&ER&
United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park NC 27711
EPA-450/3-86-015
December 1986
Air
Statistical
Analysis of
Emission Test
Data from
Fluidized Bed
Combustion Boiler
at Prince Edward
Island, Canada
-------
EPA-450/3-86-015
Statistical Analysis of Emission
Test Data from Fluidized Bed Combustion
Boiler at Prince Edward Island, Canada
Emission Standards and Engineering Division
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
December 1986
-------
This report has been reviewed by the Emission Standards and Engineering Division of the Office of Air
Quality Planning and Standards, EPA, and approved for publication. Mention of trade names or commercial
products is not intended to constitute endorsement or recommendation for use.
-------
TABLE OF CONTENTS
Page
LIST OF TABLES ...... v
LIST OF FIGURES vi
LIST OF ABBREVIATIONS vi 1
1.0 SUMMARY . 1-1
1.1 INTRODUCTION 1-1
1.2 EMISSION TEST RESULTS 1-3
2.0 PROCESS DESCRIPTION AND OPERATION 2-1
2.1 GENERAL INFORMATION 2-1
2.2 DESIGN INFORMATION 2-2
2.3 OPERATING INFORMATION 2-4
2.4 OPERATING PROCEDURES 2-9
2.5 PROCESS OPERATION 2-10
3.0 EMISSION TEST RESULTS 3-1
3.1 PROCESS AND EMISSION RESULTS 3-1
3.1.1 Methodology Used in Data Gathering and Data
Reduction 3-1
3.1.1.1 Data Gathering Techniques 3-1
3.1.1.2 Data Reduction Techniques 3-2
3.1.2 Summary Statistics for Process and Emission Data 3-5
3.2 COMBUSTION EFFICIENCY RESULTS 3-11
3.2.1 Methodology and Assumptions Used in the Analysis 3-11
3.2.2 Combustion Efficiency Results 3-14
3.3 ESTIMATION OF RECYCLE RATIO 3-15
3.3.1 Methodology and Assumptions Used in the Analysis 3-15
3.3.2 Recycle Ratio Results 3-16
iii
-------
TABLE OF CONTENTS (CONTINUED)
Page
4.0 EMISSION DATA VARIABILITY ANALYSIS 4-1
4.1 METHODOLOGY 4-1
4.2 RESULTS OF VARIABILITY ANALYSIS 4-5
5.0 REGRESSION ANALYSIS 5-1
5.1 METHODOLOGY AND MODELS USED IN THE REGRESSION ANALYSIS 5-1
5.2 REGRESSION ANALYSIS RESULTS 5-3
6.0 REFERENCES 6-1
APPENDIX A - SAMPLE PROCESS DATA SHEET AND LIST OF NOMENCLATURE A-l
APPENDIX B - DESCRIPTION AND RESOLUTION OF COAL FEED TOTALIZER
MALFUNCTION FROM MARCH 21 TO MARCH 25 B-l
APPENDIX C - COMBUSTION EFFICIENCY CALCULATIONS C-l
APPENDIX D - ESTIMATION OF RECYCLE RATIO D-l
APPENDIX E - PROCEDURES USED IN VARIABILITY ANALYSIS E-l
iv
-------
LIST OF TABLES
Table Page
2-1 BAGHOUSE DESIGN DATA... 2-6
2-2 OPERATING ANOMALIES DURING FBC EMISSION TEST AT PRINCE
EDWARD ISLAND 2-15
3-1 DAILY AVERAGE PROCESS FLOW RATES 3-6
3-2 DAILY AVERAGE PROCESS PARAMETERS AND EMISSION RESULTS 3-7
3-3 AVERAGES (STANDARD DEVIATIONS) OF OPERATING PARAMETERS AND
EMISSIONS DETERMINED FOR EACH PARAMETRIC TEST 3-8
5-1 RESULTS OF REGRESSION MODELS 5-4
5-2 S02 REDUCTION EFFICIENCY STATISTICS ". 5-6
A-l SAMPLE PROCESS DATA SHEET A-2
A-2 PROCESS DATA NOMENCLATURE GENERAL BOILER PARAMETERS A-4
A-3 PROCESS DATA NOMENCLATURE BED "A" PARAMETERS A-6
A-4 PROCESS DATA NOMENCLATURE BED "B" PARAMETERS A-7
B-l SAMPLE DAILY LOG OF TOTALIZER READINGS AND OTHER PROCESS
READINGS B-2
-------
LIST OF FIGURES
Figure Page
2-1 Foster Wheeler boiler at Prince Edward Island 2-3
2-2 Flow diagrams for coal, wood and limestone 2-5
2-3 FBC layout 2-8
2-4 Emission source test summary 2-17
3-1 Measurement points for estimating combustion efficiency 3-12
5-1 Plot of S02 emission reduction versus Ca/S ratio 5-8
5-2 Plot of SO- emission reduction versus load 5-9
5-3 Plot of S02 emission reduction versus bed temperature 5-10
B-l Boiler efficiency comparison (1986) B-3
B-2 Calcium utilization comparison B-5
vi
-------
LIST OF ABBREVIATIONS
Kg/hr =» kilogram per hour
Ib/hr - pound per hour
tonnes/day = metric tons per day
10 Btu/hr = million British thermal units per hour
lb/10 Btu » pounds per million British thermal units
psi = pounds per square inch
kPa = kilopascal
mm. H«0 - millimeter of water (pressure unit)
Btu/lb = British thermal units per pound
gr/SCF « grains per standard cubic foot
ACFM = actual cubic feet per minute
ppm - parts per million
Vol. % - volumetric percent
% * percent
°C = degree Celsius
RSO - relative standard deviation
2
R - coefficient of determination
vi i
-------
1.0 SUMMARY
1.1 INTRODUCTION
Emission testing was conducted from February 28 to March 28, 1986, on a
fluidized bed combustion (FBC) boiler (boiler No. 2) at the Canadian Forces
Base in Summerside, Prince Edward Island, Canada. The emission data
collected on boiler No. 2 have been presented in the emission test report
prepared by the test contractor. Both the sulfur dioxide (S02) and
nitrogen oxide (NO ) emission test data will be considered by the U.S.
A
Environmental Protection Agency (EPA) in developing new source performance
standards (NSPS) for steam generating units (i.e., boilers) rated at 100
million Btu/hour heat input or less. The purpose of this report is to
examine the statistical characteristics of S02 emissions from boiler No. 2.
The emphasis during the testing was to determine the S02 emission reduction
performance of this boiler. Nitrogen oxide emission data were also
characterized but were not subjected to the same rigorous analysis as the
SO- emission data.
In order to examine the SO- emission reduction performance of this
boiler, three series of tests were performed during the emission testing.
First, a 27-day test series was performed to achieve an average SO- emission
reduction of 93 percent or greater. It was estimated that this average
reduction level would be necessary for the FBC boiler to meet a 90 percent
reduction requirement when using a 30-day rolling average. The test plan
specified that the operating loads for this test series be at least
70 percent of rated capacity. The FBC boiler operated for approximately 23
days at loads ranging from 70 to 75 percent. Continuous SO- emission data
for the first 7.5 days of continuous operation were statistically
characterized with respect to central tendency (mean), variability (standard
deviation), and the independence of successive measurements
(autocorrelation). Hourly emission data were subsequently fitted to a time
series model that allows predictions of emission reduction performance over
periods of 30 days or longer.
