the mercury into the biosystem. In the worst case, the ash with the
highly contaminated carbon would be considered a hazardous waste, requiring costly, special
disposal practices.
-------
Present Status
(No Control)
650 Ibs. Hg 10 Biosysem
State-of-the-Art
(Carbon Injccuon)
Hg - Containing Coal
650 Ibs, Hg Entering Biosyscra
ADA Mcrcu-RE Process
Figure 1 - Fate of Mercury in Various Control Schemes
In a full-scale power plant application, the Mercu-RE process would involve multiple sorbent
modules treating approximately 100,000 ACFM each and would encompass the following
steps:
1. Capturing mercury for one day or one month from 100,000 ACFM of flue gas at 300ฐF to
400ฐF;
2. Taking one sorbent module off-line;
-------
3. Regenerating the sorbent module for eight hours at 500ฐF to 700ฐF, passing less than 50
ACFM of hot purge gas through the module, thereby creating a highly concentrated
mercury stream;
4. Condensing at 120ฐF the mercury contained in the purge gas;
5. Putting the sorbent module back on-line; and
6. Selling or safely disposing of the liquid mercury.
The promise of the Mercu-RE process to meet the needs of a range of applications, such as
thermal processing of hazardous waste and coal-fired utilities, derives from its ability to
capture all common forms of mercury vapors and from the variety of physical
configurations in which the technology can be practiced, some of which have the potential to
remove paniculate matter simultaneously with the mercury.
In one configuration, the sorbent could be dispersed hi the body of a high-temperature
bag filter. In this configuration, mercury is sorbed in the body of the filter after particles
are removed from the gas stream on the front face of the filter. The advantage of this
configuration is that no new equipment is needed to conduct the mercury removal.
In another configuration, the sorbent in a particulate form could be dispersed in a
cylindrical support structure that is placed inside a filter bag. This configuration
accomplishes the mercury removal inside the vessel that is used already for particle
control, allows for a greater residence time of the gas in the sorbent, and allows for
greater lifetime of the sorbent between regeneration cycles.
In a third configuration, a coated monolithic form of the sorbent in its own vessel can
be made with sufficient residence time to allow and very high mercury removal efficiency
while providing very low flow resistance.
All three of these configurations can be readily adapted into the air pollution control systems
employed for both waste and coal-fired utility applications and represent trade-offs between
pressure drop, frequency of regeneration, and mercury removal efficiency.
Based on our early work on this concept as applied to municipal waste incinerators (Roberts,
1995), we are now extending our process to the very low levels of mercury found in coal-
fired power plants (0.1 ppb to 1.0 ppb). Our current contract with DOE FETC-Pittsburgh has
afforded us the opportunity to look at these low levels both in the laboratory and with flue gas
from a pilot coal combustor.
This work has consisted of four tasks:
Task 1 - Screen Sorbent Configurations in the Laboratory
Task 2 - Design and Fabricate Bench-Scale Equipment
Task 3 - Test Bench-Scale Equipment on Pilot Combustor
-------
Task 4 - Evaluate Economics Based on Bench-Scale Results
The following sections describe the results of each of these tasks.
Screening Sorbent Configurations in the Laboratory
We examined the paniculate and the monolith form of the sorbent in the laboratory. To
make the paniculate form, we crushed commercially-available alumina beads and sieved
them to be smaller than about one millimeter. We then dispersed the noble metal on these
fine alumina beads. We subjected this paniculate form of the sorbent to two accelerated
durability tests. First, 10 grams of the sorbent was held in an oven at the regeneration
temperature of 700ฐF continuously for 180 days. In real operation the sorbent would be
exposed to the regeneration temperature for at most 50% of the time, so the 180 days of
exposure represented at least one year of operation. We removed samples of the sorbent from
the oven periodically and examined the size of the noble metal crystallites using x-ray
diffraction line broadening techniques. The size of the crystallites remained unchanged for
180 days (Figure 2) indicating that the crystallites themselves do not migrate or grow under
the elevated regeneration temperatures.
1.50
1.00
~
i
S w
"> *
a =
ฃ 0.50 -
aj O
ฃ
O
0.00
50 100 150
Time of Exposure to 370C (days)
200
Figure 2 - Stability of Noble Metal Crystallites During Exposure to High Temperature
The second accelerated durability test was to expose this paniculate form of the sorbent to
continued sorption and desorption of mercury at elevated mercury concentrations (3,000
ug/m3) in a synthetic flue gas containing air and seven volume percent water vapor at 275ฐF.
We have described the experimental apparatus elsewhere (Roberts, et al., 1996). The
-------
apparatus includes a continuous, speciating mercury analyzer developed by ADA
Technologies (Sjostrom, et al., 1997). This analyzer was used to determine when mercury
began to appear downstream of the sorbent bed (breakthrough).
The mercury concentration of 3,000 ng/m3 is much higher than that encountered in coal-fired
power plants and therefore "ages" the sorbent with respect to mercury exposure much faster
than what would occur in a coal-fired power plant. We found that after a "break-in" period
of about 20 cycles, the sorbent breakthrough time remained consistent through 56 cycles,
which is when we stopped the test (Figure 3). Because the mercury concentration in these
tests was about 300 times that expected in coal fired power plants, the sorbent has seen as
much mercury as if it had undergone 1600 cycles in a coal-fired power plant. If the mercury
itself is going to adversely affect the sorbent behavior in a coal-fired power plant application,
it would have done so in the 56 cycles that we tested. While we cannot say that Figure 3
proves that the sorbent would last for over 1500 cycles in coal-fired power plant flue gas, we
can at least say that the sorbent is robust in the presence of much higher concentrations of
mercury than will ever be encountered in the coal-fired application.
Mercury Uptake Performance
Sorbent BVI
12
10 ปซ
i= 4
2 --
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56
Cycle Number
Figure 3 - Accelerated Sorption and Desorption Test for Particulate Form of Sorbent
Once we proved the stability of the sorbent in the particulate form, we then turned our
attention to what we believed would be a more practical form of the sorbent for a coal-fired
power plant, namely, a monolithic configuration. To make this monolithic configuration, we
obtained metallic monoliths commercially and coated the inside walls of the monolith with
the sorbent. We chose metallic monoliths because of their superior heat transfer
characteristics compared with standard ceramic monoliths that are commonly used in
automobile catalytic converters.
-------
We subjected the monolithic form of the sorbent to 21 cycles of sorption and regeneration in
a synthetic flue gas containing 18 ug/m3 of elemental mercury, 4% C>2, 6% water vapor, 34
ppm HC1, 1000 ppm SO2, 7.5 % CO2, and the balance nitrogen. The sorption temperature
was 300ฐF, and the regeneration temperature followed a profile that peaked at 700ฐF. We
varied the ratio of sorption time to desorption time until we found a ratio that worked well.
The monolith showed no permanent loss of performance over these 21 cycles, and it seemed
to be refreshed by two 24-hour desorptions at 1000ฐF.
4.00
3.50-
3.00
2.50
2.00-
o
-------
Designing and Fabricating Bench-Scale Equipment
We devised a skid-mounted test unit that consisted of two sorbent units, air heaters for
regeneration, and suitable automatic control systems. Each sorbent unit consisted of 17 tubes
in a shell-and-tube heat exchanger design. There were 51 monoliths stacked three per tube in
each sorbent unit. Each unit was designed to handle 20 ACFM of flue gas taken as a slip
stream from Consol's pilot coal combustor. The superficial velocity in each of the 17 tubes
was about 1 ft/sec at 300ฐF and one atmosphere pressure, and the empty bed residence time
was about 1.5 seconds. The slip stream itself was taken downstream of an electrostatic
precipitator.
We used ADA's continuous, speciating mercury analyzer to monitor the mercury
concentration at the inlet and the outlet of the sorbent beds. During two weeks of the testing,
Consol personnel also measured the mercury concentration with a modified Ontario Hydro
method. We also took several samples with iodated carbon traps provided by Frontier
Geosciences (Seattle, WA). The system was designed for remote operation through a
computer modem link. In practice, the unit did not run well enough to operate remotely.
Each of the monoliths contained a total of one milligram of noble metal on the monolith
surfaces. With a typical inlet mercury concentration of 10 ug/m3, this amount of noble metal
could be expected to last for 15 hours before reaching breakthrough. In this way, we hoped
to achieve several sorption/desorption cycles in the 90 hours of run time in a week of
operation of the Consol pilot combustor.
The skid-mounted unit was constructed at ADA laboratories in Englewood, CO, and shipped
to Consol.
Testing Bench-Scale Equipment on Pilot Combustor
We installed the bench-scale equipment at the pilot combustion facility of Consol, Inc., in
Library, PA. Consol bums its coals in this facility for about 90 hours per week and for about
32 weeks per year to evaluate fouling, slagging, and emissions behavior of its coals in
support of its coal business. Consol's combustor burns about 150 pounds of coal per hour
(about 1.5 million Btu/hr). We installed our 20 ACFM skid downstream of Consol's
electrostatic precipitator.
The bench-scale equipment treated the flue gas from four coals over eight calendar weeks in
which we achieved approximately 700 hours of operation (Table 1). Each of the coals had
approximately 0.1 ppm by weight mercury but varied in their sulfur and chlorine contents.
We monitored the mercury concentration in the inlet and outlet of our skid using our
speciating mercury analyzer (Roberts, et al, 1996; Sjostrom, et al., 1997). We did periodic
checks of the mercury readings by sampling through iodated carbon traps provided by
Frontier Geosciences (Seattle, WA; provides total mercury concentration). During the two
weeks of the higher sulfur Pittsburgh Seam tests, Consol sampled both before and after our
skid using a modified Ontario Hydro impinger train, a derivative of EPA method 29 that is a
-------
leading contender to being a "reference method" for measuring oxidized and elemental
mercury in flue gases.
Table 1 - Coals Burned During Testing of 20 ACFM Skid at Consol
Dates Tested
1/27 through 1/31
2/1 through 2/14
3/3 through 3/13
6/4 through 6/20
Coal Name
Illinois #6 Seam,
High Sulfur
Illinois #6 Seam,
Low Sulfur
Pittsburgh Seam,
High Sulfur
Pittsburgh Seam,
Low Sulfur
Sulfur Content
3.6-3.8%
1.0-1.1 %
2.5-2.7%
1.8%
Chlorine Content
0.06%
0.42 %
0.12%
0.11%
Substantial data were obtained during the 700 hours of run time. However, operating the unit
turned out to be much more of a challenge than we anticipated. We designed the unit to
operate remotely from our offices in Englewood, CO. Problems with drift on the analyzer
and crashes of the PLC program, however, gave only intermittent data in the first three weeks
of running. We did, however, consistently see a removal of 10 (Jg/m3 of mercury as the flue
gas passed through the monolithic sorbent beds.
During the two weeks between 3/3 and 3/13, Consol sampled with the Ontario Hydro
impinger ("wet chemistry") method, and we sampled with iodated carbon traps and with our
continuous analyzer. The inlet measurements agreed rather well, but the outlet measurements
did not agree. Indeed, there were several very high outlet numbers reported by the wet
chemistry and iodated carbon traps (Table 2). In contrast, the continuous analyzer reported
complete removal of the mercury during these tests.
There were several issues that became apparent from these iodated carbon data (the wet
chemistry numbers showed similar random high outlets). We came to realize that the exact
sequencing of the valves when switching from one sorbent unit to the other was able to
introduce regeneration gas to the outlet, providing what'looks like high average outlet values.
Of course, the only way that high Outlet values were possible was because the monoliths were
sorbing the mercury and then giving it off during regeneration.
-------
Table 2 -lodated Carbon Measurements of Skid Performance
Date and Time
3/6; 10:45 am to 11:45 am
3/11; 9:45 am to 10:20 am
3/11; 2:45 pm to 3:45 pm
3/12/97; 2:20 pm to 3:20 pm
3/12/97; 6 pm to 7 pm
3/13; 5 pm to 6 pm
Inlet
18.1 ug/m3
12.6ng/m3
ll.Ong/m3
8.6 |ag/m3
14.8 |ig/m3
9.81 ug/m3
Outlet
17.5 ug/m3
53.9 pg/m3
2 1.9 ug/m3
10.0 |Jg/m3
10.0 ng/m3
10.6 ug/m3
Upon further testing, we discovered that the regeneration gas was not regularly getting hot
enough to provide a good and regular regeneration. We fixed this problem before the 250
hours of running in June. In May, 1997, (after three coals and before the fourth), when we
desorbed the bed with gas that we were sure was hot enough (700ฐF), we obtained desorption
of mercury that quantitatively equalled the amount that we expect would be sorbed on a
saturated bed (five milligrams of mercury on 51 milligrams of noble metal; Figure 6). With
this result, we became confident that the monoliths had sorbed mercury during the first 450
hours of running during exposure to three coals but that had likely not been desorbing
adequately.
We were unable to achieve routine operation of the unit for a time period long enough to get
several cycles of sorption and desorption at essentially constant operating conditions.
However, the high outlets concentrations reported by the iodated carbon traps and the wet
chemistry and the quantitative desorption shown in Figure 5 indicate that the monoliths did
sorb mercury under field conditions and at least sometimes were able to desorb it. Because at
least half of the mercury in the Consol flue gas was oxidized mercury, as reported by our
analyzer and by wet chemistry, our results confirm that the monoliths removed both oxidized
and elemental mercury under field conditions, even after seeing four coals and substantial
non-routine operating conditions.
At the end of the testing at Consol we brought back one monolith from each vessel for
laboratory testing with both elemental mercury and mercuric chloride. Each monolith sorbed
100% of the mercury, both elemental and oxidized, for three cycles of testing in the
laboratory. These monoliths were coated with a fine dust after the 700 hours of operation at
Consol. This performance after about 350 hours of exposure (each) to flue gas, even with a
fine coating of dust, was confirmation of the robust nature of the monolithic mercury removal
technology.
10
-------
30
25 -
6 20-
a is-
O 10 -
Mass of Hg Desorbed = 5.1 mg
200 400 600 800 1000 1200 1400 1600
Time (min)
Figure 5 - Desorption of Unit 1
We believe in retrospect that it would have been prudent to go from the laboratory testing of
one monolith to field testing of one monolith to avoid some of the scale-up and operating
problems of heat transfer and flow distribution that hindered the operation and compromised
our ability to get the high quality of data we would have liked to obtain. However, the
sorption behavior and the regenerability of the monoliths was demonstrated.
Evaluating Economics Based on Bench-Scale Results
The capital cost of a sorbent bed to treat a specific flue gas depends on the concentration of
mercury in the gas and on how long the bed will last between regenerations. A simple
relationship can be derived between the capital cost of sorbent and the time between sorbent
regenerations. If essentially all of the mercury vapor is captured up to the point of
breakthrough, the breakthrough time, 1^, is related to the flow rate, Q, the mercury
concentration, C, and the mass of sorbent in the bed, Ms, by
11
-------
_MS AqW
Q C (1)
where W is the mass fraction gold on the sorbent and, Aq is the difference between the mass
ratio of mercury to gold at the end of a sorption cycle and the mass ratio of mercury to gold at
the beginning of a sorption cycle. The capital cost of the sorbent per unit of flue gas flow
rate, AIS, is related to the breakthrough time by
AIS G C
Here, G is the cost of a unit mass of noble metal on the sorbent. Reasonable values of Aq
and G are 0.1 and $10,000/lb; with these values, the capital cost per unit of breakthrough
time can be written
UgHACFMMday
(3)
As an example, equation 3 states that if the mercury concentration is 10 (Jg/m3 and if the
sorbent is regenerated once every other day (T^ = 2 days), the capital cost of the noble metal
itself will be $0.1796 for each ACFM of flue gas flow rate, or $179,600 for a 1,000,000
ACFM facility (about 250 MW). Preparing the sorbent costs about as much the noble metal
cost itself, and therefore the sorbent cost for the million ACFM system is approximately
$360,000. Based on these figures, we believe the major capital expense will not be the
sorbent itself but the vessels to hold the sorbent and the ducting to connect the sorbent
vessels to the power plant flue gas ducting.
It is likely that a grass roots installation of this technology downstream of an electrostatic
precipitator would cost in the neighborhood of $70/kW to $105/kW, figures taken from pulse
jet and reverse gas baghouse installations (e.g., Sloat, et al, 1993) and adjusted for inflation.
These costs, as we will see below, would make the technology more expensive than injection
of activated carbon.
The operating costs include electricity for overcoming the system pressure drop, heat for
regeneration, and maintenance. The cost of the electricity to run the fan to push the flue gas
through the sorbent bed depends on the bed pressure drop. With superficial velocities near 1
ft/sec, the bed pressure drop can be kept in the range of 1" to 2" of water. A pressure drop of
1.5" of water corresponds to a power consumption of 2.94 kWh per million actual cubic feet
of gas flow. At 50 per kWh, this power will cost $69,450 per year in a plant that processes
1,000,000 ACFM of flue gas (one year assumed to be 7884 operating hours).
12
-------
Table 3 -- Projected Process Costs for One Million ACFM Facility
Capital Items
Sorbent
Sorbent Vessels, Installed
Duct Work, Installed
Total Capital Cost
Operating Costs
Fan Power
Heat for Regeneration
Maintenance
(2% of Initial Capital Cost)
Total Operating Costs
Total Annualized Cost
Initial Capital Cost
$360,000
$23,750,000
$1,250,000
$25,360,000
Annualized Cost
(15% of initial capital cost)
$54,000
$3,562,500
$187,500
$3,750,000
$70,000
$70,000
$507,200
$647,200
$4,397,200
The major unknown cost at this point is the cost of heat for regeneration. At a coal-fired
power plant, plenty of steam is available compared to the needs of this process, and the actual
cost of this energy may be quite small. We have estimated the regeneration energy cost to be
on the order of the fan power cost. The amount of mercury needing to be removed and
condensed in the regeneration step is so small (e.g. one liquid quart of mercury condensed
every month in a system treating 1,000,000 ACFM with 10 ug/m3 of mercury) that no
substantial cooling loads will be required. Further, the condensation downstream of the
regeneration vessel can be with standard cooling water at about 100ฐF, so no refrigeration
will be needed. Maintenance on a substantial facility such as this one may actually be the
major operating expense, so precise figures on fan power and regeneration power are not
critical at this time.
Carbon injection is the only established technology for mercury control available for flue gas
treatment today (Bustard and Chang, 1994; Chang et al., 1993; Schelkoph et al., 1995;
Sjostrom, et al, 1997). According to this literature, to achieve a mercury capture efficiency
above 75%, approximately 10,000 pounds of injected carbon are needed per pound of
mercury removed. The price of an appropriate activated carbon is about $0.55/lb in the large
quantities needed for a large flue gas application. Therefore, to treat 1,000,000 ACFM
containing 10 ug/m3 of mercury will require three million pounds of activated carbon at a
cost of $1,620,000 per year. Operating and maintenance costs for the activated carbon
system are likely an additional 20% of this figure, bringing the overall annual cost of the
carbon system to $1,944,000.
This approximate cost comparison shows that the capital expense of the vessels that comprise
the Mercu-RE process must be reduced to compete on an economic basis with injection of
activated carbon. The cost of noble metal itself is quite insignificant. Costs for both the
13
-------
Mercu-RE process and for carbon will be reduced as the mercury concentration is reduced,
but carbon works less well at lower mercury concentrations while the Mercu-RE process is
independent of the mercury concentration in the range tested to date. Therefore, the cost
comparison will be more favorable to the Mercu-RE process at lower mercury
concentrations. The Mercu-RE process has attributes not available in other methods of
mercury control such as the ability to collect all chemical forms of mercury, to generate no
secondary wastes, and to regularly remove over 95% of the mercury. We are currently
devising contacting concepts, such as high throughput monoliths or bag impregnation, that
will reduce the capital expense of the process so as to make it attractive economically.
Acknowledgements
This work is being sponsored by the Federal Energy Technology Center under contract DE-
AC22-95PC95257 (October 1, 1995 through September 30, 1997). The authors wish to
acknowledge the active participation of Mr. Tom Brown, Contracting Officer's
Representative, both in this project and in seeking to advance the state of the art in the
measurement and control of mercury found in flue gases.
References
Bustard, C. J., R. Chang, "Sorbent Injection for Flue Gas Mercury Control," presented at the
87th Annual Air and Waste Management Meeting, Cincinnati, OH, June 1994.
Durham, M. D., D. E. Hyatt, R. M. Stewart, R. J. Schlager, "Mercury Removal Apparatus
and Method," U. S. Patent 5,409,522, April 25, 1995; assignee: ADA Technologies.
Roberts, D. L., "Design of a Sorbent-Based Process for Removal and Recovery of Mercury
Vapor from Incinerator Flue Gases," 1995 International Incineration Conference,
Seattle, WA, May 8-11, 1995.
Roberts, D. L., R. Marinaro, "A Process for Capturing and Recycling Mercury Found in Flue
Gases," paper 47c, Spring National Meeting, American Institute of Chemical
Engineers, New Orleans, LA, February 25-29, 1996.
Schelkoph, G. L., S. J. Miller, D L. Laudal, R. Chang, P. D. Bergman, "Observations in
Bench-Scale Study of Sorbent Screening for Elemental Mercury and Mercury(IT)
Chloride,'' presented at the 88th Annual Meeting of the Air and Waste Management
Association, San Antonio, TX, June 18-23, 1995.
Sjostrom, S., J. Smith, T. Hunt, R. Chang, T. Brown, "Demonstration of Dry Carbon-Based
Sorbent Injection for Mercury Control in Utility ESPs and Baghouses," paper 97-
WA72A.07, 90th Annual Meeting of the Air and Waste Management Association,
Toronto, Ontario, Canada, June 8-13, 1997.
Sloat, D. G., R. P. Gaikwas, R. L. Chang, "The Potential of Pulse-Jet Baghouses for Utility
Boilers, Part 3," Air and Waste, 43, 120-28, January, 1993.
14
-------
NOVEL IN SITU GENERATED SORBENT METHODOLOGY AND UV IRRADIATION
FOR CAPTURE OF MERCURY IN COMBUSTION ENVIRONMENTS
Chang-Yu Wu1, Tai-Gyu Lee1, Elizabeth Araru and Pratim Biswas1'*
'Aerosol & Air Quality Research Laboratory
Department of Civil & Environmental Engineering
University of Cincinnati, ML #71
Cincinnati, OH 45221-0071
2National Exposure Research Laboratory
U.S. Environmental Protection Agency
Cincinnati, OH 45219
Abstract
A gas phase sorbent precursor injection methodology resulting in in situ generated sorbent titania
particles in conjunction with UV irradiation has been shown to be effective in capture of mercury
in combustor exhausts. Capture efficiencies as high as 96% were measured for mercury by titania
with UV irradiation. Results of a series of experiments performed are presented to elucidate the
mechanism of capture of mercury. A very high surface area titania sorbent was first formed,
with mercury species vapors condensing onto this surface, followed by photocatalytic oxidation
and firm binding with the sorbent particles. The process has significant potential as a low cost
methodology for mercury control in practical combustion systems. Minimal retrofitting maybe
necessary as existing particulate control devices such as electrostatic precipitators have coronas
with UV radiation present.
Introduction
Volatile metals introduced into combustion systems may be transferred to the gas phase and have
been found to be difficult to capture in conventional control devices (Linak and Wendt, 1993).
A species that is receiving significant attention is mercury, and emissions from combustion
sources are potentially a great concern (Krishnan et al., 1994; Chu and Porcella, 1995). Unlike
most other heavy metals that are emitted in particulate form, mercury has been reported to be
released in the vapor phase in the elemental state. Data for municipal waste incinerators show
the gas phase fraction ranging from 10% to 90% depending on the waste composition and
operating conditions (Lindqvist, 1986; Bergstrom, 1986; Reimann, 1986; Hall etal., 1991;
Livengood et al., 1994). Data for coal-fired power plants show an even higher fraction, in some
-------
cases over 95% (Lindqvist, 1986; Bergstrom, 1986; Reimann, 1986; Vogg etal., 1986; Meij,
1991;Larjavaef al., 1992; Morency, 1994), although there is a wide range reported in the
literature (Fahlke and Bursik, 1995; Evans and Nevitt, 1997). Gas phase elemental mercury is
not effectively captured in typical air pollution control devices.
The mostly widely proposed technique for mercury capture is activated carbon, and activated
carbon impregnated with sulfur, chlorine or iodine (Sinha and Walker, 1972; Otani et al., 1986;
Krishnan et al., 1994; Livengood et al., 1994; Morency, 1994). However, the use of activated
carbon is problematic because of its high cost, poor capacity, low applicable temperature range,
regeneration and slow adsorption rate (Otani et al., 1986; Quimby, 1993). The other technique
proposed is corona discharge (Urabe et al., 1988; Helfritch et al., 1996). Radicals such as OH
and O, and ozone are generated in such an environment which then oxidize elemental mercury.
Very high efficiency of oxidation has been demonstrated by using pulsed, positive voltage corona
with energy density 10 w/cfm (Helfritch et al., 1996). However, no data of the particle size
distribution resulting from the subsequent nucleation of mercury oxide are available. Typically,
metallic particles formed by nucleation in combustion environments are enriched in the
submicrometer regime (Linak and Wendt, 1993; Lin and Biswas, 1994), and it is well known that
most particulate control devices have a minimum in collection efficiency in this regime (Flagan
and Seinfeld, 1988). Alternative approaches to remove elemental mercury vapor therefore need
to be developed.
Sorbent particles have been demonstrated to be effective for the capture of certain toxic metals in
combustion environments (Uberoi and Shadman, 1990; Ho et al., 1992; Uberoi and Shadman,
1992; Gullet and Ragnunathan, 1994). Recently it has been shown that compared to traditional
bulk sorbent particles used in a fixed bed or fluidized bed, in-situ generated agglomerates possess
a higher capture efficiency (Owens and Biswas, 1996a; Owens and Biswas, 1996b; Biswas and
Zachariah, 1997) as well as to suppress the formation of submicrometer particles (Owens and
Biswas, 1996a). This is due to the higher surface area provided by the large number of
agglomerated particles. Due to its closed shell electronic structure, it is known that mercury is
not as reactive as the other heavy metals such as lead, cadmium and others. However, it has been
reported that radicals produced in corona environments could potentially oxidize mercury.
Therefore, titania is a potential candidate amongst the various sorbent materials for mercury
capture. It is well known that radicals are generated on titania particle surfaces when ultraviolet
(UV) radiation is applied, and these radicals can be used as oxidants chemicals (Biswas and Wu,
1997). The proposed methodology could be readily scaled up in a cost effective manner as
electrostatic precipitators have a corona region with UV irradiation.
A methodology for capture of mercury in combustion environments by using in-situ generated
titania particles with UV irradiation is described. The results of mercury capture in systems with
UV irradiation alone, with titania sorbents, and with both UV and titania are reported.
-------
Experiment
Apparatus and Materials
A flow reactor system (Figure 1) was used to study the capture of mercury by in-situ generated
titania particles with and without UV radiation. An alumina reactor tube, 91.44 cm long with an
inner diameter of 2.54 cm was used. Compressed air was used as the carrier gas and was passed
through a HEPA filter (75-62 FT-IR purge gas generator, Balston Filter Products) to assure it was
particle free. Mercury vapor was introduced into the system by passing air above liquid mercury
contained in a gas washing bottle. The gas washing bottle was wrapped by a heating tape
(Thermolyne), and the temperature was controlled with a power controller (Type 45500,
Thermolyne). To minimize mercury condensation, another heating tape was used for the Teflon
tubing (that connected the exit of the bottle to the entrance of the furnace). Titanium
isopropoxide was introduced into the system by bubbling argon (prepurified, 99.99%, Wright
Brothers) through the precursor solution (titanium isopropoxide, 97%, Aldrich) contained in a
bubbler (Midget, 30 ml, Ace Glass). The bubbler was placed in a water bath (No. 5160 wide
neck flask, 500 ml, Pyrex; TM106 heating mantle, 500 ml, Glas-Col), the temperature of which
was controlled by a power controller. The tubing before (connected to the argon source) and
after (connected to the furnace entrance) the bubbler was wrapped by a heating tape to prevent
any losses due to condensations.
A photochemical reaction cell was placed at the exit of the reactor tube which was irradiated with
UV light. The cell was 60 cm in length and 5 cm in diameter, and was made of borosilicate glass
(No.7740, Pyrex). The transmittance of the glass is 94% for 360 nm, 72% for 320 nm and 30%
for 300 nm. The UV lamp (Type XX-40, 80 W, Spectronics) was 120 cm long, and the intensity
(for 365 nm at 25 cm) was 1850 W/cm2. A glass fiber filter (No.61663, Gelman Science) was
used downstream to collect particles for composition analyses. The gas stream was then passed
through a series of sampling impingers to capture gas phase mercury species using a procedure
similar to USEPA Method 29. To measure particle size distribution in real time, tubings were
connected to the system before and after the photochemical reaction cell to direct the sample
particles to a Scanning Mobility Particle Sizer (Model 3934 DMPS, TSI Inc.).
Procedures and Measurement
Three sets of experiments were conducted. In Set I, UV irradiation was used (without sorbents)
to test if there was any effect on mercury speciation. In Set E, the capture of mercury by in-situ
generated titania sorbent particles without UV irradiation was investigated. In Set ffl, UV
irradiation in conjunction with sorbents was used to examine the effects on Hg speciation. The
experimental conditions for all these sets are listed in Table 1. In all the experiments, the total
flow rate was maintained at 1 1pm. The corresponding residence time in the furnace was
approximately 3 s (calculated for a temperature of 1000 ฐC), and that in the photochemical
reaction cell was approximately 70 s (calculated for a temperature of 80 ฐC). To understand the
effects of feed rate, different titania to mercury ratios were used by varying the titanium precursor
or mercury feed rates. The mercury feed rate was determined by measuring the mercury
-------
1
; Heating
5
ฑr
n
Hg
Power
controller
1
1
AJumina Reactor ->
Furnace
UV lamp
Water
bath
P
Ti02
recurs
DT c
o
Power
ontroll
Photoreaction
Cell
Filter
To Hoo
Impingers
DMA
o
o o
n
solution
solution
Scanning Mobility
Particle Sizer
Figure 1. Schematic diagram of the experimental system to study mercury capture.
concentration in the impingers placed after the gas washing bottles. The titania feed rate was
determined by weighing the particles collected on the filter with a titanium precursor only feed.
Particle size distribution is an important factor affecting the capture efficiency. The distribution
was determined by a scanning mobility particle sizer (SMPS), which provided information about
the mean particle size, number concentration, volume concentration and standard deviation.
Measurements were made after the system had stabilized, and at least three measurements were
averaged for each experiment. To determine the particle composition, the particles collected on
the filter were analyzed by X-ray diffraction (Siemans D5000). The morphology of the particles
was examined by Transmission Electron Microscopy (TEM, CM20, Philips). The elemental
composition of the particles was determined by Energy Dispersive X-ray Spectroscopy (EDS,
ED AX International, 9800 Plus). To determine the partitioning of mercury compounds in
different phases, both the filter extract (by acid digestion, Wu et al., 1997) and the impinger
solutions were analyzed by Cold Vapor Atomic Absorption spectroscopy (CVAA, Thermo
Instruments, Inc.). The first sampling impinger that contained 0.1M HNO3 solution was used to
determine the amount of soluble mercury species (i.e. HgO) in the gas phase (Chu and Porcella,
-------
1995; Morency, 1994). The last three impingers that contained 0.4% KMnO4 in 10% H2SO4
were used to determine the amount of elemental mercury (CFR, 1992). The three impingers
were used to ensure that all the mercury was trapped, this being confirmed by the relatively low
concentrations trapped in the last impinger. All the impingers were rinsed with a 1M HNO3
solution. The sample collection time was 1 hour for all experiments. Due to the fact that
mercury contamination of labware can lead to irreproducible results, extra precaution was taken
to ensure cleanliness. Before every experiment, the reactor was purged with clean air at a very
high temperature (1500 ฐC) to remove remnant mercury species. Critical tubing was also
changed prior to every experiment. Blank samples by running clean air through the system were
collected and analyzed to determine background levels before every experiment (and confirm
relatively low levels).
Results and Discussion
Several experiments (Wu et al., 1997) were conducted to understand the mechanism and
determine the effectiveness of the capture process; only a few of the results (Table 2) are
discussed in this paper due to space limitations. The first set of experiments were conducted
with Hg and UV irradiation, and mercury and titania only. These were followed by experiments
with Hg, titania and UV irradiation; and the results compared to those described above.
The furnace temperature, residence time and Ti/Hg ratios were varied to elucidate the mechanism
of capture. The engineering of the sorbent oxide formation process to obtain a high surface
area/mass ratio is critical for effective metals capture. Details of the characterization of the in
situ generated sorbent have been described elsewhere (Wu et al., 1997).
The capture efficiencies for the different experiments were calculated (Table 2) as the fraction of
the mercury species on the filter to the total mercury determined as the sum of the mercury on the
filter and the impingers. The results are discussed in the following sections.
Set I: Hg + UV System
As radicals may already be generated in a UV environment (Helfritch et al., 1996), experiments
were first conducted for mercury with UV only to investigate whether oxidation occurred under
this condition. However, the results of SMPS measurements showed no difference in particle
size distribution from that of the background air. In addition, the CVAA analysis of the filter
extract showed that no mercury was captured as particulates, nor was any of the gas phase
mercury (determined by the analysis of the impinger solutions) oxidized to any significant degree
(Table 2). Experiments using a much higher power of UV (450 W, Xe arc lamp, Oriel) also
provided similar results. Consequently, it is concluded that mercury is not oxidized in air
streams when UV radiation is applied alone.
-------
Table 1
List of experimental conditions
Set
#
I.
Hg+UV
n.
Hg+TiO,
m.
Hg+Ti02+
uv
Hg inlet
temp (ฐC)
68, 85,
97
85,97
68, 85,
97
Hgflow
rate
(cc/min)
100
100
100
Ti
inlet
temp
(ฐC)
75
75
Tiflow
rate(cc/min)
100
100
Ti/Hg
(molar
ratio)
17.5,
9.1
48.0,
17.5,
9.1
Total
flow
rate
(cc/min)
1000
1000
1000
Furnace
temp (ฐC)
1000
1000
1000
Table 2
Summary of key experimental results (Furnace Temp. = 1000 ฐC, Ti/Hg =9.1).
Conditions
Hg only
Hg + UV
Hg + TiO2
Hg + Ti02 + UV
% of Mercury Collected
Impingers as HgO
~ 1
-0.13
0.05 to 0.31
0.005 to 0.01
Impingers as Hg
-99
-99
-98
3.85 to 15.6
Filter
(capture efficiency)
~0
-0.22
1.6 to 2
84.4 to 96.1
-------
Set II: Hg + TiO2 System
The capture of mercury by in-situ generated titania sorbent particles without UV radiation was
investigated. No discernible difference in the particle size distributions and the X-ray diffraction
patterns from those of titanium precursor only feed could be observed. The collected particles
also had the same white color as that of titanium precursor only feed. The CVAA analyses of the
filter extract did show that some mercury was captured on the titania particles. The above results
suggested that mercury was probably physically adsorbed on the in-situ generated titania
particles. Previous studies conducted in a suspension system also showed the adsorption
capability of titania (Wu, 1996). The physical adsorption on the in-situ generated particles is a
very important step to capture mercury as will be illustrated in the next section where both TiO2
and UV radiation are present. However, it should be noted that only a small fraction ( 2 %) of
mercury is captured by the titania particles which suggests that the capture is not very effective
and that a more stable form of mercury captured by the particles is necessary. It is therefore
concluded that the mercury vapors are re-entrained from the TiO2 particle surfaces as the exhaust
gases flow past the filter.
Set III: Hg + TiO2 + UV System
In-situ UV irradiation was used to investigate the effectiveness of Hg capture by titania particles.
The X-ray diffraction patterns and EDS of the collected particles for the highest Hg feed are
shown in Figures 2 and 3, respectively. When the titanium precursor was used in conjunction
with UV irradiation, the elemental mercury was oxidized to form mercury oxide. This was
qualitatively evidenced by the color of the filter samples, which was yellowish-brown instead of
white as observed in the previous experiments (Set H). These colors agree with previous reports
on mercury oxide (Kaluza and Boehm, 1971; CRC Handbook of Chemistry and Physics, 1983).
The formation of mercury oxide is also evidenced by the X-ray diffraction pattern (Figure 2) of
the collected particles. The identified peaks match those of mercury oxide reported in the
literature (Kaluza and Boehm, 1971). EDS results also identified Ti and Hg on the same
particles. These analyses prove that mercury oxide is formed in the particulate phase.
Compared with the previous set, a much higher fraction of mercury was collected on filters when
UV was added (84 % to 96 %, Table 2, and increased with increasing Ti/Hg ratios (Wu, 1996)).
The higher fraction of mercury collected on the filter indicates that a stronger bond is formed
between Hg and TiO2.
To investigate whether the post irradiation by UV light is also effective and applicable after
particles are collected, the filter samples of Set n (Hg+TiO2 system) were exposed to UV
radiation (450 W Xe arc lamp, Oriel) for 1 hour. No change of the color of the samples could be
discerned. The X-ray diffraction patterns of these samples also showed no differences from that
of the titanium precursor only feed. Probably the thicker layer of these samples reduces the UV
penetration and therefore its effectiveness. From these results, it can be concluded that UV
radiation must be applied in-situ to be effective.
-------
Figure 2. X-ray diffraction patterns of the collected particles for (a) Ti only feed and (b) Ti,
Hg and UV experiments (set IE).
1000FS
Al
Hg
TI
TI
Cu
Hg
Hg
ft n
20.0
Figure 3. Energy dispersive x-ray spectra of the collected particles in the set in experiment.
-------
The results of the different experiments performed (Wu et ah, 1997) help elucidate the
mechanistic capture of mercury. The titanium precursor that is introduced into the high
temperatures regions, is oxidized and a highly networked sorbent agglomerate is formed (Yang et
al., 1996, Yang and Biswas, 1997). This large surface area agglomerate provides a surface for
the condensation of the mercury species, however, the bonding between the mercury and titania
particles is weak. When no UV irradiation is applied, the mercury is re-entrained into the gas
stream, thus resulting in low capture efficiencies. However, if the agglomerate is exposed to UV
irradiation, the mercury species is oxidized (to a less volatile form) and binds with the titania
particles. Thus, the mercury is effectively captured resulting in high efficiencies as determined
quantitiatively (Table 2).
The use of in situ generated sorbent material is a critical step in the overall process. First, the
production of the sorbent oxide in the high temperature environments results in a large surface
area to mass ratio agglomerate. The injection of the sorbent precursor at optimal temperature
zones is thus an important factor. If not optimal, the particles may sinter readily resulting in a
decrease in the surface area. The second important advantage of the in situ generation of the
sorbent is that the sorbent-mercury complex can be effectively exposed to UV irradiation. As
stated earlier, this could potentially occur in the corona region of existing electrostatic
precipitators, thus not requiring external UV irradiation sources. This is currently being
researched in our laboratory. The third advantage is that the in-situ generated sorbent process is
effective at relatively short residence times (less than 70 s), shorter than that reported in the
literature (1.5 - 28 hrs) to observe the formation of oxide (Kaluza and Boehm, 1971). The short.
residence time of this process indicates the feasibility for application in real combustion
facilities for removing mercury.
The proposed methodology is effective and potentially superior to activated carbon techniques
for mercury capture. First, the amount of mercury captured by activated carbon has been
reported to be approximately 0.4 mg Hg/g of carbon (Krishnan et al., 1994). In this work, the
mercury captured by in situ generated titania particles is approximately 150 mg Hg/g of TiO2.
Using reagent grade precursor prices and an additional factor of 10, an estimate of the cost is
$4000 per pound of mercury removed. This is an upper limit as the commercial grade large scale
cost of titanium precursor is expected to be lower. Titanium dioxide is used extensively as
ingredients in low cost commerical items such as paints and toothpaste. The cost of the proposed
titanium precursor is much lower than that of activated carbon. The annual capital and operating
costs of activated.carbon have been estimated to be $ 14,000 to $ 38,000 per pound of mercury
removed (Chang and Offen, 1995). Another advantage of titania particles is that it could
potentially photo-oxidize other hydrocarbon species, VOCs and toxic PAH particles.
Conclusions
Mercury capture by in-situ generated titania aerosols with UV radiation was investigated in this
work. Mercury was found to be physically adsorbed on the titania particles. In the absence of
UV irradiation, the bonding between the adsorbed mercury and titania particles was weak and
mercury was readily reentrained. When UV irradiation was used the adsorbed mercury
-------
underwent oxidation and complexation on the titania particle surface. The formation of mercury
oxide was confirmed both qualitatively and quantitatively. The process was most effective
when both titania particles were generated and the UV radiation was applied in-situ. The
proposed methodology has potential for use in industrial scale systems and is cost effective.
ACKNOWLEDGMENTS
This work was partially supported by the Department of Energy under the Grant No. DE-FG22-
95PC95222 and the Ohio Coal Development Office, Grant OCRC-97B17.
REFERENCES
Bergstrom, J. G. T. (1986). Mercury Behaviour in Flue Gases, Waste Management &
Research, 4, 57-64.
Biswas, P. and Zachariah, M. (1997). In Situ Immobilization of Lead Species in Combustion
Environments by Injection of Gas Phase Silica Sorbent Precursors, to appear, Environ.
Sci. Technol.
Biswas, P. and Wu C.Y. (1997) Enhanced Capture of Heavy Metal Emissions. US Patent
Application Filed, May 1997.
CFR, 1992, Title 40, Pan 61, Appendix B, Method 101 A.
CFR, 1992, Title 40, Part 266.
Chang R. and Offen, G. R. (1995). Mercury Emission Control Technologies: An EPRI
Synopsis, Power Engineering, 51-57.
Chu, P. and Porcella, D. B. (1995). Mercury Stack Emissions from U.S. Electric Utility Power
Plants, Water, Air and Soil Pollution, 80, 135-144.
Clarke, M. J. (1993). The Nation's First Mercury Emissions Standard: An Assessment of
New Jersey's Standard for Municipal Waste Combustors, Annual Meeting of Air &
Waste Mgmnt. Assocn., Paper 93-RP-154.01, Denver.
CRC Handbook of Chemistry and Physics, 64 ed.,(1993). Weast, R. C., Astle, M. J. and
Beyer, W. H. Ed. CRC Press, Inc., Boca Raton, FL, p. B-l 16.
Evans, A. and Nevitt, K. D. (1997). Mercury Speciation Measurement on a 10 Mwe Coal-
Fired Simulator, Annual Meeting of Air & Waste Mgmnt. Assocn., Paper 97-
WP72B.07, Toronto, Canada
Fahlke, J. and Bursik, A. (1995). Impact of the State of the Art Flue Gas Cleaning on
Mercury Species Emissions from Coal-Fired Stream Generators, Water, Air and
Soil Pollution, 80, 209-215.
Flagan, R. C. and Seinfeld, J. H. (1988). Fundamentals of Air Pollution Engineering,
Prentice Hall, Englewood Cliffs, NJ.
10
-------
Gullett, B. K. and Ragnunathan, K. (1994). Reduction of Coal-Based Metal Emissions by
Furnace Sorbent Injection, Energy & Fuels, 8, 1068-1076.
Hall, B., Schager, P. and Lindqvist, O. (1991). Chemical Reactions of Mercury in
Combustion Flue Gases, Water, Air and Soil Pollution, 56, 3-14.
Helfritch, D., Harmon, G. and Feldman, P. (1996). Mercury Vapor Control by Means of
Corona Discharge, Annual Meeting of Air & Waste Mgmnt. Assocn., Paper 96-
ES96.41, Nashville.
Ho, T. C., Chen, C., Hopper, J. R. and Oberacker, D. A. (1992). Metal Capture During
Fluidized Bed Incineration of Wastes Contaminated with Lead Chloride, Combust. Sci.
Tech., 85, 101-116.
Kaluza, U. and Boehm, H. P. (1971). Titanium Dioxide Catalyzed Photooxidation of
Mercury, / CataL, 22, 347-358.
Krishnan, S. V., Gullett, B. K. and Jozewicz, W. (1994). Sorption of Elemental Mercury by
Activated Carbons, Environ. Sci. Technol, 28, No. 8, 1506-1512.
Larjava, K., Laitinen, T., Vahlman, T., ArtMann, S., Siemens, V., Broekaert, J. A. C. and
Klockow, D. (1992). Measurement and Control of Mercury Species in Flue Gases
from Liquid Waste Incineration, Intern. J. Environ. Anal. Chem., 49, 73-85.
Lin, W. Y. and Biswas, P. (1994). Metallic Particle Formation and Growth Dynamics During
Incineration, Combust. Sci. and Tech., 101, 29-43.
Linak, W. P. and Wendt" J. O. L. (1993). Toxic Metal Emissions from Incineration:
Mechanisms and Control, Prog. Energy Combust. Sci., 19, 145-185.
Lindqvist, O. (1986). Fluxes of Mercury in the Swedish Environment: Contributions from
Waste Incineration, Waste Management & Research, 4, 35-44.
Livengood, C. D., Huang, H. S., Mendelsohn, M. H. and Wu, J. M. (1994). Mercury Capture
in Bench-Scale Absorbers, PETC's 10th Annual Coal Precapture, Utilization and
Environmental Control Contractors Conference.
Meij, R. (1991). The Fate of Mercury in Coal-Fired Power Plants and the Influence of Wet
Flue-Gas Desulphurization, Water, Air, SoilPollut., 56, 21-33.
Morency, J. R. (1994). Control of Mercury in Fossil Fuel-Fired Power Generation, PETC's
10th Annual Coal Precapture, Utilization and Environmental Control Contractors
Conference.
Otani, Y., Kanaoka, C., Usul, C., Matsui, S. and Emi, H. (1986). Adsorption of Mercury
Vapor on Particles, Environ. Sci. & Technol., 20, 7, 735-738.
Owens, T. M. and Biswas, P. (1996a). Vapor Phase Sorbent Precursors for Toxic Metal
Emissions Control from Combustors, Ind. Eng. Chem. Res. 35, 3, 792-798.
Owens, T. M. and Biswas, P. (1996b). Reactions between Vapor Phase Lead Compounds
and In Situ Generated Silica Particles at Various Lead-Silicon Feed Ratios:
Applications to Toxic Metal Capture in Combustors, J. Air Waste Mgmnt. Assocn.,
46, 530-538.
11
-------
Quimby, J. M. (1993). Mercury Emissions Control from Combustion Systems, Annual
Meeting of Air & Waste Mgmnt. Assocn., Paper 93-MP5.03, Denver.
Reimaim,.D. O. (1986). Mercury Output from Garbage Incineration, Waste Management &
Research, 4, 45-56.
Sinha, R. K. and Walker, P. L. (1972). Carbon, 10, 754-756.
Uberoi, M. and Shadman, F. (1990). Sorbents for Removal of Lead Compounds from Hot
Hue Gasesm, AIChE J., 36, No. 2, 307-309.
Uberoi, M. and Shadman, F. (1991). Simultaneous Condensation and Reaction of Metal
Compound Vapors in Porous Solids, Ind. Eng. Chem. Res., 30, 624-631.
Urabe, T., Wu. Y., Nagawa, T. and Masuda, S. (1988). Study on Hg, NOX, SOX Behavior in
Municipal Refuse Incinerator Furnace and Removal of Those by Pulse Corona
Discharge, Seiso Giho, 13, 12-29.
Vogg, H., Braun, H., Metzger, M. and Schneider, J. (1986). The Specific Role of Cadmium
and Mercury in Municipal Solid Waste Incineration, Waste Management & Research,
4, 65-74.
Wu, C. Y. (1996). Ph. D. Thesis, Department of Civil and Environmental Engineering,
University of Cincinnati.
Wu C.Y., Lee T.G., Tyree G., Arar E. and Biswas P. (1997) Capture of Mercury from
Combustion Systems by In Situ Generated Titania Particles with UV Irradiation,
Combustion Science and Technology, in review.
Yang, G. M., Zhuang, H., and Biswas, P. (1996). Characterization and Sinterbility of
Nanophase Titania Particles Processed in Flame Reactors, NanoStructured Materials,
7, No. 6, 675-689.
Yang, G., and Biswas, P. (1997). Study of the Sintering of Nanosized Titania Agglomerates
in Flame Using In Situ Light Scattering Measurements, to appear, Aerosol Science and
Technology.
12
-------
A CIRCULATING FLUID BED
FINE PARTICULATE AND MERCURY CONTROL CONCEPT
D.J. Helfritch, P.L. Feldman and SJ. Pass
Environmental Elements Corporation
3700 Koppers Street
Baltimore, MD 21227
Abstract
The concept employs a circulating fluid bed in order to achieve a high particle density, which
promotes the agglomeration of particles. The fine particles entering the system are formed
into larger agglomerates which are then readily captured by a conventional electrostatic
precipitator. In addition, activated carbon can be injected into the circulating bed in order to
adsorb mercury vapor. High residence time, due to the recirculation, allows very effective
utilization of the carbon.
The testing of a laboratory pilot system is Described. The results showed that fine particles
and mercury vapor can be significantly reduced by passage through a fluidized bed of fly ash
and activated carbon. The best fine particle reduction is obtained with a dense bed, low
temperature, and high relative humidity. It was found that one can expect about an order of
magnitude decrease in fine particle number density when optimum bed conditions are
existing.
The mercury vapor concentration in the gas was effectively reduced to zero by the addition of
activated carbon to the fluid bed. It was shown that the addition of about 1% of the bed
weight in carbon reduced the exit mercury vapor to zero for over two hours, with a carbon
utilization of 770 gm of carbon per gm of mercury removed.
Introduction
U.S. utilities are faced with economic challenges to remain competitive as a result of the
ongoing movement to allow the transmission of low cost electricity across state lines. In
addition, environmental pressures are forcing most of these utilities to be prepared to reduce
the air emissions such as NOX, SOX, fine particulates and mercury from coal-burning plants.
The proposed PM2 5 regulations will demand improved fine particle control from existing
equipment, and potential mercury vapor regulations would impose the installation of new
control equipment.
The fact that a bed of particles will cause the agglomeration of smaller particles through
collisional processes has been demonstrated. Small particles, traveling at high velocity
-------
through a fluid bed composed of almost stationery larger particles will impact with these
particles, forming agglomerates of small particles attached to larger target particles. The
resulting larger agglomerates are then more readily captured by conventional means. This can
be illustrated by considering the passage of 0.1 (im particles through a bed of 5 (j.m stationery
particles. Figure 1 shows that the efficiency of collision between the 0.1 and the 5 um
particles increases as the bed density increases. The efficiency of agglomeration also depends
upon the collided particles to stick together, and this effect is shown by the influence of
decreasing cohesion. One principle objective of the work described here is a demonstration of
this effect with a fluid bed device with commercial potential that can improve the fine particle
collection capabilities of conventional equipment.
It is also known that activated carbon adsorbs both elemental and ionic mercury, typically
found in flue gas in vapor form. Demonstrations of this phenomenon using fixed beds of
carbon and by means of the injection of powdered carbon into the duct have been
performed.3'4'5 The second principle objective of this work is the demonstration of good
mercury adsorption by powdered carbon within and part of the fluid bed.
A successful demonstration of both of these objectives would define a control device for
combined fine particulate and mercury control and would suggest that successive scale ups
could lead to a commercial product constructed from systems that are already in full scale use
for acid gas absorption. These experiments', using a pilot scale version of a commercial fluid
bed contactor for sulfur dioxide and hydrogen chloride control, were designed to accomplish
this task.
Experimental Arrangement
The 150 CFM laboratory system used for the experiments is shown in Figure 2. Ambient air
is drawn into the heater section, where it is electrically heated and steam is injected such that
the resulting gas is similar to flue gas in temperature and water content. Fly ash obtained
from BG&E's Wagner Station precipitator is then injected such that the ash loading of the gas
is approximately 1 gr./CF. The size distribution of this ash is given in Figure 3, and is seen to
have a 10 micron mass mean diameter and a standard deviation of 3.
Mercury vapor was also injected into the reactor inlet for some tests. Up to 50 (igm/m3 of
mercury vapor could be injected by means of evaporation from an ambient temperature pool
of liquid mercury which was drawn off by small flow of air and subsequently mixed with the
reactor flow.
The flow then enters the reactor vessel, where it is turned vertically upward and reduced in
velocity to approximately 4 FPS. A circulating fluid bed totaling several pounds of fly ash is
maintained in the reactor vessel. Up to 500 gr./CF of the bed material leaves the top of the
reactor vessel to be caught by an in-line cyclone and returned to the bottom of the bed. A
-------
venturi below the bed keeps it from falling by providing a high velocity throat region through
which the particles cannot fall. Figure 4 shows the reactor construction.
A two fluid nozzle, directed upward, at the bottom of the reactor vessel sprays up to 0.1 gpm
of water into the vessel and the fluid bed. The increased humidity of the gas and the bed
serves to increase the agglomerating characteristics of the fluid bed and improve the
performance of the downstream precipitator.
A wire in tube electrostatic precipitator follows the vessel-cyclone system. This is composed
of a single tube, 8" in diameter and 5' long. The precipitator was energized with full wave
rectified power of approximately 40 KV and 2 mA. The precipitator was hand rapped in
advance of each test run. The dust catch from the precipitator was not returned to the fluid
bed - only the catch from the cyclone was returned to the bed.
The fluid bed was built from a relatively coarse sample of ash obtained from an inlet
precipitator field, which was simply added as a batch. The coarse fraction of the inlet ash
would also add to the bed and make up for mass lost through the cyclone. The bed would fall
through the venturiis to the reactor bottom when the fan was turned off. It would then be
removed from the reactor and saved for the next run, when it would be reintroduced to the
reactor. In this way it was not necessary to rebuild a bed each time.
The system flow rate ranged from 50 to 200 CFM. The inlet temperature was held close to
300ฐ F, while the reactor outlet temperature varied between 125ฐ and 240ฐ F, depending upon
the water spray injection rate. The bed density could be Increased hi steps by adding
additional bed particulate. The bed density was estimated by its pressure drop and by
measuring the recirculating concentration.
The particle size distribution and total mass at the precipitator exit was measured with an
Insitec PCSV continuous particulate emissions monitor. This instrument uses laser light
scattering to calculate gas velocity and particle size distribution. The size range capability of
this instrument is 0.3 urn to 33 um.
An EPM Model 791.905 mercury vapor monitor was used for mercury vapor concentration
measurements. The instrument is a double beam photometer using UV light as a source. An
optical filter selects the narrow 253.7 nm mercury line. Elemental mercury, but not HgO or
HgCl, are measured. The instrument has a detection limit of 2 ugm/m .
-------
Test Results
The effect of the bed density on the number density of particles less than one micron is shown
in Figure 5. This figure shows that as the circulating bed is increased from 0 to 400 gr/CF, the
number of particles less than 1 urn decreases fivefold from 100,000 to 20,000. This result
confirms the theory of Figure 1 which says that as the bed density increases, the number of
targets for small particle capture increases, and agglomeration efficiency improves. The
particle mass penetration of the cyclone remained relatively constant (at about 0.3%) as the
circulating bed density increased, also demonstrating that a decrease in fine particle number
density occurs as bed density increases.
Liquid water is a cohesive agent for small particles, and one would expect that the operation
of the spray nozzle would enhance small particle agglomeration. When water is sprayed into
the fluid bed, it exists for a few seconds either as a free droplet or as a liquid film on the bed
particle surfaces before it evaporates. This evaporation results in gas cooling, and this spray
cooling can be increased until the flue gas dew point is reached.
Figure 6 shows the effect of water spray cooling (given in terms of the reactor exit
temperature) on the number density of particles less than 1 um. Three sets of data are shown,
each for a different flow rate. The number'densities were obtained at the precipitator exit,
except the system inlet value of 22E4 particles per cc less than 1 pm, which was obtained at
the reactor inlet. We see that the effectiveness of the bed is poor at temperatures above 220ฐ
F, where cohesion is poor and where there is little difference in performance with or without
the presence of a bed. The fine particle concentration decreases as the water spray is
increased (and exit temperature decreases), and an order of magnitude decrease in fine particle
concentration takes place as the temperature is decreased from 250ฐ F to 125ฐ F. We also see
that fine particle number counts for each flow rate fall approximately on the same curve,
indicating little influence of flow rate on fine particle reduction.
While the electrostatic precipitator voltage and current were held constant in the tests
described above, a variation of this parameter (voltage) was made in order to demonstrate how
outlet particulate concentrations can be effected by this parameter. Figure 7 shows that
particle number density and mass density decrease as precipitator voltage increases. We see
that the total mass density decreases by two orders of magnitude for a 20 KV increase in
voltage, while the number density of fine particles decreases by one order of magnitude as a
result of the same voltage increase.
The effect of an activated carbon spiked bed on mercury vapor adsorption is shown in Figure
8. This figure shows the results of two separate runs, alike except for the inlet concentrations
of mercury vapor. The system was initially run with no bed, 150 ACFM, 230ฐ F temperature,
and a mercury vapor inlet concentration of about 50 u-gm/m3 as measured at the precipitator
outlet with no bed. Once steady conditions were verified, approximately 4 pounds of bed
particulate was added (at approximately the 20 minute point of Figure 8). The mercury vapor
-------
levels were cut in half by adsorption by the bed material. 20 gm of iodine impregnated
activated carbon, ground to approximately 1/16 inch average size, was added to the bed at the
30 and 50 minute marks for the high and low inlet concentration cases, respectively. The
carbon addition resulted in an Immediate decrease of outlet mercury vapor, to almost zero,
and the mercury levels remained this low for more than two hours, without the injection of
additional carbon. The unit was allowed to run in this way for the low concentration case
until the carbon became saturated and mercury breakthrough could be observed. Mercury
breakthrough occurred 170 minutes after the injection of 20 gm of carbon to the circulating
bed. The mercury concentration was held to below 20 ugm/m3 even after it had saturated the
activated carbon, due to continued adsorption by the fly ash bed.
Discussion of Results
The fly ash bed results show that fine particles and mercury vapor can be significantly
reduced by passage through a fluidized bed of fly ash and activated carbon. The best particle
reduction is obtained with dense beds and low spray down temperature. This result is
predicted by Figure 1, based upon more numerous and more cohesive target particles. We
find that we can expect about an order of magnitude decrease in fine particle number density
when optimum bed conditions are existing.
Downstream precipitator performance not only benefits from the increase in particle mass
mean diameter resulting from the action of the circulating bed but also from the low
temperature, high humidity conditions created by the evaporative cooling of the fluid bed
system, which lowers the volume flow rate and also allows the precipitator voltage to be
increased.
The mercury vapor concentration in the gas can be effectively reduced to zero by the addition
of activated carbon to the fluid bed. It was shown that the addition of 20 grams of activated
carbon to the bed, which was about 1% of the bed weight, reduced the exit mercury vapor
level from 25 to 3 jigm/m3 The fly ash bed itself is also an effective mercury adsorbent,
capable of a 50% reduction in mercury vapor. The mercury concentration, gas flow rate,
carbon weight, and length of time before breakthrough, give a carbon utilization of 1250 gm
of carbon per gm of mercury. If the contribution of bed ash adsorption is also counted, then
we obtain 770 gm carbon per gm mercury.
Economic Considerations
It is known that the removal efficiency of toxic trace elements from coal-fired power plants
increases as the fine particulate fraction of power plant emissions is controlled to lower and
lower levels. Thus, the achievement of very high fine particulate collection efficiency across
the circulating bed agglomerator assures a high level of control of particulate based trace
elements. If this high level of collection can be achieved at a cost less than conventional
-------
alternatives , then the country will have a formidable tool for achieving clean air goals and In
protecting health and property. The added ability of the circulating bed agglomerator to
control mercury emissions to high levels adds even more to the appeal of the system.
The control of respirable particulates(< 10 microns)requires very high overall particulate
collection efficiency, e.g. .003 Ib/mmBTU. This level of control is not presently practiced in
utility powerplants, and the capability of conventional particulate control technologies such as
electrostatic precipitators and fabric filters to reliably achieve this level of fine particulate
control remains in question. However, assuming that they can be used successfully for this
purpose, the necessary upgrade and additions to an existing control device would be expected
to cost on the order of $ 100/KW to achieve the .003 Ib/mmBTU level. On the other hand, the
circulating fluid bed agglomerator added to an upgraded but not expanded ESP would be
expected to cost approximately S20/KW. Added to about S30/KW for the "in-footprint" ESP
upgrade the total cost of this approach would be S50/KW, or half the cost of a questionable
equivalent option.
The pressure drop across the circulating bed system is expected to be approximately 5 inches
of water. This is lower than that of a conservatively designed fabric filter and represents an
energy consumption less than the oversized ESP needed to achieve the high collection
efficiency. Thus the operating costs for the circulating bed system will be lower than the
conventional devices for this service. "
In addition to low capital and operating costs for the control of fine particulates, the fluidized
bed allows the operating cost of the activated carbon necessary for mercury removal to be cut
significantly because of the higher carbon utilization in the circulating mode, and because of
higher capacity of adsorption in the lower temperature mode of operation. We have seen that
the activated carbon utilization was 1250 gm of carbon per gm of mercury. When the
contribution of the ash adsorption is added, the utilization increases to 770 gm of carbon per
gm of mercury. This can be compared to values that range from 2xl04 gm/gm 5 to 1250
gm/gm 3'4 If we assume a 100 MW power plant, burning a 10,000 Btu/lb coal with a mercury
content of 0.1 ppm for 8000 hours per year, then the yearly cost at a carbon ($4/lb.) injection
rate of 1250 Ib carbon per Ib mercury would be $360,000 6 Utilization of carbon adsorption
in a fluid bed at a rate of 770 gm/gm would result in a yearly cost of $220,000.
Conclusions
The program objectives were designed to characterize the limits of performance of the fluid
bed concept. Two pollutants were targeted as potential applications for this device, and the
test program was directed toward defining the parameters important for the removal of these
pollutants. The test program and analysis describe the following major performance
characteristics of the fluid bed system:
1. It was shown that increasing the bed density (and hence the ash recirculation rate)
increases fine particle agglomeration and can result in an order of magnitude decrease in the
-------
exit concentration of fine particles. Figure 4 illustrates the effect. The effect of residence
time is illustrated in Figure 5, wherein two sets of data are plotted, one representing a
residence time of 3 seconds (200 cfrn) and the other 4 seconds (150 cfrn). The data for 4
seconds shows a greater reduction in the number of fine particles than does the data for 3
seconds. This result is in agreement with theory, which predicts a time dependence due to the
importance of diffusion for fine particulate mobility.
2. When water spray evaporation is employed, temperature and humidity are coupled, and it
was demonstrated that low temperatures and high relative humidity enhance fine particulate
agglomeration through cohesivity enhancement. See Figure 5. All of the runs shown in
Figure 5 were operated with the same precipitator voltage. It is known that the low
temperature operation allows precipitator enhancement through voltage increase due to the
higher gas density. Therefore even greater reduction of fine particle emissions would have
been possible through higher voltage operation of the precipitator.
3. We have seen the number density of particles smaller than 1 micron, as measured at the
precipitator exit, decrease by an order of magnitude following the action of the circulating
bed, while keeping the applied voltage constant. We also know that the low temperature, high
humidity gas exiting the circulating bed agglomerator allows the application of higher voltage
and therefore higher electric field strength because of the higher gas density. This higher field
strength can result in significant further improvement in precipitator performance. Figure 6
illustrates the importance of applied voltage in reducing precipitator effluents.
4. Tests showed that the fiyash bed itself accounted for about 50% reduction of an inlet
concentration of 50 ngm/m3 of elemental mercury vapor at 230ฐF. A single injection of
iodine-impregnated activated carbon, equaling about 1% of the bed weight, was shown to then
adsorb virtually all of the remaining mercury for longer than two hours, representing a carbon
capacity for mercury of about 770 gm carbon/ gm mercury. See Figure 7. These results are
extremely significant given the need for effective mercury control. Because adsorption
becomes more effective at lower temperatures, it is expected that lower temperature operation
will result in even higher capacities for both the fiyash and carbon.
5. A cost comparison shows that the capital cost of the addition of a fluid bed system for
particle agglomeration and subsequent capture of submicron particles is about 1/2 the cost of
the modifications needed for a conventional particulate control device in order to grant it the
same degree of submicron capture. The addition of lime to the fluid bed gives the device
sulfur dioxide removal as well at minimal extra cost. If activated carbon is used for mercury
control, then the superior utilization provided by the fluid bed, as opposed to once through
injection, allows a cost savings of to $140,000 per year.
-------
Future Work
A 1000 CFM scale up of the system described here is currently undergoing testing at
PSE&G's Mercer Station. In addition to characterizing the operation of the reactor on actual
flue gas, the effects of cohesivity additives and additional activated carbons will be evaluated.
Acknowledgments
This work is being carried out under a Department of Energy SBIR Grant. The DoE Project
Officer is Henry W. Pennline.
References
1. H. Arastoopour, et al, "Effect of Temperature and Gas Velocity on the Fluidization of
Sticky Particles", International Conference on Fluidization, May, 1986.
2. M Benlyamani, et al, "Agglomeration of Particles During Roasting of Zinc Sulfide",
Metallurgical Transactions, December, 1986.
3. R. Chang and C. J. Bustard, "Sorbent Injection for Flue Gas Mercury Control", Paper 94-
WA68A.01, 87th Annual Meeting of the A&WMA, June, 1994.
4. S. J. Miller, et al, "Laboratory-Scale Investigation of Sorbents for Mercury Control", Paper
94-RA114A.01, 87th Annual Meeting of the A&WMA, June, 1994.
5. C. D. Livengood, et al, "Experimental Evaluation of Sorbents for the Capture of Mercury
in Flue Gases", Paper 94-RA114A.04, 87th Annual Meeting of the A&WMA, June, 1994.
6. K.S. Kumar and P.L. Feldman, "Control of Air Toxics from Coal Burning Plants", 1996
ICAC Forum, March, 1996.
-------
1
0.9-
0.8-
0.7-
> 0.6-
1 o.5H
uJ 0.4-
0.3-
0.2-
0.1-
0
0 100 200 300 400 500 600 700 800 900 1000
BED DENSITY (GRAINS PER CUBIC FOOT)
'Figure 1.
Agglomeration Efficiency Versus Bed Density
0.1 micron Particles on 5 micron Targets
PRECIPITATOR
1
DRAW OFF
FLUID BED TOUER HEATER/HUMIDIFIER
]VALUE
Figure 2.
The Laboratory System
-------
LU
r\i
CO
Q
LU
C_J
5
CO
CO
LU
LJ
LU
CL
1 10
PARTICLE SIZE, MICRONS
100
Figure 3.
Inlet Dust Size Distribution
-------
i-t!^-12'ซI rT=p Cover
4'OJJ
Uppw Rซoctor Cylhdor -
r CySnder
Venhri, Single,
W_i
s ฅ
'
Wei Section-
Figure 4.
The Circulating Fluid Bed Reactor Vessel
-------
o
Cฃ
O
ce
D_
LL.
O
CC
The
CIRCULATING BED DENSITY, GR/CF
Figure 5.
Effect of Bed Density on Fine Particle Concentration
o
Cd
CJ
_J CJ
_ I CJ
co -
LU (S
Q_
L!_
O
cr
LU
CD
25
20
15
10
SYSTEM INLET
150 CFM, 300 DEG IN
o NO BED
. 200 CFM, 325'DEG IN
D NO BED
* 100 CFM, 383 DEG IN
0 50 100 150 200 250
FLUID BED OUTLET TEMPERATURE, 'DEG F
Figure 6.
The Effect of Evaporative Cooling on Fine Particle Concentration
-------
fjl
en
-=c
3.1
3.01
10 15 20 25 30 35
PRECIPITATOR VOLTAGE, KV
Figure 7.
Effect of Precipitator Voltage on Performance
o:
O
en e_>
en e_>
uj
s>
CO G>
LU -
11 X
[
o:
o.
fe
Cฃ
LU
CO
ADSORPTION BY ASH ONLY
50 100
150 200 250
MINUTES
300 350 400
Figure 8.
The Effect of Carbon Addition on Outlet Mercury Concentration
-------
Thursday, August 28; 1:00 p.m.
Parallel Session B:
Air toxics Control - Mercury Capture by Sorbents: Pilot Scale Studies
-------
DEMONSTRATION OF DRY CARBON-BASED SORBENT INJECTION FOR MERCURY
CONTROL IN UTILITY ESPS AND BAGHOUSES
S. Haythornthwaite
S. Sjostrom
T. Ebner, J. Ruhl, R. Slye, and J. Smith
ADA Technologies, Inc., 304 Inverness Way South, Suite 365, Englewood, CO 80112
T.Hunt
Public Service Company of Colorado, 550 15th St. Suite 800, Denver, CO 80202
R. Chang
Electric Power Research Institute, PO Box 10412, Palo Alto, CA 94303
T.D. Brown
U.S. Department of Energy, Federal Energy Technology Center, PO Box 10940, Pittsburgh, PA 15236
Abstract
Domestic coal-fired power plants emit approximately 40 to 80 metric tons of mercury to the atmosphere
annually, about 30% of all mercury emissions from human activities. However, the mercury
concentration in utility flue gas is in the extremely dilute range of 0.1 to 1 part per billion. The EPA is
assessing whether such low concentrations for mercury emissions from coal-fired utilities pose any
significant health risk and whether mercury regulations would be necessary or appropriate. In
anticipation of possible mercury control regulations, DOE has funded Public Service Company of
Colorado (PSCo) to evaluate carbon-based sorbents for mercury control at utility coal-fired power
plants.
Initial investigations of the use of dry carbon-based sorbent injection for mercury control on utility
applications have shown that carbon-based sorbents are capable of removing gaseous phase mercury.
Because of the difficulty in capturing and measuring mercury, however, it is important to evaluate these
technologies extensively on actual utility flue gas. Testing has been conducted over the past year on a
slipstream of flue gas from PSCo's Comanche Station in Pueblo, Colorado. The test fixture is a 600
acfm particulate control module that can be configured as an electrostatic precipitator, a pulse-jet
baghouse, or a reverse-gas baghouse. Sorbent is injected into the flue gas slipstream prior to the
particulate control module, and is removed by the module. Flue gas temperature and sorbent residence
time can be changed to evaluate a range of plant operating conditions. In addition the effect of flyash on
mercury capture can be evaluated because the flue gas slipstream can be taken from either upstream or
downstream of Comanche Station's full-scale reverse-gas baghouse. This paper presents the preliminary
results of a two year test Phase I program that is being funded by the U.S. Department of Energy, PSCo,
and the Electric Power Research Institute (EPRI). Phase n of this program will further establish the
reliability of carbon injection at coal-fired utilities.
-------
Introduction
Interest in determining economical methods of mercury removal on coal-fired utility boilers has been
increasing due to potential future regulatory requirements. This project tests the injection of activated
carbon in a unique test facility located at Public Service Company of Colorado's Comanche station. The
test facility may be configured as a pilot-scale ESP, pulse-jet baghouse, reverse-gas baghouse, or in an
EPRI-patented TOXECON configuration (a high air-to-cloth ratio, pulse-jet baghouse operating
downstream of a primary particulate collector). Preliminary results indicate that carbon injection can
obtain from 30 to 90% mercury removal at this site depending on operating conditions. However,
baseline mercury concentrations are so low that sampling methods are very important to collect accurate
data. Another interesting result is that significant mercury removal is obtained with no carbon injection
due to the properties of Comanche's flyash and other operating conditions.
Most trace metals in flue gas are normally present in the solid state, in the flyash particles, and can be
removed effectively by an efficient particulate collector. However, mercury forms a number of volatile
compounds and the extent of their removal across existing emission control devices is poorly
understood. Currently mercury removal is completed at municipal solid waste (MSW) incinerators using
various technologies. Sorbent injection of activated carbon has been used with good success at MSW
units (1,2) but mercury concentrations in MSW process gases are generally several orders of magnitude
higher than found in flue gas from utility boilers. The gas conditions and mercury species found at the
two types of facilities can also significantly differ. Factors such as gaseous HC1 concentration, SC>2
concentration, and oxidized mercury concentration have been found in laboratory studies to affect the
adsorption capacity of Norit Darco FGD carbon (3). Preliminary studies funded by DOE and EPRI on
utility flue gas indicate that sorbent injection may also be effective on these units (4,5). Sorbent
injection also represents a technology that is relatively easily retrofitted into an existing pollution control
system and may be installed at a lower cost than other technologies .
Before sorbent injection can be used on utility units, further data is required to document performance of
available sorbents using a variety of particulate control systems. Data is required to determine the
mercury removal possible, the economics, and the effect on operations for sorbent injection. The many
differences between utility flue gas and MSW flue gas necessitate research to confirm the commercial
viability and reliability of sorbent injection in the coal-fired flue gas environment. PSCo was selected by
the DOE to complete a pilot program and work began in October 1995. The project is approaching
completion of the first phase. This paper presents a description of the test facility and summary of the
results that have been evaluated to date.
Approach
The primary objective of this two-phase program is to provide a sound technical basis for the viability of
carbon sorbent injection for mercury control at coal-fired power plants. This objective was met in Phase
I by engineering, optimizing, and demonstrating dry sorbent injection on a coal-fired utility slipstream at
the pilot-scale. The project focus was to evaluate mercury removal performance of sorbents on actual
flue gas at a field test facility, assess the impact of the sorbent and the injection system on existing power
plant equipment, and analyze collected sorbent/ash to determine teachability.
The project team investigated the performance of mercury removal sorbents injected into a flue gas
stream from a coal-fired power plant. This was done by routing a slipstream of flue gas through a
-------
specially-designed particulate control test fixture where candidate sorbents were injected into the flow.
Gas samples drawn upstream of injection and downstream of the collection device were used to quantify
the removal rate of mercury from the slip stream. A separate mercury vapor injection system was
developed to permit a wider range of mercury concentrations to be tested. The test fixture has been
tested while configured as four different particle control devices: an electrostatic precipitator, a pulse-jet
baghouse, a reverse-gas baghouse, and TOXECON. The test matrix included experiments to evaluate
the effects of flue gas temperature, sorbent concentration, and sorbent type on mercury removal. In
addition, the impact of sorbent injection on the performances of the particulate control devices was
assessed.
Description of Test
Facility
The particulate control
module (PCM), its
connecting ductwork, and
associated instrumentation
were installed at PSCo's
350 MW Comanche Station
Unit 2 which burns Powder
River Basin coal from the
Belle Ayre mine in
Wyoming. The total vapor
phase mercury
concentration was expected
to be 7 ng/Nm3 in flue gas
at the test location.
Although the HC1
concentration was not
measured during the
program, little HC1 was
expected in the gas. The
SO2 concentration (@ 3%
O2, dry) was 275 to 325
ppm and the NOX (@ 3%
62, dry) concentration was
180 to 250 ppm.
Flue Gas
Inlet Sample 2
Carbon
Injection
(sevaral ports)
Q.
CD
ID
Variables
Velocity
Residence Time
Temperature
Outlet
Sample
Figure 1. Schematic of laboratory-scale test fixture.
The lab-scale test facility was designed and fabricated to permit significant control over the operating
conditions during sorbent evaluation tests. In addition to changing the particulate control configurations,
operating parameters such as flue gas flow rate, duct temperature, flue gas moisture content, sorbent duct
residence time, and flue gas mercury concentration could be controlled over ranges of interest. Sorbent
effectiveness was evaluated for temperatures from 200ฐF (expected cold weather baseline at Comanche)
to 325ฐF. Duct cooling was achieved by an air-to-air heat exchanger. Flue gas was heated with a duct
heater for selected tests. The sorbent injection ports were located to provide for in-duct sorbent
residence times of 0.75 to 1.5 seconds to allow evaluations of the impact of duct residence time on
sorbent effectiveness. An overall schematic of the test fixture is shown in Figure 1.
-------
The 30 foot high main collection vessel was incorporated
into the 8-foot by 10-foot framework shown in the photo
in Figure 2. The injection section and collection section
were built within the framework and were accessible
from platforms. A mast was installed on the tower to
allow configuration changes in the PCM without the
assistance of a crane. At most, two people were required
for major configuration changes in the PCM, such as
conversion from the electrostatic precipitator to the pulse-
jet baghouse.
Control System
The PCM control system was designed to allow manual
or automatic operation of the pilot. The primary control
elements for the pilot consisted of a Programmable Logic
Controller (PLC) and an intelligent data-logger.
Pneumatic actuators on the inlet, outlet, bypass, flow
control, purge, and hopper discharge valves permitted
automatic flow control, off-line cleaning, and isolation of
the pilot for shut-down. The control system was
programmed to bring the pilot off-line, clean the bags or
rap the plates, and purge the system for alarm trip
conditions. Trip conditions included low boiler load and
low duct temperature for all configurations and high duct
Figure 2. Photograph of paniculate control temperature and high tubesheet pressure drop during the
module installed at Comanche Station fabnc fllter tests' The baง cleamng or Plate raPPinง
sequence could be initiated automatically or controlled
manually at the control panel. Monitored and recorded
parameters included: gas temperatures, flowrate, pressures, boiler load, secondary voltage and current
(ESP), cleaning/rapping frequency, and pulse pressure or reverse gas flow (fabric filters). Data were
stored in time-stamped arrays for transfer to other software for analysis and graphical presentation.
Electrostatic Precipitator Configuration
The ESP was a wire-tube type unit designed to treat 620 acfm. This flowrate resulted in a velocity of 5
ft/sec through the 20 foot long ESP collection section. The specific collection area (SCA), a standard
measure of collection area to total gas flow, at these operating conditions was 327 ft2/Kacfm. This SCA
was selected because it was representative of many ESPs installed at utilities in the United States.
Four 10-inch diameter collection tubes, the gas passages for the ESP, were hung from a tubesheet at the
top of the 28-inch diameter collection vessel housing. Four high-voltage electrodes, one situated on the
centerline of each gas passage, were attached to a rigid frame and powered from a single transformer-
rectifier (T/R) set. The lower frame was weighted to keep the wires straight and a pneumatic vibrator
was attached for cleaning ash from the electrodes. The top frame was attached to the high voltage bus at
the feedthrough insulators. The T/R set was controlled by an automatic voltage controller (AVC) and
was set to simulate conditions in a full-scale wire-plate ESP. For these tests, the T/R set was operated in
-------
the 40 KV, 15 mA range (at 15 mA, the current density is 80 nA/cm2). Comanche Station bums a
Powder River Basin coal, which typically causes problems with back corona on full-scale ESPs because
of the high resistivity flyash it produces. The ESP T/R controls were set for intermittent energization,
which successfully quenched the back corona in the pilot.