1-1
-------
The other two test series performed on boiler No. 2 were parametric
tests lasting less than one day. The first of these, lasting about 8 hours,
was performed to determine the effects of varying limestone particle size on
S02 reduction. Limestone with particles sized at 0.25 inch and less
(1/4 x 0) was injected into the boiler instead of the typical limestone with
particles sized between 0.03 and 0.09 inches (8 x 20 mesh). For the second
test series, lasting about 20 hours, the boiler was operated to achieve the
maximum S02 emission reduction possible while maintaining stable operation
of boiler and auxiliary equipment. The purpose of this series of tests was
to investigate whether mass transfer effects might preclude achieving 90
percent SO- removal, or greater, for this boiler on low sulfur coal.
The major objectives of this report then, are as follows: (1) to
statistically evaluate the SO- emission control performance of this FBC
boiler over the 7.5-day period and to predict performance using a 30-day
rolling average, (2) to summarize the emission results from the two
parametric tests, and (3) to relate SO* emission reductions to important
operational parameters such as calcium-to-sulfur (Ca/S) molar ratio, bed
temperature, load, and limestone particle size. This report is a companion
to the emission test report prepared by the emission test contractor.
Section 1.2 summarizes the emission test results for the FBC boiler
(boiler No. 2) at Prince Edward Island. Section 2.0 discusses the boiler's
process design and operational characteristics of this boiler. This section
also summarizes the operating anomalies which occurred during the emission
test. Section 3.0 summarizes the results of the emission test including
estimates of the combustion efficiency and solids recycle ratio observed
during testing. Section 4.0 discusses S02 emission reduction performance
based .on continuous emission monitoring over the 7.5-day period. This
section also discusses the statistical model and methodology for predicting
30-day rolling average performance results. The results of regression
analyses relating S02 removal efficiency to key bperating parameters are
presented in Section 5.0.
1-2
-------
1.2 EMISSION TEST RESULTS
For the first 7.5-day period of continuous operation, the FBC boiler
No. 2 achieved 94.2 percent S02 reduction operating at an average Ca/S molar
ratio of 3.7; the boiler operated at 72.4 percent of full load and fired a 6
weight percent sulfur coal, on average. The mean SCL reduction performance
of the boiler during the entire 30-day test period, including process upsets
and power outages, was 93.5 percent. Variability analysis of data collected
during the 7.5-day period showed that this boiler could meet a 90 percent
S02 reduction requirement on a 30-day rolling average basis by operating at
a mean level of 91.3 percent S02 reduction, assuming variability
characteristics remained unchanged. If the mean SO- reduction performance
of 93.5 percent cited above were maintained, the FBC boiler No. 2 would be
in compliance with a 90 percent S(L reduction standard using a 30-day
rolling average method.
The results from parametric testing using coarser limestone particles
than those used during the 7.5-day period indicated that mean S0« reduction
performance dropped to 91 percent. The boiler was operating at an average
Ca/S ratio of 4.5 at near the same load and fuel sulfur content as that
during the 7.5-day test. The results from the parametric test aimed at
achieving very high SO- reduction levels showed that this boiler achieved a
99.4 percent mean SO* reduction when operating at an average Ca/S ratio of
7.2. The boiler was operated at 56 percent of full load and fired a 5.7
percent sulfur coal during this period. These results indicate that mass
transfer effects in the FBC unit do not limit the outlet S02 concentration
levels that can be achieved.
Regression analysis was performed using the entire emission dataset to
predict S02 reduction efficiency as a function of key operating parameters
such as Ca/S ratio, load, and bed temperature. The best-fit model (with a
coefficient of determination, R , of 0.62) that ts consistent with observed
FBC boiler behavior is the following:
EFF - 100 [ 1 - exp (0.036 Ca/S - 0.88) Ca/S]
1-3
-------
where: EFF = S02 reduction, in percent; and
Ca/S = Calcium-to-sulfur molar ratio
2
The low R value of this model is attributed partly to the lack of
variation among the SCL reduction data. In addition, only two parametric
tests, each lasting less than one day, were performed during the entire
source testing. A better fit of the data would have been realized if more
parametric tests had been performed or if these tests had been of longer
duration.
1-4
-------
2.0 PROCESS DESCRIPTION AND OPERATION
This section presents a process description of the fluidized bed
combustion (FBC) test unit.- Included are characterization of the boiler
and auxiliary equipment design and operation and discussion of the boiler
operating history, control procedures, and process test conditions.
2.1 GENERAL INFORMATION
The two FBC boilers located at the Summerside, Prince Edward Island,
site were built to provide 40,000 pounds/hour (Ib/hr) of steam each for
space heating at the Canadian Forces Base and are operated as full scale
demonstration units by the Canadian government. The two identical FBC
boilers (Units No. 1 and No. 2) are field-erected, bubbling bed units
manufactured by Foster Wheeler (FW) Canada Ltd. The emission test data were
gathered on Unit No. 2.
Two dump and grate stoker boilers, each more than 25 years old, are
also located at the Summerside site. These two stoker units have a combined
steam production capacity of 20,000 Ib/hr. The two FBC units were purchased
to meet the additional steam needed for Base heating and are operated as
peak load units. Fluidized bed combustion was chosen because the Canadian
government wanted to demonstrate on a commercial scale that FBC is a viable,
clean, "off oil" technology.
During early operation of the FBC units, attention was drawn to the
great amount of make-up water required by the boilers. This concern led to
a steam leak repair program in the steam tunnels of the Base, resulting in a
significant reduction in the steam demand. The reduction was so large that
typically only one FBC unit is now needed on-line to accommodate the total
Base steam demand.
2-1
-------
2.2 DESIGN INFORMATION
A side view of one of the FBC boiler units located at the site is shown
in Figure 2-1. Each of the-.atmospheric fluidized bed (AFB) boilers is
designed to produce 40,000 Ib/hr of saturated steam at 110 to 140 pounds per
square inch (psi). The heat input capacity of each unit is rated at 50
million Btu/hr. The fluidized beds are operated at temperatures near 1560°F
(850°C). The units were guaranteed by FW Canada Ltd. to achieve greater
than or equal to 80 percent boiler efficiency. In actual testing, the
boiler efficiency has been measured at 83 percent. Each FBC boiler consists
of two combustion beds, A and B, each fed by its own overbed spreader
stoker. In Figure 2-1, combustion bed A is indicated as the preferential
bed, while combustion bed B is indicated as the secondary bed. The two beds
are divided by a waterwall which has a one square-foot opening located at
the bottom center of the wall. This opening allows bed materials to
circulate between the two beds when both are operating. This two-bed system
allows for greater flexibility in turndown. A turndown ratio of 8:1 has
been achieved during regular operation by taking bed B off-line.
As shown in Figure 2-1, tubes are located both in and above the bed.
2
The total heating surface is 9,930 square feet (ft ). In-bed tubes are
mounted vertically along the walls and then diagonally across the unit.
Above-bed tubes are mounted vertically along the wall. When fluidized, the
bed depth is approximately 4.5 feet (ft) and the freeboard extends 22 ft
above the bed. The slumped (non-fluidized) bed height is 2 ft. Bed A is 4
ft wide and 9.5 ft long, while bed B is 4.5 ft wide and 9.5 ft long.
Both FBC units are equipped with fly ash reinjection systems. All of
the fly ash collected in the convective tube bank and by the mechanical dust
collector is recycled back to bed A. The mechanical dust collector is a
multi-tube cyclone housing five rows of small-diameter cyclone tubes which
operate in parallel. The multi-tube cyclone was'manufactured by Enviro
Systems and Research and was designed to capture particles greater than 40
microns in diameter.
2-2
-------
Steam Drum
Observation Door
(Front)
Flnea Reinfection
Ptpea
Feeder Pipe
(LH.S.)
Damper '' Bed DralrTpIpe F««0*P|P«» Bed Drain Pipe
(LH.S.) (L.H.S.)
Figure 2-1. Foster Wheeler boiler at Prince Edward Island.
Figure 2-1 from CANMET Division Report ERP/ERL 82-10 (TR). Used by
permission of CANMET (Canada Centre for Mineral and Energy Technology)
on behalf of the copyright owner.
2-3
-------
Each FBC unit is also equipped with a baghouse unit designed to capture
nearly all the particulate matter remaining in the flue gas exiting the
cyclone. The baghouse was also manufactured by Enviro Systems and Research.
The design data for the baghpuse are shown in Table 2-1.
2.3 OPERATING INFORMATION
Annual operation occurs from early September to late May and the
typical load over this heating period is 75 to 80 percent of maximum design
load. During operation of the boiler, coal and limestone are trucked to the
facility and dumped into separate underground hoppers as shown in
Figure 2-2. The coal and limestone are sifted by screens and the larger
sizes are re-crushed before being carried by bucket elevators to on-site
storage bunkers. Bucket elevators transport the coal and limestone from the
storage bunkers to the respective coal and limestone day bins located above
the boiler unit. From the day bins, the coal and limestone flow by gravity
to weigh feeders. The coal feed system is equipped with one weigh feeder
which delivers the coal to a pant leg where the coal flows by gravity and is
distributed to separate stoker feeders, one feeder for each bed. The
limestone feeder system is equipped with a separate weigh feeder for each
bed. From the weigh feeders, the limestone flows by gravity through
separate feed pipes to beds A and B.
The coal normally fired in the FBC units, as well as during the
emission testing period, is an unwashed eastern bituminous coal which has
been sized to 1 inch x 0. The coal has an average heating value of 11,500
Btu/lb and an average sulfur content of 5.5 percent. This corresponds to
potential SOg emissions of 9.57 Ib/million Btu on an uncontrolled basis.
Typical ash, nitrogen, and moisture contents are 19.3, 1.0, and 0 to 10
percent, respectively. The average values for fixed carbon and volatile
matter are 47.2 and 33.5 percent, respectively. '
The limestone normally used at the site is Havelock limestone with a
calcium carbonate (CaC03) content of 94 to 95 percent. Currently, it is
sized to 0.0934 inch x 0.0331 inch (8 x 20 mesh) with certain percentages at
2-4
-------
v
CM
V
I Bur
v
t
kara
t
V-
NO. 1 Bollar >
z^- f=;
|
I
W1
Coal I |
B,n. l_J
\ to
righFaadara
t
^M
U
Coa
V
Bun
V
t
Kara
V
I C3~ < '
J No. 2 Boiltr
Row Diagram * Coal and Wood
Bunk*
pi.
Y
I I
J l""
V V
^-A\ /fr*n
^z;
Y
K
1
Bunk*
Flow Diagram Umaaton*
NATERIM.S HANDLING - RATES AND CAPACITIES FOR EACH F8C UNIT
Trtnsfcr to rtctlvlng hopptrs
to bunktrs
Bunktrs capacity
Tnnjfar to bins
Bins cipadty
Burn ratt
COAL
25 tons/hour
4 t 82.5 tons
5 tons/hour
2 x 5 tons
2.5 tons/hour
LIMESTONE
25 tons/hour
2 * 75 tons
2.5 tons/hour
2 i 2.5 tons
1.13 tons/hour
Figure 2-2. Flow diagrams for coal, wood and limestone.
Figure 2-2 from CANMET Division Report ERP/ERL 82-10 (TR). Used
by permission of CANMET (Canada Centre for Mineral and Energy
Technology) on behalf of the copyright owner.
2-5
-------
TABLE 2-1. BAGHOUSE DESIGN DATA
Flue Gas
Inlet gas volume
Inlet gas moisture
17,278 ACFM 0 350°F
8 wt. percent
Dust
Loading
Particle size
2.4 gr/SCF
40 percent less than 10
microns
Baas
Number of cells
Number of bags
Total cloth area
Air-to-cloth ratio
12/unit
36/cell ,
4,898 ft*
3.52 (with all cells on-line)
Operating Information
Collector design gas-side
pressure drop
Collector casing design
pressure
Particulate matter emissions
(Baghouse outlet)
Collection efficiency
+ 3 inches water (gauge)
+ 20 inches water (gauge)
0.02 gr/SCF
99.2 percent
Collection efficiency calculated on the basis of the inlet loading and
outlet emissions.
2-6
-------
10, 12, and 15 mesh. The FBC unit can also use limestone sized to 0.25 Inch
x 0, which contains a higher percentage of fines than does the 8 x 20 mesh
limestone. Limestone is used as the sorbent for purposes of sulfur dioxide
($02) removal. As will be discussed in Section 2.4, limestone is also added
to the bed as one method to control bed level. During normal operation, the
11mestone-to-coal feed ratio is maintained near 1 to 5, which corresponds to
a calcium-to-sulfur (Ca/S) ratio of approximately 1.1:1 based on an average
coal sulfur content of 5.5 percent. Sulfur dioxide emission reductions
usually range from about 40 to 50 percent at this operating condition. This
unit was designed to achieve 83 percent S02 emission reduction at Ca/S
ratios ranging from 2.4 to 4.1, depending on the fuel burned.
The material handling system shown 1n Figure 2-2 can also handle wood
chips. The FBC units can combust wood but only 50 percent load can be
achieved due to size limitations of the wood feed system. (The original
design called for 340 percent of the heat input to be from wood chips.)
Screw conveyors are operated under the beds to remove ash and sulfated
limestone (i.e., bed material). The desired bed level 1s maintained by
adjusting either the limestone feed rate or the screw conveyor discharge
rate. Each bed 1s equipped with a separate drain pipe which allows bed
material to drain by gravity when the screw conveyors are operated.
The FBC layout Is shown 1n Figure 2-3. Combustion air passes through a
forced draft (FD) fan and enters the plenum. The air flow 1s distributed
across the two beds at fluidlzing velocity by means of distributor nozzles
which are uniformly spaced across the base of each bed. Fluldlzed bed
material freely circulates through the door in the water wall which
separates beds A and B. This movement of material, as the two beds seek a
common level, promotes fuel and sorbent mixing and Improves the operational
stability of the boiler.
Hot flue gas leaves the bed at approximately 1560°F (850°C), passes
through the freedboard section, and flows across the convectlve tube bank
before exiting the boiler. All of the fly ash collected 1n the convectlve
tube bank 1s pneumatically conveyed into bed A.