Pulse-Jet Baghouse Configuration
The pulse-jet was designed to filter 628 acfm flue gas at an air-to-cloth ratio of 4 ft/min. To achieve this
ratio, six 20-foot long full-scale bags were hung from the pulse-jet tubesheet. Full-scale bags were used
to better simulate the filtering and cleaning characteristics experienced in a full-scale unit. The bags
were sealed to the tubesheet by means of a metal snap band and a double beaded gasket sewn into the
top of the bag. A rigid steel wire cage was inserted into each bag. Flue gas entered the bag compartment
at the bottom and passed through the bags from outside to inside, depositing the particulate matter on the
outside of the bags. The flue gas then flowed out of the compartment through the outlet plenum on the
clean side of the tubesheet. Flue gas in the outlet duct passed an annutube flow sensor, the flow control
damper, and through a section of duct located beneath the ash hopper. Ash from the hopper was fed into
this duct section by a rotary valve. The particulate-laden gas then returned to the host duct. The bags
were cleaned with a pulse of compressed air from pulse pipes located above each row of three bags.
Reverse-Gas Baghouse Configuration
The PCM was configured as a reverse gas baghouse by installing a cell plate with seven 8-inch diameter
holes near the bottom of the PCM housing. Each bag was attached to the cell plate by a metal snap band
and a fiberglass double-beaded gasket sewn into the bottom of the bag. The 21-foot long, 8" diameter
full-scale fiberglass bags were sealed at the top by a metal bag cap. The caps were attached to tensioning
springs at the top of the PCM and the bags were pre-tensioned to a load of approximately 35 Ibs. Flue
gas entered the bag compartment from the bottom and passed through the cell plate into the interior of
the bags. The gas then flowed from inside to outside of the bags, deposited the ash on the inside of the
bags, and exited the compartment via the outlet plenum. The bags were cleaned by reversing gas flow
across the bags from outlet to inlet causing the bags to gently collapse, thus breaking off the dust cake
collected on the inside of the bags. The ash fell into the ash hopper at the bottom of the compartment.
The PCM system used Comanche's hot, clean, dry preheat air for reverse-gas. During a clean, automatic
valves were actuated to close the outlet duct of the PCM and open the reverse gas line. This allowed
reverse gas to enter the compartment through the outlet plenum. Cleans were initiated when pressure
drop across the bags exceeded a threshold level.
TOXECON Configuration
TOXECON is a pulse-jet baghouse with sorbent injection for air toxics removal operating at a high A/C
ratio downstream of a primary particulate collector. EPRI has patented the TOXECON technology.
This configuration for the PCM was designed to filter 633 acfm of flue gas at an air-to-cloth ratio of 16
ft/min. The target operating air-to-cloth ratio for these tests was 12 ft/min, which meant that the flow
was somewhat below the design value. To achieve this ratio, two 15-foot long bags were hung from the
TOXECON tubesheet and the PCM filtered 470 acfm of flue gas. An annulus was installed to increase
the can velocity (upward gas velocity in the vessel on the dirty-side of the tubesheet) to approximately
900 ft/min at an A/C ratio of 12 ft/min to better simulate the flows in a full scale unit. The injected
-------
sorbent was collected on the exterior of the felted bags, as in a full-scale unit. Because TOXECON is
designed for installation downstream of a primary paniculate collector, the PCM version was configured
to draw flue gas downstream of the existing Comanche baghouse. The operation of TOXECON is
similar to a conventional pulse-jet baghouse except that cleaning is initiated by a timer and the bags are
cleaned off-line.
Mercury Measurement Technique
When sampling particulate-laden gas in all configurations except the ESP, an isokinetic sampling system
was used to collect a representative sample of ash along with the gas. This system consisted of a
standard particulate filter, used in EPA methods 5 and 29, with the addition of a glass cyclone upstream
of the sampling filter to collect the flyash. The cyclone removed a large fraction of the flyash, thus
minimizing the contact of the flue gas and flyash. The vapor mercury measurements were made with a
modified Mercury Speciation Adsorption (MESA) method drawing a particulate-free sample from
downstream of the isokinetic ash sampling system. The modified MESA train consisted of the last two
traps (primary and backup) of the full MESA train. These final two traps were filled with iodated
carbon, which adsorbed all forms of mercury at the temperatures tested. Following sampling, the traps
were analyzed for mercury content using cold vapor atomic fluorescence spectroscopy (CVAFS) (6).
The mercury concentrations measured with the traps were reported as total vapor phase mercury.
Use of a particulate filter for sampling is not an ideal arrangement to quantify the fraction of mercury on
the flyash and in vapor phase because Comanche flyash adsorbs mercury and it is likely that forcing the
flue gas through a fixed bed of this flyash (i.e. EPA method 29 or similar sampling filter) will probably
increase the amount of mercury collected on the ash and bias low the vapor phase measurements.
Therefore, mercury measurements are reported as total mercury and include the sum of the mercury
captured with the flyash and the mercury captured in the iodated carbon trap. Samples collected during
sorbent injection were compared to baseline (no sorbent injection) to assess a sorbent's effectiveness as
an incremental removal of mercury from the flue gas stream. For some test conditions, significant
removal of mercury by native Comanche coal flyash was observed.
Mercury Doping System
A mercury doping system was designed to increase the elemental mercury vapor into the duct feeding
flue gas to the particulate collector. The fundamental design of this doping system was based on
previous systems designed on other DOE programs. Nitrogen was passed at a constant rate over liquid
mercury in a temperature controlled container. The mercury concentration in the gas exiting the vessel
was determined by the temperature of the vessel and the nitrogen flowrate. This system was used on
some pulse-jet, reverse-gas and TOXECON tests when there was little flyash present in the inlet gas
stream and the inlet mercury concentration was expected to be below 2 |ig/m3.
Mercury Sorbents
Both physical and chemical properties of the sorbent are important. The sorbent particle size is
important for mass transfer. Effective sorbents have large specific surface area, which provides
extensive sites for adsorption on the surfaces of sorbent particles. Specific surface area, expressed as
area per unit mass, is typically measured by the B-E-T (Brunauer - Emmitte - Teller) method.
-------
The two primary sorbents selected for testing were Darco FGD activated carbon from American Norit
(referred to in this report as Norit carbon) and an experimental carbon identified as AC-1. These two
sorbents showed the most promise in EPRI-sponsored laboratory research. A third sorbent, iodine-
impregnated carbon, was evaluated in two test configurations. This sorbent is much more expensive to
produce than Norit or AC-1, and was selected for comparison with the primary sorbents.
Norit is an activated carbon derived from lignite and is used to remove mercury in municipal solid waste
(MSW) combustors in Europe and the United States. It has also been used in a few utility mercury
removal tests including previous tests at Comanche Station. The second tested sorbent, AC-1, is an
activated carbon prepared from a bituminous coal. The mass mean diameter of the sorbent samples, as
measured by aMicrotrac laser scattering particle sizer, were 14.6 microns for Norit, 13.1 microns for
AC-1, and 14.7 microns for the iodine-impregnated activated carbon. The BET surface area of the
sorbents as measured by the manufacturers were 600 m2/g for Norit and 400 - 600 m2/g for AC-1. The
size distributions of the three sorbents were very similar and should provide reasonable comparison data.
Preliminary Results
Testing was conducted with four particulate collectors. Collector temperature during testing ranges from
200 ฐF to 317 ฐF. Activated carbon sorbents were injected at sorbent concentrations in the flue gas of 0 -
6 Ib/MMacf, which correspond to injection ratios up to 16,000:1 wt carbon:wt mercury. The inlet
mercury concentration during "dirty" flue gas testing, where the gas was extracted for upstream on
Comanche's full-scale baghouse, was measured at 3.8 -10.7 |0.g/Nm3 during the 12 months of testing.
When testing on "clean" flue gas from downstream of Comanche's baghouse, an elemental mercury
doping system was used. The inlet mercury concentrations downstream of mercury doping were
measured to be 1.8 -17 u.g/Nm3 during testing.
The typical test duration was two to four hours, during which duplicate gas samples for mercury
concentration were obtained at the inlet and outlet of the test facility. Test conditions were held constant
during each test. The facility had the capability to heat or cool the flue gas stream and to vary the flow.
As much as possible, test conditions were duplicated for the several particle control device
configurations in order to facilitate comparisons.
A brief test summary and the results are presented for each configuration in which the PCM was tested.
The mercury removal as a function of sorbent concentration in the flue gas (Ibs/MMacf) is plotted for
each configuration. Another important variable noted on many of the graphs is the flue gas temperature.
ESP Configuration
Two sorbents, Norit and AC-1 activated carbon, were tested at sorbent concentrations from baseline
conditions (no sorbent) to 2.0 Ibs/MMacf at temperatures from 200 to 290 ฐF. Isokinetic ash sampling
was not employed during these tests.
Two concerns with carbon injection upstream of an ESP were possible detrimental changes in ESP
power levels and potential increases in outlet particulate emissions. Although no long-term carbon
injection tests were conducted, results from the short-term tests indicate that the ESP power levels did
not change significantly as a result of carbon injection. These results should be interpreted with caution
until longer-term data is available. Predictions about increased outlet particulate emissions due to
-------
carbon injection from these during short-term tests are difficult. Longer term testing is being planned as
part of the Phase n portion of this program.
Another issue with carbon injection into an ESP is an increase in the measured loss on ignition (LOT) of
the flyash. The inlet particulate loading to the pilot ESP at this site was approximately 1 gr/acf. A
carbon sorbent concentration of 1 Ib/MMacf would increase the measured LOT by nearly 1%. Some sites
sell their flyash and the price of the ash is often determined by the LOI. An increase in LOI may reduce
the market price of the ash or make it difficult to sell the ash. A possible method to reduce the LOI of
ash, whether due to carbon injection or unburned coal, is the emerging technology of ash separation. At
least one of the new separation technologies will be explored in Phase n of this program.
ESP evaluations were conducted in three temperature regimes: 200-220 ฐF, 230 - 260 ฐF, and 275 - 290
ฐF. The sorbent concentrations in the flue gas are shown as pounds per million actual cubic feet
(MMacf) for better comparison with tests conducted at other facilities and to facilitate calculations for
scale-up. These sorbent concentrations shown correspond to injection ratios (wt carbon per wt mercury)
of 800-8000:1.
Preliminary mercury removal results using the Norit and AC-1 sorbents with the ESP are summarized in
Figure 3, which presents mercury removal as a function of sorbent concentration in the flue gas in
Ib/MMacf. The data presented indicates a general increase in mercury removal with increased sorbent
concentration. Note that the mercury removal shown in the graph is for the overall vapor phase mercury
across the PCM, and not just the mercury removal that occurred due to sorbent injection. The presence
of flyash contributes to the mercury removal as indicated by the baseline data (zero injected sorbent)
which shows significant mercury removal. It is believed that the modified MESA method without
isokinetic ash sampling used for this set of tests did not accurately report the total particulate and vapor
phase mercury in the extracted sample. Therefore, the mercury removal shown is considered to be only
that in the vapor phase at the sampling point plus any mercury that had been sorbed onto the flyash
collected in the sampling train.
ฃ
n
o
1
DC
e-
o
s
i uu -
90 -
80
70 -
60
50 -
40
30 -
20 -
10
n
A
X *
g
*
1 1 1 1
Baseline
Norit (MMD - 25 urn)
A AC-1 Fine (MMD - 3 urn)
x AC-1 Course (MMD - 13 urn)
0.0 0.5 1.0 1.5 2.0
Sorbent Concentration (Ib/MMacf)
2.5
Figure 3. Summary of ESP mercury removal results.
-------
Baseline (no sorbent injection) removals of vapor phase mercury of 11 to 36% (average approximately
30%) are shown in Figure 3. The mercury fractions found on flyash captured by the MESA paniculate
filter were 2 to 37% of the total mercury collected in the sampling train. This flyash was collected non-
isokinetically and may not have been representative of the paniculate matter present in the flue gas.
Although the mercury concentration in the paniculate fraction may be biased due to the sampling
method, the data strongly suggests that mercury is adsorbing onto the flyash. Thus, while the mercury
removal totals provide data for analysis, they probably do not reflect an accurate measurement of the
sorbent's ability to remove mercury. In order to obtain a rough estimate of the sorbent's mercury
removal capability, an average baseline (no sorbent) vapor phase mercury removal of 30% was used to
estimate the ability of the sorbent to remove mercury. For example, the 1.2 Ib/MMacf test 15 data point
reports about 58% vapor Hg removal. For an assumed 30% flyash Hg removal, the sorbent was
calculated to have removed 40% of the remaining vapor phase mercury.
Mercury removal was not reported as a function of temperature for the ESP tests. Later testing showed
that the uptake of mercury by flyash in the flue gas is a strong function of temperature. However, the
absence of isokinetically collected particulate-bound mercury samples masks the effect of temperature
on mercury measurements for this series of tests. Several tests were recently repeated with isokinetic
sampling, however, this data is not yet available.
Pulse-Jet
Due to the observed affinity of Comanche's flyash for vapor-phase mercury, the project team decided to
test with both "clean" and "dirty" flue gas in the pulse-jet configuration by using flue gas obtained from
Comanche's full-scale fabric filter's outlet and inlet, respectively. Flyash was collected isokinetically
during mercury sampling tests with particulate-laden flue gas, and the ash was analyzed for mercury
content. These tests were conducted at temperatures from 210 to 275 ฐF and sorbent concentrations up
to 2.1 Ib/MMacf. Elemental mercury vapor was doped into the flue gas for the "clean" gas tests.
"Clean" gas tests were conducted at temperatures from 260 to 300 ฐF with sorbent concentrations up to
5.7 Ib/MMacf. Acrylic bags were used during all tests conducted at temperatures below 275 ฐF during
the "clean" flue gas tests. Ryton bags were installed in the unit to facilitate testing at temperatures above
275 ฐF and remained in service during dirty gas tests.
Pulse jet performance data indicated that carbon sorbent injection did not affect tubesheet pressure drop
or cleaning frequency. This is not surprising during "dirty" gas testing since the sorbent, injected at 1
Ib/MMacf, comprised only 1% of the total particulate matter in the flue gas. In addition, the sorbent
particle size had a mass mean diameter of 13-14 microns, which was comparable to flyash. A very fine
sorbent may result in higher pressure drop at lower concentrations due to infiltration into the fabric and
the formation of a less permeable dustcake. The presence of sorbent during "clean" testing did not result
in an increased tubesheet pressure drop either. This was also likely due to the low total particulate
loading (0.007 grains/act equals 1 Ib/MMacf) and the associated filtering air-to-cloth ratio of 4 ft/min. A
common fabric filter pressure drop model based on the Darcy equation suggests that the increase in
pressure drop between cleans is less than 0.005 inches HiO per minute for this operating condition.
Testing on Low Ash, "Clean" Flue Gas. Figure 4 is a summary of the mercury removal data as a
function of sorbent concentration taken during "clean" pulse-jet tests. The figure includes baseline data
(no sorbent injection), and injection of Norit and AC-1 carbons. Three constant-temperature contours
are shown on the graph to mark trends for the average removal data points at the specified temperature
with no injection, and at various Norit sorbent concentrations in the flue gas. The data suggest that the
-------
mercury removal improves at lower temperatures. There was little difference noted in the performance
of the AC-1 carbon as compared to the Norit carbon during these tests.
As shown in the figure, 0 to 25% mercury removal was measured during baseline conditions (no sorbent
injection). At carbon sorbent concentrations below 1 Ib/MMacf, 28 - 78% of the mercury was removed
at temperatures below 260ฐF. A maximum mercury removal of 93% was measured at 5.7 Ib/MMacf.
The high sorbent concentration corresponds to a nominal injection ratio (carbon:mercury weight ratio) of
16,000:1.
A linear regression analysis was conducted on the Norit data obtained from "clean" gas testing. The
parameters included in the regression included temperature, sorbent concentration, and pressure drop
across the fabric. Temperature and sorbent concentration had a statistically significant influence on
mercury removal. Differential pressure was not found to be statistically significant to the removal of
mercury during these tests. The pressure drop across the tubesheet was quite low, less than 1 inch B^O,
for all "clean" tests and it is believed that pressure drop would influence mercury removal when a more
representative dustcake is present on the fabric.
^^
0^
*" '
flj
g
rcury Re
CD
IVJU -
90 -
80 -
70 -
60 -
50 -
40 -
30 -
<
20 -
1Qc
0 if
260 ฐFป A 275 ฐF o>
* -* **ซ* _ *"~ " ~
*'~~~ * ~"~ ^
" . '* ~ '""
-i* X ฐ ฐ
I I '
-\ ^
I 1 '" 0
1 *
-1 /
' / ,' 300 ฐF
f / "
1 /
i /
i/
[/
\
O i i i ,
^ Baseline 260
o Baseline 275
o Baseline 300
A Norit 260
0 Norit 275
o Norit 300
+ AC-1, 270
xAC-1, 280
0123456
Sorbent Concentration (Ib/MMacf)
Figure 4. Total Mercury Removal, Pulse-Jet Configuration, No Flyash
Testing on Typical Ash, "Dirty" Flue Gas. During "dirty" pulse-jet testing, mercury samples were
collected isokinetically at the inlet to the pulse-jet; this conventional "dirty" flue gas was extracted from
the plant ductwork prior to Comanche's full-scale fabric filter. Each sample was analyzed for both
paniculate and vapor-phase mercury. Sampling at the inlet of the pulse-jet indicated that 25 to 58% of
the mercury captured in the sampling train was present on flyash collected in the cyclone and on the
filter of the isokinetic sampling system. The cyclone/filter assembly was maintained at 250 ฐF for these
tests.
-------
The data set from pulse-jet testing with "dirty" gas showed almost no data scatter in repeated tests under
similar conditions. Removals from 73 to 78% with 0.71 Ib/MMacf sorbent injected and from 86 to 90%
with 2.1 Ib/MMacf sorbent injected were obtained over four tests at each sorbent concentration. When
no sorbent was injected (baseline), and the flue gas temperature was below 250 ฐF , 65 to 67% removal
was measured in three tests. It is notable that there is very little data scatter for both sorbent
concentrations of Norit and AC-1, collected at 230 ฐF, and that the data trend defined by the sorbent
injection tests is consistent with the baseline data (taken at 200 and 250 ฐF). It is estimated that the
resulting mercury removal attributed to sorbent injection alone (flyash contribution removed
mathematically) is 20 to 35% for the lower sorbent concentration tested and 60 to 70 % for the higher
sorbent concentration. Baseline data collected at 275 ฐF indicate 10 and 17% mercury removal during
these duplicate samples. This data suggests that mercury adsorption onto flyash is a function of
temperature, as appears to be the case with the Norit and AC-1 carbon sorbents.
Comparing data collected with "dirty" flue gas and "clean" flue gas suggested that the fractional removal
of mercury as a function of sorbent concentration appeared to be independent of the presence of
particulate matter in the flue gas, or of the starting gas-phase mercury concentration. This observation
implied that the mercury uptake was controlled by gas phase mass transfer. During low-ash tests, the
elemental mercury concentration in the flue gas was increased by doping. The resulting mercury
concentrations collected on the iodated carbon sampling traps (representing the gas-phase mercury
concentration) were as much as a factor of ten times higher than for the tests conducted in the presence
of flyash. Based on data presented in Figure 4 and that discussed above under "dirty" flue gas testing, it
appeared that under similar temperature conditions, the mercury removal efficiency across a pulse-jet
fabric filter in the presence of 100 to 150 Ib/MMacf of Comanche's flyash (0.7 to 1.1 gr/acf) and no
sorbent was comparable to the mercury removal in the presence of little flyash and 0.3 Ib/MMacf Norit
or AC-1.
Reverse-Gas
Mercury removal evaluations were conducted on "clean" gas at PCM temperatures from 271 - 278 ฐF
with Norit activated carbon sorbent concentrations of 0 - 1.7 Ib/MMacf. Testing was conducted on
"dirty" gas at PCM temperatures from 269 - 317 ฐF and with Norit and AC-1 sorbent concentrations of 0
- 4.8 Ib/MMacf.
With a clean-initiate setpoint of 5 in. H2O tubesheet differential, the PCM was cleaning every 2 hours.
The addition of activated carbon did not affect the rate of increase of differential pressure across the
bags. This was expected because the carbon injected comprised a small fraction of the total particulate
entering the baghouse.
"Clean" Flue Gas Testing. Testing in the reverse-gas configuration with little flyash present was
somewhat unusual because the dustcake itself forms the primary filter in a reverse-gas design. To
minimize the amount of sorbent passing through the fabric, a commercially available precoat material
made from inert alumina silicate was used to form a filtering dustcake for these tests.
The few data sets suggest that up to 90% mercury removal can be achieved with a sorbent concentration
in the flue gas of 1 Ib/MMacf Norit activated carbon. However, lower removal was recorded at higher
carbon sorbent concentrations. The PCM operating conditions were quite similar during these seven
tests. It is difficult to draw any conclusions from the available data as to why the mercury removal
-------
appeared to decrease with increased sorbent concentration. Further testing would be required to
determine if this is a repeatable condition.
"Dirty" Flue Gas Testing. Following "clean" flue gas tests, the inlet dampers were adjusted to
provide the PCM with flue gas from upstream of Comanche's full-scale baghouse. Carbon injection of
0.5 - 5 Ib/MMacf at temperatures from 294 - 320 ฐF resulted in 47 to 90 % mercury removal across the
baghouse. This data is presented in Figure 5.
Analysis of the inlet mercury samples (ash and MESA trap) indicated a significant fraction of the total
mercury was particulate-bound at inlet sampling location temperatures below 280 ฐF. Samples collected
at temperatures at or below 286 ฐF contained particulate-bound mercury concentrations above 4.3
|ig/Nm3. The largest particulate-bound mercury concentration measured at sampling temperatures above
286 ฐF was 1.96 |ig/Nm3 Vapor phase mercury measurements ranged from 1.0 - 4.3 ug/Nm3 during the
lower temperature tests and 1.9 - 9.5 (Ig/Nm3 during the higher temperature tests.
iuu -
90 -
-7 80 -
6?
r 70 -
ra
g 60 -
ฃ
ง 50 -
DC
ฃ> 40 -
3 |
ฃ 30 -
0)
S 20 -
10 -
O
*
OA
4*
i
0 -1
0
8
A
A
A
A
i
A
Baseline 268F
Baseline 31 OF
A Norit 294F
A Norit 316F
uAC-1 296F
AC-1 31 8F
1 1 1 1 1
2 3 4 5 e
Sorbent Concentration (Ib/MMacf)
Figure 5. Mercury removal during "dirty" gas reverse-gas tests.
A stepwise linear regression analysis was performed on this data set to determine the factors influencing
mercury removal. The analysis showed that sorbent concentration and PCM temperature were the
predominant effects, as expected from previous tests. Higher temperatures resulted in lower mercury
removal and higher carbon sorbent concentrations resulted in higher mercury removals. Another
parameter that was included in the regression was the pressure drop across the bags, which was
influenced by the amount of ash and carbon on the bags. This pressure drop also produced a statistically
significant effect on mercury removal.
TOXECON
There is very little fly ash present in the flue gas downstream of Comanche's full-scale baghouse. When
the PCM was configured as a TOXECON unit operating on this "clean" flue gas, the dustcake on the
-------
bags accumulated slowly. The cleaning logic was set to clean the bags once per day which maintained
the pressure drop across the fabric below 2 inches H2O. When a sorbent such as carbon or sodium was
injected into the baghouse, a distinct change in the rate of pressure drop increase across the fabric was
noted. The baseline AP/At approached 0 inches H2O/min. At a Norit sorbent concentration of 0.5
Ib/MMacf Norit, the AP/At was an average of 0.002 inches H2O/min. This is roughly an additional 0.1
inch H2O per hour. At a Norit sorbent concentration of 2.0 Ib/MMacf, the pressure drop increased 0.007
inches H2O/min, or an additional 0.4 inches H2O per hour. At these carbon sorbent concentration on a
"clean" flue gas stream, minimal cleaning would be required to maintain the pressure drop across the
fabric below a reasonable 5 inches H2O.
Carbon-based sorbent s were evaluated in three PCM temperature ranges of 240-250 ฐF, 270 - 290 ฐF,
and 310 - 313 ฐF. During the lowest temperature tests, the mercury removal reached an average of 94%
at a sorbent concentration of 0.47 Ib/MMbtu. Little additional removal was achieved by increasing the
sorbent concentration at these temperatures.
During the 270-290 ฐF tests, a single data pair suggested that significant removal can also be achieved at
a low Norit sorbent concentration of 0.49 Ib/MMbtu. However, a closer examination of this data
suggested that the high removal may be a result of residual carbon remaining on the filter bags from an
earlier test. This data suggested potential for intermittent carbon injection with TOXECON where little
cleaning is required because of the extremely low inlet flyash load. At a sorbent concentration of 2.4
Ib/MMacf, 84% mercury removal was measured in this temperature range.
The results from Norit injection at the higher temperatures indicated little mercury removal was possible
at a sorbent concentration of 0.6 Ib/MMbtu. A maximum average mercury removal of 56 % was
achieved at an maximum sorbent concentration of 2.47 Ib/MMbtu at the higher temperatures.
Iodine-impregnated activated carbon was tested in the 270-290 ฐF temperature range and the data
indicates that mercury removal with this sorbent is similar to the removal achieved with Norit activated
carbon.
0.5
1.0 1.5 2.0
Sorbent Concentration (Ib/MMacf)
2.5
3.0
o None 250 F
None 280 - 290 F
n Norit < 250 F
Norit 270-280 F
Norit > 300 F
A lod C 280 - 290 F
Figure 6. Mercury removal in TOXECON.
-------
Mercury in Coal and Ash
In addition to measuring the mercury in the flue gas, two coal samples and two PCM hopper ash samples
were collected and analyzed for mercury concentration. The average mercury concentration in the coal
samples tested was 60 ng/g. Calculations based on plant operating conditions suggested that when this
coal was burned, if all the mercury were deposited on the flyash, the mercury concentration in the ash
would be around 2000 ng/g. The two ash samples contained mercury concentrations of 3074 ng/g and
1819 ng/g. These were collected in the ESP hopper at temperatures below 210 ฐF. This was not an
attempted mass balance for mercury in the plant, but was used to indicate that the mercury collected with
the flyash represents a significant fraction of the mercury in the flue gas at these lower temperatures. In
addition, a grab sample was collected from the full-scale reverse-gas baghouse hopper and analyzed for
mercury. Results from the analysis showed 1035 ng/g mercury in the ash sample. This indicated
significant mercury removal was occurring within the full-scale baghouse.
Waste Characterization
The EPA classification of the collected sorbent and flyash mixture is of great concern in identifying
disposal options. If the combined sorbent-flyash product collected in the particulate collector hopper
remains in a non-hazardous category, it can be handled and disposed of using methods currently
employed to dispose of flyash. Samples collected and analyzed during ESP, pulse-jet, reverse-gas and
TOXECON testing at Comanche indicate that the sorbent-flyash material remains non-hazardous. The
samples were analyzed by TCLP (toxicity characteristic leaching procedure) which showed that all 8
RCRA elements of concern were well below regulatory limits. In fact, the levels of most metals
including mercury were below analysis detection limits.
Preliminary Conclusions
ESP testing resulted in mercury removal of up to 74% with injection ratio (mass carbon: mass Hg) of
less than 10,000:1. This result may be unique to the Comanche facility, given some of the
characteristics of the operating conditions and ash. Further tests in Phase n will help determine
whether these results can be scaled up, and whether similar performance can be obtained at other
sites.
Baseline (no sorbent injection) mercury removal during ESP testing is attributed to a combination of
sorption by Comanche's flyash and a temperature drop across the particulate control module.
Sorption of mercury by the flyash significantly increases the importance of collecting a
representative particulate samples and analyzing the particulate and vapor-phase mercury to
determine mercury removal efficiency. We devised an isokinetic sample train for use on the high-
ash inlet tests which reduced data scatter in the results.
Comanche's flyash appears to adsorb mercury at temperatures below 280 ฐF, with mercury removal
increasing with decreasing temperature.
Data from baghouse tests (pulse-jet, reverse-gas and TOXECON) at temperatures from 260 to 305 ฐF
suggest that mercury removal improves significantly with lower flue gas temperatures in this range.
This is apparent from total mercury measurements across the fabric filter and from particulate-bound
mercury fractions collected at the inlet to the PCM.
-------
Mercury removal up to 90% was achieved in all baghouse tests. The carbon sorbent concentration
required for this mercury removal approaches 5 Ib/MMacf (-25,000:1) at temperatures above 280 ฐF.
At temperatures near 250 ฐF, this mercury removal may be achieved at ratios near 2 Ib/MMacf
(closer to 10,000:1).
Comanche's flyash was found to sorb mercury under certain temperature and operating conditions.
In the pulse-jet baghouse configuration, the normal inlet loading of 100 to 150 Ib/MMacf flyash
performed similarly to sorbent concentration of 0.3 Ib/MMacf of carbon sorbent (with no flyash).
No significant differences were noted in the mercury removal performance of AC-1 and Norit
activated carbons. One test with iodine impregnated carbon in the TOXECON configuration also
resulted in mercury removal rates similar to Norit and AC-1. This infers that the reaction is gas-
phase mass transfer controlled.
During these short-term PCM evaluations, activated carbon injection did not impact power levels in
the ESP or tubesheet pressure drop in the pulse-jet or reverse-gas configurations.
Acknowledgments
The author's would like to recognize the contributions of George Crater and Norm Haberkorn of ADA
for their assistance designing and fabricating the pilot facility. Jean Bustard and Michael Durham of
ADA Environmental Solutions were key technical advisors during the project. Jim Butz assists with
engineering and technical editing.
The author's appreciate the assistance of Jim Weller, plant manager, Mark Andorka and the remaining
staff at Public Service Company of Colorado's Comanche Generating Station.
References
1. Brna, T.G. (1991). "Toxic Metal Emissions from MWCs and Their Control", in Proceedings of the
2nd International Conference on Municipal Waste Combustors, Tampa, FL, April.
2. Nebel, K.L., and D.M. White (1991). "A Summary of Mercury Emissions and Applicable Control
Technologies for Municipal Solid Waste Combustors", a report prepared for the Standards
Development Branch of the US EPA, September.
3. Carey, T.R., O.W. Hargrove, C.F. Richardson, R. Chang, F. Meserole (1997). "Factors Affecting
Mercury Control in Utility Flue Gas Using Sorbent Injection," presented at the 90th Annual Meeting
of the Air and Waste Management Association, Toronto, Ontario, June 8-13.
4. Chang, and Bustard (1994). "Sorbent Injection for Flue Gas Mercury Control", Paper No. 94-
WA68A.03, presented at the 87th Annual Meeting of the Air & Waste Management Association,
Cincinnati, OH, June 19-24.
5. Miller et al, (1994). "Laboratory-Scale Investigation of Sorbent for Mercury Control", Paper No. 94-
RA114A.01, presented at the 87th Annual Meeting of the Air & Waste Management Association,
Cincinnati, OH, June 19-24.
6. Prestbo and Bloom. "Mercury Speciation Adsorption (MESA) Method for Combustion Flue Gas:
Methodology, Artifacts, Intercomparison, and Atmospheric Implications", Water, Air, and Soil
Pollution, 1995, Volume 80: 145-158.
-------
Mercury Control in Utility ESPs and Baghouses
through Dry Carbon-Based Sorbent Injection
Pilot-Scale Demonstration
E. G. Waugh, B. K. Jensen, L. N. Lapatnick, F. X. Gibbons
Public Service Electric and Gas Company
80 Park Plaza, 16G
Newark, New Jersey 07102
S. Sjostrom, J. Ruhl, R. Slye
ADA Technologies, Inc.
304 Inverness Way South, Suite 365
Englewood,CO80112
R. Chang
Electric Power Research Institute
Abstract
The mercury concentration in utility flue gas is in the range of 0.1 to 1 part per billion.
The EPA is assessing whether such low concentrations of mercury emissions from coal-
fired utilities pose significant health risk and whether mercury regulations would be
necessary or appropriate. In anticipation of possible regulations that would impose
control on mercury emissions, Public Service Electric and Gas Company (PSE&G) has
joined with the Electric Power Research Institute (EPRI) to evaluate carbon-based
sorbents for mercury control at PSE&G's coal-fired power plants.
Activated carbon is currently injected into municipal solid-waste combustor (MWC) flue
gas streams to reduce vapor phase mercury concentrations. This technology has not been
rigorously tested on flue gas from coal-fired utility boilers. Differences in flue gas
composition, and difficulties encountered in capturing and measuring the low
concentrations of mercury present in utility flue gas prevent the direct transfer of
knowledge from MWC applications to coal-fired utilities. Testing of sorbent injection
technology is currently underway on a slipstream of flue gas from Hudson Unit 2 located
in Jersey City, New Jersey. The boiler is a dry-bottom, supercritical, once-through
design, firing a low sulfur bituminous coal. The test facility (pilot) includes a 160 acfm
wire-tube electrostatic precipitator (ESP) and a 4000 acfm transportable pulse-jet
baghouse operating in the COHPAC configuration . This paper describes the results of
initial pilot testing of the activated carbon injection technology for mercury capture.
-------
Introduction
Electric Power Research Institute (EPRI) surveys and pilot plant demonstrations have
shown the applicability of pulse-jet baghouses to a variety of coals. Upon reviewing
available test data, Public Service Electric & Gas (PSE&G) expressed interest in
evaluating the potential of pulse-jet baghouses to reduce particulate and mercury
emissions at the Hudson Generating Station in Jersey City, New Jersey. A research
project jointly-funded by EPRI and PSE&G was created to conduct the evaluation. ADA
Technologies was contracted to take the lead in operating the pilot plants at the Hudson
station.
The objective of the project described in this paper was to assess promising options for
fine particulate and air toxics (including mercury) control utilizing particulate collectors
such as an electrostatic precipitator and pulse-jet baghouse. Activated carbon sorbents
were evaluated for mercury control in an electrostatic precipitator and a pulse-jet
baghouse in the COHPAC configuration (CQmpact Hybrid RArticulate Collector, a
patented EPRI technology). The EPRI Transportable Pulse-Jet 1 MW(e) pilot unit (TPJ)
and a 160 acfm EPRI ESP pilot unit were utilized as the particulate control devices. This
paper presents a summary of the operating results and performance analysis completed
through March 1997.
The Hudson Generating Station consists of three generating units, one coal and/or gas-
fired, one oil and/or gas-fired, and a combustion turbine unit. Testing was conducted on
Unit 2, a 620 MW load-following coal-and/or gas-fired unit. This unit is equipped with a
four field electrostatic precipitator manufactured by Research Cottrell. The fuel typically
burned at the Hudson Station is pulverized, Eastern bituminous coal from West Virginia.
This low-sulfur (<1%) coal typically contains 0.1% chlorine. The high chlorine content
is likely a contributor to highly speciated mercury fractions.
An analysis of the coal ash is shown in Table 1 (the coal is ashed at the lab before
analysis). In addition, flue gas conditioning with SO3 and NH3 is used upstream of the
ESP to improve performance. The normal injection set points for NH3 and SO3 are 3 ppm
and 9 ppm, respectively. This additional SO3 has the potential to affect test results.
Table 1: Elemental Analysis of Ashed Hudson Coal
Element
(as oxide)
SiO2
A1A
TiO2
FeA
CaO
%
53.46
29.34
1.06
6.82
1.58
Element
(as oxide)
MgO
NajO
K2O
PA
SO3
%
0.94
0.94
3.47
0.26
0.28
-------
COHPAC compartment temperatures fluctuate with the flue gas temperatures in the
Hudson 22 ESP and were recorded from 250ฐ to 295ฐF. At high load conditions, the
temperatures were typically between 280ฐ and 290ฐF. The ability of activated carbon to
adsorb mercury decreases with increasing temperature in this operating range. Therefore,
flue gas temperature during carbon injection testing was monitored closely to assess the
impact of temperature on mercury removal at Hudson. Other parameters monitored
continuously include differential pressure across the collector bags (tubesheet pressure
drop), pulse cleaning frequency, cleaning pulse pressure, pilot-plant flue gas flow rates,
and Hudson's boiler load, boiler oxygen and stack opacity. Pilot ESP performance was
monitored in terms of electrical operating parameters.
Inlet and outlet flue gas mercury concentrations, particulate concentrations and particulate
size distribution were measured manually. The majority of the mercury measurements
were made with a modified Mercury Speciation Adsorption (MESA) method designed to
measure only total mercury. The size distribution tests were conducted with University
of Washington Mark V impactors. The manual mass concentration tests were performed
with a modified EPA Method 17 sampling system.
Test Facility Description
EPRJ's 1 MW(e) TPJ fabric filter pilot plant has been proven to supply reliable
performance data that can be used to predict full scale operation.1'2 The design and
operation of the pilot have been described in detail previously.3
A 3700 acfin slipstream for the pilot testing is extracted downstream of the full-scale
ESP at Hudson in order to duplicate the particulate loading that would be supplied to a
"polishing" unit. This is the configuration for which EPRI's COHPAC pulse-jet
baghouse is intended. Provisions were also made to test COHPAC at higher grain
loadings by providing a flue gas off-take ahead of the station ESP. Figure 1 is a
photograph of the TPJ and ESP pilot plants installed at Hudson Station.
The TPJ pilot at this installation was configured to clean with the low pressure (LP)
cleaning design.3 In order to attain the higher air-to-cloth ratio in the LP COHPAC
configuration, only 12 bags and cages were installed. An annulus is placed inside of the
filter vessel to decrease the effective interior volume of the filter vessel and simulate full-
scale upward velocities around the bags.
The COHPAC bags are cleaned by a single large pulse-valve through a rotating manifold
arm. A commercial full-scale LP compartment bag bundle typically has over 300 bags in
multiple concentric circles. The bags for this evaluation were full-scale bags made from
an 18 oz/yd2 Ryton felt with a Ryton scrim, singed on the inside only. The total
fabric area during COHPAC testing with 12 bags installed was 310 ft2.
-------
Figure 1. Photograph of transportable pulse-jet and ESP pilot plants.
The EPPvJ pilot-scale ESP is a wire-tube-type unit designed to treat nominally 160 acfrn
of flue gas moving at 5 ft/sec through a single gas passage. The main collector tube is
constructed of 10 inch diameter stainless steel tube. The specific collection area (SCA), a
standard measure of collection area to total gas flow, is 287 ft2/Kacfrn. The electrode is
powered by a 50kV, 5 mA power supply controlled with an automatic voltage controller.
The maximum possible current density with this power supply in this configuration is
lOOnA/cnr.
COHPAC Operation - Results
A total of 1400 operating hours were accumulated in this test configuration over a 6
month period from July through December 1996.
-------
Filter Cleanability
The operating A/C ratio for these tests was set at 12 ft/min. Tubesheet pressure drop
initially increased with time in operation, and stabilized at about 4 inches of water after
approximately 800 hours of cumulative operation. Cleaning frequency was extremely
low throughout the test, with the bags being cleaned once every 18 hours. The results
from the COHPAC operation are summarized in Table 2.
The extremely low cleaning frequency indicated that little ash was being collected on the
bags during filtering. Other COHPAC pilots and full-scale installations typically
underwent more frequent cleanings. Preliminary analyses of the particulate test results
indicate an average particulate concentration at COHPAC inlet of 0.004 grains/acf or
0.016 Ib/MMBTU. This particulate concentration is lower than recent Hudson stack test
measurements and indicates that the COHPAC pilot may not be receiving a representative
particulate mass loading. The mass mean diameter of the particulate collected at the inlet
was 2.8 micrometers.
Emission Rate
The data collected over the testing covered in this paper produced an average COHPAC
outlet particulate concentration of 0.0002 grains/acf or 0.0007 Ib/MMBTU.
Table2. COHPAC Evaluation: Operating Parameters and Results
Parameter
A/C (ft/min)
Number of bags
Cloth Area (ft2/bag)
Bag Length (inches)
Bag Flat Width (in.)
Can Velocity (ft/min)
Avg. Pressure Drop (inches/H2O)
Cleaning Freq. (cleans/bag/hour)
Outlet Emissions (Ib/MMBTU)
Hours of Operation
Value
12.0
12
25.8
242-1/2
7-11/16
1200
4.0
1/18
0.0007
1400
The initial results of COHPAC testing were quite promising. However, the low inlet
particulate loading may not be representative of a full-scale installation at this site,
especially during ESP upset conditions. A common baghouse pressure drop model
indicates that the tubesheet pressure drop is proportional to the time between cleans
multiplied by the inlet mass loading (constant air-to-cloth ratio and flyash
-------
characteristics). This model suggests that at a representative loading of 0.046
Ib/MMBTU, the time between cleans should be about three to four hours, which is an
excellent value for COHPAC installations. Further tests have been conducted at Hudson
to evaluate COHPAC performance with higher particulate loads. The results from these
tests are being evaluated while testing continues at the site.
Mercury Removal Evaluation
The objective of this phase of the testing was to evaluate the effect of the injection of
sorbents, namely activated carbon products, in controlling mercury emissions. Particulate
collector performance was monitored when the sorbent was injected upstream of the
COHPAC or ESP pilot plant. The test conditions identified for each particulate
collection device (COHPAC and ESP) provided information to allow an initial evaluation
of: 1) the effect of sorbent injection rate on mercury removal for two sorbents, and 2) the
effect of flue gas temperature on mercury removal for two sorbents. The parameters
included in the test matrix were sorbent type, injection rate, and temperature.
Mercury sorbent effectiveness is determined by the amount of sorbent needed to achieve
a specific removal efficiency. This efficiency is dependent on many factors including
operating conditions (such as temperature and residence time), flue gas composition
(including mercury species present and flyash properties), and sorbent properties (both
physical, such as size and surface area, and chemical).4
Due to the low mercury concentration in flue gas from coal combustion, the ratio of the
mass of sorbent injected to the mass of mercury present is expected to be very high 4>5(on
the order of 10,000:1) when high removal efficiencies (79%) are required. Mercury mass
transfer (removal effectiveness) can be improved by long residence times, small sorbent
sizes, and good sorbent-gas mixing. Residence times greater than 1 second and sorbent
particle size below 10 jim are desirable. In general, baghouses are expected to be more
effective for mercury capture than ESPs due to the increased contact time between
sorbent and mercury. When residence time is limited (such as in the ductwork
connecting an ESP to the air heater outlet), much larger amounts of sorbent or very small
sorbent particle sizes are needed to compensate for the lack of contact time.
Mercury Sorbents
The two sorbents selected for testing were Darco FGD activated carbon from American
Norit (referred to in this paper as Norit carbon) and an experimental carbon identified as
AC-1. Norit is an activated carbon derived from lignite and is used to remove mercury in
municipal solid waste (MWC) combustors in Europe and the United States.6'7 The mass
mean particle diameter of a sample of the Norit carbon, as measured by a Microtrac
-------
particle analyzer, was 14.6 micrometers. The B-E-T surface area as reported by the
manufacturer is 600 m2/g.
The second sorbent, AC-1, is an activated carbon prepared from a bituminous coal.
Initial laboratory and field evaluations of AC-1 have shown promising results as
compared with other carbon-based sorbents, including Norit activated carbon. AC-1 was
chosen for testing because of the promising technical results and projected cost savings.
The mass mean particle diameter of the AC-1 sorbent, as measured by a Microtrac
instrument, was 13.1 micrometers and the B-E-T surface area was 400 - 600 m2/g,. A
finer grind of the AC-1 was also evaluated at Hudson. The mass mean particle diameter
of the AC-lF(ine) sorbent, as measured by the manufacturer, was 3 micrometers.
Mercury Measurement Technique
Mercury measurements were made with a modified Mercury Speciation Adsorption
(MESA) method. This is an extractive process where a flue gas sample is drawn from the
duct through a process train. The modified MESA system employs two solid sorbent
traps assembled in series, where a quartz probe with a glass wool plug is installed
upstream of the traps. During sampling, flue gas passes through the glass wool in the
quartz probe where particulate is removed. Particulate-free flue gas then passes through
the two sorbent traps, which each contain an iodinated carbon, a sorbent which adsorbs
all forms of mercury. In the standard MESA train two other traps designed to capture
non-elemental forms of mercury are inserted upstream of the iodinated carbon traps.
Recently, some researchers have found that some elemental mercury may be adsorbed in
these first two traps and reported as oxidized mercury. Therefore, the majority of the
MESA trains used during testing at Hudson contained only the iodinated carbon traps and
the results were reported as total mercury. Following sampling, the traps are sent to a lab
to be analyzed for mercury content using cold vapor atomic fluorescence spectroscopy
(CVAFS).8
Mercury Testing Results
Sorbent injection evaluations were conducted in two distinct temperature ranges: between
288 and 298 ฐF and between 220 and 250 ฐF. The inlet mercury concentration during the
higher temperature tests varied from 2.9 to 6.5 (ig/Nm3. The inlet mercury concentration
varied from 1.4 to 5.2 p.g/Nm3 during the lower temperature tests. A coal sample
collected in May 1996 was analyzed to measure mercury content. The results indicated
0.056 |ig mercury per gram of coal. Using typical plant operating conditions, total
mercury concentration in the flue gas was estimated to be approximately 4.8 (J.g/Nm3.
Inlet mercury measurements made during testing were consistent with this estimated
value as calculated from the coal measurement.
-------
COHPAC Mercury Removal Results. The primary performance measurements
compiled during COHPAC testing were mercury removal efficiency, tubesheet pressure
drop, and cleaning frequency. Table 4 shows the completed COHPAC tests and the range
of results obtained. A summary of the results from the AC-1 and Norit injection tests
with COHPAC are shown in Figures 2 and 3. Figure 2 presents the mercury removal as a
function of sorbent concentration for 22 data pairs (inlet/outlet samples). Figure 3
presents mercury removal efficiency as a function of sorbent injection ratio for the same
data points. Baseline (no sorbent injection) mercury removal by COHPAC was less than
15 % for all temperatures tested, showing that the flyash at this site did not adsorb
significant mercury. As expected the mercury removal effectiveness increased with
carbon injection rate (expressed as carbon concentration in flue gas). The effect of
temperature on mercury removal is indicated by the data labels on the figure. In general,
as temperature increased, the mercury removal efficiency decreased for a constant sorbent
concentration. At temperatures > 288 ฐF, the maximum removal efficiency achieved was
40 % at a sorbent concentration of 3 Ib/MMacf (carbon-to-mercury mass ratio of 21,000:1
based on inlet mercury during testing). At 280 ฐF, removal efficiencies >50% with Norit
or AC-1 were achieved with sorbent injection concentrations > 4 Ib/MMacf (carbon-to-
mercury mass ratio > 14,700:1 based on inlet mercury during testing). Removal
efficiencies up to 87% were observed at temperatures of 280 ฐF, but the data scatter
makes accurate predictions difficult without further testing.
The data collected during Norit evaluations at temperatures below 255 ฐF suggest an
effect of sorbent buildup on the filter bags. For example, an initial sample was collected
30 minutes after Norit injection began at 1.4 Ib/MMacf. Immediately following this test,
approximately 45 minutes after beginning the initial sample, a second sample was
collected. The initial sample indicated 26% mercury removal and the subsequent sample
indicated 57% removal. This test sequence was repeated with Norit concentration at 2.4
Ib/MMacf. The first sample indicated 57% mercury removal and the subsequent sample
indicated 77% removal.
The mercury removal effectiveness for AC-1 and Norit appear quite similar from the
results at all tested temperatures. A single sample pair (inlet/outlet) was collected for the
AC-lF(ine) carbon. The mercury removal measured by this sample was greater than that
achieved under similar test conditions with Norit carbon. Although it is difficult to draw
any conclusions from one data point, greater removal can be expected for two primary
reasons: 1) laboratory tests indicate AC-1 has similar mercury removal properties to Norit
carbon, and 2) the median particle size of the AC-1F sorbent tested at Hudson was much
smaller than that of the Norit carbon tested. The mass mean diameter of the AC-1F tested
at Hudson was 3 micrometers and the Norit carbon was 14 micrometers. Although some
agglomeration of the smaller sorbent is expected during feeding, the AC-IP sorbent
particles present in the duct were likely much smaller than the Norit sorbent. Smaller
diameter particles present more surface area per unit mass for mercury adsorption.
-------
Table3.
COHPAC Test Matrix and Data Summary
Sorbent
None
None
None
Norit
Norit
Norit
Norit
Norit
Norit
AC-1
AC-1
AC-1
Injection Rate
(Ib/MMacf)
0
0
0
1.4
2.4
4.8
2.1-2.4
4.3-4.9
3.0
1.5
1.5-1.9
3.1-3.2
Collection Temp
(OF)
220 - 240 ฐF
260 T
295 - 300 ฐF
235 - 245 ฐF
235 - 245 ฐF
235 - 245 ฐF
280 - 282 ฐF
275 - 280 ฐF
288 ฐF
244 ฐF
280-285 ฐF
280 ฐF
Hg Removal Range (%)
0-14
5-13
0
26-57
57-77
84-97
10-26
55-87
39
73
33-35
75-80
COHPAC operation was monitored during sorbent injection tests. Critical parameters
included flue gas temperature, flow and pressure drop across the filters. The flows were
maintained at approximately 3720 acfin for an air-to-cloth ratio of 12 ft/min. The
pressure drop increase through the COHPAC unit due to sorbent injection is important
when evaluating the feasibility of mercury sorbent use in conjunction with a COHPAC
polishing unit. The tests conducted at Hudson were short-term tests and definitive
projections for long-term operations cannot be made. However, an indication of the
effect of sorbent injection on the pressure drop across the bags can be measured. During
Norit injection, the additional pressure drop due to the sorbent was 0.1 to 0.17 in H2O/hr
per Ib/MMacf sorbent. For example, if 5 Ib/MMacf Norit were injected upstream of the
COHPAC unit, an additional 0.5 - 0.85 inches H2O could be expected per hour. The
pressure increase due to the AC-1 was higher at 0.22 in H2O/hr per Ib/MMacf. The
higher pressure drop is likely a result of the finer particulate matter.
-------
s?
"5
o
o:
o
<5
Figure
g
i
Cฃ.
3
ฃ
a
5
100
90 -
80 -
70
60 -
50 -
40 -
30 -
20 -
10 -
0 -I
240 ฐF ,280ฐF
" * N
fl 244 ฐF " ป 28ฐ ฐF * ฐne
242 ฐF
240 ฐF m 28o op Norit
288 ฐF ai~ 1
290 OF i280ฐF 4ACM
238*F aAC-1F
280 ฐF ฐ
\J -f 1 1 1 | - |
0123456
Sorbent Concentration (Ib/MMacf)
2. Mercury removal as a function of sorbent concentration in COHPAC.
100
90 -
80 -
70 -
60 -
50 -
40 -
30 -
20 -
10 -
0
280 "F a 240ฐF
280ฐFi.242ฐF
*244ฐF
_ 240 ฐF . 242 ฐF
280 ฐFf
280 ฐF *
^ 288 ฐF
-238V "29ฐฐF
'_ m 280 ฐF
i 1 \ 1 \
ซ None
, Norit
AAC-1
A AC-1 F
10000 20000 30000 40000
Sorbent Injection Ratio (wt C: wt Hg)
50000
Figure 3. Mercury removal as a function of sorbent injection ratio in COFEPAC.
10
-------
ESP Mercury Removal Results. The primary performance measurements during
ESP pilot testing were mercury removal efficiency and secondary voltage and current.
Table 4 shows the completed ESP tests and the range of mercury removal achieved. A
summary of the results from the ESP tests are shown in Figures 4 and 5. Figure 4
presents the data in the same format as the COHPAC results: mercury removal as a
function of injection rate for sixteen data pairs (inlet/outlet samples). Figure 5 presents
mercury removal efficiency as a function of sorbent injection ratio for this data set. As in
the COHPAC experiments, there is little mercury removal during baseline tests (no
sorbent injection), indicating that the flyash present in the flue gas probably did not
contribute to mercury removal during sorbent injection.
Table4.
ESP Test Matrix and Summary of Results
Sorbent
None
None
Norit
Norit
Norit
Norit
Norit
AC-1
AC-1
AC-1
Injection Rate
lb/MMacf
0
0
1.3
2.3
4.8-4.9
3
4.9
1.8-2.2
4.8
3.2
Collection Temp
(op)
255 ฐF
268 - 278 ฐF
240 - 255 ฐF
240 - 255 ฐF
220 - 235 ฐF
275 - 280 ฐF
270 - 275 ฐF
240 - 250 ฐF
240 - 250 ฐF
280 ฐF
Hg Removal Range
(%)
3
0
13-17
41-42
76-83
14-38
28-35
33-45
56-58
28
The data shown on the figure indicate a distinct reduction in mercury removal efficiency
with increased temperature. The maximum mercury removal achieved was 83% at 221 ฐF
in the ESP with a Norit concentration of 4.8 lb/MMacf. Full-scale ESP operation at this
low temperature is not practical, due to potential problems with acid condensation. The
highest mercury removal achieved at the higher temperature range was 35% at 272 ฐF and
a carbon concentration of 4.9 lb/MMacf (carbon-to-mercury mass ratios near 45,000:1).
Three sets of MESA samples were collected for the AC-1 carbon. The mercury removal
during these tests was comparable to the removal achieved during the Norit injection
tests. The duplicate data runs appear less consistent for the AC-1 tests than for the Norit
tests. This is likely due to difficulties in maintaining a consistent feed of the AC-1
sorbent. The sorbent feedrate was much more constant during the Norit injection tests.
Pilot ESP operation was monitored during sorbent injection tests. No significant change
in secondary voltage or current was noted during testing, indicating that there was no
noticeable impact on electrical ESP performance from carbon injection. The secondary
11
-------
voltage during testing was 38-39 KV and the secondary current was 1-2 mA (current
density of 20 - 40 nA/cm2).
-e
^
"ra
S
o
a
ฃ
3
o
k-
0)
5
90 -
80 -
70 -
60 -
50 -
40 -
30 -
20 -
10 -
0
220 - 230 ฐF
B
ป 250 ฐF
A
! 246ฐFi.F241ฐF
* 1 280 ฐF . 272 ฐF
275 ฐF
255 ฐF
i ..._i.... . ...J 1
, None
. Norit
AAC-1
0123456
Sorbent Concentration (Ib/MMacf)
Figure 4. . Mercury removal as a function of sorbent concentration in pilot ESP.
. .
~
o
E
o
o:
a
a>
&
90 -
80 -
70 -
60 -
50 -
40 -
30 -
20 -
10 -
0 -
220 - 230 ฐF
"
250 ฐF
246 ฐF
% 241 ฐF
246 ฐF m 272 OF
255 ฐF
1 1 ! 1
ซ None
. Norit
AAC-1
0 20000 40000 60000 80000
Sorbent Injection Ratio (wt C:wt Hg)
Figure 5. Mercury removal as a function of sorbent injection ratio in ESP.
12
-------
Preliminary Conclusions
Pilot testing at Public Service Electric & Gas' Hudson Unit 2 is continuing. Results of
COHPAC testing through December 1996 and mercury removal testing through March
1997 from this pilot plant test program suggest the following preliminary conclusions:
COHPAC technology is viable for Hudson Unit 2
Particulate Emissions:
The COHPAC pilot maintained the outlet emissions well below the New Source
Performance Standard of 0.03 Ib/MMBTU. The measured outlet emissions
averaged less than 0.001 Ib/MMBTU. It is expected that COHPAC can meet all
current and anticipated primary particulate emission limits.
Filter Cleanability:
The LP COHPAC design was able to maintain a pressure drop of 4 inches H2O at
12 fiVmin with a cleaning frequency of 1 clean/bag/18 hours, which is significantly
below the normal observed cleaning frequency of 1 to 3 cleans/bag/hour.
The low cleaning frequency should lead to significantly longer bag life than
normally observed for COHPAC baghouses.
Testing was conducted with low inlet particulate loading that may not be
representative of full-scale operation or ESP upset conditions. Additional
COHPAC testing is underway.
Mercury removal effectiveness via activated carbon injection at Hudson is
dependent on sorbent injection rate temperature, sorbent residence time, and
sorbent size.
Sorbent Loading and Temperature
Mercury removal effectiveness increases with sorbent loading and decreases with
temperature . For COHPAC at temperatures 275 -285 ฐF, sorbent injection
concentrations >2.5 Ib/MMacf (carbon-to-mercury mass ration 14,000:1) can
achieve >50% removal efficiency. Higher removal efficiencies (up to 87%) were
measured but the data scatter make accurate predictions of removal achievable
difficult without more testing. At temperatures > 285 ฐF, > 50 % removal may not
be achievable at sorbent injection concentrations < 3 Ib/MMacf (carbon-to-mercury
mass ration 21,000:1). For ESP at >275 F, the results show that mercury removals
>35% are unlikely at sorbent injection concentrations < 51b/MMacf (carbon-to-
mercury mass ratio 45,000:1).
13
-------
At lower temperatures (<250 ฐF), high mercury removals (80 to 90%) are achievable
at sorbent concentrations of 5 Ib/MMacf (carbon-to-mercury mass ratio > 50,000:1)
while more modest mercury removal levels can be achieved with sorbent
concentrations < 1.5 Ib/MMacf (carbon-to-mercury mass ratio 8,400:1) for
COHPAC and 2.5 Ib/MMacf (carbon-to-mercury mass ratio 12,000:1) for ESPs.
However, lower temperatures are likely cause corrosion and bag life problems due
to acid condensation.
Sorbent Residence Time
COHPAC mercury removal effectiveness is better than that of the ESP. This is
further confirmation that sufficient residence time is needed for good mercury
removal. With an ESP, it is important to inject the sorbent as far ahead of the ESP
as possible. For a baghouse, prolonging the cleaning interval and keeping a thicker
dustcake will likely increase mercury removal efficiency.
Other Considerations
The limited test data show that smaller sorbent size improves mercury removal
effectiveness. This is consistent with theoretical predictions. However, finer
sorbents will be more difficult to feed and collect.
The flyash at Hudson removes very little mercury.
The novel sorbent AC-1 has a similar removal effectiveness to commercially
available Norit carbon.
Carbon injection produced acceptable increases in COHPAC cleaning frequency
and has no effect on ESP power levels.
With carbon injection, baghouse pressure drop varied between 4 and 7 niches H2O
at cleaning frequencies from 1 clean/bag/3 hours to 1 clean/bag/6 hours. This
cleaning frequency is still significantly below normal COHPAC operating
conditions and shows that carbon injection can be used with COHPAC even at
fairly high sorbent concentrations (5 Ib/MMacf, carbon-to-mercury mass ratio
27,000:1). This reflects the fact that although the ratio of sorbent to mercury in the
flue gas is significant, the minute mercury concentration level requires only modest
quantities of sorbent to be injected.
Sorbent injection did not affect power levels in the ESP pilot during these short-
term tests. However, the impact on ESP collection efficiency and outlets emissions
is unknown.
The combination of COHPAC and activated carbon injection technologies can be an
effective means of control for particulate Hazardous Air Pollutants(HAPs) and
vapor phase HAPs such as mercury.
14
-------
Acknowledgments
This project is supported through the cooperation of several organizations. The roles of
each organization and the key personnel are gratefully acknowledged. The Electric
Power Research Institute provides both financial and technical support. Public Service
Electric & Gas Company provides co-funding and technical guidance for the project
through the Air Toxics Team. Steve Marbaise provides on-site liaison support at Hudson
Station operation. ADA Technologies, Inc. is responsible for pilot operation, testing
services, data analysis, and report writing. Jean Bustard is a key technical advisor; Jim
Butz assists with technical editing and engineering services and Tim Ebner, Gary
Anderson, and Paul Anderson assist with on-site testing.
REFERENCES
1. J. Bustard, S. Sjostrom, R. Chang (1997). "COHPAC Model Prediction". Presented
at the EPRI-DOE-EPA Combined Utility Air Pollutant Control Symposium,
Washington B.C., August 25-29.
2. A.K. Hindocha, B. Brown, R. Chang (1993). "Commercial Demonstration of
COHPAC". Presented at the Tenth Particulate Control Symposium & Fifth
International Conference on Electrostatic Precipitation, Washington, D.C, April 5-8,
1993.
3. S.M. Sjostrom, C.J. 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 Chen, S., M. Rostam-Abadi, and R. Chang (1996). "Mercury Removal from
Combustion Flue Gas by Activated Carbon Injection: Mass Transfer Effects," In Proc
Am. Chem. Soc., New Orleans, La, March 23-28.
5 Carey, T. et al. (1997). 'Tactors affecting mercury control in utility flue gas using
sorbent injection", Paper 97-WA72A.05, 90th Annual Meeting Exhibition of the Air
and Waste Management Association, Toronto, Canada, June 8-13.
6 Brna, T.G. (1991). 'Toxic Metal Emissions from MWCs and Their Control", in
Proceedings of the 2nd International Conference on Municipal Waste Combustors,
Tampa, FL, April.
7 Nebel, K.L., and D.M. White (1991). "A Summary of Mercury Emissions and
Applicable Control Technologies for Municipal Solid Waste Combustors", a report
prepared for the Standards Development Branch of the US EPA, September.
8 Prestbo and Bloom (1995). "Mercury Speciation Adsorption (MESA) Method for
Combustion Flue Gas: Methodology, Artifacts, Intercomparison, and Atmospheric
Implications", Water, Air, and Soil Pollution, Volume 80: 145-158.
15
-------
NOVEL VAPOR PHASE MERCURY SORBENTS
M. Rostam-Abadi1-2, S. G. Chen1, H-C Hsi2, M. Rood2, R. Chang3, T. Carey4, B. Hargrove4, C.
Richardson4, W. Rosenhoover5, F. Meserole6
1 Illinois State Geological Survey, 615 E. Peabody Dr., Champaign, IL 61820;2 University of
Illinois, Environmental Engineering Program, 205 N. Mathews Ave., Urbana, IL 61801;
3 Electric Power Research Institute, 3412 Hillview Ave., Palo Alto, CA 94403;4 Radian
International, LLC, 8501 N. Mopac Blvd., Austin, TX 78759;5 CONSOL, 4000 Brownsville Rd.,
Library, PA 15129;6 8719 Ridgehill Dr., Austin, TX 78759
Abstract
Research to date on understanding the fundamentals of mercury control via sorbent injection has
shown that sorbent effectiveness can reach the highest levels achievable in a given flow
environment i.e., the levels set by mass transfer limitations if the sorbents have sufficient capacity
and activity. Several coal-based sorbents prepared in this study showed promising capacity and
activity in bench and pilot-scale testing. This paper describes results of bench-scale sorption
testing and compare these with pilot sorbent injection test data.
Introduction
Injection of activated carbon upstream of a particulate control system has the potential of
providing a low-cost option for control of mercury emissions from utility flue gas1. In several
bench2"5, pilot6'7 and full-scale tests8'9 of the method, the influence of carbon type, carbon structure
and carbon surface chemistry, injection methods (dry or wet), amount of carbon injected, and flue
gas temperatures on mercury removal have been examined. The low concentrations of mercury in
the flue gas, and limited exposure time (<3 seconds) of the sorbent, generally required large
amounts of activated carbons in these sorbent injection tests.
The high C/Hg ratio needed to achieve adequate mercury removal could be a result of either mass
transfer limitations due to the extremely low concentration of mercury in the flue gas, or low
mercury capacity of the carbon due to the low reactivity of the carbon. To reduce the operating
cost of the carbon injection process, either a more efficient sorbent that can operate at a lower
C/Hg ratio, or a lower-cost sorbent, or both are required.
Mass Transfer Calculations
Film Mass Transfer
When injected into flue gas, fine carbon particles will suspend and flow with the gas stream. In
absence of internal (intraparticle) diffusion, the equation describing the transfer of mercury
molecules from the bulk flue gas to the surface of the carbon per unit volume of duct is10:
N = kg(aAO(Cg-C<) (1)
-------
where N=mass flux (g/cm2.s); kg = mass transfer coefficient (cm/s); a = total interfacial area in the
duct (cm2/cm3) ; V = total volume of the duct (cm3); and Cg = mercury concentration (g/cm3) in
the bulk flue gas, and C* = mercury concentration in equilibrium with the adsorbed mercury on
the carbon surface (CJ. Cg -C* is considered as the driving force for mass transfer.
If no strong turbulence or back mixing occurs in the duct, the gas-solid phase can be modeled as
a plug flow system. The mass balance equation for a plug flow system is:
kgl(Cg-C')Sdz=-FgdCg (2)
where S is the cross section area of the duct (cm2), Fp is the flue gas flow rate (NnrYs), and dz is
the differential length of the duct (cm).
The velocity of the particles relative to the flue gas is practically zero. If the carbon particles are
well dispersed and do not agglomerate during the process, the mass transfer coefficient at the gas-
solid interface could be calculated by the following equation:
2*0.261 0.522
-
p P
where dp = particle size (cm) and D^ = diffiisivity of the mercury molecule in flue gas (cm2/s),
which is 0.261 (cm2/s) for the diffusivity of mercury in air at 140 ฐC. Equation (3) shows that the
mass transfer coefficient increases with decreasing carbon particle size. Any attempts to
introduce turbulence to the flow may not have any significant effects on the mass transfer
coefficient.
To examine the role of film mass transfer (the maximum mass transfer flux), assume C*ซ Cg at
all positions in the duct (this means that mercury adsorption capacity of the carbon and the
carbon reactivity are not limiting the mass transfer rate). Equation (4) is obtained by integrating
equation (2) using the following boundary conditions: 1) at z = 0 ( entrance), C =C0" 2) at z = L
(outlet), Cg=Cg.
i i ฐ\ i a SL
ta(F)=*ซ? ~ (4)
where L = length (cm) of the duct and SL/Fg=t is the residence time of carbon particles in the
duct. Equations (3 ) and (4) can be used to solve for the minimum activated carbon interfacial
area required by mass transfer to remove 90%of the mercury, i.e Cg = 0.1 C0) from oneNm3 of
utility flue gas:
-------
a 2.303 Fg 2.303 dp
= ฃ= ฃ=
V kg SL 0.522 /
Because only the external surface area of carbon particles serves as the gas-solid interfacial area
the minimum interfacial area needed for mass transfer implies that a certain minimum amount of
carbon is required to achieve the desired mercury removal. For spherical particles the external
surface area per gram of activated carbon is 6/ dp pc where pc is the carbon particle density in
g/cm3. The amount of carbon required for mercury removal from one Nm3 of utility flue gas
therefore is:
(6)
The carbon /mercury ratio can be calculated from the following relationship:
Carbon =0.7353 P# (7)
Mercury t (CQ-Cg) ^ '
Equation (7) shows that the C/Hg ratio depends strongly on the particle size and on the mercury
concentration in the flue gas (C0-Cg). Table 1 represents the carbon/mercury weight ratios
required for 90% mercury removal from flue gas under mass transfer limited conditions, with
activated carbon ranging in size from 1 to 10 urn, assuming the particle density of activated
carbon pc= 0.5g/cm3, contact time of 2 seconds, and an inlet mercury concentration of 10ug/Nm3.
The required C/Hg weight ratio for a 1 urn carbon particle is 204 but it increases to 20,417 for a
10 urn carbon particle. This analysis indicates that a high C/Hg weight ratio is required when the
carbon particle size is larger than 10 |j.m.
Table 1 also lists the required mercury capacity of the sorbent predicted by the mass transfer
analysis. It indicates that under a mass transfer limited process, a low capacity sorbent is
required. For example for a 5 \im carbon particle size, the mass transfer capacity is only 196 |ig/
g carbon. However, when the mercury capacity of a sorbent is comparable to that of the mass
transfer capacity, the C/Hg ratio is determined by both mass transfer parameters and sorbent
capacity. Under some extreme conditions, the mercury capacity of the sorbent could limit the
removal efficiency and the C/Hg ratio is determined by the sorbent capacity rather than the mass
transfer capacities shown in Table 1. In a future publication13, we will show that for a 5 |im
sorbent particle the required mercury capacity of the sorbent is in the range of 500 to 1,000 ug/g
carbon.
Figure 1 shows the dependence of the minimum C/Hg weight ratio on the concentration of the
mercury in the flue gas (for 90% removal and 2 seconds residence time). Equation (7) and
Figure 1 can be used to compare the minimum C/Hg weight ratios required to remove 90%
mercury from utility and incineration flue gases. MSW incinerator flue gas has a mercury
concentration of around 200-1000 ng/Nm3 which is almost two orders of magnitude higher than
-------
that of utility flue gas. The C/Hg ratio, according to equation (7), for the MSW flue gas will be
about two orders of magnitude lower than that of the utility flue gas. The theoretical C/Hg ratio
required for a MSW flue gas containing 600 ug/Nm3 mercury is about 60 times lower than that of
a utility flue gas containing 10 ug/Nm3 mercury for the same level of mercury removal.
These calculated C/Hg weight ratios are the minimum needed and assume mass transfer
limitations. At low C/Hg ratios (such as for the MSW flue gas at a C/Hg ratio of 2.3), it is
possible that the carbon will actually have reached its equilibrium capacity. Also at higher
temperatures, most sorbents have very low capacity for mercury. Under these conditions much
more sorbent than that predicted by mass transfer limitations will be needed.
Intraparticle Diffusion
Because diffusivities in mircroporous materials vary broadly, depending on their pore structure
and pore size, it is difficult to estimate diffusivity when the diffusion is in the configuration^
range. Configurational diffusion only occurs when the micropore size is comparable to the
molecular size of the adsorbate. In this study, calculations were made for a single spherical
carbon particle dispersed in a flue gas. The details are given elswhere12 The carbon particle was
assumed to be exposed to a step change in mercury concentration at its external surface at t = 0
(corresponding to the injection location). For a particle size of dp=10 |^m, t=3 s, and 90%
mercury uptake , the mercury diffusivity in activated carbon was calculated to be 1.52xlO"8 cmVs.
This value of diffusivity is in the range of Configurational diffusion. The calculations indicate that
with a 10 um activated carbon particle, the intraparticle diffusion will be important only when the
pore diameter is about 3 A, i.e., the molecular diameter of mercury. Because the micropore size
of the activated carbon is generally larger than 3 A, it can be concluded that intraparticle diffusion
is unlikely to be the controlling step in the carbon injection process.
Summary and Conclusions from Mass Transfer Studies
The analysis presented above indicates that under certain carbon injection conditions, mercury
removal from coal-fired flue gas is film mass transfer controlled. For example, Miller et al.6 used
a C/Hg ratio greater than 3000 for an activated carbon with a weight-averaged particle size of
5.5urn, to remove about 90% mercury from a flue gas. In the same study, for an iodine-
impregnated activated carbon with a weight-averaged particle size of 3 um, the C/Hg ratio was
about 1000. Such C/Hg ratios are comparable to those listed in Table 1. Full-scale tests of the
carbon injection process in MSW incinerator flue gas also confirmed the results shown in Table
1 Licata et al.12 reported that the equilibrium mercury capacity of an activated carbon used in
their tests was about 0.33 gHg/gAC, which corresponds to a C/Hg ratio of 3 (temperature was
not mentioned). However, in full-scale MSW tests with the same carbon, a C/Hg ratio of more
than 300 was used to reduce mercury concentration in the flue gas from 600 to 70 ug/Nm3 at 135
ฐC. This ratio corresponds to 0.0033 gHg/gAC, which is only 1% of the equilibrium capacity of
the carbon. In still another field test, White et al.9 found that carbon injection methods (dry or
wet) had a significant effect on mercury removal while the type and surface chemistry of the
activated carbon had none. These data suggest that mass transfer was controlling the mercury
removal. For conditions where mercury adsorption is mass transfer limited, measures should be
-------
activated carbon had none. These data suggest that mass transfer was controlling the mercury
removal. For conditions where mercury adsorption is mass transfer limited, measures should be
taken to increase the mercury mass flux (from the bulk gas to the surface of carbon) rather than
using a carbon with high capacity. To increase the mass transfer, either the mass transfer
coefficient, kg, or the interfacial area, a/V, should be increased. According to equation (3) the
mass transfer coefficient increases with decreasing carbon particle size. Reducing carbon particle
size also increases the interfacial area, without increasing carbon dosage. The most effective way
to reduce the C/Hg ratio is therefore to decrease the carbon particle size.
Mass transfer limits only apply when the carbon has sufficiently high reactivity and capacity.
Under certain injection conditions, the mercury capacity of the carbon may become limiting.
To evaluate the effect of mercury capacity of carbon on the required C/Hg weight ratio, an
adsorption isotherm which relates C* and CB is required. This isotherm can be used in conjunction
with the mass balance equation (2) to develop an expression which relates C/Hg weight ratio to
mass transfer parameters, exposure time, and mercury capacity of the carbon. The details are
given elsewhere13
Figure 2 shows the dependence of C/Hg weight ratio on the mercury capacity of carbon for
several carbon particle sizes. This figure was prepared by using a Henry's law isotherm (mercury
capacity is represented by the Henry's law constant, H) and employing conditions used to
generate Table 1. Figure 2 shows three regions. In region I, the C/Hg weigh ratio is solely
controlled by the mercury capacity of the carbon. An injection process could fall in this region
when the mercury capacity or particle size of carbon is small. In region n, C/Hg weight ratio is
determined by both the mass transfer parameters and mercury capacity of the carbon. An
injection process could fall into this region when mercury capacity of the carbon is comparable to
the mass transfer capacity. In region HI, the C/Hg weight ratio is controlled only by the mass
transfer parameters.
Figure 2 could provide some guidance for designing a carbon injection process. For example, to
remove 90% of the mercury from flue gas using a carbon with a Henry's law constant of 0.01 g
Carbon/ Nm3 (corresponding to a mercury capacity of 1000 ug/g Carbon at 10 ng/Nm3 gas phase
mercury concentration), the particle size of the carbon should be between 5 and 10 um. If the
particle size is larger than lOum, then mass transfer will limit the injection process and only a
small fraction of the carbon capacity will be utilized. If the carbon particle size is less than 5 um,
then the mercury capacity of the carbon is smaller than the mass transfer capacity. Under this
condition, the mercury removal efficiency can not be increased further by reducing the particle
size of the sorbent.
Carbon Development Studies
Characteristic of Activated Carbon for Mercury Vapor Capture
To design and produce a suitable, low-cost sorbent for use in an injection process, the
characteristics of the sorbent for optimum performance must be identified. Because the typical
concentrations of mercury in utility flue gas is about 1 to 10 ug/Nm3 and sorbent exposure time in
the duct is about less than 3 seconds, it is likely that only a small fraction of the carbon pore
-------
surface area is used in the mercury removal process. The high C/Hg weight ratios calculated from
the mass transfer confirms this observation. It is, therefore, important to determine the carbon
surface area required for monolayer mercury coverage. For example, if a large fraction of the
pore surface area of carbon is utilized in the process, then increasing the pore surface area of the
carbon would be desired (i.e., high capacity). On the contrast, if a small fraction of the pore
surface area is occupied by mercury, then a low surface area carbon which exhibits a high
mercury sorption rate and would be saturated during the limited exposure time to flue gas is
desirable (i.e., high reactivity). In the extreme case, the surface area needed by the monolayer
mercury coverage is less than the interfacial area needed for the mass transfer. In this situation,
the surface chemistry of carbon would be critical and the internal structures of the carbon, such as
pore size, pore volume and pore surface area will not be critical.
Assuming the same conditions used in the mass transfer analysis and further assuming that each
Hg molecule occupies 1 0 A2 of surface area and average carbon particle size of 1 0 urn, then the
pore surface area of the carbon required for removal (monolayer coverage) of 90% of the
mercury molecules in one Nm3 of utility flue gas is :
(10-l)(ng/Nm3)/200(g)x6.02 xio23 xio xl(r16(cin2)= 27.1 cmVNm3
The interfacial area required by film mass transfer is:
tkg 2tDHg 2*2*0.522
= 1103
Comparing this value with the surface area needed by the monolayer mercury coverage indicates
that only 2.45% (27. 1/1 103) of the carbon pore surface area is required for monolayer mercury
coverage. This analysis suggests that a carbon with low surface area (low capacity) and a high
sorption rate is desirable for the carbon injection process.
Activated Carbon Production
A carbon development program was initiated at the Illinois State Geological Survey (ISGS) to
investigate the effects of different carbon types, carbon structures, and carbon surface functional
groups on mercury sorption rate and capacity. In this paper only the results from a study to
prepare coal-based activated carbons are presented. Carbon products were made both in bench-
and pilot-scale reactors.
Bench-Scale Carbon Production. Bench-scale experiments were performed in a 5-cm ID
fluidized-bed reactor (FBR). These experiments were intended to: 1) identify optimum
conditions for producing activated carbon samples with desired properties for removal of mercury
species from utility flue gas, and 2) obtain scale-up date for producing larger quantities of carbon
products in a pilot-scale FBR. Gram-sized quantities (10 to 50 grams) of more than 20 activated
carbon samples were prepared from three high- volatile bituminous Illinois coals. The activated
carbon samples were characterized for their physical properties (surface area, porosity, pore size
distribution) and chemical compositions (ash, total carbon, elemental analysis) to gain' additional
-------
insight into the fundamentals of preparation and properties of the products, and to help explain
their interactions with mercury species. Selected properties of the activated carbon samples are
given in Table 2. Coals A and C2 were obtained from different sources but have similar chemical
properties. Some chemical properties of coal B were different from those of coals A and C2. The
Illinois coal-derived activated carbons (ICDAC) prepared under various processing conditions
had N2-BET surface areas ranging from 500 to 750 m2/g. The ash contents varied from 15 to 27
wt%.
Pilot-Scale Production. In July 1996, ISGS worked with Svedala Inc., a chemical process
equipment firm located in Oak Creek, WI, to produce more than 120 pounds of activated carbon
from C2 coal. For these production runs, a batch, 18-inch ID pilot-scale FBR was used. Five
pilot production tests were made. The processing conditions employed in the pilot production
runs were those identified during the bench-scale studies of coal C2. This pilot production
encountered little difficulty in scaling-up the process. The properties of the pilot-produced
ICDAC are presented in Table 2. This product was obtained by blending the activated carbon
samples produced during two of the pilot production tests. The carbon products were ground to
two particle size ranges, 5.2 and 8.0 um mass mean diameter. The finer particle size sample, AC-
F, had a slightly higher ash content (24.7 wt%) than the coarser sample AC-C (18.8wt%). The
surface areas of the carbon samples (671 and 688 m2/g) were comparable to the activated carbon
(C-36) produced from C2 coal in the bench FBR. The properties of the pilot ICDAC were
comparable to those of Darco FGD carbon manufactured by American Norit FGD carbon is
designed for removal of vapor phase mercury from combustion flue gases. Data obtained from a
similar pilot activated carbon production test indicate that the production costs of the ICDAC for
use in an injection process are considerably lower than those of commercial activated carbons.
Mercury Removal Testing
The mercury removal performance of the activated carbon samples produced in this study were
measured both in a bench-scale test apparatus and in a sorbent injection pilot-plant.
Bench-Scale, Fixed-Bed Tests
Bench-scale mercury testing was conducted at Radian International's Austin laboratory. The
bench-scale screening tests focused on evaluating the mercury breakthrough characteristics of the
ICDAC samples and the Darco FGD carbon, under a wide variety of test conditions. Detailed
descriptions of the test set-up, conditions and data interpretation have been provided elsewhere3.