2-7
-------
-I» NORTH
I
/
1. Control Room
2. F.D.Fan
3. Boiler *
4. Mechanical Dust Collector
5. Economizer
8. Baghouse
7. I. D. Fan
8. Stack
9. Coal Receiving Hopper
10. Coal Bunker
11. Coal Bin
12. Limestone Receiving Hopper
13. Limestone Bunker
14. Limestone Bin
15. Wood Chip Bin /
18. De-Aerator
17. Diesel Generator
18. Light Oil Tank
19. Office
20. Ash Silo
Figure 2-3. FBC layout.
Figure 2-3 from CANMET Division Report ERP/ERL 82-10 (TR). Used by
permission of CANMET (Canada Centre for Mineral and Energy Technology)
on behalf of the copyright owner.
2-8
-------
The flue gas exiting the boiler enters the mechanical dust collector
where fly ash is captured and gravity fed through a straight leg reinjection
system back into bed A. Spargers inject air into the reinjection leg at the
reinjection point to assist'fly ash recycle and prevent plugging of the
reinjection leg.
The flue gas then passes through the finned-tube economizer where it is
cooled before entering the baghouse. In the economizer, the feed water
temperature is raised from 220°F (105°C) to 275°F (135°C). In the baghouse,
nearly all remaining particulate matter is removed. From the baghouse, the
flue gas is drawn through an induced draft (ID) fan and exhausted to the
stack.
The ash which is collected from under the bed and in the baghouse is
transported in a vacuum-type pneumatic handling system to the ash silo.
From the ash silo, the waste is trucked to disposal trenches located on the
Base.
2.4 OPERATING PROCEDURES
During normal operation, the boiler plant master controller operates
automatically and regulates the steam output from the operating boilers.
The controller modulates the total steam flow from the boiler plant to
maintain constant pressure in the common steam header to the Base. During
peak heating periods, when more than one boiler is operating, the control
strategy is to operate one unit at constant load conditions while the peak
load unit is allowed to track the excess steam demand. This control
strategy is accomplished by setting the boiler master controller for the
constant load unit to operate manually and operating the boiler master
controller for the peak load unit automatically.
As mentioned earlier, the FBC boilers are normally operated as peak
load units. Operating the FBC master controller "automatically adjusts the
coal feed rate to the unit to compensate for swings in the plant steam
demand. The combustion air controller, sensing the coal rate change,
adjusts the total air flow to the unit to maintain a constant air-to-coal
2-9
-------
ratio. The total air flow is further adjusted by an oxygen trim controller,
as necessary, to maintain the bed at optimum sulfation temperature.
The FBC unit design control scheme calls for the limestone feed
controllers to automatically track the stack gas SCL concentration and
adjust the sorbent feed rates to beds A and B based on an SCL emission
set-point. However, this loop was never sufficiently tuned to allow fully
automatic control and, therefore, has been by-passed as part of the
overall FBC unit control scheme. During normal operation, the limestone
feed controllers are operated manually; the limestone feed rates to beds A
and B are adjusted by operating personnel to maintain desired bed levels.
The boiler feed water flow responds to the steam output requirements.
The boiler feed Water flow rate is also subject to control by the steam drum
level, which is a slower control loop. The steam drum level is controlled
at the mid-level height of the drum.
2.5 PROCESS OPERATION
The primary operating objective of the emission source test was to
demonstrate long-term FBC system performance of approximately 93 percent S02
reduction. Secondary operating objectives of the test were (1) to obtain
system performance data demonstrating the effect of limestone particle size
on sulfation capacity, and (2) to determine the maximum SCL reduction
achievable during stable two-bed unit operation.
During the first three weeks of the testing period, the FBC unit was
operated at near 70 percent load and the limestone feed rate was manually
adjusted to maintain the S02 removal efficiency at the targeted 93 percent
level. As discussed in Section 2.4, the control system between the S02
outlet flue gas concentration and the limestone feeders was never
sufficiently tuned to allow fully automatic control of S02 emissions using
the limestone feed rate. As a result, the limestone feeders to Beds A and B
were controlled manually by plant operators.
2-10
-------
A target S02 emissions level was established consistent with the feed
coal sulfur and heating values and desired S0« reduction levels. While It
would have been physically possible for the plant operators to adjust the
limestone feed rates on a near-continuous basis to match the target S02
emissions level and load changes, this would have required an operator to be
dedicated to limestone feed adjustments. This was Impractical and was
inconsistent with normal plant operation in which limestone feed rates were
set manually and only adjusted infrequently.
Another approach considered and eventually chosen was to hold the unit
load at a steady level so that a relatively constant Ca/S feed ratio could
be maintained. This was the control approach used for all but the last week
of the emission source test. It was desired that the FBC unit operate at as
high a load as was practicable in order to fully test the capabilities of
the limestone feed system, solid waste withdrawal system, and sulfur capture
reactions in the bed. Unfortunately, weather conditions during the
emission test period were unseasonably mild such that Base steam demand was
below normal. To operate the FBC system at loads in the range of 80 to 90
percent would have required the venting of steam during much of the test
period, at considerable cost to the plant. A unit load at or above 70
percent was selected as a reasonable compromise between the desire to
demonstrate FBC system performance under the adverse conditions of
relatively high load and the economic penalties of venting large quantities
of steam. During short periods when the steam demand exceeded that
available from the FBC unit operating at or above 70 percent load, the dump
and grate stokers were quickly brought on-line to meet the extra demand.
The actual steam output of the FBC boiler was not known because the
steam flow rate recorder was out of calibration and was considered
unreliable for measuring actual flows. However, the recorder was considered
reliable for indicating flow trends. The unit load target of 70 percent was
maintained on the basis of the coal feed rate, the heating value of the
coal, and the known maximum heating input capacity of the FBC boiler.
For these reasons, the control strategy used during emissions testing
was to set the coal feed rate to the test unit at a level approximating a
2-11
-------
70 percent load condition and to set the limestone feeders to achieve a
target SO* emissions level. During the emission testing, the bed level was
maintained by adjusting the screw conveyor bed drain discharge rate. The
dump and grate stoker Toad or the steam vent rate was then adjusted as
necessary to allow the FBC unit to operate at the steady load condition
during the test period. The boiler combustion air and feed water controls
were operated automatically during the test period to ensure smooth boiler
operation at the constant load condition.
Variations from the desired load condition during testing were due to
fluctuating coal characteristics such as moisture content and bulk density.
At constant controller settings, the bulk density affected the coal delivery
rate to the unit while increased coal moisture content lowered the effective
heat input to the unit. Operating personnel monitored flue gas SO*
concentration trends during the test period and adjusted the limestone feed
rates stepwise as necessary to maintain desired S02 reduction performance.
During the last week of the source testing period, two short-term
parametric tests were conducted. During the first parametric test, system
performance data were gathered over an eight-hour period while 0.25 inch x 0
limestone was fed to the unit. The 0.25 inch x 0 limestone evaluated during
this period was obtained from the same mine as the 0.0934 inch x 0.0331 inch
(8 x 20 mesh) limestone used during the remainder of the source testing.
The data gathered during this period were used to compare the SO* reduction
efficiency of the two sizes of limestone at equivalent calcium to sulfur
(Ca/S) ratios.
During the second parametric test, the unit was operated at high Ca/S
ratios (by increasing the limestone feed rate) to determine if mass transfer
restrictions would prevent the FBC unit from operating at SO* flue gas
concentrations in the range of 20 to 40 ppm(v). For a unit firing low
sulfur coal (<2 percent sulfur), the flue gas SO* concentration must be
maintained in this range to achieve approximately 93 percent SO* reduction.