Pilot-Scale Mercury Tests
The mercury removal performance of pilot-scale ICDAC and of Darco's FGD carbon were
determined in a 500 scfrn (0.25 MWe) pilot plant operated by CONSOL, Inc. The pilot plant
can simulate flue gas conditions downstream of the air preheater in a coal fired utility power plant;
the flue gas mercury concentration studied (10-15 ug/m3) is typical of utility flue gas
concentration. Mercury removals were evaluated in the flue gas duct, which provides a gas
residence time of approximately 2 seconds, and in the baghouse, where the solids retention times
can be as high as 30 min. Common test conditions were: flue gas flow, 350 scfrn; flue gas wet
-------
bulb temperature, 121-128 ฐF; flue gas composition, 1000 ppmv dry SO2, 10 vol% dry O2, and 10
vol% dry CO2. All tests were conducted with a fly ash obtained from a coal-fired utility boiler
firing an eastern bituminous coal. The fly ash feed rate was 10-11 Ib/hr (solids loading of 3.2-3.7
gm/scf). Mercury removal was determined from the mercury feed rate, the solids (carbon and fly
ash) feed rate, and mercury analysis of the feed and recovered solids (by combustion followed by
cold vapor atomic absorption spectroscopy). Except where noted, all mercury removal results
discussed in this paper include mercury removal by the carbon sorbent and the fly ash.
Results from Mercury Testing
Bench-Scale Tests. Table 3 summarizes the initial and 100% breakthrough mercury capacities
of the activated carbon samples. The initial breakthrough capacity is defined as the capacity of
the sorbent at the time when mercury is first detected at the outlet. The 100% breakthrough
(equilibrium) capacity is the capacity at the time when the outlet mercury concentration is first
equal to the inlet concentration. The initial and 100% mercury capacities of the carbon provide
the information needed to determine the extent of mercury removal from flue gas by a sorbent
injection process. The mercury concentrations used in these tests ranged from 45 to 73 |ig/Nm3.
Studies conducted at Radian indicate that equilibrium mercury capacity of activated carbon
increases with increasing mercury concentration in the simulated flue gas. The initial
breakthrough capacities ranged from 0 to 2243 (ig/gC for elemental mercury and 0 to 514 ng/gC
for the mercuric chloride. Sample AC-2 and AC-43 had comparable pore surface areas and ash
contents, but their initial elemental mercury breakthrough capacities were 0 (Jg/gC and 851
jig/gC, respectively. Sample AC-2 also had no initial breakthrough capacity for mercuric
chloride. Some of the inherent chemical properties of the Illinois coal from which samples AC-2
was made were different from those of the coals from which samples AC-1, AC-43 and AC-36
were made. This suggests that certain properties of the precursor coal affected the reactivity of
the product carbons for mercury capture. The data also suggest an increase in the elemental
mercury capacity with increasing surface area for samples AC-1, AC-43, and AC-36. The largest
elemental mercury capacity was observed for sample AC-1 with a surface area of 742 m2/g. The
mercuric chloride capacity of the ICDAC samples does not appear to be influenced by their
surface areas.
The initial elemental mercury breakthrough capacities of the ICDAC were generally greater than
that of the Darco FGD carbon. However, their saturation capacities were comparable. Darco
FGD carbon had much greater mercuric chloride capacity than the ICDAC samples. The average
pore diameter of the FGD carbon is about 38.5 A which is about twice that of the ICDAC
samples. Pore size distribution data for the ICDAC samples indicate that almost all the pores are
smaller than 100 A, and a large fraction of the pore area is contained in pores with diameters
smaller than 17 A. In contrast, almost half the pore area of the Darco FGD carbon is in pores
between 100 and 1000A. The molecular diameter of mercuric chloride is about 5A as compared
to about 3A for the elemental mercury. Therefore it is possible that the larger mercuric chloride
molecule is more accessible to the Draco FGD carbon than the ICDAC samples. The mercury
capacities of the pilot FBR activated carbon are also listed in Table 3. Both the elemental
mercury and mercuric chloride capacities of the pilot ICDAC are comparable to samples
prepared in the bench FBR. There are no significant differences between the AC-F and AC-C
samples. These data confirmed that scale up of the carbon production was successful and that
-------
the ICDAC could be manufactured on a commercial scale.
Pilot-Scale tests. The Darco FGD carbon had a mass median particle size of 15 |im as
compared to 5 and 8 |im for samples AC-F and AC-C, respectively. At all test conditions
studied, both the ICDAC samples were as effective as the FGD carbon, and in many cases were
significantly more effective. As shown in Figure 3, the effect of mercury species on removal at a
C/Hg weight ratio of 3,000 and 225ฐF flue gas temperature was small. System removals of Hgฐ
and HgCl2 with AC-F were 52 and 47%, respectively. Removals with the Darco FGD were 44%
with both mercury species. Figure 4 shows system mercury removals at flue gas temperatures of
275 and 325 ฐF and a C/Hg weight ratio of 10,000. At both temperatures, AC-C and AC-F
achieved significantly higher mercury removals than the Darco FGD carbon. With AC-F,
Hg"removal was 77% and HgCl2 removals were 64 to 69%. With the Norit FGD carbon, Hgฐ
removals were 53% to 57% and HgCl2 removals were 34 to 44%. Hgฐ and HgCl2 removals at
275 ฐF with AC-C were about 84%. The data shown in Figure 4 do not demonstrate a
pronounced effect on Hgฐ removal as flue gas temperature rose from 275 to 325 ฐF. HgCl2
removal with AC-F and Norit FGD carbon decreased with increasing temperature. The absolute
changes however, were small.
Test results from injecting fly ash alone at 275 ฐF showed Hgฐ and HgCl2 removals of 9 (Hgฐ)
and 15 (HgCl2) duct and 31 (Hgฐ) and 23 (HgClj) system (duct+baghouse), and 8 (Hgฐ) and 15
(HgCy duct and 25 (Hgฐ) and 33 (HgCL,) system at 325 ฐF. If the fly ash mercury removal is
assumed to be constant, then the incremental removal due to the activated carbon can be
estimated by subtracting the fly ash contributions from the combined carbon plus fly ash mercury
removal tests. AC-F Hgฐ system removal at flue gas temperatures of 275 and 325"F and at
10,000 C/Hg weigh ratio were 46 and 52%, respectively, compared to 22 and 32% with the Norit
FGD carbon. AC-F HgCl2 system removals at these conditions were 46 and 31%, compared to
Norit FGD removals of 21 and 1%, respectively. AC-C Hgฐ and HgCl2 system removals at
275 ฐF were 53 and 60%, respectively. Duct mercury removals were low with both AC-F
(7%Hgฐ and 2% HgCy and FGD carbon (3% Hgฐ and 3% HgCl2). AC-C duct removals were
somewhat higher, 15% (Hgฐ) to 21% (HgCl2).
The pilot plant duct provided a 2 seconds residence time. As shown in Figures 3 and 5, mercury
removal in the duct is limited by bulk gas mass transfer. For most tests, the duct mercury removal
was between 9 and 19%, and was independent of carbon type, carbon feed rate, temperature, or
mercury species. AC-C showed a somewhat higher in-duct HgCl2 removal (36%), Figure 5.
Reducing the mass median particle size of the carbons from 13-24 urn in the previous work11 to as
low as 5|im in this work did not significantly increase mercury removal in the duct.
Conclusions
This study shows that the minimum amount of carbon needed to achieve a specific mercury
removal efficiency via sorbent injection can be predicted by assuming mass transfer limitations.
Mercury removal effectiveness can be increased by decreasing the size of the carbon injected,
increasing the residence time, or the amount of carbon injected. If mercury removal is limited by
the reactivity and capacity of the carbon (i.e. not mass transfer limited), then significantly more
carbon than the amount predicted by mass transfer limitations may be needed for effective
-------
mercury removal unless the reactivity and capacity of the carbon can be improved through
structural and surface chemistry changes. Intraparticle diffusion is not important because of the
small carbon sizes normally used for injection.
Ulinois coal-derived activated carbon (ICDAC) samples with desired properties were prepared
both in bench- and pilot-scale reactors. The pilot production showed no difficulty with the scaling
up of the process. The results from pilot-scale mercury tests showed ICDAC has mercury
removal capacity comparable to or greater than a commonly used commercial product.
Acknowledgment
This research is supported by the Electric Power Research Institute and the Illinois Clean Coal
Institute
References
1. Chang, R.; and Offen, G., "Mercury Emission Control Technologies: An EPRI Synopsis,"
Power Engineering ,Volume 5, November, 1995.
2. Krishnan, S. V.; Gullett, B. K. and Jozewicz, W., Environ. Sci. Technol. 1994, 28, 1506.
3. Carey, T. et al, " Factors affecting mercury control in utility flue gas using sorbent injection",
Paper 97-WA72A05, 90th Annual Meeting & Exhibition of the Air & Waste Management
Association, Toronto, Canada, June 8-13,1997.
4. Senior, L. S., et al., "A fundamental study of mercury partitioning in coal fired power plant
flue gas", Paper 97-WA72B.08, 90th Annual Meeting & Exhibition of the Air & Waste
Management Association, Toronto, Canada, June 8-13,1997.
5. Vidic, R. D., and McLaughlin, J.D., "Uptake of Elemental Mercury by Activated Carbons",
Journal of AWMA, Vol. 46, March, 1996.
6. Miller, S. J., Laudal, D. L.; Chang, R. and Bergman, P.O., In Proc. of 87th Annual Meeting &
Exhibition of the Air & Waste Management Association, Cincinnati, Ohio, June 19-24, 1994.
7. Sjostrom, S. et al., "Demonstration of dry carbon-based sorbent injection for mercury control
in utility ESPs and baghouses", Paper 97-WA72A.07, 90th Annual Meeting & Exhibition of
the Air & Waste Management Association, Toronto, Canada, June 8-13,1997.
8. Felsvang, K., Gleiser, R; Juip, G. and Nielsen, K. K., Fuel Proc. Technol., 1994, 39, 417.
9. White, D. M.; Kelly, W. E.; Stucky, M. J.; Swift, J. L.; and Palazzolo M A US EPA,
EPA/600/SR-93/181. January, (1994).
10. Chen, S., Rostam-Abadi, M., Chang, R, Prep. Am. Chem. Soc., Div. Fuel, Chem., 41, 1996,
pp. 442-44.
11. Stouffer, M.R, Rosenhoover, W. A, Burke, P.P., A & WM 89th Annual Mtg Nashville TN
June 1996. ' ' '
12. Licata, A; Babu, M.; and Nethe, L.; Proceedings of the 1994 National Waste Processing
Conference. Boston, Massachusetts, June 1994.
13. Chen, S., M. Rostam-Abadi, and R. Chang, "Mercury Removal from Flue Gas by Carbon
Injection: Mass Transfer Analysis", in preparation.
-------
Table 1
Mass Transfer C/Hg Weight Ratios and Mercury Capacities for Different Carbon Particle Sizes
Particle size
(mm)
10
5.5
5
3
1
k*
(cm/s)
522
949
1044
1740
5220
Interfacial area
(m2/g)
1.2
2.18
2.4
4.0
12.0
Wt Ratio of C/Hg
Utility
20417
6176
5103
1836
204
MSW
341
103
85
30
3.5
Mercury Capacity
((ig mercury /g C)
Utility
49
162
196
545
4902
MSW
2932
9708
11765
33333
285714
Table 2
Properties of Activated Carbon Samples
Coal
Activated Carbon
ID
Ash
(Wt %)
Surface
Area
K/g)
Pore
Volume
(cc/g)
Porosity
(%)
Ave. Pore
Diameter
(A)
Mass Mean
Particle Size
(urn)
Bench FBR Activated Carbon Samples
A
B
C2
C2
AC-1
AC-2
AC-43
AC-36
27.0
14.3
15.8
18.9
742
563
533
680
Pilot FBR Activated Carbon Samples
C2
C2
AC-C
AC-F
18.8
24.7
688
671
0.382
0.423
46.2
48.7
8.0
5.2
Commercial Activated Carbon
Lignite
Darco FGD
32.1
546
0.611
57.9
38.5
15.3
-------
Table 3
Mercury Removal Performance of AC Samples at 275 ฐF
ID
Mercury Capacity (ug/gC)
Hgฐ
Inlet
Concentration
(ug/Nm3)
Initial
Breakthrough
100%
Breakthrough
HgCl2
Inlet
Concentration
(Hg/Nm3)
Initial
Breakthrough
100%
Breakthrough
Bench FBR Activated Carbon Samples
AC-1
AC-2
AC-43
AC-36
AC-C
AC-F
61
73
55
54
2243
0
851
1883
2718
1114
1005
2091
61
46
55
63
514
0
357
510
596
19
450
604
Pilot FBR Activated Carbon Samples
76
64
1939
1721
2188
1958
42
56
431
397
450
438
Commercial Activated Carbon
FGCT|| 45
510
2590 || 60
1330
1570
10
100
Mercury Concentration,
Figure 1
Dependence of Carbon/Mercury Ratio on Mercury Concentration
-------
10000-p^
~ 1000-r=
CD
o
e
CO
o
100
1 um
^
100
1000
10000 100000 1000000
(Nm3/gCarbon)
Figure 2
Dependence of Carbon/Mercury Ratio on Mercury Capacity of Carbon
100-1
80-
5 40-
20-
Carbon/Mercuiy =3040-4030
HgCb
I
T I
AC-F FGD
BAGHOUSE
AC-F FGD
R] DUCT
FigureS
Effect of Contact Time on Mercury Removal at 225ฐF and 3000 C/Hg Weight Ratio.
-------
100 -i
80-
40-
20-
Carbon/Mercury=9180-11180
4* ACC
o ฐ Hgฐ-ACF
Hg2+-ACF
Hgฐ-FGD
Hg2+-FGD
250 275 300 325 350
Temperature, ฐF
Figure 4
Effect of Temperature on System Mercury Removal
100-1
80-
1 60-
3" 40-
20-
HgCl2
AC-F AC-C FGD AC-F AC-C FGD
| BAGHOUSE UJ DUCT
Figure 5
Effect of Contact Time on Mercury Removal at 275ฐF
-------
Thursday, August 28; 3:00 p.m.
Parallel Session A:
High Gas-to-Cloth Ratio Baghouses
-------
ALABAMA POWER COMPANY
E.C. GASTON 272 MW ELECTRIC STEAM PLANT - UNIT NO. 3
ENHANCED COHPACI INSTALLATION
Richard L. Miller
Research-Cottrell, Inc.
Somerville, NJ
Wallis A. Harrison
Southern Company Services
Brrmingham, AL
David B. Prater
Alabama Power Co.
E.C. Gaston Electric Steam Plant
Wilsonville, AL
Ramsay Chang
Electric Power Research Institute (EPRI)
Palo Alto, CA
Abstract
In an effort to reduce the outlet particulate emission levels from their existing hot-side
electrostatic precipitators, Alabama Power contracted with Research-Cottrell, Inc. to install an
enhanced, COmpact Hybrid PArticulate Collector (COHPAC) on Unit No. 3 at their E.C.
Gaston Electric Generating Plant located on Highway 25, about one mile northeast of
Wilsonville, Alabama, which is approximately forty miles southeast of Birmingham, Alabama.
This hybrid system utilizes Research-CottrelFs Low Pressure/High Volume (LPHV) pulse jet
technology, installed after the existing hot-side ESP collectors and air-preheaters. Due to the
limited available real estate, this unique COHPAC system was retrofitted into the original
abandoned cold-side ESP casing, which had previously been utilized as ductwork.
The system is designed to treat a total flue gas volume of 1,070,000 ACFM, and reduce the
remaining particulate emission levels exiting the hot-side ESPs down to outlet loadings of 0.005
gr/dscf (0.02 Ib/MMBtu) and 5% opacity. This system will operate as an on-line cleaning pulse
jet system using 23-foot-long bags. An aggressive construction schedule began in early August,
1996, with system start-up in late December. A short 10 week construction window was
provided by Alabama Power during a planned outage. The E.C. Gaston project is the third full
scale COHPAC installation installed to date on a utility coal fired boiler. This brings the total
installations currently being treated by COHPAC to nearly 1,500 MW, all provided by Research-
Cottrell, Inc. Details on the overall design, construction, start-up, operating history and test data
will be presented.
-------
Introduction
Alabama Power Company, a Southern Company, owns and operates the E.G. Gaston Electric
Generating Plant located near Wilsonville, Alabama, approximately 40 miles southeast of
Birmingham Alabama. This plant consists of four (4) 272 MW balanced draft and one (1) 900
MW forced draft coal fired boilers, each outfitted with hot-side electrostatic precipitators for
particulate control. All five boilers currently bum low sulfur, Eastern Bituminous coals.
Alabama Power considered the positive experience of the first full-scale COHPAC installation at
TU Electric's Big Brown station. In September, 1995, EPRI and Southern Company Services
(SCS) installed a 1-MW COHPAC pilot plant at their Miller Steam Electric Generating facility
in anticipation of using this technology at their own facility.
Pilot Plant Description
The Miller Station pilot plant was installed downstream of the existing Unit 2 hot-side ESP
utilizing a LPHV pulse cleaning head, similar to the Research-Cottrell design that was ultimately
used at E.G. Gaston. This pilot facility, while originally designed to filter approximately 5,000
cfm at a nominal air-to-cloth ratio of 4.0:1, was modified to simulate the next generation in
COHPAC I operation. The pilot plant was modified to operate with interstitial can velocities of
less than 1,000 fbm while filtering at nominal A/C ratios of 8.5 to 10.0 fpm, and utilizing an on-
line cleaning mode of operation.
In the past, all other operating COHPAC designs, including TU Electric's Big Brown Station
installation used off-line cleaning. As it was Research-CottrelFs and Southern Company's intent
based upon prior COHPAC operating experience to optimize and simplify the operation of future
installations by utilizing an on-line cleaning mode of operation. EPRI and SCS wanted to ensure
that this method of operation was viable and reliable prior to implementing it on a full-scale
basis. Full length 24-foot-long Ryton felt filter bags were installed to achieve the desired
filtration rates. Both timer and pressure drop initiation was utilized to maintain desired pressure
drop levels across the pilot unit.
Since its initial operation, the COHPAC pilot facility has operated exceptionally well, with no
significant problems. Utilizing an on-line cleaning mode of operation, the Low Pressure/High
Volume pulse cleaning technology successfully maintained tube sheet pressure drops of 4.0"
W.C. while at an 8.5 to 10.0:1 air-to-cloth ratio and maintaining few cleaning cycles per day. As
discussed later in this paper, this successful operation was transferred over to the full-scale B.C.
Gaston COHPAC system.
E.C. Gaston Full-Scale COHPAC Installation
Based in part on the success of Alabama Power's Miller Station COHPAC pilot testing, and of
TU Electric's Big Brown Station's 2 X 575 MW COHPAC units, Alabama Power decided to
install a full scale COHPAC installation on E.C. Gaston Station's, Unit #3, 272 MWe coal fired
boiler. Alabama Power contracted with Research-Cottrell to install a hybrid COHPAC pulse-jet
cleaning system downstream of an existing hot-side electrostatic precipitator.
-------
This system was installed in two abandoned cold-side ESP casings located on the cold-side of the
existing Ljungstrom air pre-heaters. Due to extremely tight real estate limitations, the COHPAC
collectors were installed in the abandoned cold-side ESP casings which previously had been
gutted, hoppers removed and used as sections of ductwork. These units were located directly
under the existing hot-side precipitator, and in between Units 2 and 4 which were in full
operation during the construction of the system. These unusual conditions made this one of the
most difficult retrofit installations ever attempted, especially during a relatively short outage
window.
The COHPAC system is designed to maintain Unit 3's outlet opacity levels below 5% on a 6
minute average basis with maximum particulate loadings not to exceed 0.005 gr/dscf and 0.02
Ib/MMbtu with all operating conditions and while firing the specified coals. The system is
designed to operate at a maximum gross air-to-cloth ratio of 8.5:1 utilizing on-line cleaning.
Scope of Supply
The general scope of supply furnished to Alabama Power for this project by Research-Cottrell
included the design, supply, freight and field erection of a complete COHPAC pulse jet fabric
filter system which consisted of a LPHV Pulse Jet baghouse. Additional scope items included
any new and modified inlet and outlet ductwork, casing steel, hoppers, tube sheets, required
expansion joints, compartment inlet, outlet, bypass and purge dampers, any additional required
support steel and associated foundations, access, supply of Imtech doors, lighting, hatch doors
and associated hoists, test ports, instrumentation, PLC controls including required programming,
MCCs and control house. The scope items also included the low pressure clean air blowers &
piping, LP cleaning assembly's, bags & cages, and supply and installation of a new ash removal
system, including any required demolition to the existing abandoned ESP casing and the flue gas
model study.
Alabama Power removed any existing asbestos insulation, provided and installed all new
insulation and lagging, modified existing I.D. Fans and motors, grounding, modified their DCS
system programming, supply and programming of their ash handling system controls, and
performance testing.
COHPAC Baghouse Installation Description
The COHPAC system is designed to treat a total flue gas volume of 1,070,000 acfin at 290ฐ F,
producing a design gross filtration rate of 8.5 fbm with all compartments in operation.
-------
The design fuel and ash analysis under which all guarantees are to be met are as follows:
Proximate analysis (% by weight as received)
Moisture 12% maximum
Ash 17% maximum
Volatile matter 18% to 3 6%
Fixed carbon 50% to 65%
HHV 10,000 BTU/lb minimum
Ultimate analysis (% by weight as received)
Moisture 4% to 12%
Carbon 65% to 75%
Hydrogen 3% to 5%
Nitrogen 1.1% to 1.9%
Chlorine 0% to 0.17%
Sulfur 0.6% to 1.5%
Ash 8% to 17%
Oxygen 4% to 9.0%
Ash Mineral Analysis (% by weight)
Silicon Dioxide 48% to 62%
Aluminum Oxide 29% to 35%
Titanium Dioxide 1.2% to 2.3%
Iron Oxide 4.6% to 7.6%
Calcium Oxide 1.8% to 5.8%
Magnesium Oxide 0.6% to 1.6%
Potasium Oxide 0.04% to 2.9%
Sodium Oxide 0.0% to 2.5%
Sulfur Trioxide 0.04% to 2.4%
Phosphoric Pentoxide 0.04% to 1.3%
Strontium Oxide 0.06% to 0.7%
Barium Oxide 0.07% to 0.3%
Manganese Oxide 0.05% to 0.12%
The COHPAC collector consists of a total of four (4) isolatable compartments, two
compartments per air-preheater identified as either Side A or B. Each compartment consists of
two bag bundles, each having a total of 544, 23-foot-long, Ryton felt filter bags, 18 oz./yd2
nominal weight. This results in a total of 1,088 bags per compartment, or 4,352 bags per casing.
To allow inspection and/or maintenance of any given compartment, due to the few number of
compartments per air-preheater, either the compartments are isolated during a scheduled reduced
load period or a partial bypass condition is initiated on the side being entered.
-------
Due to the limited size of the abandoned cold-side ESP casing, to be able to handle the design
flue gas volumes and provide the specified air-to-cloth ratios, the ESP casing had to be extended
by approximately 12 feet in length. Each COHPAC baghouse casing is separated into two (2)
isolatable compartments, separated by a central flue gas bypass section. The bypass system can
either be 100% opened or regulated to maintain design gas-to-cloth ratios in the on-stream
compartment. Each compartment is equipped with a multi-louver inlet damper and a guillotine
outlet damper to allow isolation during periods of inspection and/or bag replacement.
In addition to these dampers, Research-Cottrell outfitted each compartment with individual low
leakage, purge/ventilation poppet dampers designed to allow the introduction of ambient air into
the compartments and flush out any SO3 laden flue gas. They also can prevent the build-up of a
tenacious filter cake on the bags when isolated for an extended period of time. These dampers
also serve as a vacuum breaker to aid in opening the roof hatches, and to also help ventilate the
isolated compartments while the roof hatch is in the open position. The dampers automatically
open when a single compartment is isolated or when the entire baghouse is either bypassed or
shut down. Purge air is drawn into this damper via the use of the system I.D. fans for a preset
time period.
This type of damper system allows the isolated compartments to remain at atmospheric pressure,
thus preventing flue gas from leaking into the isolated compartment/collector and condensing on
the filter cake/bags.
CLEANING AIR MANIFOLD DRIVE
& AIR RECEIVER TANK (TYP.)
GUILLOTINE DAMPER
Side-View
E.G. Gaston
COHPAC System
EXISTING
DUCT
ESP to FF Conversion Downstream of Hot-Side ESP.
Figure 1 - B.C. Gaston COHPAC System Elevation View
-------
Due to the limited available height clearance below the existing hot-side precipitator casing,
access to the individual compartments was provided through roof mounted hatches which are
located over each individual bag bundle. Each hatch is separated into two (2) hinged halves.
They are opened with the use of an overhead hoist and are locked into the open position to allow
easy access to the tube sheet area of the compartment for inspection and/or bag replacement. A
short ladder allows access to the top of the tube sheet. Research-Cottrell's LPHV pulse jet
technology allows full access to the bags without the need to remove individual blow pipes.
Figure 2 - Roof/Tube Sheet Access System
Pulse cleaning air is provided via three (3), 50% low pressure positive displacement blowers (2
operating, 1 spare). The blowers are designed to deliver large volumes of up to 1200 ICFM of
cleaning air at low pulse pressures of 9 to 14 psig which was required to maintain guaranteed
pressure drop, bag life and emission levels. During the first few months of operation, only one
(1) clean air blower was required to maintain required system pressure drop levels. Pulse
pressures during this time period were maintained at 9 psig and reduced to 2.5 psig during non-
pulsing time periods.
Gas Flow Model Study
Due to the complexity of retrofitting a pulse-jet system into an existing precipitator casing, a
three dimensional gas flow model of the COHPAC system, (l/12th scale), was fabricated of clear
plexiglass in Research-CottreH's Gas Dynamics Laboratory. A complete series of gas flow
modeling tests demonstrated key gas flow criteria, from the air heater outlet, through the
baghouse casing, up to the I.D. fan inlet, including all associated dampers and ductwork. A
single baghouse casing was modeled, including both isolatable compartments, bypass duct and
the four individual bag bundle areas utilizing fabric to simulate the actual bag material.
-------
Inlet and outlet vanes and perforated plates were also modeled to simulate full scale gas flow
patterns and system pressure drops. Figure 3 shows a photo of the model study performed for
Alabama Power.
Figure 3 - COHPAC Model Study
All of the flow control devices developed during the course of the model study were incorporated
into the full scale installation. The model was a major contributor towards achieving low
mechanical pressure drops. All specified gas flow goals and pressure drops were achieved in
both the model and in the full scale installation.
COHPAC System Construction
This unusual COHPAC I system presented Research-Cottrell with a very difficult and
challenging construction opportunity. Due to extremely tight real estate restrictions, Research-
Cottrell proposed to convert the existing Joy/Western cold-side precipitator casing, which had
previously been gutted by Alabama Power (including the removal of the ESP hoppers) and was
subsequently used as a section of ductwork located between the air heater outlet to the I.D. fan
inlets. These boxes are located at grade, below the later-vintage Research-Cottrell hot-side
precipitators and in between Units 2 & 4, which would remain in operation during construction.
The project goal was to re-utilize as much of the existing casing and ductwork as possible in
order to minimize the cost of materials, and to also reduce the overall construction time period.
In addition to conversion of the existing box, a 12 foot extension was added to the existing
-------
casing to support the required amount of bags & cages to achieve a specified gross air-to-cloth
ratio of 8.5 to 1. A major obstacle was to complete the conversion in a 10-week boiler outage
window.
To achieve these goals, Research-Cottrell sub-contracted with Birmingham Industrial Services,
Inc. (BISCO), a union erection contractor, in association with Southern Energy Constructors,
Inc., of Birmingham, Alabama, for all the construction services, including demolition, erection of
the new equipment, painting, and electrical. In addition, Research-Cottrell provided an erection
advisor during construction, as well as start-up personnel. As much pre-outage construction as
possible was completed prior to the outage start. Due to the limited construction area available,
most of the steel panels and system components were installed in small sections, thus increasing
the amount of required man hours. Portions of the existing casing had to be cut into pieces small
enough to allow removal via wheel barrows.
BISCO and Research-Cottrell completed the majority of the construction on time, allowing for
on-time start-up of the system. Figure 4 shows the tight space and roof of the abandoned ESP
casing, including some of the many obstructions that had to be dealt with. Figure 5 is a front
view showing the modified I.D. Fans with the abandoned ESP casing in the background, which
is located under the hot-side ESP. The photo also shows that crane access to this area is highly
restricted.
Figure 4 - Top of Abandoned ESP Casing
-------
Figure 5 - Front of Construction Area Facing I.D. Fans
Startup and Initial Operating History
Prior to initial flue gas entry into the compartments, filter bags in both the A & B sides of the
system were precoated with an alumina silica material (pearlite) which, due to its light density,
provides uniform coating, and protection of the bags from potential upsets during initial start-up.
On December 21, 1996 as part of normal start-up procedures, the A-Side casing was precoated
and inspected for loose or improperly installed bags by injecting an ultraviolet leak detection
powder into the gas stream at the inlet to the casing. A few bags were found to be improperly
installed and were reseated into the tube sheet. Due to several system operating problems
experienced following the extended boiler outage, the B-Side casing was not precoated until
December 27th. After the bags were leak tested, only 6 bags were found to be improperly seated
and were re-installed.
Initial boiler startup and flue gas entry into the compartments occurred on December 27th, 1996
with gas flow entering both the A-Side and B-Side of the COHPAC collector at 1400 hours at a
boiler load of 130 MW. The cleaning system was initially programmed to allow cleaning to be
initiated based upon reaching a preset pressure drop level of 3.0". After the system had stabilized
the baghouse would be re-programmed to initiate cleaning based upon a preset filter cake drag
basis. The goal was to slowly raise the boiler load over the next few hours to develop a good
base filter cake. By December 28th, the unit was operating at 170 MW with baghouse inlet gas
temperatures of 256ฐ F on the A-Side and 267ฐ F on the B-Side. Initial tube sheet pressure drops
were measured at approximately 1" W.C. By the end of the day, the load was raised to 215 MW,
which provided inlet temperatures of around 275ฐ F. This load range was maintained for several
-------
days to allow proper seasoning of the filter bags. On December 29th, the bags were only required
to be pulsed twice during the early morning for the first time since start-up. Tube sheet pressure
drops were maintained at approximately 2" W.C. at this load level.
On January 2nd, the pressure drop cleaning cycle initiation set points were modified and raised
from 3.0" to 5.0", which allowed pulse cleaning down to 4.5" (1/2" reduction). This reduction in
pressure drop required a pulse duration of only 2-3 minutes. Initial pulse cleaning pressure set
points were maintained at 7.5 psi. On January 6th, the unit was released to Alabama Power to
allow boiler loads to be raised above 215 MW. Shortly thereafter, loads were increased to about
245-250 MW and pressure drops were maintained around 5" flange-to-flange. On January 15th, a
timer override was installed to force the cleaning cycle to initiate cleaning every 3 hours to avoid
a thick cake from forming and potentially limiting rapid load increases due to high pressure
drops and/or filter cake drags.
By January 20th, after 570 hours of operation, enough operating time and experience had
accumulated to allow the cleaning system to operate under a drag initiated cleaning mode (tube
sheet pressure drop divided by the air-to-cloth ratio). The initial drag set point was established at
0.5 in.H2O/ft/niin., which was automatically calculated by adjusting the flange-to-flange pressure
drop for duct losses to create a simulated tube sheet pressure drop in the program. This drag
initiation value was adjusted twice over the next couple of weeks to the final set point of 0.4 as of
January 30th. By this time, the unit had accumulated 849 hours of operation. The timer override
set point was also increased to initiate cleaning every 5 hours in the event a low load operating
period and/or fuel change allowed cleaning cycles to stretch out before reaching preset drag
initiation set points.
COHPAC Remote Monitoring System
For any fabric filter installation, the initial operation is the most critical time period. It can
ultimately determine whether a system is successful or not. During the initial operating period,
the system must be monitored around the clock to be able to react to system changes and allow
fine tuning of the baghouse system controls. This kind of monitoring can be costly, especially if
the job site is in a remote location or as was the case on this project, start-up occurred around the
Christmas holidays. With this in mind, and as part of this project, The Electric Power Research
Institute (EPRI), Southern Company Services and Alabama Power, contracted with Southern
Research Institute (SRI) to develop and install a remote monitoring system which can be
accessed via modem to allow real-time monitoring of the system operation, along with current
and long-term trend analysis.
Alabama Power found this system to be extremely useful, especially during the first month of
operation. They found comfort in knowing that their COHPAC system was being monitored by
Research-Cottrell and SRI on a regular basis, especially over the holidays. If a problem had
developed, or something looked abnormal, a simple phone to the site allows someone to look at
the system in a matter of minutes and diagnose the problem and recommend appropriate system
changes. The system operation is to be monitored by SRI for a minimum of one (1) year. More
details of this system and capabilities will be presented by SRI in a separate paper.
-------
Current Operating History
To date, the COHPAC system at E.G. Gaston Station, Unit #3 has operated exceptionally well.
Other than a few initial mechanical problems, the unit has operated better than anticipated. The
COHPAC I technology has provided Alabama Power with a high degree of system flexibility; it
eliminating ESP opacity spikes and reduces average stack outlet opacity levels. This increases
the amount of acceptable fuels available, and reduces overall operating costs.
On many hot-side electrostatic precipitator installations, to prevent or delay sodium depletion in
the ash from the boiler, many utilities, including Alabama Power, add sodium onto the coal feed
belts to help control ash resistivity and reduce the rate of plate and wires buildup. This buildup
can reduce the efficiency of the precipitators and allow higher opacity excursions, requiring the
utility in many cases, to reduce boiler generation output to remain in compliance. This buildup
can force the owner to periodically shut down the unit and wash or blast the precipitator
internals. The frequency of these outages varies with the specific coal properties being burned.
The cost of continuously using of sodium can be expensive and can greatly increase the
operating cost of the utility.
Even with ESP outlet opacity spikes higher than originally designed, the COHPAC system has
worked extremely well. During periods when the fuel characteristics have not been the best and
inlet opacity spikes have been high (which normally would have forced a load reduction on the
unit to maintain required stack opacity levels), the COHPAC system has maintained stack
opacities at low levels while maintaining full plant generating output. During episodes of high
inlet dust loadings, the COHPAC system automatically adjusts the cleaning frequency to handle
the upset condition. The COHPAC system normally cleans every 3-5 hours, however, when the
inlet particulate loadings are high, the PLC automatically adjusts to clean more often and may
move towards continuous cleaning to maintain preset drag set points. As soon as the upset
condition has ended, reducing inlet dust loadings to the bags, the cleaning cycle automatically
stretches back to normal time periods.
In addition to fuel variations, which have caused higher inlet loadings than expected, high inlet
dust loadings also occur during times when the plant overfires the Unit 3 boiler to provide steam
for the startup of Unit 5. Prior to the installation of the COHPAC system, Alabama Power had to
overfire all (1-4) units to supply startup steam to unit 5. Now only unit 3 is used with no increase
in stack opacity. Inlet dust loadings to the bags have increased by as much as 10 times during
this time period, which will automatically initiate a cleaning cycle depending upon the current
operating pressure drops. The length of cleaning varies depending on the amount of tune it takes
to reach the lower drag set point which will stop the cleaning cycle. This is usually equal to a
0.5" pressure drop reduction across the filter bags. Figure 6 shows the relationship to inlet
loadings vs cumulative hours over a few days of operation for COHPAC casing B.
-------
Differential Pressure
Calc. Drag x 10
3B1-Outlet Temperature
382-Outlet Temperature
InletTemperature
7/10
7/11
7/12
7/13
7/14
7/15
Figure 6 - Inlet Mass Concentration For COFTPAC Casing B
In April there were several days when high inlet dust loadings to the baghouse occurred. These
occurred on April 1-3, 9-11, 17-18 and 21-22. In response to these occurrences, and to prevent
the system from operating at long periods of continuous cleaning, the system drag cleaning
initiation point was adjusted from 0.4 to 0.44 in. FfO/ft/min on each casing (April 3rt) after 2,322
hours of operation, and the termination point was also raised from 0.32 to 0.36 in. H2O/ft/min.
The cleaning system parameters have successfully remained at these levels since that time.
Baghouse Acceptance Testing
Prior to conducting the formal COHPAC system performance/acceptance testing, the individual
compartments were inspected by both Alabama Power and Research Cottrell for signs of bag
leaks and/or bags that may have slipped out of the tube sheet hole after startup. Sample bags
were also pulled and shipped to Grubb Filtration for analysis. Only one bag had slipped out of
its tube sheet opening and needed replacement. All the rube sheet surfaces were clean and ash
free. The filter bags tested by Grubb were in excellent condition, with no noticeable strength
loss, nor wire impressions after over 3,000 hours of operation. It is anticipated that additional
bags will be periodically removed and sent out for similar analysis to allow tracking of bag life.
-------
Formal acceptance testing occurred on May 7th, 8th and 9th, with a total of five separate tests
conducted with test periods ranging from 3 to 4 hours per test. Simultaneous inlet and outlet
testing was conducted on both COHPAC casings while the boiler operation was held at full load
(270 MWe) during each test. As shown in the chart, the average outlet emission rates from
Casings A & B calculated based upon the use of an O2 "F" Factor ranged from 0.0028 to 0.0075
Ib/MMBtu for the five test periods, with the average of all tests equal to 0.0046 Ib/MMBtu
(0.002 gr/dscf).
The higher spike in outlet emissions during Test 5 was a result of an unusual test period when
both front comer and back center TR sets (four total) were shut off to demonstrate the effect on
the COHPAC system performance from higher than design inlet ash loadings.
The emission levels increased as a result of requiring a more aggressive cleaning cycle (shorter
time interval between cleaning cycles) to maintain the same system pressure drops. The required
pulses/bag/hour required to maintain a programmed pressure drop level of approximately 5.5"
Outlet Emission Test Results
Test 1 Test 2 Test 3 Test 4 Test 5
A-Side
B-Side
W.C. were as much as 2.93 on the B-Side casing during this test. During this test, the B-Side
inlet dust loading was measured at 0.206 gr/dscf versus an inlet dust loading of 0.085 gr/dscf on
the A-Side casing. The difference in inlet loadings to the casings may be attributed to a
difference hi hot-side ESP performance on the two sides, and/or balancing hi the boiler. Most of
the operating experience to date shows that there is a higher dust loading to the B-Casing,
although gas volumes are fairly close to each other. The PLC automatically adjusts for these
differences and controls the required cleaning cycles independently for each casing.
While the pulse frequencies during Test 5 were higher than normal, the average pulse frequencies
for the entire testing period over both casings were 0.89 pulses/bag/hour. Normal cleaning
frequencies average 0.5 pulses/bag/hour during normal system operation.
-------
Conclusion
Alabama Power Company's decision to purchase and install an advanced version of COHPAC I
at their B.C. Gaston steam plant from Research-Cottrell was an excellent decision. The system
has operated exceptionally well from initial startup with few mechanical problems and has
provided Alabama Power with a highly efficient system that is reliable and forgiving.
Performance to date shows that a properly designed COBS'AC system can operate with both on-
line cleaning and long filter bags at filtration rates of 8.5 fpm while providing low outlet
emission levels (<0.01 Ib/MMBtu) and reduced pressure drops, even with occasional high inlet
ash loadings. While long-term bag life still needs to be demonstrated, to date the Ryton felt bags
are in excellent shape, with little if any strength loss and high permeabilities (low pressure
drops).
Acknowledgments
The authors wish to acknowledge the support and assistance of the following individuals who
each has contributed to the success of this advanced, full scale COHPAC I installation. Tom
Frazer, Byron Corina, Johnny Rush and Brian Barham (now at Miller Station) of Alabama
Power's B.C. Gaston station, Gary McGrath, Michael Kelaher, Gerry Smith and Steve Kozma of
Research-Cottrell, Ken Gushing of Southern Research Institute (SRI), Theron Grubb of Grubb
Filtration, Inc., Ray Wilson of Ray Wilson Consulting and finally Rusty Jones of Southern
Energy Construction (Birmingham Industrial Services) for his efforts in ensuring the success of
this difficult construction effort.
References
1. R.L. Chang, "COHPAC compacts emission equipment into smaller, denser unit", Power
Engineering, July 1996, pp. 22-25.
2. R. L. Miller, "Advanced Technology Development Dry Filtration Systems", Council of
Industrial Boiler Owners - ADVANCED TECHNOLOGIES E ROUNTABLE, July
1996, Washington, DC.
3. W. A. Harrison, K. M. Gushing, R. L. Miller, R. L. Chang, "Recent COHPAC Data for
Fine Particulate Matter & Air Toxics Removal from Coal-Fired Power Plants", In
Proceedings: Power-Gen International 96, Orlando, FL, December 4-6, 1996.
4. K. M. Gushing, "SRI remote baghouse monitoring system for Plant Gaston", Clear
Stacks, Reinhold Environmental Ltd., December, 1996.
5. R. Jones, "Gaston Rebuild Project tops site difficulty", Clear Stacks, Reinhold
Environmental Ltd., December, 1996.
-------
PREDICTING COHPAC PERFORMANCE
C. J. Bustard
ADA Environmental Solutions
7931 S. Broadway, #349
Littleton, CO 80122
S. M. Sjostrom
2416 Emerson St.
Denver, CO 80205
Ramsay Chang, Ph.D.
Electric Power Research Institute
P.O. Box 10412
Palo Alto, CA 94303
Executive Summary
COHPAC (Compact Hybrid Particulate Collector) is a patented Electric Power Research Institute
(EPRI) technology developed to assist coal-fired utilities in cost effectively meeting future, more
stringent emission regulations. COHPAC is showing promise as a technology for polishing
particulate emissions, and hi combination with sorbent injection, may also be a viable technology
for mercury and SO2 control.
With COHPAC, a pulse-jet baghouse is operated at high air-to-cloth ratios because of the
relatively low inlet grain loadings downstream of existing electrostatic precipitators. The high
air-to-cloth ratio reduces the size of the baghouse and therefore the capital expenditure.
Performance data compiled from pilot and full-scale Installations over the past seven years have
provided insight into the sensitivity of tubesheet pressure drop to the primary performance
variables: air-to-cloth ratio, inlet grata loading, time between cleans, and specific dustcake/ash
characteristics.
An empirical model was developed from actual performance data obtained from COHPAC pilot
tests and full-scale systems. This model predicts operating drag (tubehseet pressure drop/air-to-
cloth ratio) as a function of dustcake accumulated between cleans. It also provides a method to
estimate a target air-to-cloth ratio when design parameters such as tubesheet pressure drop and
cleaning frequency are known.
-------
Introduction
The Electric Power Research Institute (EPRI) began development of the COHPAC (Compact
Hybrid Particulate Collector ) technology in 1989 as a cost-effective alternative for coal-fired
utilities to meet future, more stringent emission regulations. COHPAC is a "high" air-to-cloth
ratio pulse-jet baghouse installed downstream of an existing electrostatic precipitator. COHPAC
can operate at high air-to-cloth ratios (typically between 8 and 12 ft/min) because of the
relatively low Inlet grain loadings downstream of existing electrostatic precipitators. Because of
the reduced size of the baghouse, capital expenditure is less than other options such as a reverse
gas baghouse or a new electrostatic precipitator.
Early COHPAC development took place at the pilot scale using EPRI's 1 MW(e) transportable
pulse-jet baghouses. Based on promising performance results, a 145 MW demonstration was
conducted at TU Electric's Big Brown Station1, followed by commercial installations at Big
Brown and Alabama Power's Gaston Station2. Performance data from these installations were
compiled and analyzed for trends to develop a predictive model for COHPAC performance. A
statistically significant trend was found between drag and dust loading deposited on the bags
between cleans. This trend can be expressed as a form of Darcy's equation, which is often used
to predict baghouse pressure drop.
A semi-empirical model would be a valuable design tool because performance at the COHPAC
installations has widely2'3 This model would provide better estimates of COHPAC pressure
drop at different air-to-cloth ratios, dust loadings, and cleaning frequencies.
COHPAC Database
The database is made up of operational and performance data from COHPAC pilot tests and full-
scale installations. Only data from COHPAC installations that were filtering flyash from coal-
fired PC boilers, and had no other sorbent material injected for either acid gas or air toxics
control, were used. In all cases, COHPAC was downstream of the primary particulate control
device and in all cases this was an electrostatic precipitator. The database consists of five
installations at three different sites and represents nearly 26,000 hours of operation over seven
years.
Table 1 provides descriptive information about each of the three sites used in the database.
Information includes coal type, electrostatic precipitator type and Specific Collection Area
(SCA). Both eastern and western low sulfur coals have been evaluated, as well as Texas lignite.
There has been limited testing on either medium- or high-sulfur coals because the electrostatic
precipitators at these sites generally perform within the precipitator design basis. As these sites
consider either switching to or blending with low sulfur coals to reduce sulfur emissions,
precipitator performance will degrade and upgrade options such as COHPAC will have to be
considered.
-------
Table 1
Database Plant Descriptions
Plant Name
TU Electric
Big Brown Station
Southern California
Edison
Mohave Station
Alabama Power
Gaston Station
Coal Type
Texas Lignite
Black Thunder -
western low sulfur
subbutminous
Jim Walter -
Alabama low sulfur
ESP Manufacturer Specific Collection
Area
Research Cottrell,
weighted wire,
chevron
Research-Cottrell,
weighted wire,
chevron
Buell BA weighted
wire, hot-side
156
112
280
Table 2 presents operational and performance data from each installation and represent typical
average values after COHPAC operation was stable. Operational data include air-to-cloth ratio,
inlet dust loading, and mass median size of the particulate matter. Measured performance
parameters are drag and cleaning frequency. Areal dust loading on the bags between cleans is
calculated from the performance parameters. All installation used an 18 oz/yd2 Ryton felted
fabric. For the full-scale installations, these values were taken at full-load conditions.
Table 2
Station
Big Brown
Big Brown
Big Brown
Mohave
Gaston2
Installation
Units 2 & 3
Units 1 & 3
Pilot
Pilot
Units
Average
Air-to-
Cloth
Ratio
(ft/min)
11.3
11.3
18
12
8
Average
Drag
(inches
HzO/ft/rnin)
0.67
0.70
0.33
0.38
0.41
Inlet Dust
Loading
(gr/acf)
0.130
0.130
0.023
0.022
0.051
Cleaning
Frequency
(cleans
per hour)
3.3
3.3
2
1.7
1.27
Mass
Median
Diameter
(microns)
4
4
6.51
6
0.009
Area!
Loading
0.0038
0.0038
0.0017
0.0013
0.0027
Model Development
For an engineer trying to determine if COHPAC is appropriate for a specific site and the design
basis for unit size, the most appropriate method available is to use a derivation of Darcy's
equation to predict pressure drop for a range of operating conditions4 This equation is
-------
traditionally used to predict the pressure drop across a filter bag, equation 1. Or by dividing by
velocity, Darcy's equation predicts drag, equation 2:
APd = (Kf+K2Wr)V + K2CtV2/7000 (1)
or APd/V = Kr + K2CtV/7000 (2)
where
APd = differential pressure across baghouse tubesheet (inches H2O);
APd/V = drag (inches H2O/ft/min)
Kf = fabric/dust resistance coefficient (inches H2O*min/ft)
Wr = residual dustcake weight (Ib/ft2)
V = air-to-cloth ratio (ft/min)
K2 = specific dustcake resistance coefficient (inches HiO-ft-min/lb)
C = dust loading (grains/acf)
t = filtration time between bag cleaning (min)
Kr = (K,+K2Wr)
The residual drag (Kr) typically increases over time and is influenced by the cleaning frequency
and permeability of the fabric and residual dustcake layer. K2 is the specific dustcake resistance
coefficient, a function of flyash particle characteristics that include size and morphology. The
term CtV represents the areal dust loading and is the mass of dust buildup on the filter bag
between cleans.
Figure 1 shows that the drag is indeed linearly correlated to the areal dust loading with an R2
value is 0.91. This graph predicts that the residual drag, Kr, for COHPAC is 0.13 Inches
H2O/ft/min and the specific dustcake resistance coefficient, K2, is 140 inches H2O*ft*min/lb. K2
values for reverse-gas baghouse are normally around 10 inches H2O*ft*min/lb. The significant
differences in the K2 values of COHPAC versus conventional baghouses is possibly the much
finer particle sizes collected by COHPAC. Although it is encouraging that the COHPAC
database follows Darcy's Law, there are discrepancies between field measured Kr and the model
predicted Kr.
Examples of Predicting COHPAC Performance
A series of graphs was developed for a range of estimated COHPAC operating parameters and
acceptable performance conditions using the semi-empirical data fit model. Figure 2 presents a
graph that estimates drag as a function of inlet grain loading for 3 different air-to-cloth ratios,
assuming a cleaning frequency of 2 cleans per hours. This graph shows that for a desired drag of
0.5, an air-to-cloth ratio of < 12 ft/min should only be specified for inlet grain loadings up to 0.05
gr/acf (0.1 Ibs/MMBtu). If inlet grain loadings up to 0.1 gr/acf (0.2 Ib/MMBui) are expected, the
design air-to-cloth ratio should be 8 ft/min.
Figure 3 shows drag as a function of air-to-cloth ratio for 4 different inlet grain loadings. The
dramatic increase hi slope with increased grain loading illustrates the sensitivity of drag to air-to-
-------
cloth ratio at high dust loadings. At a grain loading of 0.02 gr/acf, drag increases by less than
0.05 inches I^O/ft/min when air-to-cloth doubles from 8 to 16 ft/min. At a grain loading of 0.2
gr/acf, drag increases by 0.5 inches J^O/ft/min, which significantly impacts tubesheet pressure
drop and cleaning frequency.
Figure 4 shows how drag is affected by cleaning frequency for 3 inlet grain loadings and an air-
to-cloth ratio of 8 ft/min. On this graph, the curves are relatively flat until a point where larger
areal loading between cleans causes sharp increase in drag. It is noteworthy that if COHPAC
operation is on the steep part of this curve, small decreases in cleaning frequency can
significantly impact drag.
Figure 5 provides a quick method to determine acceptable inlet grain loadings to maintain a
specific drag point at different air-to-cloth ratios and clean frequencies, hi this graph, inlet grain
loading is plotted as a function of cleaning frequency at 3 different air-to-cloth ratios and a drag
of 0.5 inches HjO/ft/min. If it is desirable to design COHPAC at an air-to-cloth ratio of 12
ft/min, the maximum inlet grain loading at 2 cleans per hours is 0.057 grains/acf. If 3 cleans per
hours is acceptable, the maximum grain loading increases to 0.08 gr/acf. For an air-to-cloth ratio
of 8 ft/min at 2 and 3 cleans per hours, the maximum inlet grain loadings are 0.072 and 0.115
gr/acf, respectively.
0.70
0.65
0.60
1c
I ฐ'55
I 0.50 i
c
g> 0-45
Q
0.40-
0.35
0.30
y= 140.59X+0.1288
R2 = 0.9058
0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004
Areal Loading (Ib/ft2)
Figure 1
Linear Fit of Drag Versus Areal Loading
-------
1.4 --
0.8 +
=- 0.6 --
D)
D 0.4 - -
0.2 --
0.0
12Wmin
10ft/min
SfVmin
0.05 0.1 0.15
Participate Loading (gr/acf)
0.2
Figure 2
Drag as a Function of Inlet Grain Loading for 3 Different Air-to-Cloth Ratios
1.4
1.2
9, 0.8 --
X
i- 0.6 -
01
Q 0.4 - -
0.2 -
0.0
0.2 gr/acf
0.1 gr/acf
" "6.5s gr/acf
0.02 gr/acf
2 cleans/bag/hour
10 12
Air-to-Cloth Ratio (ft/min)
14
16
Figure 3
Drag as a Function of Air-to-Cloth Ratio for Different Inlet Grain Loadings
-------
1.4 -- A/C = 8ft/min
1.2 --
I 1.0 --
I 0.8-
c
= 0.6 - -
o>
Z
a 0.4 - - v
0.2 --
0.0 -I 1
0.2 gr/acf
'-._ 0.05 gr/acf
0.02 gr/acf
-t-
0.5 1 1.5 2 2.5
Cleaning Frequency (cleans per hour)
Figure 4
The Effect of Cleaning Frequency on Drag at 3 Inlet Grain Loadings
0.130
w 0.110
o
0.090 -
0.010
Drag = 0.5 in l-bO/fl/min
= 8fVmirvx A/C=10ft/min
0.5
1.0 1.5 2.0 2.5 3.0
Cleaning Frequency (cleans per hour)
Figures
Grain Loading as a Function of Cleaning Frequency at 3 Different Air-to-Cloth Ratios
-------
Examples of Predicting Air-to-Cloth Ratio for Specific Conditions
To illustrate how the linear data fit model can be used to estimate air-to-cloth ratio and
subsequent foot print of a COHPAC baghouse, design parameters were chosen for a hypothetical
station with an electrostatic precipitator that has a maximum emission rate of 0.17 Ib/MMBtu.
At this site it is assumed that the operating drag should be 0.5 inches HiO/ft/min and the
maximum cleaning frequency should be 2 cleans per hour. The maximum flue gas temperature
is 340ฐF, which is a range where Ryton felted fabrics can be used. This baghouse will be
designed for on-line cleaning. Table 3 summarizes the design basis and presents the calculated
air-to-cloth ratio for these conditions. The COHPAC model predicts that this baghouse should
be conservatively designed for a maximum, net (one compartment off-line for maintenance) air-
to-cloth ratio of 8.2 ft/min.
Parameter
Table 3
Design Parameter
Calculated Design Value
Maximum Inlet Loading
Maximum Tubesheet Pressure
Drop
Maximum Cleaning Frequency
Average Drag (Design)
Cleaning Logic
Maximum Air-to-Cloth Ratio
0.075 gr/acf (0.17 Ib/MMBtu)
6.0 inches H2Oa
2 cleans per hour
0.5 inches H2O/ft/min
On-Line Cleaning
8.2ft/min
a. By definition of drag, maximum air-to-cloth ratio is 12 ft/min. Model will provide predicted maximum air-to-
cloth ratio to maintain the design drag value.
To verify the model, actual operating conditions from the Big Brown COHPAC installation were
used to estimate operating drag. At this site the inlet grain loading is 0.13 gr/acf, the cleaning
frequency is 3.3 cleans per hour, and the air-to-cloth ratio is 11.3 ft/min. At these conditions, the
predicted drag is 0.66 inches HaO/ft/min. After over a year of operation, the average operating
drag at Big Brown is 0.68 inches HiO/ft/min. The model accurately predicts drag for this
installation. There is, however, an inconsistency between the field measured and the predicted
residual drag at Big Brown that is still being investigated.
-------
Future Model Development
There are several areas where model development is continuing in addition to resolving the issue
of measured versus predicted residual drag. These include incorporating the impact of particle
size distribution and particulate matter other than flyash on drag prediction.
Particle Size Distribution
The data presented in Table 2 does not reflect a significant difference in particle size distributions
among the 3 sites. The range in mass median diameter was from 4.0 to 6.5 microns. It has been
demonstrated that on a reverse gas baghouse, pressure drop increases as particle size decreases5
Tests performed under EPRI funding have confirmed that particle size can significantly impact
COHPAC performance. The impact of particle size distribution on rate of increase in pressure
drop will be evaluated in the near future.
Using COHPAC to Scrub Acid Gases or Air Toxics
A COHPAC type configuration can be used with sorbent injection to scrub acid gases or air
toxics. This configuration can only be used when the grain loading exiting the existing
particulate control device is at a level where additional grain loading from the sorbent will not
cause high pressure drop. EPRI has patented this concept for air toxic control and it is referred to
as TOXECON.
TOXECON was developed for advanced control of vapor phase mercury, hi TOXECON, carbon
sorbents are injected upstream of a high air-to-cloth ratio pulse-jet baghouse. The primary
advantages of this configuration are that carbon is not mixed with the bulk of the flyash where it
could adversely impact flyash sales and that there is the potential to recycle carbon for better
utilization.
Several test programs evaluating TOXECON have been conducted with results presented at this
conference. These data will be evaluated to determine if baghouse performance can be predicted
as a function of sorbent type and quantity. The initial approach would be similar to one
addressing flyash particle size distribution, as mentioned earlier.
Test programs evaluating the injection of calcium and sodium based products upstream of
COHPAC were conducted at four different utility sites. Again, these data are being presented at
this conference and will be evaluated in terms of baghouse performance modeling in the near
future.
-------
Conclusions
1. The traditional baghouse model based on Darcy's Law appears to correlate COHPAC
performance well. However, the Kr and K^ values are only applicable to coal flyash as it is
expected that particle properties such as size and shape will strongly influence Kr and K^
values. Thus, new data will be needed to determine Kr and K^ values for carbon or sodium
sorbent injection. These data are specific to COHPAC and it should not be assumed that they
can be used for conventional baghouses.
2. The semi-empirical model predicts COHPAC performance accurately for existing units,
however, there are still some issues to be resolved. Because of inconsistencies between
predicted and field measured residual drag, efforts are continuing to resolve and understand
this issue.
3. A series of graphs were developed for easy evaluation of COHPAC performance at a
range of operational variables. These graphs illustrate the sensitivity of COHPAC
performance as a function of each of the primary variables: inlet grain loading, air-to-cloth
ratio, and cleaning frequency.
4. For installations where dust loading is < 0.1 gr/acf (0.2 Ib/MMBtu), a maximum air-to-
cloth ratio of 8 ft/min should be considered. Figure 2 shows how drag varies as a function
of particulate loading for 3 different air-to-cloth ratios and a maximum cleaning frequency of
2 cleans per hour. At a target drag of 0.5 inches H2O/ft/min, an air-to-cloth ratio of 8 ft/min
is recommended for an inlet grain loading of 0.1 gr/acf (0.2 Ib/MMBtu).
References
1. A.K. Hindocha, B. Brown, R. Chang. "Commerical Demonstration of COHPAC" Presented
at the Tenth Particulate Control Symposium & Fifth International Conference on Electrostatic
Precipitation, Washington, D.C, April 5-8, 1993.
2. R. Miller, W. Harrison, D. Prater, R. Chang, K. Gushing, R. Wilson. "Installation and
Performance Monitoring of COHPAC I and COHPAC I Plus on Alabama Power
Company's E.G. Gaston Unit 3." Presented at the EPRI-DOE-EPA Combined Utility Air
Pollutant Control Symposium, Washington D.C, August 25-29, 1997.
3. B. Browning, A.K. Hindocha, A. Casey, J Bustard, T. Grubb. "Operation and Performance
of COHPAC at TU Electrics' Big Brown Station". Presented at the EPRI-DOE-EPA
Combined Utility Air Pollutant Control Symposium, Washington D.C., August 25-29,1997.
4. J.D. McKenna and J.H. Turner. "Fabric Filter - Baghouse I, Theory, Design, and Selection
(A Reference Text)" Copyright, 1989 By ETS, Inc., Roanoke, Virginia. Library of Congress
Catalog Card Number: 89-84672.
-------
5. C.J. Bustard, K.M. Gushing, I. Nicol. "Monthly Technical Report No. 89 - Paniculate
Emission and Operating Characterization of a Fabric Filter Pilot Plant". Prepared for the
Electric Power Research Institute Contract No. 1129-8. September, 1987.
-------
Performance Response of a COHPACI Baghouse During Operation
with Normal and Artificial Changes in Inlet Fly Ash Concentration and
During Injection of Sorbents for Control of Air Toxics Emissions
by
Kenneth M. Gushing
Southern Research Institute
P. O. Box 55305
Birmingham, AL 35255-5305
Wallis A. Harrison
Southern Company Services
P.O. Box 2625
14N8195
Birmingham, AL 35202-2625
Ramsay L. Chang
Electric Power Research Institute
P. O. Box 10412
Palo Alto, CA 94303
Abstract
Both the Clean Air Act Amendments of 1990 and the proposed National Ambient Air
Quality Standards for Particulate Matter and Ozone set emission standards that challenge
the capabilities of existing utility emissions control devices downstream of coal-fired
power boilers, especially electrostatic precipitators (ESPs), at many sites around the
United States. Fuel switching for control of chemical emissions (SOX and NOX) and
injection of various sorbents for chemical and air toxics emissions control have been
shown to severely affect the particulate collection efficiencies of ESPs.
As particulate removal requirements become more stringent, successful control
technologies will be required to combine high removal efficiencies with system flexibility
and cost-effectiveness. The use of advanced COmpact Hybrid P Articulate Collector
(COHPAC) technologies can meet and/or exceed New Source Performance Standards
(NSPS) and provide consistent clear stack conditions while providing cost savings of as
much as 60% in capital cost requirements in comparison to more traditional designs.
To better understand the capabilities of COHPAC technologies to meet these stringent
requirements, EPRI has funded research on a pilot-scale COHPAC I unit at Alabama
-------
Power Company's Miller Electric Generating Station near Birmingham. Recently tests
were conducted to evaluate COHPAC I performance changes during the injection of fly
ash and activated carbon. Fly ash was injected to simulate operation when the
performance of the upstream control device (hot-side ESP, for example) degraded.
Activated carbon was injected to study COHPAC baghouse performance changes when
the unit was challenged with various concentrations of this fine paniculate material used
for control of mercury emissions. Data were also gathered during the normal operation of
the COHPAC unit during natural changes in the inlet fly ash concentration as the
performance of the upstream particulate control device changed.
Data are presented showing the effect of fly ash and sorbent injection on baghouse
performance (air-to-cloth ratio, pressure drop, pulsing frequency). These data
demonstrate the sensitivity of COHPAC baghouse performance (i.e., pulse frequency) to
changes in inlet mass concentration and particle size at high filtering air-to-cloth ratios.
Introduction
Electrostatic precipitators and low-ratio fabric filters (baghouses) have been used
extensively by U. S. utilities over the last 35 years to control particulate emissions at coal-
fired power plants. These control devices, nominally 1,200 ESPs and 110 baghouses,
have a good performance record and have enabled utilities to improve air quality and meet
applicable emissions and opacity standards. Currently, four factors are challenging utilities
to improve their particulate control capabilities:
the affect of the federal 1990 Clean Air Act Amendments (CAAA) on ESP
performance
more-restrictive particulate emission requirements by various state and local agencies
the aging of ESPs currently in service, and
the possible new limits on air toxics emissions due to CAAA Title HI or state
regulations.
Most utilities are considering ways to cost effectively improve particulate control, given
the factors described above and the accelerating push toward deregulation which will
demand reduced operating costs in order to remain competitive.
There are a number of ways in which ESP performance can be improved. These include
flue-gas conditioning, precharging, intermittent and pulsed energization, increasing the
size of the ESP, and replacement of ESPs with low-ratio fabric filters. There are site-
specific situations, however, in which these various options may not provide adequate
performance improvement. Also, these methods of improvement might not be applicable
because of space constraints or they may not be economical. The Compact Hybrid
Particulate Collector (COHPAC), developed in the early 1990s by the Electric Power
Research Institute (EPRI), can provide a cost-effective alternative to ESP upgrade
-------
options. COHPAC technology can be configured in various ways to meet plant-specific
conditions and objectives.
Implementation of COHPAC technology is simple. A filtering system, typically a pulse-jet
baghouse, is installed between the existing ESP, or other poorly performing participate
control device, and the stack. It collects any particles passing the ESP. The existing ESP
significantly reduces the particulate concentration in the flue gas and imparts an electrical
charge to those particles entering the filter. The charged particles then help to form a
more porous filter cake on the fabric thereby providing lower flow resistance. A
consequence of the lower particulate loading and charge imparted to the fly ash is that the
flue gas can pass through the filter bags of a COHPAC system at significantly higher
velocity (four to eight times) than in comparable baghouses (low ratio) used as primary
particulate collectors.'
Pulse-jet fabric filters are ideal for COHPAC systems because a large number of bags can
be housed in a relatively small space, filtering air-to-cloth ratios can be high because of the
reduced particulate loading in the flue'gas, and on-line cleaning reduces the need for a
significant degree of compartmentalization. This corresponds to a reduction in baghouse
size and cost when compared to normal low-ratio baghouses. There are two types of
COHPAC systems available. COHPAC I utilizes a stand-alone PJFF unit downstream of
the existing electrostatic precipitator. In a COHPAC n system the pulse-jet fabric filter is
housed in the last field of the existing ESP.
There are currently two plants where full-scale applications of COHPAC I technology are
in use. The first is at TU Electric's Big Brown Plant (Units 1 and 2, each 575-MW) near
Fairfield, Texas. The second is at Alabama Power Company's E. C. Gaston Electric
Generating Plant (270-MW Unit 3) near Birmingham, Alabama. Prior to each full-scale
system being built, a pilot-scale COHPAC I pilot plant was utilized to evaluate system
performance under similar operating conditions, plus evaluate the ability of COHPAC I to
control emissions of fine particulate and hazardous air pollutants. For Plant Gaston
simulation testing, one of EPPJ's 1-MW Transportable Pulse-Jet baghouses was installed
adjacent to Unit 2 at APC's Miller Electric Generating Plant outside Birmingham. At
Plant Gaston the COHPAC I baghouse was installed downstream of Unit 3's hot-side
electrostatic precipitator and air preheater. The pilot unit at Plant Miller was installed
similarly, downstream of Unit 2's hot-side ESP and air preheater. Both plants bum similar
coals, so the flue gases treated at Plant Miller have been similar to those at Plant Gaston.
Description of the Pilot Plant
The EPRI Transportable Pulse-Jet Baghouse was originally sized to filter approximately
5,000 acftn at a nominal air-to-cloth ratio of 4 ft/min. To simulate COHPAC I operation,
the pilot plant was modified to provide interstitial can velocities of < 1,000 fpm while
operating at an air-to-cloth ratio of 8.5 ft/min. A total of twelve, 24 foot long Ryton felt
bags were installed to achieve the desired filtration rates. The bags are cleaned with Low
-------
Pressure/High Volume (LPHV) pulse jet technology which has been successfully utilized
on COHPAC I pilot and full scale i3A5 applications in the past and would be installed on
the E. C. Gaston full scale COHPAC facility. The pulse-jet baghouse was set for pressure
drop initiated cleaning. We used on-line cleaning with pulse pressures of approximately
11 psi. The pilot plant began operation in September 1995 and operated successfully in
excess of 8,000 hours on Ryton bags. In October 1996 P84/Ryton composite bags were
installed in the pilot facility. Through June 1997 these bags had experienced over 5,000
service hours.
During periods of normal operation the average air-to-cloth ratio ranged from 8 to 10
fh/min, the average tubesheet pressure drop varied from 4 to 6 in. B^O, depending on the
cleaning set points, and the average pulse frequency ranged from 0.05 to 0.5 p/b/h. The
inlet mass concentration ranged from 0.004 to 0.04 Ib/MMBtu, depending on the
condition of the upstream hot-side ESP, while the outlet emissions ranged from 0.00007
to 0.001 Ib/MMBtu.
Test Program
One of the many observations during this test program has been the sensitivity of pulse
frequency on inlet paniculate mass concentration. Since long bag life is an important
economic consideration in the application of baghouse technology, it is important to
understand how COHPAC PJFFs respond to an increase in fly ash loading and the
introduction of sorbent materials to the inlet flue gas stream. For this particular test series,
we were interested in how the COHPAC I baghouse would respond to the injection of
activated carbon, which might be utilized in the future at full scale for control of mercury
emissions. Of concern also was COHPAC I response to increased inlet fly ash loadings
that could occur upon the degradation of the performance of the upstream primary
paniculate control device.
To conduct these tests we obtained a quantity of activated carbon ("FGD Carbon" from
Norit America, Inc.) and several barrels of fly ash. The fly ash was removed from the
outlet field hoppers of the Plant Miller Unit 3 ESP. We selected this site so as to simulate
the fine paniculate expected from the outlet of a hot-side ESP on Unit 2. To feed the
materials at a controlled rate we used an AccuRate Dry Materials Feeder and a K-
Tron Dry Materials Feeder. Each feeder could provide injection rates up to 500 grams
per hour. During our tests with activated carbon, we evaluated COHPAC I performance
with and without co-injection of fly ash. The materials were injected through one of the
inlet sampling ports on the pilot baghouse which were located approximately 10 feet
upstream of the baghouse's hopper inlet transform. The materials were educted into the
inlet flue gas stream using ambient air as a carrier gas.
-------
Test Results
The results of the tests are described in the following two subsections. The first describes
the tests with activated carbon, both with and without co-injection of fly ash. The second
subsection describes the tests with fly ash only.
Activated Carbon
We evaluated COHPAC I PJFF performance under controlled conditions during injection
of fly ash alone, with the simultaneous injection of fly ash and activated carbon, and finally
with activated carbon alone. These tests were performed to simulate COHPAC I
operation downstream of a poorly performing electrostatic precipitator both with and
without injection of activated carbon as a mercury sorbent. Several graphs have been
prepared to show the effects on performance caused by the injection of these materials.
Figure 1 shows data for Test Block #1 when we were injecting fly ash and carbon
simultaneously. The bottom graph shows the effective baghouse inlet mass loading, as
lb/106 Btu, for each of the nine test periods. During the first four test periods there was
no injection of ash or activated carbon. We estimated the inlet mass loading from the
Miller Unit 2 opacity data using a previously determined relationship. During Test Period
5 we injected fly ash at a rate of about 350 grams per hour. For the next three Test
Periods, 6, 7, and 8, the fly ash injection rate remained the same. During Test Period 6 we
injected carbon at a rate of 100 grams per hour. For Test Periods 7 and 8 we increased
the carbon injection rate to 200 grams per hour. For Test Period 9 we stopped injection
of both ash and carbon. The total mass loading into the baghouse ranged from as low as
about 0.01 lb/106 Btu during Test Periods 1 through 4 to as high as 0.165 lb/106 Btu
during Test Period 8.
The upper graph in Figure 1 shows the average drag and pulse frequency during each of
the nine test periods. During the first 7 test periods the drag varied between 0.65 and 0.7
in. EkO/ft/min. For the first 4 test periods the pulse frequency remained low, however, it
was sensitive to small changes in the inlet mass loading, increasing by a factor of 4 as the
inlet mass loading increased from 0.006 to 0.012 lb/106 Btu. When ash injection began
during Test Period 5 the pulse frequency increased by a factor of 10 from 0.03 to 0.3
pulses per bag per hour (p/b/h). When carbon injection started during Test Period 6 the
pulse frequency increased by a factor of 2 from 0.3 to 0.6 p/b/h. For Test Period 7 the
carbon injection rate was doubled; the pulse frequency increased by a factor of 2.5 to 1.5
p/b/h. To control the pulse frequency we raised the drag from 0.7 to 0.86 in. HzO/ft/min
for Test Period 8. The pulse frequency fell to 0.5 p/b/h, one-third of its previous value.
For Test Period 9 we stopped both ash and carbon injection and lowered the drag back to
its former value. The pulse frequency returned to a value similar to those experienced
prior to the injection of ash and carbon, 0.17 p/b/h.
-------
Figure 2 shows COHPAC I performance data during Test Block #2 when we only injected
activated carbon. The bottom graph shows the effective baghouse inlet mass loading
during each test period. We started with a period of operation with no injection. For Test
Periods 2 and 3 we injected carbon at a rate of about 100 grams per hour. This was
followed by two test periods with a carbon injection rate of 200 grams per hour. After a
weekend break with no injection, Test Period 6, we resumed carbon injection at 200
grams per hour. This was followed by the two final test periods where the carbon
injection rate was raised to 300 grams per hour.
The drag and pulse frequency data for Test Block #2 are shown in the upper graph in
Figure 2. To make the changes in drag easier to discern, the values of drag were plotted
at 10 times their actual value. For this test period we began with the baghouse operating
in a more aggressive manner. The drag was reduced to 0.47 in. BbO/ft/min for the start of
this test block. During the first test period with no injection the average pulse frequency
was 0.23 p/b/h. When carbon was injected at a rate of 100 grams per hour during Test
Period 2 the average pulse frequency increased by almost a factor of 10 to 2.1 p/b/h. To
lower the pulse frequency we increased the average drag to 0.62 for Test Periods 3 and 4.
For Test Period 3 the average pulse frequency fell to 0.6 p/b/h. We then increased the
carbon injection rate to 200 grams per hour during Test Period 4. The average pulse
frequency rose by almost a factor of 6 to 3.5 p/b/h. For Test Period 5 we raised the drag
to 0.73 in. FtzO/ft/min. By doing this we were able to reduce the average pulse frequency
to 2.7 p/b/h. We did not experience a large decrease in pulse frequency between these
two test periods because we increased drag only 18%, 0.62 to 0.73 in. H^O/ft/min
During the previous change in drag, 0.47 to 0.62 in. H2O/ft/miri, the increase was 32%.
For Test Period 6 we turned off the carbon injection system. The average pulse frequency
fell to 0.13 p/b/h.
We resumed carbon injection during Test Period 7. The average pulse frequency rose
only to 0.6 p/b/h. During Test Period 5 the pulse frequency had been 2.7 p/b/h. Two
factors could have caused this. We increased the drag from 0.73 to 0.77 in. H2O/ft/min by
lowering the air-to-cloth ratio from 8.6 to 8.2 ft/min and we operated for 70 hours
without carbon injection. This extended period without injection may have cleaned the
bags and allowed a fresh layer of fly ash to form on the fabric surface. This ash layer may
have aided in the formation of a more permeable cake once carbon injection resumed. For
Test Period 8 we increased the carbon injection rate to 300 grams per hour. The average
pulse frequency rose by a factor of 3.5 from 0.6 to 2.2 p/b/h. The only change we made
for Test Period 9 was to lower the drag back to 0.73 in. FJbO/ft/min by raising the air-to-
cloth ratio from 8.2 back to 8.6 ft/min. The average pulse frequency rose from 2.2 to 5.7
p/b/h. This again shows the sensitivity of baghouse performance to small changes in
baghouse flow rate at these high air-to-cloth values accompanied by a relatively high inlet
mass loading (~ 0.1 lb/106 Btu). During Test Period 9 we used up all of the activated
carbon. This completed our test program with carbon and ash injection.
-------
Fly Ash
Our detailed studies of COHPAC IPJFF performance with higher than normal inlet
particulate mass concentrations occurred in 1997 following the change in filter bags in
October 1996. Prior to this testing we demonstrated that these P84/Ryton combination
bags were performing similarly to the Ryton bags used during the fly ash/activated carbon
studies described above. Our goal during this study was to document pulse frequency
values for a range of air-to-cloth ratios (7.7 to 9.5 ft/min), average tubesheet pressure
drop values (3.5 to 5.4 in. H2O), and over a range of inlet mass loadings (0.002 to 0.2
Ib/MMbtu). We then compared these data to those obtained from the full-scale Plant
Gaston COHPAC I facility.
Figure 3 summarizes data collected during testing at an average tubesheet pressure drop
of 4.3 in. H2O, while Figure 4 summarizes data collected during testing at an average
tubesheet pressure drop of 5.4 in. H20. The total inlet mass loading was calculated by
adding the normal inlet loading for the test period (predicted from Unit 2 outlet opacity
data) to the inlet loading calculated from the ash injection rate for that test. The lines
represent linear regression fits to the data. Each data set is assumed to pass through the
origin. As mentioned above, all tests represented in Figures 3 and 4 were conducted with
average tubesheet pressure drop values maintained at 4.3 or 5.4 in. HjO, respectively. As
expected, the pulse frequency increases for higher air-to-cloth values for the same inlet
mass loading. Figure 5 shows data obtained from Plant Miller at an average tubesheet
pressure drop value of 3.5 in. H2O. This value is approximately the same as that being
experienced at Plant Gaston. Plant Gaston data are shown in this figure for two average
A/C values, 7.6 and 8.3 ft/min. The data match well with the test results from the pilot
unit at Plant Miller, lying between the Plant Miller data sets at A/C values of 8.7 and 7.6
ft/min..
Conclusions
Tests were performed at the Plant Miller COHPAC I pilot plant to evaluate pulse-jet
fabric filter response to higher than normal inlet fly ash concentrations and to injection of
activated carbon. These are important issues for COHPAC I performance because of the
high air-to-cloth ranges in which they operate. These data should help COHPAC I users
better understand how their system might respond to upsets in their primary particulate
control device, or if they eventually use activated carbon for the control of mercury
emissions. Fly ash test results demonstrate the usefulness of pilot-scale systems for
prediction of performance of full-scale COHPAC I fabric filtration systems.
Performance tests with fly ash and activated carbon provided the following results.
At the typical air-to-cloth values used in COHPAC I PJFFs, average pulse frequency is
very sensitive to the concentration and properties of the material being filtered.
-------
For fly ash alone, performance is sensitive to even slight changes in inlet loading to the
baghouse.
At typical COHPAC I fly ash inlet mass concentrations (0.001 to 0.01 Ib/MMBtu)
performance (pulse frequency) is much more sensitive to injection of activated carbon
than fly ash.
Co-injection of fly ash with activated carbon reduces the sensitivity of pulse-jet
performance from the activated carbon.
There were no long-term residual affects caused by the injection of fly ash or activated
carbon. The pulse-jet baghouse was able to "clean itself up" following periods of
injection of fly ash or activated carbon, returning to previous low average pulse
frequency values.
Performance tests at higher than normal concentrations of fly ash provided the following
results.
Pulse frequency response was as expected, higher pulse frequency at higher air-to-
cloth ratios for the same inlet mass concentration; higher pulse frequency at higher
inlet mass concentrations for the same air-to-cloth ratio.
For air-to-cloth ratios in excess of 9 ft/min, pulse frequency is very sensitive to inlet
mass loading changes.
Pulse jet performance recovery is very good following periods of high inlet mass
concentration.
Full-scale data from Plant Gaston COFtPAC I baghouse matches very well with the
data generated at pilot scale at Plant Miller.
Data developed at pilot scale for COECPAC I applications can be very useful and
reliable in predicting performance at full-scale.
Acknowledgments
The authors wish to thank the staff of Alabama Power Company's Plant Miller for their
support and assistance in the timely completion of the work described herein. We would
specifically like to thank Mr. Jack Walker, Superintendent, Mr. Fabert Davis, and Mr.
Randy Alexander. We also acknowledge the help of several Southern Research staff
members: Mr. David Smith, Mr. Bill Page, and Mr. Sammy O'Neal.
References
1. R. L. Chang, "COFEPAC compacts emission equipment into smaller, denser unit."
Power Engineering, July 1996, pp. 22-25.
2) R. L. Miller, R. L. Chang, A. D. Casey, G. A. Anderson, "First Commercial Use of
COHPAC Technology at a United States Power Plant, " Power-Gen 93, November 17-19
1993, Dallas, TX.