The boiler was operated at near 50 percent load during this 12-hour test to
avoid overloading the baghouse at the high Ca/S ratios. The dump and grate
stokers were brought on-line as necessary to handle any additional steam
demand.
2-12
-------
Due to Inclement weather conditions, electrical power to the Base was
interrupted four times during the source testing period. When a power
outage occurred, a diesel-fired generating system at the plant was started
and brought on-line to provide electricity for the FBC boiler operation.
The switching process from main power to back-up power required a few
minutes to execute; as a result, all feed flows to the unit were interrupted
during this period. As the back-up power was brought on-line, controller
settings were restored to their pre-outage positions. However, the
limestone feed was discontinued until the coal and combustion air rates had
been adjusted to bring the bed temperatures back within desired operating
ranges.
The system operating performance and SO* reduction performance were
unsteady while the unit was operated on back-up power due to the process
upsets (e.g., interruption of feed flows). Also, the electricity generated
by the diesel system was at a slightly different frequency from that of the
main power system. This affected the performance and reliability of
sensitive electronic equipment such as the feed totalizers.
Once main power was restored to the facility, the main switching
process was repeated as the back-up power supply was taken off-line and main
power to the boiler was restored. The fuel and combustion air feed flows
were again interrupted for a short period, and the limestone feed was
discontinued until bed temperatures stabilized within operating ranges.
After the return to main power, a period of 8 to 15 hours was normally
required for coal and limestone feed rates to stabilize and for the FBC
system to return to near steady operation.
Due to the frequency of power outages at this facility during winter
storm periods, a more advanced back-up system will be installed in the
future to provide electricity instantaneously to the unit during power
outage periods. The new back-up system will allow the unit to operate
smoothly during the transition from main to back-up power supply.
For these reasons, the data gathered during periods of back-up power
generation were not considered to be representative of stable FBC unit
operation. However, data from these periods were used in the regression
2-13
-------
analyses as an Indication of FBC unit performance under operating conditions
which were significantly different from those encountered during main power
operation. Once the unit was restored to main power and the system
performance steadied to set'conditions, testing was recommenced. However,
the data gathered during process upset periods .caused by unavoidable
equipment malfunctions and operating anomalies other than power outages were
considered representative test data. Data gathered during these periods
were evaluated in order to obtain an estimate of achievable long-term S02
reduction performance under conditions which were considered typical for
this FBC unit at the time of the emission test. A summary of the operating
anomalies encountered during the emissions source test is included in
Table 2-2.
A summary of the FBC unit status for the source testing period is shown
in Figure 2-4. Power outages occurred on the following dates: March 6,
March 9, March 15, and March 18, 1986. On March 7, a severe winter storm
forced the Base to be closed to all non-essential personnel. The ash
hauling trucks, therefore, were not available to transfer waste ash from the
ash silos to disposal trenches located on the Base. To avoid overloading
the ash silo during this period, the boiler was operated at low limestone
feed rates. The unit was returned to test conditions on the morning of
March 8. The March 18 power outage resulted in the formation of clinkers in
bed B of the test unit. The clinkers prevented fluidization of bed B,
making this bed inoperable. The unit was taken off-line and allowed to cool
so that operating personnel could remove the clinkers. The test unit was
restarted on March 21 and was operating at test conditions by approximately
12:30 p.m. During the week after restartup, the parametric tests associated
with the secondary operating objectives were conducted.
2-14
-------
TABLE 2-2. OPERATING ANOMALIES DURING FBC EMISSION TEST AT PRINCE EDWARD ISLAND
From:
To:
Date0 Time Date*
Time
Duration
(Days)
Operating anomaly
ro
2/28 2300 2/29 . 0800 0.38
3/5 0900 3/6 0830 0.98
3/9 1800 3/10 0900 0.63
3/16 0800 - 3/16 1045 0.11
3/16 2200 3/16 2315 0.05
Operators bypassed flue gas around the baghouse. The
baghouse was operating at high differential pressures
(6 to 6.5 inches of H20). Operators were manually
cleaning the ash from the baghouse.
General inspection. Coal rate to the boiler was
lowered such that no steam was vented; load was reduced
to about 50 percent. During the early morning of
March 6, load was at 66 percent.
Limestone feed rate did not respond to increased
limestone control settings. Consequently, S02
emissions were high (300 to 400 ppm).
One of the pins holding a coal bucket came loose in
primary bucket elevator. While switching coal flow to
the other bucket elevator, the coal supply was
exhausted. Began feeding coal from emergency
stockpile. Coal weigh feeder jammed and bed
temperature varied more than normal after switching to
emergency coal due to high moisture and fines content.
Very wet coal from emergency stockpile extinguished
furnace flames. Limestone feed was discontinued until
furnace fire was restarted and the bed temperature
returned to 1560°F (850°C).
-------
TABLE 2-2. OPERATING ANOMALIES DURING FBC EMISSION TEST AT PRINCE EDWARD ISLAND (CONCLUDED)
From:
To:
Date0
Time Date Time
Duration
(Days)
Operating anomaly
3/16 2315 3/18 0945 1.44
3/21 1330 3/22 0900 0.81
IS}
3/25 1345 3/25 1400 0.01
Operated with emergency coal until new coal shipment
arrived on Base. Operating problems occurred with coal
handling and feeding systems.
Discontinued limestone feed due to artifically'high
baghouse differential pressures (between 7 and.7.5
inches of H»0). The high differential pressure was
caused by plugging in the pneumatic line of the
controller.
Replaced faulty electrical board in coal weigh feeder
totalizer. Coal feed rate readings from March 21 &
0945 to March 25 @ 1400 were artifically low.
'1986.
-------
UJ
2
0000
0200
0400
0600
0800
1000
1200
1400
1600
1800
2000
2200 H
2400
DATE (1986)
2/27 2/28 3/1 3/23/3 3/4 3/5 3/B 3/7 3/83/9 3/10 3/11 3/12 3/13 3/14 3/15 3/16 3/17 3/18 3/19 3/20 3/21 3/22 3/23 3/24 3/25 3/26 3/27 3/28
Unit Operating at Test Conditions B Unit Operating but not at
H Test Conditions
Unit not Operating down for Repairs
Figure 2-4. Emission source test summary.
-------
3.0 EMISSION TEST RESULTS
This chapter summarizes the results of the emission test conducted on
boiler No. 2 at Prince Edward Island, Canada, from February 28 to March 28,
1986. Section 3.1 presents the process and emission results and discusses
the methodology used to gather and reduce the process and emission data.
Combustion efficiency was also estimated and the results are discussed in
Section 3.2. Section 3.3 discusses the results of estimating the recycle
ratio during the test.
3.1 PROCESS AND EMISSION RESULTS
3.1.1 Methodology Used in Data Gathering and Data Reduction
3.1.1.1 Data Gathering Techniques. Two main datasets were created
from data collected during the source testing period. The first dataset
contained instantaneous process and emission data recorded automatically by
a tape recorder every 5 minutes and by an on-site computer/printer every
15 minutes. The readings from the tape recorder were used as back-up in
case the computer/printer malfunctioned during the test. The recorded
process parameters included: the total coal feed rate, limestone feed rates
to beds "A" and "B", total air flow rate, bed temperatures, average
freeboard temperature, boiler exit gas temperature, stack gas temperature,
steam flow rate and pressure, economizer inlet oxygen (02) content on a wet
basis, and pressure drop across the beds. Other process-related parameters
recorded were: fan suction and discharge pressures, flue gas operating
pressures, boiler feed water temperature and pressure, and steam drum level.