-------
3) C. Jean Bustard et al, "Experience with 145 MW COHPAC Demonstration,"
EPRI/DOE International Conference on Managing Hazardous and Paniculate Air
Pollutants, August 15-17, 1995, Toronto, Canada.
4) Richard L. Miller, "Combining ESP's and Fabric Filters for Particulate Collection,"
ICAC Clean Air Technology News, Winter 1994.
5) R. Miller, W. Harrison, K. Gushing, R. Chang, "Recent COmpact Hybrid PArticulate
Collector (COHPAC) Data For Fine Particulate Matter And Air Toxics Removal From
Coal-Fired Power Plants," ICAC Forum '96: Living with Air Toxics and NOX Emissions
Control, March 19-20, 1996, Baltimore, MD.
-------
Miller COHPAC I Performance with Injected Ash & Carbon
34567
Test Period (August & September 1996)
Miller COHPAC I Performance with Injected Ash & Carbon
34567
Test Period (August & September 1996)
Figure 1. Changes in Miller COHPAC I performance during injection of activated carbon with and
without co-injection of fly ash.
-------
Miller COHPAC I Performance with Injected Carbon
0 Drag, x 0.1 in. H2O/ftVmin
H Pulse Frequency, p/b/h
5 6
Test Period (September 1996)
Miller COHPAC I Performance with Injected Carbon
B Injected Carbon
Q Normal Fly Ash
4567
Test Period (September 1996)
Figure 2. Changes in Miller COHPAC I performance during injection of activated carbon.
-------
Miller COHPAC I Performance During Ash Injection
A/C = 7.7 ft/min, dP = 4.3 in. H2O
A/C = 8.8 ft/min, dP = 4.3 in. H2O
A A/C = 9.5 ft/min, dP = 4.3 in. H2O
A/C = 9.8 ft/min, dP = 4.3 in. H2O
0.02 0.04 0.06 0.08 0.1 0.12
Average Inlet Mass Loading, MMBtu
0.14
0.16
Figure 3. Miller COHPAC I performance response during articficial increases in inlet mass loading at an
average tubesheet pressure drop of 4.3 in. H2O.
-------
Miller COHPAC I Performance During Ash injection
1.8
1.6
A/C = 7.6 ft/min, dP = 5.4 in. H2O
A/C = 9.2 ft/min, dP = 5.4 in. H2O
0.02 0.04 0.06 0.08 0.
Average Inlet Mass Loading
1 0.12 0.14 0.16
, MMBtu
Figure 4. Miller COHPAC I performance response during artificial increases in inlet mass loading at an
average tubesheet pressure drop of 5.4 in. H2O.
-------
2.5
Miller and Gaston COHPAC I Performance Data
o Miller, A/C = 8.7 ft/min, dP = 3.5 in. H2O
Gaston, A/C = 8.3 ft/min, dP = 3.6 in. H2O
A Gaston, A/C = 7.6 ft/min, dP = 3.2 in. H2O
n Miller, A/c = 7.6 ft/min, dP = 3.5 in. H20
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Average Inlet Mass Loading, MMBtu
Figure 5. Comparison of COHPAC I performance at pilot scale (Plant Miller) and full scale (Plant Gaston) at
similar air-to-cloth ratios and inlet mass loadings for an average tubesheet pressure drop of 3.5 in. H2O.
-------
OPERATION AND PERFORMANCE
OF COHPAC AT TU ELECTRIC'S BIG BROWN STATION
B. J. Browning
A. D. Casey
A. K. Hindocha
TU Electric
1601 Bryan St.
Dallas, TX 75201-3411
C. J. Bustard
ADA Environmental Solutions
793 IS. Broadway, #349
Littleton, CO 80122
W. T. Grubb
Grubb Filtration Testing Services
P.O. Box 1156
DelranN.J. 08075
Executive Summary
COHPAC (Compact Hybrid Particulate Collector), an EPRI patented technology, is in operation
at TU Electric's Big Brown Station. COHPAC was chosen as a retrofit technology to allow Big
Brown to meet more stringent 20%, six minute average site specific opacity regulations that
became effective in 1996. Big Brown has two 575 MW, supercritical boilers that fire a locally
mined lignite coal. The original pollution control equipment consisted of 156 SCA electrostatic
precipitators, latter enhanced with a flue gas conditioning system; downstream of which
COHPAC has been installed.
Each COHPAC unit consists of four individual pulse-jet baghouses that are designed to operate
at a maximum (net-net) air-to-cloth ratio of 15.5 ft/min with off-line cleaning. After over a year
of operation performance is below expectations. Although it is believed that operating
conditions have not exceeded the design basis of COHPAC, differential pressure across the
baghouse and bag life have not met expectations. Bag failures have resulted in higher than
anticipated stack opacity.
Mechanisms believed to be contributing to bag failure include flue gas conditions and cleaning
frequency. High pressure drop is a function of the air-to-cloth ratio, inlet grain loading and
residual dustcake conditions. Both temporary and permanent measures to alleviate performance
problems are being evaluated and several have been implemented.
-------
Introduction
In 1996 the two 575 MWe generating units at Big Brown were required to meet a more stringent
six-minute average 20% opacity standard. The particulate control technology chosen to achieve
this new opacity compliance was a Compact Hybrid Particulate Collector (COHPAC), a
technology patented by the Electric Power Research Institute. COHPAC is a pulse-jet baghouse
installed downstream of an existing electrostatic precipitator. It is designed to operate at air-to-
cloth ratios higher than conventional boiler pulse-jet installations because of relatively low inlet
mass concentrations.
Although COHPAC was a new technology that was still being evaluated at the time that a
technical decision was required in 1992, this approach was chosen because pilot and full-scale
evaluations showed promising performance results, and economic studies showed significant
cost savings over other viable technologies. The pilot tests had showed that COHPAC could
operate at air-to-cloth ratios greater than 12 ft/min while maintaining pressure losses less than 6
inches water across the bags. Average emission levels were <0.01 Ib/MMBtu.
Given the success of the pilot, TU Electric and the Electric Power Research Institute collaborated
on a 145-MW full-scale demonstration unit treating about 735,000 acfm of flue gas from Big
Brown Unit 2. The demonstration unit was designed and supplied by Research-Cottrell, using
the same low-pressure/high-volume pulse-jet cleaning technology used in the pilot study. This
unit went into operation in May 1992. The full scale unit performance goals were met in terms
of opacity limits, pressure drop across the baghouse, and overall cost; however, the major
economic goal of 2 year baglife was not met. There was also a period of high pressure drop
caused by a heavy dustcake buildup from moisture condensation during numerous
shutdowns/startups.
Recommendations to solve baglife problems included a reduction in flue gas temperature, a
decrease in cleaning frequency through optimization, redesign of cleaning arms to more evenly
distribute pulse energy from inner to outer rows, and to further evaluate a novel perforated cage
design. To prevent a heavy dustcake buildup, recommendations included singeing the filtering
surface of the fabric and implementing shutdown procedures that cleaned the bags and purged
the compartments of flue gas. Based on results and recommendations, but prior to the
completion of the demonstration, a commitment to add seven (7) more COHPAC baghouses was
made. The 8 COHPAC baghouses, 4 on each boiler, treat all of the flue gas leaving Unit 1 and 2
boilers.
Startup of Unit 2 COHPAC occurred on November 6, 1995. On this unit, 3 new baghouses were
installed to complement the 145-MW demonstration baghouse. New bags were installed in all
baghouses; however, existing cages were used in the demonstration baghouse. The 4 baghouse
on Unit 1 COHPAC were put into service April 28, 1996.
The primary performance goals in the first year were to optimize the COHPAC cleaning cycles
to increase baglife and to develop operating procedures to prevent the formation of a heavy
dustcake. Initial performance was very encouraging, with opacity less than 3% and pressure
-------
losses within the design basis. However after 10-11 months of operation on Unit 2 COHPAC, a
significant percentage of the bags failed in baghouses 2-1 and 2-2. At the time these bags were
replaced in March 1997, virtually all bags had failed. In baghouses 2-3 and 2-4 the failures
occurred at a slower rate with 12% of the bags failed after 19 months. After one year, Unit 1
COHPAC also experienced a significant number of failed bags in baghouses 1-1 and 1-2.
Pressure drop performance on both units became marginal at high load boiler operation within
the first 6 months. Residual drag of the fabric stabilized after 3500 service hours at a level
higher than anticipated, resulting in periods when high pressure drop across the baghouse
prevented the boiler from meeting load demand. There has been no incidence of heavy dustcake
formation.
Many factors contribute to the less than predicted performance of COHPAC at Big Brown,
including the learning curve associated with new technologies. This paper provides an overview
of the original design conditions, early performance results, and strategies being developed to
improve performance.
COHPAC Design
The boilers at Big Brown are fired using locally mined low sulfur lignite coal with a typical ash
content ranging between 10% and 20% by weight.
Figure 1 presents a layout of the electrostatic precipitators and COHPAC baghouses for one unit
at Big Brown. Units 1 and 2 are identical, except that the original demonstration baghouse, 2-1,
was left in place. Flue gas exiting each boiler travels either through the A-side or B-side air pre-
heaters and pollution control equipment.
The electrostatic precipitators are a Research-Cottrell design that were installed with the boilers
in 1971. The ESP's consist of a chevron arrangement of 4 sections each with one inlet and one
outlet field. They have an SCA of 156 ft2/kacfm and are undersized for the boiler flows that they
receive. The ESP's should remove about 98% of the flyash with the baghouse removing the
remaining 2%.
Each boiler unit at Big Brown consists of 4 baghouses, each baghouse containing 8
compartments, and each compartment containing 312 bags for a total 9,984 bags. The bags are
arranged in 11 concentric rows in each compartment. Each bag is oval in shape, approximately
3"x6", and 20' in length. The material is felted RytonR (Ryton) fiber that is singed on both sides
prior to tubing into bags. The design net-net air-to-cloth ratio is 15.5 ft/min, with two
compartments per baghouse out of service, one for cleaning and one for maintenance. The
average pressure loss flange-to-flange is designed to be not greater than 8.5 inches H2O.
The inlet dust loading design is 1.0 Ibs/MMBtu (0.42 gr/dscf) with an allowable outlet loading of
0.03 Ibs/MMBtu.
-------
ADC 03/14/97
Figure 1
Schematic of COHPAC Installation on Big Brown Unit 2
The maximum allowable inlet temperature to the baghouses is 385ฐF continuous, which is above
the traditional 375ฐF upper limit for Ryton fiber. As part of the baghouse design to compensate
for high boiler exit gas temperatures, an atmospheric damper was added to each compartment
inlet plenum that opens whenever the inlet gas temperature reaches 385ฐF. This allows ambient
air to be drawn in and mixed with the hot gas to cool it prior to reaching the bags.
-------
This COHPAC baghouse was designed with off-line cleaning because of the combination of high
air-to-cloth ratio and high compartment can velocities (upward velocity in the compartment).
After initial startup was completed, the baghouse cleaning logic was set for continuous, off-line
cleaning (one compartment always off-line). The off-line time duration for each compartment to
clean was nominally 115 seconds. The maximum cleaning frequency at this setting was
approximately 4 pulses per bag per hour.
Prior to introducing flue gas to the baghouses, the bags were precoated with an alumina/silica
material. Flue gas was introduced at a relatively low flow (air-to-cloth ratio = 8 ft/min). The
first clean was initiated when tubesheet pressure reached nominally 3.0 inches H2O. Cleans were
initiated manually for the first several days of operation to allow a dustcake to form and
minimize ash penetration into the fabric during the first cleans.
COHPAC Performance
The primary success of COHPAC can most directly be measured by the ability of Big Brown to
follow load demand while maintaining opacity compliance. COHPAC did not meet full
expectations during the first year, as there were periods when the boiler was derated because of
high pressure drop or high opacity. This inability to meet load demand was caused by many
interrelated factors. This section discusses fundamental performance parameters and their impact
on COHPAC operation.
The primary goals defined for the Big Brown COHPAC installation included:
1. Stack opacity should be maintained below a six-minute average 20% value;
2. Pressure losses across the baghouse should be less than 8.5 inches H2O;
3. Filter bags should have a minimum life span of 2 years; and
4. Life cycle evaluation of combined capital and operating costs should be lower than any other
particulate control alternative available.
Pressure Drop/Drag
Baghouse pressure drop performance can be discussed in terms of flange-to-flange pressure drop,
tubesheet pressure drop, or drag. Flange-to-flange pressure drop is measured from the inlet to
outlet flange of the baghouse and includes pressure loss across the tubesheet and losses attributed
to the ductwork. Tubesheet pressure drop is measured in each compartment between the dirty
and clean side of the tubesheet. This parameter most directly measures pressure drop across the
accumulated and residual dust deposited on the bags. Drag provides a means to compare
tubesheet pressure drop at different air-to-cloth ratios. Drag is defined as tubesheet pressure drop
divided by the air-to-cloth ratio. At Big Brown, drag is calculated for each baghouse by
averaging tubesheet pressure drop of compartments in service and dividing by the air-to-cloth
ratio for that baghouse.
-------
Results from the 145 MW demonstration showed that flange-to-flange pressure drop across the
baghouse could be maintained below 8.5 inches H2O at air-to-cloth ratios as high as 15.5 ft/min.
Inlet grain loading during these tests varied between 0.05 and 0.15 gr/dscf with a corresponding
inlet opacity range between 10 and 35%. The operating drag during the demonstration was less
than 0.45 inches H2O/ft/min, except for a period when a heavy dustcake accumulated on the bag
surface.
Based on these results, it was believed that COHPAC was conservatively designed at a
maximum air-to-cloth ratio of 15.5 ft/min with two compartments out of service (one for
cleaning and one for maintenance). In practice the highest air-to-cloth experienced at full load
with one compartment off-line has been 11.3 ft/min (if two compartments were off-line the air-
to-cloth ratio would have been 13.2 ft/min).
Similar to performance during the demonstration, both COHPAC units consistently operated at a
flange-to-flange pressure drop of less than 8.0 inches H2O at full load for the first 3000 hours
after start-up. However, shortly after this length of operation pressure drop began to increase to a
flange-to-flange pressure drop near 9.0 inches H2O at full load with inlet opacities between 30
and 35%. If inlet conditions could be maintained at these levels, then COHPAC could operate
and not restrict load. However, changes in coal quality, flue gas temperatures greater than 375ฐF,
and non-ideal electrostatic precipitator performance each have a negative effect on COHPAC
performance and are discussed below. Many of these conditions occur frequently and often
simultaneously. It is important to note that even during periods with very poor inlet flue gas
conditions, all indications are that the inlet grain loading is still well below the predicted
maximum design value of 0.42 gr/dscf.
Coal Quality: Heating value of the coal typically varies between 6000 and 7000 BTU/lb
while ash contents varies between 10 and 20%. This variation results in a change in both
flow and flyash mass loading to the electrostatic precipitator. If the precipitators maintain a
constant removal efficiency, then when grain loading increases to the precipitators, grain
loading also increases to COHPAC, resulting in an increase in pressure drop.
Flue Gas Temperatures: To maintain flue gas temperature within the recommended
operating range for Ryton, air dilution dampers were incorporated into the baghouse design
to cool the gas whenever inlet gas temperatures reached 385ฐF. These dilution dampers
allow ambient air at nominally 10% of the total volume to enter the compartment just
upstream of the inlet plenum. This dilution air cools the gas by 25ฐF when ambient
temperatures are near 100ฐF. Unfortunately, the added volume of dilution air also adds about
1.0 inch H2O of additional pressure drop across the baghouse.
Electrostatic Precipitator Performance: The electrostatic precipitators at Big Brown are
early 1970s vintage two-field, weighted wire, chevron design precipitators. The design SCA
is 156 ft2/kacfm. At full load, flow is higher than the ESP design basis and the actual SCA is
closer to 120 ft /kacfin. Because of the small collection area of these precipitators, any upset
condition can severely impact performance. Since startup of COHPAC, the precipitators
have been maintained in good mechanical condition. Upset conditions experienced to date
-------
include: failed catalyst in the flue gas conditioning system, inlet temperatures above 400ฐF
which results in increased flow and decreased effectiveness of SO3 flue gas conditioning, loss
of one electrical field, and excessive rapping. If an electrical field is out of service for an
extended period or if two fields fail, COHPAC is bypassed until the precipitator is repaired.
Design conditions established for COHPAC were sufficiently conservative that we believe
less than optimum ESP performance has not resulted in COHPAC inlet dust loading greater
than design.
Drag Discussion: The inability of COHPAC to operate at less than 8.5 inches H2O at
maximum design conditions can be traced to both the residual and dynamic components of drag.
Residual fabric drag also has two components: 1) the contribution to drag from dustcake that
remains attached to the fabric and 2) the contribution from dust that has penetrated into the
fabric. Demonstration results indicated that if drag from the attached dustcake could be kept at a
minimum with aggressive cleaning, than residual drag would stay at a level less that 0.4 inches
H2O/ft/min. At this drag value and an air-to-cloth ratio of 11.3 ft/min, the corresponding
tubesheet pressure drop would be 4.5 inches H2O and the flange-to-flange pressure drop would
be approximately 6.5 inches H2O (allowing 2 inches H2O for duct losses at full load).
Unfortunately, measurements show that residual fabric drag stabilizes at higher value of
nominally 0.52 inches H2O/ft/min between 3500 and 4500 hours of operation. The major portion
of this residual drag is due to dust that has penetrated into the fabric. At this level and an air-to-
cloth ratio of 11.3 ft/min, flange-to-flange pressure drop from the residual drag is predicted to be
7.8 inches H2O.
The impact of the dynamic portion of drag as a result of varying dust loading at COHPAC air-to-
cloth ratios may not have been fully appreciated in the design phase. TU Electric believed that a
design air-to-cloth ratio of nominally 12 ft/min with all compartments in service was
conservative; however, this design value may be excessive for the range of grain loadings
possible at Big Brown.
The equation most often used to estimate both residual and dynamic pressure drop across a filter
bags is a derivation of Darcy's equation is shown below:
APd = K.V+ K2CtV2 /7000
where
APd = differential pressure across baghouse tubesheet (inches H2O);
Kf = residual fabric dust and dustcake weight coefficient (inches H2O-ft-min/lb)
V = air-to-cloth ratio (ft/min)
K2 = specific dustcake resistance coefficient (inches H2O-ft-min/lb)
C = dust loading (grains/acf)
t = filtration time between bag cleaning (min)
This equation shows that the dynamic portion of pressure drop increases as the square of the air-
to-cloth ratio and is linear to the inlet grain loading. For example, using a K2 of 80 , an air-to-
cloth ratio of 11.3 ft/min, a time between cleans of 15 minutes, and an inlet grain loading of 0.15
-------
gr/acf, the predicted increase in pressure drop between cleans is 3.2 inches H2O. This does not
include the contribution from the residual pressure drop at these conditions of nearly 8.0 inches
H2O. A COHPAC design at a lower air-to-cloth ratio would have significantly lessened the
impact of the residual pressure drop and the rapid increase in dynamic pressure drop from the
large range of inlet grain loadings encountered at Big Brown.
Drag Measurements. Residual pressure drop of the bags is tracked at Big Brown by manual
measurements made periodically. During the 145 MW demonstration, the Electric Power
Research Institute funded the design and fabrication of a drag measuring device. This simple
device consists of a portable blower, throttling valve, venturi flowmeter, and pressure
manometer. During an outage, the drag of hundreds of bags can be measured in one or two days.
This device has been essential in identifying the cause of higher than anticipated pressure drop
and in evaluating test bags and cages that are being considered as alternatives to those currently
in use.
Tables 1 and 2 present a summary of drag measurements made on bags in compartment A of
each of the eight baghouses. These drag values are the average of 22 measurements on bags
equally distributed from inner to outer circular row in a compartment. In Unit 2, average residual
drag for all the compartments was nominally 0.52 inches H2O after 3800 hours. In Unit 1 this
drag value was reached between 2160 and 4658 hours. Baghouses 2-3 and 2-4 have shown an
increasing trend in drag throughout their operation and the most recent measurement showed
residual drag to be greater than 0.6 inches H2O/ft/min.
The importance of residual drag cannot be overemphasized. If this seasoned value had been
known prior to designing COHPAC at Big Brown, options to increase filtering surface area
would have been considered. The reason this value was not confirmed during the demonstration
was that bag failures, equipment failures and operating problems limited the maximum run time
of each of the four test phases to less than 3500 hours and residual drag was still increasing.
Opacity
Acceptance tests performed less than 5 months after startup of Unit 2 COHPAC showed that
outlet emissions were less than 0.005 gr/acf. Average stack opacity at this time was 3% and no
visible emissions could be seen. Unit 1 had similar performance. These results were well within
performance goals and guarantees.
Big Brown must maintain less than 20% opacity on a 6-minute basis. Because of operational
problems related to high pressure drop, an unanticipated mode of operation was implemented
that increased stack opacity. This mode is referred to as "partial bypass'' and is used to prevent
having to derate the boiler when pressure drop is close to the maximum value. Partial bypass can
be loosely defined as one bypass damper open on 2 of the 4 baghouses of a particular unit. Each
baghouse has 4 bypass dampers for a total of 16 per unit. With 2 dampers open, nominally one-
eighth of the flow is bypassed. This lowers pressure drop significantly while simultaneously
-------
increasing opacity. However, there were times when maximum load capacity was still limited
during partial bypass because of opacity.
Table 1
Unit 1 COHPAC Drag
Compartment August 7-8,1996 Decembers, 1996
(2160 hours) (4658 hours)
March 10-11, 1997
(6170 hours)
1-1-A
1-2-A
1-3-A
1-4-A
NA
NA
0.30
0.20
0.54
0.45
0.54
0.46
0.57
0.54
0.35
0.48
Compartment April 23, 1996
(3812 hours)
Table 2
Unit 2 COHPAC Drag
August 6-8, 1996 November 6-7,1996 March 10-11,1997
(5616 hours) (7616 hours) (10,100 hours)
2-1-H
2-2-A
2-3-A
2-4-A
0.64
0.55
0.41
0.47
NA
0.48
0.45
0.53
0.53
0.50
0.48
0.57
Ryton failed
Ryton failed
0.59
0.74
Partial bypass provided some relief during periods when system conditions taxed COHPAC
performance. Unfortunately, the ability to use partial bypass decreased once the bags began to
fail and baseline opacity increased. Each baghouse has an individual outlet opacity monitor.
When more than 50% of the bags are failed, outlet opacity can be as high as 20%. If two
baghouses have a significant number of bags failed, load can be limited by stack opacity.
-------
Bag Life
Bag failures at Big Brown are occurring due to the combination of the following:
Loss of Strength
Embrittlement
Number of pulse cleaning cycles
Cleaning Energy
Although two year baglife was not achieved in the demonstration tests, it was believed that bag
life could be extended by lowering flue gas temperatures, optimizing cleaning logic to reduce the
number of pulses, using modified cleaning arm designs that reduced cleaning forces on the outer
rows, and using alternative cage designs.
COHPAC equipment problems delayed the implementation of a cleaning optimization process
on Unit 2 COHPAC. By the time mechanical problems were solved, pressure drop was already
unacceptably high and partial bypass operation had begun. Unit 1 start-up had far fewer
equipment failures problems and an optimum cleaning logic was used from the outset. In both
cases significant bag failures occurred after one year of operation, showing that so far cleaning
optimization has not improved bag life.
One phenomenon that is puzzling and under investigation is that bag failures occur earlier on the
A-side of each boiler than the B-side. Baghouses 1-1, 1-2, 2-1, and 2-2 are all on the A-side of
the boilers. Nearly 50% of the bags had failed in baghouse 2-1 by November 1996 while
baghouse 2-3 and 2-4 had less than 4% failures after the same period of operation. All bags were
replaced in baghouses 2-1 and 2-2 in April 1997. In anticipation of bag failures in 2-3 and 2-4,
new bags were ordered so that replacement bags would be in stock before summer operation
began. Preliminary strength evaluations on the Unit 1 bags showed that they were less damaged
than the 2-1 and 2-2 bags for the same number of service hours, and the possibility that these
bags would last through the summer; therefore no new bags were ordered. This was not the case,
however, as massive bag failures occurred in baghouses 1-1 and 1-2 in May 1997, exactly one
year after start-up. New bags intended for baghouses 2-3 and 2-4 were installed into baghouses
1-1 and 1-2 beginning in June 1997. An additional 2 baghouses worth of bags had to be ordered
in June 1997 for baghouses 2-3 and 2-4.
The original bag failure mechanism seen in baghouses 2-1 and 2-2 was from cage wire-to-fabric
abrasion. The failures began as slits on the small radius of the oval bag about 2 ft below the bag
top. With time these slits often became 6 to 8 ft in length and eventually the bag became tattered
from repeated pulsing.
In addition to a loss of strength of the fabric, there has been a problem with shrinkage (or loss of
elasticity) of the Ryton bags that occurs after one year of service. This phenomenon has forced
the cages to rise off the tubesheet and interfere with rotation of the cleaning arm. When this
occurs, the compartment must be isolated and the problem bags replaced or other corrective
-------
actions taken. This problem is labor intensive, requires bag replacement even if the bags have
not physically failed, and causes performance problems as pressure drop increases with isolated
compartments.
A new failure mechanism occurred in the Unit 1 bags in May 1997. In this case the bags failed
at the bottom disc. This failure was caused by the cage pushing through the fabric as the bags
shrunk, in contrast to the bag raising the cage off of the tubesheet. It is speculated that the fabric
strength was so weak that pressure from the cage weight ripped the bottom disk. This failure was
even more devastating hi terms of outlet opacity than the slits because once the disc was ripped
the bag bottoms moved up to several inches above the cage bottom allowing dirty gas to pass
directly through the bottom of the bag.
Because of the unusual fabric shrinkage, an intensive effort to analyze the cause of shrinkage and
strength loss was begun. Preliminary results suggest that both flue gas constituents and
temperatures may be contributing factors. Ryton is a polymer that becomes brittle when a
phenomenon called cross-linking occurs. Cross-linking can be caused by exposure to high
temperature, oxidation or chemical attack. It appears that a combination of all three mechanisms
could be occurring at Big Brown.
Operation and Maintenance Costs
Economic analysis showed that COHPAC provided a 25% cost savings over other options,
which were a new and larger precipitator or a conventional pulse-jet baghouse. This analysis
assumed 100% Ryton felt bags and a 2 year bag life, with a potential bag life of 3 years. With 1
year baglife COHPAC still has a substantially better value than the other options.
Strategies to Improve Performance
Cleaning Logic Optimization
Cleaning logic at startup was programmed to use a drag setpoint to control cleaning frequency.
The maximum frequency was 4 pulses/bag/hour at a pressure near 12 psig. Cleaning frequency
increased or decreased with respect to the drag setpoint.
After the majority of mechanical problems were fixed, cleaning logic was optimized in August
1996. Efforts were successful in that both pressure drop and cleaning frequency were decreased.
Cleaning frequency was reduced by varying the settling time as a function of the drag setpoint.
Settling time occurs after the bags are pulsed and the compartment is still isolated. This time
allows suspended flyash a period to settle prior to reintroducing flue gas. The settling time was
modified from a constant 25 seconds to be a variable with a range between 50 and 300 seconds.
At the longest settling time (usually at low load) a compartment will clean once every 45
-------
minutes, at the shortest once every 16.3 minutes. Similar to the startup logic, there was always
one compartment off-line for cleaning.
Electrostatic Precipitator Performance
Several independent experts have inspected the electrostatic precipitators and have concluded
that they are performing as expected. Since COHPAC is so far unable to meet design criteria, it
is important that the precipitators and the flue gas conditioning system are kept in peak working
condition. Big Brown has committed to increased monitoring and maintenance of the
precipitator and ancillary systems.
Evaluation of New Cleaning Arm
Results from the demonstration tests showed that the majority of bag failures occurred in
the outer rows and that the heaviest dustcakes were found in the inner rows. This correlates to
the sizing of the pulse nozzles on the cleaning arm, where the inner nozzles have a much smaller
cross-sectional open area than the outer nozzles. The pulse energy imparted to each radial row of
bags was measured with accelerometers. These tests confirmed that the outer rows were
receiving cleaning pulses that were much higher in energy than the inner rows. Based on these
results, Research Cottrell redesigned their existing cleaning arm design for better pulse energy
distribution from inner to outer row. A modified cleaning arm was installed in compartment 2-
1B for testing purposes prior to initial startup in November 1995.
To measure the effectiveness of the modified arm, manual measurements of drag and bag
inspections were scheduled periodically throughout the first year. The first two inspections did
not provide any data because the cleaning arm had failed because of poor welds. The first
opportunity to evaluate the new cleaning arm was in November. At this time over 40% of the
bags in this compartment were failed. The failures were mapped and compared to the pattern in
the other compartments in the same baghouse. In this compartment the failures were more
uniformly distributed from inner to outer rows, where in the other compartments the majority of
failures were in the outer 4 rows.
The conclusion from this test is that the modified cleaning arm had better distribution of
pulse cleaning forces than the original design. However, the modified cleaning arm did not
reduce premature bag failure.
Evaluation of Alternative Cage Design
Alternative cage designs were evaluated in the demonstration baghouse. The most successful
alternative in reducing bag wear at the lowest incremental cost was to use the existing bottom
half of the cages and replace the top half with cages made from perforated plate. The perforated
plate that had the best performance and was easiest to handle had 33% open area. In theory the
-------
perforated cage has two advantages: 1) the smooth surface eliminates wire-to-bag wear, and 2)
the perforated cage helps to better distribute the cleaning pulse along the length of the bag
resulting in lower accelerations at the top and increased accelerations in the lower half.
Perforated cages were installed on a row of new bags and a row of used bags in compartment 2-
3A in November 1996. For baseline comparison, a row of new bags were installed on original
wire cages and a row. Preliminary data look encouraging; however, only 3000 hours of
operation were accumulated on the new bags at the time of the inspection.
Evaluation of Alternate Fabrics
Alternative fabrics have been evaluated in both the demonstration and the permanent installation.
The fabrics that are still under consideration are Tefaire (Tefaire), alternate PPS, Ryton with
Rastex and PTFE scrim, Ryton with a chemical resistant coating, and a novel star-shaped Ryton
bag supplied by Albany International.
Tefaire Bags. Tefaire bags have been tested in several phases. The first Tefaire bags were
installed in Phase I of the demonstration. Three of these bags are still in service and have
accumulated over 25,000 hours of operation. Two other Tefaire bags have been in service for
18,000 hours. These bags show no loss of strength due to flue gas exposure, no unusual wire
abrasion and no shrinkage. The Phase I bags, however, were very loose on undersized cages for
the first 13,000 hours of operation and are finally showing signs of wear. The measured drag of
these bags is comparable to the Ryton bags in the same compartment.
Tefaire bags and 39 Ryton bags were installed in 2-1A for start-up in November 1995. There
have been no Tefaire failures, no loss of fabric strength, and no shrinkage after over 11,000
service hours. All Ryton bags in this compartment have failed. A commitment to purchase
Tefaire bags has not been made because their cost is nearly 3.5 times greater than Ryton and
there is some concern that the relatively fuzzy surface of the these bags may be conducive to
formation of a heavy dustcake, under certain upset conditions. To be cost competitive with a 2
year Ryton bag, a 7 year baglife would be necessary with the Tefaire bags.
Ryton with PTFE Scrim. Ryton felt with PTFE scrim can provide greater durability than a
100% Ryton felt because the PTFE scrim is usually unaffected by flue gas exposure and can
extend the life of the bag even when the Ryton fiber is highly degraded. Sample bags of Ryton
with PTFE scrim were evaluated in Phase I of the demonstration. Although the Ryton batt of
these bags failed after the same length of service as the 100% Ryton bags, the PTFE scrim
remained entirely intact. Premature bag failures in Phase I were caused by flex fatigue of the
fabric, which occurred because the bags were much larger than the cages. There is optimism that
longer bag life will be achieved with test bags to be installed in July 1997, since the bags will be
sized to fit the cages properly. These bags are approximately 1.5 to 2 times the cost of the
standard Ryton bags, depending on the weight of the composite fabric and PTFE scrim used.
-------
Chemical Resistant Coatings. Recent fabric analyses indicate that chemical attack may be
accelerating loss of strength of the Ryton fabric. Although chemical resistant finishes are not
typically used with Ryton fabrics, such finishes are being considered and if appropriate will be
installed for evaluation.
Star Bags. Albany International has recently introduced a star-shaped Ryton bag that
increases the filtering area of each bag. This bag is supported by a custom star-shaped cage.
This bag offers the advantage of reducing the air-to-cloth ratio by increasing the effective cloth
area.
Temporary Measures
Because of the lack of a definitive explanation for the "shrinking" bags, and the necessity to
order replacement bags, bag length has been increased. Bags ordered as replacements for
baghouses 2-1, 2-2, 1-1 and 1-2 were made an additional 2 inches longer to allow for as much as
4 inches of shrinkage in length. The bottom reinforcement of these bags was increased to 7
inches from 4 inches to ensure that it would enclose the cage bottom over a wide range of
changes in bag length. In addition, the bottom disc of the bags was made out of Tefaire fabric to
prevent flex fatigue failure, which has been experienced with Ryton discs when the bag is more
than 1.5 inches longer that the installed cage.
Conclusions
In the first year of operation the primary performance goals were to optimize COHPAC cleaning
cycles to increase baglife and to develop operating procedures to prevent a heavy dustcake from
forming. Following is a summary of the conclusions associated with both primary and secondary
goals.
1. Opacity should be maintained below a 6 minute 20% limit: Goal achieved prior to bag
failures. Prior to bag failures and high differential pressure, stack opacity was well below
the 20% limit and usually below 3%.
2. Filter bags should have a life span of 2 years: Goal not achieved. As stated, 50% of the
bags failed and required replacement after 1 year of operation. Bags were replaced in
baghouses 2-1 and 2-2 in April and in baghouses 1-1 and 1-2 in June 1997. Baghouses 1-3,
1-4, 2-3 and 2-4 also have failures but not to the degree of the other baghouses. Cleaning
optimization was implemented in August 1996 by Research-Cottrell but did not increase bag
life on baghouses 1-1 and 1-2.
It is believed that the major factors contributing to loss of fabric strength are flue gas
temperatures that often approach the allowable upper limit for Ryton felt (up to 385ฐF), flue
gas constituents and high bag cleaning frequencies
-------
Specific changes that could help prolong baglife will either have a cost or performance
impact associated with them. Examples of these options include:
Perforated cages;
Alternative bag materials; and
Lower flue gas temperatures.
3. Prevent buildup of heavy dustcake: Goal Attained. The combination of singeing the
filtering side of the fabric and implementing shutdown procedures that include pulsing the
bags and purging the compartments of flue gas have prevented the formation of the heavy
dustcake. Future candidate fabrics should also have a singed or "non-fibrous" filtering
surface and shutdown procedures should be kept in place.
4. Flange-to-flange pressure drop should be below 8.5 inches H2O: Goal not achieved.
Pressure drop was higher than expected and prevented boiler operation at maximum capacity
during the summer season. To minimize pressure drop it was necessary to operate with one
bypass damper open on 2 baghouses. Even in this mode of operation boiler capacity was
limited by opacity.
The cause of this higher-than-expected pressure drop was a combination of higher residual
dustcake drag and higher inlet grain loading than experienced during the 145 MW
demonstration test. Operation to date indicates that residual drag of both the Ryton and
Tefaire bags will stabilize around 0.53 inches H2O/ft/min. At this value it is estimated that
inlet opacity greater than 30% will result in higher than acceptable pressure drop. If high
inlet opacity does occur, it will be necessary to operate in a partial bypass mode.
5. Optimize cleaning logic: Implemented. A modified cleaning logic was implemented in
August 1996 by Research Cottrell. This logic reduced the maximum number of cleaning
pulses per hour by varying the settling time as a function of the drag setpoint. This logic
initially appeared to reduce pulse energy while still providing adequate cleaning for pressure
drop and dustcake control. Efforts to further reduce cleaning frequency and pulse pressure
are desirable; however, increasing differential pressures prevent any additional reduction in
cleaning energy at this time.
6. Evaluate modified cleaning arm: Preliminary evaluation completed. A cleaning arm
that was modified for better distribution of pulse energy was installed in compartment 2-1B.
Results showed that indeed this arm did provide greater cleaning energy to the inner rows
and reduced cleaning energy to the outer rows. This cleaning arm did not prevent or reduce
bag failures when compared to other compartments in the same baghouse.
7. Integrating COHPAC into overall system: Efforts still continuing. The first year of
operation has shown that COHPAC requires an experienced engineer to oversee system
operation and how it impacts COHPAC performance. It is important that the primary
particulate control equipment is well maintained and that COHPAC performance is
monitored closely. Big Brown has committed to this level of support, and this will improve
overall long term performance.
-------
Thursday, August 28; 3:00 p.m.
Parallel Session B:
Air Toxics Control - Mercury Capture by FGD
-------
Factors Affecting Control of Mercury by Wet FGD
O. W. Hargrove, Jr
T. R. Carey
C. F. Richardson
R. C. Skarupa
Radian International, LLC
P.O. Box 201088
Austin, TX 78720-1088
F. B. Meserole
Meserole Consulting
8719 Ridgehill Dr.
Austin, TX 78759
R. G. Rhudy
Electric Power Research Institute
P. O. Box 10412
Palo Alto, CA 94303-0813
Thomas D. Brown
Federal Energy Technology Center
Department of Energy
P.O. Box 10940
Pittsburgh, PA 15236
Abstract
Bench-scale and pilot-scale mercury control research activities are being funded by the Department
of Energy and the Electric Power Research Institute to investigate options for utilities to reduce
mercury emissions in the event that such .controls become required. This paper discusses the initial
results of this project which is designed to increase the control efficiency of mercury and other air
toxics across wet flue gas desulfurization processes.
Mercury can exist in two forms in utility flue gasas elemental mercury and as oxidized mercury
(predominant form believed to be HgCl2). Previous test results have shown that wet scrubbers
effectively remove the oxidized mercury from the gas but are ineffective in removing elemental
mercury. Recent improvements in mercury speciation techniques confirm this finding. The cur-
rent work focuses on developing processes to improve elemental mercury removal around an FGD
system. Potential catalysts and additives that can capture elemental mercury or convert it to a form
easily controlled by a wet FGD system were screened in the laboratory. Attractive additives and
catalysts were then tested on the 4MW pilot at EPRI's Environmental Control Technology Center.
Catalyst and additive performance results from both the laboratory- and pilot-scale studies are pre-
sented.
-------
Introduction
This project is funded by the U.S. Department of Energy's Federal Energy Technology Center
(DOF7FETC) under a cost-sharing PRDA with Radian International. The Electric Power Research
Institute (EPRI) is providing cofunding and technical oversight. The project is part of FETC's Ad-
vanced Power Systems Program, whose mission is to accelerate the commercialization of afford-
able, high-efficiency, low emission, coal-fueled electric generating technologies.
The 1990 Clean Air Act Amendments mandated the Environmental Protection Agency to study the
health effects caused by hazardous air pollutants (HAPs) from electric utility plants. The act also
mandated a separate study on the effect of mercury emissions. Most of the HAPs in power plants
occur in the particulate phase at flue gas exit temperatures. However, mercury, while emitted in
extremely low concentrations, is primarily in the vapor phase at most plants. The goal of this
PRDA research was to explore the development of advanced concepts for removing toxic sub-
stances from flue gases using wet flue gas desulfurization (FGD) systems.
During this program, catalytic oxidation of vapor-phase elemental mercury was investigated in
three phases. During the first phase, bench-scale tests were conducted to screen potential catalyst
and fly ash types for catalytic activity'. Based on the results from these tests, the most promising
catalysts were tested at EPRTs Environmental Control Technology Center (ECTC) in Barker, New
York, using a 4-MW pilot FGD system. Following the pilot tests, a catalyst field test unit was de-
veloped, and additional catalyst testing was conducted at a utility power plant which bums lignite
coal. Results from tests conducted at the full-scale facility were still being reduced and evaluated at
the time of this paper's preparation.
Objective
The overall objective of this project was to leam more about controlling emissions of hazardous air
pollutants (HAPs) from coal-fired power plants that are equipped with wet flue gas desulfurization
(FGD) systems. The project was included by FETC as a Phase I project in its Mega-PRDA pro-
gram. Phase I of this project focused on three research areas. These areas in order of priority were:
Catalytic oxidation of vapor-phase elemental mercury;
Enhancedparticulate-phase HAPs removal by electrostatic charging of liquid droplets; and
Enhanced mercury removal by addition of additives to FGD process liquor.
Mercury can exist in two forms in utility flue gasas elemental mercury and as oxidized mercury
(predominant form believed to be HgCl2). Previous test results have shown that wet scrubbers
effectively remove the oxidized mercury from the gas but are ineffective in removing elemental
mercury. Recent improvements in mercury speciation techniques confirm this finding.
Catalytic oxidation of vapor-phase elemental mercury is of interest in cases where a wet scrubber
exists or is planned for SO, control. If a low-cost process could be developed to oxidize all of the
elemental mercury in the flue gas, then the maximum achievable mercury removal across the exist-
ing or planned wet scrubber would increase. Other approaches for improving control of HAPs in-
cluded a method for improving particulate removal across the FGD process and the use of additives
1 Tested materials are referred to throughout this paper as "catalyst" or fly ash samples. Testing of these materials has
not been performed long enough to determine whether the material reacts directly with elemental mercury or instead
catalyzes the oxidation of elemental mercury. The term "catalyst" is used for convenience.
-------
to increase mercury solubility. This paper discusses results related only to catalytic oxidation of
elemental mercury.
Laboratory Tests
Approach
The bench-scale unit used to test the reactions of gas-phase elemental mercury with potential cata-
lysts is shown in Figure 1. The general test approach consisted of passing a simulated flue gas
containing elemental mercury across a fixed-bed reactor containing a mixture of catalyst material
and sand. The gas exiting the fixed bed was analyzed semi-continuously to determine the fraction
of inlet elemental mercury oxidized across the bed.
Mamai On/Off Valves (^ BcoromeaUy Controlled Valvo
Figure 1
Elemental mercury conversion bench-scale test apparatus.
The simulated flue gas was prepared using reagent gases and calibrated flow meters. Elemental
mercury was added to the simulated flue gas by passing nitrogen carrier gas across a mercury dif-
fusion cell which contained a Hgฐ permeation tube. The amount of diffused mercury was con-
trolled by controlling the flow of nitrogen through the diffusion cell and the temperature of the dif-
fusion cell. The mercury-containing nitrogen was then mixed with other flue gas components
(SO2, HC1, O2, CO2, and H2O) at constant temperature before the gas entered the fixed-bed reac-
tor.
The fixed-bed reactor consisted of a mixture of catalyst or fly ash material and sand placed in a
temperature-controlled, vertical Pyrex column (typically yielding a bed length of about 1.75
inches). Gas exiting the fixed-bed was analyzed to determine the percentage of inlet elemental mer-
cury that was either removed or oxidized across the bed. Oxidized forms of mercury exiting the
bed were captured in a 1M Tris buffer solution. This impinger solution has been shown in other
studies to effectively capture oxidized mercury while allowing elemental mercury to pass through
the solution. Elemental mercury passing through the Tris solution was measured semi-
continuously using a gold amalgamation unit and cold vapor atomic absorption (CVAA) unit. The
inlet Hgฐ concentration was also measured using the gold amalgamation/CVAA unit at the begin-
ning and end of each test.
-------
Ideally, the total mercury concentration (oxidized plus elemental) exiting the fixed-bed reactor
should be equal to the inlet concentration. The outlet oxidized mercury concentration was deter-
mined by analyzing the Tris buffer solutions. By comparing this concentration to the inlet mercury
concentration, the fraction of mercury oxidized across the fixed-bed could be determined. The out-
let elemental mercury concentration was determined based on the gold amalgamation/CVAA analy-
ses to provide the fraction of mercury that passed through the bed unchanged. The sum of these
two analyses was compared to the inlet concentration to determine the total fraction of inlet mercury
that was detected at the outlet. In many cases, the total outlet concentration was lower than the inlet
concentration. These differences were attributed to adsorption of mercury by the catalyst material.
During the bench-scale tests, fourteen different catalyst and eleven different fly ash samples were
tested to determine their ability to oxidize elemental mercury. Each sample was tested at the base-
line conditions shown in Table 1. In addition, several parametric tests were conducted which
evaluated the effect of temperature, SO2 concentration, HC1 concentration, and NOX concentration
on the oxidation of elemental mercury. The materials tested are shown in Table 2. During each test,
a ten gram mixture of sand and catalyst was placed in the fixed-bed reactor. Most materials were
tested using a bed loading of 100 mg/g. Lower loadings, shown in Table 2, were used for some
catalysts thought to be more reactive.
Most of the laboratory tests were conducted for only a few hours, not long enough for the adsorp-
tion capacity of the catalyst to be depleted. Therefore, elemental mercury oxidation was not meas-
ured at steady-state. These short tests were intended to be catalyst screening tests to indicate rela-
tive differences in performance between catalysts. A few long-term tests were conducted until
steady-state was obtained. Elemental mercury oxidation measured during these tests provides a
better estimate of catalyst performance.
The catalyst samples shown in Table 2 can be classified as either commercial catalysts, laboratory-
prepared catalysts, or fly ashes. Several commercial catalysts were obtained from a catalyst sup-
plier. These included samples consisting of both palladium and nickel on an alumina (A12O3) sub-
strate. Other samples included a zinc catalyst, alumina powder, a SCR catalyst, and a carbon-based
catalyst. Seven of the fourteen catalysts were iron-based. Some of these iron-based catalysts were
obtained commercially while others were generated in the laboratory.
Table 1
Baseline Gas Conditions for Catalyst Screening Tests
Parameter Baseline Condition(s)
fixed-Bed Temperature 300 and 700ฐF
[Hgฐ] 45-60mg/Nm3
02 7%
CO, 12%
H26 7%
SO, 1600 ppm
HC1 50 ppm
Gas Flow Rate 1 L/min
Previous testing has shown that fly ash at the ECTC is capable of oxidizing Hgฐ in flue gas. To
determine if this property is shared by other fly ashes, several fly ash samples generated from dif-
ferent coal sources were obtained for bench-scale testing. The goal was to determine if particular
-------
fly ash sources are more effective at oxidizing mercury. Ash samples from bituminous, subbitumi-
nous, and lignite coals were obtained. An additional sample from an oil-fired system was also ob-
tained for comparison. Three different samples of Bituminous #2 ash, each collected from a differ-
ent process location, were obtained. The locations included the first and fifth fields of the ESP as
well as a cyclone connected upstream of a gas sampling train.
Table 2
Catalyst and Fly Ashes Tested in Bench-Scale Reactor
Catalyst Samples
Alumina
NiO
ZnO
Pd#l
Pd #2 (20 mg/g)
NQx Catalyst
Carbon (2 mg/g)
Fe #1(100, 20 mg/g)
Fe #2 (20 mg/g)
Fe #3 (20 mg/g)
Fe#4
Fe#5
Fe#6
Fe#7
Fly Ash Samples
Subbituminous Ash #1
Subbituminous Ash #2
Bituminous Ash #1
Bituminous Ash #2 - Cyclone
Bituminous Ash #2 - ESP field 1
Bituminous Ash #2 - ESP Reid 5
Bituminous Ash #3
Lignite Ash #1
Lignite Ash #2
Lignite Ash #3
Oil-RredAsh
Subsequent to testing at the ECTC, some additional bench-scale tests were run to determine the
effect of SO3 and HC1 on the oxidation of the catalyst samples. Results from the ECTC had indi-
cated that SO3 could be deactivating or inhibiting catalyst activity. SO3 was added to the gas by
purging a solution of fuming sulfuric acid (30% free SO3) with a nitrogen stream. The HC1 bench-
scale tests were repeated as long-term tests since the initial, short-term bench tests were inconclu-
sive.
Lab Results
Figure 2 shows results from the 300ฐF (baseline) and 700ฐF catalyst tests for the most active of the
catalyst materials. The percentages of inlet elemental mercury oxidized and adsorbed by the catalyst
samples are shown. As shown in Table 2, most of the catalysts were tested using 1000 mg of
catalyst in a 10 gram mixture of catalyst and sand. Materials tested using a smaller mass were car-
bon (20 mg) and Fe #2 and #3 (200 mg).
Catalysts with the greatest combined adsorption and oxidation in the baseline tests were the two
Pd-based catalysts, Fe #1, Fe #2, Fe #3, Fe #4, and carbon. The activity of carbon was considered
high since a small mass was used relative to the other catalysts. Generally, both oxidation and ad-
sorption across the catalysts were higher at 300ฐF than at 700ฐF. These data suggest that elemental
mercury is first adsorbed to the catalyst surface and then reacts. Since physical adsorption gener-
ally decreases as temperature increases, less of the mercury is also oxidized.
Results from the baseline tests (300ฐF) with fly ash samples are summarized in Figure 3. (The ef-
fect of temperature is similar to that with the catalyststhe fly ash activity at300ฐF is generally
greater than at 700ฐF so 700ฐF results are not shown.) Both Subbituminous ashes and Bituminous
-------
#1 showed greater oxidation than any of the lignite ashes. Lignite #1 showed greater oxidation than
the other two lignites which exhibited greater adsorption than oxidation at 300ฐF.
390T 700ฐF 300ฐF 700ฐF 300T 708ฐF SWF 700ฐF 300ฐF 700ฐF 300T 700"F 30TF 700ฐF 360ฐF
Feซ Pd#l Fe*2 Fe#3 NOx Catalyst Fc#4 Pd#2 Carbon
1000 mg 1000 mg
200 mg
200 mg
1000 mg 1000 mg 1000 mg
20 mg
Figure 2
Catalyst Activity at Baseline Conditions
Figure 3. Fly Ash Activity at Baseline Conditions
-------
Three different samples of Bituminous #2 fly ash were tested. These samples were collected from
ESP Field 1, ESP Field 5, and a cyclone pre-cutter in a sampling train. The highest oxidation ob-
served was from the Hopper E ash, the smallest-sized material. The lowest oxidation observed was
from the cyclone catch. Since all of these samples were tested at the same loading (1 gram), the
effect may be due to size differences or to surface chemistry differences.
The effects of SO2, HC1, and NOX concentration on oxidation activity were also tested during the
bench-scale tests. Variability in the data resulted in no distinct trends being identified during the
first series of tests. Because of the perceived importance of HC1 in the oxidation step and because
catalyst testing at the ECTC indicated that some of the materials were being deactivated by SO3,
additional laboratory parametric tests were conducted at the conclusion of the pilot program to de-
termine the effect of these flue gas constituents on oxidation.
Testing of the metal-based catalysts at the ECTC indicated that oxidation quickly decreased as ex-
posure to flue gas increased. These observations suggested that a flue gas component present at the
ECTC was deactivating the catalysts. This component was suspected to be SO3.
To confirm the suspected effect of SO3, SO3 was added to the matrix gas during tests with fresh Fe
Catalyst #1 samples. The sample size was reduced to 0.02 grams (in 0.98 g of sand) so that effects
could be observed over a reasonably short time period. These results are summarized in Figure 4.
After adding about 10 ppm SO3 to the gas, Hgฐ oxidation decreased from 29% to 4% and the per-
centage of Hgฐ passing through the column unchanged increased from 8% to 96%. Apparently, the
SO3 reacts with the metal oxide-based catalyst in some manner, thereby preventing mercury reac-
tions from occurring. The deactivation mechanism, however, is unknown.
100
ec
S3
90--
80--
70--
60.-
50.-
40--
30--
20--
10-.
0--
DHgฐ adsorbed
Hgฐ oxidized
with SO,
Figure 4. Effect of SO$ on Catalyst Fe #l-Lab Results
The effect of HC1 on Hgฐ oxidation with a carbon-based catalyst in the laboratory reactor is shown
in Figure 5. These results show that Hgฐ oxidation is essentially zero at HC1 concentrations below
10 ppm. It is emphasized that these results are valid for the simulated flue gas used in these tests.
Interaction with other flue gas constituents such as NOX may have important effects as well. How-
-------
ever, the fact that the HC1 concentration is 50 ppm at the ECTC may be one reason that high Hgฐ
oxidation occurs naturally there.
100 ,
90.:
o,
90
20 30 40 50 60 70 80
Flue Gas HCI Concentration (ppm)
Figure 5. Effect of HCI Concentration on Oxidation of Hgฐ
Based on the baseline test results and anticipated catalyst material costs for full-scale application,
the following samples were recommended for pilot-scale testing: Fe #1, Fe #2, Fe #4, carbon,
bituminous #2 fly ash, and lignite #3 fly ash.
Pilot Tests
Approach
Based on the bench-scale results, the most promising catalyst samples were to be tested on the 4-
MW pilot system at EPRI's ECTC. The configuration used for testing catalysts on the pilot-scale at
the ECTC is shown in Figure 6. Gas containing fly ash passed through the dirty raw gas (DRG)
duct to the pilot-scale electrostatic precipitator (ESP). After removing over 99% of the fly ash, the
gas passed through the clean raw gas (CRG) duct to the spray-dryer absorber (SDA) and pulse-jet
fabric filter (PJFF). The SDA is used only as a means to convey gas to the PJFF (i.e., the atomizer
wheel is not in service). After passing through the PJFF, the gas was returned to the Kintigh stack
through the treated gas return (TGR) duct.
Flue gas for the pilot system was drawn from New York State Electric and Gas' Kintigh Station.
The Kintigh flue gas contained about 90% oxidized mercury. In order to test the oxidation of ele-
mental mercury, the elemental mercury concentration had to be artificially increased by injecting
elemental mercury into the duct. Mercury was injected at several different locations. To determine
the concentrations of both oxidized and elemental mercury in the flue gas, a modified Method 29
-------
gas sampling train was used. In this train, the Method 29 peroxide/nitric acid impingers were re-
placed with Tris impingers (M29T). The Tris solution effectively captures oxidized forms of mer-
cury while allowing elemental mercury to pass through the solution to the permanganate impingers.
The M29T train is shown in Figure 7.
Rgure6
Simplified flow diagram for pilot-plant Hgฐ oxidation testing
Thermometer
Gloss QIQS5 Probe
Probe Tip Liner
Glass Filter
Thermometer
Pitot Tris/EDTA 5%HNO3/ 4% KMnO.,/
Manometer 10%H2O2 10%H.,SO,,
Orifice
Figure?
Schematic of Method 29-Tris Sampling Train
-------
Using the configuration shown in Figure 6, the ability of various catalyst samples to oxidize ele-
mental mercury was determined by coating the catalyst material on the PJFF bags. The bags were
pulsed clean before adding the desired mass of catalyst and then were not cleaned again until test-
ing of the given sample was complete. By injecting elemental mercury upstream of the PJFF, ele-
vated concentrations of elemental mercury contacted the catalyst material as the gas passed through
the PJFF. Mercury samples were collected upstream of mercury injection, at the PJFF inlet, and
PJFF outlet By comparing the PJFF inlet and outlet elemental mercury concentrations, the fraction
of inlet elemental mercury oxidized was determined. By collecting the prespike sample, the actual
amount of elemental mercury spiked into the duct was determined (this sample is not absolutely
necessary to determine the oxidation fraction across the PJFF).
The above method of testing catalysts was generally effective at the ECTC. However, operational
difficulties caused by unexpectedly high oxidation across the original Ryton bags and failure of
acrylic bag material (already available on site) limited catalyst testing on the pilot equipment. Dur-
ing periods when the PJFF was out of service, testing was conducted by placing catalysts or bag
materials in the filter holder location of a Method 29-Tris (M29T) sampling train to maximize data
collection.
Pilot Results
Initial testing using the configuration shown in Figure 6 indicated that the elemental mercury spike
was nearly 100% oxidized at the PJFF outlet even with no catalyst material added to the PJFF.
Previous testing at the ECTC has shown that elemental mercury oxidation in the duct is a strong
function of the residence time. Several changes were made in order to decrease the duct residence
time of the mercury spike; however, after decreasing the residence time from the spike location to
the post-PJFF sampling location from 51 seconds to 12 seconds, a significant fraction of the ele-
mental mercury spike was still oxidized.
Ryton bags were used in the PJFF for the residence time testing described above. Several tests,
including those at short residence times, indicated that these bags were responsible for oxidizing a
significant fraction of the elemental mercury (20-30%). These results indicated the need to study
the oxidation of elemental mercury across different bag materials. For these tests, different bag
materials (some new and some previously contacted with fly ash) were placed on filters in a
Method 29T gas sampling train, and the oxidation of mercury across these materials was compared
to a baseline Method 29T train sampling along side the test train. Based on the bag materials tests,
Gore-Tex on fiberglass bags resulted in very little oxidation of elemental mercury; therefore, a
set of these bags was ordered to rebag the PJFF.
To continue testing catalyst samples while the new bags were being purchased, Method 29T sam-
pling trains were used to test various catalyst samples. Catalyst samples tested on the Method 29T
sampling filters included Fe #1, Fe #2, Fe #4, Pd #1, carbon-based material, and the SCR catalyst.
The results from these tests are shown in Figure 8. A decrease in KMnO4 capture indicates the
amount of elemental mercury oxidized and/or adsorbed. An increase in Tris capture indicates the
amount of elemental mercury oxidized.
The results show that the carbon-based catalyst was the most active by a significant margin, even
though only 0.6 g of carbon was used compared to 2.5 g of the next most reactive material (Fe
#2). It is also important to note that the carbon-based material oxidized a greater fraction of the
mercury (measured as increase in Tris) as the run time was extended and adsorption decreased.
This could indicate that the carbon-based catalyst may oxidize a high percentage of Hgฐ after steady
state is reached.
10
-------
90%
D Decrease in KMnO4
Increase in Tris
Carbon
0.6 g
Carbon 0.6 g
extended contact
Figure 8
Effect of Hgฐ Oxidation "Catalysts" at 300ฐF (M29T Trains)
During the Method 29T tests, most of the catalyst materials appeared to be deactivated during con-
tinued exposure to flue gas. As exposure time to flue gas increased, Hgฐ oxidation decreased. One
of the major differences between the ECTC flue gas and the simulated laboratory flue gas was the
presence of SO3 at the ECTC. These observations led to the laboratory testing of SO3 as discussed
previously.
Once the new Gore-Tex on fiberglass bags were installed in the PJFF, catalyst CT-9 (a carbon-
based catalyst) was coated on the PJFF bags to determine its ability to oxidize mercury. The cata-
lyst was introduced to the PJFF downstream of the ESP to ensure that a minimal amount of fly ash
would be present with the catalyst on the bags. (The initial test was with 20 pounds of CT-9, al-
though subsequent tests included higher and lower catalyst loadings.) The PJFF was then operated
continuously for ten days with no pulse-cleaning, and samples were taken each day to determine
elemental mercury oxidation and removal. Although all ESP fields were placed in service and the
fly ash loading was below 0.03 Ib/MMBtu, the PJFF continued to remove fly ash. The effect was
that the oxidation of Hgฐ increased over time with just the native fine fly ash accumulation. This
effect of fly ash accumulation on oxidation is shown in Figure 9. The baseline oxidation increased
from 30-50% immediately after bag cleaning (baseline) to 50-80% after several days. This makes it
somewhat difficult to interpret the effect of catalyst materials added to the PJFF.
The effect of CT-9 was estimated by using the data in Figure 9 to "correct" for the background
oxidation due to the fine ash accumulation. The results of this analysis are shown in Figure 10.
The results indicate that 20 pounds of CT-9 accounted for about an 85 % decrease in the outlet Hgฐ
concentration after correcting for fine ash oxidation. Ten and 5 pounds of carbon accounted for
about 65 and 35% oxidation. It should be noted that average values over several days of operation
were used to generate Figure 10. With the difficulty in measuring very low mercury concentrations
and the daily variations in oxidation values, the uncertainty in Figure 10 data is quite large despite
the smooth curves shown.
11
-------
D Decrease in KMnO4
Increase in Tris
Baseline Day 1 Day 2 Day 3 Baseline Day 1 Day 2 Day 3 Day 4 Day 5
Figure 9. Effect of Fine Ash Accumulation on PJFF Bags
J5
"v
si
TJ
o
O)
J-;
100
90
80
70
60
50
40
30
20
10
0
o"-.
-Reduction in Hg(0)
- Outlet Hg(0) Cone.
-tIIII1IIII1I-
0 5 10 15 20
Amount of Carbon Added to PJFF (Ib)
20
O
c
90
10
-.5
o
n
25
Figure 10. Effect of Carbon on Hgฐ Decrease and Hgฐ Outlet Concentration
12
-------
Extended Catalyst Testing at Other Sites
While results from the ECTC indicate that a carbon-based catalyst can be used to oxidize elemental
mercury, there are several issues that limit the usefulness of further Hgฐ oxidation test work there.
First, the naturally-occurring mercury consists of 90% oxidized mercury with only about 1 (Xg/Nm3
Hgฐ in the flue gas. This limitation requires that Hgฐ be spiked into the flue gas to increase the ac-
curacy of the Hgฐ determinations. However, even with no catalyst, the injected Hgฐ is easily oxi-
dized which complicates the determination of catalyst material performance.
A second limitation is the amount of catalyst material that can be coated on the PJFF without ex-
ceeding the allowable pressure drop. Forthe carbon-based catalyst, the maximum amount was
about 30 pounds for a 5-10 day run. The pressure drop continues to increase with time of opera-
tion because most of the fine ash exiting the ESP is collected by the PJFF. This limits the maxi-
mum length of a test to about 10 days. For a catalyst to be economically attractive, it must stay ac-
tive for considerably longer than 10 days. Therefore, more useful results could be generated at dif-
ferent utility sites with higher concentrations of Hgฐ in the flue gas using a device designed to ex-
pose the catalyst to flue gas for long periods of time.
Initial short-term tests have been conducted at a power plant that burns lignite coal. While prelimi-
nary data indicate that the carbon-based catalyst oxidized the Hgฐ fairly efficiently, the data from
this test are still being interpreted. A test unit has been designed and will be built to allow extended
catalyst testing. This extended catalyst testing will be part of the proposed Phase II PRDA activi-
ties.
Acknowledgments
This research is being sponsored by the U.S. Department of Energy's Federal Energy Technology
Center (Pittsburgh) under contract DE-AC22-95PC95260. We would like to thank our key sub-
contractors ADA Technologies and Parsons Power, the operations contractor for the ECTC.
The work presented in this paper is the result of research carried out at EPRI's Environmental
Control Technology Center (ECTC) located near Barker, NY. We wish to acknowledge the sup-
port of the ECTC cosponsors: New York State Electric and Gas, Empire State Electric Energy Re-
search Corporation, Electric Power Development Corporation, and the U.S. Department of En-
ergy.
13
-------
IMPROVED MERCURY CONTROL IN WET SCRUBBING
THROUGH MODIFIED SPECIATION
C. David Livengood
Marshall H. Mendelsohn
Argonne National Laboratory
9700 South Cass Avenue
Argonne, Illinois 60439
Abstract
Integration of control functions for multiple pollutants within a single system can offer a number
of economic and operational advantages. While recent field studies have shown that typical
sulfur-dioxide scrubbers are relatively ineffective in controlling elemental mercury (Hgฐ) in flue
gas, research at Argonne National Laboratory has been focused on enhancing the capture of
mercury in such systems through conversion of Hgฐ to more soluble mercury compounds. The
results have shown that such a change in speciation may be accomplished through a combination
of various chemical agents and native flue-gas components. In some cases, nitric oxide has been
found to have a strong beneficial effect on Hgฐ conversion while being simultaneously removed,
providing indications that a combined control process for sulfur dioxide, nitric oxide, and
mercury may be feasible. Initial bench-scale experiments have investigated different
combinations of chemical agents, flue-gas components, and gas-liquid contacting, while process
development research is being initiated to investigate the mercury-control potential of scrubbing
with modified speciation under realistic conditions of flue-gas temperature, composition, and
residence time.
Introduction
The control of hazardous air pollutant (HAP) emissions was addressed in Title III of the Clean
Air Act Amendments of 1990, which provided an initial list of 189 elements and compounds of
concern. The combustion of coal has the potential to produce a number of those species, either
directly as a result of the trace elements found in coal, or as products of chemical reactions
occurring in combustion. However, field studies conducted by the U.S. Department of Energy
(DOE), the Electric Power Research Institute (EPRI), and others have shown that the actual
emissions are very low and that effective particulate-matter capture can control most of the
inorganic species. The most significant exception is mercury, which has also been singled out
-------
for particular regulatory attention because of its behavior in the environment (bioaccumulation)
and the potential for deleterious health effects.
In anticipation of possible regulations regarding mercury emissions, research efforts sponsored
by DOE, EPRI, and others are investigating the risks posed by mercury emissions, improved
techniques for measuring those emissions, and possible control measures. The focus in the
control research is on techniques that can be used in conjunction with existing flue-gas-cleanup
(FGC) systems in order to minimize additional capital costs and operational complexity. The
very small amount of mercury (on the order of a few micrograms per cubic meter) in flue gas, its
occurrence in several chemical forms that vary from system to system, the very low solubility of
the elemental form, and the fact that it is usually in the vapor phase combine to make the
achievement of cost-effective control a challenging task.
Argonne National Laboratory has supported the DOE Fossil Energy Program for over 15 years
with research on advanced environmental control technologies. The emphasis in Argonne's work
has been on integrated systems that combine control of several pollutants. Specific topics have
included spray drying for sulfur dioxide and particulate-matter control with high-sulfur coal,
combined sulfur dioxide and nitrogen oxides control technologies, and techniques to enhance
mercury control in existing FGC systems. The latter area has focused on low-cost dry sorbents1
for use with fabric filters or electrostatic precipitators and techniques for improving the capture
of mercury in wet flue-gas desulfurization (FGD) systems. One way of improving that capture is
to convert relatively insoluble elemental mercury (Hgฐ) to a more soluble form, such as mercuric
chloride (see Table 1 for a list of selected chemical formulas). This paper presents results from
recent work that has studied the effects of several oxidizing agents in combination with typical
flue-gas species (e.g., nitrogen oxides and sulfur dioxide) on the oxidation of Hgฐ. The results of
initial laboratory studies indicate that conversion of Hgฐ to an oxidized form is feasible, and that
a combined sulfur dioxide/nitrogen oxides/mercury scrubbing process may be possible. Further
process development research is designed to determine both the technical and economic potential
of such a process.
Background
The fate of trace elements liberated in the combustion process can be influenced by the type of
boiler, the operating conditions, other species present in the flue gas, and the FGC system.
Mercury is a particular problem because it belongs to a group of elements and compounds
denoted as Class III, which remains primarily in the vapor phase within the boiler and subsequent
FGC system. Wet scrubbing is commonly used to control gas-phase species, but in field tests
conducted by DOE and EPRI, mercury removal in wet FGD systems has been highly variable.
Removal values have ranged from about 10% to over 80%.2 Much of this variation may be
caused by differences in the chemical form of the mercury. The presence of chlorine in coal
means that mercury can be found in both the elemental and oxidized forms, with the relative
amounts depending on such factors as the ratio of chlorine to mercury, the gas temperature, and
the gas residence time at various temperatures.3 While other species are also possible and may
-------
be present in small amounts, Hgฐ and mercuric chloride appear to be the most significant for
control considerations. The huge difference in solubility between the two species is particularly
important in wet scrubbing applications.
Argonne's research has focused on the control of Hgฐ. Initial experiments4 used a laboratory-
scale wet scrubbing system that had been well characterized in previous work on combined
sulfur dioxide/nitrogen oxides control.3 For those tests, the feed-gas stream consisted of nitrogen
with about 40 ng/m3 of Hgฐ. The scrubber was first operated as a partially flooded column with
water, a calcium hydroxide solution, or a calcium hydroxide solution plus potassium polysulfide
as the scrubbing liquor. No appreciable mercury removal was found in any of those cases.
Some improvement was found when ceramic-saddle packing was added to the scrubber, but the
removals were still under 10%. More promising results were found when the packing was
changed to stainless steel and used in conjunction with potassium polysulfide hi the scrubbing
liquor. Removals of up to 40% were obtained. However, the use of the polysulfide in FGD
systems could be precluded by the fact that a very high pH is required to maintain its stability.
At that point in the program, the emphasis was shifted to the study of techniques for changing the
chemical form of mercury hi order to produce a more soluble species. Tests were conducted with
several additives that combine strong oxidizing properties with relatively high vapor pressures
(e.g., chlorine). Tests with minimal gas-liquid contacting yielded mercury removals as high as
100% and indicated that the removal reactions were occurring in the gas phase above the
scrubber liquor. However, tests with the addition of sulfur dioxide to the gas stream showed the
additives to be very reactive with that species as well, which could result in excessively high
additive consumption in order to realize effective mercury control.
Next, tests were conducted with a chloric-acid-based chemical, NOXSORB, which is a
product of the Olin Corporation. Typical feed-gas compositions included 1,000 ppm sulfur
dioxide, 200 ppm nitric oxide, 15% carbon dioxide, and 33 ug/m3 of Hgฐ. For a batch test with a
dilute (4%) solution of the as-received NOXSORB concentrate, an outlet reading of zero was
obtained for Hgฐ for approximately 24 min. During that period, the nitric oxide outlet
concentration decreased rapidly to near zero and then rose gradually to where it was almost equal
to the inlet value. The breakthrough in the outlet Hgฐ concentration (the point at which the
concentration rose above zero) appeared to coincide with the point at which the nitric-oxide
outlet concentration leveled off. The apparent correlation between the two removals indicated
that the mercury could be reacting with a product or intermediate of the nitric-oxide removal
process. Subsequent tests with and without nitric oxide in the flue gas again suggested that nitric
oxide promoted Hgฐ removal by NOXSORB.
The results of those tests indicated that not only could effective mercury removal be achieved via
this approach, but that a combined process that also removed nitric oxide might be feasible. To
explore in more detail the interactions among Hgฐ, oxidizing additives, and the various flue-gas
-------
species, a series of experiments using bubblers was designed. The following sections of this
paper describe the experimental setup, procedures, and results of those tests.
Experimental Setup and Procedures
For these experiments, a simulated flue gas was passed through a series of three bubblers for 30
min. A solution of the reactive chemical to be tested was placed in the first bubbler and the
degree of Hgฐ conversion was determined by comparing the amount of mercury found in the
bubbler solutions with the total amount of Hgฐ fed in the flue gas. The Hgฐ concentration in the
gas was typically 45 ug/m3.
The source of Hgฐ was a calibrated and certified permeation tube from VICI Metronics, which
was placed in a constant-temperature water bath controlled to ฑ 0.5 ฐC. Bottled, high-purity
(99.998%) nitrogen gas flowed around the permeation tube to produce a gas stream with a
constant concentration of Hgฐ. This stream was then combined with another gas stream
containing nitrogen and other gaseous components, such as oxygen, carbon dioxide, nitric oxide,
and sulfur dioxide. Carbon dioxide was used as a carrier gas for the nitric oxide. Oxygen was
obtained from a laboratory air line without further purification. Carbon dioxide, nitric oxide, and
sulfur dioxide were used from bottled gases without further purification. The nominal purities
for these gases were as follows: carbon dioxide, 99.5%; nitric oxide, >99.0%; and sulfur
dioxide, >99.98%. After blending, the initial gas composition was checked by using standard
flue-gas analyzers from Beckman instruments: oxygen, Model 755 Oxygen Analyzer; carbon
dioxide, Model 864 Infrared Analyzer; nitric oxide, Model 951A NO/NOX Analyzer; and sulfur
dioxide, Model 865 Infrared Analyzer. Typical concentrations of the various gas components
were as follows: oxygen, 5%; carbon dioxide, 15%; nitric oxide, 250 ppm; and sulfur dioxide,
l,000ppm.
After the feed-gas composition was measured and stabilized, a valve was turned to admit the gas
mixture to the series of three bubblers, each containing 150 mL of solution. While the first
bubbler contained the solution to be studied, the second and third bubblers usually contained
distilled water. Commercial solutions of iodine, bromine, chlorine (sold as sodium
hypochlorite), and chloric acid (NOXSORB ) were used without further purification.
Following the 30-min test, liquid samples from each bubbler were saved for total mercury
analysis. Analyses were performed by a standard cold-vapor atomic absorption
spectrophotometric method (U.S. EPA Method 7470A, SW-846). The estimated accuracy for
this method is ฑ 10% orฑ 0.02 ug/L, whichever is greater.
One objective of the bubbler experiments was to determine the degree to which the important
reactions were occurring in the gas phase, the liquid phase, or both. However, this was difficult
to determine from the initial arrangement. We could presume that mercury found in the second
bubbler was due to gas-phase reactions, assuming that liquid carryover from the first bubbler was
negligible. However, the situation in bubbler #1 was more complex. We could not distinguish
between gas-phase reactions occurring inside the gas bubbles followed by rapid dissolution of
-------
products into the liquid phase versus gas dissolution at the gas-liquid interface followed by a
rapid liquid-phase reaction. To gain some insights into this issue, a second series of tests was
performed, wherein the open-end inlet tube in bubbler #1 was replaced with a fritted-glass
cylinder having a coarse porosity. This gas diffuser greatly decreased the size of the bubbles (by
about an order of magnitude) passing through the liquid in the first bubbler, thereby increasing
the available gas-liquid contact area. Thus, we expected that the results of the small-bubble tests
would be more influenced by gas-liquid interactions than the corresponding large-bubble tests.
Experiment Results
Before any tests were performed with various solutions in bubbler #1, a number of baseline tests
were carried out with only distilled water in all three bubblers. When no Hgฐ was added to the
gas stream, no mercury was found in any of the bubbler solutions. This result demonstrated that
the system was free of mercury contamination to the detection limit (0.02 ug/L or 0.003 (ig
mercury in 150 mL of water) of the analytical method. Such baseline tests were run periodically
to ensure that no mercury contamination had built up during testing. Two tests were performed
with Hgฐ added to the feed-gas stream, but again with only distilled water in all three bubblers.
For one test with large bubbles, amounts of mercury barely above the detection limit (0.004-
0.005 ug in 150 mL) were found in each of the three bubblers. This amount may be compared
with the calculated amount of Hgฐ in the gas stream of 1.9 ug for the 30-min test. This result
showed that the amount of Hgฐ removed using only distilled water was less than 0.3%. For a
similar test with small bubbles, no mercury was found in either bubbler #1 or bubbler #2.
Therefore, any amount of mercury found in the bubblers above these baseline amounts must be
from reactions of Hgฐ with components of the various solutions tested in bubbler #1.
To verify that the effects of other flue-gas components were due to interactions with the test
solutions, additional experiments were run in which the bubblers contained only water. The
synthetic flue gas contained either nitrogen, oxygen, sulfur dioxide, and Hgฐ, or nitrogen,
oxygen, sulfur dioxide, carbon dioxide, nitric oxide, and Hgฐ. In both cases, less than 1% of the
inlet mercury was found in the liquid phase at the end of the experiment. Thus, these flue-gas
species alone are not capable of significantly affecting the solubility of mercury.
The following sections discuss the results of experiments with several different reactive agents in
the bubbler #1 solution. The discussion is divided into two sections, with the results for the
large-bubble, open-ended tube presented first, followed by the results for the small bubble,
coarse-fritted-cylinder inlet. Most of the results are presented to three significant figures.
However, recent analysis of nine small-bubble experiments with NOXSORB indicated that a
systematic change in the results occurred during the experimental program and a correction
factor of about 33% has been applied to some of the data. Those values are only presented to
two significant figures.
-------
Large Bubble Tests
Iodine Solutions. A commercial preparation of 0.100 N iodine solution was diluted to make
up various iodine concentrations. Note that commercial iodine solutions generally contain
potassium iodide as a stabilizer, as well as dissolved iodine. Previous experiments in our
laboratory had shown that iodine solutions react rapidly with mercury vapor in gas streams
containing only Hgฐ and nitrogen.6 However, those tests were performed with only a gas-phase
Hgฐ analyzer to measure the mercury concentration in both the feed and effluent gas streams.
That analyzer was later proven unreliable for measuring Hgฐ in some complex gas mixtures. The
tests reported here are the first ones in which liquid-phase mercury was measured.
Using a solution of about 125 ppb iodine in bubbler #1 (the same concentration employed in
earlier scrubber experiments), we found that more than 90% of the inlet mercury was retained in
the bubbler solutions when only nitrogen and Hgฐ were in the feed-gas stream. (Note that all of
the percentages presented in this paper represent the total mercury found in the liquids from all
three bubblers, although the water in bubbler #3 was not analyzed for mercury unless a
significant amount was found in bubbler #2.) However, when oxygen and carbon dioxide were
added to the gas mixture, Hgฐ removal was reduced to about 6%. In an effort to improve the
removals, further tests were run with an iodine concentration of 250 ppb. The result for a gas
stream containing only Hgฐ and nitrogen was not substantially different from the lower
concentration test (-81% Hgฐ removal), but when either nitric oxide or sulfur dioxide (or both)
were added to the gas mixture, the amount of mercury found in the bubbler solutions went either
to zero (for nitric oxide or for nitric oxide plus sulfur dioxide) or close to zero (-1% for sulfur
dioxide).
Two additional tests were performed with a still higher iodine concentration of 12.7 ppm, which
was chosen such that the reagent cost would be approximately equal to that for a 250 ppm
bromine solution or a 2,500 ppm chlorine solution. The results are shown in Table 2. While no
experiments were run for just Hgฐ and nitrogen in the gas stream, it can be safely assumed that a
mercury capture of over 90% would have been obtained. When just oxygen was added to the
gas, the removal was reduced to 41%, and when carbon dioxide and nitric oxide were added, the
removal fell still further to about 35%.
In earlier unpublished work in our laboratory, we found that the concentration of gaseous Hgฐ
was substantially reduced simply by passing Hgฐ vapors (mixed with nitrogen gas only) over an
agitated iodine solution (-250 ppb). Those tests showed that the most likely Hgฐ removal
mechanism was a rapid gas-phase reaction between iodine vapors and Hgฐ, probably yielding
mercuric iodide. This gas-phase reaction may explain why a considerable portion (-35%) of the
total mercury found in the liquid phase was found in bubblers #2 and #3. However, as discussed
above, we cannot exclude the possibility that some of the mercury found in bubbler #1 was from
a liquid-phase reaction at the gas-liquid interface. Indeed, a published report on the reaction of
iodine in solution with dissolved Hgฐ stated that "the rate was too fast to measure."7 The product
of the solution reaction was shown to be mercuric iodide by its ultraviolet spectrum. Therefore,
without further modeling studies and knowledge of the rate constants (no experimental data on
-------
the gas-phase reaction of Hgฐ with iodine could be found in the literature), we cannot specifically
analyze how much of each reaction (gas phase versus liquid phase) is responsible for the mercury
found in bubbler # 1.
Bromine Solutions. A commercial solution containing 3% bromine (by weight) was used to
prepare a solution containing about 250 ppm bromine. This solution was tested with three
different gas compositions. As can be seen in Table 2, the highest Hgฐ removal was obtained in
the test without either nitric oxide or sulfur dioxide. Adding nitric oxide to the gas mixture
resulted in a significant decrease in the Hgฐ removal, while adding sulfur dioxide resulted in an
even larger decrease in the removal performance. This decrease was presumably caused by the
consumption of bromine in reactions with the two species, which in the case of dissolved sulfur
dioxide would yield bromide and sulfate ions.
Chlorine Solutions. Solutions of molecular chlorine are more complex than those of iodine
because of the greater tendency of chlorine to disproportionate in aqueous solution to
hypochlorous acid and chloride ions. Bromine demonstrates intermediate characteristics in this
regard. Commercial chlorine solutions are sold as sodium hypochlorite because, in alkaline
solutions, the equilibrium between molecular chlorine and hypochlorite ions greatly favors the
latter. Nonetheless, because of the various equilibria involved, detectable amounts of chlorine
will exist in both the gas and liquid phases. Table 3 presents the results for the removal of Hgฐ
by chlorine solutions, which appears to depend both on the composition of the feed-gas mixture
and, in some cases, on the concentration of the chlorine solution.
For gas mixtures containing only oxygen and nitrogen, the Hgฐ removal did not change
appreciably for different chlorine solution concentrations. To help us understand the observed
behavior, we surveyed the literature for data on the gas-phase reaction of Hgฐ with chlorine.
Recent modeling work has assumed the rate constant to be very small.8 However, laboratory
experiments have yielded conflicting results. Some workers have found this reaction to be slow,9
while others have found it to be relatively fast.10 Still other work has shown the reaction of Hgฐ
with chlorine to be surface catalyzed." It appears from these conflicting results that one must be
very careful in interpreting data for this reaction. Our data suggest that the rate of reaction
between Hgฐ and chlorine is not fast, because not much change in Hgฐ removal was observed
with increasing chlorine concentration. Our conclusion on the gas-phase reaction of Hgฐ with
chlorine is that it is slow unless there is an appropriate surface available to catalyze the reaction.
For the gas mixtures containing nitric oxide or nitric oxide plus sulfur dioxide, Hgฐ removal
increased with increasing chlorine concentration. However, the rate of increase differed for the
two gas mixtures. Addition of nitric oxide to the feed-gas mixture appeared to have a definite
positive effect on the amount of mercury transferred to the liquid phase, as compared with the
removals obtained with only oxygen and nitrogen present. An explanation for this behavior
might be that nitric oxide reacts with chlorine to yield nitrosyl chloride. This reaction has been
described in the literature and appears to occur rapidly at room temperature.12 Although we
-------
could not find a literature reference to the reaction of nitrosyl chloride with Hgฐ, one paper did
report that nitrosyl chloride oxidizes mercurous chloride to mercuric chloride, as well as
oxidizing elemental zinc and copper.13 Our conclusion for the reaction of Hgฐ in the presence of
nitric oxide is that nitrosyl chloride probably reacts faster with Hgฐ than chlorine does.
The results in Table 3 also show that when sulfur dioxide is added to the feed-gas mixture, Hgฐ
removal is much lower at the lower chlorine concentrations than when sulfur dioxide is not
present. However, at the highest chlorine concentration studied (5,000 ppm), the Hgฐ removal
performance with sulfur dioxide present was actually slightly higher than the performance
without sulfur dioxide. It is difficult to postulate a mechanism in which the presence of sulfur
dioxide could actually increase the oxidation of Hgฐ by chlorine, because it is well known that
sulfite ions will reduce molecular halogens to their corresponding halides. Perhaps the
improvement in Hgฐ removal with chlorine concentration can be understood as simply the result
of an excess of chlorine that swamps the reaction between dissolved bisulfite (from absorbed
sulfur dioxide) and chlorine and/or hypochlorite ions in solution.
Chloric-Acid Solutions. Chloric-acid solutions were prepared from concentrated
NOXSORB, which has a nominal composition of 17.8% chloric acid and 22.3% sodium
chlorate. The results for tests with concentrations of 0.71% chloric acid (25:1 dilution of the
concentrated stock solution) and 3.56% chloric acid (5:1 dilution) are shown in Table 4. The
primary vapor-phase species above these solutions is thought to be chlorine dioxide. However,
chlorine dioxide is very reactive and readily photolyzes to molecular chlorine and oxygen. Also,
in the presence of moisture, chlorine dioxide can produce a number of different chlorine
oxyacids, such as hypochlorous acid and chlorous acid. Therefore, a large number of different
species may be present in the vapor above a chloric-acid solution. To the best of our knowledge,
there has been no previous research addressing the reaction of Hgฐ with either chlorine dioxide or
chlorate anions.
As shown in Table 4, the change in Hgฐ removal from a solution of 0.71% chloric-acid
concentration to one with about a five times higher concentration is about the same for each of
the three different feed-gas mixtures. In each case, the Hgฐ removal was about a factor of two
higher with the 3.56% chloric-acid solution. Next, we note that gas mixtures containing nitric
oxide showed a substantially higher Hgฐ removal than the gas mixtures without nitric oxide.
This result is similar to what was observed with chlorine solutions. However, a different
chemical mechanism is probably responsible in this case. The reaction of nitric oxide with
NOXSORB solutions may produce hydrochloric and nitric acids among its products.14
Because nitric acid dissolves liquid elemental mercury, we propose that this gaseous by-product
causes the improved Hgฐ removal when nitric oxide is present in the gas stream. The results of
these tests suggest that the gas-phase reaction of Hgฐ with nitric acid might be rapid and should
be examined further.
Contrary to the behavior observed with chlorine, we found that for both concentrations studied,
the presence of sulfur dioxide in the feed-gas stream reduced the Hgฐ removal by about 30%
-------
from the level without sulfur dioxide (but with nitric oxide). Also in contrast to the behavior
observed with chlorine solutions, it appears that within this range of chloric-acid concentrations,
the reduction in Hgฐ removal when sulfur dioxide is present cannot be overcome with higher
chloric-acid concentrations. This result again points to the possibility that a mechanism different
from chlorine oxidation of Hgฐ is operating for these solutions.
Small-Bubble Tests
The small-bubble tests were performed to learn more about the relative importance of gas-phase
versus liquid-phase mechanisms in the Hgฐ removal processes. For a constant volume of gas, the
smaller bubbles produce a larger gas-liquid contact area, which is proportional to the total gas-
bubble surface area. Thus, it was expected that the small-bubble tests might be influenced more
by gas-liquid interactions, in contrast to the large-bubble tests where gas-phase interactions
would be more important.
Bromine Solutions. Only two bromine tests were performed with the small-bubble apparatus.
Both tests used a solution of 250 ppm bromine. One test used a feed-gas mixture of oxygen plus
nitrogen (plus Hgฐ) and gave a Hgฐ removal of 69%. A similar large-bubble test that gave a
removal of 71.1 %. A second test used a feed-gas mixture of oxygen plus nitrogen plus carbon
dioxide plus nitric oxide and gave a Hgฐ removal of 38%, as compared with similar large-bubble
tests that gave a Hgฐ removal of 46.3% (average of two tests). Thus, there did not appear to be
any significant differences between the results for the two bubble sizes.
Chlorine Solutions. Results for the small-bubble tests using chlorine solutions are shown in
Table 5. The remarkable result found for these tests is that approximately the same Hgฐ removal
performance (22% ฑ 3%) was observed regardless of chlorine concentration (from 100 ppm to
2,500 ppm) or gas composition. This finding indicates that a single mechanism, most likely
involving gas-liquid contact, is dominating Hgฐ removal under these conditions. By comparing
the data in Tables 3 and 5, it appears that gas-phase interactions become more important for
higher chlorine concentrations (>250 ppm) and for feed gas mixtures containing either nitric
oxide alone or nitric oxide plus sulfur dioxide.
Chloric-Acid Solutions. Table 6 gives results for the small-bubble tests using NOXSORB
solutions in bubbler #1. The pattern for those results is similar to what was obtained in the
chlorine tests, with little variation in Hgฐ removal regardless of chloric-acid concentration or gas
composition. Presumably, this result indicates that a single mechanism was dominating Hgฐ
removal. By comparing the data in Tables 4 and 6, it again appears that gas-phase interactions
become more important at higher chloric-acid concentrations and with either nitric oxide alone or
nitric oxide plus sulfur dioxide present in the gas stream.
-------
Conclusions
Previous research at Argonne and elsewhere has indicated that typical wet FGD systems are not
effective in controlling emissions of elemental mercury. Although more intensive scrubbing
with the addition of certain types of packing may yield some degree of control, a more effective
approach could be to convert the mercury into a form that is readily absorbed. Our research thus
far has shown that halogen-containing solutions are capable of promoting the removal of
gaseous, elemental mercury by aqueous solutions, presumably by oxidizing the Hgฐ to soluble
compounds, such as mercuric chloride. The specific behavior of these solutions depends on
many factors, including the reactive species, the gas-phase composition, and the degree to which
gas-gas or gas-liquid reactions are important.
An FGD process modified for mercury control could involve upstream injection of a reactive
species, with Hgฐ conversion depending upon gas-phase reactions. Iodine solutions can be
effective in oxidizing Hgฐ, even at very low concentrations (< 1 ppm). This appears to be due to
a rapid gas-phase reaction. However, that effectiveness is lost when species other than nitrogen
and Hgฐ are in the gas stream. Whether this interference is caused only by reaction of iodine
with the other gaseous components, or whether another mechanism is responsible for the
interference, cannot be determined from these tests. In any case, iodine does not appear to be an
attractive option for the oxidation of elemental Hgฐ in the presence of gases other than nitrogen.
For bromine, the similarity of the results for large and small bubbles indicates that gas-liquid
reactions are not the dominant removal mechanism, which is in agreement with the results for
iodine. While substantial conversion of Hgฐ by bromine was obtained when only oxygen and
nitrogen were in the gas stream, the addition of nitric oxide and sulfur dioxide again diminished
that conversion significantly. Thus, neither iodine nor bromine is likely to be cost-effective in a
commercial system.
A different pattern of behavior was found for solutions containing chlorine or chlorine
compounds. Chlorine solutions showed no dependence on concentration when nitric oxide and
sulfur dioxide were absent, indicating that the mercury-chlorine reaction is probably slow
without the presence of a catalyst. Addition of nitric oxide to the gas stream greatly increased
the amount of Hgฐ removed. This increase in removal may have been due to the formation of an
intermediate compound, such as nitrosyl chloride, which could react rapidly with the Hgฐ. On
the other hand, sulfur dioxide depressed the Hgฐ removal, at least at lower chlorine
concentrations. Nevertheless, the removal appeared to increase with chlorine concentration when
either nitric oxide alone or nitric oxide plus sulfur dioxide were added to the gas stream.
Mercury removal with chloric-acid solutions also appeared to increase with increasing chloric-
acid concentration regardless of gas composition. In a similar manner to chlorine, the presence
of nitric oxide greatly increased Hgฐ removal. In this case, the important gas-phase reaction may
involve nitric acid formed from the reaction of nitric oxide and chloric acid. The presence of
sulfur dioxide decreased Hgฐ removal, but it remained intermediate to that with and without nitric
oxide.
10
-------
For the small-bubble tests, gas-liquid interactions should play a larger role. For both chlorine
and chloric-acid solutions, Hgฐ removal appeared to be constant regardless of solute
concentration or gas-phase composition. The average removal with chlorine solutions was about
22%, while the average removal with chloric-acid solutions was about 40%. Both of these
values were somewhat higher than the "baseline" large-bubble results obtained with only
nitrogen or nitrogen plus oxygen in the gas stream, but they were lower than the removals
obtained at higher solute concentrations with nitric oxide in the gas stream. Thus, it appears that
both gas-gas and gas-liquid reactions are operating in these systems, with the gas-phase reactions
involving nitric oxide becoming increasingly important as the solute concentration is raised. In
that situation, some degree of nitric-oxide removal may also be obtained as part of the reaction
mechanism. Soluble oxidation products could then be removed in a downstream aqueous
scrubber system.
Continuing Research
The results of the bubbler experiments with chlorine and chloric acid indicate that these species
may promote Hgฐ removal in wet scrubber systems and that there may be synergistic effects that
can promote even higher levels of removal in the presence of nitric oxide. While sulfur dioxide
clearly seems to exert a negative influence on Hgฐ removal, this effect may be overcome or
avoided through appropriate process design. In order to explore process possibilities under more
realistic operating conditions, larger-scale laboratory experiments have been designed to explore
the potential for an integrated sulfur dioxide/nitrogen oxides/mercury control technology.
The design of the new experimental system is based on the premise that the key reactions for a
combined removal process take place predominantly in the gas phase. Initially, the system has
been set up to simulate injection of a reactant into the ductwork upstream of a scrubber. A
solution of the reactive species (e.g., chlorine or chloric acid) is sprayed into a flowing stream of
synthetic flue gas by an ultrasonic atomizer. A residence time of about 1-2 sec is allowed for
reactions to occur before the gas passes through a scrubber vessel to capture any soluble reaction
products. Samples of the scrubber liquor are analyzed to determine the extent of mercury
capture. The removals for other species, such as nitric oxide, are determined using conventional
flue-gas analyzers.
In the experimental program that is now underway, the effectiveness of Hgฐ and nitric-oxide
removal will be evaluated for different gas compositions, reactant addition rates, and gas
residence times. If the results indicate that sulfur dioxide exerts such a negative effect that
excessive amounts of reactant are required, other process concepts incorporating prescubbing of
sulfur dioxide may be evaluated. Data obtained from these experiments will help determine key
parameters for further tests at the scale of a process development unit.
11
-------
Acknowledgments
This work was supported by the U.S. Department of Energy, Assistant Secretary for Fossil
Energy, under contract W-31-109-ENG-38, through the Federal Energy Technology Center. We
gratefully acknowledge the support and guidance provided by the Contracting Officer's
Representative, Peter Botros, as well as by Perry Bergman and Tom Brown of the Center.
References
1. Livengood, C.D.; Huang, H.S.; Mendelsohn, M.H.; Wu, J.M., 1996, "Enhancement of
Mercury Control in Flue-Gas Cleanup Systems," Proc. U.S. DOE/PETC First Joint Power
& Fuel Systems Contractors Conference, Pittsburgh, Perm., July 9-11.
2. Chang, R.; Hargrove, B.; Carey, T.; Richardson, C.; Meserole, F., 1996, "Power Plant
Mercury Control Options and Issues," Proc. POWER-GEN '96 International Conference,
Orlando, Fla., Dec. 4-6.
3. Huang, H.S.; Wu, J.M.; Livengood, C.D., 1995, "Development of Dry Control
Technology for Emissions of Mercury in Flue Gas," Proc. The Fourth International
Congress on Toxic Combustion Byproducts, Berkeley, Calif. June 5-7.
4. Mendelsohn, M.H.; Wu, J.; Huang, H.; Livengood, C.D., 1994, "Elemental Mercury
Removals Observed in a Laboratory-Scale Wet FGD Scrubber System," Emerging Clean
Air Technologies and Business Opportunities, Toronto, Canada, Sept. 26-30.
5. Mendelsohn, M.H.; Harkness, J.B.L., 1991, "Enhanced Flue-Gas Denitrification Using
Ferrous'EDTA and a Polyphenolic Compound in an Aqueous Scrubber System," Energy
& Fuels, 5(2):244-247.
6. Livengood, C.D.; Mendelsohn, M.H.; Huang, H.S.; Wu, J.M., 1995, "Development of
Mercury Control Techniques for Utility Boilers," 88th Annual Meeting Air & Waste
Management Association, San Antonio, Texas, June 18-23.
7. Warrick, Jr., P.; Wewerka, E.M.; Kreevoy, M.M., 1962, "The Reactions of Iodine in
Solution with Elemental Mercury," J. Am. Chem. Soc., 85:1909-1915.
8. Pleijel, K.; Munthe, J., 1995, "Modelling the Atmospheric Mercury Cycle - Chemistry in
Fog Droplets," Atmospheric Environment, 29:1441-1457.
9. Skare, I.; Johansson, R., 1992, "Reactions Between Mercury Vapor and Chlorine Gas at
Occupational Exposure Levels," Chemosphere, 24:1633-1644.
12
-------
10.
11.
12.
13.
14.
Skripnik, V.A.; Fedorovskaya, L.F.; Kravetskii, L.I.; Umanskaya, I.M., 1979,
"Mechanism and Kinetics of Mercury Oxidation by Chlorine-Containing Solutions," Zh.
Prikl. Khim. (Leningrad), 52:1233-1237 (Engl. trans. 1169-1172).
Medhekar, A.K.; Rokni, M.; Trainor, D.W.; Jacob, J.H., 1979, "Surface Catalyzed
Reaction of Hg + C12," Chem. Phys. Letters, 65:600-604.
Stoddart, E.M., 1944, "The Kinetics of the Reaction between Chlorine and Nitric Oxide "
J. Chem. Soc., 388-393.
Partington, J.R.; Whynes, A.L., 1948, "The Action of Nitrosyl Chloride on Some Metals
and Their Compounds," J. Chem. Soc., 1952-1958.
Kaczur, J.J., 1996, "Oxidation Chemistry of Chloric Acid inNOx/SOx and Air Toxic
Metal Removal from Gas Streams," Environmental Progress, 15(4):245-254.
The submitted manuscript has teen authored
by a contractor at trie U.S. Government
under contract No. W-3M09-ENG-38.
Accordingly, 'he U. S. Government retains a
nonexclusive, royalty-free license lo publish i
or reproduce the published form ol this
contribution, or allow Others to do so, for
U. S. Government purposes.
13
-------
Table 1
List of Selected Chemical Formulas
Chemical Name
Bromine
Carbon Dioxide
Calcium Hydroxide
Chloric Acid
Chlorine
Chlorous Acid
Hypochlorous Acid
Iodine
Mercury (elemental)
Mercuric Chloride
Nitric Oxide
Nitrogen
Nitrogen Oxides
Nitrosyl Chloride
Oxygen
Sodium Chlorate
Chemical Formula
Br2
CO2
Ca(OH)2
HC1O3
C12
HC1O2
HOC1
I2
Hgฐ
HgCl2
NO
N2
NOX
NOC1
02
NaClO3
Table 2
Summary of Hgฐ Removal Results for Large-Bubble Tests with Iodine and Bromine Solutions
Hgฐ Removal (%)
Feed-Gas Composition
02 + N2 + Hgฐ
O2+N2 + NO + CO2 + Hgฐ
O2 + N2 + NO + CO2 + S02 + Hgฐ
I (12.7 ppm)
41.4
34.9
Br (250 ppm)
71.1
50.9,41.6
11.8
14
-------
Table 3
Summary of Hgฐ Removal Results for Large-Bubble Tests with Chlorine Solutions
Chlorine Concentration (ppm)
Feed-Gas Composition
02 + N2 + Hgฐ
02 + N2 + NO + C02 + Hgฐ
O2 + N2 + NO + C02 + SO2 + Hgฐ
2.5
11.6
19.0
0.5
250
14.4, 13
20.6"
13.8
1,000
Hgฐ removal (%)
.3
35.4,28.1
35.1,34
2,500
9.3
37
35.4,41
5,000
14.3
44.5
52.2
a!5-min test as versus 30 min for other data.
Table 4
Summary of Hgฐ Removal Results for Large-Bubble Tests with HC1O3 Solutions
HC1O3 Concentration (%)
0.71 3.56
Feed-Gas Composition Hgฐ removal (%)
14.0 26.9
, + N2 + NO + CO2 + Hgฐ 33.9 69.6
, + N2+NO + CO2 + SO2 + Hgฐ 22.8 48.2
15
-------
Table 5
Summary of Hgฐ Removal Results for Small-Bubble Tests with Chlorine Solutions
Chlorine Concentration (ppm)
100 250 500
2,500
Feed-Gas Composition Hgฐ removal (%)
O2 + N2 + Hgฐ - 23
O2 + N2 + NO + CO2 + Hgฐ 23 20 22
O2+N2 + NO + CO2 + SO2 + Hgฐ 18
22
20
27
Table 6
Summary of Hgฐ Removal Results for Small-Bubble Tests with HC1O3 Solutions
HC1O3 Concentration (%)
0.356 0.71 1.78
3.56
Feed-Gas Composition Hgฐ removal (%)
O2 + N2 + Hgฐ -- 44.6,38
O2 + N2 + NO + CO2 + Hgฐ -- 34.5
O2 + N2 + NO + CO2 + SO2 + Hgฐ 38 39.9 47
40.8
38
38.7
16
-------
MERCURY ABSORPTION IN AQUEOUS HYPOCHLORITE
Lynn L. Zhao
Gary T. Rochelle
Department of Chemical Engineering
The University of Texas at Austin
Abstract
The absorption of elemental Hg vapor into aqueous hypochlorite was measured in a stirred tank
reactor at 25 and 55ฐC. NaOCl strongly absorbs Hg even at high pH. Low pH, high Cl~ and high
temperature favor mercury absorption. Aqueous free C\2 was the active species that reacted with
mercury. However, chlorine desorption was evident at high Cl~ and pH < 9. Overall second order
reaction was observed between mercury and chlorine with an apparent second order rate constant
of 1.7xl015 JVHs-1 at 25ฐC and 1.4 x 1017 lvHs-1 at 55ฐC.
Gas phase reaction was observed between Hg and C\2 on apparatus surfaces. Strong mercury
absorption in water was also detected with C\2 present. Results indicate that the chlorine
concentration, moisture and surface area contribute positively to mercury removal.
Introduction
The capability of hypochlorite to dissolve mercury has long been recognized. Parks and Baker1
described the recovery of mercury by contacting mercury-containing material with a hypochlorite
solution maintained at pH of 4.5 to 9.5. Nguyen2 demonstrated the use of hypochlorite to treat
mercury-containing waste water from chlor-alkali plants. Nene and Rane3 reported quantitative
results on mercury absorption in 5.8-27 mM sodium hypochlorite and 2.6-9.5 mM hypochlorous
acid. They concluded that hypochlorous acid was much more reactive than hypochlorite. With
hypochlorite, reactivity further increased in the presence of sodium or potassium chloride.
Potassium hypochlorite was more reactive than sodium hypochlorite.
Medhekar et al.4 reported surface reaction of mercury with chlorine at 250ฐC to form a compound
with a stoichiometric formula of (HgCl2)n- Cylindrical test cells made from four different materials
were tested. Their initial observations rated the reactivity in the sequence: Teflon-coated stainless
steel > stainless steel > quartz > Inconel. However, it was also observed that after several runs the
reaction rate became faster until it eventually became independent of the surface material. They
concluded that it was likely that the reaction products formed on the surface (HgClz) coated the
surfaces and thus made all surfaces essentially the same. McCannon and Woodfin5 reported that a
reduced amount of mercury vapor was detected by atomic absorption if the gas mixture contained
more than a few ppm chlorine. The reduction in mercury vapor concentration was attributed to the
formation of reaction products between mercury and chlorine. Menke and Wallis6 studied the gas
phase reaction of mercury with chlorine in a cylindrical quartz tube with four levels of Cl2
concentration (0, 0.5, 1.5 and 3.8 ppm) and two levels of relative humidity (13% and 80%). They
concluded that at constant mercury concentration, the rate of the formation of Hg-Cl2 reaction
product (possibly HgCl2) increased with chlorine concentration and relative humidity. They also
observed adsorption of Hg-Cl2 reaction product on the quartz walls.
-------
Experimental Method
All experiments were performed in the well-characterized stirred cell reactor with Teflon surfaces
described in previous papers (Zhao and Rochelle7; Zhao8). A gas rate of 1 1/min was used. The
inlet Hg concentration was 97 ppb.
In a typical experiment, 1.06 liters of distilled water was put into the reactor while elemental
mercury in nitrogen bypassed the reactor. After the mercury analyzer gave a stable reading, the
mercury stream was passed over the distilled water inside the reactor. This condition was used to
calibrate the analyzer. After the analyzer gave a stable reading, known amounts of sodium
hypochlorite solution (Fisher Scientific, Purified Grade, 4-6 wt%) were sequentially injected into
the water using a syringe with a long needle. In some experiments, a small amount (< 1 ml) of
0.01 or 0.1 M HC1 or NaOH solution was injected to modify pH while Hg-N2 was passed over
the solution. The outlet concentration of elemental mercury was analyzed continuously by a cold
vapor atomic absorption Hg analyzer (LDC Analytical).
The rate of mercury absorption was calculated from the gas phase material balance. A three-point
calibration was performed before and after each run to account for the effect of base line drifting of
the mercury analyzer. Hypochlorite addition to the liquid phase was determined by iodometric
titration (Lagowski9). When a dilute amount of hypochlorite was used, its concentration was
determined by weighing the syringe before and after the injection.
During the studies of gas phase reaction of mercury and chlorine (972 ppm C12 in N2), a variety of
apparatus configurations were used. The largest Teflon surface area was exposed with all of the
tubing and the reactor in the gas flow path. When the gas stream bypassed the reactor, then the
surface area of the reactor was excluded. When the gas stream bypassed the reactor and all of the
tubing before the analyzer, the least amount of surface area was exposed to the Hg-Cl2 stream.
Mass Transfer with Simultaneous Chemical Reaction
In aqueous hypochlorite solution, the distribution of OC1", HOC1 and C12 depends on solution pH
and Cl~ concentration. Since lower pH gives higher Hg removal, it is probable that free Cl2 is the
active ingredient that reacts with Hg. The concentration of free C12 can be obtained from the two
equilibria:
HOC1 < KI ) H+ +OC1- (1)
C12 + H2O ( K2 ) HOC1 + Cl- + H+ (2)
Due to the wide concentration ranges of NaCl and NaOCl in the solution, the ionic strength has a
significant impact on the interpretation of experimental results. The activities of the aqueous
components were used instead of concentrations:
[Total NaOCl] = [C12] + [OCJ-] + [HOC1] (3)
(4)
(5)
Thus:
-------
~
[Total NaOCl]
K2 KI K2
with assumptions:
7a2 = y*ป = 1 (7)
Ifcr = Yocr ~ TN^CI (8)
and:
a t = iQ(-measuredpH) (9)
Equation (6) can be simplified to:
n __ [Total NaOCl] _
(10)
YNaC, is determined according to the effective Nad concentration, where:
[NaCl]effective = [NaCl] + [NaOCl] (11)
JNaa is 0.996, 0.903, 0.779 and 0.657 for 0.001, 0.01, 0.1 and 1 M NaCl at 25ฐC, respectively.
YNaCl is 0.768 for 0.1 M NaCl and 0.655 for 1 M NaCl at 55ฐC (Robinson and HarnediO; Lobo
andQuaresma11).
The temperature dependence of the equilibrium constants is given in table 1.
Table 1. Equilibrium constants at two different temperatures
Equilibrium Constant 25ฐC 55ฐC
KI (M)* S.OOxlO-8 4.46x10-8
K2 (M2)* 3.94X10-4 7.10X10-4
*obtained from Atkins12 'Obtained from Connick and Chia13
We have found that the rate of reaction between elemental mercury vapor and hypochlorite is given
by the mechanism:
Hg + Cl2(aq) _> Products (12)
The reaction rate is given with the second order rate constant, k2:
reaction rate = k2 [Hg] [C12] (13)
Using surface renewal theory with fast reaction near the gas-liquid interface (Danckwerts14), the
flux of elemental mercury, Nng, should be given by:
-------
k2 DHg-H20 [Cl2]i (14)
where:
(15)
(16)
The physical constants and mass transfer coefficients (kgjHg and k\ Hg"""). are given in Zhao and
Rochelle15 and their values are listed in appropriate table captions, if applicable.
For most experiments the results are reported as normalized flux, Kg', analogous to the overall gas
phase mass transfer coefficient:
Kg' = gH* (17)
B PHgi
The Kg** concentration at the gas-liquid interface [Kg**]; is calculated from the experimental flux
by:
[Hg++ ]j = [Kg""- ]b, injtja! injected + [Hg** lb, absorbed + ko S++ (18)
The portion of [Hg^+Jb resulting from mercury absorption was obtained by integration of the gas
phase material balance. The nominal concentration of injected mercuric chloride was used for the
portion of [Hg^+Jb resulting from mercuric chloride injections.
Results and Discussion
Mercury Absorption in Sodium Hypochlorite
Although there is no positive or negative indication of the presence of NaCl in NaOCl by the
manufacturer, it is reasonable to assume that for every mole of NaOCl, there is one mole of NaCl
in NaOCl solution (discussed in detail later). This assumes that NaOCl was manufactured by
bubbling chlorine through NaOH:
C12 + 2NaOH <-ป NaOCl + NaCl + H2O (19)
Injections of NaOCl resulted in step decreases in the outlet mercury concentration. This indicates
that NaOCl (with the presence of NaCl) absorbs mercury readily even at high pH. All the systems
that we have studied previously absorb Hg well only at low pH with the presence of strong acid
and oxidizer. Hypochlorite is the only oxidizer we have tested that substantially removes mercury
at high pH.
The effect of NaCl was investigated by injecting NaOCl into 0.1 and 1 M NaCl. The results are
given in figure 1 . With a trace amount of NaOCl, 1 M NaCl greatly enhanced mercury absorption.
At 10~5 M NaOCl, the normalized flux was increased by a factor of 70. Controlled pH
experiments described later in the paper indicate that lower pH favors mercury absorption. From
the chlorine hydrolysis reaction:
C12 + H2O <-ป HOC1 + H+ + Cl- (20)
It is obvious that lower pH and higher Cl" produce more free C12 in the solution. Therefore it is
probable that free chlorine is the active species reacting with mercury.
-------
Normalized flux, K
(mole/s-cm2-atm)
H- * H^ t I H
o o o e
^ -J O\ t/l
=1
: PH 7*9 -
n NaCI = NaOCl 1
- 1 M NaCI 1 -
V o
94 10.3
c '
E 8.6 E
8.0
7.9
7 10'5 10"3 10
-1
Total injected NaOCl (M)
Figure 1. The effect of NaCI on Hg absorption in NaOCl at 25ฐC.
1.7 Hg.
/ / in
ฃ
B.
a.
3
o
-------
Because of the positive effect of NaCl on mercury absorption in NaOCl, two levels of high NaCl
were tested. Figure 2 gives the results with 0.1 and 1 M NaCl at controlled pH. The low pH was
obtained by adding HC1. At 0.1 M NaCl, injections of both 9.2xlQ-7 and 2.9xlQ-6 M NaOCl
caused the outlet Hg to decrease immediately. However, the outlet Hg increased shortly after
NaOCl was injected. The outlet mercury concentration continuously increased until it approached
the inlet mercury concentration. With 1 M NaCl, injections of NaOCl in the same range caused the
outlet mercury concentration to decrease and stabilize at the reduced value. The results indicate that
the presence of NaCl either helped absorbed mercury stay in the solution, or it produced enough
free chlorine to continuously absorb mercury, or the produced free chlorine desorbed and reacted
with Hg on apparatus surfaces.
Figure 3 gives the effect of pH on mercury absorption in NaOCl with 1 M NaCl. Lower pH
favors mercury absorption. This further indicates that aqueous free chlorine is the active ingredient
reacting with Hg. At very high pH, such as pH 11, a large amount of NaOCl needs to be injected
to reach or approach gas phase control. However, at pH 10 or lower, gas phase control was fairly
easy to achieve. At intermediate pH, there were some values of Kg' that exceeded kg. This extra
amount of Hg removal might have been caused by surface reaction of Hg with desorbed chlorine,
which will be discussed later.
us
oฃ
a
c
CB
U
O
Z
E
ra
10
-4
10
-5
I 10
10
'*
-7
10"
| ' ' ' f 1.0 m 3.0 ฃ
- pH A 7.5
1 0 9.0 10.2 3
ปป 1^^^^^^
- k *
: g m
- i O A
" * :
: o . ฐ * G ฐ E
D
.0
11. 1
10'
10--
10
-3
10
-1
Total injected NaOCl (M)
Figure 3. The effect of pH during Hg absorption in 1 M NaCl with sequential
injections of NaOCl at 25ฐC. Data points for pH < 9 represent the
maximum fluxes associated with each injection of NaOCl.
Figure 4 gives the dependence of normalized flux, Kg', on the estimated activity of free aqueous
Cl2 using equation (10). 9xlO'7 to 1 M NaCl and a pH range of 4.9 to 11.1 were used in these
experiments. By assuming a 1:1 ratio of NaOCl to NaCl, results without externally adding NaCl
to NaOCl were fitted successfully with controlled NaCl-controlled pH results. The data are
tabulated in tables 3a, 3b and 3c.
-------
0 9x10-0.15 M NaOCl=NaCl, pH=4.9-ll.l, 0-0.6 mM HgCl2
1 M NaCl, pH = 9.0-11.1, no HgCl2
ซ S
s-t
*tt C5
s
"3
I*
O
10
tu
"o
s
-5
I I io-ซ
10"
10
-8
10
-18
10
-16
10"
14
Activity of CL (M)
10
-12
Figure 4. Hg absorption in NaOCl-NaCl at 25ฐC. The wide pH range was
obtained by adding NaOCl, HgC12 or NaOH.
Table 3a. Mercury absorption in NaOCl with or without HgC12 injection at 25ฐC.
The inlet Hg was 97.7 ppb. kg, Hg = 0.39 mole/s-atm-m2 and kฐl,
Hg++ = 2.2x10-5 m/s. NaCl concentration was assumed to be the same
as that of the cumulative amount of NaOCl
PHgi
xlOป
atm
8.8
8.2
7.8
7.6
6.8
6.0
5.0
2.8
8.4
7.9
6.9
6.5
8.0
7.8
NHg
x!0fo
mole
s-m2
10.2
14.2
16.7
18.4
23.9
29.7
36.4
51.8
12.8
16.7
23.0
25.8
15.7
17.0
[Hg+^i
M
5.2E-08
7.7E-06
2.3E-05
6.3E-05
1.8E-04
1.7E-04
3.9E-04
6.0E-04
6.2E-08
1.1E-07
1.4E-07
1.8E-07
8.4E-08
9.1E-08
Injected
[NaOCl]
M
3.71E-02
3.71E-02
3.70E-02
3.70E-02
3.69E-02
7.00E-02
6.99E-02
6.98E-02
2.02E-02
4.13E-02
7.83E-02
1.50E-01
3.78E-02
3.90E-02
I NaCl
0.80
0.80
0.80
0.80
0.80
0.76
0.76
0.76
0.84
0.79
0.75
0.72
0.80
0.80
aci2
M
4.3E-15
1.2E-14
1.4E-14
2. IE- 14
5.3E-14
9.1E-14
1.7E-13
1.1E-12
8.3E-15
2.3E-14
4.4E-14
4.8E-14
2.3E-14
2.5E-14
pH
11.12
10.90
10.86
10.77
10.57
10.71
10.57
10.16
10.73
10.80
10.91
11.16
10.76
10.76
k2
xlO'15
1
M-s
2.1
1.7
2.2
1.8
1.5
1.8
2.0
2.0
1.8
1.3
1.6
2.2
1.1
1.3
-------
6.1
5.5
4.4
8.3
8.2
8.1
8.3
8.1
8.0
7.2
1.9
8.4
7.9
7.5
7.2
7.2
7.2
6.4
6.3
10.2
9.5
9.0
8.3
7.9
9.8
9.6
8.9
7.3
5.0
2.6
28.9
32.9
41.2
14.0
14.3
15.1
13.8
15.2
15.5
21.3
58.5
12.7
16.5
19.1
21.0
21.2
21.3
26.7
27.7
0.5
5.2
8.6
13.3
16.1
3.1
4.3
9.0
20.4
36.4
53.2
3.0E-05
1.6E-04
4.0E-04
7.7E-08
8.1E-08
9.2E-08
7.1E-08
4.9E-06
4.9E-06
1.3E-05
1.3E-04
6.8E-08
9.6E-08
1.2E-07
1.4E-07
1.4E-07
1.5E-07
1.8E-07
2.1E-07
2.7E-09
2.6E-08
5.4E-08
8.6E-08
l.OE-07
1.6E-08
9.4E-06
3.0E-05
7.7E-05
2.0E-04
4.4E-04
9.04E-02
9.04E-02
8.98E-02
3.90E-03
3.90E-03
3.63E-03
2.58E-03
2.58E-03
2.58E-03
2.58E-03
2.35E-03
1.17E-03
1.17E-03
1.17E-03
7.22E-03
7.22E-03
7.22E-03
2.72E-02
6.63E-02
9.04E-07
1.23E-05
9.08E-05
9.34E-04
1.13E-02
8.60E-06
8.59E-06
8.56E-06
8.56E-06
8.53E-06
8.52E-06
0.74
0.74
0.74
0.92
0.92
0.92
0.93
0.93
0.93
0.93
0.93
0.95
0.95
0.95
0.89
0.89
0.89
0.82
0.76
0.97
0.97
0.97
0.95
0.87
0.97
0.97
0.97
0.97
0.97
0.97
7.6E-14
1.3E-13
3.3E-13
1.1E-14
2.2E-14
1.8E-14
8.8E-15
1.5E-14
2.3E-14
4.8E-14
4.4E-12
l.OE-14
1.8E-14
2.2E-14
3.5E-14
4.0E-14
4.2E-14
9.5E-14
7.2E-14
8.6E-18
1.1E-15
3.2E-15
8.8E-15
1.9E-14
4.4E-16
9.7E-16
6.5E-15
5.2E-14
2.8E-13
2.3E-12
10.85
10.74
10.53
10.00
9.84
9.86
9.87
9.75
9.66
9.50
8.46
9.50
9.38
9.33
10.00
9.97
9.96
10.32
10.74
7.86
7.95
8.64
9.44
10.31
8.00
7.80
7.25
6.50
5.80
4.89
2.0
1.9
1.8
1.8
0.9
1.3
2.1
1.6
1.1
1.2
1.5
1.5
1.6
1.9
1.6
1.4
1.4
1.2
1.8
2.0
1.8
1.9
1.9
1.4
1.5
1.4
1.0
1.0
1.3
1.2
Table 3b. Hg absorption in NaOCl with 1 M NaCl at 25ฐC with pH varied from
9.0 to 11.1 (obtained by adding NaOH). JNaa is 0.657. kg, Hg was
0.39 mole/s-atm-m2 and k\ Hg++ was 2.2xlQ-5 m/s. Total NaOCl
represents initially cumulatively injected concentration. No external
injection was made
PHg,i
xlO8
atm
4.3
2.6
9.8
9.5
7.8
6.6
4.7
3.8
2.9
NHgxl010
mole
sec in2
41.6
53.3
2.8
5.2
16.8
25.7
38.6
44.8
51.2
Injected
[NaOCl]
M
1.5E-05
3.9E-05
1.6E-06
1.2E-05
9.1E-05
3.3E-04
1.7E-03
4.3E-03
l.OE-02
act
M
4.5E-13
1.6E-12
3.3E-16
2.4E-15
1.8E-14
5.8E-14
2.7E-13
5.9E-13
1.2E-12
pH
9.04
8.97
10.12
10.13
10.13
10.16
10.18
10.21
10.24
k2xlO'15
1
M-s
1.4
1.7
1.7
0.8*
1.7
1.8
1.7
1.5
1.6
-------
10.1
10.0
9.8
9.5
8.8
7.7
1.0
1.9
3.4
5.1
10.0
17.6
1.3E-05
5.8E-05
1.8E-04
4.4E-04
2.0E-03
8.8E-03
3.5E-17
1.5E-16
4.8E-16
1.2E-15
5.0E-15
2.0E-14
11.07
11.07
11.07
11.07
11.08
11.10
1.7
1.7
1.7
1.7
1.7
1.7
represents data not included in rate constant regression
Table 3c. Hg absorption in NaOCl with 0.1 or 1 M NaCl at 55ฐC with pH
varied from 9.3 to 10.1 (obtained by adding NaOH). JNaCl is 0.655
(1M NaCl) or 0.768 (0.1 M NaCl). kg, Hg was 0.39 mole/s-atm-m2
and kฐi, Hg++ was 3.9-4.0xlO-5. Total NaOCl represents initially
cumulatively injected concentration. No external HgCl2 injection
was made.
PHg,i
xlO8
atm
9.6
9.1
8.2
5.5
3.4
6.4
4.4
NHgxl010
mole
sec m^
4.2
7.1
13.3
30.7
44.2
24.6
37.9
Injected
[NaOCl]
M
9.1E-07
3.0E-06
1.1E-05
3.3E-05
8.9E-05
1.2E-05
3.1E-05
[NaCl]
M
1
1
1
1
1
0.1
0.1
act
M
9.4E-17
3.0E-16
1.2E-15
5.1E-15
1.6E-14
5.2E-15
9.5E-15
pH
10.06
10.07
10.03
9.97
9.94
9.31
9.39
k2x!0-17
1
M-s
1.2
1.2
1.2
3.6*
6.1*
1.6
4.6*
represents data not included in rate constant regression
For controlled pH and NaCl experiments with pH > 9 and Cl~ > 0.1 M, an overall second order
reaction between Hg and Cl2 was observed. The combination of high pH and low Cl" also gave
good results. In addition, the same second order behavior was obtained with low Cl" and low pH,
such as with pH = 4.9 and Cl' = 8.5xlQ-6 M.
Figure 5 gives the results of experiments conducted at 25 and 55ฐC. High temperature favors
mercury absorption. At 25ฐC, all data points gave reasonably good fit to second order kinetics.
However, high chlorine points at 55ฐC tended to give higher mercury flux, as shown in figure 5.
The data of high pH at 55ฐC are given in table 3c.
The second order rate constant is 1.7xl015 M'V1 at 25ฐC and 1.4xl017 M-V1 at 55ฐC.
The effect of HgC^ can be determined from table 3a. Since results obtained with external HgCh
injections (ranged from 4.8xlQ-6 to l.OxlO'3 M) gave the same second order rate constant as those
without external HgCh injection, it was concluded that the additions of HgCl2 did not affect Hg
absorption in NaOCl-NaCl. External HgCl2 injections were made in low and intermediate pH
experiments with Cl" > 0.1 M.. The results indicate one more time that HgC^ did not affect Hg
absorption in NaOCl-NaCl.
When a small concentration of HgCl2 was injected into relatively large amount of NaOCl, the net
effect of HgCl2 was not apparent. However, when relatively large changes in HgCl2, such as 1O4
M, were made in 0.1 M NaOCl, the net effect of HgCl2 was significant, as indicated by the step
decreases in the outlet Hg concentration. HgCl2 itself did not help to absorb more Hg, but rather it
reduced the solution pH. Lower pH favors Hg absorption in NaOCl. Small HgCl2 addition to a
-------
large amount of NaOCl did not lower solution pH significantly. On the contrary, large addition of
HgCl2 to relatively dilute NaOCl lowers solution pH significantly and thus absorbs Hg more
readily. Lower pH resulted in more free aqueous chlorine and thus absorbed more Hg. It is only
in this sense that HgCla helped to absorb more Hg.
At pH > 9, no C\2 desorption was apparent. Results with pH as low as 4.9 and low Cl" (8.5 l
also agreed well with those of pH > 9 (no C\2 desorption was experienced at low pH with low Cl".
This is expected since such a low amount of Cl" (8.5 (iM) did not result in significant amount of
Cl2). However, at pH < 9 and Cl" > 0.1 M, Kg' was lower than expected, as shown in figure 5.
During these experiments at low pH, the outlet mercury decreased immediately after NaOCl was
injected and increased shortly after. Therefore, it is probable that severe chlorine stripping was
occurring. With intermediate pH, Cl2 desorption was experienced only at low NaOCl injections,
such as 2.6xlO"6 M total NaOCl. After each injection of NaOCl, the outlet Hg concentration
decreased then increased to a steady-state value which was less than the inlet Hg concentration.
With high NaOCl injection, solution pH was always higher than intermediate pH. So this was the
same situation as that of pH > 9 and no chlorine desorption was experienced. Due to the
complicated chlorine desorption process, it is possible that the free chlorine concentration estimated
by equation (10) overpredicted the actual free chlorine concentration, especially at the gas -liquid
interface.
e . ,
S3 E
'S ?
(S
* S
B V
N O
=3 1
o
Z
io-5
io-7
1ft-9
!' o ป 9.0-11.1 1 " "1 '1 "1 "
: o 6.3-8.0 :
a 1.0-5.3 no n
r ^ * ^* .8 g " ^
: o^ o" " :
r o c -=
: o -
I Mimil I UJJllJ I nihn! l lll/ml _l_UlllJ lllnnl irm innml i ,,l nnnnl J , ninni i MMn|
10
-18
1Q-
14
10
-10
Activity of CI2 (M)
Figure 5. Hg absorption in NaOCl-NaCl at 25 and 55ฐC. Low and intermediate pH
results are all with Cl" > 0. 1 M. High pH data at 25ฐC included some low
to intermediate pH results with low Cl". All data were those of without
external HgCl2 injection. Data points for low and intermediate pH
represent the maximum fluxes associated with each injection of NaOCl.
At low and intermediate pH, NaOCl injection resulted in Cl2 desorption, with lower than expected
Hg flux. High NaOCl injection gave results near or at gas phase control, as shown in figure 6.
There are even several points that exceeded the gas phase control line. This may have been caused
-------
by additional reaction of mercury with chlorine in the gas phase catalyzed by the Teflon surface.
Because of C\2 desorption, the pH effects, and the high reactivity of C\2 with Hg, both in the
aqueous solution and gas phase, the kinetics was complicated at intermediate and low pH with high
ci-
*
e
N
o
in-4
in-5
in-6
io-7
10'8
in-9
- k . r-n n :
- *Jr * c "u -
: @ฐ ^ -
O A
r o n -
: ฐ :
^ o 9.0-11.1 '_
\ A A 6.3-8.0 D =
: a 1.0-5.3 :
10-18 jQ-16 lfl-14 jQ-12 lfl-10 jQ-8 JQ-ซ
Activity of C12 (M)
Figure 6. The effect of gas phase control on Hg absorption in NaOCl-NaCl at 25
and 55ฐC. Data points for low and intermediate pH represent the
maximum fluxes associated with each injection of NaOCl.
Gas Phase Reaction of Elemental Mercury and Chlorine
Due to chlorine desorption at Cl~ > 0.1 M and pH < 9, kinetic information was difficult to obtain
under these conditions. In order to understand this phenomenon, various amounts of chlorine
were added to elemental Hg in the gas phase. It was estimated from previous experiments that the
desorbed C\z during Hg absorption in NaOCl would not exceed 1 ppm. Thus it was assumed that
adding a fixed amount of C\2 greater than 1 ppm to the gas phase would minimize the effect of
chlorine desorption.
Hg-Cl2 was bypassed the reactor, absorbed into pure water and absorbed into 1 M NaCl at
different pH values. Table 4 gives the summarized results.
With the coexistence of chlorine and elemental mercury in the gas phase, less elemental mercury
was detected at the outlet. Furthermore, higher chlorine concentration gave lower outlet elemental
mercury concentration. This was experienced in every experiment. Thus it was concluded that
more chlorine resulted in more reaction between elemental mercury and C\2.
-------
Table 4. Summarized results of chlorine reaction with elemental mercury at
25ฐC. The inlet Hg concentration was 97 ppb
Mode
1
2
3
4
Mode Description
Hg & Cl2 bypassed the reactor and
connected directly to Hg analyzer
Hg & C\2 bypassed the reactor
only Hg absorbed in H2O
C\2 added after the reactor
Hg & Cla absorbed in H2O
Estimated Teflon
Surface Area
(m2)
0.0025
0.019
0.034
0.034
Inlet Cl2
(ppm)
0
5
11
38
0
5
30
0
1.5
10
0
1
10
Outlet Hg
(ppb)
97
97
89
17
97
14
1-5
86
85
81
87
71
9-14
At the same inlet C\2 concentration, the degree of surface contact of the Hg-Cl2 gas mixture affects
the results. It was shown that more solid surface area gave less elemental mercury in the outlet.
When Hg-Cl2 bypassed the reactor and Teflon tubes upstream of the analyzer and flowed directly
to the mercury analyzer, the Hg-Cl2 was only exposed to the surface of the gas blending tube and
of the analyzer. This is the first mode of operation in table 4. It gave the least amount of surface
area exposure. Comparison of this mode to the second mode of normal bypassing shows that
solid surface area, even with Teflon coated reactor surface and Teflon tubes, contributes to more
reaction between elemental mercury and chlorine.
The third mode in table 4 indicates that only Hg absorbed into water. Chlorine was added to the
gas stream after the reactor. Under this mode of operation, chlorine was not exposed to a lot of
moisture. The less amount of Hg removal was expected.
The fourth mode in table 4 is Hg and chlorine absorbed in water. In this mode of operation, Hg
was blended with chlorine before the reactor and absorbed into water with all the normal tubing
system. In this mode, more mercury loss was observed. This indicates that moisture contributes
positively to Hg reaction with Cl2- It is also expected that chlorine react with Hg quite readily at
the water surface.
The history of the experiments also seems to influence the results. When a very high level of
chlorine was introduced at the beginning of the experiment, it tended to give more reaction between
mercury and chlorine even if a very small amount of chlorine was present later. This indicates that
once the reactor or surface was coated with high levels of the reaction product of chlorine and
mercury, more reaction could be expected.
Acknowledgments
This study was supported by Electric Power Research Institute Contract RP 3470-02.
-------
Notation
A = gas - liquid contact area (m2)
DHg-HzO = liquid film diffusion coefficient of Hg (m2 sec'1)
HHป = Henry's constant of Hg (atm M'1)
[Hgli = Hg concentration at the liquid side interface (M)
KI or K2 = equilibrium constant
k2 = second order rate constant (M~l sec'*)
kg = gas film mass transfer coefficient (mole sec-1 atar1 nr2)
Kg' = normalized Hg flux, 5^- (mole s"1 arm"1 nr2)
nHgi
kฐi, Hg = physical liquid film mass transfer coefficient of Hg (m sec"1)
NHS = Hg flux (mole sec'1 m"2)
Pflgb = partial pressure of Hg in the bulk gas phase (atm)
PRgi = partial pressure of Hg at the gas side interface (atm)
PHgin = partial pressure of Hg in the inlet gas (atm)
T = temperature (ฐK)
aa_ = activity of Cl-(M)
da = activity of aqueous C\2 (M)
aH. = activity of H+ (M)
ama= activity of HOC1 (M)
a = activity of OCJ- (M)
Ycr = activity coefficient of Cl~
Ya2 = activity coefficient of aqueous C\i
YHOCI~ activity coefficient of HOC1
Yfiaa = activity coefficient of NaCl
Yoa- ~ acuvity coefficient of OC1"
References
1. G. A. Parks, and R. E. Baker. Assigned to Mountain Copper Company of California,
Martinez, California. "Mercury process." U. S. Patent 3,476,552. November 4,1969.
2. X. T. Nguyen. Assigned to Domtar Inc., Montreal, Canada. "Process for Mercury
Removal." U.S. Patent 4,160,730. July 10, 1979.
3. S. Nene, and V. C. Rane. "Kinetics of the Absorption of Mercury." Indian J. Technol.,
19(1), 20-5 (1981).
4. A. K. Medhekar, M. Rokni, D. W. Trainor, and J. H. Jacob. "Surface Catalyzed Reaction
of Hg + C12." Chem. Phys. Lett., 65(3), 600 (1979).
5. C . S. McCannon, Jr., and J. Woodfin. "An Evaluation of a Passive Monitor for Mercury
Vapor." Am. Ind. Hyg. Assoc. J., 38, 378 (1977).
6. R. Menke, and G. Wallis. "Detection of Mercury in Air in the Presence of Chlorine and
Water Vapor." Amer. Ind. Hyg. Assoc. J., 41, 120 (1980).
-------
7. L. L. Zhao, and G. T. Rochelle. "Hg Absorption in Aqueous Permanganate." AIChEJ.,
42(12), 3559-62 (1996).
8. L.L.Zhao. "Mercury Absorption in Aqueous Solutions." Ph.D. Dissertation, Department
of Chemical Engineering, The University of Texas at Austin, May 1997.
9. Lagowski. "Bleach Analysis by Redox Titration." Experiences in Chemistry, 124-6,
McGraw-Hill, Inc., 1995.
10. R. A. Robinson, and H. S. Earned. "Some Aspects of the Thermodynamics of Strong
Electrolytes from Electromotive Force and Vapor Pressure Measurements." Chem. Rev.,
28, 419.
11. V. M. M. Lobo, and J. L. Quaresma. Handbook of Electrolyte Solutions. Elsevier
Science Publishing Company, New York, 1989.
12. P. W. Atkins. Physical Chemistry, 4th Ed. W. H. Freeman and Company, New York,
1990.
13. R. E. Connick, and Y.Chia. "The Hydrolysis of Chlorine and its Variation with
Temperature." J. Am. Chem. Soc., 81, 1280 (1959).
14. P. V. Danckwerts. Gas-Liquid Reactions. McGraw Hill Book Company, New York,
1970.
15. L. L. Zhao, and G. T. Rochelle. "Mercury Absorption in Aqueous Oxidants Catalyzed by
Mercury(II)." Submitted to Ind. Eng. Chem. Res. (1997).
-------
MERCURY EMISSIONS CONTROL IN FGD SYSTEMS
Kevin E. Redinger
Amy P. Evans
Ralph T. Bailey
McDermott Technology, Inc.
1562BeesonSt.
Alliance, OH 44601
Paul S. Nolan
Babcock & Wilcox, a McDermott International Company
20 South Van Buren Ave.
Barberton, OH
Abstract
In cooperation with the U.S. Department of Energy, the Ohio Department of Development's Ohio
Coal Development Office, and Babcock & Wilcox, McDermott Technology, Inc. has characterized
trace element emissions from the combustion of Ohio bituminous coals and control of these
emissions using conventional particulate and SO2 emissions control equipment. In response to
industry concern over potential regulation of mercury emissions from utility boilers, testing in Phase
II of the Advanced Emissions Control Development Program has focused on measurement of the
quantity and species distribution of mercury downstream of the boiler and emissions control
equipment. This paper presents the results of mercury emissions testing on pilot-scale facilities at
the Alliance Research Center including a wet limestone SO2 scrubber, a lime spray dryer, the
condensing heat exchanger based Integrated Flue Gas Treatment system (IFGT) and the Enhanced
Limestone Injection Dry Scrubbing (E-LIDS) system.
Introduction
Under the Clean Air Act Amendments of 1990, the United States Environmental Protection Agency
(US EPA) was mandated to evaluate emissions of hazardous air pollutants (HAPs) from fossil fuel-
fired electric generating units and to provide a summary report to Congress on mercury emissions
sources, controls and health impacts. Field measurements sponsored by the United States
Department of Energy (US DOE) and the Electric Power Research Institute (EPRI) have
characterized HAP emissions from a variety of boiler types and emissions control equipment
configurations. The results have indicated that existing particulate emissions control equipment -
electrostatic precipitators (ESPs) and baghouses - provide high efficiency removal of most of the
trace elements generated by coal combustion. However, for mercury the data revealed that a wide
-------
range of removal efficiencies exist for commercial particulate and SO2 emissions control equipment.
McDermott Technology, Inc. is using pilot testing to evaluate causes of the observed performance
variation and optimize the use of conventional systems to provide near-term solutions for enhanced
control of mercury emissions from coal-fired boilers.
The US EPA, state environmental agencies and regional associations continue to evaluate the need
for regulation of mercury emissions from coal-fired boilers to reduce human exposure to this
persistent, bio-accumulative trace element. Mercury is emitted from coal-fired boilers in very low
concentrations. Based on field sampling at utility sites, uncontrolled mercury emissions from coal
combustion are generally in the range of 5 to 30 ug/dscm, already well below the regulated
emissions limit of 80 p.g/dscm for municipal solid waste (MSW) boilers. Annual mercury emissions
from a coal-fired unit not equipped with S02 emission controls are on the order of one third to one
pound of mercury per MW of generating capacity1. However, as a group, coal-fired boilers represent
a major unregulated source of mercury emissions to the environment. The US EPA and EPRI
estimate that coal-fired utility boilers emit 50 to 55 tons of mercury per year in the U.S.2
A wide range of mercury emissions control performance for wet scrubbers in bituminous coal
applications (0 to 96%) appear in the literature with a number of factors contributing to this
variability2-3-4-5'6. Significant differences in the mercury content of U.S. coals result in a wide range
of mercury concentrations in the flue gas from the boiler. The form or species of mercury (elemental
mercury or an oxidized compound such as HgCl2) in the flue gas is thought to affect flue gas
desulfurization (FGD) system mercury removal efficiency. Mercury speciation in the flue gas is
believed to be influenced by the type of coal fired, with sub-bituminous coals generating a higher
relative proportion of elemental mercury than bituminous coals. EPRI pilot data indicates that at a
flue gas temperature of 300 ฐF, 68% of the total vapor phase mercury was present as elemental
mercury for the sub-bituminous coal compared to 6% as elemental mercury for the specific
bituminous coal evaluated7. The coal chlorine content and ash characteristics may also influence
partitioning between the solid and vapor phases and the mercury species in the vapor phase. The
scrubber spray tower configuration, liquid-to-gas ratio (L/G), and slurry chemistry may also impact
the reported mercury emissions control.
The draft US EPA report. Study of Hazardous Air Pollutant Emissions from Electric Utility Steam
Generating Units - Interim Final Report, presented a median mercury emissions control efficiency
of only 17% for wet scrubbers with a range of 0 to 59% based on sampling at five commercial
plants. The wet scrubber sample population used as the basis for this report does not appear to
reflect mercury emissions control performance by existing commercial FGD units in the eastern U.S.
Only two of these five utility boilers fire bituminous coal. Four of these scrubbers have an open
spray tower design and the fifth is the only U.S. installation of the Chiyoda jet bubbling reactor
system. Three of the five units are designed for 60% SO2 removal or less and include bypassing a
portion of the flue gas around the scrubber for reheating the gas upstream of the stack. All of the
units were designed to operate at an L/G of less than 70 with one unit designed for an L/G of 22.
In current commercial practice, an L/G of 90 to 100 is typical for a limestone forced oxidation FGD
system designed for 90 to 95% S02 removal efficiency.
-------
Underestimating mercury removal in the existing population of FGD systems and the potential for
additional mercury emissions reductions in new FGD installations may result in an over-estimation
of U.S. utility boiler mercury emissions. Wet FGD systems are currently installed on about 25% of
the coal-fired utility generating capacity in the U.S., representing about 15% of the number of coal-
fired units. FGD systems provide a cost-effective, near term mercury emissions control option with
a proven history of commercial operation. For boilers already equipped with FGD systems, the
incremental cost of any mercury removal achieved is zero. The extent of the publicly available
information base concerning the impact of basic wet scrubber design and operating conditions on
mercury emissions control for bituminous coal applications needs to be expanded to provide a
representative sampling of commercial FGD systems.
The variation associated with the reported mercury emissions control efficiency of commercial FGD
systems and the continuing development of mercury speciation measurement methods suggest that
additional research is necessary to better define causes for the observed performance variation and
maximize mercury emissions control performance of FGD systems. In cooperation with the US
DOE and the Ohio Coal Development Office (OCDO) of the Ohio Department of Development,
Babcock & Wilcox (B&W) is evaluating mercury emissions control performance of commercial
FGD systems as well as advanced systems under development by B&W. The Advanced Emissions
Control Development Program (AECDP) is directed toward demonstration of practical, cost-
effective strategies for reducing HAP emissions from coal-fired boilers using conventional
particulate and SO2 control equipment. The Integrated Flue Gas Treatment (IFGT) project is
demonstrating the application of B&W's condensing heat exchanger technology for enhanced boiler
system efficiency and emissions control. The Enhanced-Limestone Injection Dry Scrubbing (E-
LIDS) system has been demonstrated under the Low Emissions Boiler System (LESS)
Development project.
B&W and the project co-sponsors have invested in the 10 MW equivalent Clean Environment
Development Facility (CEDF) at the Alliance Research Center of McDermott Technology, Inc.
(formerly the Research and Development Division of B&W). The CEDF provides a test site for
demonstrating emissions control solutions to the concerns facing the power generation industry. The
lOOxlO6 Btu/hr facility includes a furnace sorbent injection system, an ESP, a spray dryer/baghouse
system and a slipstream wet scrubber permitting testing of a wide array of emissions control options.
The 5 million Btu/hr small boiler simulator (SBS) facility provides additional capacity for
development work at a smaller scale.
Mercury Partitioning and Speciation
The form or species of mercury present in the flue gas impacts the performance of emissions control
equipment. Mercury is generally present either as elemental mercury, Hgฐ, or as oxidized
compounds such as HgCl2 and HgO. Industry experience to date suggests that Hgฐ and HgCl2 are
the dominant species in the flue gas from coal-fired boilers. The oxidized form of mercury is much
more soluble in the aqueous solution present in FGD systems than elemental mercury and is
-------
therefore more likely to be removed from the flue gas. Elemental mercury tends to remain in the
vapor state at the operating temperature of conventional emissions control equipment. A relatively
higher proportion of oxidized mercury present as HgCl2 would be expected to result in higher
removal efficiency in a FGD system.
The flue gas temperature and compounds such as HC1, SO2 and NOX present in the flue gas may
impact the actual forms of mercury present or interfere with the speciation measurement. The
concentrations of these compounds are different at the inlet and outlet of FGD systems and this
difference may also impact the relative speciation measurements.
EPA Method 29 is a validated method for measuring total mercury emissions and is used as a
benchmark for comparison of alternative speciation measurement methods. Much of the early
mercury emissions testing cited in the draft US EPA Mercury Study Report to Congress was
performed using EPA Method 29. However, the method has been shown to report a significant
fraction of the elemental mercury as oxidized mercury8 There is currently no US EPA-validated
method for measuring individual mercury species in flue gas. The Ontario Hydro Method is a
modification of EPA Method 29 in which an alternative reagent is used in the initial impingers to
prevent the oxidation of elemental mercury. This method uses the same basic hardware as EPA
Method 29.
The US DOE and EPRJ are sponsoring efforts to evaluate various measurement techniques for
quantifying the relative amount of elemental and non-elemental or oxidized forms of mercury in the
flue gas8. The on-going work is currently focusing on two impinger train methods - the TRJS buffer
method and the Ontario Hydro Method. The latter has been used to characterize the relative
distribution of mercury species in pilot testing at McDermott Technology, Inc., with EPA Method
29 used as a check on the total mercury. All of the mercury speciation methods continue to evolve
as more field and laboratory experience is obtained. The most recent modification of the Ontario
Hydro Method involves the addition of a KMnO4/H2SO4 solution to the KC1 impingers immediately
following the post-sampling leak check of the impinger train. This stabilizing agent prevents the loss
of mercury from these impingers during the recovery procedure and improves the total mercury
recovery8
Vapor/Particulate Phase Mercury Partitioning
Partitioning of the total mercury emissions between the vapor phase and the particulate phase
measured in pilot tests at the Alliance Research Center is summarized in Table 1. The flue gas was
sampled using the Ontario Hydro Method downstream of the combustion air pre-heater before the
particulate collection equipment. Mercury emissions data for the Ohio 5&6 coal blend was obtained
while firing the 5xl06 Btu/hr Small Boiler Simulator (SBS). The remaining data in Table 1 was
obtained while firing the coals in the lOOxlO6 Btu/hr CEDF. The measured loss-on-ignition (LOI)
at 800 ฐC of fly ash sampled isokinetically from the flue gas stream for each coal is noted.
-------
Table 1 - Partitioning of Mercury Emissions
Coal
Ohio 5&6
Mahoning 7
Ohio6A
Meigs Creek
Flue Gas
Temperature
(ฐF)
333
328
342
358
Total
Mercury
(ug/dscm)
17.6
22.3
20.1
11.2
Vapor
Phase
(%)
88.8
74.4
93.8
95.2
Particulate
Phase
(%)
11.2
25.6
6.2
4.8
Fly Ash
LOI
(%)
2.5
5.7
5.0
1.8
On average, approximately 12% of the total mercury was present on the particulate collected in the
sampling train for the bituminous coals fired. For the narrow range of relatively low LOI values in
these tests, the distribution of mercury between the vapor phase and the particulate did not appear
to be strongly correlated with fly ash LOI.
Vapor Phase Mercury Speciation
The distributions of vapor phase mercury species in the flue gas at the air pre-heater outlet were
measured for the four coals presented in Table 1 at the indicated flue gas temperatures. The
speciation measurements using the Ontario Hydro Method are summarized in Figure 1. For these
bituminous coals, the vapor phase mercury is primarily present as oxidized species following the
boiler, upstream of the particulate emissions control equipment.
Wet FGD Mercury Emissions Control
Wet FGD Pilot System
The pilot wet scrubber system includes the absorber tower, a slurry recirculation tank, a reagent feed
system, and a mist eliminator wash system. The 50 ft high by 2 ft diameter absorber tower is
constructed of plexiglass to permit visual observation of the slurry sprays. Pre-pulverized limestone
is mixed with make-up water in the reagent feed tank. The solid content of the recirculating slurry
is maintained at 12 to 15%. To achieve the desired L/G, any combination of four levels of single
spray nozzles may be used. The pilot is equipped with a removable gas flow distribution plate to
simulate both tray tower and open spray tower scrubber designs. An air sparger ring in the bottom
of the recirculation tank is used for forced oxidation operation. Spent slurry from the scrubber is
dewatered using a hydroclone circuit. The hydroclone underflow is discharged to settling tanks
where the solids settle out and water is decanted to the clarified recycle water tank for re-use in the
-------
100 -p
Oxidized Mercury Species
Elemental Mercury
Ohio 5&6
(CEDF)
Ohio 5&6
(SBS)
Ohio 6A
(CEDF)
Meigs Creek
(CEDF)
Figure 1 - Mercury Speciation Downstream of Air-Preheater
Note: An early version of the Ontario Hydro Method which did not include the addition of a preservative to the initial impingers
immediately after sampling was used in the Ohio 5&6 CEDF test series noted in Figure 1 This sampling technique was shown to
result in some loss of oxidized mercury during recovery of the sample train resulting in a high bias of the reported elemental fraction11.
The other tests were all completed with KMnO4 preservation of the KC1 impingers which has been shown to improve the retention
of mercury absorbed from the flue gas by the Ontario Hydro Method sampling train9.
scrubber. A variable speed ID fan located downstream of the scrubber is used to control the gas flow
rate through the scrubber. Typical scrubber operating conditions are summarized in Table 2. The
pilot scrubber was run at a higher oxidation air stoichiometry than a commercial unit to maintain the
desired level of near complete oxidation because of the limited available height in the recirculation
tank.
Initial tests in 1996 with an Ohio 5&6 coal blend showed mercury emissions were reduced to an
average of 1.0 ug/dscm across the combined ESP/wet scrubber system representing a total reduction
of 93% from the uncontrolled boiler emissions. The scrubber was operated at an L/G of 90 gal/1000
acf. Three AECDP pilot test programs have been performed to characterize the mercury emissions
control performance of wet scrubbers over a range of operating conditions for several coals.
Following a brief series of tests to demonstrate that variation of scrubber operating conditions can
impact mercury removal efficiency, an extensive program to characterize the impact of key scrubber
design and operating parameters on mercury emissions control was completed. These tests covered
the range of operating conditions reported in the field measurements summary used by EPA as a
basis for the draft Mercury Study Report to Congress. The impact of inlet vapor phase mercury
speciation on scrubber mercury emissions control performance was evaluated in the third test series.
-------
Table 2 - Wet Scrubber Pilot Operating Parameters for Mercury Emissions Control Testing
Operating Parameter
Inlet Flue Gas Flow (acfm)
Nominal Slurry pH
Nominal L/G (gal/ 1000 acf)
Slurry Spray Flux (gpm/ft2)
Oxidation Air Stoichiometry
(mol O,/mol SO2 absorbed)
Range of Operation
2000 to 3000
5 to 6
35 to 130
20 to 70
Oto8
Nominal Conditions
for Speciation
Impact Comparison
2050
5.4
125
67
6
Impact of Absorber Configuration
Most of the existing U.S. wet FGD capacity may be classified as open spray tower or tray tower
designs. Packed towers and venturi scrubbers represent smaller segments of the market. B&W
markets the tray tower absorber design for controlling utility SO2 emissions and has approximately
27,000 MW of wet FGD systems installed or under contract.
The pilot scrubber was operated as both a tray tower and an open spray tower downstream of the
baghouse while firing the Ohio 5&6 coal blend in the SBS. Scrubber operations covered a wide
range of slurry spray flux rates with a common tower velocity representative of conventional
commercial scrubber operation. Operation with the gas flow distribution tray installed enhanced
both SO2 and mercury emissions control over a wide L/G range as illustrated in Figures 3 and 4. The
error bars shown represent the range of the triplicate measurements for each set of operating
conditions. For all of the tests presented in Figures 3 and 4, the oxidation air Stoichiometry was
greater than 5 mol O2/mol of SO2 absorbed to maintain near complete oxidation and the absorber
slurry pH was maintained at 5.4 to 5.5. Additional tests to compare the tray tower and open spray
tower configurations at nominal scrubber operating pHs of 5.0 and 5.9 showed comparable relative
performance to that presented in Figures 3 and 4.
The tray tower configuration provided more consistent SO2 and mercury emissions control than the
open spray tower over the two week test period. The tray significantly improved mercury emissions
control at the lower L/G operating condition. At a nominal L/G of 100 gal/1000 acf, the average tray
tower mercury emissions were 38% lower than the average measured for the open spray tower on
a ug/dscm basis.
-------
A
si?
.2 =
ซ 'o
E ฃ
HI LU
02
w ฃ
o
o
100
90 .
80 .
70 -
60 .
50 -
40 .
30
D
D i
T I
r^i
J_
f
x D Tray Tower
1 A Open Tower
0 50 100
UG Ratio (gal/1000 acf)
150
Figure 3 - Impact of Absorber Tower Configuration on SO, Emissions Control
100
ซ _ 90 -
O 5?
E =
ill .2 70 .
o UJ 60 .
o "o
E * 50-
3 ฐ
ฐ ฐ 40.
30 .
n
r, D "T 1
it
1
A
Q Tray Tower
A Open Tower
0 50 100 150
UG Ratio (gal/1000 acf)
Figure 4 - Impact of Absorber Tower Configuration on Mercury Emissions Control
-------
Impact of Scrubber Operation
The influence of slurry pH and L/G on mercury emissions control was evaluated for both scrubber
configurations while firing the Ohio 5&6 coal blend in the SBS. Figure 5 presents the impact of
LSFO scrubber operation on total (vapor phase and particulate phase) mercury emissions for the tray
tower. The superficial flue gas velocity was maintained at a steady value and the slurry spray flux
was varied to obtain a range of L/G operating conditions. Total mercury concentration downstream
of the baghouse at the scrubber inlet averaged 14.8 u.g/dscm. The distribution of mercury species
at the scrubber inlet was 94% oxidized species and 6% elemental mercury as measured using the
Ontario Hydro Method.
g
o
II)
E
ฃ
3
0)
ฃ
|
4.0 . pH = 5
3.5.
3.0 .
ฃ2.5.
|2-0.
3. "
1.0 .
0.5.
0.0 .
J a PH = 5.4
A pH = 5.9
: I .
5
D
E ' '
TL
0 50 100 150
L/G Ratio (gal/1000 act)
Figure 5 - Impact of Scrubber Operating Conditions on Total Mercury Emissions - Tray Tower
following Baghouse for Ohio 5&6 Coal Blend
The vapor phase oxidized and elemental mercury emissions for each test condition in Figure 5 are
presented in Figures 6 and 7 for the tray tower configuration. Emissions of oxidized mercury were
reduced as the L/G was increased. The slurry pH did not appear to have a significant impact on
oxidized mercury emissions. Elemental mercury emissions following the scrubber remained fairly
consistent over the range of operating conditions and were approximately the same as the inlet
elemental mercury concentration shown as the dashed line in Figure 7. The relatively higher
elemental mercury emissions at the low pH, low L/G condition were also observed in an earlier test
series.
-------
to
o 3.5 --
.2 3.0..
LU -ง-2.5 ..
||2-ฐ"
ป 5* 1 e
^ 3 I .O - -
T3
V
N
;o
'x
O 0.0
1.0 --
0.5 ..
Average Inlet Oxidized Mercury
14.0 ฑ 2.1 ug/dscm
D pH = 5.4
A pH = 5.9
-+-
-+
50 100
L/G Ratio (gal/1000 acf)
150
Figure 6 - Vapor Phase Oxidized Mercury Emissions - Tray Tower
'ฃ
HI
ฃ
u
OJ
15
'c
o>
u
HI
Average Inlet Elemental Mercury
4-0 T 0-8 ฑ 0.3 ug/dscm
3.5 .
3.0 .
-ฃ-2.5 .
E
u
to o 0
Hi -5 -
1.0 .
0.5 -
0.0 -
T nH ~ 5
P ~
li DpH = 5.4
1 4 pH = 5.9
2 T T
f r
0 50 100 150
L/G Ratio (gal/1000 acf)
Figure 7 - Vapor Phase Elemental Mercury Emissions - Tray Tower
With the open tower configuration, higher total mercury emissions were observed at each
combination of L/G and pH relative to the tray tower configuration. Both higher oxidized and
elemental mercury emissions contributed to the higher total mercury emissions. Over the practical
-------
operating pH range evaluated, slurry pH did not have a significant impact on mercury emissions at
the higher L/G operating condition for either tower configuration. There is an indication that
mercury emissions may be more sensitive to pH at the lowest L/G operating condition with
elemental and therefore total mercury emissions increasing as the slurry pH was reduced to 5.
Inlet Mercury Speciation
Impact of Paniculate Control System on Vapor Phase Mercury Speciation. The paniculate
emissions control equipment upstream of the FGD system may impact the Speciation of mercury in
the flue gas. Emissions measurements by B&W following a baghouse treating a flue gas slipstream
and an ESP processing the remaining bulk of the gas flow indicate a marked difference in mercury
Speciation. The fraction of the total vapor phase mercury measured as elemental mercury upstream
and downstream of the particulate control equipment using the Ontario Hydro Method for three coals
is presented in Figure 8. Mercury emissions were measured with the Ohio 5&6 coal blend fired in
two different test boilers, the CEDF and the Small Boiler Simulator (SBS). The baghouse operating
temperature was generally 50ฐF lower than the ESP operating temperature as a result of heat loss
from the slipstream flue work. The baghouse was operated with Gore-Tex membrane fabric bags
installed for each of the test programs.
40 -r
rjAir Heater Outlet
B ESP Outlet
Q Baghouse Outlet
Ohio 5&6
(CEDF)
Ohio 5&6
(SBS)
Ohio 6A
(CEDF)
Meigs Creek
(CEDF)
Figure 8 - Impact of Particulate Control Equipment on Vapor Phase Mercury Speciation
-------
Figure 8 suggests that the fly ash may have a catalytic impact on the conversion of elemental
mercury to oxidized mercury species as the flue gas passes through the ash filter cake in the
baghouse. Significant species transformation was not generally observed across the ESP. The total
vapor phase mercury concentrations were comparable before and after the particulate collection
devices. In the U.S., precipitators are used for particulate collection upstream of the majority of
commercial wet FGD systems.
Impact ofSpeciation on FGD Mercury Emissions ControL The impact of the mercury species
distribution on wet scrubber mercury removal was measured in a series of tests with Ohio
bituminous coals. The scrubber was configured as a tray tower and operated at the same conditions
for each coal. Nominal scrubber operating conditions included a slurry spray flux of 67 gpm/ft2,
a slurry pH of 5.4 and an oxidation air stoichiometry of 5 to 6 mol O2 / mol SO, absorbed. The
impact of the fraction of vapor phase mercury present as oxidized mercury at the scrubber inlet on
total mercury emissions control is illustrated in Figure 9. For the Ohio 6A and Meigs Creek coals,
the scrubber was operated downstream of the ESP first, and then downstream of the baghouse for
each coal to provide two levels of oxidized mercury to the scrubber. Only the baghouse was used
for the Ohio 5&6 coal blend test data shown in Figure 9.
o
o
O
co S?
c
o >,
VI fZ
ซ g
E o
HJ it
3
a
s
100 T BOhio 6A
80 .
60 .
40 .
20 .
0 .
D Meigs Creek
A Ohio 5&6
' i
0.75 0.80 0.85
* * 5
-r
I
T
5
i i
i i
0.90 0.95
I
1
1.00
Fraction of Vapor Phase Mercury
as Oxidized Species at Scrubber Inlet
Figure 9- Impact of Inlet Speciation on FGD Mercury Emissions Control - Tray Tower
Mercury emissions control across the scrubber for the same coal was significantly different if the
ESP or baghouse was used for particulate emissions control. The difference in scrubber performance
can not be completely attributed to the higher fraction of elemental mercury in the vapor phase at
the scrubber inlet following the ESP relative to the baghouse. With the ESP in service, emissions
of elemental mercury from the scrubber were greater than the level of elemental mercury at the
-------
scrubber inlet for each of the individual measurements which make up the average mercury
emissions control efficiencies presented in Figure 9. In the limestone forced oxidation system,
control of elemental mercury emissions appears to be related to the scrubber chemistry.
A higher concentration of elemental mercury in the flue gas has been measured at the scrubber outlet
than at the inlet over a range of operating conditions for different coals. This phenomenon may
reflect mercury reactions in the wet scrubber or result from reactions in the sample train impinger
solutions. The apparent increase in the concentration of elemental mercury across the scrubber
suggests that oxidized mercury absorbed by the scrubbing reagent may be reduced and off-gassed
as elemental mercury. This explanation assumes that the speciation measurements at the scrubber
inlet and outlet adequately reflect the actual speciation at these two locations. The US DOE and
EPRI are continuing research into development of alternative mercury speciation methods which
address interferences from other flue gas constituents such as SO2 and provide total mercury
measurements comparable to the validated baseline, EPA Method 29. This on-going work is
important to the continued evaluation of mercury emissions control technologies.
Spray Dryer FGD System
The CEDF spray drye^aghouse system (dry scrubber) processes the full 30,000 acfm flow of flue
gas from the boiler. The once-through lime spray dryer is equipped with a dual-fluid B&W
Durajet atomizer. Flue gas enters the vertical downflow scrubber at approximately 300 to 350ฐF
and is cooled by evaporation of the water in the lime slurry. Evaporation is sufficient to produce a
dry mixture of fly ash, SO2 reaction products and unreacted lime at the baghouse inlet. The absorber
outlet temperature is controlled to the desired baghouse operating temperature which is typically in
the range of 160 to 230 ฐF. SO2 removal occurs in the spray dryer and in the baghouse.