The recorded emission data included the stack gas concentrations of 0- in
percent, carbon dioxide (C02) in percent, carbon.monoxide (CO) in parts per
million (ppm), sulfur dioxide (S02) in ppm, and nitrogen oxide (NO ) in ppm,
all on a dry basis. Sample computer printouts of these parameters are
included in Appendix A.
3-1
-------
The second dataset contained coal and limestone flow rates recorded by
electronic totalizers and results from coal and limestone composition
analyses. Coal and limestone totalizer readings were recorded on an hourly
basis during the day shift operation in order to compute average hourly feed
rates during these periods. Coal and limestone totalizer readings were also
recorded at approximately midnight each day by operating personnel in order
to compute the total flow for the 24-hour period. The coal analyses were
performed on 8-hour composite samples. Limestone analyses [i.e., calcium
carbonate (CaC03) and sulfur contents] were performed on 24-hour composite
samples from samples collected every hour for each operating day.
3.1.1.2 Data Reduction Techniques. The first dataset, containing
15-minute readings on computer printouts, were keypunched onto a tape. The
data were later transferred from the tape to a mainframe computer for
subsequent data analyses. In order to reduce the amount of keypunching
required, only data for important process and emission parameters which
might be needed in the statistical analyses were keypunched. Data were
keypunched for the following operating parameters: steam flow rate;
economizer inlet 0- content in flue gas; combustion air flow rates to, and
bed temperatures in, both beds; and stack gas concentration analyses.
Because hourly SO* reduction values were used in the statistical
variability analysis, it was necessary to represent the hourly coal and
limestone analyses using the composite analyses discussed above. Hence, in
the second dataset, the composite coal analysis results were assigned to
each hourly coal analysis value in the corresponding 8-hour period. In the
same manner, each hourly limestone value was the same as the 24-hour
composite result. These hourly values for coal and limestone analyses were
keypunched and loaded into a file on the mainframe computer. When
calculating hourly S02 reduction values and Ca/S molar ratios, these hourly
coal and limestone analysis results were used in'combination with the hourly
S02 emission data.
Coal and limestone feed rates from the second dataset were considered
to be more representative and accurate than those from the first dataset.
3-2
-------
Coal feed flow rates from the first dataset varied greatly due to the "stop
and start" operation of the belts on the coal weigh feeder, which is typical
of normal plant operation. The belts on the weigh feeder stop moving at
certain times because the coal accumulates in the pant leg (located between
the spreader stoker and the weigh feeder) and its level reaches the high
level indicator thereby signaling the weigh feeder to stop sending coal to
the spreader stoker. During this time, the totalizer does not record any
flow reading because the totalizer records only the weight of coal being
passed onto the weigh feeder belts. Therefore, the instantaneous coal feed
rate reading in the on-site computer would indicate little or no flow during
periods in which the weigh feeder stops feeding coal to the spreader stoker.
In actuality, coal is still being fed by gravity from the coal
accumulated in the pant leg to the spreader stoker during times when the
weigh feeder belts are not moving. The weigh feeder belts will
automatically start moving coal to the boiler when the coal level in the
pant leg is below the high level indicator. Because only four instantaneous
coal rate readings are recorded every hour in the first dataset, the average
flow rate recorded in any hour could be lower than the actual average flow
rate depending upon whether the belt in the coal weigh feeder was moving
when the instantaneous readings were recorded.
In order to measure the coal feed rate more accurately, totalizer
readings were taken every hour so that an average coal flow rate could be
calculated. This method is preferred since it incorporates all of the stops
and starts of the weigh feeder belts during this hour period. Limestone
flow rates were also based on totalizer readings in order to be consistent
with the coal flow rate measurements. The plant personnel also considered
the limestone totalizer readings to be more accurate than what was recorded
onto the computer/printer. The limestone weigh feeders for both beds were
operated continuously during the test in contrast to the operation of the
coal weigh feeder. Neither limestone weigh feeder had a high level
indicator, since limestone never accumulated in the feed pipes.
Once both datasets were keypunched and loaded onto the mainframe
computer, the datasets were compiled into Statistical Analysis System^
3-3
-------
(SAS) datasets. SAS is a computer software system for data analysis
provided by the SAS Institute Incorporated. Each dataset was checked for
typing errors. Extraneous points, such as calibration points and faulty
readings due to equipment malfunctions, were deleted. However, during the
very high S02 reduction test period, S02 emissions measured by the monitors
were in some cases logged as negative values. For these cases, the SCL
concentrations were assumed to be zero. The first dataset containing
15-minute readings was reduced to one-hour averages. This dataset was
merged with the second dataset to form the complete dataset of hourly values
used in the statistical analyses.
Additional operating parameters were calculated on the basis of the
data in the merged dataset. The boiler heat input rate for every hour was
calculated using data on the coal flow rates and heating values. The
average bed temperature was determined by averaging individual bed
temperatures measured at various locations within both beds. The Ca/S molar
feed ratio was determined from coal and limestone feed rates, coal sulfur
content, and limestone CaCO, content. Outlet NOV and SO, emissions (in ppm)
£
were converted to a heat Input basis (pounds per million Btu or lb/10 Btu)
using the F-factor method. To calculate an F-factor, the fuel ultimate
analysis is required. Therefore, the F-factor for each hour during a test
day was based on the ultimate analysis for the first 8-hour composite coal
sample. The proximate analyses for the other two 8-hour composite samples
for each day did not provide adequate information to calculate an F-factor.
"Uncontrolled S02 emissions," in units of lb/10 Btu, were calculated for
each hour based on the heating value and sulfur contents of the coal and the
assumption that 100 percent of the coal sulfur is converted to S02. (The
term "uncontrolled S02 emissions" refers to the potential amount of S02 that
would be generated by the combustion of sulfur-bearing coal in the FBC unit
and that would be emitted to the atmosphere in the absence of limestone in
the FBC bed and sulfur retention of the ash.) It'should be noted that the
sulfur contained in the limestone was not considered in calculating the
uncontrolled S02 emissions. From the standpoint of an NSPS, uncontrolled
S02 emissions are based only on those emissions produced by the fuel. Once
the uncontrolled and outlet S02 emissions were determined, the S02 reduction
efficiency was then calculated.
3-4
-------
3.1.2 Summary Statistics for Process and Emission Data
Table 3-1 summarizes the daily average coal, limestone, and steam flow
rates for the testing period. The flow rates were calculated from the
midnight totalizer readings recorded by operating personnel and, therefore,
represent average hourly rates based on total daily flows. On test days
interrupted by power outages, the average coal and limestone rates shown in
this table were computed from the midnight and latest available day shift
readings prior to the power outages. The coal and limestone rates,
therefore, represent hourly flow rate averages indicative of test operation.
However, the average steam rates computed during test periods interrupted by
power outages also include steam output during off-test conditions. These
data, therefore, were not included in the Table 3-1 summary. As mentioned
in Section 2.5, the recorded steam rates, although considered unreliable on
an exact rate basis, were considered useful in determining operating trends.
In addition, coal feed rates from March 21 to March 25 were considered
unreliable due to a faulty electronic circuit board in the coal totalizer.
Therefore, these daily average coal flow rates are considered to be
artifically low. Appendix B discusses the nature of this problem and
discusses the treatment applied to the data in question for use in
subsequent statistical analyses.
Table 3-2 presents daily averages for other process parameters
affecting emissions and emission results during the source test period.
During the entire source test period, SO- removal efficiency averaged 93.5
percent at an average Ca/S molar feed ratio of 4.0. The average Ca/S molar
feed ratio included data at test conditions and data from process upsets and
power outages but excluded data from five days with unreliable coal flow
rate readings. Outlet SOg and NO emissions averaged 0.63 and 0.64 lb/10
Btu, respectively. Boiler load averaged 35.3 million Btu/hour, or about 70
percent of rated capacity, during the testing. Although not stated in Table
3-2, the coal fired in the boiler contained an average of 5.9 weight percent
sulfur.
Table 3-3 presents averages and standard deviations of the operating
parameters and emissions for the three parametric tests. For the data from
the first test period, which are used in the variability analysis discussed
3-5
-------
TABLE 3-1. DAILY AVERAGE PROCESS FLOW RATES
a
Date3
2/27
2/28
3/1
3/2
3/3
3/4
3/5
3/6
3/7
3/8
3/9
3/10
3/11
3/12
3/13
3/14
3/15
3/16
3/17
3/18
3/19
3/20
3/21
3/22
3/23
3/24
3/25
3/26
3/27
3/28
a!986.
"D~i 1 --
Daily hours
of test
operation
24
24
24
24
24
24
24
12.5
0
16
9
24
24
24
24
24
20
24
24
9'§
°b
Ob
11.5
24
24
24
24
24
24
12
.*« A l» |4» ft MI.1M
Average
coal feed
rate (Ib/hr)
3,010
3,080
3,180
3,120
3,080
3,160
2,960
3,240
-
3,080
3,260
3,010
3,230
3,150
3,030
2,670
3,280
2,810
3,040
3,120
-
f+
2,370^
2,380^
2,300^
2,230^
2,740C
2,800
2,760
2,460
^J 1 M ^ M A 1 * M L* M Uk £ f* \
Average
limestone feed
rate (Ib/hr)
2,190
2,120
2,280
2,380
2,290
2,290
1,960
2,260
-
2,120
2,040
2,270
2,270
2,150
2,100
2,060
2,010
1,910
2,010
2,070
-
-
1,520
2,230
2,470
2,440
2,680
2,680
2,900
3,290
. ,
ABM * ^ 4 A M
Average steam
production
rate (Ib/hr)
37,400
37,600
38,200
38,700
37,900
37,100
35,100
-
-
-
-
36,800
38,100
37,900
38,600
36,900
-
36,900
36,600
-
-
-
-
42,200
35,300
35,400
39,500
36,300
33,000
-
cCoal feed rates are considered only as estimates due to faulty coal
totalizer readings.
3-6
-------
TABLE 3-2. DAILY AVERAGE PROCESS PARAMETERS AND EMISSION RESULTS
CO
I
Date3
2/27
2/28
3/1
3/2
3/3
3/4
3/5
3/6
3/7
3/8
3/9
3/10
3/11
3/12
3/13
3/14
3/15
3/16
3/17
3/18
3/19
3/20
3/21
3/22
3/23
3/24
3/25
3/26
3/27
3/28
Weighted
Average
Load
(10° Btu/hr)
35.6
36.4
37.4
36.5
36.0
37.1
34.0
36.1
.
36.1
35.9
35.2
37.9
37.5
36.2
31.4
36.7
32.7
35.2
36.7
_
_
27. 6b
27. 8b
27. 7b
26 -5K
32. 2b
33.3
32.3
28.1
35. 3C
Bed
temperature
1,535
1,542
1.558
1.539
1.530
1.516
1.525
1.566
-
1.531
1.529
1.498
1,526
1.535
1.532
1.534
1.537
1.535
1.521
1,534
m
_
1,534
1,533
1,527
1,544
1,545
1.530
1.519
1,463
1.531
Calclum-to-sulfur
ratio
3.9
3.4
3.7
4.1
3.9
3.8
3.3
3.6
-
2.7
3.4
3.9
3.6
3.7
3.6
4.4
3.3
4.0
3.6
3.3
-
_
3.7b
5.1b
5.8b
5.6b
5.1b
4.8
6.0
7.7
4.0C
SO, Removal
efficiency
(Percent)
94.2
93.1
93.3
95.2
94.1
95.9
93.1
91.3
88.3
73.2
89.0
93.4
97.1
96.6
95.4
-
94.7
92.8
93.6
95.0
-
_
_
94.1
95.2
95.4
93.6
89.6
95.8
-
93.5
Outlet SO,
emissions
(lb/10° Btu)
0.58
0.71
0.66
0.46
0.61
0.42
0.75
0.88
3.53
2.03
1.11
0.66
0.31
0.33
0.47
-
0.53
0.68
0.61
0.53
-
_
_
0.57
0.47
0.48
0.62
1.08
0.40
-
0.63
Outlet NO
emissions.
(lb/106 Btu)
0.58
0.63
0.63
0.65
0.65
0.67
0.56
0.62
0.66
0.67
0.74
0.71
0.65
0.65
0.62
-
0.59
0.59
0.61
0.62
-
_
.
0.69
0.69
0.69
0.65
0.63
0.68
-
0.64
Stack
02,( Percent) C02
10.45
10.55
10.45
10.39
10.64
10.58
10.39
10.20
11.11
11.11
10.85
10.85
10.47
10.68
10.57
10.50
10.34
10.49
10.95
10.76
-
_
11.95
11.43
10.92
10.86
11.00
10.99
10.74
11.46
10.73
results
.(Percent)
9.52
9.57
9.80
10.03
9.64
9.47
9,66
fr.65
7.18
9.05
8.17
9.27
9.98
10.29
10.04
9.88
9.89
9.50
9.32
9.60
-
_
8.24
9.17
9.59
9.23
9.69
9.17
10.18
10.16
9.58
CO.(ppm)
466
420
395
452
464
442
570
411
420
431
465
513
420 .
408
411
426
449
494
502
371
_
_
434
444
492
450
456
576
789
1,072
485
a!986
These results are considered only as estimates due to faulty coal totalizer readings.
"-Excluding data from March 21 to March 25.
-------
TARIE 3-3 AVERAGES (STANDARD DEVIATIONS) OF OPERATING PARAMETERS
AND EMISSIONS DETERMINED FOR EACH PARAMETRIC TEST
I
00
Stack gas analysis
Test conditions
First Test Period
7.5 days of operation
used In variability
analysis
Second Test Period
0.25 x 0 limestone
Third Test Period
high SO. removal
efficiency
Coal
Test sulfur
duration, content,
days Percent
7.5 6.0
(0.2)
0.63 6.5a
0.17 5.7b
Operating
fi Load0.
10° Btu/hour
36.2
(1.95)
33.1
(0.62)
28.1
(0.13)
Bed
temperature
°F
1.538
(62)
1,540
(67)
1,470
(155)
Calclum-to
-sulfur
ratio
3.7
(0.5)
4.5
(0.65)
7.2
(0.97)
SO
Removal
efficiency.