Mercury emissions at the CEDF dry scrubber inlet and outlet have been measured using the Ontario
Hydro Method for two Ohio bituminous coals fired in the CEDF. Comparable mercury emissions
control was observed in both test programs:
When firing the Ohio 5&6 coal blend, total mercury emissions were reduced by 63%
with the dry scrubber operating at a flue gas outlet temperature of 165 ฐF to maintain
SO2 emissions below 1.0 Ib SO2/106 Btu (82% SO2 emissions reduction). Mercury
speciation measurements at the inlet indicated 73% of the vapor phase
mercury was present as oxidized species.
In testing with the Mahoning #7 coal, total mercury emissions were reduced by an
average of 64%. The dry scrubber outlet temperature varied from 205 to 230ฐF for
these tests and the SO2 removal efficiency averaged 68%. Oxidized mercury species
represented 68% of the total mercury at the scrubber inlet.
-------
Niro-U.S.A. has characterized mercury emissions control at several full-scale spray dryer
installations on bituminous coal-fired boilers in the U.S.10 The observed reduction in mercury
emissions ranged from 55 to 96% and appeared to be correlated with the chlorine content of the coal.
Higher removal efficiencies were observed at higher coal chlorine levels. Niro reported significantly
lower mercury emissions reduction efficiency for sub-bituminous coal applications with a range of
6 to 23% and an average of 15%'ฐ
Advanced FGD Systems
McDermott Technology, Inc. continues to develop advanced systems for controlling emissions of
SO2, HC1 and trace elements from coal-fired boilers. The Integrated Flue Gas Treatment system
combines enhanced heat recovery and emissions control in a single process unit. The Enhanced-
Limestone Injection Dry Scrubbing process has been developed as part of B&W's Low Emissions
Boiler system (LESS) under a US DOE project.
Integrated Flue Gas Treatment (IFGT)
B&W is developing the IFGT process for recovering waste heat from fossil fuel combustion flue gas
and reducing emissions of SO2, SO3, fine particulates and trace elements including mercury" A
condensing heat exchanger comprised of a two-pass, counter-flow, shell and tube heat exchanger is
the basis for the IFGT system. Hot flue gas enters at the top and flows downward through the first
cooling stage, across a horizontal transition section and upward through the second cooling stage.
An alkali reagent is sprayed downward from the top of the second stage for acid gas emissions
control. High efficiency SO2 emissions control of 97% has been demonstrated in pilot scale tests
resulting in emissions as low as 0.18 Ib SO2/106 Btu when firing a high sulfur bituminous coal.
IFGT mercury removal efficiency has been evaluated in 5x106 Btu/hr pilot tests at the Alliance
Research Center for a number of coals under work co-sponsored by the US DOE, OCDO, EPRI and
B&W. Total mercury removal efficiency for the IFGT has generally ranged from 31% to 71%. In
separate tests firing eastern bituminous coals and a Powder River Basin sub-bituminous coal, control
of emissions of oxidized mercury species in the flue gas ranged from 75 to 85% removal. Negligible
removal of elemental mercury has been observed. The IFGT reduced particulate phase mercury
emissions by 50 to 80% from the low concentrations (generally less than 1 (ag/dscm) at the system
inlet.
Enhanced Limestone Injection Dry Scrubbing (E-LIDS)
The E-LIDS process combines furnace limestone injection with dry scrubbing to achieve high
efficiency SO2 particulate, and trace element emissions control. Dry, pulverized limestone is
injected into the upper furnace region of the boiler. The limestone is calcined to lime and a portion
-------
of the sorbent reacts with SO2 in the flue gas. The flue gas passes through a particulate collector to
remove some of the solids from the gas stream ahead of the dry scrubber. The solids are mixed with
material collected in the baghouse to produce the SO2 scrubbing reagent for the spray dryer.
Application of the E-LIDS system when firing an Ohio bituminous coal in the CEDF has shown
very efficient emissions control performance. SO2 emissions generated from firing the nominal 3%
sulfur coal were reduced by more than 99% to less than 0.10 Ibs SO2/106 Btu. Total mercury
emissions were reduced from an uncontrolled level of 17.6 (ig/dscm to less than 0.2 ug/dscm for an
average total removal efficiency of greater than 95% from the as-fired coal mercury content12 The
measured performance confirmed earlier results obtained in the 5xl06 Btu/hr SBS facility. Mercury
measurements upstream of the dry scrubber indicated both the limestone injection and operation of
the spray drye^aghouse system at a close approach to the saturation temperature contributed to the
observed total mercury emissions reduction. The furnace limestone injection alone reduced mercury
emissions to an average of 3.1 ng/dscm.
FGD Systems Mercury Emissions Control Summary
Pilot testing at McDermott Technology, Inc. has demonstrated that high efficiency mercury
emissions control can be achieved with conventional FGD systems and new FGD systems under
development. Figure 10 presents a comparison of SO2 and mercury emissions control when similar
Ohio 5&6 coal blends were fired in facilities at the Alliance Research Center for 4 different FGD
systems.
Wet scrubber mercury removal efficiencies measured over a wide variety of operating conditions
for several bituminous coals in the AECDP pilot tests are consistent with that reported for
commercial installations and other pilot operations13 The wet scrubber FGD system research
completed to date has demonstrated that many factors impact the overall system mercury emissions
control efficiency. The particulate emissions control upstream of the FGD system as well as the
absorber tower design and operating conditions can have a significant influence on mercury
emissions for a given coal. Although the distribution of mercury species at the scrubber inlet is a
key variable influencing mercury control, it may not be the dominant factor in predicting overall
mercury emissions control efficiency.
Commercial and pilot data indicate that high efficiency mercury emissions control can be achieved
with a wet FGD system. Pilot-scale tests indicate that a tray retrofit of an existing open tower
scrubber may be a cost-effective means of enhancing both SO2 and mercury removal efficiency.
FGD system design and operation impact mercury removal performance. Application of an average
mercury emissions modification factor to predict mercury emissions based on measurements of
mercury in the coal does not differentiate the measured influences of scrubber design and operation
on emissions control. McDermott Technology, Inc. and B&W continue to evaluate various aspects
of wet scrubber design, operation and scrubber chemistry to develop techniques for enhancing
mercury removal in FGD systems.
-------
g%SO2 Reduction
B% Total Mercury Reduction
Wet Dry
Scrubber Scrubber
FGD System
IFGT
E-LIDS
Figure 10 - Comparison of SO, and Mercury Emissions Control Efficiency of FGD Systems
Acknowledgments
Funding support for the air toxics emissions control research has been provided through cost-sharing
agreements among Babcock & Wilcox, the U.S. Department of Energy's Federal Energy Technology
Center and the Ohio Department of Development Ohio Coal Development Office. The guidance and
support of the project managers from the latter organizations, Thomas J. Feeley III of DOE-FETC
and Richard Chu of OCDO, is gratefully acknowledged.
References
1.
2.
3.
4.
R.D. Vidic and J.B. McLaughlin, "Uptake of Elemental Mercury Vapors by Activated
Carbons," Journal of the Air & Waste Management Association, Vol. 46, (1996).
Electric Utility Trace Substances Synthesis Report - Volume 3:Appendix O, Mercury in the
Environment. Palo Alto, CA: Electric Power Research Institute, November, 1994.
TR-104614-V3.
J.G. Noblett, "Control of Air Toxics from Coal-Fired Power Plants Using FGD
Technology," presented at the EPRI Second International Conference on Managing
Hazardous Air Pollutants, Washington, D.C. (July, 1993).
P.V. Bush, E.B. Dismukes, and W.K. Fowler, "Characterizing Mercury Emissions from a
Coal-Fired Power Plant Utilizing a Venturi Wet FGD System," presented at the Eleventh
Annual Coal Preparation, Utilization, and Environmental Control Contractors Conference,
Pittsburgh, PA (July, 1995).
-------
5. R. Meij, "Trace Element Behavior in Coal-Fired Power Plants," Trace Element
Transformations in Coal-Fired Power Systems, Fuel Processing Technology, 1994,
pp 199-217.
6. O. W. Hargrove, "A Study of Toxic Emissions From a Coal-Fired Power Plant Demonstrating
The ICCT CT-121 FGD Project," presented at the Tenth Annual Coal Preparation,
Utilization, and Environmental Control Contractors Conference, Pittsburgh, PA (July,
1994).
7. R. Chang and D. Owens, "Developing Mercury Removal Methods for Power Plants." EPRI
Journal. July/August (1994).
8. A State-of-the-Art Review of Flue Gas Mercury Speciation Methods, Energy &
Environmental Research Center, October, 1996. EPRI Report TR-107080.
9. A.P. Evans and K.D. Nevitt, "Mercury Speciation on a 10 MWe Coal-Fired Boiler
Simulator," Paper No. 97-WP72B.07, presented at the Air & Waste Management
Association's 90th Annual Meeting & Exhibition, Toronto, Ontario, Canada (June 1997).
10. K. Felsvang, et.al., "Control of Air Toxics by Dry FGD Systems," presented at Power-Gen
'92, Orlando, FL (November, 1992).
11. R.T. Bailey, B.J. Jankura, and K.H. Schulze, "Preliminary Results on the Pollutant
Removal Effectiveness of the Condensing Heat Exchanger," presented at the First Joint
DOE-PETC Power & Fuel Systems Contractors' Conference, Pittsburgh, PA (July
1996).
12. D.A. Madden and W.F. Musiol, "Enhanced Limestone Injection Dry Scrubbing (E-LIDS)
Development as Part of B&W's Combustion 2000 LEBS," presented at the 22nd International
Technical Conference on Coal Utilization & Fuel Systems, Clearwater, FL (March 1997).
13. K.E. Redinger and A.P. Evans, "Mercury Speciation and Emissions Control in FGD
Systems," presented at the 22nd International Technical Conference on Coal Utilization &
Fuel Systems, Clearwater, FL (March 1997).
-------
Friday, August 29; 8:00 a.m.
Parallel Session A:
Particulate Control - Engineering Studies
-------
Particulate Control for Year 2000 and Beyond for Power Plants
Willard Howard
Rajendra P. Gaikwad
Air & Water Quality Division
Sargent & Lundy
55 East Monroe Street
Chicago, Illinois 60603
Abstract
As a response to the Clean Air Act Amendments (CAAAs) of 1990, utilities will be faced with, a host of
new emission limitations that will or could have a major impact on their current paniculate control
methodology. A large number of power plants will be switching to lower sulfur coals as part of Title IV
Phase n SO2 rules and these plants will need to upgrade their existing precipitators to continue to meet
the current standards. The Title m of CAAAs of 1990 also identified 189 pollutants as hazardous air
polhitants, also known as air toxics, which have been viewed as risk to human health. Many of these air
toxics exist as particulate matter and a large portion are associated with particles that are less than 10mm
(PMio). EPA's My 1997 revisions to the National Ambient Air Quality Standards (NAAQS) places
more significance on particulate matter less than 2.5mm (PM^s). Although these fine aerosols will
primarily originates as vapor phase gaseous emissions such as SO2 and NOX followed by precipitation as
sulfates and nitrates hi the atmosphere, further testing in the stack gases may identify aerosols actually
created in the power plant and therefore controllable prior to emission. This may result in further
tightening of SO?, SOs, NOx, and particulate regulations for power plants.
With such issues all cresting in the next two years, power plants owners need to understand the technical
options to help them make the tough decisions about how to improve the particulate control at their plant
at the lowest impact on cost of electricity. This paper will discuss the potential impact of the regulations,
the effect on the existing precipitators, and identify the specific range of solutions for plants in need of a
preciphator upgrade or application of other technologies such as new baghouses. The paper will also
present general approaches that will guide the owner to the cost effective solution for their situation to
lower the emissions using various technologies such as addition of fields with SOs conditioning,
baghouses, COHPAC, and wet electrostatic precqutators. Case studies wM be used to demonstrate some
probable outcomes and the economics of increased particulate control
Introduction
As a response to the Clean Air Act Amendments (CAAAs) of 1990, utilities will be faced with new
emission limitations on SC>2 and NOx that could have a major impact on their current particulate control
equipment. Most of the plants affected already have electrostatic precipitators for particulate control
Page 1
-------
A large number of power plants will be switching to lower sulfur coals as part of the CAAA Title IV
Phase n SO2 rules. The lower sulfur coal ash usually does not collect well in a small precipitator so the
existing precipitator will need to be upgraded to continue to meet their current emission limitations. Title
IV also requires plants to lower NOx emissions which is often accomplished by modifications to the
boiler such as retrofitting low NOx burners. These boiler modifications can increase the amount of
unbumed carbon in the ash as well as reduce the size distribution of the fly ash. Both of these changes
can negatively impact a precipitator's performance.
Finally, the CAAA Title DI identifies 189 pollutants as hazardous air pollutants, also known as air toxics,
which have been viewed as a risk to human health. Many of these air toxics exist as particulate matter and
a large portion are associated with particles that are less than 10mm (PMio). EPA's revised places more
significance on particulate matter less than 2.5mm (PM^s). These fine aerosols will primarily originate as
vapor phase gaseous emissions such as 862 and NOx followed by precipitation as sulfates and nitrates in
the atmosphere and therefore will not be controllable with a particulate control device. However, further
testing in the stack gases may identify additional 2.5mm aerosols actually created in the power plant and
therefore controllable prior to emission. This may result in further tightening of particulate regulations for
power plants and place an increased burden on the existing particulate control equipment.
Air Toxics
During the combustion process, trace metals in the coal are transformed into both particles and gases in
flue gas. These trace metals are normally concentrated in the fly ash leaving the boiler with the exception
of few species that remain in the gaseous form. Trace element volatility plays a significant role in their
partitioning in coal-fired systems and the degree of emissions control that can be expected from various
technologies. Individual trace elements are classified into three classes based on their volatility (ref 1).
The classification of a number of these trace elements are shown in Figure 1. Class I trace elements are
the least volatile and are found to be equally distributed between bottom ash and fly ash. These would be
collected in the conventional particulate control devices. Typically, their collection efficiencies will match
the overall particulate collection efficiency. Class n trace elements are somewhat more volatile resulting
in bottom ash depletion and fly ash enrichment as a result of initial vaporization and subsequent
condensation. These would be collected in the conventional particulate control devices with an efficiency
slightly lower than the overall particulate collection efficiency. Class IH trace elements are the most
volatile and Witt not condense under the operating conditions of the conventional power plant and will
only partially be adsorbed on the fly ash particles. These trace elements will have low removal efficiency
in conventional particulate control devices. Due to overlapping in the classification, a number of trace
elements can be categorized in two classes.
Most of the trace elements enriched in fly ash can be removed in a high efficiency electrostatic
precipitator (ESP) or a baghouse. However, most of the currently available technologies employed by
U.S. utilities have much lower efficiency for controlling the particles lower than 2.5 micron. Typically,
for most U.S. coals, collection efficiency of 99.7% to 99.9% will be required to collect greater than 99%
of the 2.5 micron-size particles.
Page 2
-------
Case Studies
This paper evaluates the technologies available to comply with these new and more stringent particulate
control requirements. To demonstrate the retrofit alternatives available to meet these new particulate
emission requirements, two hypothetical case studies are discussed. These two cases are based on
retrofitting technology on a coal fired unit with existing precipitators 150 SCA and 275 SCA SCA is the
collecting plate area in sq. ft. divided by 1000 acfin of gas flow. The 150 SCA precipitator relatively
small The 275 SCA is a moderately sized precipitator. The following scenarios are considered:
Case 1: Existing Precipitator - 275 SCA (9" Plate Spacing)
Conversion to PRB or Appalachian Low Sulfur Coal
To meet 0.03 Ib/MBtu consider the following control options:
> Addition of a field
> Addition of SOs/NHs Conditioning
> Addition of a field and SOs Conditioning
To meet 0.01 Ib/MBtu consider the following control options:
> Replacement with of a Pulse-Jet Baghouse
> Addition of COHPAC
> Addition of Wet ESP
Case 2: Existing Precipitator - 150 SCA (9" Plate Spacing)
Conversion to PRB or Appalachian Low Sulfur Coal
To meet 0.1 Ib/MBtu consider the following control options:
> Addition of a field
> Addition of SCVNHs Conditioning
To meet 0.03 Ib/MBtu consider the following control options:
> Addition of a New Precipitators
> Addition of a fields/SOs Conditioning
To meet 0.01 Ib/MBtu consider the following control options:
> Replacement with of a Pulse-Jet Baghouse
> Addition of COHPAC
> Addition of Wet ESP
In both cases, we assumed that the current coal is medium-to-high sulfur content, 10% ash, and that 80%
of the coal ash becomes fly ash.
The retrofit of these and other technologies have been previously presented (ref 2, 3).
Page 3
-------
General Description
Addition of Field(s) in the direction of gas flow increases the precipitator plate area, increases the
residence time, increases aspect ratio, and improves the existing performance. The number of additional
fields depends on the area available between the ESP and chimney and the desired level of improvement
in the performance of existing precipitator. Adding a field is not as effective if the gas velocity through
ESP is significantly greater than 5 ft/s. When adding one or two fields they could be made a part of the
existing casing. When adding more than two fields a stand alone precipitator in series would be a Kkefy
choice.
SOj/NHj Conditioning is used to enhance precipitator performance by lowering the ash resistivity. A
number of process variations to produce SOs are available including: EPRICON, molten sulfur, granular
sulfur, in-situ SQj production etc. Water vapor and SOs are two components of the flue gas being
adsorbed on the surface of fly ash particle that increase the surface conduction and lower ash resistivity.
The addition of SO3 to the flue gas in the range of 5-15 ppm (by volume) results hi ash that has higher
surface conductance and an additional resistivity of approximately 2 x 1010 ohm-cm (reฃ 4). NHs
conditioning is used for the difficult to condition ashes due to high silica and alumina content. The NH3
makes the ash stickier and more collectable.
Wet precipitators are widely used in industrial applications for collection of fine particles or mists, such
as sulfuric acid mists. Plate-type designs were widely used for industrial applications involving organic
emissions, plastic curing, food processing, printing, textile finishing, and heat treating industries. Tubular
wet precipitators have very high collection efficiency hi submicron region. Unlike dry precipitators, wet
precipitators do not require a rapping mechanism for removing the accumulations on the electrodes. The
separated mists form a liquid film on the vertical collecting electrodes that flows downwards, removing
the solid particles in suspension form. In the case of high loadings, atomizing nozzles are used to
continuously spray water into the precipitator to prevent sludge deposits on the collecting electrodes.
The major advantages of wet ESPs in utility applications lies in the removal of E^SO/i, HC1, micron-size
particles, and other mists contained in the gas.
A large number of wet ESPs have been installed in various industrial applications. However, there is no
significant experience in the utility industry. Recently, Southern Environmental built a large scale wet
ESP (100 MW) at a power plant owned by Northern States Power to control sulfuric acid mist from the
wet scrubber.
Pulse Jet Baghouse (PJBHs) /Compact Hybrid Paniculate Collector (COHPAC) PJBHs are used in
the utility industry world-wide. In the U.S., a number of IPPs have installed PJBHs. Recently, Wisconsin
Electric has installed PJBH on Valley Power plant (2 x 400,000 acfin). COHPAC is a novel, low-cost,
retrofit particulate collection concept developed by EPRI. Recently, EPRI and TU Electric have
conducted full-scale testing of a COHPAC pulse-jet baghouse at Big Brown. The success of this
demonstration has led TU Electric to install COHPAC baghouses treating 100% of the flue gas at their
2X575 MW lignite fired Big Brown station (ref 5). These full scale COHPAC baghouses have been in
successful operation for one year.
Page 4
-------
Case Study - Design Premises
The base design premises were used to retrofit an existing ESP at a greenfield site. A 250-MW unit size
(950,000 acfin gas, SOOT) was used as a base case unit.
Retrofiit with Additional Field/s, SO3 Conditioning, Wet ESP, andPJBH/COHPAC
The scope of the work included the following:
A complete system including:
Additional Field(s)
Additional casing including internal flow devices such as baffles, vanes, division plates, and/or perforated
plates and electrodes, ESP roof, insulation, galleries, penthouses, transformer-rectifier (T-R) sets,
controls, all electrical work within battery limits as appropriate, structural steel supports for ESP
equipment, and any additional foundation requirements.
SOs/NHs Conditioning
A complete SOyTVHi conditioning system including molten sulfur storage tank, SO2 generator, ammonia
storage tank, blower, catalytic converter, injection nozzle assembly, insulation, structural steel support
and foundations and state of the art control system with energy management.
Wet ESP
New wet ESP casing including internal flow devices such as baffles, vanes, division plates, and/or
perforated plates and electrodes. Casing is fabricated from fiberglass reinforced plastic (FRP) with acid
resistant coating. Collecting surfaces for wet ESP were fabricated fromporyvinyl chloride (PVC).
Discharge electrodes for wet ESP were fabricated from acid resistant material, such as C-276 or C-22.
Also included were the insulation, galleries, penthouses, transformer-rectifier (T-R) sets, controls, all
electrical work within battery limits as appropriate. The quench tower for wet ESP was fabricated from
acid-resistant material, with 2 x 100% recirculation pumps, and spray nozzles fabricated from FRP or
equivalent material
PJBH/COHPAC
New PJBH/COHPAC casing included internal flow devices, hoppers, structural frame, inlet and outlet
plenums with gas distribution devices, Ryton fabric filter bags, bag cages, galleries and walkway, access
doors, insulation, dampers with pneumatic operators, bypass ducts, baghouse penthouse roof enclosure,
painting, pulse air supply and distribution system including blower, cooler, and compressors,
instrumentation and control
Common to All
Erections included all services, materials, equipment and facilities required to expedite, ship, route,
receive, unload, store, transport, assemble, protect from weather, and fully erect for commercial
operation.
The balance-of-plant cost: which including demolition of outlet plenum, ductwork, waste handling
system, and a booster fan.
-------
Economic Design Premises
The economic criteria used in this study conforms to EPRTs guidelines. The costs are based on December
1997 dollars with a January 1, 1998, plant start-up.
Capital Cost Estimation. The capital costs were estimated based on the current Sargent & Lundy's
current data base, vendor input on recent jobs, and previous studies performed on particulate control
technologies.
Operating and Maintenance Cost. The operating costs for the particulate collection systems were
separated into fixed and variable components. The incremental fixed and operating costs were developed
for afl retrofit options. The fixed operating costs include O&M labor, filter bags and cages, and
administrative-phis support-labor. A bag life of three years was assumed for the PJBH/COHPAC options.
The maintenance cost was estimated from the process capital cost of subsystems and utility industry s
experience. The cost of bags and cages was linerized over the life of bags and cages.
The variable operating costs are those that depend on the unit's capacity factor and include consumables
such as chemicals, power, and waste disposal It is assumed the waste generated from wet ESP can be
disposed of along with the current bottom or fly ash. However, in some cases, the waste may have to be
sent to waste water treatment plant for further treatment. The cost of such treatment is not included in
this analysis.
Levelized Cost. The levetized cost of the particulate control devices is based on a 30-year plant life.
Other economic parameters include a 9.8% discount rate after tax, 16.5% capital levelized fixed charge
rate, 3% general rate of inflation. The total levelized bus-bar cost is determined by summing the levelized
total capital requirement, the levelized fixed operating cost, and the levelized variable operating cost.
Economic Comparison Of Options
Case 1: Retrofitting Existing 275 SCA Precipitator
Typical low resistivity coals (Illinois coal) will meet the 0.1 Ib/MBru emission level with existing ESP, the
0.03 Ib/MBtu emission level with an addition of a nine-foot field, and the 0.01 Ib/MBtu emission level
with an addition of two, nine-foot fields (ref 2,3). For typical high resistivity coals (PRB or Appalachian
low sulfur), the existing ESP will not be able to meet even 0.1 Ib/MBtu emission level without SOs
conditioning. Addition of two fields and SOs conditioning will be required to achieve the 0.03 Ib/MBtu
emission level and three fields with SOs conditioning will be required to meet the 0.01 Ib/MBtu emission
levels. PJBH, COHPAC, and wet ESP installations will result in 0.01 Ib/MBtu emission.
The capital and levelized costs for various options are presented in Figures 2 and 3 respectively for two
typical coals. It should be noted that the baghouse options include a 10% process contingency and the
wet ESP option includes a 20% process contingency to cover then- applicability at the 0.01 Ib/MBtu.
Figure 2 indicates that for low resistivity coals with emission levels between 0.1 and 0.01 Ib/MBtu, and
for high resistivity coals with emission levels between 0.1 and 0.03 Ib/MBtu, ESP upgrades will be lower
in capital cost than baghouse or wet ESP options. However, in Figure 3, the levelized cost indicates that
only COHPAC option would be an alternative for high resistivity coals required to meet the 0.01 Ib/MBtu
emission level This is primarily because of high operating cost of baghouses due to bag replacement and
auxiliary power. Since the COHPAC is much smaller than conventional PJBHs, the lower operating cost
makes it comparable with the ESP upgrades option.
Page 6
-------
Case 2: Retrofitting Existing 150 SCA Precipitator
A large number of old utility units have precipitators with 150 SCA or lower. It has been found that
typical low resistivity coals (Illinois coal) will meet the 0.1 Ib/MBtu emission level with one additional
field to existing ESP, the 0.03 Ib/MBtu emission level with an addition three fields, and the 0.03 Ib/MBtu
emission level with an addition of four fields (2,3). For typical high resistivity coals (PRB or Appalachian
low sulfur), addition of two fields and SOs conditioning will be able to meet 0.1 Ib/MBtu emission level
Addition of four fields and SOs conditioning will be required to achieve the 0.03 Ib/MBtu emission level
and five fields with SOs conditioning will be required to meet the 0.01 Ib/MBtu emission levels. PJBH,
COHPAC, and wet ESP installations will result in 0.01 Ib/MBtu emission.
The capital and levelized costs for various options are presented in Figures 4 and 5, respectively for two
typical coals. Figure 4 indicates that, for low resistivity coals with emission levels between 0.1 and 0.03
Ib/MBtu and for high resistivity coals with emission level of 0.1 Ib/MBtu, ESP upgrades will be lower hi
capital cost than baghouse or wet ESP options. However in Figure 3, the levelized cost indicates that
only COHPAC option would be an alternative for low and high resistivity coals required to meet the 0.01
Ib/MBtu emission level and baghouse will be an alternative for high resistivity coal It should be noted
that this analysis does not take into account the cost of the outage required to incorporate the retrofit
changes. If the cost of outage is taken into account, then COHPAC and the PJBH options would be more
cost effective than the ESP upgrade options to meet 0.01 Ib/MBtu emission leveL
Conclusions
The options are numerous and heavily dependent on the coal quality and the emission level required
and the species of the p articulate matter being controlled.
For medium to high sulfur coals with low resistivity, upgrading the existing ESP may be the lowest
cost retrofit option to meet emission level between 0.1 and 0.01 Ib/MBtu.
For low sulfur coals with high resistivity, COHPAC and PJBH offer economically attractive
alternatives to ESP upgrades if the emission level is 0.01 Ib/MBtu.
Upgrades with SOs conditioning can offer a significant cost advantage for high resistivity coal
References
1. G. F. Weber et.aL, "A Summary of Utility Elements Emissions Data from the DOE Air Toxics
Study," Presented at EPRI/DOE International Conference on Managing Hazardous and Particulate
Air Pollutants, Toronto, Ontario, Canada (Aug. 16, 1995)
2. R.P. Gaikwad, D.G. Sloat, R. Altman, and R. Chang, "Engineering Evaluation Novel Fine
Particulate Technologies," Presented at EPRI/DOE International Conference on Managing
Hazardous and Particulate Air Pollutants, Toronto, Ontario, Canada (Aug. 16, 1995)
3. R.P. Gaikwad, D.G. Sloat, R. Airman, and R. Chang, "Economic Evaluation of Electrostatic
Precipitator Retrofit Options," Presented at Tenth Particulate Control Symposium and Fifth
International Conference on Electrostatic Precipitation, Washington, DC (April 1993)
Page 7
-------
4. "A Manual on the ESP of Flue Gas Conditioning for ESP Performance Enhancement," EPRICS-
4145, Project 724-2, August 1985.
5. AK Hindocha, B. Brown, andR. Chang, "Commercial Demonstration of COHPAC," Presented
at Tenth Paniculate Control Symposium and Fifth International Conference on Electrostatic
Precipitation, Washington, DC (April, 1993)
Class m -
Class nj As,Pb,Cdetc.
Be,Ba,Cฃ,Co,Nieic.
Class I -
Figure 1: Classification of Trace Elements According to Volatility
Pageg
-------
Figure 2: Case 1 - Capital Cost for Various Emission Limits and Coals
n 1 Iks/mo*,, i-innraih/nnofTi nn rx IWAJD*,, All t*O3IS
2
ป-12-
ง 1ฐ'
5 8-
ฃ ,.
Total Capital
D ro ^ a
Note - Existing ESP internals are replaced
~'
PRB Coal
Illinois No. 6 Coal
'
. ,
1
1
'
~a
en
o-o
en
Qo-o
3-
Figure 3: Case 1 - Levelized Cost for Various Emission Limits and Coals
All Coals
JC ^ "
Levellzed Cost, mil
0 -
D cn -^ ui
Note - Existing ESP internals are replaced
PRB Coal
l
I
Illinois No. 6 Coal
i n
r
I
I
i
i '
-=
is
Page 9
-------
Figure 4: Case 2 - Capital Cost for Various Emission Limits and Coals
16 j
14-
12-
10-
8 -
6 -
4 -
2-
n .
I
1
0.1 Ib/MBtu no.03lb/MBtu Q0.01 Ib/MBtu
Note - hxisting hSH ,
internals are replaced PRB Coal
Illinois No. 6 Coal
I
r-.
E
B
3
o
Figure 5: Case 2 - Levelized Cost for Various Emission Limits and Coals
2.5 -
2 -
1 5-
1 -
0.5-
n -
0.1 Ib/MBtu DOOSIb/MBtu DO 01 Ib/MBtu
Note - bxistmg ESP pRB CQ.
All Coals
>l ^
iiieinaib are lepiaceu i
. i
I
Illinois No. 6 Coal
1
1
1
1
(
~
Sl
Pagelfl
-------
Baghouse Evaluation and Optimization
at the Bonanza Power Plant
Randy L. Merritt
P Vann Bush
Southern Research Institute
2000 Ninth Avenue South
Birmingham, AL 35255
Michael D. Goddard
Deseret Generation & Transmission Co-operative
Bonanza Power Plant
12500 East 25500 South
Vernal, UT 84078
Abstract
Southern Research Institute conducted an evaluation and optimization of the baghouse
system at the Bonanza Power Plant in 1996 with the objective of lowering the
baghouse pressure drop and extending the service life of the bags. At the beginning of
the evaluation, the baghouse was operating with bags that were over 10 years old and
the pressure drop ranged from 6.5 to 8.5 in. H2O at full load conditions. The bags had
attained weights in the range of 150 to 175 pounds.
We found no significant difference in the mass loading from the front to the back of the
baghouse. Therefore, a specific bag cleaning protocol should have a consistent effect
across the entire baghouse. We found that the delivery of reverse-gas for cleaning
was deficient at high boiler loads due to the increased pressure loss across the
reverse-gas fans. Consequently, the bags did not clean as effectively at higher boiler
load conditions. The tensioning springs supporting the filter bags were very
compressed due to the weight of the bags and a loss of resilience of the springs.
Deseret performed manual cleaning of the bags by shaking each bag (removing
approximately 90 to 100 pounds of ash per bag), and replaced the existing conical
springs with new conical springs. The manual cleaning was performed on a
compartment-by-compartment basis from June 10 through August 20. The baghouse
pressure drop was reduced by 2.0 in. H2O due to the manual cleanings.
Following the manual bag cleanings, the delivery of reverse-gas was at consistently
high levels at all boiler load conditions and the effectiveness of the sonic horns was
improved. We evaluated the various cleaning parameters to determine optimum
settings and preferred ranges of operation. We extended the duration of the sonic
horns' operation to 30 seconds and the reverse-gas period to 90 seconds, and have
decreased the frequency of bag cleaning from every 45 minutes to periods as long as 2
to 4 hours (at full boiler load conditions).
1
-------
Over the four months since the manual cleaning, the baghouses have operated at a
stable pressure drop between 4.3 and 5.7 in. H2O at full boiler load. We are optimistic
that the baghouse pressure drop can be controlled and that bag failures will be
maintained at reasonably low levels. We believe that the reduction in the cleaning
frequency should be beneficial in the extension of bag life. With continued success in
the performance of the baghouse operation, we expect that the bags will provide
several more years of service with acceptable pressure drops. It is likely that Bonanza
will achieve one of the longest bag lives in the utility industry.
One of the main benefits resulting from this evaluation program is that there has been
close monitoring and scrutiny of the baghouse performance by personnel at Bonanza
and Southern Research Institute. This increased level of attention given to the day-to-
day baghouse operation has resulted in more stable and efficient performance.
Background
The Bonanza Power Plant, located 35 miles southeast of Vernal, Utah, is owned and
operated by Deseret Generation and Transmission Co-operative. Construction of the
400 MW Unit 1 at Bonanza began in 1981. A reverse-gas fabric filter (baghouse)
system was installed to remove particulate matter from the flue gas stream. A wet
scrubber was installed downstream of the baghouse for the reduction of sulfur dioxide
emissions. Unit 1 was placed into commercial operation in December 1985.
The reverse-gas baghouse system was supplied by Ecolaire Environmental Company.
The baghouse is divided into two casings, each with the ability to be operated
independently. Each casing has 12 compartments, with 450 bags per compartment, for
a total of 10,800 bags. Fabric Filters fiberglass bags with a Teflon-B finish and a
weight density of 14 oz/yd2 were installed in the baghouse.
For the first 10 years of operation, the baghouse experienced a gradual increase in
pressure drop, which is typical of the experience of many electric utilities. Several
actions were taken to control the rising baghouse pressure drop. A change to the BHA
conical spring was made in August 1988 to increase bag tensioning as the bags
became heavier. Four BHA sonic horns were installed in each compartment in
September 1992 to augment reverse-gas cleaning. The utility performed manual bag
cleanings three times during the period of 1988 through 1992. The reduction in
pressure drop achieved from these manual bag cleanings was only temporary. (During
our initial on-site activities in April 1996, the typical baghouse pressure drop was in the
range of 6.5 to 8.5 in. H2O at full boiler load conditions.)
The baghouse at Bonanza has operated with the original bags, with the exception of
those bags replaced due to failure. At the end of 1995, approximately 10.5% of the
original bags had been replaced. Deseret considered a complete rebagging as a
means to lower pressure drop, due to the age of the bags. The bags are approaching
11 years of service, which is longer than the typical experience in the utility baghouse
industry. (In a survey of the electric utility baghouse users that we conducted in 1995,
we found that the average bag life for the 102 utility baghouses was 7.5 years,
although about 20% of the population have achieved more than 10 years of bag life.)
-------
Objectives of the Baghouse Evaluation & Optimization Program
We presented a program to Deseret with the objective of reducing baghouse pressure
drop and extending the service life of the existing bags. Reducing baghouse pressure
drop would allow more margin of operation (with more capacity from the ID fans) and
would yield cost savings due to lower fan horsepower requirements. It was estimated
that a 2 in. H2O reduction in baghouse AP would save the utility about $80,000 per
year. Extending the service life of the bags would allow the bag costs to be prorated
over a longer period. Each year of additional service with the existing bags would save
approximately $100,000 per year.
Even though the filter bags were over 10 years old, we believed that the bags could
provide service for several more years based on the bag failure history, fabric strength
analyses, and the physical appearance of the bags. Therefore, the primary objective of
a baghouse evaluation program would be to reduce the accumulated dustcake, which
would result in a lower baghouse pressure drop.
Deseret decided to proceed with our program of baghouse evaluation and optimization
at the Bonanza Station. The evaluation period began on April 1 and continued through
mid-December 1996.
Performance Criteria
In most utility baghouse control rooms, the only information available to the operator to
gauge baghouse performance is a charting of the baghouse pressure drop. This
charting is often combined with the baghouse opacity (as it is at Bonanza) for the
purpose of identifying compartments with suspected bag failures. In addition to this
charting, the baghouse control and display panel shows the current status of the
cleaning cycle and the positions of the various compartment poppets, dampers, and
other devices. Beyond these necessary operational functions, the control panel
provides minimal information to the operator on how well the bag cleaning system is
performing.
We have devised measurement capabilities and techniques over the past 15 years that
are specific to baghouse evaluations. Each of these measurement techniques provide
an insight into the baghouse operation at different scales (entire baghouse casing,
specific compartments, or an individual bag). Figure 1 shows a schematic of the
baghouse system, and the locations of the various measurement devices.
Boiler Load
Load (in gross MW) was monitored continuously so that the baghouse performance
could be correlated with boiler operations.
Baghouse AP
This parameter provides one of the best overall measures of performance. Changes in
the baghouse operating or cleaning protocol which have a significant effect will be
reflected in the level of baghouse AP. We installed a pressure transducer on the
baghouse pressure taps for the 1-2 baghouse casing.
-------
L: Baghouse (Flange) DP
TS: TubesheetDP
OP: Outlet Poppet
(To Scrubber)
t
F1
E1
D1
C1
B1
A1
F2
E2
02
C2
B2
A2
(1-1 Baghouse)
Baghouse/Scrubber /- Load
Control Room
Computer }
(To Scrubber)
t
F1
E1
C1
B1
FL
E2
C2
B2
- TS
OP
(1-2 Baghouse)
t
(From Boiler)
t
(From Boiler)
Figure 1 Layout of Measurement Locations at the Bonanza Plant
-------
Compartment TubesheetAPs
During a compartment's reverse-gas period, the pressure drop across the tubesheet
gives a direct measure of the cleaning magnitude.
Compartment Relative Drag
Relative Drag is a performance parameter that normalizes the tubesheet AP of a
specific compartment by gas flowrate through the compartment. Drag is the measure
of the permeability of the filter and dustcake. Drag is calculated as the tubesheet AP
divided by the square root of the compartment outlet poppet AP. The poppet AP was
derived by a combination of the baghouse and tubesheet AP measurements and was
monitored by a separate pressure transducer for each monitored compartment. A
schematic of the measurement technique is shown in Figure 2.
Baghouse ( DP
Figure 2. Locations of Baghouse Pressure Measurements
This measurement technique is an improvement over baghouse and tubesheet AP
measurements, although it is usually limited to applications of moderate-to-high boiler
load conditions. Measurement of drag at Bonanza provided credible results at all boiler
load conditions for compartment 1-2 A2. An example of how the drag measurement
"normalizes1' the performance for a period of varying boiler load conditions is shown in
Figure 3. Relative drag is shown to be fairly stable while the boiler load and baghouse
AP are changing over a couple of hours.
Individual Bag Weights
We recently developed a continuous bag weight monitor that can be installed on an
individual filter bag. We believe that the knowledge of bag weight (before and after
cleaning) is an important indicator of performance, since it gives direct evidence of the
effectiveness of the cleaning process.
For measurement of bag weight, the bag was equipped with load cells at the top and
bottom of the bag. The bag weight is calculated as the difference in tension at these
two points. This measurement technique provides reliable data at all boiler operating
conditions. We installed bag weight monitors on two selected bags at Bonanza, one
-------
bag each in compartments 1-2 A2 and 1-2 F2. The selection of these compartments,
at opposite ends of the baghouse, provided information regarding differences in mass
loading and cleaning characteristics with respect to compartment location.
2:00 AM 2:30 AM 3:00 AM 3:30 AM
Figure 3. Behavior of Relative Drag Measurement for September 22, 1996
Data Acquisition & Analysis Techniques
Signal cables from the pressure transducers and bag weight monitors were run to the
scrubber/baghouse control room and connected to the computer data acquisition
system. A boiler load signal and a dedicated phone line were also connected to our
computer system.
A live tabular and graphical summary of the cleaning parameters was displayed on the
computer monitor that could be used by control room personnel. We were able to
communicate with and control the on-site host computer via modem link.
Assessment of Baghouse Operations
Each casing (of 12 compartments) has a separate cleaning system that is operated
independently. Cleaning is prescribed by a AP-initiate protocol and was set to initiate
cleaning when baghouse AP exceeded 4.0 in. H2O, which kept the baghouse in
continuous cleaning for boiler loads above 250 MW. The time between cleaning
consecutive compartments was approximately 4 minutes. Therefore, a complete
cleaning cycle had a duration of approximately 45 minutes. During periods of low boiler
load, the baghouse would typically clean on a 4- to 6-hour interval.
-------
Manual Bag Weights
Measurement of bag weight provides a "report card" of the baghouse performance.
The history of bag weights reveals the effectiveness of the cleaning process. Control
of this parameter is essential to minimize AP, minimize bag failures, and to achieve the
longest service life from a set of bags.
A selection of bags were weighed in each of the designated tests compartments, A2
and F2. Three bags were weighed in compartment A2 on 4/1/96 (average of 155
pounds) and four bags were weighed in compartment F2 on 4/3/96 (average of 166
pounds). The heaviest bag that we weighed was 176 pounds. This level of weight was
higher than had been anticipated, although these weights were not at record levels for
the utility baghouse industry.
Bags had been weighed several times at Bonanza. A record of these bag weights and
other significant events are shown in the timeline in Figure 4.
180
160
140
o
100
80
60
20
0 i
-;ฅ-
~JS-
CO
""C~
Q
(0
c
o
-O-
All bags were hand-shaken
86 87 88 89 90 91 92 93 94 95
Figure 4. History of Bag Weights at Bonanza
97
Mass Loading Profile
The bag weight monitors provide a real-time indicator of mass loading by quantifying
the amount of ash that is added to the bags since the last cleaning period. The bag
weight monitors are uniquely capable of this determination, as there is no other known
sampling method that can provide this measurement directly. The major reason that
the two bag weight monitors were located at opposite ends of the baghouse (in
compartments A2 and F2) was to be able to discern the extent of stratification in ash
loading across the baghouse. There is evidence that at some baghouses the "front"
compartments may receive a heavier mass loading than those compartments furthest
from the baghouse inlet
7
-------
Figure 5 presents the calculated mass loading to the test bags for several days in April
1996. During this period, the boiler was cycled between high and low load each day,
which was typical of the boiler operation of that time period. There was no significant
difference in the mass loadings for the bags in compartments A2 and F2, as
demonstrated by the close agreement at all boiler load conditions.
The mass loading was approximately 2 to 3 pounds per hour per bag at full load, and
was approximately 0.25 to 0.75 pounds per hour at low load (140 MW). Mass loading
to the test bags was proportional to boiler load. This consistent relationship suggests
that there was minimal dropout of ash (in the ducting to the baghouse or in the
baghouse hoppers) at the low boiler load conditions.
1100
1000
900
800
700
600
500
400
300
200
100
CD
4/9
4/10
4/11
4/12
4/13
4/14
Figure 5. Dustcake Added to the Test Bags
It is interesting that no difference in the mass loading was measured from the front to
the back of the baghouse. Consequently, a specific bag cleaning protocol should have
a consistent effect across the entire baghouse. Had a significant difference in the
mass loading been measured, it may have been necessary to design a cleaning
protocol that would have considered the difference in mass loading.
Variation in Bag Cleaning Efficiency
During the first weeks of our on-site evaluation, we discovered several relationships
that led to an important finding. It was fortuitous that the boiler cycled between high
and low load during this period. We observed that the bag weights, that is the residual
weights after cleaning, cycled according to the level of boiler load. The test bags
showed an increase in weight during high load conditions and a decrease during lower
loads, changing as much as 10 pounds per day. Figure 6 shows this relationship of
boiler load and bag weight. Another perspective of this relationship is shown in Figure
7 which shows that there was a net gain in the bag weight at high load, and a net loss
in the bag weight at low load.
-------
155
150
| 145
ฃ 140
f 135
I 130
ro
m 125
120
115
Bag Weight - Before Clean
Bag Weight - After Clean
900
800
700
600
500
400
300
200
100
o
CD
4/9 4/10 4/11 4/12 4/13 4/14 4/15
Figure 6. Relationship between Boiler Load and Bag Weight
2
I 1
I 0
-1
X -2
-3
8
p
-5
ฃ -6
Dustcake Added less
Dustcake Removed
900
800
700
600
500
400
300
200
100
-
I
o
m
4/9
4/10
4/11
4/12
4/13
4/14
4/15
Figure 7. Net Gain / Loss in Dustcake Weight vs. Boiler Load
Of the cleaning cycle parameters that would directly affect the bag cleaning efficiency,
the majority were not very likely to offer varied effects, since they were not subject to
change with load changes. For instance, sounding the horns for 10 seconds would not
act differently at high and low loads. Although the reverse-gas was performed for a
fixed period of 60 seconds, the magnitude of the reverse-gas, measured at the
compartment tubesheet was found to vary according to boiler load.
Figure 8 presents a charting of reverse-gas magnitude (as measured at the
compartment tubesheet) over several days and shows that the level of reverse gas was
measurably less at high load conditions. Figure 9 shows a composite summary of this
relationship for one month of operation and shows a definitive drop-off in reverse-gas
magnitude with increasing boiler loads. The reverse-gas fan at Bonanza was operating
beyond its specified range (with the baghouse operating at the higher pressure drops)
and was unable to deliver a consistent performance with the higher pressure losses
across the fan. Since reverse-gas flowrate is the primary mechanism that provides bag
cleaning, the lower levels of reverse-gas had a significant effect on bag cleaning under
these operating conditions.
-------
-1.6
-1.8
-2.0
- -2.2
-2.4
9 -2.6
ID
% -2.8
ฃ -3.0
-3.2
Reverse-Gas AP
Boiler Load
900
800
700
600
500
TJ
CD
O
400 2
300 &
200
100
7/18
7/19
7/20
7/21
7/22
Figure 8. Variation in Reverse-Gas AP due to Changes in Boiler Load
100
150
200
250 300
Boiler Load, MW
350
400
450
Figure 9. Reverse-Gas AP vs. Boiler Load
When we began our evaluation, the baghouse cleaning cycle was initiated at a
pressure drop of 4.0 inches. By this protocol, the baghouse was cleaned continuously
at full load conditions, when the reverse-gas magnitude was at its minimum levels.
When the boiler was at low load conditions and the reverse-gas magnitude was at its
maximum levels, cleaning occurred every 4 to 6 hours. To maximize the benefit of
higher levels of reverse gas, we suggested that the cleaning setpoint be lowered, so
that the baghouse would also clean continuously at low boiler loads. This change was
made for both baghouse casings.
10
-------
We found that there was a slight improvement in the effectiveness of bag cleaning, but
most of the benefits achieved at low load conditions were eliminated when the boiler
returned to high loads (with the reduced reverse-gas flowrate).
Another example of the importance of reverse gas was observed when the reverse-gas
damper had been inadvertently lowered, which produced reverse-gas AP values of
approximately -0.10 in. H2O during cleaning. Figure 10 shows that the relative drag
and bag weight increased significantly during the period of lower reverse-gas flowrate.
20
6/14
6/15
6/16
6/17
Figure 10. Effect of Lower Reverse-Gas Flowrate on Drag and Bag Weight
Bag Tensioning Issues
During our inspections of the baghouse compartments, we observed that the tensioning
springs for the majority of the bags were compressed to a point which allowed very little
movement of the bag during reverse-gas cleaning. During the 8 years of operation with
the conical springs, the springs had lost some of their resiliency due to the bag weight
(currently in the range of 150 to 175 pounds) and exposure at 300 ฐF, which further
reduced the spring's effective movement during cleaning.
We experimented with several old conical springs and found that with no compression
force, these springs had lost approximately 1 inch of their original spring height. A new
5-inch conical spring, under the design tension of approximately 70 pounds, has a
remaining distance of 1.5 to 2.0 inches to full compression. Therefore, the old springs
had lost about half of their useful capacity.
We experimented with tensioning springs of alternate designs to compare with the
performance of a new BHA conical spring. Four alternate spring types (linear, conical,
and double conical) were installed on bags in the vicinity of the test bag in compartment
A2 that could be seen from the observation viewport. These four bags were manually
shaken down to approximately 70 pounds on May 13. With a reduction in bag weight
from the manual shaking, and use of new springs, the bags were able to move and
deform better during the cleaning period. After 5 weeks of normal operation, these
bags were reweighed, which showed that the bags had gained from 8 to 16 pounds.
11
-------
These tests did not provide conclusive results of the preferred type of tensioning spring.
However, the tests of these four bags showed that the bag weight increased much less
than had been expected since they were now handling a higher-than-average
percentage of the flow in the compartment because of their lower dustcake thickness.
Effectiveness of Sonic Horns
The use of sonic horns has been widely applied throughout the utility baghouse
industry, and has been demonstrated to be very effective in the control of the growth of
the dustcake. Bonanza installed 4 BHA horns in each compartment in September
1992. During our evaluation, we observed that the horns were operating with an air
supply pressure of 70 to 90 psi, which was within the design range of the manufacturer.
We believe that the performance of the horns were dampened with the original status
of the heavy bags. A significant portion of the sonic energy would be absorbed by the
accumulated mass of ash in the compartment. (With 450 bags per compartment with a
average weight of 160 pounds each, the total weight of ash in a compartment was over
70,000 pounds.) As a consequence of the bags being so heavy, we believe that the
sonic horns were not as efficient as they would be with bags of less weight.
Throughout the evaluation program, we monitored the baghouse operation to develop a
composite history of performance. Initially, we evaluated the baghouse under varying
boiler load conditions to determine a range of operation for the various performance
indicators. We then began to alter the level of the cleaning parameters to determine
the effect on bag cleaning efficiency.
Over the course of this evaluation phase, we were able to identify problems and
limitations to the bag cleaning operation. These problems and limitations were:
RIG was insufficient to clean bags adequately at high load conditions
Existing springs limited the amount of bag movement during R/G cleaning
Sonic horns offered minimal benefits to control bag weight
Corrective Actions
From the beginning of our evaluation, we were hoping to discover better settings of the
cleaning cycle parameters that would deliver higher cleaning energies and would strip
away the heavy accumulated dustcake in an automated fashion. We did not achieve
this goal due to the limitations of the existing bag cleaning system (reverse-gas and
sonic horns). Given the limitations of these cleaning parameters, and concern that the
heavy bags could damage the baghouse structural supports, we discussed an
alternative plan to manually shake all of the bags in each compartment.
Manually shaking all of the bags had been performed three times at Bonanza as shown
in Figure 4. The baghouse pressure drop was reported to be reduced by 0.2 to 0.9
inches as a result of these actions, although the benefits were temporary. It was
reported that the baghouse pressure drop had returned to its previous levels over a
period of several weeks. Being aware of these earlier actions, we were cautious in
assessing the benefit of shaking all of the bags.
12
-------
Five test bags that had been manually shaken and had new tensioning springs showed
minimal increases in weight over the subsequent 5-week period. This minimal increase
in bag weight seemed contrary to what had been experienced in the previous bag
shakings at Bonanza. We believe that our ability to maintain these test bags at lower
sustainable bag weights was due to the increased movement of the bags during
cleaning because the bags were lighter and were mounted on new springs.
Manually shaking all of the bags to remove sufficient dustcake as to lower baghouse
AP would effectively increase the capacity of the reverse-gas fans at the higher boiler
load conditions. The consistent delivery of reverse-gas at sufficient levels was deemed
to be essential in controlling the increase in bag weight in long-term operation. We
experimented with bag shaking techniques and found that a bag could be shaken at
the top and bottom of the bag repetitively for approximately one minute to remove 90 to
100 pounds of ash from each bag.
Deseret undertook a program to manually shake all of the bags (to yield reductions of
90-100 pounds per bag) and to install new BHA conical springs. This work was
performed on a compartment-by-compartment basis from June 10 through August 20.
Based on shaking 90 pounds of ash from a bag, the cumulative amount of ash
removed from the baghouse was almost 1 million pounds. During the compartment
activities, bags which had ring cover failures (due to a defective sewing technique by
the manufacturer) and other failed bags were replaced. Of the 620 bags that were
replaced during this period, 95% of the bags were replaced due to ring cover failures.
Following the completion of the manual bag cleanings on August 20, the baghouse AP
was reduced by over 2 inches as a result of the reduction in bag weight. We had
achieved the objectives of delivering a consistent level of reverse gas to the
compartments at all boiler load conditions due to the lower levels of baghouse pressure
drop. Figure 11 shows the stability of the R/G AP after the manual bag cleanings were
completed.
-0.6
-0.8
-1.0
I
I I Iff 'I , III!
-2.4
500
400
300
200
100
0
-100
-200
-300
4/1 5/1 6/1 7/1 8/1 8/31 10/1 10/31 12/1 12/31
Figure 11. Boiler Load and Reverse-Gas AP for April through December 1996
13
-------
Optimization Activities
The last phase of the project was to determine optimum settings for the baghouse
cleaning system that would maintain the low level of baghouse AP achieved from the
manual bag cleaning. The deficiency of reverse gas was now remedied by the
reduction of the baghouse AP. With this improved baseline of operation, we began to
make systematic changes in the cleaning protocol to define the effects and
relationships for the cleaning parameters.
Figure 12 presents the 1-2 baghouse AP for the evaluation period, and shows a long-
term perspective of the subsequent operation following the manual bag cleanings. The
data shown includes only periods of full boiler load operations (420 to 440 MW) with all
compartments in service. This figure shows that prior to the manual bag cleanings, the
baghouse AP ranged from 6.1 to 7.8 in. H2O for boiler operation over 420 MW. Since
August 20, the baghouse has operated within a stable range of 4.3 to 5.7 in. H2O.
Data shown are for boiler
operation over 420 MW, and with
all compartments in service I
4/1 5/1 6/1 7/1 8/1 8/31 10/1 10/31 12/1 12/31
Figure 12. 1-2 Baghouse AP for April through December 1996
Manual Bag Cleaning
in Compartment A2
4/1 5/1 6/1 7/1 8/1 8/31 10/1 10/31 12/1 12/31
Figure 13. Relative Drag for Compartment 1-2 A2 for April through December 1996
14
-------
Another perspective of the change in baghouse operation is shown in Figure 13, which
shows the "after-dean' relative drag for compartment 1-2 A2 during the evaluation
period. This figure supports the trend of the baghouse AP shown in Figure 12. Each of
these two figures shows a total reduction of approximately 25% in their respective
performance parameters.
We considered the changes to the cleaning protocol that may offer more effective bag
cleaning. Since we were operating the sonic horns at the maximum delivery pressure
and the reverse-gas system at its maximum capacities, we altered the duration that
these activities were performed. In addition to these tests, we evaluated the effects of
extending the filtration cycle by raising the cleaning setpoint. These various tests that
were conducted during this optimization phase are listed below.
Increased sonic horn duration from 10 to 20 seconds September 12
Extended filtration period September 24
Extended R/G period from 60 to 90 seconds October 10
Increased sonic horn duration from 20 to 30 seconds October 28
Effects of these changes to the cleaning protocol are shown in Figure 14. This figure
presents the baghouse AP and relative drag (before and after cleaning) for
compartment 1-2 A2 during the optimization period.
12/7
12/21
8/17 8/31 9/14 9/28 10/12 10/26 11/9 11/23
Figure 14. Baghouse DP and Relative Drag for Compartment 1-2 A2 for August 20
through December 20, 1996
Sonic Horn Operation
Following the manual bag cleanings, we performed several tests to evaluate the sonic
horns. On several occasions, we valved out the sonic horns for the test compartments
for periods of several days. During the periods that the horns were turned off, the bag
weights increased from 5 to 15 pounds, although the gain in weight was completely
reversible after the horns were turned back on. Figure 15 presents the bag weight for
the test bag in compartment 1-2 A2 during this test period. Figure 16 shows the
relative drag for compartment 1-2 A2 for this same time period. These tests verified the
15
-------
benefits of the sonic horns and provided a reference for assessing improvements with
the sonic horns.
70
6/18 6/28 7/8 7/18 7/28 8/7 8/17 8/27 9/6 9/16
Figure 15. Compartment 1-2 A2 Bag Weight for June 18 through September 25, 1996
Manual Cleaning of
Bags Completed
Increase Horn Duration
from 10 to 20 seconds
6/18 6/28 7/8 7/18 7/28 8/7 8/17 8/27 9/6 9/16
Figure 16. Compartment 1-2 A2 Relative Drag for June 18 through September 25, 1996
The timers for the sonic horns were initially set for 10 seconds, and were activated,
coincident with the beginning of the reverse-gas period. The actual sounding of the
horns had a duration of approximately 8 seconds, due to the lag period until the high-
pressure air reached the horns. We evaluated the benefits of operating the horns for a
longer time period. We increased the sonic horn duration from 10 seconds to 20
seconds for the entire baghouse on September 12. Figure 17 presents the baghouse
AP for a 6-hour period during this test and shows an immediate reduction of
approximately 0.5 in. H2O as a result of the increase in the horn duration. During this
6-hour period, the unit operated at steady, high-load conditions.
16
-------
490
480
* 5.0
410
13:30 14:30 15:30 16:30 17:30 18:30 19:30
Figure 17. Baghouse AP and Boiler Load on September 12, 1996
The effect on bag weights attributed to this increase in horn duration is shown in Figure
15. We measured a reduction in bag weight coincident with this increase in the horn
duration. There was a reduction of less than 5 pounds in the weight of the test bag in
compartment 1-2 A2 and a 10 pound reduction in the test bag for compartment 1-2 F2.
These data suggests that the increase in the sonic horn duration from 10 to 20 seconds
provided more energy, cleaning the bags more effectively.
To further test this parameter, we increased the sonic horn duration from 20 to 30
seconds on October 28. The effects of this change were indeterminate. We believe
that operating the horns with a 20- to 30-second duration would be an effective range
of operation.
Filtration Period
One of our initial hypotheses was that the filtration period could work in a counter-
intuitive manner, in that better cleaning may be achieved with longer time periods
between cleanings. We speculated that if the bag accumulated a sufficient.amount of
ash, due to an extended filtration period, the bag may clean more effectively during the
reverse-gas period. An avalanche of ash within the bag may enable more vigorous
scouring of the dustcake as the ash falls down the 35-foot bag.
Following the manual bag cleanings, the baghouse pressure drop was reduced by a
couple of inches, which allowed more flexibility in the settings of the cleaning
parameters. Specifically, the length of the filtration period could be extended without
risking a problem with high AP. (During normal, full-load operations with no
compartment in cleaning, the amount of ash added to the bags increases the baghouse
pressure drop by 0.3 to 0.5 in. H2O per hour.). On October 1, we began to extend the
filtration period by gradually increasing the setpoint for the cleaning initiation to an
eventual setting of 5.8 inches. At this setting, the baghouse compartments cleaned at
an interval of 2 to 4 hours (at full load conditions). Figure 18 shows a history of the
filtration period duration for the 1-2 baghouse casing since mid-September.
17
-------
9/15
9/30
10/15
10/30 11/14
11/29
12/14 12/29
Figure 18. Filtration Period Duration for the 1-2 Baghouse from September
through December 1996
Since October 1, when the baghouse began operating with an extended filtration
period, the baghouse has operated at a steady pressure drop and relative drag as
shown in Figure 14. We can not determine if the extended filtration period has assisted
in the control of these parameters, although there is no evidence that this protocol has
been detrimental. We believe that operating with an extended filtration period should
be considered. We predict that there would be a tangible benefit for the durability of
the bags by extending the filtration period. Assuming that a 3-hour filtration period is
achievable in long-term operation, the bags would be cleaned one-fourth as often as
with continuous cleaning. This reduction in the normal flexing of the bags during the
reverse-gas cleaning process should have a positive effect in the extending bag life.
Another benefit from operation with an extended filtration period would be derived from
the elimination of one of the reverse-gas fans, by utilizing one fan for both baghouse
casings. The time required to complete one cleaning cycle for a baghouse casing is 45
minutes, therefore the time to clean both casings consecutively would be approximately
90 minutes. We have operated for the past 3 months with filtration periods in the range
of 120 to 300 minutes; therefore, there is sufficient time in the current cleaning protocol
to permit one reverse-gas fan to service both baghouse casings. Currently, both R/G
fans (one for each casing) are operated continuously, regardless of whether the
baghouse is in the cleaning cycle. The savings due to fan horsepower costs, and
periodic maintenance on one less fan should provide a financial incentive to use a
single reverse-gas fan for both casings. Plant personnel have verified that this change
in procedure could be facilitated by the logic in the current control system.
Reverse-Gas Period
Since the function of the reverse-gas period is primary in the role of cleaning, we
increased the period from 60 to 90 seconds on October 10. Although there was a slight
improvement in the performance parameters shown in Figure 14, the variations in these
parameters over a long timeframe obscures the benefit from this change.
18
-------
THE USE OF TREATMENT TIME AND EMISSIONS
INSTEAD OF SCA AND EFFICIENCY
FOR SIZING ELECTROSTATIC PRECIPITATORS
Robert A. Mastropietro
Research-Cottrell, Inc.
Route 22 West & Station Road
Branchburg, NJ 08876
Introduction
The first commercial electrostatic precipitator (ESP) was developed in 1907. Since that time many
types of mathematical models have been used to calculate ESP sizes relative to desired performance.
This paper discusses several of these historical approaches, before describing a treatment time vs.
emissions approach to ESP sizing. The use of treatment time vs. emissions for sizing ESPs has
become a common approach to this calculation, due to changes in ESP designs and performance
requirements.
Historical Trends in ESP Sizing
Many simple mathematical models have been used over the years to size ESPs. These include simple
charts such as "CFM/PIPE vs Efficiency'1 or "CFM/Duct vs. Efficiency" type models. Other
approaches used charts to select precipitation rate parameters vs. operating conditions such as fuel
sulfur, gas temperature, gas moisture, etc. These latter charts were simply graphical means of
predicting resistivity. Volume effects on ESP performance were typically predicted using
mathematical equations.
The most common ESP sizing approach for the era 1920-1960, was the Deutsch-Anderson equation;
Efficiency = 1 - e"
-(/Wxw)
or = 1 - e'
-{SCA/16.67 x w)
-------
Where A is collecting electrode (C.E.) surface area, V is gas volume, w is the precipitation rate
parameter, and SCA is the ratio of C.E. surface area to gas volume in 1000 ACFM. However it
should be noted that the above equation was really Deutsch's equation (circa 1922), and was given the
name Deutsch-Anderson Equation in respect to Anderson's earlier equation. Anderson's ESP sizing
equation in 1919 was;
Efficiency = 1 - k'
In this equation, k was a precipitation rate constant and t was the treatment time. Thus some
researchers in this field used treatment time, before the concept of SCA or ratio of collecting surface
to gas volume was recommended by Deutsch in the early 1920s.
It is important to subsequent calculations to note that the Deutsch Equation has been modified in
recent times to reflect a flatter slope when plotting SCA vs. efficiency at very high efficiencies. This
newer equation had been called the modified Deutsch equation;
Efficiency = 1 - e x y
Qr _ j _ g-(SCA/16.67xwk)
Where the exponent y typically takes values between 0.5 and 0.8. The exact value of y is determined
by performing multiple linear regression on test data for a specific application. Typically for utility
applications, R-C uses a value of this exponent, y, of 0.6.
Problems With The Efficiency and w/wk Approach
In using the above historical approach to ESP sizing several problems can occur. The first problem is
confusion over just what value of w/wk to use. Figure 1 shows test data from three different utility
boiler installations; low sulfur P-C, SO3 conditioned P-C, and oil fired. Even within a three set test
sample, there is often a wide variation in w/wk. Note that further discussion will be limited to w^
values for simplicity. The value of wk for the best tests compared to worst tests in these three
examples were 16%, 21%, and 190% greater respectively. This is a wide variation; there must be
some explanation for why the ESP worked well on some tests and much poorer on others?
Figure 2 shows the same data sets, but with the inlet and outlet dust concentrations also shown. This
data shows that the outlet emissions were actually pretty consistent for each of the sets of data. Thus
the tests that we viewed to be poor from Figure 1 were actually good, and in some cases actually the
best test when evaluated on the basis of outlet emission. The parameter that was confusing this
evaluation was the inlet dust loading. The ESP was basically operating as a constant emission device,
and the efficiency or wk calculation was responding to the inlet loading.
-------
The second problem in using the historical approach deals with the use of empirically obtained Wk
values in designing to judgementally specified design values. For example, the following set of test
data may have been obtained for a utility boiler with a specific low sulfur coal:
SCA INLET OUTLET EFF. wk(0.6Exp.)
FT2/KACFM LB/MMBTU LB/MMBTU % FT/SEC.
350 6.0 .029 99.52 0.78
350 6.7 0.027 99.60 0.82
350 5.8 0.026 99.55 0.79
0.80 Avg.
In this case the data was quite uniform in inlet loading and w^ value. The problem arises in taking this
empirical data and applying it to a specification for a new ESP. The specification will contain coal
information such as;
Coal Ash Content 4-16 %
Coal Heating Value 11,000-12,500 BTU/LB
Thus if an inlet loading is calculated based upon the worst case conditions of coal;
1,000,OOOMMBTU/ 11,OOOBTU/LB * 0.16 Ash * 0.85 Carryover
= 12.4 LB/MMBTU
To achieve NSPS of 0.03 LB/MMBTU, this gives a required efficiency of (12.4-0.03)/12.4=99.76%.
Applying the average wk from above gives a required SCA of 417 FT2/KACFM. This is 20 % larger
than the size of the original installation which operated below the 0.03 LB/MMBTU requirement. Is
this additional 20 % really required because the ESP is a constant efficiency device, or is it not
required because the ESP is more of a constant emissions device? Current findings indicate that the
latter approach is more correct. For most variation within a given process, the ESP acts more as a
constant emissions device. Although we have found that if the inlet loading increase is dramatic
enough to impact the ESP electrical operation, the ESP will allow a greater penetration.
Treatment Time Sizing Approach
The use of treatment time in modem sizing is an approach that has come about due to necessity. Over
the last 20 years R-C has used collecting electrode spacing of 9", 10", 11", 12", 13", 13.75", 14",
15.75", and 16" Applications engineers originally tried to treat these different C.E. spacings by using
SCA credits. Laboratory and full-scale studies have shown that ESP performance is similar with
different C.E. spacings, at "equivalent" SCAs (providing corona suppression is not encountered). For
example, an SCA of 400 FT2/KACFM at 9" was equivalent to 400x9/12=300 FT2/KACFM at 12" or
400x9/16=225 FT2/KACFM at 16" Once the SCAs were corrected to an
-------
equivalent collecting electrode spacing, w^ values could be compared on an equivalent basis. While
this is one possible way to address this situation, this method is extremely unwieldy. It also
creates confusion when a customer or A/E is comparing different vendors at different plate spacings,
some of whom even have different spacings from field to field within their own designs. Even more
confusing are vendors who want to include discharge electrode surface or collecting plate projections
as collecting surface. All this confusion disappears, by simply evaluating ESP size on the basis of
treatment time.
The above discussion points out that comparison of ESP performance on the basis of uncorrected
SCA is not valid when the C.E. spacing differs. Figure 1 for example does not state a plate spacing
with the SCA values, and is confusing without this parameter. Therefore, it is suggested that SCA
take more of a "back-seat" in modem sizing. This suggestion, that SCA be ignored, may be
uncomfortable to a generation of ESP engineers accustomed to using SCA in the Deutsch-Anderson
Equation. The treatment time approach, however, is a much better method for comparison and sizing.
Note that mathematically the two values are directly related. For example at 12" gas passage spacing
(i.e. 1' spacing makes the chamber width equal to the number of gas passages);
SCA = (#P*#C*#gp*2*C.E. Ht. * C.E. Length)/(Volume/1000)
Re-arranging terms;
SCA = ((C.E.Len.*#P*#C*%5*C.E.Ht.*60)/Volume)x2xlOOO/60
Separately;
Treatment Time = C.E.Length/(Volume/(#P*#C*#gp*C.E.Ht.*60))
Re-arranging terms;
Treatment Time = C.E.Length*#P*#C*#gp*C.E.Ht.*60/Volume
Therefore, by substitution
SCA = Treatment Timex2xlOOO/60 = Treatment Time x 33.3 (at 12")
By similar calculations the constant multiplier can be obtained for other plate spacings. Several
common spacings and multipliers are shown below;
Plate Spacing Multiplier
9" 44.4
12" 33.3
16" 25.0
-------
For example, 300 FT2/KACFM at 16" has 300/25=12 seconds of treatment time.
Thus the two values, SCA and treatment time, are directly proportional at any given plate spacing.
Since SCA and treatment time are directly proportional, the concept of predicting ESP performance
based on either of the two is essentially similar; the larger the ESP, whether measured by SCA or
treatment time, the higher the efficiency and the lower the emission.
Emissions Only Sizing Approach
Another major change that has occurred in the industry, is that performance requirements are typically
no longer stated in terms of efficiency. All governmental codes are now stated in terms of emissions
(i.e. GR/SDCF, GM/NM3, LB/MMBTU, etc.). This departs from the historic requirements of
efficiency or outlet stopper. For example, utility boilers require particulate emissions less than 0.03
LB/MMBTU regardless of the inlet dust concentration and resultant required efficiency. Efficiency
remains of technical interest, but is not used in performance guarantees. More than half of recent
performance tests performed by R-C have been on the outlet emissions only. Collection efficiencies
are never even determined.
Because efficiency is becoming a value that is infrequently measured (and not used in performance
guarantees), it is less useful in the actual sizing calculations. More often the specified inlet loading is
used to size the ash handling equipment, and not to calculate efficiency for the ESP sizing. Especially
if the customer/AE has obviously added conservatism to this inlet value.
Treatment Time and Emissions Approaches
The end result of the changes in design philosophy in terms of treatment time and emissions
guarantees, has been a change in the way ESPs are sized. Where sizing curves historically took the
form of efficiency vs. SCA (Figure 3), the new sizing curves are taking the form of treatment time vs.
emissions (Figure 4). Findings show that this latter approach is valid, in that efficiency will fluctuate
depending on inlet dust concentration. However emissions remain relatively constant for a given
treatment time even though efficiency varies. This is attributed to carry over of larger particles from
the same process causing higher inlet dust concentrations. This rationalization may not be absolutely
correct, but the curves such as Figure 4 follow smooth downward trends. There are no high emissions
outliers that can be attributed to higher inlet loadings.
There is also a theoretical basis for the concept of ESP sizing by treatment time vs. emissions. For
example, if point A on Figure 5 is a typical performance level, 9 seconds of treatment (SCA = 9 x 33.3
= 300) and 0.01 GR/ACF. And assuming an inlet of 2 GR/ACF. We can then use the modified
Deutsch equation to calculate a WK;
0.6
,-(300/16.67 xwK)
Eff. = (2-0.01)72 = 0.995 = 1-e
WK =0.895
Then using this WK value at other treatment times/SCA's, we can calculate a series of emissions values
predicted by the modified Deutsch Equation. This can be used to develop a plot of the predictions of
-------
INLET
GR/ACF
2
2
2
2
EXPONENT
0.6
0.6
0.6
0.6
SCA
FT2/KACFM
300 (9 sec)
167(5 sec)
200(6 sec)
233 (7 sec)
WK
FT/SEi
0.895
0.895
0.895
0.895
the modified Deutsch Equation within the format of the treatment time vs. emissions plot;
EFF.
%
99.5
97.6
98.43
98.95
2 0.6 400 (12 " ) 0.895 99.82 0.004
2 0.6 500(16") 0.895 99.925 0.0015
Plotting these values on Figure 5, it can be seen that there is quite good agreement between the
predictions of the modified Deutsch Equation and the plot of treatment time vs. emissions. Thus the
method of sizing from treatment time vs. emissions agrees with historical theory on prediction of ESP
performance.
Exactly how treatment time vs. emissions curves are used in predicting performance where
performance risk is an issue, is dependent on the risk level that is assumed. Obviously a mean line
based upon a least squares approach could be constructed through the center of the data (after
transformation to a linear form). Or if so desired a level of conservatism can be added by calculating a
line at a desired confidence level. However, it should be noted that the data displays non-homogeneity
of variance (i.e. the variance is small at high treatment times and large at low treatment times). Thus
the level of conservatism would differ depending on the performance range.
Summary
The above discussion describes one of the sizing methods that can be used to address the current
design requirements. This method is a combination of scientific facts and commercial necessities. R-C
has found these techniques to be the best way to respond to the marketplace that exists in the 1990s.
-------
FIGURE 1
ESP TEST DATA COMPARED ON THE BASIS OF SCA, EFFICIENCY, AND w/wk
(ACTUAL 3 TEST SAMPLES FROM FULL SCALE UTILITY INSTALLATIONS)
Precipitation Rate Parameters
Application
P-C
P-C, S03
OIL FIRED
Test
Number
1
2
3
1
2
3
1
2
3
SCA
FT2/KACFM
755
765
764
216
216
216
263
266
268
Collection
Efficiency
m
99.91
99.86
99.85
97.99
97.87
96.92
94.55
91.49
78.95
Std. Deutsch
w.CM/S
4.72
4.36
4.32
9.19
9.05
8.19
5.60
4.71
2.95
Mod. Deutsch
wk, FT/SfO.6 exp.)
0.57
0.50
0.49
0.75
0.73
0.62
0.38
0.28
0.13
Remarks Based
on Efficiency
or w/wk
BEST TEST
MEDIUM
MEDIUM
BEST TEST
GOOD
POOR
BEST TEST
MEDIUM
VERY POOR
-------
FIGURE 2
ESP TEST DATA COMPARED ON BASIS OF TREATMENT TIME vs. EMISSIONS
(ACTUAL 3 TEST SAMPLES FROM FULL-SCALE UTILITY INSTALLATIONS)
Dust Concentrations
Application
P-C
P-C, SO3
OIL FIRED
Test
1
2
3
1
2
3
1
2
3
Treatment
Time Seconds
22.7
23.0
22.9
4.9
4.9
4.9
5.9
6.0
6.0
Inlet
LB/MMBTU
10.74
6.15
5.24
9.42
6.59
6.23
0.128
0.109
0.044
Outlet
LB/MMBTU
0.01
0.009
0.008
0.20
0.13
0.19
0.007
0.009
0.009
Collection
Efficiency
ฃ%}
99.91
99.86
99.85
97.99
97.87
96.92
94.55
91.49
78.95
Remarks Based
on Outlet
Emissions
GOOD
GOOD
BEST TEST
MEDIUM
BEST TEST
MEDIUM
BEST TEST
GOOD
GOOD
-------
Figure 3
TYPICAL SCA VS. EFFICIENCY SIZING CURVE
009
99.8
99.0
98.0
#
>-"
z 95-ฐ
LU
O
u.
lL
UJ
90.0
85.0
80.0
70.0
Research - Cottrell
7"
s
,x~
/
s
s
\s
/
x
f
t
t
s
s
/
S 8
SPECIFIC COLLECTION AREA, ft2/1000 acfm
-------
Figure 4
TYPICAL ELECTROSTATIC PRECIPITATOR DATA
TREATMENT TIME VS. EMISSIONS
O.Ob
u. ฐ-ฐ4
O
5:
o:
CD
to"
ง ฐ'ฐ3
c/2
ID
LU
5 0.02
O
UL
CL
0.01
0
n
D
n
n
n
a
a
n a
n D
D a
n
m nn _ a
Da
n ._. a u
Du n _
n
5 7 9 11 13 15 17 19
ESP TREATMENT TIME, SECONDS
11
-------
a:
(D
z
o
CO
UJ
til
I
o
o:
0.05
0.04
0.03
0.02
0.01
Figure 5
TYPICAL ELECTROSTATIC PRECIPITATOR DATA
TREATMENT TIME VS. EMISSIONS
a
I I I I I
I I I I
I
11 13 15 17 19
ESP TREATMENT TIME, SECONDS
!LJL
13
-------
Computational Fluid Dynamics Applications for Power Plants
Rajendra P. Gaikwad
William DePriest
Air & Water Quality Division
Sargent & Lundy
55 East Monroe Street
Chicago, Illinois 60603
Abstract
Use of Computational Fluid Dynamics (CFD) modeling realized a significant growth in the utility
industry during low-NOx burner retrofit installations to assist engineers in predicting NOx emissions
from boilers. Since these initial applications of CFD, a number of utilities have also accepted the
CFD modeling as an effective tool for predictions of velocity, temperature, and NOx concentration
profiles. Also, apart from gas side modeling, there are numerous multimedia applications of CFD in
the utility industry.
As examples, Sargent & Lundy has successfully used CFD modeling to predict such values as the
temperature and velocity profiles in a complicated assembly of ductworks, reentrainment in a
chimney, flow distribution in an electrostatic precipitators, and flow distribution through an FGD
precooler. In conjunction with these investigations, we used the technique to identify methods to
improve flow to reduce the pressure drop and improve the effectiveness of these devices. In one
application, the ductwork included the inlet and outlets of scrubbers, flue gas bypasses, and chimney.
The velocity profiles provided the locations to add flow-guiding vanes and identified possible
locations of solids build-up; the temperature profiles provided the locations and extent of mixing
zones that may be subjected to high rates of corrosion. The results of this study helped redesign the
outlet ductwork to minimize the build-up in the ductwork and reduce the pressure loss. Several CFD
applications including FGD systems, ESPs, and ductwork routing are presented in this paper.
Introduction
The use of Computational Fluid Dynamics (CFD) involves the solution of the Navier-Stokes
equations for fluid flow, heat transfer, and chemistry at several thousand discrete points on a
computational grid (or within cells) in the flow domain. The use of CFD enables engineers to obtain
solutions for problems with complex geometries and boundary conditions without building a
prototype. A CFD analysis results in values for fluid velocity, fluid temperature, and concentration
throughout the solution domain. Based on this analysis, an engineer can optimize fluid flow patterns,
by installing flow-guiding vanes, or temperature distribution, by adjusting either the geometry or the
Page 1
-------
boundary conditions. However, if the baseline validation is not performed or can not be performed,
the results from CFD shall be treated as guidelines rather than absolute results.
CM) modeling was introduced to the utility industry during low-NOx burner retrofit installations to
assist engineers in predicting NOx emissions from boilers. In most cases, baseline NOx data were
necessary to predict the impact on NOx due to changes in burner geometry or the introduction of an
overfire port. Overall, the uses were limited as the utilities still relied on the OEM's performance
guarantees rather than validating the performance by using CFD. Since these initial NOx applications
of CED, a number of utilities have also accepted the effectiveness of CFD modeling as an effective
tool for predicting velocity and temperature profiles in various geometries. Apart from NOx
prediction, there are a numerous applications of CFD in the utility industry.
Sargent & Lundy has licensed FLOW3D, a CFD code developed by AEA Technology, Pittsburgh.
Several models are supplied with this code, which enables a user to effectively simulate the problem
as close to reality as possible. The code has undergone more than 15 years of development and
refinement to reach its current state. The code has been extensively used by various industries
inchidiag the chemical industry, aeronautical industries, nuclear industry, and auto industry.
As examples, Sargent & Lundy has successfully used CFD modeling to predict such values as the
temperature and velocity profiles in a complicated assembly of ductworks, reentrainment in a
chimney, flow distribution in electrostatic precipitators, and flow distribution through an FGD
precooler. In conjunction with these investigations, we identified the methods to improve flow to
reduce the pressure drop and improve the effectiveness of these devices.
From among the number of studies performed, this paper summarizes the following five applications:
Application of CFD to Inlet Vanes of ESP for Uniform Flow Distribution
Application of CFD to Scrubber Outlet Ductwork and Chimney Breaching
Application of CFD to How Distribution at the Inlet to an ID Fan
Application of CFD to Achieve Uniform Flow Distribution at the Inlet of a baghouse
Other - Application of CFD to Predict Erosion in a Circulating Water Pipe
In a typical application, it is important to obtain a precise drawing and dimensions of the geometry to
be studied. Due to limitations with the computational facilities used, the applications were limited to
60,000 grid cells. The geometry is built and the grid is generated to connect all the blocks with a
fairly uniform grid. Once the grids are generated, the command file is written to simulate the
boundary conditions. The convergence may take anywhere from 300 to 700 iterations. Typically,
multiple profiles such as temperature and velocity take more time than a single profile such as only
velocity.
Application of CJb'l) to Inlet Vanes of ESP for Uniform Flow Distribution
In this application, an ESP inlet duct, ESP inlet nozzle, perforated plates, ESP body, and the outlet
nozzle were simulated to analyze the current flow distribution and suggest the possible modifications
to the vanes and/or distribution plates to achieve uniform flow distribution (1). As shown in Figure 1,
Page 2
-------
the flue gas from a air-preheater flows downward into an electrostatic precipitator nozzle. The
velocity exiting the air preheater toward the inlet nozzle was 60 ft/s. The flue gas had to make a 90ฐ
turn to enter the Met nozzle of the precipitator. This model had approximately 30,000 cells and used
constant flow boundary conditions across the rectangular ductwork at the inlet to the model
By viewing the results of the analysis in detail, it was possible to gam an understanding of the
operation of the ESP that would have been impossible by viewing tie actual system in operation. The
analysis results showed flow direction, velocity, and pressure at every point within the model The
cause of the poor flow distribution hi the precipitator was readily apparent as shown in Figure 1. The
90ฐ bend in the inlet ductwork and a small distribution pkte in the Met nozzle resulted in an increase
in velocity at the top 2/3 of the precipitator. Instead of the expected uniform flow distribution of
15% rms, the flow distribution was noted to be 40% rms. A perforated pkte used downstream
helped to redistribute the flow to a certain extent, however, the poor distribution remained.
The recommended changes included the reinstallation of a distribution plate smaller than the original
to distribute the flow across the height of the precipitator. An additional perforated pkte was also
recommended to further distribute the flow. A series of iterations were performed to determine
precisely what pkte configuration would provide uniform flow at the least cost in terms of gas side
pressure drop. The results are shown in Figure 2. The analysis showed that these changes would
reduce, but not eliminate, the flow unbalance problem. The flow distribution with the modification
was estimated to have an rms value of 22% in lieu of 40% without these minor changes. It was
concluded that the configuration of the Met duct as well as the outlet duct going to chimney would
have to be significantly modified to achieve a 15% rms flow.
Typically a I/IO"1 scale physical model is used in a conventional analysis at a cost of $40,000 to
$50,000. However, the use of CFD analysis, including the analysis time, use of hardware, and cost of
licensing, was less than $10,000. Another advantage of the CFD model is that it can be changed very
quickly to evaluate alternate design scenarios and configurations. Engineers can quickly evaluate the
impact of possible design changes without the cost and time involved in building new prototypes.
Application of CFD to Scrubber Outlet Ductwork and Chimney Breaching
In this application, three scrubber outlet ducts and two hot bypass ducts were connected to a
common duct. The common duct was connected to chimney using another duct. The common duct
was used to mix the hot and cold gases so that the gases going into the stack would be mixed
together and the saturated gases from the scrubber would be heated to at least 170ฐF. This heating
results in dry stack operation. If the gases are not mixed property, however, flow stratification may
occur. This ductwork did not have any flow-guiding vanes and the units were operated in a random
fashion. This model had approximately 50,000 cells and used variable flow boundary conditions
across the rectangular ductwork at the Met to the model The common Met ductwork was circular,
the ductwork combining the circular duct and circular chimney was rectangular. The configuration of
the system is shown in Figure 3.
Several flow variations were analyzed. In some cases, for example, the different amounts of gas were
passed through different scrubbers, predominantly due to flow-balancing problems that were the
Pages
-------
result of mist eliminator plugging or damper malfunctioning. A typical velocity profiles are shown in
Figures 4 and 5. In these figures, E and D are the bypass entrance ducts and A, B, and C are the
scrubber outlet ducts entering a common duct. Figures 4 and 5 depict the velocity profiles when each
of the bypass duct is carrying 10% of the flow and scrubbers A, B, and C are in operation. As
individual flows enter the common duct, the velocity drops significantly and flow tends to slug
towards the wall because there are no flow-guiding vanes in this ductwork. The ductwork
connecting the chimney and the common duct as well as the bottom part of the chimney is shown in
Figure 6. As the flow enters the breaching, it tends to slug to the top of the ductwork and enters with
higher velocity in the top half of the breaching. The flow strikes the opposite wall of the chimney and
distributes across the cross-sectional area of the chimney. If the uneven distribution persists all along
the chimney, the flow measurement devices tend to give either lower or higher flow than the actual
flow. It is therefore necessary to put the flow-guiding vanes as the flow enters the chimney.
Figure 5 shows the temperature distribution across the ductwork as the saturated gas at 130ฐF enters
along with 348CF bypass flue gas. As the hot gases from duct D enter and move in the duct, they
strike saturated gases from ductwork A At the boundary of the hot and cold flow interface,
extremely corrosive conditions occur. Such portions of the ductwork have to be protected by using
proper material to avoid corrosion. In this case, approximately 63% of the flow is coming from the
left side through ducts B, C, and E whereas only 37% of the flow is coming from the right side
through ducts A and D. When the flow enters the combination ductwork, the flow from left side
enters with higher velocity, which results in flow stratification in this ductwork and in the chimney. It
is therefore important that the guiding devices be placed in this ductwork to avoid any temperature
stratification. The flow-guiding vanes not only helped the flow distribution but also helped reduce the
pressure drop across the ductwork assembly.
Application of CFD to Flow Distribution at the Inlet to an ID Fan
The main objective of this flow modeling was to provide uniform flow to the ID fan inlet in order to
improve fan performance. Uniform flow would also lower the pressure drop in the ductwork. The
inlet duct configuration shown in Figure 7 was simulated to establish a baseline geometry for the
flow analysis. This geometry represents half of the symmetrical duct. The flue gas flow rate through
the simulated assembly was approximately 361,000 acfin (-170 m3/s). Initially the inlet ductwork
was simulated without any flow-guiding vanes to represent the current design. The 11-tube rows in
the direction of flow were simulated as 6 rows of tubes along the X-axis and the 18-tube rows were
simulated as 8 rows along the Z-axis. This model had approximately 30,000 cells and used constant-
flow boundary conditions across the rectangular ductwork at the Met to the model
Figures 7 and 8 represent the velocity profiles in the X and Y direction in the ductwork without
turning vanes. X-direction is a profile across the duct and perpendicular to the flow. As can be seen,
the velocity profiles in the ductwork are not symmetrical even though the multiclone outlet velocities
were assumed to be uniform. The imbalances in flow occur mainly because high velocities from the
tubes tend to push the gas to the upper part of the duct. This situation results in extremely high
velocities in the upper part of the duct and low velocities in the bottom part of the duct. When
compared with the actual velocity distribution measurement, the distribution in Figure 8 (center
profile) showed the distribution of O.SFto 2.1K(Fisthe average velocity) compared to 0.45Fto
Page 4
-------
1.74V. After various flow distribution trials, it was found that a 60:40 flow distribution (60% flow
through first half of tubes and 40% through the second half of tubes closer to ID fan) was the most
representative and would simulate actual velocity measurements.
Based on these results, further work was performed to improve the flow by using various vane
configurations. Based on the 60:40 flow distribution, the arrangement shown in Figure 9 was
recommended. Once the vanes are installed, however, the uniform flow profiles will also result in
eliminating the circulation zone that was forming in the duct close to the muMclone. The vaning
arrangement, therefore, may result in uniform flow distribution across the muMclone tubes. After
various trials, the number of throat area vanes was reduced to two and the vane over the last tube
row was kept just to cover it up. Figures 10 shows the result of uniform flow distribution from the
multiclone. With uniform flow distribution across the tubes, the vaning arrangement led to more flow
in the bottom part of the duct. This vaning arrangement was simulated to guide more flow in the
bottom to overcome the low flow from the last row of tubes. A part of the jet from the last row,
therefore, does not strike the plate and is diverted slightly upwards to uniformly distribute the flow.
The arrangement shown in Figure 10 encompasses possible flow distribution after the installation of
vanes. These modifications are currently being installed, and the modeling results will be verified with
flow measurements.
Application of CFD to Achieve Uniform Mow Distribution at the Inlet of a Baghouse
The main objective with this flow modeling was to provide uniform flow to the baghouse inlet
manifold in order to utilize all of the baghouse modules equally. Uniform flow will minimize
baghouse pressure drop and will maximize the bag life. The Industrial Gas Cleaning Institute (IGCI)
standard was selected to define uniform flow. IGCI EP-7 is written for electrostatic precipitators
which are more sensitive to flow uniformity than the baghouses, so if we meet these EP-7
requirements we are confident that the flow is uniform enough for the baghouse. The standard
defines uniform flow as follows:
The velocity pattern shall have a minimum of 85% of the velocities not more than 1.15 times the
average velocity, and
99% of the velocities shall not be more than 1.4 times the average velocity.
The inlet duct configuration shown in Figure 11 was simulated to establish a baseline. This model
had approximately 25,000 cells and used constant-flow boundary conditions across the rectangular
ductwork at the inlet to the model Initially, the inlet ductwork was simulated without any flow-
guiding vanes. Figures 11 represent the velocity profiles in Y direction in the ductwork without
turning vanes. The Y-direction is a profile across the duct and perpendicular to the flow. The velocity
profiles in the ductwork were not symmetrical even though the economizer outlet velocities were
assumed to be uniform. The imbalances in flow occur mainly due to two 90ฐ turns after the gases exit
the economizer. The path preferred by the gas results in a recirculation zone in the top left corner of
Figure 11. The flow tends to hug the ceiling and therefore develops a high to low velocity profile
across the ductwork from top to the bottom. The flow distribution at the inlet of the baghouse
revealed the velocity ranging from 45.7 ft/s (13.9 m/s) to 81.3 ft/s (24.8 m/s) with an average
velocity of 60 ft/s (18.3 m/s). This results in a velocity variation of minus 25% and phis 35% from
PageS
-------
the average velocity. The resultant distribution had only 75% of the velocities less than 1.15 times
the average velocity and 93% velocities less than 1.4 times the average velocity, both of which
violated the IGCI guidelines.
Based on these results, further work was performed to improve the flow uniformity of the baseline
configuration by using various vane configurations. The configuration that gave the best flow
characteristics is shown in Figure 12. Again, it was assumed that the velocity profile across the outlet
of the economizer was uniform. The velocity profiles throughout the new ductwork were fairly
uniform. The flow distribution at the inlet of the baghouse revealed a velocity variation of 53.5 ft/s
(16.3 m/s) to 70.8 ft/s (21.6 m/s) with an average velocity of 60 ft/s (18.3 m/s). This results in the
velocity variation of approximately 18% from the average. However, 95% of the velocities were less
thijti 1.15 times the average velocity and 100% of the velocities were within 1.4 time the average
velocity. In fact, 100% of the flow was within 18% of the average flow. The recommended
configuration is being installed, and the inlet flow distribution will be tested during the
commissioniag of the baghouse.
Other Applications
The CFD code has also been used to predict the erosion in a circulating water pipe, the flow
stratification across a feed water pipe, to predict the flow distribution across the boiler building so
that the top of the boiler can be used as the air-intake for the FD fan to utilize the heat losses from
the boiler, to predict the flow distribution in a scrubber due to mist eliminator location, to predict the
flow distribution in a scrabber-precooler that did not have any flow-guiding devices, to predict
droplet reentrainment and trajectories, and to predict flow distribution across an ESP to achieve
IGCI guidelines.
Application of C'FD to Predict Erosion in a Circulating Water Pipe
In this application, two circulating water pipes from two condensers were connected into one large
pipe. As shown in Figure 13, the water from first pipe is expanded and the second pipe enters the
large common pipe after the expansion. This model had approximately 15,000 cells and used
constant flow boundary conditions across the inlets of both pipes. The geometry results in high-
velocity water from the second pipe entering into low-velocity water from the first pipe. Due to the
differences in velocity, a recirculation zone is created at the intersection of the pipes. As shown in
Figure 13, the recirculation creates very high velocities at the bottom of the pipe. For example, the
velocities entering the pipes are 2.5 m/s, however, the velocities in the bottom of the recirculation
zone can reach 4.5 m/s. The continuous formation of the recirculation zone and the high velocities at
the bottom of the pipe may result in massive erosion at that location. As a matter of fact, one utility
discovered a large hole in the pipeline at the section shown hi Figure 13 The resolution of this
problem was to reinforce the pipe at the junction to avoid such problem in the future.
Page 6
-------
Conclusions
The major conclusions drawn from Sargent & Lundy's experience are as follows:
CFD is a very effective way to visualize and optimize the flow and temperature profiles in various
geometries in a power plant
CFD can be used in various applications related to gas flow, water flow, and heat transfer
equipment in the power plant
CFD lowers the cost and significantly reduces the schedule and increases the flexibility of a
model study when compared to model study performed using a plexiglass model
References
1. R.P. Gaikwad, "Computer Simulation Helps Engineers Devise Low Cost Modifications," The
Chief Engineer, October 1996.
Page?
-------
Figure |: Velocity Profile Across the Last Distribution Plate ia the Existing Precipitator with Flow
Guiding Viines, First and Second Perforated Plate with 39% Open Area, and Distribution Plate with
60% Open Area (X - direction)
The velocity profiles show Hui the flow is not uniformly distributed across the cross sectional area
and the high velocities are seen ia upper 70% cross-scctionaJ area.
Figure 2-: Velocity Profile Across the Last Distribution Plate in the Modified Precipitator with How
Guiding Vanes, First Perforated Plate, 1/2 Height aud 39% Open Area, Secoud Plate - FuU Across
Width and 39% open area, and Distribution Plate with 40% Opeu Area (X - direction)
The velocity profiles show that the flow is fairly uniform across the cross sectional area.
-------
Figure 3: Scrubber Outlet Ductwork and Chimney Configuration
-0"
Figure 4: Velocity Profiles in the Center of the Common Ductwork.
Profiles show uneven flow distribution across the ductwork.
-------
Figure 5: Temperature Profiles in the Center of the Common Ductwork
Profiles show uneven temperature distribution across the ductwork
k
4 6502E+02
4 2ub8EH>2
4 0750E+U2
3 8833E4-02
3 69I6E+02
. 1^
-It')''1-
Figure 6: Velocity Profiles in the Chimney and Breaching Ductwork
Profiles show uneven flow distribution across the ductwork and chimney
-------
Figure^: Velocity Profiles in the Ductwork without Flow Guiding Vanes (Zr
direction, Center of the assembly). Uniform velocities assumed through all the
multiclone tubes.
Profiles show the high velocity region in the top and extremely low velocity region in
the bottom of the duct. A large dead zone is developed for from the last tube to test
port location.
Figure^: Velocity Profiles in the Ductwork without Flow Guiding Vanes (X-
direction). Uniform velocities assumed through all the multiclone tubes.
Profiles show the high velocity region in the top aud extremely low velocity region in
the bottom/center of the duct. y
2 1 170E+OI
1.7G42E + OI
1 41I3E+01
1 05S5E+OI
7 0567E+00
3.S283E + 00
0 OOOOE + OCJ
2 1 IVOEi-01
I.7S42E + 01
1.41 13E+01
I.0585E + OI
7.0567E-I-00
3 5283E+00
O.OOOOE + 00
-------
Figure i? : Velocity Profiles in (lie Ductwork willi Flow Guiding Vanes (Z- direction,
Center of the Duct). 60% Flow is assumed through first half of tubes and 40% flow
through the second hall' of tubes closer to II) fan inlet. Four vanes in throat area.
Four vanvs on the tubes and a splitter vane in tube area duct.
Profiles show the better velocity distribution compared to shown in Figure 14.
f
Figure |0 Velocity Profiles in (he Ductwork with Klo\v (iuiding Vanes (/.- direction,
Center of the Duct), liniform flow is assumed through nil the tubes. Two vanes in
throat area. Four vanes on the lubes anil a splitter vane in tube area duct. Last Tube
is partially covered.
Profiles show the good velocity distribution with slightly lower velocity in the top of
the duct. V
3 30IOE-M)!
2 75I7E+OI
2 201 3E-KJ1
1 6510E+OI
1. 1007E+OI
5 5033E+00
() OOOOE+OO
2 7720E+OI
2 3 100E + 01
I 84SOE+OI
I 3860E+OI
9 2400E4-OU
4 02UUE + 00
U.OOOUE + 00
-------
Figure/I :Velocity Profiles in the Ductwork without Flow Guiding Vanes (X
direction across streamline)
Figure*! Velocity Profiles in the Ductwork with Flow Guiding Vanes (X - direction
across streamlines)
Profiles show the high nnd low velocity regions and non-uniform Dow
distribution nil along the ductwork
Profiles show even flow distribution all across the ductwork.
Note: Extreme Velocity Gnidient
12?
102
81
61
11
20
0
m/s
1 714 I GUI I
; ID i')t nil
j ^ 1 I E+ll I
.11 J 1 E HI I
-------
Figure/? Velocity Profiles in the Circulating Water Pipes
Profiles show the recirculation zone with extremely high velocities at the bottom of the joint.
4 5484E+OO
3 7 90 3 E + 00
3 0322E+00
1 .5161E+00
7 5&06E-01
O OOOOE + OO
-------
Friday, August 29; 8:00 a.m.
Parallel Session B:
Air Toxics Control - General Topics
-------
TRACE METALS REMOVAL FROM RESIDUAL FUEL OILS
H.B. Lange, Ph.D. J.G. Reynolds, Ph.D.
Carnot Lawrence Livermore National Laboratory
15991 Red Hill Avenue, Suite 110 7000 East Avenue
Justin, California 92780-7388 Livermore, California 94550
J. Pirkey B. Toole-O'Neil
Empire State Electric Energy Research Corp. Electric Power Research Institute
c/o Consolidated Edison Co. of New York, Inc. 3412 Hillview Avenue
4 Irving Place Palo Alto, California 94303
New York, New York 10003
Abstract
Potential methods to remove trace metals from residual fuel oils used to fire electric utility
boilers are being assessed. Title III of the 1990 Clean Air Act Amendments may potentially
require control of the emissions of certain metals from oil-fired boilers. Available control means
would include installation of a baghouse or electrostatic precipitator. However, removal of
selected metals from the oils before firing offers a potentially lower-cost alternative. Two little
known and relatively simple oil washing procedures, both of which employ aqueous chemicals,
show promise of achieving significant metal removals at low cost. Initial testing at bench scale
has shown a number of chemicals to be effective to varying degrees in extracting various metals.
More tests are planned to explore the effects of important process parameters. Depending upon
what control requirements are finally enacted pursuant to Title III, further development of this
technology by the electric utility industry may be merited.
Introduction
As implementation of Title III of the Clean Air Act Amendments of 1990 gradually becomes a
reality in various parts of the country, concern about emissions of trace substances from utility
combustion equipment is increasing. Whereas it was previously thought that only the well
known by-products of combustion such as NOX, SOX, particulate matter, CO, and CO2 were
important and would need to be restricted, it is now being recognized that other, often inorganic,
trace substances can be emitted. Federal, state, and local regulators and various industry groups
A054B277.DOC
-------
are assessing whether any significant problems exist as well as potential control methods that
could be applied. As part of this effort, the Empire State Electric Energy Research Corporation
(ESEERCO) and the Electric Power Research Institute (EPRI) are co-sponsoring a program to
evaluate technologies that may be cost-effective in removing trace metals from residual oils used
as boiler fuels by electric utilities.
Although oil use in fossil-fired electricity generation has decreased in recent years, oil-fired
generating capacity remains high, and the utility industry expects oil firing to continue as an
important part of its operation for many years. This expectation is reasonable in light of
historical data. Electric utility residual fuel oil consumption has been cyclical, reflecting the
economics of competing fuel prices. Should current fuel prices shift and natural gas prices rise, a
dramatic increase in oil consumption is likely.
Phase I. Background Research and Technology Screening
Phase I of the program consisted of gathering and analyzing information on the technical
requirements of Title III as they relate to oil-fired utility boilers, the nature and concentrations of
various trace metals in residual fuel oils, and technologies having capability to remove metals
from oils.
Potential Effect of Title III on Oil-Fired Utility Boilers
Title III of the Clean Air Act Amendments of 1990 designates 189 substances as hazardous air
pollutants and mandates that the federal government, through EPA, undertake a series of
regulatory actions. These actions are intended to reduce the amounts of the substances being
emitted and the exposure levels experienced by the general population. The Title III list (which
EPA can amend as necessary) includes metals, volatile organic compounds, halogenated
hydrocarbons, inorganic compounds, pesticides, and radioactive materials.
Many of the 189 listed substances are hydrocarbon compounds whose formation may be
minimized by good combustion practices. However, a number of Title III species are metallic
and hence cannot be destroyed even by a well-adjusted combustion process. Table 1 shows
metallic species from the Title III list.
In general, the emissions standards set under Title III for significant emissions sources must
reflect the maximum degree of emissions reduction achievable considering the cost to achieve
the reduction and other health, environmental, and energy impacts. Since conventional practice
to reduce particulate emissions involves such devices as baghouses and electrostatic precipitators
(ESPs), a utility affected by Title III requirements could expect to be required to have these
devices installed on its oil-fired boilers or to utilize other technologies providing similar
emissions reductions. This is significant to many utilities using oil fuel since less than 20
percent of oil-fired utility boilers in the U.S. are equipped with back-end particulate control
devices.
A054B277.DOC
-------
Table 1
Title III Metallic Species
Antimony compounds
Arsenic compounds
Beryllium compounds
Cadmium compounds
Chromium compounds
Cobalt compounds
Lead compounds
Manganese compounds
Mercury compounds
Nickel compounds
Phosphorous
Selenium compounds
Trace Metals Found in Crude Oils and Relationship to Residual Fuel Oils
Although crude petroleum consists primarily of hydrocarbons, most oils contain measurable
amounts of many metals. Nickel and vanadium usually are the most abundant metals, but many
others are frequently present in parts per billion (ppb) to parts per million (ppm) concentrations.
The nature of the metals and their relative concentrations can give information about the origin
and maturation of the petroleum.
In a petroleum refinery, crude oil is distilled and otherwise processed to produce various
products (gasoline, jet fuel, etc.). In general, distillation separates higher value, lower boiling
point (distillate) materials from higher boiling point (residual) materials. Residual fuel oil used
by electric utilities contains the least valuable, highest boiling point liquid fractions from crude
oil. Varying amounts of distillate material may be blended back into the residual material to
yield residual fuel oil capable of meeting various fuel purchase specifications.
Analyses of crude and residual oils are widely available for such common parameters as
hydrogen, nitrogen, ash, and sulfur contents, specific gravity, viscosity, and heat content. Data
pertaining to trace metals also are available but are more limited. Typical values for residual oils
in particular are difficult to establish because of the variety of crudes and refining processes that
are used as well as various amounts of blending oils (cutter stock) used to produce a finished
residual fuel oil with specific physical and chemical properties.
Table 2 presents neutron activation analysis results of some trace element constituents in crude
oils from four widely separated oil fields'12 It is clear that orders-of-magnitude differences can
be found for the same element in different crudes.
A054B277.DOC
-------
Table 2
Trace Elements in Some Area-Specific Crude Oils (ppm)
Element
Arsenic
Beryllium
Cadmium
Chromium
Cobalt
Copper
Lead
Manganese
Mercury
Nickel
Phosphorous
Selenium
Zinc
California
0.655
-
0.640
13.5
0.93
1.20
23.1
98.4
-
0.364
9.76
Libya
0.777
--
0.0023
0.032
0.19
0.79
-
49.1
-
1.10
62.9
Venezuela
0.284
--
0.430
0.178
0.21
0.21
0.027
117.0
-
0.369
0.692
Alberta
0.0024
--
--
0.0027
--
0.048
0.084
0.609
--
0.0094
0.670
The Chemistry of Metals in Petroleum
To understand the difficulties in removing trace metals from petroleum and to evaluate the
possible technologies for doing so, an understanding of the chemistry of metals in oil is
necessary3 The chemical forms and environment of the metallic species determine the
effectiveness or ineffectiveness of removal treatments and the collateral damage done by the
treatment to the rest of the petroleum fraction. Removing the contaminant metals is a complex
problem for several reasons:
The metals are chelated or complexed in large organic molecules that are completely miscible
in the petroleum, making their separation and removal difficult;
The amounts of metals are very small, making it difficult to pinpoint the metals for removal
without simultaneously removing or altering other parts of the petroleum;
The metals can exhibit catalytic effects during processing, possibly interfering with the
removal process;
Some of the metals are catalyst poisons, making catalytic demetallation complicated.
Porphin, shown below, is derived from chlorophylls and hemoglobin in the plants and animals
that contributed to the formation of crude oils . Porphyrins, of which many examples are found
in petroleum, differ in how the basic porphin structure is modified5
A054B277.DOC
-------
Porphin
The two major classes of importance for this work are the porphyrin metals (metalloporphyrins)
and the nonporphyrin metals (transition metal complexes/chelates)6. It has generally been
observed that metals are concentrated in the most polar or least distillable petroleum fraction, but
short of removing this entire fraction, there is no simple single way to remove metals from both
the porphyrin molecules and the nonporphyrin molecules. Even the most common metals (nickel
and vanadium) appear to exist in porphyrin and nonporphyrin forms, so even if, for example, a
method could be developed specifically for removing metals from metalloporphyrins, this would
not completely remove all of a given metal from the oil.
The significance of the chemical environment of trace metals to attempts to demetallize
petroleum is two-fold. First, the processing of crude oil to produce refined products removes the
lighter molecular weight components in distillate fractions, concentrating the heavier molecules
into the residuum. This residuum, with'or without cutter oils being added, becomes the residual
fuel oil used by utilities. Since the trace metals tend to occur in the least distillable segments of
the crude oil, the metals in the crude largely are transferred into the residual fraction. Second,
very little of the metals in petroleum exist as inorganic salts and solid particles, making simple
physical and mechanical separation processes ineffective for demetallation. Effective
demetallation requires either (1) the removal of metal atoms from molecules in which they are
chemically bound, or (2) the removal of the entire molecular matrix that contains the metal
atoms. Option (1) is chemically difficult, and option (2) is expensive due to the loss of valuable
hydrocarbon with the metals.
Technology Screening
Phase I of this study was directed toward identifying commercial or near-commercial
technologies and processes that might be used to demetallize crude or residual fuel oils.
A054B277.DOC
-------
Thirty-nine technologies were reviewed during the screening process. These technologies fell
into three major categories: catalytic processes, solvent extraction processes, and miscellaneous
processes. Results in each category were as follows:
Catalytic processes. No promising new catalytic processes or technologies for residual fuel
oil demetallation were discovered. Not surprisingly, the catalytic processes are oriented
toward upgrading the heavy petroleum fractions to lighter products. Due to the elevated
temperature and pressure needed for catalytic processing, there always will be some cracking
and hydroconversion occurring whether that is desired or not. A number of heavy oil
upgrading processes that are known to remove nickel and vanadium are commercially
available. Two such processes that are widely used are the ABB-Lummus LC-Finingฎ
process and the IFF (France) HYVAHLฎ process.
Extraction processes. The variety of solvent types available is well represented in the
different approaches found in extractive demetallation processes. Conventional paraffmic
solvents, binary solvent systems, and supercritical extraction all are represented. Several
commercially available extraction process are known to remove nickel and vanadium, and
one in fairly common use is the Kerr-McGee Roseฎ process. Two other extraction processes
were identified as having promise for residual or crude oil demetallation using mild
processing conditions and relatively simple equipment: a process that has been commercially
demonstrated (Process 'A') and a process currently under development (Process 'B').
Miscellaneous technologies. These included a biological process using bacteria to remove
sulfur, nitrogen, and nickel from heavy oil, polymeric membrane filtration, ion exchange, and
others. None of these processes involve commercial or near-commercial technologies
appropriate for large-scale demetallation of residual or crude oil.
Processes A and B, both of which are relatively simple aqueous washing processes, are similar
but different. Process A was developed primarily to remove calcium (and secondarily to remove
iron) from certain heavy crude oils, although it also has been tested in the laboratory on vacuum
residuum material. In practice, it can be implemented as a simple modification to conventional
desalting operations in a refinery.
Desalting is a common crude oil processing step to remove inorganic (especially Na) salts that
can poison catalysts and cause corrosion and fouling of downstream equipment. Mechanical,
chemical, and electrical approaches have been taken by various process developers, but all'
essentially involve heating the oil at pressures sufficient to prevent vapor loss (typically 200-300
ฐF and 50-500 psig), then separating the aqueous and hydrocarbon phases, possibly by settling in
a large vessel. Chemicals or an electrostatic field may be used to break stable emulsions in the
oil.
Following development of Process A in the laboratory, a 50,000 bpd unit processing crude oil
was installed as one of several planned process additions and changes in a Chinese refinery. The
process achieved 50-80 percent calcium removal (20-30 ppm inlet; 3-10 ppm outlet) and 50
A054B277.DOC
-------
percent iron removal (5-15 ppm inlet; 4-8 ppm outlet). The unit ran successfully for several
weeks, but the client stopped using it because the other refinery modifications were cancelled.
Less is known about Process B, which is still being developed. However, limited information
indicates that it is both technically and economically similar to Process A.
Economic Analyses of Candidate Technologies
Since conventional technologies for removing trace metals from flue gases use baghouses or
ESPs, these options are the benchmarks against which any competitor must be compared.
Comparative economic analyses were performed for Processes A and B as well as three
commercially available technologies identified in the technology screening phase.
In an earlier EPRI project, detailed cost analysis of particulate control technologies (baghouses
and ESPs) had been performed according to the EPRI Technical Assessment Guide (TAG)
methodology Analysis of the oil demetallation technologies used assumptions and economic
parameters consistent with those in the previous analysis to permit the results to be compared on
a consistent basis; however, the cost figures in the older analysis were updated. Detailed
explanations of the methodology and parameter values are provided in Reference 7.
The results of the comparative cost analyses for the various technologies are summarized in
Table 3. Baghouses and ESPs both can provide particulate control at a cost of about 2.6-3.3
mills/kWh (an increase of 5-6 percent in the base cost of electricity generation). The
commercially available oil cleaning options have significantly higher total costs than baghouses
or ESPs; in addition, the capital costs for a single installation can be very high. These higher
costs result from high capital and operating costs and from losses of feedstock material during
processing (either direct physical losses, as in solvent extraction, or losses from the residual oil
boiling range, as in hydroconversion).
Processes A and B show very attractive total costs, perhaps an order of magnitude lower than the
costs for baghouses and ESPs. In addition, they use common process equipment at unremarkable
pressures and temperatures, making them feasible for use by utilities or intermediate fuel
suppliers if oil refiners will not produce demetallized fuel oils.
The cost comparison presented in Table 3 necessarily is preliminary in nature. Many questions
about the performance of the alternative technologies in removing trace metals other than the
ones for which they were intended must be answered. If their demetallation potential is
confirmed, significant process development and demonstration work will be needed before they
can be considered commercially viable choices for utility use.
Phase II. Bench-Scale Tests
In Phase II of the project, which is presently in progress, bench-scale tests of Processes A and B
are being performed. Each of these processes can use any one (or more) of a number of different
A054B277.DOC
-------
chemical reagents. Other important processing parameters include temperature, contact time and
acidity. These tests are focusing on seven trace metals that are considered to be representative of
metals that may be potentially significant with regard to possible Title III regulations. These are
arsenic, cadmium, chromium, mercury, nickel, selenium, and vanadium. The primary objective
of the bench-scale tests is to assess, for each process, the trace metals removal effectiveness of a
number of reagents considered most likely to be successful. A secondary objective is to
determine the effects of important parameters that will affect process economics. A first round
of testing has been completed, and the following will report on the methodology and results of
those tests as well as planned future testing activities.
Table 3
Comparative Costs of Trace Metals Control Technologies
October-1996 Dollars
Technology 1
Baghouse (250 MW)
ESP (SCA=400 ft2/103 acfin)
Process A (applied to crude* 1})
Process B
ROSEฎ process (solvent extraction)
LC-Fining (expanded cat. bed)
IFF Hyvahlฎ (fixed cat. bed)
First Year Costs
tf/gal fuel oil mills/kWh
2.1-2.9
2.7
0.3-1.3 0.2-0.9
0.3-2.2 0.2-1.6
7.9 5.6
>16.1 >11.5
30-Year Levelized Costs
^/gal fuel oil
0.6-1.9
0.4-3.6
10.0
>17.4
>2.3
mills/kWh
2.8-3.3
2.6
0.2-0.8
0.2-1.5
4.2
>7.4
>9
(1) Processing residual oil might reduce costs substantially.
Note: The processes shown in this table include mature technologies and processes that are still under
development. The cost estimates use data from equipment manufacturers, literature, and process licensers
and developers. The overall cost uncertainties correspond to EPRI TAG Class I or Class II cost estimates.
Method Used to Analyze Trace Metals in Oils
The method used to analyze the concentrations of trace metals hi the oils before and after
washing is the foundation of a project of this nature, in which metals must be measured in a
relatively difficult heavy oil matrix at levels as low as a few parts per billion (ppb). Based on
years of experience in analyzing trace metals in heavy oils on the parts of EPRI, ESEERCO and
Carnot, the instrumental neutron activation analysis (INAA) method was chosen. The INAA
A054B277.DOC
-------
analyses were performed by the Massachusetts Institute of Technology (MIT) Nuclear Reactor
Laboratory under the direction of Dr. Ilhan Olmez.
Selection of a Residual Fuel Oil for Testing
In selecting a residual fuel oil sample on which to perform the bench-scale tests, a goal was to
obtain an oil sample having concentrations of the seven trace metals of interest at least ten times
their respective analytical detection limits. This would permit a removal of up to 90 percent of
any metal to be verified. Residual fuel oil samples were solicited from a number of oil-using
utilities. Preference was given to higher-sulfur oils, thought to generally contain higher levels of
trace metals. Pre-cleaned sample bottles, meeting an EPA specification for trace constituents
sampling, were sent to each participating utility along with instructions for acquiring an oil
sample representative of as-received oil while avoiding contamination. Five oil samples were
obtained. Their analyses for the seven trace metals of interest are shown in Table 4. Oil B had
either the highest or a comparable level of each of the seven trace metals, and was thus selected
for the tests. Unfortunately, concentrations of cadmium in all five oils were below the analytical
detection limit for this metal. Thus the bench-scale tests were inconclusive for cadmium.
Table 4
Trace Metal Concentrations in Residual Fuel Oil Samples Received from Utilities
Sample ID A B C D E
Arsenic (ppb)
Cadmium (ppb)
Chromium (ppb)
Mercury (ppb)
Nickel (ppm)
Selenium (ppb)
Vanadium (ppm)
67ฑ6
<3.7
180ฑ70
0.74ฑ0.22
60ฑ22
150ฑ30
155ฑ10
82ฑ8
<9
280ฑ20
4.2ฑ1.0
215ฑ60
120ฑ30
230ฑ15
57ฑ6
<7.2
84ฑ12
1.3ฑ0.8
<30
130ฑ40
82ฑ6
110ฑ10
<5.9
240ฑ20
3.4ฑ0.9
<30
82ฑ22
160ฑ11
10ฑ1
<3.4
190ฑ20
1.1ฑ0.8
<30
13ฑ10
185ฑ13
Oil Washing Tests
The oil washing tests were performed using laboratory glassware commonly used for mixing and
phase separation. The tests were designed to achieve good dispersion of the oil and aqueous
reagent phases. Acidity of the aqueous phase was controlled at levels known to be conducive of
good metal removal in each case. Liberal conditions of reagent dosage, temperature and contact
time were used, i.e., not necessarily those that would ultimately be selected for acceptable
process economics. For Process A, four reagents (Al through A4) were tested; and for Process
B, six reagents (B 1 through B6) were tested. Each test was run in triplicate.
A054B277.DOC
-------
Results
Trace metal removals achieved in the bench-scale tests are summarized in Table 5. Each metal
for which one or more reagents was effective is included in the table. As mentioned above, the
tests were inconclusive for cadmium due to that metal being below its analytical detection limit
in the pre-wash oil. None of the reagents was effective in removing arsenic. For each metal
listed in the table, the pre-wash concentration in the oil is shown followed by post-wash
concentrations and percentage removals for each effective reagent. Except as noted on the table,
the post-wash concentrations are each the average of three triplicate tests. The range shown for
the percentage removal in each case results from the compounding of the uncertainties in the pre-
wash and post-wash analyses, i.e., the lower percentage removal figure results from dividing the
post-wash concentration less its uncertainty by the pre-wash concentration plus its uncertainty,
and the higher figure results from applying the uncertainties in the opposite senses.
It is encouraging that significant removals are indicated for five of the six metals that were
detected in the pre-wash oil (i.e., excluding cadmium, which was not detected). Unfortunately,
the Process B tests were inconclusive for arsenic, mercury, nickel and selenium due to apparent
contamination problems and/or high bias of the post-wash analyses. This matter is now being
investigated.
Conclusions
Two little-known and relatively simple aqueous oil washing processes show considerable
promise of being able to remove significant fractions of trace metals from residual fuel oils.
Depending upon which trace metal emissions are ultimately affected by Title III regulations and
what removal efficiencies are required, this approach may prove to be a cost-saving alternative to
retrofitting back-end particulate control devices such as baghouses or electrostatic precipitators.
However, only one round of bench-scale testing of these processes has thus far been completed,
and much remains to be learned about their removal efficiencies for various trace metals and the
effects of important process variables.
Planned Future Activities
Near-term activities will include: (1) investigation of Process B results for arsenic, mercury,
nickel and selenium to eliminate apparent effects of contamination and/or high analytical bias, if
possible, and (2) a second round of tests to examine the effects of important process parameters
on removal efficiencies of the most promising reagents.
Longer-term, since the economics of this approach to meeting the requirements of Title III are
dependent upon which metals will ultimately have to be controlled and what degrees of control
will be required, the extent and direction of future development of this technology will probably
depend mainly on the requirements of Title III regulations as they take shape. As this picture
becomes clearer, if the economics of these processes in meeting Title III requirements appear
attractive relative to retrofitting back-end control devices, additional bench-scale testing followed
by pilot-scale testing would probably be merited.
A054B277.DOC 10
-------
Table 5
Summary of Trace Metal Removals Achieved in Bench-Scale Testing"
Trace Metal Pre-Wash
Analysis
Chromium (ppb) 280ฑ20
Mercury (ppb) 4.2ฑ1.0
Nickel (ppm) 215ฑ60
Selenium (ppb) 120ฑ30
Vanadium (ppm) 23 0ฑ 1 5
Reagent
Al
A2
B4b
A3
A4
Al
A2
A3
A4
Alb
A2
A3
A4
Al
A2
A3
A4
Bl
B2
B3
B4
B5C
B6b
Post-Wash
Analysis
230ฑ40
257ฑ27
143ฑ100
1.39ฑ0.69
2.15ฑ0.94
77.0ฑ23.5
137ฑ33.3
109ฑ27.0
87.7ฑ25.3
80.7ฑ21.0
97.7ฑ18.0
75.3ฑ14.0
91.0ฑ19.0
133ฑ10
137ฑ10
130ฑ10
137ฑ10
202ฑ12
203ฑ16
211ฑ16
143ฑ6
84ฑ4
94ฑ4
Removal, %
-4 to 37
-9 to 23
7 to 86
35 to 87
3 to 77
35 to 81
-10 to 62
12 to 70
28 to 78
-13 to 60
-29 to 47
Ito59
-22 to 52
33 to 50
32 to 48
35 to 51
32 to 48
Oto22
-2 to 24
-6 to 20
31 to 44
59 to 67
54 to 63
Notes:
a. None of the reagents in either Process A or Process B was effective in removing arsenic.
b. Average of two tests.
c. Singular test.
A054B277.DOC
11
-------
Acknowledgement
The Lawrence Livermore National Laboratory work was performed under the auspices of the
U.S. Department of Energy, Contract W-7405 ENG-48.
References
1. Edwards et al., Trace Metals and Stationary Conventional Combustion Sources. EPA
Contract No. 68-02-2608 (IERL/RTP) (March 1979).
2. Battelle-Columbus Laboratories, Fuel Contaminants - Volume I. Chemistry. EPA 600/2-76-
177a, EPA-ORD (IERL/RTP) (July 1976).
3. Speight, J.G. The Chemistry and Technology of Petroleum. 2nd Edition, Marcel Dekker,
Inc., New York (1991).
4. Tissot, B.P., and D.H. Welte, Petroleum Formation and Occurrence, Springer-Verlag, Berlin
(1978).
5. Quirke, J.M.E. in Metal Complexes in Fossil Fuels. R.H. Filby and J.F. Branthaver, Eds.,
American Chemical Society Symposium Series, Vol. 344 (1987).
6. Yen, T.F. in The Role of Trace Metals in Petroleum. T.F. Yen Ed., Ann Arbor Science,
Michigan (1975).
7. Gaikwad, R.P. and D.G. Sloat, Economic Evaluation of Particulate Control Technologies.
Volume I: New Units. EPRI Report No. TR-100748 (September 1992).
A054B277.DOC 12
-------
THE USE OF COAL CLEANING
FOR TRACE ELEMENT REMOVAL
Clifford E. Raleigh, Jr. and David J. Akers
CQInc.
One Quality Center
Homer City, Pennsylvania 15748
Barbara Toole-O'Neil
Electric Power Research Institute
Palo Alto, California 94303
Abstract
Under Title HI of the 1990 Clean Air Act Amendments, the U.S. Environmental Protection
Agency is required to assess the release of toxic substances by electric steam generating
stations. Title El defines 189 substances as hazardous air pollutants (HAPs), including 14
elements that are commonly found in coal in trace concentrations. Coal cleaning is an effective
method for decreasing the trace element content of coal, thereby reducing HAPs. The key to
realizing the full potential of precombustion air toxic control technology is to first learn the
fundamentals of trace element mode of occurrence and the mechanisms of removal during
cleaning.
Introduction
Coal cleaning is a technology that can solve a broad array of environmental problems associated
with older, state-of-the-art, and future electric generating stations. By increasing thermal
efficiency and reducing parasitic power requirements, the use of coal cleaning can reduce all
types of power plant emissions per unit of electricity produced, including SO2, NO,, CO2, and
hazardous air pollutant precursors (HAPs). Currently, more sulfur (and related SOj) is removed
via coal cleaning than by all post-combustion technologies combined.
Coal cleaning offers a number of advantages as a HAPs control technology. It is an effective
and relatively inexpensive method of controlling HAPs emissions. The technology suits all
power generation systems because it addresses the feedstock and not plant hardware,
eliminating the need for direct capital investment by coal users. Also, cleaning increases the
healing value of delivered coal while reducing transportation, handling, maintenance, and ash
disposal costs and it may be combined with other emissions reduction technologies to further
reduce the quantity of HAPs in flue gas.
-------
While coal cleaning is a mature technology, in the past it has been used only for the
comparatively simple purpose of removing ash-forming and sulfur-bearing minerals. The
application of this technology to HAPs control will require a sophisticated approach based on
understanding the mechanisms of trace element removal during cleaning.
The trace elements named as HAPs in the 1990 Amendments to the Clean Air Act can occur in
coal in numerous forms and have various associations. For example, arsenic may be primarily
associated with late-stage (epigenetic) pyrite; chromium may be associated with clays; mercury
may occur predominately in epigenetic pyrite, but also may have some level of organic
association; and selenium may be organically-bound or associated with pyrite or accessory
minerals such as clausthalite. In addition to having multiple modes of occurrence, the range of
concentration of trace elements in coals varies remarkably. The United States Geological
Survey (USGS) has reported that the concentrations of arsenic in coal range from not detected
(ND) to 2,200 ppm; for chromium - ND to 250 ppm; for mercury - ND to 5 ppm; and for
selenium - ND to 150 ppm [1].
Fortunately, most of the trace elements of concern in coal are either associated primarily with
mineral matter or exist as minerals, both of which can be removed very effectively using coal
cleaning technologies. As a by-product of cleaning to remove ash-forming and sulfur-bearing
minerals, the reduction of trace element content of coal is well documented. For example,
arsenic and mercury reductions during commercial cleaning have been reported to be as high as
85% and 78%, respectively [2].
The extent of trace element removal that may be attained during cleaning will depend on the
physical characteristics of the coal being cleaned (e.g., mode of occurrence of the trace
elements, size distribution, and degree of liberation of mineral matter), the method of cleaning,
and the relative intensity of cleaning [3]. Current research is focusing on these parameters in
order to better understand the fundamentals and the mechanisms of trace element removal
during cleaning.
Recent Results of Trace Element Removal Studies
In work funded by the U.S. Department of Energy's - Federal Energy Technology Center
(DE-AC22-95PC95153), the Electric Power Research Institute, and industry participants, four
coals were characterized in the laboratory for geochemical and mineral processing
characteristics such as coal mineralogy, theoretical washability, and liberation behavior. These
coal samples are from major producing seams in the following regions: Northern Appalachian,
Powder River Basin, Southern Appalachian, and Eastern Interior.
Samples of each raw coal, as well as cleaned minus 28 mesh samples of the bituminous coals
and natural minus 28 mesh fines from the Powder River Basin coal sample, were submitted to
the USGS for determination of the mode of occurrence of the HAPs elements. This
determination was accomplished through a series of leaching steps (see Figure 1) followed by
analysis of residues and leachates. Scanning electron microscopy, microprobe analysis, and x-
ray diffraction studies complemented this work.
-------
Chemical Fractfonatton Procedure
Ammonlm HytixJtO. HydnAnt: NbfcAcU
AeoWolN AdUZNlMcM: **ซ%tetoK ZNfcadm:
tadปKrซป tactadaal kBctadcod tatadcool
ml mm tปp 1 iramlBpZ
stool
On**ป% Bound AcUSobbKSals SbM SulBe.
Umw (CMonta.SuM> Mudto
eUKUti quMtz
ป) ohrt
(CMonta.SuM> Mudto pyrtean
ndieUKUti quMtzmi Soin
Figure 1
Leaching Procedure for Trace Element Mode of Occurrence Determination
Significant advancements have been made in understanding how trace elements occur in coal:
Primary Mซteป of Occurowco of Trace Elements . Most of the twelve elements studied are
Anttmonjr -, ' - Organic and Pyrfte associated with sulfide and silicate minerals. For
Baryta!." - ' - StotScS^Si example, arsenic is associated with pyrite while
suffides chromium occurs with illite clay.
Cadmium Sphateftte
Chromkปn PBte clay
Cnปปซ . , Oxides, Ofganc, SScatss, Many of the elements in these coals have multiple
andSulfides ,J , . , ,. . r
fluorine ', daysmuorites" modes ot occurrenceincluding organic.
Lead , x Catena and Organic
Manganese Ca?bonatesandSi8cates , .
Mercwy - Pyrite and Oiganfc Mercury and selenium may be associated with
f**61 " a^ffltoanffi'SBk!ateS' very fine (unliberated) pyrite, organic fractions of
Seteniuai Pyrits, Orga^ SBieates, the coals, Or both.
andSelemtes
In general, the coal characterizations in this work have generated the highest quality trace
element washability database yet developed. For example, the poorest mass balance closure for
the uncrushed size and washability data for mercury on all four coals is 8.44% and the best is
0.46%. This indicates an extremely high level of reproducibility of the data.
Physical coal cleaning data from this work indicate that the level of reduction of arsenic,
mercury, and selenium during cleaning are related to the mode of occurrence of the trace
element and the way it is cleaned by a specific device. As illustrated in Figure 2, the level of
arsenic reduction that may be attained when the bituminous coals are cleaned will be higher
than that when the Powder River Basin coal is cleaned. This is probably because about 70% of
the arsenic in the bituminous coals is found with pyrite as compared to only about 25% in the
subbituminous coal. And since arsenic is found primarily with pyrite in the bituminous coals,
there is a marked difference in the level of arsenic reduction that may be achieved through
cleaning in gravity-based versus surface property-based devices.
-------
El
"
il
NA SA El
|50
ฃ40
i30
| ao
110
NA
Q] Froth FloBtlon Cteanug
NA-NoittwmApiJalKMan SA-Southem Appstachbn Et-ฃnum Marto PRB-Powdar Rtar Baaki
Figure 2
Trace Element Reduction at 90% Energy Recovery during Physical Coal Cleaning
(28 Mesh x 0 Size Fraction of Uncrushed Coal Samples)
For selenium reduction, surface property-based cleaning actually outperforms gravity-based
cleaning on the Southern Appalachian coal. This is the only coal of the four that has selenium
associated with silicates, and froth flotation can sometimes remove fine, sheet-like particles such
as silicates more efficiently than can gravimetric devices.
Knowing the concentration, theoretical washability, and mode of occurrence of various trace
elements in coals will allow engineers to design specific coal cleaning flowsheets to remove the
elements of concern efficiently and economically [3]. Figure 3 serves to illustrate the
importance of this type of information. Note that an area low in arsenic (as indicated by the blue
shading) is visible in the upper portion of a pyrite grain from the Southern Appalachian coal
sample. Not only does this provide direct evidence that arsenic is associated with pyrite (which
can be removed very effectively by physical coal cleaning), but it also shows that the arsenic is
not evenly distributed throughout the host mineral. Thus, a specific type of pyrite (euhedral
rather than framboidal in this case) must be removed by the cleaning process if arsenic reduction
is to occur.
Figure 3
Arsenic Map of a Pyrite Grain in the Southern Appalachian Coal Sample
-------
The reduction of mercury content from the Northern Appalachian, Southern Appalachian, and
Eastern Interior (Illinois Basin) coals using conventional cleaning techniques is promising. As
shown in Figure 4, cleaning may reduce the mercury content of the SA coal by as much as 50
percent without sacrificing energy recovery excessively. A high mercury reduction via coal
cleaning can be obtained for the NA and El (ILL) coals; however, the trade-off between mercury
reduction and energy recovery is less favorable. As with arsenic, physical cleaning of the PRB
coal to reduce its mercury content is probably not practical. In fact, it is possible that cleaning
may actually concentrate mercury in this subbituminous coal.
Interestingly, Figure 4 also shows that all of the coals examined in this work have a limit to
which mercury content can be reduced by conventional physical processesabout 0.08 to
0.09 ppm. This may represent an incapability of conventional physical coal cleaning techniques
to reduce mercury content beyond this level, unless the coal is first crushed to liberate additional
mercury-bearing minerals. The use of advanced coal cleaning techniques, possibly in
combination with crushing, may extend this limit, especially since mode-of-occurrence work at
the USGS suggests that about two-thirds of the mercury in these four coals is bound with mineral
matter.
0.25
0.20
E
r
I
o
I
03
0.10
0.05 -
0.00
10 20 30 40 50 60 70 80
Energy Recovery (%)
90 100
0.25
0.20
0.15
0.10
0.05
0.00
Figure 4
Mercury Washability of the Four Program Coals
-------
Other laboratory-based work on these coal samples included initial investigations of the use of a
number of chemical solvents and biological treatments for reducing the trace element content of
the project coals. The solvents and treatments screened in this study included acid and alkaline
solvents, oxidative and chelating agents, bacterial and algal agents, and ultrasonics.
Bench-scale testing at Howard University revealed that the levels of mercury and other HAPs
elements in the project coals may be decreased over 50% using a new chemical cleaning
approach. While bench-scale test data are very encouraging, the development of accurate cost
and performance data requires testing in a larger-scale, continuous system. Initial estimates
place the cost of this process at less than $3.00/ton. This technology, in combination with the
improved use of physical cleaning technologies, could greatly reduce HAPs emissions in the
U.S. at less cost and lower environmental risk than post-combustion control measures.
Also during this recent work, sets of algorithms were developed to predict the amount of trace
element reduction that may be expected during conventional and advanced coal cleaning to
enable improved plant design and operation. Using regression analyses, project engineers
produced equations that predict how the concentrations of arsenic, chromium, cobalt, fluorine,
lead, manganese, mercury, nickel, and selenium relate to parameters such as ash content, sulfur
content, sulfur forms, initial trace element content, trace element mode of occurrence and
particle size during gravity-based and surface property-based coal cleaning operations.
Engineers also investigated the impact of crushing the coals to enhance the removal of trace
element-bearing minerals during cleaning.
All of the equations have ash content parameters and many include sulfur and trace element
mode of occurrence as predictors. The equations were analyzed to assess accuracy and fit and
to compare the predicted values to the applicable ASTM analytical reproducibility bandthis
indicates that the values were predicted as accurately as can be measured. As shown in
Figure 5, most of the low-end values for mercury and selenium (where clean coal analyses
typically fall) are found within the ASTM band, indicating that the equations have good
accuracy.
Additional work is required to:
Verify the applicability and performance of the predictive equations.
Test the chemical process for mercury removal at larger scale in continuous mode to obtain
data on the dynamics of the system and to allow evaluation of the process on larger-size coal.
Evaluate the relative leaching stability of trace elements in samples of cleaning plant waste
and power station ash to avoid dealing with environmental issues in a piecemeal fashion.
This last point is important. If a trace element is less leachable in coal cleaning refuse than in
fly ash (which is likely to be the case since the element has existed in its naturally-occurring
form over geologic time), shifting a trace element from a coal cleaning refuse matrix to a fly ash
may create additional groundwater pollution.
-------
3500
J300
1200
E
1-100
100 200 300 400 500
Measured Mercury Concentration (ppb)
600
2345
Measured Selenium Concentration (ppm)
Figure 5
Analysis of the Accuracy of the Mercury and Selenium Equations for Predicting Trace
Element Reduction During Cleaning of Uncrushed Coals.
-------
Summary
Based on an assessment of post-combustion mercury control options by the Electric Power
Research Institute [4], reliable and cost-effective mercury control methods for utility boilers
have not yet been developed. However, work by CQ Inc. and others is demonstrating that in
some cases, conventional methods of cleaning coal can remove over 50% of the mercury. The
use of advanced coal cleaning methods can remove even more mercury. Given the long lead
time likely required to develop cost-effective, post-combustion mercury control technologies
and the relatively high effectiveness of existing cleaning technologies, coal cleaning is likely to
be a very important part of any near-term efforts to reduce mercury emissions from coal-fired
boilers.
Building the knowledge to accurately and reliably predict the extent to which specific cleaning
processes will remove trace elements from any given coal will allow coal producers and users to
control the disposition of trace elements, ensuring that these elements do not cause air or
ground-water pollution. It will also aid in the selection of effective, economical cleaning
methods for these coals by allowing coal cleaning engineers to identify proven and promising
methods for the design of new or the retrofit of existing coal cleaning plants that will control
HAPs precursors.
Acknowledgements
The authors wish to extend thanks to the Department of Energy's-Federal Energy Technology
Center, the Electric Power Research Institute, and our numerous industry partners for their
continued support in evaluating the potential of trace element removal from coal via cleaning.
References
1. R. B. Fmkelman, "Modes of Occurrence of Environmentally-Sensitive Trace Elements in
Coal," Environmental Aspects of Trace Elements in Coal, Kluwer Academic Publishers,
1995, pp. 24-50.
2. D. J. Akers, "Coal Cleaning Controls HAP Emissions," Power Engineering, Pennwell
Publishing Company (1996).
3. C. E. Raleigh, Jr. and D. J. Akers, "Coal Cleaning: An Effective Method of Trace Element
Removal, Proceedings of the 11th Annual International Pittsburgh Coal Conference,
Pittsburgh, Pennsylvania, pp. 96-101 (1994).
4. Change, R. & G. Offen, "Mercury Emissions Control Technologies: An EPRI Synopsis,"
Power Engineering, Vol. 99, Issue 11, pp 51-57 (1995).
-------
PREDICTION OF TRACE ELEMENT PARTITIONING
IN UTILITY BOILERS
F. J. Frandsen
Department of Chemical Engineering, Technical University of Denmark,
DK-2800 Lyngby, Denmark
J. J. Helble
Chemical Engineering Department, University of Connecticut, Storrs, CT, USA
T. A. Erickson and V. Kuhnel
Energy and Environmental Research Center, University of North Dakota,
Grand Forks, ND, USA
Abstract
The phase partitioning and speciation of the trace elements As, Cd, Cr, Hg, Ni,
Pb, and Se, in a model system based on a demonstration-scale pressurized
FBC, firing a Pittsburgh #8 bituminous coal, is mapped as a function of
temperature, by use of Global Equilibrium Analysis (GEA). Ideal gas and pure
condensed phases are assumed and the presence of a calcium sorbent material
for sulfur capture, is taken into account.
Results from four thermodynamic packages (MINGTSYS, NASA-CET89, FACT,
and SOLGASMIX) are compared over the temperature range [700 - 2000 K], at
1 and 20 atm.. For the trace elements, As, Hg, and Se almost identical
equilibrium distributions were predicted by the four packages, while for Cd, Cr,
Ni, and Pb, differences in the output from the four packages were observed,
primarily due to a lack of thermochemical data for important chemical species,
and/or scattering in the four thermochemical input-data sets utilized, by the four
thermodynamic packages. Calculations in which the standard Gibbs energy of
formation of different cadmium species was varied confirmed that differences in
the thermochemical input-data affect the final distribution of species.
This paper contains an outline of the trace elements chemistry mapping based
on the output from the four thermodynamic packages. A comparison between
the results and a detailed discussion of the application and limitations of GEA for
trace element chemistry mapping in combustion systems is provided.
-------
1. Introduction
The elements contained in fossil and biomass fuels can be grouped into three
concentration levels: 1) the major elements, C, O, H, S, and N, comprising the
combustible organic matrix of the fuel; 2) the ash forming elements, Al, Ca, Fe,
K, Mg, Na, and Si, typically present in the concentration range from about 1000
ppmw to a few %(w/w) on a dry fuel basis; and 3) the trace elements (including,
e.g., As, B, Cd, Cr, Hg, Ni, Pb, Se and others) present typically in concentrations
below 1000 ppmw [1]. Several of these trace elements may be vaporized during
devolatilization and pyrolysis to later nucleate or recondense on the surface of fly
ash particles during subsequent cooling of the flue gases [2 3]. Full-scale
measurements have revealed significant amounts of some trace elements (e.g.,
B, Hg, and Se) in the flue gas leaving the stack [4 - 8]. Direct gaseous emission
of these trace elements is undesired because of suspected toxicological effects
on the environment and potential genetic or biological changes in living creatures
[1,3,5,9].
A mechanistic model for trace element transformations in thermal fuel conversion
systems will require detailed information on the release of trace elements (i.e.
heat-up, devolatilization and pyrolysis of the fuel), residual ash and aerosol
formation (i.e., char burn-out and fragmentation, coalescence and fragmentation
of mineral grains, gas phase supersaturation, cluster formation, nucleation and
coagulation), gas phase reactions of trace element species and subsequent
recondensation of trace elements (i.e., gas- and particle-particle interactions, and
mass transfer from the bulk gas to fly ash particles).
Only a very few kinetic data for trace element transformations in hot flue gases
are available. Thus, as a first approach, Global Equilibrium Analysis (GEA) has
been used for several years in order to understand the various subprocesses in
thermal fuel conversion systems. Frandsen et al. [10] have provided an
introduction to the application of GEA on combustion systems, by utilizing the
Gibbs energy minimization code MINGTSYS and reporting equilibrium
distributions for 18 trace elements of concern with respect to coal utilization [1].
Reducing and oxidizing conditions were considered, and the results were
compared qualitatively to experimental results published in the literature [8, 10
11].
Numerous thermodynamic packages have been developed for the purpose of
minimizing the total Gibbs energy of mass balance constrained systems
containing ash forming and trace element species [12 13]. Differences in the
output from these packages may be due to: 1) a lack of thermochemical data for
important chemical species, 2) the use of different or inconsistent thermo-
chemical data or 3) the use of different numerical solution techniques and/or
convergence criteria. Thus, a GEA test was performed in order to compare the
output from different thermodynamic packages.
-------
2. Description of the Modeling Approach
The total Gibbs energy, Gl, of a chemical system is given by:
nt N
R-T
R-T
(1)
where G symbolizes a Gibbs energy, superscript t denotes total, R is the
universal gas constant, T is the absolute temperature, N is the total number of
chemical species, n; is the number of moles of species i, superscript (ฐ) denotes
the standard state, subscript fi denotes the formation of species i, and a, is the
activity of species i. The function G* is combined with the mass balance
constraints of the system and minimized, using the method of undetermined
Lagrangian multipliers [14 -15].
Gaseous phase:
C02, H20, 02)N2
S02, CO, H2, H2S,
Me, MeO, MeCln,
etc
Sorbent -
02, N2 -
C, H, O, N,
S, Me, etc.
Equilibrium
reactor
T
P
Condensed phase:
MeO, MeCO3,
MeSO4, MeS, etc.
Figure 1:
The model system used in Global Equilibrium Analysis.
A GEA model, based on an actual operating combustion system, has been set
up. The temperature, pressure, total elemental composition, and a list of
chemical species to be considered, are specified, see Figure 1. The gas is
assumed ideal, and all condensed phases are considered pure. Among the ash
forming elements, only calcium, Ca, is taken into account as an As-, Cr-, and S-
capturing sorbent. The trace elements, As, Cd, Cr, Hg, Ni, Pb, and Se, are
considered one by one. The temperature is varied between 700 K and 2000 K
-------
and the pressure is set equal to 1 atm. and 20 atm. The elemental composition
of the system considered is presented in Table 1.
Table 1:
Reactor feedstock mass flows and compositions used in this study.
Feedstock mass flows (kg/hr): Coal : Sorbent: Moisture = 15.2 : 8.7 : 5.2
Major elements. %(w/w) (db):
Carbon (C)
Oxygen (O)
Hydrogen (H)
Sulfur (S)
Nitrogen (N)
Calcium (Ca)
Trace elements, ppmw:
Chlorine (Cl)
Arsenic (As)
Cadmium (Cd)
Chromium (Cr)
Mercury (Hg)
Nickel (Ni)
Lead (Pb)
Selenium (Se)
Coal:
83.3
9.0
5.0
1.2
1.5
1200.0
45.0
0.11
16.0
0.15
13.0
6.3
1.8
Sorbent:
12.0
48.0
40.0
320.0
1.9
0.14
2.8
0.0012
12.0
6.9
0.79
Four different thermodynamic packages, MINGTSYS, NASA-CET89, FACT, and
SOLGASMIX, have been used to minimize the total Gibbs energy of a system
with the elemental composition presented in Table 1. Preliminary results of the
GEA round robin, i.e. comparison of the output from the four packages were
reported previously [13], A summary, with special focus on the results at 1 atm.,
is provided below.
3. Summary of GEA Round Robin Test Results
A list of the major combustion products and trace element chemical species
taken into account when minimizing the total Gibbs energy of the system
considered in this study (see Table 1), is provided in Table 2. For the trace
elements, As, Hg, and Se, all four packages have predicted equilibrium
distributions equal to those reported by [10], while for Cd, Cr, Ni, and Pb
differences were observed in the calculated equilibrium distributions from the
four packages [13],
-------
Table 2:
Major combustion, sorbent and trace element species included in this work:
C0(g), C02(g), COS(g), Cl(g), CIO(g), CI2(g), CI2O(g), HCN(g), HCI(g), H2O(g),
NH(g), NH2(g), NH3(g), NO(g), NOCI(g), NO2(g), NO2CI(g), N2(g), N2O(g), O(g),
OH(g), 02(g), SO(g), SOCI2(g), SO2(g), SO2CI2(g), SO3(g).
CaO(cr.l), CaSO4(cr).
Arsenic:
As(g), AsCI3(g), AsO(g), As2(g), As3(g), As4(g), As4O6(g), As2O3(cr,l), As2O4(cr),
As2O5(cr,l), Ca3(AsO4)2(cr).
Cadmium:
Cd(g), CdCI2(g), CdO(g), CdCI2(cr,l), CdO(cr), Cd(OH)2(cr), CdSO4(cr).
Chromium:
Cr(g), CrCI2(g), CrCI2O2(g), CrCI3(g), CrCI4(g), CrO(g), Cr(OH)(g), CrO2(g),
Cr(OH)2(g), CrOOH(g), CrO3(g), Cr(OH)3(g), CrO(OH)2(g), Cr(OH)4(g),
Cr02OH(g), CrO(OH)3(g), Cr(OH)5(g), CrO2(OH)2(g), CrO(OH)4(g), Cr(OH)6(g),
CrCI2(cr,l), CrCI3(cr), Cr2O3(cr), Cr2(SO4)3(cr).
Mercury:
Hg(g), HgCI(g), HgCI2(g), HgO(g), HgCI2(cr,l), Hg2CI2(cr), HgO(cr), HgSO4(cr).
Nickel:
Ni(g), Ni(CO)4(g), NiCI(g), NiCI2(g), NiO(g), NiCI2(cr,l), NiO(cr), NiSO4(cr).
Lead:
Pb(g), PbCI(g), PbCI2(g), PbCI4(g), PbO(g), Pb2(g), PbCI2(cr,l), PbO(cr.l),
PbSO4(cr,l)
Selenium:
Se(g), SeCI2(g), SeO(g), SeO2(g), Se2(g), Se2CI2(g), Se3(g), Se5(g), Se6(g),
Se7(g), Se8(g), SeCI4(cr), SeO2(cr).
3.1. Arsenic (As), Mercury (Hg) and Selenium (Se) Results
Arsenic forms the stable crystalline calcium arsenate, Ca3(AsO4)2 (cr), and
gaseous arsenic(ll) oxide, AsO(g), below respectively above 1400 K, see Figure
2. No significant amount of AsCI3(g) is formed. Mercury forms the stable gaseous
mercury dichloride, HgCI2(g), at temperatures up to 800 K. Above 850 K, Hg(g)
and minor amounts of HgO(g) are the stable forms of mercury. Increasing the
temperature above 900 K, HgO(g) is gradually decomposed, forming Hg(g) and
O2(g), see Figure 2. No condensed forms of mercury are formed. In analogy with
sulfur, selenium forms the stable SeO2(g) in the temperature range [700 - 2000
K], see Figure 2. Above 1300 K, the equilibrium form of Se shifts toward SeO(g)
and to a minor extent to Se(g).
-------
100
80
60
40
20
0
MINGTSYS: Arsenic, 1 atm.
-AsO(g)
Ca3(As04)2(cr)
700 900 1100 1300 1500 1700 1900
Temperature (K)
100
80
60
40
20
MINGTSYS: Mercury, 1 atm.
1100 1300 1500
Temperature (K)
1 700
1 900
MINGTSYS: Selenium, 1 atm.
700 900 1100 1300 1500
Temperature (K)
1900
Figure 2:
Equilibrium distribution of As, Hg, and Se, as predicted by MINGTSYS.
-------
3.2. Cadmium (Cd) Results
MINGTSYS has found cadmium to exist as CdCI2 at 1 atm. and temperatures up
to 1100 K: CdCI2(cr) below, respectively CdCI2(g), above 750 K. At temperatures
above 1200 K, Cd(g) and CdO(g) are the stable species, Cd(g) accounting for
more than 95 %(mol/mol) of the Cd present at all temperatures. Gaseous
cadmium oxide, CdO(g), is gradually decomposing as the temperature is raised
above 1280 K, forming Cd(g) and O2(g), see Figure 3.
NASA-CET89 has found CdCI2(cr) to be stable below, and Cd(g) to be the major
stable form of cadmium above 750 K. Small amounts (< 2 %(mol/mol)) of the Cd
were present as CdO(g) in the temperature range 750 -1000 K.
FACT has found CdCI2(cr) to be stable up to 750 K, above which temperature
CdO(cr) is stable up to 1000 K, see Figure 3. Formation of Cd(g) and CdO(g)
begins at 850 K, Cd(g) being the major species above 1000 K, where CdO(g)
gradually decomposes forming Cd(g) and O2(g).
SOLGASMIX has found CdO(cr) to be the stable form of cadmium up to 1400 K.
Formation of Cd(g) begins at 1200 K. Above 1600 K, Cd(g) is the major stable
form of Cd present. In addition, minor amounts of CdO(g) and CdCI2(g) were
formed in the temperature ranges [1300 2000 K] and [700 1500 K],
respectively, see Figure 3.
3.3. Chromium (Cr) Results
At 1 atm., MINGTSYS and NASA-CET89 have found Cr2O3(cr) to exist at
temperatures up to approx. 1200 K, where it is decomposing, forming
CrO2(OH)2(g). The latter has a maximum in occurrence around 1300 K. Above
1300 K, chromium showed a very complex equilibrium chemistry, forming the
gaseous components, CrO(OH), CrO2(OH), CrO(OH)2, CrO2, and CrO3 [13].
In contrast, different chromium species distributions were obtained with FACT
and SOLGASMIX. This is a result of the inclusion of a stable mixed oxide
compound, CaO.Cr2O3(cr), and the exclusion of the high temperature gaseous
chromiumoxides and - hydroxides discussed in [16]. FACT found CaO.Cr2O3(cr)
to be thermodynamically stable between 800 and 1800 K [13], whereas
SOLGASMIX found CaO.Cr2O3(cr) to be the major Cr-species stable in the
temperature range [700 2000 K]. These results clearly demonstrate that the
distribution is sensitive to the inclusion of species that are not obviously
important at first glance (e.g. the hydroxyoxides); care must therefore be taken
when using commercially available databases that may not contain all of the
important species.
-------
MINGTSYS: Cadmium, 1 atm.
ป 80-
co 60
* 404-
20-
900 1100 1300 1500 1700
Temperature (K)
1900
700
FACT: Cadmium, 1 atm.
Figure 3:
Equilibrium distribution of Cd as predicted by MINGTSYS, FACT and
SOLGASMIX.
-------
3.3 Nickel (Ni) Results
MINGTSYS and FACT have found Ni to form NiO(cr) at 1 atm. and temperatures
up to approximately 1500 K, where NiO(g), Ni(g), and NiCI(g) were formed.
Formation of NiCI2(g) begins at 1000 K. NiO(g), NiCI(g), NiCI2(g) were stable up
to 1800 K, where the last trace amount of NiO(cr) disappeared. Above 1825 K,
existing equilibrium among NiO(g), NiCI(g), NiCI2(g), and Ni(g) were shifted
towards Ni(g) as the temperature was increased, see Figure 4. Calculations
performed with FACT indicated that implementation of Ni(OH)2(g), which was not
included in the common list of species taken into account in this study, could
affect the stability of NiO(cr).
NASA-CET89 has also predicted NiO(cr) to be stable at temperatures up to 1500
K, where formation of NiO(g), Ni(g), and NiCI(g) started, but NASA-CET89 has
predicted a much higher concentration of Ni(g) and a correspondingly lower
concentrations of NiO(g), NiCI(g), and NiCI2(g) above 1750 K, than MINGTSYS
and FACT.
SOLGASMIX has found NiO(cr) to be the major stable form of Ni at all
temperatures between 700 K and 2000 K. Trace amounts of various gaseous Ni-
species (Ni, NiO, NiCI, and NiCI2) were formed at temperatures above 1600 K.
As with FACT, the output from SOLGASMIX has indicated that Ni(OH)2(g) may
affect the stability of NiO(cr).
MINGTSYS: Nickel, 1 atm.
U)
0)
'o
o
Q.
Z
0)
if
0)
Q_
IUU
80
60
40
20
" - N
\
*
rx
. Ni(g)
NiCI(g)
- - - NiCI2(g)
NiO(g)
- NiO(cr)
700 900 1100 1300 1500
Temperature (K)
1700
1900
Figure 4:
Equilibrium distribution of Ni, as predicted by MINGTSYS.
3.4. Lead (Pb) Results
At 1 atm., MINGTSYS and NASA-CET89 have found PbCI4(g) to be the major
stable form of Pb, below 1100 K, see Figure 5. Above 1300 K, PbO(g) was the
major stable Pb-species, but it was gradually decomposed with increasing
-------
temperatures above 1500 K, forming Pb(g). Peaks of PbCI2(g) and PbCI(g) were
in the temperature ranges [1000 1500 K] and [1100 2000 K], respectively.
PbCI2(g) has a maximum occurrence of approx. 48 % (mol/mol) at 1200 K, while
PbCI(g) has a maximum occurrence of approx. 8 % (mol/mol) at 1250 K.
FACT and SOLGASMIX have showed an equilibrium distribution of Pb, similar to
that reported above, but the relative amounts of PbCl(g) and PbCI2(g) formed
were smaller than predicted by MINGTSYS and NASA-CET89. The output from
SOLGASMIX has indicated formation of significant amounts of PbO(cr) in the
intermediate temperature range [1050 1800 K]. No condensed phases of Pb
were found below 1000 K, see Figure 5.
NASA-CET89: Lead, 1 atm.
900 1100 1300 1500 1700 1900
Temperature (K)
SOLGASMIX: Lead, 1 atm.
80
60
40 -
20
oi!
7t
a , -
\n D /'
a,
^ 'ฐv ^
-' ^s^r
0 900 1100 1300 1500 1700 1900
ll
- PbCI4(g)
PbO(g)
Pb(g)
PbCI(g)
- - - PbCI2(g)
O PbO(cr)
Temperature (K)
Figure 5:
Equilibrium distribution of Pb, as predicted by NASA-CET89 and SOLGASMIX.
-------
4. Summary and Discussion
Global Equilibrium Analysis (GEA), is an easy-to-use and simple way of
qualitatively predicting phase partitioning and speciation of trace elements in
thermal fuel conversion systems. Phase partitioning data are important when
evaluating the potential trace element enrichment of the fly ash separated in the
electrostatic precipitator (or other particulate removal device). Speciation of trace
elements in the different phases (i.e., flue gas, bottom ash, fly ash) is important,
since the properties and reactivities of the trace elements depend strongly on
their oxidation state. Oxidized forms of trace elements are more likely to deposit
locally or regionally [17]. Another important aspect is the mobility of oxidized
trace element species during water leaching in fly ash disposal areas [18].
The equilibrium speciation of As, Cd, Cr, Hg, Ni, Pb, and Se have been predicted
at 1 and 20 atm. in the temperature range [700 2000 K], by use of the
thermodynamic packages MINGTSYS, NASA-CET89, FACT, and SOLGASMIX.
For the trace elements, As, Hg, and Se, almost identical equilibrium distributions
were predicted by the four packages, while for the elements, Cd, Cr, Ni, and Pb,
differences in the output from the four packages were observed. The greatest
differences were due to the presence or absence of chemical species, such as
Ni(OH)2(g) and the chromium oxyhydroxides, in the respective databases of the
four thermodynamic packages. Differences due to the thermodynamic data used
for individual chemical species were also significant. For example, the amount of
Ni(g) at 2000 K predicted by NASA-CET89 suggests that other thermodynamic
data have been used than in the remaining packages.
In the case of cadmium, where four different equilibrium distributions were
predicted, a systematic comparison of the thermochemical input-data (i.e. CP(T),
H(T), S(T), and G(T) data) utilized for major combustion and Cd species, in the
four packages, are currently being performed. As a first step, MINGTSYS was
used to perform a simple pertubation analysis. In Table 3, the number of moles
of Cd(g), CdCI2(g) and CdO(g) formed at 1000 K are listed as a function of the
Gfiฐ(Cd(g),1000K) value. The tabulated G^-value for Cd(g) at 1000 K was
multiplied by a pertubation factor of 0.5 (left column), 1.0 (i.e. the tabulated
value, middle column) and 2.0 (right column), respectively. The values of 0.5 and
2.0 were chosen to illustrate extremes in the uncertainty of thermochemical
input-data.
Notice in Table 3 that the speciation of Cd between Cd(g) and CdCI2(g) at 1000
K is strongly affected by the value of Gfiฐ(Cd(g)). Thus, scientists and
researchers that utilize Global Equilibrium Analysis (GEA) on thermal fuel
conversion systems should be aware of the quality of the thermochemical input-
data: Poor quality thermochemical input-data may give poor quality, unreliable
results.
-------
Table 3:
Impact of the quality of Cd(g) thermochemical input-data on MINGTSYS output:
The number of moles of Cd(g), CdCI2(g) and CdO(g) formed at 1000 K are listed
as a function of the Gfiฐ(Cd(g),1000K) - value. The tabulated Gfiฐ-value for Cd(g)
at 1000 K was multiplied by a pertubation factor of 0.5 (left column), 1.0 (i.e. the
tabulated value, middle column) and 2.0 (right column), respectively.
Pertubation fact:
Cd(g)
CdCI2(g)
CdO(g)
Sum:
0.5
1. 610201 E-11 mol
1. 69991 2E-06mol
7.189899E-11 mol
1.700E-06mol
1.0
8.792107E-10mol
1 .699049E-06 mol
7.186247E-11 mol
1.700E-06mol
2.0
1.031 801 E-06 mol
6.681709E-07 mol
2.825777E-11 mol
1.700E-06mol
Pertubation factors of 0.9 and 1.1, corresponding to a 10 % variation of the Gfiฐ-
value for Cd(g), and of other gaseous Cd species, at 1000 K confirmed that even
small variations in the thermochemical input-data may affect the final distribution
of species.
When GEA is used on a thermal fuel conversion system, one has to be aware of
the following [10, 19]:
1. All relevant chemical species occuring in the thermal fuel conversion
system must be taken into account, otherwise the output from the GEA will
be misleading.
2. Consistent thermodynamic data must be used.
3. Appropriate mixing models (pure phases, ideal or non-ideal mixing) should
be applied in the condensed phase(s).
4. In thermal fuel conversion systems, mixing phenomena and/or boiler design
characteristics may introduce local conditions (e.g., temperature and/or
composition gradients) not taken into account in the GEA.
Item 1) above, has been illustrated through the comparison of thermodynamic
packages performed in this study. All relevant chemical species in an actual
operating thermal fuel conversion system must be included when performing a
GEA, otherwise the result and conclusion will be misleading. Some trace
element gas phase reactions in hot flue gases are expected to be kinetically
limited, i.e. they do not reach equilibrium within the time scale actual in a fuel
conversion system. Thus, a certain creativity in the selection of chemical species
considered in the GEA, based on expected kinetic limitations, could be usefull.
-------
Acknowledgement
The authors are grateful to the Combustion and Harmful Emission Control
(CHEC) research program at the Department of Chemical Engineering,
Technical University of Denmark, for financial support. The CHEC research
program is cofunded by ELSAM (The Jutland-Funen Electricity Consortium),
ELKRAFT (The Zealand Electricity Consortium), the Danish Technical Research
Council, the European Union, and the Danish and Nordic Energy research
programs.
Dr. William P. Linak, National Risk Management Research Laboratory, U.S.
Environmental Protection Agency, Research Triangle Park, NC, USA, is grately
acknowledged for providing the NASA-CET89 calculations presented in this work
and for numerous valuable discussions and comments to this manuscript.
The authors are also grateful to: the National Center for Excellence on Air Toxic
Metals at the Energy and Environmental Research Center, University of North
Dakota, and the Environmental Research Institute, University of Connecticut.
Grateful acknowledgement is also made to Dr. Bonnie McBride, NASA Lewis
Research Center, and to Dr. Michael Michelsen, Department of Chemical
Engineering, Technical University of Denmark, for providing the NASA-CET89
respectively the MINGTSYS codes, and other assistance.
References:
1. D. J. Swaine. Trace Elements in Coal. Butterworth and Co. Ltd., London,
UK, 1990.
2. R. M. Davidson, D. F S. Natusch, J. R. Wallace, and C. A. Evans, Jr.,
"Trace Elements in Fly Ash Dependence of Concentration on Particle
Size", Environ. Sci. Technol., Vol. 8, No. 13, pp. 1107 -1113 (1974).
3. D. J. Swaine and F Goodarzi, Environmental Aspects of Trace Elements in
Coal. Kluwer Acad. Publishers, Dordrecht, The Netherlands, 1995.
4. R. Meij, J. Van der Kooij, J. L. G. Van der Sluys, F. G. C. Siepman, and H.
A. Van der Sloot, "Characteristics of Emitted Fly Ash and Trace Elements
from Utility Boilers Fired with Pulverized Coal", KEMA Sci. & Technol.
Reports, Vol. 2, No. 1, pp. 1 - 8 (1984).
5. Trace Element Emissions from Coal Combustion and Gasification. London,
UK: IEA Coal Research, 1992. Rep. No. IEACR/49.
6. B. Sander, "Measurements of Trace Element Mass Balances in Coal-Fired
Power Plants Equipped with Different Types of FGD Systems", presented
at the 2nd Int. EPRI Conf. on Managing Hazardous Air Pollutants,
Washington D. C. (July 1993).
7. G. F. Weber, S. R. Ness, S. J. Miller, T. D. Brown and C. E. Schmidt, "A
Summary of Utility Trace Element Emissions Data from the DOE Air Toxics
-------
Study1', presented at the Int. EPRI Conf. on Managing Hazardous and
Particulate Air Pollutants, Toronto, Canada (August 1995).
8. L. Aunela-Tapola, F J. Frandsen, E. Hasanen, "Trace Element Emissions
from Estonian Oil Shale Fired Power Plant", manuscript submitted to Fuel
Proces. Techn., Jan. 1997.
9. Trace Elements in Coal. London, UK: IEA Coal Research, 1996. Rep. No.
IEAPER/21.
10. F. J. Frandsen, K. Dam-Johansen, and P Rasmussen, "Trace Elements
from Combustion and Gasification of Coal An Equilibrium Approach",
Prog. Energy Comb. Sci., Vol. 20, 115-138(1994).
11. F. J. Frandsen, Trace Elements from Coal Combustion. Ph.D.-Thesis, Dept.
Chem. Eng, Techn. Univ. of Denmark. ISBN-87-90142-03-9.
12. Trace Element Emissions Literature Review. Perten, The Nteherlands:
ECN, 1994. Rep. No. ECN-C-94-096.
13. F J. Frandsen, T. A. Erickson, V. Kuhnel, J. J. Helble and W. P Linak,
"Equilibrium speciation of As, Cd, Cr, Hg, Ni, Pb, and se in oxidative
thermal conversion of coal A comparison of thermodynamic packages",
presented at the 3rd Int. Symp. on High Temperature Gas Cleaning,
Karlsruhe, Germany (September, 1996).
14. G. Eriksson, "Thermodynamic Studies of High-Temperature Equilibria XII",
Chimica Scripta, Vol. 8, p. 100 (1975).
15. H. Greiner, "The Gibbs Energy of a Chemical Reaction System Considered
as a Function of its Elemental Abundancies", CALPHAD, Vol. 12, No. 2, pp.
143-154(1988).
16. B. B. Ebbinghaus, "Thermodynamic of Gas Phase Chromium Species: The
Chromium Oxides, the Chromium Oxyhydroxides, and Volatility
Calculations in Waste Incineration Processes", Combust. Flame, Vol. 93,
pp. 119-137(1993).
17. S. A. Benson, M. L. Jones, J. H. Pavlish, C. J. Zygarlicke and S. J. Miller,
"Trace Element Behavior in Coal Combustion Systems", presented at the
EPRI conf. on Effects of Coal Quality on Power Plants, Kansas City,
Missouri, USA (May, 1997).
18. K. Sandelin, "Water-Solubility of Trace Elements in Ashes and Other Solid
Residues from Coal Combustion", presented at 4th Nordic Ash Workshop,
Alabama, Data City, Aabo, Finland (June, 1997).
19. W. P Linak and J. O. L. Wendt, "Trace Metal Transformation Mechanisms
during Coal Combustion", Fuel Process. Techn., Vol. 39, pp. 173 198
(1994).
-------
SUMMARY OF KEY AIR TOXIC RESULTS FROM THE CENTER FOR AIR TOXIC
METALS (CATM)
John Pavlish
Steve Benson
Energy & Environmental Research Center
15 North 23rd Street
Grand Forks, ND 58203
Abstract
CATM is a multiyear multimillion dollar program between the U.S. Environmental Protection Agency
(EPA) and the Energy & Environmental Research Center (EERC) at the University of North Dakota
(UND) focused on conducting fundamental and practical research and development on air toxic
elements to provide a theoretical and scientific basis for preventing and controlling air toxic emissions
and evaluating the fate and impacts of air toxic elements on human health and the environment.
CATM's mission is to provide a nationally coordinated and practically oriented multidisciplinary
research, development, and training program on the prevention, transformation, behavior, and control
of potentially toxic element emissions from energy-producing and incinerating systems and on the
prevention and minimization of the effect of these metals on the environment through partnerships with
industry, academia, and government. This paper will present some of the key findings and
experimental results that are pertinent and critical to air toxic metals and industry.
Introduction
The Center for Air Toxic Metals (CATM) at the University of North Dakota (UND) Energy &
Environmental Research Center (EERC) was established in 1992 by the U.S. Environmental Protection
Agency (EPA) Office of Environmental Engineering and Technology Demonstration (OEETD) to focus
national research efforts addressing air toxic trace element emissions, which have become a matter of
worldwide concern as well as a regulatory issue in the United States.
To assist EPA, CATM provides a nationally coordinated and practically oriented multidisciplinary
research, development, and training program on the prevention, transformation, behavior, and control
of potentially toxic metal emissions from energy-producing and incinerating systems and on the
prevention and minimization of the effect of these metals on the environment through partnerships
developed with industry, research institutions, and government.
The goal of CATM is to develop key information on air toxic metal compounds such that pollution
prevention strategies can be developed and implemented to reduce air toxic metal emissions (1).
Specific objectives of CATM are to 1) elucidate air toxic transformation mechanisms and pathways in
energy-producing and incinerating systems, 2) develop and demonstrate technologies to control metals
behavior and emissions, 3) develop and demonstrate environmentally sound methods to utilize and
dispose of residuals, 4) develop and validate methods to sample and analyze air toxics, 5) develop
predictive tools and databases, 6) develop partnerships with industry, 7) develop environmental
awareness and pollution prevention programs through education, and 8) commercialize results and
technologies.
To accomplish the mission and goals of CATM, new partnerships and existing relationships with
government, industry, and research institutions are encouraged (2). The needs of both industry and
government are addressed through the CATM Affiliates Program (2). The Affiliates Program allows
-------
outside entities to participate and direct CATM research activities. To encourage industry participation,
a portion of CATM funds are set aside to perform research projects with industry. These funds are
designated for research projects that address specific concerns or interests of the Affiliate members.
CATM is the focal point of these partnerships that are key to effective research and development
programs (3-5).
To systematically address the many complex issues associated with air toxic metals, CATM is
organized into five integrated program (focus) areas. Program Area 1 - Air Toxic Metals
Transformation Mechanisms focuses on the determination of the chemical and physical transformations
of air toxic metals as a function of the association and abundance of the metal in fuel and system design
and operating conditions. Program Area 2 - Analytical and Sampling Methods Development involves
the verification and enhancement of existing methods and application of new methods to effectively
speciate and determine the abundance of air toxic metals in fuels and process streams. Program
Area 3 - Control Technologies focuses on pollution prevention and evaluating and enhancing currently
used emission control technologies for air toxic trace elements. Program Area 4 - Modeling and
Database Development involves the development of tools to predict the fate of air toxic metals in
combustion, gasification, and incineration systems. Program Area 5 - Technology Commercialization
and Education involves the development of partnerships through the transfer of information from
CATM to industrial sponsors as well as to academic and government partners through newsletters,
education programs, and annual meetings.
Results
EPA has been mandated by Congress through the 1990 Clean Air Act Amendments to study and
regulate, if necessary, air toxic metals. Eleven trace elements that are of particular concern are
beryllium, chromium, manganese, cobalt, nickel, arsenic, selenium, cadmium, antimony, lead, and
mercury. To address EPA needs, CATM has and will continue to conduct research that leads to a
better understanding of the trace element forms (species) and quantity produced and potentially emitted
from energy conversion systems. Since CATM has been an ongoing program since 1992, it is beyond
the scope of this paper to discuss all research findings. Consequently, only a few key results are
selected from each program area and are summarized below.
Transformation Mechanisms
The chemical form of trace elements in fossil fuels and wastes transform and partition during
combustion into solid, liquid, or vapor components throughout the system and beyond. A unique, state-
of-the-art combustion system (referred to as the conversion and environmental process simulator
[CEPS]) for performing tests regarding air toxic metal formation, emissions, and control (6) was
designed and built, (refer to Figure 1). CEPS is designed primarily for assessing the effects of fuel type
and combustion conditions on trace element emissions, partitioning, and transformations. Shakedown
testing of the CEPS and trace element partitioning was evaluated and compared to the partitioning
results obtained in other bench- and pilot-scale combustion systems. A blank test on the CEPS indicated
that contamination from hot surfaces and injected gases within the CEPS was insignificant. The
partitioning of As, Pb, Ni, Se, and Hg between ash paniculate and gaseous combustion by-products
was generally consistent as compared with three other experimental combustion systems (6).
The distribution, or partitioning, of As, Pb, Ni, Se, and Hg in the combustion flue gas and paniculate
for four coals was tested (7). A five-stage multicyclone and final filter assembly with cut points of 6.95,
3.85, 2.01, 1.42, 0.59, and 0.295 /on were used. Figure 2 shows that Hg is nearly entirely partitioned
to the vapor phase. Se shows a significant portion of its mass in the vapor with the remaining falling
into the fine paniculate fraction. Se trends more toward the smaller size ranges because of its
-------
Ceramic
Tube
Baghouse
Bypass
Typical Furnace
Section
Convective
Pass
Section
Main
Furnace
Section
Heat
Exchangers
EBKCZ11KS.CDR
Figure 1
Design Sketch of the CEPS Showing the Main and Convective Pass Section, Heat
Exchangers, and Baghouse
condensation onto the smaller paniculate or its nucleation from a vapor. On a total mass basis, most of
the Pb, Ni, and As are found in the supermicron size. On a concentration basis, Ni and Pb seem to
concentrate in the finer particulate fraction (6, 7). Arsenic seems to concentrate more in the 2-4-/zm
range, with a few exceptions noted. It is not entirely understood why the As is not more associated with
-------
100
Figure 2
Panitioning of Trace Elements for Different Size Fractions
the fine fraction similar to Ni and Pb, since the thermal stability of As vapor species should account for
higher concentrations in the fine paniculate/filter fraction compared to less stable Ni and Pb forms.
Implications of these data are that significant vapor or fine paniculate near 1-jrni or < 1-ftm fractions of
Hg, Se, Ni, Pb, and As may escape collection via conventional control devices. Proposed federal
regulations of particulate to a new PM2.5 level would force improved fine paniculate collection,
thereby capturing significant quantities of the fine particulate and its associated air toxic metals.
Previous work by Pavlish and others (8, 9) using a bench-scale drop-tube furnace system for coal
combustion showed similar relative concentrations of trace elements collected in a multicyclone and
EPA Method 29 sampling train (10).
Volatile trace elements, such as Se and Hg, as discussed above, are partially or completely vaporized at
combustion temperatures (=1500ฐC). As flue gases cool, it is possible for a significant fraction of the
vaporized trace elements to physically or chemically adsorb onto fly ash particles, especially those with
large surface area-to-volume ratios. Consequently, fly ash plays a major role in partitioning and in the
formation of trace metal species (8). Hg, because of its volatility is initially transformed during
combustion to gaseous elemental mercury, Hgฐ (g). Subsequent oxidation reactions with flue gas
components at lower temperatures produce inorganic mercuric compounds that are primarily divalent
anion species (Hg2+X) or the elemental species (Hgฐ). Other data in Hg speciation analysis (11) suggest
that generally greater than about 50% of the Hgฐ(g) reacts with oxidants in combustion flue gases. EPA
Method 29 was used primarily to speciate Hg by capturing oxidized or ion forms of Hg in a hydrogen
peroxide-nitric acid impinger series, followed by elemental Hg capture in a potassium
permanganate-sulfuric acid impinger series. The results of this work, however, must be used with
caution, since EPA Method 29 has been shown to overestimate the oxidized Hg form by approximately
10%-30%, depending upon fuel type; other speciation methods are currently being developed and
evaluated (10).
Several combustion tests with accompanying flue gas sampling using EPA Method 29 were performed.
From these tests, it was observed that on the average 43 % of the Hg vapor fraction was captured in the
peroxide impingers (i.e., mercuric form (Hg2+), with the remaining vapor-phase Hg collected in the
permanganate impingers and presumed to be the elemental (Figure 3). The proportions of different Hg
-------
forms in the flue gas can vary considerably. In these tests, the combustion and sampling parameters
were held fairly constant, and major differences in Hg speciation are believed to be due to fuel
components. The low-chlorine lignite and Powder River Basin (PRB) coals show significant differences
in Hg speciation, compared with results acquired for the relatively high-chlorine (790 ppm) Pittsburgh
No. 8 coal. From the data, it appears that chlorine may be a dominant factor in Hg speciation, since the
chlorine-rich flue gas associated with the bituminous coal is consistent with the collection of a relatively
large proportion of the divalent Hg form, most likely HgCl2(g), as shown in Figure 3. The
condensation of Hg vapor onto abundant fine clay particles, calcium-rich particles, or Hg vapor being
sorbed by residual carbon particles in the Lignite B ash (= 1.3 % loss on ignition) may account for the
strong paniculate association by that coal ash indicated in Figure 3.
Analytical and Sampling
Part of the analytical and sampling program involves determining the best techniques to use for a given
measurement. Several techniques used for measuring the trace elements As, Be, Cd, Co, Cr, Hg, Mn,
Ni, Pb, Sb, and Se in coal are compared in Table 1. Visible and UV light spectroscopic techniques
such as atomic absorption spectroscopy (AAS) and inductively coupled plasma atomic emission
spectroscopy (ICP-AES) are versatile and widely used. Results are obtained more quickly with
ICP-AES, but the LLQ (lower limits of quantification) are better with AAS. Two x-ray fluorescence
(XRF) techniques, energy-dispersive XRF (EDXRF) and wavelength-dispersive XRF (WDXRF), are
available for some metals, including Cr, Ni, and Pb. The XRF techniques do not require sample
digestion, which represents an advantage over AAS and ICP. A more direct form of trace element
speciation analysis, x-ray absorption fine structure (XAFS), is currently being evaluated.
Method development efforts focused on determining repeatability and reproducibility of various
analytical techniques used to measure trace element content of coal and other bulk samples.
Repeatability and reproducibility were determined for WDXRF and three types of AAS: graphite
furnace AAS (GFAAS), hydride generation AAS (HG-AAS), and cold-vapor AAS (CV-AAS). The
two issues of repeatability and reproducibility were investigated independently. Control charts were
JERC CZ1X4&CDR
PRB
A
PRB
B
Lig.
A
Figure 3
Comparison of Hg Speciation for Combustion Flue Gas from Several Coals
-------
Table 1
Comparison of Trace Element Measurement Techniques used to Measure Trace Element Levels in Coal
Precision, ppm
(95% confidence level)
Element
As
Be
Cd
Co
Cr
Hg
Mn
Ni
Pb
Sb
Se
Technique
GFAAS
ICP-AES
ICP-AES
GFAAS
ICP-AES
ICP-AES
GFAAS
ICP-AES
EDXRF
WDXRF
CV-AAS
ICP-AES
GFAAS
ICP-AES
EDXRF
WDXRF
GFAAS
ICP-AES
WDXRF
GFAAS
Hydride AA
ICP-AES
LLQ, ppm
0.5
30
0.5
0.01
1
10
0.05
5
1.8
14
0.01
10
0.5
10
2.8
4
0.5
30
7
0.2
0.2
50
Sampling
+0.35
NA1
NA
+0.01
NA
NA
+0.40
NA
NA
NA
NA
NA
+0.21
NA
NA
NA
+0.66
NA
NA
NA
+0.03
NA
Overall
+0.48
NA
NA
+0.06
NA
NA
+ 1.40
NA
+2.18
+0.88
+0.005
NA
+2.18
NA
+ 1.76
+0.58
+ 1.20
NA
+ 1.34
NA
+0.17
NA
Not available.
used to provide a visual determination of reproducibility and repeatability. The main reason for using a
control chart is to determine whether the analytical process is in control (i.e., whether there are sources
of bias or assignable errors due to imprecision or inaccuracy of the overall analysis) (6).
Repeatability. The elements analyzed by each instrument were As, Cd, Cr, Pb, Ni, Se, and Hg, and
the LLQ and LLD (lower limit of detection) for each element were set, with the LLQ being 2 to 3
times higher than the LLD. For each element analyzed by an instrument, the instrument repeatability
indicated that results obtained were not significantly different from trial to trial. Sample preparation
produced more varied outcomes. The type of analyte (i.e., coal), the element analyzed, the preparation
technique, the laboratory preparing the sample, and the instrument used to analyze the element all
influenced the repeatability of the sample measurement. Sample preparation variability was easily seen.
The WDXRF sample preparation results were influenced by such variables as the laboratory preparing
the sample, the subdivision of the sample, the grinding time, and the grinding medium. The results for
the WDXRF pellets suggest that if sample preparation is performed in a consistent manner, both the
sample preparation and instrumentation results can be within statistically repeatable limits for Ni, Cr,
and Pb. Preliminary results show that the concentrations of Ni and Pb decreased with an increase in
grinding time when no grinding medium was used. The Cr concentration appeared to decrease with an
increase in grinding time when alcohol was used as the grinding medium.
Control charts for the AAS techniques showed that the overall process was generally in control. When
statistical tests were done on the repeatability of the AAS instruments, the results showed no significant
differences. The greatest difference in the results for AAS was caused by sample preparation.
-------
Eventually, the AAS instruments, like the XRF instruments, can be calibrated to produce repeatable
results for the trace elements analyzed, but great care must be taken in sample preparation.
Reproducibility. The WDXRF instrument used certified standards that were pressed into pellets.
These certified standards were not used in the calibration of the instrument. The WDXRF values for
Pb, Cr, and Ni values fell between the UCL (upper control limit) and LCL (lower control limit) of the
accepted value.
The AAS techniques used liquid standard solutions to calibrate the instrumentation so that the issue of a
sufficient number of standards was not as critical as for WDXRF. For this evaluation, only one coal
standard was used to test the reproducibility of the AAS techniques for two reasons: a) trace element
standards are expensive and b) a large amount of sample was required to complete one digestion
procedure. Most of the experimental values were found to be within the acceptable limits of the
standard. Problems were noted with false low Pb and high Ni measurements. In microwave digestion,
the vessel is sealed during the heating process but is opened separately for venting. This could account
for some loss of Pb but not enough to completely account for the low Pb readings. Since the Pb values
were all less than the LCL for the standards, a more plausible explanation for the results is that the Pb
combined with the sulfate ions to form a precipitate, thus reducing the concentration level that is
registered by the instrument, which can only analyze liquids. This precipitate could form in the
digestion environment given the use of sulfuric acid, which forms sulfate ions. A similar explanation
can be used for the Cr values. Because of these results, the EERC no longer uses sulfuric acid in the
coal digestion process, and the results for Pb and Cr are now within the acceptable limits of the
certified value. For Ni, the majority of experimental values obtained were above the accepted reference
values. Discoloration of the microwave vessel, during this study, may have been caused by elements
being left behind from a previous digestion and then removed from the lining of the vessel into the
sample solution through the corrosive nature of the acids. One possible element that could be removed
from the lining of the vessel is Ni. The instrumentation provided repeatable results but the sample
preparation did not. Thus few steps were taken to reduce the sample preparation problems encountered
when using microwave digestion on coal. First, new vessels were purchased. No discoloration occurred
in these vessels which reduced the possibility of contamination from one sample to another. Second,
sulfuric acid is no longer used to digest the coal. This eliminates the formation of the precipitates noted
for Pb and Cr.
In summary, WDXRF and AAS techniques can be calibrated to provide acceptable results, but again
the key to consistent repeatable and reproducible results lies with sample preparation. In addition, the
analyst must also make careful determinations such as the LLQ and LLD for each element for a
particular sample.
A general statement is that standard reference methods for ah- toxic metals that are accepted for
oxidizing environments may not necessarily work without modification for reducing environments.
Control Technologies
With the exception of Hg and Se, and assuming paniculate regulations remain the same, most of the
trace elements are expected to be removed effectively by conventional particulate control devices (12,
13). Some of the research performed through CATM and documented in other sources (6-8) support
this conclusion. In the event that more strict controls are required, CATM performed research that
demonstrated that ultrahigh collection (above 99.99%) of fly ash from combustion sources is attainable
with the use of advanced filter media (14, 15). However, advanced filter media cost significantly more
than what is currently used today.
-------
Mercury, on the other hand, because of its volatility, is known to be difficult to collect from
combustion systems using conventional gas paniculate control systems and, through CATM, has been
demonstrated to have an even higher volatility in highly reducing gasification systems (16). In an
attempt to develop an effective mercury control technology, CATM has taken several approaches
ranging from precombustion to postcombustion strategies. For example, a combined technology that
removes both organic sulfur and hazardous air pollutant (HAP) precursors has been preliminarily
developed and tested (17). With the financial support of the Illinois Clean Coal Institute, the U.S.
Department of Energy (DOE), and EPA CATM, an advanced coal-cleaning procedure using subcritical
water to remove sulfur and HAPs from coal has been developed. The procedure consists of
pressurizing a coal-water slurry mixture, followed by heating the mixture to temperatures approaching
370ฐC. Analysis performed on the clean fraction indicates that high-sulfur Illinois coal was cleaned
from over 3wt%to<0.8wt% sulfur. Scanning electron microscopy (SEM)-wavelength-dispersive
spectrometry (WDS) analysis indicated that the extraction process had significantly reduced the
numerous sulfur-bearing minerals (gypsum [CaSOJ, pyrite [FeSJ and sphalerite [ZnS]) present in the
raw coal. The difference demonstrated that the extraction process is effective at removing inorganic
sulfur. The extraction process breaks down certain minerals (particularly sulfates and sulfides) and
allows other minerals such as iron oxide form to take up the inorganic materials remaining behind.
Preliminary evidence also indicated that HAP elements such as Hg, As, and Se are associated with
sulfur components. For the two coals tested, Hg levels were reduced by 95%, As by 85%, and Se by
60%.
Another approach taken by CATM to control mercury emissions is the utilization of existing equipment
to enhance capture of volatile trace elements (i.e., Hg and Se). Typically, this involves designing the
optimum sorbent that will capture the volatile elements and will also be easily collected by the
paniculate control device. To date, results of tests on sorbent effectiveness in improving fabric filter
(baghouse) trace element removal efficiencies (8, 14, 18) are encouraging (refer to Figure 4). Two
precombustion sorbents, zeolite and kaolinite, were found to be only marginally effective for
controlling air toxic metal emissions, and a lime precombustion sorbent was found to have some control
with regard to Hg and Se emissions for certain coals (8). Activated carbon has proved to be effective
for capturing Hg for a lignite-fired coal combustion system, and preliminary combustion tests with
activated carbons reveal that they are the most promising sorbent material for capturing volatile air
toxic elements such as Se and Hg. Unfortunately, not all sorbents work effectively given the range of
fuel types, mercury concentration, and operating conditions under which they must work. Obviously,
temperature plays a key role as well as the composition of the fly ash. Fly ash was discovered to
behave differently relative to its ability to collect potential air toxic metals. Some coal ashes can
effectively collect up to 98% of the Hg released from the coal during combustion, while others may
collect only 5% (8).
Modeling and Database
CATM is involved in predicting the fate of trace elements utilizing various computer-based tools
including fundamental mechanistic stochastic modeling and empirically derived modeling.
Thermochemical equilibrium modeling is one predictive method that appears to show promise for
predicting trace element species formation during fuel conversion. Several methods or models are
currently in use for trace element thermochemical equilibrium modeling, and CATM has been
scrutinizing these models. The four models compared were FACT, SOLGASMIX, CET-89, and
MINGTSYS (19). Preliminary results comparing the four models for Pb are shown in Figure 5 (6).
The main difference in thermochemical models is in the database they employ to calculate phase
equilibria. The numerical approaches used by the four tested models may be different, but since the
-------
-------
percent Pb-ซpecies
o 3 ฃ S S 8
MINGTSYS: Pb species at p=1
^ ."---.
* ' ' .
W
SoKgyjoEncj
-COO*- D(9l
PbCI(g)
PbCI4
-------
i.e., speed. The only transfer of information over the Internet is the SQL (structured query language)
statement sent to the server and the resulting data retrieved from the server and sent back to the clients'
PCs. All of the other processing is executed on the individual PC (see Figure 6).
Technology Commercialization and Education
This program area focuses on commercializing products developed from research performed under
CATM. As an example, through CATM, the EERC is working toward developing a catalyst for
reducing vaporous, halogenated Hg compounds to elemental Hg vapor. A polyphenyl sulfide polymer
was found to effectively reduce HgCl2 vapor to elemental Hg vapor, which is a necessary step for
current continuous emission monitor (CEM) technologies which are reliable only for measuring
elemental Hg on-line (21).
Within the mission of CATM, there is a strong emphasis on information dissemination and education
related to research on air toxic metals. The following list a few events that were coordinated with the
help of CATM:
CATM coordinated an international workshop on the fate of trace elements in power plants that was
cosponsored by EPRI and DOE. The peer-reviewed papers and discussion have been published by
Elsevier Science Publishers for worldwide distribution (1).
CATM, through cosponsorship with DOE and industry, coordinated an international conference in
Prague, Czech Republic, entitled "Energy and Environment: Transitions in East Central Europe,"
which focused on energy and environmental problems, including the effects of air toxic metals.
CATM has developed a home page on the Internet by which outside organizations can access up-to-
date information on air toxics.
- The CATM database containing over 4000 data results is accessible through the CATM home
page.
- CATM newsletters are posted and made accessible through the CATM homepage. CATM has
published and distributed six newsletters to over 1000 organizations worldwide.
- CATM has conducted three annual meetings and Program Affiliates meetings.
CATM
application
resides on
client's computer.
CATM Database
at the EERC
EERC WP14028.CDR
Figure 6
Architecture of Data Access Via the Internet
-------
Research results generated from CATM have resulted in over 50 published papers and
presentations in the United States and internationally.
A 1-day short course on air toxic metals, entitled "Trace Metals in Industrial Applications," was
developed to provide a blend of fundamental and practical information related to the power
industry. The goal of the short course is to provide an overview of trace metal behavior and control
in systems utilizing coal, oil, and natural gas or alternative fuels such as biomass, municipal solid
waste, and refuse-derived fuels. The course covers topic areas such as trace elements in fuels,
sampling and analysis, trace element transformations and partitioning, control technologies,
prediction and modeling, ash by-products, and a regulatory overview (22).
Conclusions
Transformation Mechanisms
Extensive bench- and pilot-scale combustion testing has shown that, on a mass basis, As, Ni, and
Pb are partitioned into the supermicron size fraction according the particle-size distribution. On a
concentration basis, Ni and Pb tend to be more concentrated in the submicron size and As in the
l-3-/im size range. Se partitions into a vapor and solid phase with a larger solid-phase fme
paniculate component, depending on the fuel type and combustion conditions. Hg, in almost all
cases, partitions primarily to the gas fraction.
Mercury in fossil fuels is initially transformed to gaseous elemental mercury, Hgฐ(g), during the
combustion process. Subsequent reactions with flue gas components at lower temperatures produce
inorganic mercuric compounds (HgX[s, g], where X is C12, O, SO4, etc.). Kinetic data and
speciation analysis results indicate that not all of the Hgฐ(g) is oxidized. In some cases, elemental
mercury may remain predominantly. Data in the literature report that generally greater than about
50% of the elemental Hgฐ (g) reacts with oxidants in combustion flue gases. From the tests
performed through CATM, it was observed that an average of 43% of the Hgฐ(g) reacted.
A unique, state-of-the-art combustion system for performing tests regarding air toxic metal
formation, emissions, and control was designed, built, and tested. System results compared
favorably to those from other experimental facilities.
Fly ash plays a major role in postcombustion transformations.
Analytical and Sampling
Standard reference methods for air toxic metals under oxidizing environments may not work
without modification for reducing environments.
EPA Method 29 has been shown to overestimate the oxidized Hg form by approximately
10%-30%, depending upon fuel type; other speciation methods are currently being developed and
evaluated (10).
The sample preparation step introduces the most error with respect to accurate trace element
measurement. Procedures must be followed consistently and carefully for each type of analyte and
element.
WDXRF and AAS techniques can be calibrated to provide acceptable results, with precautions
taken during the sample preparation procedure.
-------
Careful determinations must be made for the LLQ and LLD for each element for a particular
sample prior to analysis.
Control Technologies
Hg, because of its volatility, is known to be difficult to collect in combustion systems using
conventional gas paniculate control systems but was also demonstrated to have an even higher
volatility in highly reducing gasification systems.
Most trace element emissions, with the exception of Hg and Se, are associated with paniculate and
thereby controllable using conventional paniculate collection devices. New PM2.5 standards may
require additional controls.
Ultrahigh collection (above 99.99%) of fly ash from combustion sources is attainable with the use
of advanced filter media. Advanced filter media cost significantly more than what is currently used
today.
Two precombustion sorbents, zeolite and kaolinite, were found to be only marginal in effectiveness
for controlling air toxic metal emissions, and a lime precombustion sorbent was found to have some
control with regard to Hg and Se emissions for certain coals. Preliminary tests with activated
carbons reveal that they are the most promising sorbent material for capturing volatile air toxic
elements such as Se and Hg. Unfortunately, not all sorbents work effectively given the range of fuel
types, mercury concentration, and operating conditions under which they must work.
With the financial support of Illinois Clean Coal Institute, DOE, and EPA CATM, an advanced
coal-cleaning procedure using subcritical water to remove sulfur and HAPs from coal has been
developed. The procedure consists of pressurizing a coal-water slurry mixture, followed by heating
the mixture to temperatures approaching 370ฐC.
Coal combustion fly ash was discovered to behave differently relative to its ability to collect
potential air toxic metals. Some coal ashes can effectively collect up to 98% of the Hg released
from the coal during combustion while others may only collect 5 %.
Modeling and Database
CATM serves as a centralized repository of integrated information related to air toxics. CATM has
developed an internationally accessible database containing over 4000 air toxic measurements that
can be easily retrieved via a graphical user interface to assist in health risk assessments and
developing preventive and control emission strategies. The data are maintained in a relational
database engine at the EERC; however, the application has been designed such that the data are
accessible through the Internet.
CATM has evaluated several thermochemical equilibrium computer models, their advantages and
disadvantages, and their applicability to various industrial processes.
Technology Commercialization and Education
Through CATM, the EERC is working toward developing a catalyst for reducing vaporous,
halogenated Hg compounds to elemental Hg vapor.
-------
CATM coordinated an international workshop on the fate of trace elements in power plants that was
cosponsored by EPR1 and DOE.
CATM, through cosponsorship with DOE and industry, coordinated an international conference in
Prague, Czech Republic, entitled "Energy and Environment: Transitions in East Central Europe,"
which focused on energy and environmental problems, including the effects of air toxic metals.
CATM has developed a home page on the Internet by which outside organizations can access up-to-
date information on air toxics.
A 1-day short course on air toxic metals, entitled "Trace Metals in Industrial Applications," was
developed and is available to provide a blend of fundamental and practical information related to
the power industry.
Acknowledgments
The authors express their thanks to the following sponsors: North Dakota Industrial Commission, Basin
Electric Power Cooperative, Cooperative Power Association, Minnesota Power, Montana-Dakota
Utilities Company, Electric Power Research Institute, U.S. Department of Energy Federal Energy
Technology Center, EERC Center for Air Toxic Metals Program sponsored by EPA under Assistance
Agreements CR821518 and CR823173, Otter Tail Power Company, Northern States Power Company,
Mitsubishi Heavy Industries, Ltd., and Consolidated Edison/ESEERCO.
References
1. Benson, S.A.; Steadman, E.N.; Wixo, C.Y. "Toxic Trace Elements in Coal-Fired Power Plants:
Identifying Current Research Needs and Direction," topical report for the period Sept. 29, 1992 -
Dec. 31, 1994, for U.S. DOE DE-FC21-86MC10637, Task 7.46, and Electric Power Research
Institute RP9002-07; EERC Publication 94-EERC-01-9, Jan. 1994.
2. Pavlish, J.H. CATM Prospectus, Revised February 1995.
3. CATM Staff. "1995-1996 Annual Report of the Center for Air Toxic Metals," EERC publication,
Jan. 1997.
4. CATM Staff. "1994-1995 Annual Report of the Center for Air Toxic Metals," EERC publication,
Jan. 1996.
5. CATM Staff. "1993-1994 Annual Report of the Center for Air Toxic Metals," EERC publication,
Jan. 1995.
6. Benson, S.A.; Pavlish, J.H.; Zygarlicke, C.J. Erickson, T.A.; Galbreath, K.C.; Schelkopf, G.L.;
O'Leary, E.M.; Timpe, T.C.; Anderson, C.M. "Center for Air Toxic Metals Years 2 and 3 Final
Technical Report (Draft)," EPA Assistance Agreement C R 823173, EERC publication, Jun. 1997.
7. Zygarlicke, C.J.; Pavlish, J.H. "Partitioning of Trace Element Species in Coal Combustion Systems
and Impacts on Control," In Proceedings of the 22nd International Technical Conference on Coal
Utilization and Fuel Systems; Clearwater, FL, March 16-19, 1997, pp 903-914.
8. Pavlish, J.H.; Gerlach, T.R.; Zygarlicke, C.J.; Pflughoeft-Hassett, D.F. "Mitigation of Air Toxics
from Lignite Generation Facilities," final report prepared for Lignite Research Council, Basin
Electric Power Cooperative, Cooperative Power Association, Minnesota Power, Montana-Dakota
Utilities, Co., Electric Power Research Institute, U.S. DOE, and EPA, Center for Air Toxic
Metals; EERC publication, Oct. 1995.
9. Zygarlicke, C.J.; Pavlish, J.H. "The Fate and Control of Mercury Emissions from Coal-Fired
Systems," In 13th Annual International Pittsburgh Coal Conference Proceedings: Coal - Energy
and the Environment; 1996; Vol. 1, pp. 241-246.
-------
10. Laudal, D.L.; Heidt, M.K.; Brown, T.D.; Nott, B.R.; Prestbo, E.P. "Mercury Speciation: A
Comparison Between EPA Method 29 and Other Sampling Methods," Presented at the 89th Annual
Meeting of the Air & Waste Management Association, Nashville, TN, Paper 96-WA64A.04, June
1996.
11. Galbreath, K.C.; Zygarlicke, C.J. "Mercury Speciation in Coal Combustion and Gasification Flue
Gases," Environmental Science & Technology 1996, 30 (8), 2421-2426.
12. Miller, S.J. "Review and Assessment of Results from Comprehensive Characterization of Air Toxic
Emissions from Coal-Fired Power Plants," quarterly technical progress report for the period
Jan. - March 1994 for U.S. DOE DE-FC21-93MC30097; EERC publication, April 19, 1994.
13. Miller, S.J.; Dunham, G.E.; Laudal, D.L.; Heidt, M.K. "Measurement of Reentrainment Effects
in Electrostatic Precipitators and Fabric Filters," PARTEC 95, International Congress for Particle
Technology, 3rd European Symposium on Separation of Particles from Gases, Numberg,
Germany, March 21-23, 1995.
14. Benson, S.A.; Pavlish, J.H.; Erickson, T.A.; Katrinak, K.A.; Miller, S.J.; Steadman, E.N.;
Zygarlicke, C.J.; O'Leary, E.M.; Dunham, G.E.; Pflughoeft-Hassett, D.F.; Galbreath, K.C.
"Center for Air Toxic Metals Final Technical Report," EPA Assistance Agreement C R 821518-01,
EERC publication, Dec. 1995.
15. Miller, S.J.; Heidt, M.K. "Air Toxic Fine Paniculate Control," quarterly technical progress report
for the period Jan. - March 1994 for U.S. DOE DE-FC21-93MC30097; EERC publication,
April 29, 1994.
16. Brekke, D.W; Erickson, T.A. "Assessment of Hazardous Air Pollutants for Advanced Power
Systems," final topical report; EERC Publication 95-EERC-12-03, Dec. 1995.
17. Anderson, C.A.; Timpe, R.C. "Organic Sulfur and HAPs Removal from Coal with Subcritical
Water," final technical report prepared for Dlinois Clean Coal Institute and EPA, Center for Air
Toxic Metals, EERC publication, Sept. 1996.
18. Young, B.C.; Pavlish, J.H.; Gerlach, T.R.; Zygarlicke, C.J. "Mitigation of Air Toxic Elements
from the Combustion of Low-Rank Coals in Power Generation Plants," Presented at the Air &
Waste Management Association 89th Annual Meeting and Exhibition, Nashville, TN, June 23-28,
1996.
19. Frandsen, F.; Erickson, T.A.; Kuhnel, V.; Helble, J.J.; Linak, W.P. "Equilibrium Speciation of
As, Cd, Cr, Hg, Ni, Pb, and Se in Oxidative Thermal Conversion of Coal: A Comparison of
Thermodynamic Packages," In Proceedings of the 3rd International Symposium on Gas Cleaning at
High Temperatures, Karlsruhe, Germany, Sept. 18-20, 1996, pp 462-473.
20. Peck, W.D.; O'Leary, E.M.; Erickson, T.A. "Application of the Center for Air Toxic Metals
(CATM) Database," In Proceedings of the 13th Annual International Pittsburgh Coal Conference:
Coal - Energy and the Environment; Chiang, S.H., Ed.; Pittsburgh, PA, Sept. 3-7, 1996; Vol. 2,
pp 1350-1355.
21. Schelkoph, G.L.; Pavlish, J.H.; Zygarlicke, C.J. "Evaluation of a Catalyst for the Conversion of
Oxidized Mercury Compunds of Elemental Mercury in Vapor Form," Presented at the Air & Waste
Management Association 90th Annual Meeting & Exhibition, Toronto, Ontario, Canada,
June 8-13, 1997.
22. CATM Staff. "Trace Metals in Industrial Applications," Short Course, EERC Publication,
Oct. 1996.
-------
FIELD VALIDATION OF SAMPLING PROCEDURES FOR THE SPECIATION OF
MERCURY IN FLUE GAS
Franklin M. Stevens Jr, William R. Froberg, and W. Steven Lanier; Energy and Environmental Research
Corporation 1001 Aviation Parkway, Suite 100 Morrisville, North Carolina 27560
Scott Rauenzahn, US EPA, Office of Solid Waste Management, Waste Management Division (5302W)
2800 Crystal Drive, Crystal City, 6th Floor, Arlington, VA 2202
Dan Bums, (WSRC-DOE), Savannah Technical Center, Bldg.. 676T, Room 11, Aiken, SC 29808
Peter Grohse, and James A. O'Rourke; Research Triangle Institute, Cornwallis Road, Research Triangle
Park, NC 27709
ABSTRACT
In order to assess the performance of continuous emissions monitoring systems (CEMS) in
measuring total mercury, it was necessary to determine the species of the mercury present in the flue gas.
It was therefore necessary to make reference measurements with a sampling train capable of speciating
mercury emissions. Such a train was developed by the EPA office of Research and Development; and
their contractor, Research Triangle Institute. A quad train was used. Measurements were made on the
stack, but at a single point, not traversing. Two of the trains were dynamically spiked with mercury and
mercuric chloride generated by a permeation device. The spiking was carried out by introducing gas
phase mercury and mercuric chloride sequentially into the probes of two of the four trains while
sampling from the stack. The spike amounts were on the order of the amount of mercury collected from
the stack gas during the sampling period, and were verified by sampling the spike flow at the point of
introduction into the MM29 probe with a midget impinger Method 29 train. A fresh M29 verification
train was used for each spike to allow accurate determination of the amounts of mercury and mercuric
chloride spiked. Nine (9) one to two-hour runs of the quad trains were made, with spiking of two of the
trains carried out each time.
INTRODUCTION
The U.S. Environmental Protection Agency (EPA) regulates the burning of hazardous waste in
incinerators under 40 CFR Part 264/265, Subpart O, and in boilers and industrial furnaces under 40 CFR
Part 266, Subpart H. The Agency has proposed revised regulations applicable to hazardous waste
burning incinerators (HWIs), cement kilns (CKs), and light weight aggregate kilns (LWAKs). (See 61
FR 17358, April 19, 1996.) These devices are collectively known as hazardous waste combustors
(HWCs). This rule is scheduled to be promulgated in December 1996. Included in the proposed
regulations are draft performance specifications for total mercury (Hg) continuous emissions monitoring
systems (CEMS) and requirements for their use. In support of these proposed monitoring requirements,
1
-------
EPA needed to test commercially available Hg CEMS to ensure that such devices can meet the proposed
performance specifications and data quality objectives.
In order to assess the performance of the CEMS in measuring total mercury, it was necessary to
know the speciation of the mercury in the stack gases sampled during the test. It was, therefore,
necessary to make reference measurements with a sampling train capable of speciating mercury
emissions. Such a train has been developed by the EPA Office of Research and Development (ORD)
and their contractor, Research Triangle Institute (RTI). This train has been developed based on
information from the extensive testing of various sampling trains carried out at the Energy and
Environmental Research Center at the University of North Dakota for the Electric Power Research
Institute (EPRI) and the Department of Energy (DOE). The speciating train proposed by ORD is based
on Method 29 (M29) and consists of two additional water filled impingers prior to an acidified peroxide
impinger. This proposed draft method, herein called Proposed Draft Method 101B (M101B) has been
extensively laboratory tested by RTI. Energy and Environmental Research Corporation (EER)
performed the field validation for the Proposed Draft Method 101B train according to EPA's "Protocol
For The Field Validation of Emission Concentrations From Stationary Sources" (EPA Method 301) as
part of the Hg CEMS demonstration project.
EPA Method 301 provides procedures for determining and documenting systematic error (bias)
and random error (precision) at a permissible concentration of the sources emissions. This protocol, as
specified in the underlying regulations, is to be used whenever a new method is proposed to meet a
United States Environmental Protection Agency (EPA) requirements in the absence of a validated
procedure. The procedure involves introducing known concentration of an analyte to determine bias and
collecting multiple simultaneous samples to determine precision. For this study, nine (9) runs were
made using a quadruplet train configuration. A quadruplet train bundles the probes from four (4)
separate sampling trains so that the tips of the sampling nozzles were in a 6x6 centimeter square area
measured from the inside edge of the probe tip with the pilot tube in the center. Two (2) of the four (4)
trains were spiked with known concentrations of mercury (Hg) and mercuric chloride (HgCl2) using a
permeation device. Spiking occurred at the base of the probe prior to the filter assembly.
SITE DESCRIPTION
The site selected for the Hg CEMS demonstration is Kiln number 2 at the Holly Hill, Inc.,
cement manufacturing facility located in Holly Hill, South Carolina. This cement kiln co-fires
hazardous waste with various other fossil and waste fuels.
Holnam operates two (2) wet process kilns at the facility. Kiln #1 is 12'6" in diameter and 500'
long, with a design capacity of 1,300 tons of clinker per day. The larger Kiln #2 is 18'6" in diameter
and 580' long, with a design capacity of 2,100 tons of clinker per day. The cement production process
involves: quarrying and crushing; grinding and blending; clinker production in the kiln; finish grinding;
and packaging.
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, off-site 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 kirns 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 MM Btu/hr and 600 MM Btu/hr for Kiln #1 and Kiln #2,
respectively.
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), which finally fused 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 2,700ฐF, while the gaseous material reaches a
temperature in excess of 3,000ฐF. The clinker produced from this process is conveyed through a grate
cooler and is cooled by air from forced draft fans to about 175ฐF. 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 assembled for Holnam.
SAMPLING METHOD
Historical data from the site suggested that 99% of the mercury was gaseous phase. Since the
quad train setup is so massive hi size it was felt that single point sampling would be the best option for
this test program. This would be possible as long as the majority of the mercury was actually in the gas
phase. To verify this fact we ran a method 201A sampling train and a M29 sampling system prior to
performing the Method 301 study. Less than 1% of the mercury was found on the particulate fractions
and the M29 sampling train substantiated the emissions levels at 15-30 ug/dscm . Since particulate
matter concentrations were not determined, no acetone rinse was used during the recovery procedures.
The front half of the sampling train was recovered with 0.1 N Nitric acid and analyzed for mercury only.
The total mercury sampling train was configured as specified in 40 CFR Part 60 Appendix A
Method 29. The nozzle, probe liner, and filter housing were constructed of quartz glass with a Teflon
filter support. The filter housing will be followed by a series of seven quartz impingers. Six (6) of the
impingers were modified Greenburg-Smith impingers and one was a standard Greenburg-Smith
impinger. Table I lists the impinger contents for M29.
Quartz fiber filters (3 inch diameter) were used in both sampling trains for all testing. The filters
were blanked for mercury prior to use in the field. Each filter contained less than 1.3 ug Hg / square
-------
inch of surface area. American Standards Testing Materials (ASTM) specification Dl 193-77, Type II
water is specified for the sampling trains. However, since Type II water was not commercially
available, J.T Baker Ultra Resi-Analyzed water was used as an equivalent reagent. The nitric acid,
hydrochloric acid, and hydrogen peroxide (30% (v/v)) solutions used were J.T Baker Ultex II ultrapure
reagents. Concentrated sulfuric acid was used to produce the acidified potassium permanganate
solutions from potassium permanganate crystals. All reagents were blanked for mercury prior to use.
Blank values were less than 1 ng/ml.
The acidified peroxide solution was made by adding 50 ml of concentrated nitric acid to 500 ml
Ultra Resi-Analyzed water with stirring. 333 ml of 30% hydrogen peroxide was then added to this
mixture. The final volume was adjusted to 1000 ml with Ultra Resi-Analyzed water. Reagent blanks
and train blanks were taken in the field to verify that the solution contained less than 2 ng mercury /ml.
The permanganate solution was made by adding 100 ml of concentrated sulfuric acid to 800 ml
Ultra Resi-Analyzed water with mixing. The final volume was adjusted to 1000 ml with Ultra Resi-
Analyzed water to produce 10% (v/v) sulfuric acid. The potassium permanganate reagent was made on
a daily basis. Forty (40) grams of the potassium permanganate crystals were dissolved into the 10%
sulfuric acid solution in a separate container. The final volume was adjusted to 1000 ml using the 10%
sulfuric acid solution. The solution was then filtered through Whatman 541 filter paper. The storage
container was to be vented to remove the potential of explosion due to reactions between the
permanganate and acid. A blank of the final solution was taken in the field, after the samples were
collected. Samples were recovered immediately after each test run and placed in cold storage until
delivery to the analytical lab.
Proposed Method 101B consisted of a standard Method 29 sampling train modified to separate
the oxidized and elemental forms of mercury during sampling. The first acidified peroxide impinger
was replaced by two water filled impingers. The water reagent captures oxidized mercury species
(Hg+2),
while the acidified peroxide solution absorbs SO2 (to prevent SO2 from collecting in the permanganate
impingers, which compromises their mercury capture efficiency). Elemental mercury (Hgฐ) passes
through the water and acidified peroxide solutions and is collected in the acidified permanganate
solution. Laboratory testing (Grohse. 1996) has shown that >99% of Hg+2 is retained in the water
impmgers (testing with HgCl2 only), and >99% of Hgฐ is retained in the acidified permanganate.
Recovery of the Ml 01B train was modified to provide individual fractions for the first 5 impingers.
Each fraction was analyzed separately to verify the collection efficiency of oxidized forms of mercury
across these impingers.
Table II lists the impinger contents for proposed Method 10IB. Figure 1 gives a schematic
representation of the M101B sampling system.
Sampling runs were conducted as specified in EPA 301, Section 5.3, "Procedure for
Determination of Bias and Precision in the Field (for Analyte Spiking)" This procedure requires a
minimum of six runs of a quad train, and provides twenty four total samples. Two of the trains from
-------
each run were spiked dynamically while sampling stack gas. The spiking procedure outlined in Section
5.1.2.1 of M301 which requires sampling of the gas phase spike followed by sampling of the exhaust
gas stream. A more stringent test is to spike the impingers while sampling the exhaust gas stream, which
was the approach used for this validation. This approach measured the effects of the stack gas matrix on
the collection efficiency of the impinger solutions.
Each sampling run consisted of four simultaneous M101B sampling trains. Sampling runs were
single point with a duration of one (1) hour for each single point run . Sample volumes were
approximately one (1) dscm. Each individual train was operated as specified by EPA Method 29. Two
of the four sample trains were spiked dynamically with Hg and HgCl2 during sampling. The volume of
the spikes were small in comparison to the sample volume, and required no adjustments to maintain
isokinetic sampling.
Recovery of the proposed Ml 01B was similar to M29 with the following exceptions:
The first five (5) impingers were recovered, stored, and analyzed individually. Each
impinger was rinsed three tunes with 0.IN nitric acid. The rinses were collected and
stored in the corresponding impinger recovery containers.
Ten fractions were recovered from each sample train: Front-half (F/H); Back-half(B/H);
Impingers 1 -5 were recovered separately; Impingers 6 and 7, combined; and an 8N HC1
rinse of impingers 6 and 7. Table III identifies the contents of each sample fraction.
Figure 2 illustrates the recovery procedures for the proposed M101B sampling train.
The test procedures utilized a quad train with four (4) co-located probes. Variations from
Method 301 protocol included the following differences:
1. Spiking was performed online simultaneously with stack gas sampling.
2. Spiking was performed at the expected stack concentrations of 10-30 ng/dscm, not at the
emission standard level.
3. Spiking was not exact due to the systematic bias related to the generation of the reference
material. Spike concentrations were verified with a separate Method 101A sample train.
4. Statistical analysis was modified from the procedures outlined in M301. The variations
in the spiking procedures which were used for this program required a multi-variable
analysis of the data instead of the single variable analysis normally performed with
M301.
-------
Gas phase mercury (Hgฐ) and mercuric chloride (a representative Hg+2 compound which is
thought to be a major component of total mercury emissions) were injected sequentially into the front
half of the filter housing assembly of proposed M101B while sampling stack gas. The spike gas was
maintained at a temperature > 120ฐC. The concentration of each spike was verified using a modified
Method 101A sampling train that sampled spike gas from a distribution manifold immediately preceding
the Method 301 quad train. While nine runs of the quad train were completed, only six of the runs
yielded data valid for statistical analysis. A single fixed sample point was used for all sampling, since
the particulate was less than 10 microns in size and previous site data demonstrated that less than 1% of
the Hg was found on the paniculate.
Spiking of gas phase mercury and mercuric chloride was accomplished using Dynacalฎ
permeation tubes supplied by Research Triangle Institute, and manufactured by VICI Metronics Inc.
The permeation devices are small, inert capsules containing pure chemical compounds in a two phase
gas/liquid equilibrium. At a constant temperature, the devices emit a gas phase compound through a
permeable membrane at a constant rate. The permeation rate is as constant as the heat source, and
remains stable until the entire mass of the reference material is completely consumed.
It was preferred that permeation rates could be certified gravimetrically to NIST traceable
standards. However, this certification procedure is not available for mercuric chloride permeation tubes,
due to the physical properties of the gas. Even if NIST traceable standards were available for both
mercury and mercuric chloride, the complexity of the dynamic spiking system prevented the use of
NIST traceable flow and temperature control throughout the system. Consequently, permeation tubes
with nominal flow rates of 1000 r|g/min were used for reference material standards, and concentrations
were verified with a modified Method 101A sample train.
Spike concentrations were initially estimated from the permeation rates demonstrated by RTI
under laboratory conditions. Actual concentrations of the spike gases used in the field validation
procedures were produced by dilution, and verified with a separate modified Method 101A sample train.
The concentrations measured by the Ml01A 'Verification Train' for each individual run were used as
the reference values for Method 301 statistical analysis of the final data.
TEST RESULTS
The test program was conducted between July 10 and July 23, 1996. The verification of total
mercury emission concentrations was determined on July 10, 1996. Following the completion of the
run, all field equipment was broken down and stored onsite due to the approach of Hurricane Bertha.
The Holnam facility was partially shutdown, and put on stand-by until the hurricane had passed through
the area. The test site equipment was re-assembled on July 17, 1996. Method 301 validation procedures
were conducted between July 18andJuly22, 1996. Nine (9) runs of the quad proposed Method 101B
were completed during this period. Kiln operation remained stable for the duration of the testing. The
data from the initial run was used to verify operation of the quad sampling and spiking system . The
actual test matrix is shown in Table IV.
Results from the analyses of each fraction of the proposed Method 101B quad trains are
-------
presented in Table V. Since six (6) runs are required for Method 301 statistic, three runs were not
evaluated during statistical analysis. The choice of runs which were evaluated was based on the
continuity of testing and other QC criteria. Spike results from the analyses of the Method 101A midget
impinger spike verification trains are presented in Table VI. The percent recovery of spikes in each
quad train is presented in Table VII. The total mass of mercuric chloride detected is the sum of the
masses detected in the back-half rinse, and Impingers 1,2, 3,4, and 5. The total mass of mercury
reported is the sum of the masses detected in Impingers 6, 7, and the HC1 recovery rinse. Data from the
front half rinses and filter weights were not considered in the recovery of mercury or mercuric chloride,
since the catches are not speciated by the front half of the sampling train. However, this data was
considered when evaluating the sample train for total mercury.
As shown in the test matrix, ten (10) quad train runs were completed between July 10 and July
23, 1996. During Run 1, a leak check difficulty was encountered which required modification of the
filter housing glassware components before testing could proceed. Modifications were completed
during the partial plant shutdown caused by Hurricane Bertha. Spiking data from trains B and C of Run
1 were not used in further statistical analysis. Run 6 was excluded due to a leak in one of the quad train
components. Run 7 was scratched due to extreme weather conditions which prevented completion of
the spiking sequence.
Of the remaining seven quad trains, only data from six trains were considered for statistical
analysis, as required by the method. Examination of the results indicates a potential outlier during Run
2. The filter data of Run 2, Train C was extremely high, 21.795 |ig of total mercury, almost 100 time
higher than the average filter mercury gain for the other qualified runs. The corresponding mercuric
chloride recovery was also low, indicating that the loss of mercuric chloride had occurred before the first
impinger. The cause may have been a cold spot in the filter housing, which is very possible since this
was the first run following remobilization, and sufficient tune may have not been allowed for thermal
conditioning of the glassware. Temperature is a critical variable when handling mercuric chloride, due
to its tendency to sublimate at moderate temperatures. Data from Run 2 was excluded from Method 301
statistical analysis.
The experimental procedure was a modification of Section 5.3 of Method 301 for Analyte
spiking. The statistics that apply are those of Section 6.3 with the following changes. The bias equation
given in Method 301 (Eq. 6-13) is
B=Sm-Mm-CS
where B = Absolute Bias at the spike level
i n
Sm= mean of the spiked samples, 2J-S.
n ,=i '
Mm= mean of the unspiked samples, ^,Mj
n ,=i
-------
CS= calculated value of the spike level.
Equation 6-13 can be derived by averaging the individual biases calculated for spiked and
unspiked pairs of trains:
= -X>, = -E (s, ~Mr cs)
n , = 1
S^-M^ --ECS.
'"ซ,=!
The statistics in Method 301 assumes that the spike values are known and equal, therefore;
i "
Cs, = CS,so ECS. =n CS , thus -Ecs, = cs and Eq.. 6-13 is recovered.
n , = i
The general case of Eq. 6-13 is thus (from the last line off the equation above):
B = Sm-Mm-CSm (1)
where 05^=^ CS: Thus Equation (1) should be used to calculate the bias of speciating mercury train
n ,=i
validation, with CS being the spike values determined by the verification train. The values of CS are
simply averaged. The data is not paired.
This is not the case for the precision calculation of the spiked trains required to assess the
statistical significance of the bias. In Method 301 the standard deviation of the spiked samples is
determined with Eq. 6-14:
,
Erf,.2
2n
where d, are the differences of the paired values (A and B) from the spiked trains for each run: d, = A,
B,.
-------
The bias was evaluated for statistical significance with a t-value calculated as follows;
B
SDM
<; 2.201
where:
SDM =
A subsequent June 1996 revision of Method 301 recognized the fact a significant variation may
also exist in the unspiked samples due to changes in mercury concentration in the stack effluent. A new
approach was taken which used pooled statistics. The variance of both the spiked and unspiked samples
was included in the calculation of the t-value;
SD
pooled
Ss2 + Sm 2
2.228
/( S* + Sf ) /12
Where;
Variance of the unspiked samples
Variance of the spiked samples
While this approach considered the variability of the spiked and unspiked samples of the quad
train, it still assumes that the spike value is a known and constant value. However, the data in Table VI
shows that the spike value was not constant, especially in the case of mercuric chloride. The variability
was attributed to the precision and bias of both the permeation tube spiking system that was utilized to
generate the in-situ spike gas, and the uncertainty associated with Method 101A measurements of the
spike concentrations.
-------
In order to evaluate the variability of the spiked samples, the unspiked samples, and the spike
concentration, the pooled statistics approach was expanded to include three variables.
n -1
SDpoaM
s 1
upooled \
<. 2.131
n
The resultant value of t3 was compared to a value obtained from a t-distribution with (3n - 3)
degrees of freedom and a significance level of a = 0.05. Note that this test is valid if the value of n
paired sample sizes is equal for the spiked trains, the unspiked trains, and the spike values. There is no
need to perform a test on equality of variance to judge whether a t-test with unequal variances should be
performed.
Note that this calculation of precision is used for a statistical test of the bias, so it could include
all sources of measurement variability (from the measurement of speciating train A and B values and
M101 verification train values (AsandBs)), which it does. However, spike level variability is removed, as
wanted.
The results of the Method 301 analysis are shown in Tables VIII and IX. Table VIII represents
the statistical approach from the June 1996 revision of Method 301. Since the flue gas concentrations
and the spike levels had a fairly large variation we also performed a three variable pooled statistic that
was worked out by SAIC. Table IX summarizes the results using three variable pooled statistic. Results
are shown for a) Mercuric chloride spiking, b) Mercury spiking, and c) Total mercury species spiking.
Six runs of the quad train were evaluated: Runs 3, 4, 5, 8, 9, and 10. The total mass of each spike
amount to the quad train, CS, was calculated from a mass balance of the dynamic spiking system and
the verification train flow rate and concentration. This quantity was then converted to a concentration
basis for to account for differences in sample volumes between the spike verification runs and each
proposed Method 10IB train .
Examination of the results in Table VIII shows the difficulty of delivering precise amounts of
mercuric chloride at a low concentration. The mean values of the spike and the spiked samples are
almost identical. The bias of the spike level was very large when compared to the mean value of the
spike. The variance of the spiked samples was also large. This resulted in a large t-value, poor
10
-------
precision, and a correction factor outside of the acceptable range for HgCl2.
Precision and bias results were excellent for mercury. The calculated t-value was well within the
critical value for a 2-sided t-distribution for eleven degrees of freedom. The precision and bias values
reflect the higher recovery rates and consistency of the data when compared to the mercuric chloride
results. A correction factor is not required for the mercury results .
Total mercury results include the analytical results of the front half washes and filter weights
were used for these calculations. However, as indicated in Table VII, the difference in the analyte
recoveries with and without the front half and filter fractions was typically in the range of 1% to 3%,
indicating that very little particulate bound mercury material was recovered from the sample trains. The
calculated t-value for analysis of total mercury species falls between the t-values for mercury and
mercuric chloride spiking individually.
Table IX shows the results of the multi-variant analysis developed by EER and SAIC. This
approach removes the variability associated with the spike values and the mercury concentrations in the
flue gas. The elimination of these variances results in a t-statistic less than the critical value for mercuric
chloride, mercury and total mercury. The relative standard deviation was still high for mercuric chloride
however, it is not reflected in the t-statistics.
CONCLUSIONS
Based on the results of the study we believe that there is a need to evaluate the statistics in M301.
The M301 statistics currently assume that the concentration of the pollutants being measured are
constant and that the spiking value is a constant. With this being true, it is possible to perform a single
variable analysis of the data. Experience indicates that inherently there can be substantial heterogeneity
in stack gas concentrations of a pollutant in both location and time. Since the sampling procedures
sample discreet points in time, this heterogeneity will translate itself into sampling variability. There
was also significant variability in the generation of a Hg and HgCl2 in the gas phase. When performing
a single variant analysis all these variabilities become lumped into one and inflate the variability of the
sampling procedure being evaluated. Thus in practice you have a multi-variant situation where a single
variant analysis no longer applies.
The t-values calculated for the proposed Method 10IB sampling train by the multi-variant
statistical analysis approach show the Bias at the spike levels is statistically insignificant for mercury,
mercuric chloride, and total mercury for the flue gas conditions at this facility. Performing the single
variant approach in M301 for a pooled statistic results in a very high t-values and poor precision for the
mercuric chloride and total mercury. The sampling train passes the single variant analysis for mercury
speciation.
We have not finalized the results of this tests at this time. We are continuing to look at how the
multi-variant analysis is performed. We have gotten support from SAIC's Statistical group in the
development of the mult-variant approach to analyzing the data. We are working to ensure that the basis
that is used for the final evaluation will be the best approach for the type of procedures being utilized.
11
-------
TABLE I. EPA METHOD 29IMPINGER CONTENTS
Impinger Number
1
2
3
4
5
6
7
Impinger Type
Modified G/S
Modified G/S
G/S
Modified G/S
Modified G/S
Modified G/S
Modified G/S
Contents
Empty
5%HNO3/10%H2O2
5%HNO3/10%H2O2
Empty
4% KMnCV 10% H2SO4
4% KMnO4/ 10% H,SO4
Silica Gel, Indicating
Volume (ml)^
100
100
100
100
~200g
TABLE II. PROPOSED METHOD 101B IMPINGER CONTENTS
Impinger Number
1
2
3
4
5
6
7
8
Impinger Type
Modified G/S
Modified G/S
Modified G/S
G/S
Modified G/S
Modified G/S
Modified G/S
Modified G/S
Contents
Empty
H2O
H2O
5%HNO3/10%H2O2
Empty
4%KMnO4/10%H2SO4
4% KMnO4/ 10% H2SO4
Silica Gel, Indicating
Initial Volume
(ml)
0
100
100
100
0
100
100
~200g
12
-------
TABLE HI: PROPOSED DRAFT METHOD 101B SAMPLE FRACTIONS
Container #
1
2
3
4A
4B
4C
4D
4E
5A
5B
5C
Name
Filter
N/A
Front Half
Back Half
Imp#l
Imp #2
Imp #3
Imp #4
Imp #5
KMnO4
8NHC1
Contents
Filter
Probe Rinse
Nozzle, probe liner, front Yi of
filter housing rinses
Filter support, back V4 of filter
housing rinses
Condensate & rinse
Impinger catch & rinse
Impinger catch & rinse
Impinger catch & rinse
Impinger catch & rinse
Impingers 6 & 7 combined
recovery
Impinger 6 & 7 rinse
Recovery Rinse
Type
None
Acetone
0. IN Nitric
Acid
0. IN Nitric
Acid
0. IN Nitric
Acid
0.1N Nitric
Acid
0. IN Nitric
Acid
0. IN Nitric
Acid
0. IN Nitric
Acid
KMnCX,
solution
8NHC1
Volume
N/A
N/A
100ml
100ml
33ml
33ml
34ml
100ml
100ml
100ml
25ml
Comments
Analyzed for Total Mercury: but
not used in statistics
No PM analysis
Analyzed for Total Mercury: but
not used in statistics
Analyzed for Mercuric Chloride
Analyzed for Mercuric Chloride
Analyzed for Mercuric Chloride
Analyzed for Mercuric Chloride
Analyzed for Mercuric Chloride
Analyzed for Mercuric Chloride
Analyzed for Total Mercury
13
-------
TABLE IV. TEST MATRIX
Run No.
1
2
3
4
5
Date
7/10/96
7/18/96
7/18/96
7/19/96
7/19/96
Test Method
M101B
M101A
M301
M101A
M301
M101A
M301
M101A
M301
M101A
Clock Time
1545- 1790
1545- 1600
1645- 1700
1700- 1715
1725- 1740
1131 - 1457
1331 - 1346
1346- 1401
1421 - 1436
1442- 1457
1726- 1845
1726- 1741
1746- 1801
1810- 1825
1830-1845
1245-1416
1245-1300
1300- 1315
1342- 1357
1401 - 1416
1701 - 1827
1701 - 1716
1720- 1735
1755- 1810
1812-1827
Sampling
Duration
(min)
60
15
15
15
15
60
15
15
15
15
60
15
15
15
15
60
15
15
15
15
60
15
15
15
15
Comments
Pre-screening.
M301 Testing delayed
due to plant shutdown
for
Hurricane Bertha.
Hg spike
HgCl2 Spike
Hg spike
HgCl2 Spike
Hg spike
HgCl, Spike
Hg spike
HgCl2 Spike
Hg spike
HgCl2 Spike
Hg spike
HgCl2 Spike
Hg spike
HgCl, Spike
Hg spike
HgCl2 Spike
14
-------
TABLE IV. TEST MATRIX (cont.)
Run No.
6
7
8
9
10
Date
7/20/96
7/20/96
7/21/96
7/21/96
7/22/96
Test Method
M101B
M101A
M101B
M101A
M101B
M101A
M101B
M101A
M101B
M101A
Clock Time
1013- 1133
1013- 1021
1024-1031
1032-1047
1100-1115
1118-1133
n/a
1043- 1158
1043 - 1058
1103- 1118
1127- 1142
1143- 1158
1347- 1457
1347- 1402
1404- 1419
1425 - 1440
1442 - 1457
1041- 1147
1041 - 1056
1058-1113
1116-1131
1132-1147
Sampling
Duration
(min)
60
15
15
15
15
25
15
15
60
15
15
15
15
60
15
15
15
15
60
15
15
15
15
Comments
{1021- 1024, Leak
repaired; M101B,
Train D}
Hg spike
HgCL, Spike
Hg spike
HgCl2 Spike
Run scratched due to
electrical storm during
testing. Data not used
forM301 statistical
analysis.
Hg spike
HgCl2 Spike
Hg spike
HgCl, Spike
Hg spike
HgCl2 Spike
Hg spike
HgCl2 Spike
Hg spike
HgCl2 Spike
Hg spike
HgCl2 Spike
15
-------
TABLE V. TOTAL HG/HgC12 DETECTED IN ALL ANALYZED FRACTIONS OF QUAD
TRAIN
Run*
2
3
4
5
6
7
8
9
10
Samplin
g Train
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
F/H
Acid
(ug)
0.0000
0.3944
0.4026
0.0002
0.3402
1.6042
0.2121
0.6040
0.0332
1.3523
0.1709
0.0686
0.9478
0.0179
0.0672
0.0473
0.2123
0.0556
0.0231
0.5552
0.0088
0.0103
0.0009
0.0000
0.0080
0.0083
0.0072
0.0068
0.0194
0.0171
0.0181
0.0071
0.0216
0.1728
0.2425
0.0263
Filter
(ug)
0.2347
3.4418
3V?930
0.6865
0.0540
0.0236
0.1056
0.0810
0.0215
1.5293
0.1656
0.0313
0.0303
0.0506
0.1455
0.0329
0.0302
0.0380
0.0392
0.0359
0.0120
0.9488
0.0272
0.0203
0.0292
0.1337
0.2506
0.0190
0.0270
0.1091
0.2582
0.0159
0.0231
0.0813
0.3584
0.0494
B/H Acid
(ug)
0.0113
0.3884
0.0285
0.0159
0.2981
0.3230
7.8952
0.0081
0.6585
0.2090
0.3557
0.1952
0.0020
0.5504
1.3959
0.3522
0.0074
1.4652
0.7447
0.0036
0.3856
4.4499
1.2451
0.5790
0.0231
0.1728
0.2936
0.4570
0.0105
0.3272
0.8549
0.4648
2.8785
1.9040
9.1345
5.0595
Imp 1
Empty
(ug)
3.6050
13.5294
9.1397
3.9733
6.7256
7.5976
28.0079
5.6185
14.0538
23.3508
27.0766
9.3112
5.3818
23.7717
39.3409
3.8610
10.5642
43.4325
21.7451
13.0140
5.8090
14.9530
9.3643
5.9800
12.6519
31.1952
30.9579
17.4093
21.2014
29.4892
28.1386
7.7065
10.1018
23.5177
24.1742
9 1810
Imp 2
H20
(ug)
1.0014
2.6051
0.8539
0.6846
0.7370
0.4528
3.0674
09342
1.3740
3.7960
2.0358
2.6411
0.5398
4.7148
1.4369
1.0946
4.1856
3.4630
5.5499
3.0219
1.7223
1.1781
1.5231
1.1163
1.6853
5.7224
2.7222
2.8773
3.9549
6.4110
3.9172
1.9525
0.7187
2.8389
2.7377
1.9858
Imp 3
H20
(ug)
0.1569
1.0819
0.1237
0.0830
0.1891
0.1398
0.8060
0.1146
0.1985
0.4848
0.3489
0.5049
0.0925
0.9109
0.4719
0.4153
0.5615
0.5719
0.8634
0.9598
0.3422
0.4015
0.1366
0.1438
0.4259
0.6941
0.5187
0.7081
1.5015
0.3554
0.5437
1.4564
0.3679
0.4647
0.2840
0.4898
Imp 4
HNO3/H2O2
(ug)
0.2722
0.5367
0.1963
0.1904
0.5374
2.0908
1.4237
0.4125
0.2344
0.6722
0.5910
0.9800
0.2752
2.1491
1.7430
0.5492
0.2666
0.5632
1.1614
0.7446
0.7735
0.8787
2.0183
0.3114
0.6401
0.6859
0.9805
0.5384
0.8908
1.0484
1.0289
1.3907
0.5168
0.5351
0.3117
0.4927
Imp 5
Empty
(ug)
0.0231
0.0396
0.0234
0.0270
0.1612
0.1180
0.3086
0.0989
0.0580
0.1023
0.1266
0.0586
0.1440
0.2520
0.5128
0.1343
0.1160
0.1611
0.2736
0.1678
0.0439
0.0839
0.0907
0.0460
0.1255
0.1241
0.2655
0.0722
0.1140
0.0602
0.0338
0.0420
0.1032
0.2093
0.1484
0.0698
Imp 6&7
KMn04
/H2SO,
(ug)
6.4271
14.1423
25.5659
6.9113
11.2781
32.7750
14.7913
11.0456
7.2058
19.4005
17.8848
8.6267
9.1769
19.1190
23.1871
6.0401
7.5833
22.8202
21.1084
10.7176
5.6666
7.6156
6.3345
5.9518
7.6645
12.6265
10.5163
7.3752
10.7344
22.5267
19.4756
8.2427
9.8798
18.4642
15.5781
82854
HC1 Rins<
(ug)
0.1637
0.1190
0.1593
0.0915
0.1610
0.0740
0.0303
0.0993
0.0321
0.1371
0.0438
0.0431
0.0286
0.0337
0.0864
0.0393
0.0233
0.0587
0.0648
0.0828
0.0902
0.0857
0.0237
0.0810
0.0722
0.0742
0.0558
0.1039
0.0649
0.0720
0.0595
0.0534
0.0576
0.0983
0.0714
00555
16
-------
TABLE VI. TOTAL MASS OF HG SPECIES SPIKED TO METHOD 101B TRAIN PROBES
B&C
Test Id
MID-R2**-B
MID-R2**-C
MID-R3**-B
MID-R3**-C
MID-R4**-B
MID-R4**-C
MID-R5**-B
MID-R5**-C
MID-R6**-B
MID-R6**-C
MID-R7**-B
b) MID-R7**-C
MID-R8**-B
MID-R8**-C
MID-R9**-B
MID-R98**-C
MID-R10**-B
MID-R10**-C
Total Hg spike to
quad train
(ug)
13.7622
25.6731
10.7672
10.4429
8.9952
9.9875
14.1009
9.1083
22.5680
11.5575
8.6455
- ttfa-i
10.1220
9.4628
5.2033
8.8855
10.0268
10.4726
Total HgCl2 spike to quad
train
(ug)
11.2905
19.7787
29.0670
29.2814
100.4279
21.5973
31.7110
29.8684
57.4229
47.9391
45.9049
' ^ , -, , .-Ufa -
38.6351
37.6752
21.0985
38.3170
10.2026
22.2146
Total Hg species spike to quad train a)
(ug)
25.0527
45.4517
39.8342
39.7243
109.4231
31.5849
45.8119
38.9767
79.9909
59.4966
54.5505
ซ " '< '^.-;'. Va" -,-^H -,-, ฃ;/
48.7571
47.1380
26.3018
47.2025
20.2294
32.6873
Note: a) Spike concentrations as determined from Method 101A midget impinger trains.
b) No spikes completed for Run 7- Probe C
17
-------
TABLE VII. % RECOVERY OF MERCURY AND MERCURIC CHLORIDE SPIKES IN
PROP. METHOD 101B QUAD TRAINS
Run*
2
3
4
5
6
7
8
9
10
Sampling
Train
B
C
Avg
B
C
Avg
B
C
Avg
B
C
Avg
B
C
Avg
B
C
Avg
B
C
Avg
B
C
Avg
B
C
Avg
Total HgCI2 HgC12; no bh rinse
116.42% 113.10%
27.03% 26.95%
71.72% 70.03%
9.69% 9.11%
114.11% 87.67%
61.90% 48.39%
13.38% 13.6%
70.99% 71.3%
42.19% 42.46%
81.41% 80.2%
127.58% 123.5%
104.49% 101.87%
57.17% 54.6%
27.95% 26.4%
42.56% 40.52%
;. - -. . '.;
28.93% 20.3%
n/a n/a
28.93% 20.29%
51.15% 51.3%
44.50% 44.4%
47.82% 47.84%
82.31% 81.9%
37.11% 35.5%
59.71% 58.69%
130.99% 151.2%
93.39% 70.1%
112.19% 110.68%
Total Hg
54.34%
73.72%
64.03%
200.10%
35.51%
117.80%
129.27%
100.56%
114.91%
82.43%
175.71%
129.07%
60.68%
104.71%
82.69%
21.35%
n/a
21.35%
50.56%
33.09%
41.82%
250.48%
111.90%
181.19%
95.21%
62.76%
78.98%
Total Hg Species Total Hg Species
no filter/ fh / bh
95.79% 80.82%
101.23% 53.37%
98.51% 67.10%
63.89% 60.73%
92.89% 73.96%
78.39% 67.34%
25.47% 23.11%
81.16% 80.57%
53.31% 51.84%
80.72% 80.91%
138.02% 135.70%
109.37% 108.31%
57.76% 56.34%
42.27% 41.62%
50.01% 48.98%
29.45% 20.45%
n/a n/a
29.45% 20.45%
! i:-*.* :;" ''''" ''"r ';"":.;; .."" ' :'.'./: f
51.26% 51.17%
42.69% 42.09%
46.97% 46.63%
115.93% 115.24%
51.70% 49.88%
83.81% 82.56%
114.21% 123.46%
85.23% 67.77%
99.72% 95.62%
18
-------
TABLE VIII. SUMMARY OF METHOD 301 STATISTICS for PROPOSED DRAFT METHOD
101B
Statistical Approach: Method 301, June 96 revision
The mean of the spiked samples (Sm)
The mean of the unspiked samples (Mm)
The mean value of the spike, (Cm)
Variance of Spiked Samples (Ss2)
Variance of Unspiked Samples (Su2)
Ratio: unspiked / spiked Variance (F)
Pooled Standard Deviation (SD pooled)
Calculated t value
Correction factor (CF)
ug/dscm
ug/dscm
ug/dscm
(ug/dscm)2
(ug/dscm)2
0.139
-------
All glass sample exposed surface to here.
Thermocouple
Probe
Reverse-type
TL_ Stack
Thermometer
I Glass
Glass Probe (j) / filter
Thermocouple
Check
Impingers with absorbing solutions
Vacuum
Modified G/S
Modified G/S: 100 ml H2O
Modified G/S: 100 ml H2O
G/S: 100ml 5% HNO3/ 10% H2O2
Modified G/S: Empty
Modified G/S: 100 ml 4% KMnO4/10% H2SO4
Modified G/S: 100 ml 4% KMnO4/10% H2SO4
Modified G/S: Silica Gel
Dry gas
(Q
ง
-o
O
-o
o
C/l
to
O.
O
55
o
D.
CD
CO
D)
CD
0)
-------
Probe liner, nozzle Fitter support and 1st Implnger 2nd
Front half of Titter and flexible back half of filter (empty at impinger
(H,0)
3rd 4th
Impinger Implnger
(H20)
Measure
impinger
contents
Measure
Impinger
contents
Measure
Implnger
contents
Empty contents
Empty contents
Empty contents
Rinse
three limes with
Container
Rinse
three limes with
i
>
Container
Rinse
three times with
i
r
Container
Rinse
three times with
i
i
Container
5th Implnger (empty) &
6th and7lh Implngers
(Addtfied KMnO.)
Last Implnger
Figure 2. Proposed Method 101B sampling train field recovery scheme.
-------
NOTES
-------
NOTES
-------
NOTES
-------
NOTES
-------
NOTES
------- |