Percent
94.2
(2.3)
91.0
(2.3)
99.4
(0.7)
Outlet SO.
emissions,
lb/10° Btu
0.60
(0.26)
1.0
(0.18)
0.06
(0.06)
Outlet NO
emissions.
lb/10° Btu
0.62
(0.06)
0.60
(0.04)
0.70
(0.03)
Oxygen,
Percent
10.46
(0.30)
11.06
(0.28)
11.55
(0.43)
CO .
Percent
9.74
(0.43)
9.23
(0.30)
10.13
(0.62)
CO,
ppm
453
(107)
507
(114)
1.007
(114)
'The mean sulfur content of the coal for the 15-hour test were assumed to be the sane as those from the two 8-hour composite samples collected during this
time period. Standard deviation of the sulfur In the coal MBS not calculated since It 1s based on analyses from only two samples.
The sulfur content of the coal for the 4-hour test was assumed to be the same as that from the one 8-hour composite sample collected during this time
period. No standard deviation could be calculated.
°Fu11 load Is at 50 million Btu/hour heat Input.
-------
in Section 4.0, the S02 removal efficiency averaged 94.2 percent at an
average Ca/S molar feed ratio of 3.7. Outlet SO, and NO emissions averaged
6
0.60 and 0.62 lb/10 Btu, respectively. The boiler operated at 72.4 percent
of full load and fired a 6 percent sulfur coal during this test period.
Based on these results, the major objective of operating at or above 93
percent S02 removal was achieved for this test while maintaining at least 70
percent of full load.
For the test using the coarser limestone with particle sizes of 0.25
inch or less, the S02 removal efficiency was 91.0 percent at an average Ca/S
molar feed ratio of 4.5. Comparing this result to that from the first test
period using the finer limestone (0.093 inch x 0.033 inch), the coarser
limestone resulted in lower S02 removal efficiency than did the finer
limestone, even at the higher Average Ca/S ratio. This lower performance
can be attributed in part to the larger particle sizes but may also be
attributed to the larger percentage of fines which may have been readily
elutriated from the bed. The limestone particle size distribution analysis
supports this contention. Eleven percent of the coarser limestone contained
particles smaller than 0.023 inches compared to only 2 percent for the finer
limestone. The CaCO, content of both limestones was 98 weight percent.
6
Outlet S02 and NOX emissions during this test averaged 1.0 and 0.60 lb/10
Btu, respectively. The boiler operated at 66.2 percent of full load and
fired a 6.5 percent sulfur coal during this test. Prior to the test using
the coarser limestone particles, the boiler was operating using the finer
limestone particles. The test using the coarser limestone particles did not
begin until all of the particles in both beds of the boiler were of the
coarser type.
Since both limestone types came from the same quarry source, there is
no reason to expect a difference in sorbent reactivity between them.
However, a difference in sorbent reactivity may have resulted from a
difference in on-site storage techniques. The coarser limestone had been
stockpiled at the plant for a long time; however, the finer limestone was
used in the operating boiler shortly after delivery from the quarry. It was
observed that the coarser limestone particles were somewhat moist and tended
to form clumps. The finer limestone was dry and flowed easily. Although
3-9
-------
the CaCO, content of both limestone types was the same, the weathering
effects of long-term stockpiling on the coarser limestone may have
contributed to its lower S02 reduction performance (i.e., lower reactivity)
compared to the finer limestpne* A sample of the coarser limestone was not
submitted for sorbent reactivity analysis due to the relatively high cost of
the analysis and the fact that operation with this limestone was a secondary
objective of the emission test.
Sorbent reactivity is the ability of a sorbent to react with S02 to
form sulfurous solid compounds. For highly reactive sorbents, less sorbent
is needed to achieve a given SO- reduction than for sorbents having low
reactivity. Westinghouse has developed a procedure to test the sorbent
reactivity of over 150 limestones and dolomites used at various FBC
Q
operating conditions. A sample of the finer limestone was submitted for
reactivity analysis and was judged by Westinghouse to be a high reactivity
g
stone relative to other stones evaluated. The basis for this comparison is
the mass of sorbent predicted as being required for 90 percent S02 reduction
for a 4 percent sulfur coal under typical FBC operating conditions specified
by Westinghouse. These operating conditions are: 1545°F (840°C) bed
temperature, 20 percent excess air, and 0.5 second gas residence time in the
11 12
bed. ' The coarser limestone was not analyzed for reactivity.
For the third test period, S02 removal efficiency averaged 99.4 percent
at an average Ca/S molar feed ratio of 7.2. As indicated by the very high
average Ca/S ratio, the limestone feed rates were very high during this
test. Outlet S02 and NOX emissions averaged 0.06 and 0.70 lb/10 Btu,
respectively, while the boiler was operating at 56.2 percent of full load
and fired a 5.7 percent sulfur coal. The boiler's heat input was lowered
below 70 percent of full load to minimize the potential for overloading both
the baghouse and ash handling system due to the higher limestone feed rate
and concomitant ash production rate. These problems would have occurred if
operation was maintained at or above 70 percent of full load.
The results of this test indicate that at this boiler load nearly all
of the S02 can be removed at very high Ca/S molar feed ratios. The results
also indicate that mass transfer effects did not preclude achieving very
high S02 emission reductions or very low outlet S02 concentrations in the
3-10
-------
stack gas. Since outlet S02 concentration does not appear to be limited by
mass transfer effects, it also should be feasible for this system to attain
90 percent S02 reduction or greater on a low sulfur coal.
Based on observations made during this test, discussions with plant
personnel, and design constraints of the auxiliary equipment (e.g., the ash
handling equipment and baghouse), it is unlikely that this level of
performance could have been maintained at or near full load due to increased
difficulty in removing the ash from the FBC system firing a high sulfur
coal. However, the baghouse and ash handling system were not designed to
handle such high particulate matter (PM) loadings as were experienced during
this test period. For a low sulfur coal operation, the limestone feed rate,
and hence PM loading, would be reduced if the Ca/S ratio were maintained
constantly at the 7.2 level observed during the test period. A 1 percent
sulfur coal, for example, would require only one-fifth the limestone
required by a 5 percent sulfur coal at the same Ca/S ratio and, presumably,
would result in flue gas PM loadings one-fifth those of the higher sulfur
coal. The baghouse and ash handling system for boiler No. 2 were capable of
handling such reduced PM loadings, even at full load. If the outlet SO,
c f.
emissions under these circumstances remained near 0.06 lb/10 Btu, the
corresponding SO* reduction would be approximately 96 percent, assuming a
heating value of 12,500 Btu/lb for the 1 percent sulfur coal.
3.2 COMBUSTION EFFICIENCY RESULTS
3.2.1 Methodology and Assumptions Used in the Analysis
Combustion efficiency was determined from data collected on March 12
over a six-hour period and from data collected over a one-hour period on
March 10. During both periods, the boiler was operated to meet the test
objective of achieving S02 removal efficiencies of at least 93 percent.
Combustion efficiency was calculated based on the combustible carbon balance
of those streams entering and leaving the FBC boiler as illustrated in
Figure 3-1. Combustible, or organic, carbon enters with the coal and leaves
with the flue gas (as C02 and CO), with the bed drain material, and with the
3-11
-------
co
i
ro
C In Coal
Limestone Bin
PI Limestone Mechanical
Bin Dust
Collector
C In Paniculate Matter
t
Stack
C In Bed Drain Material
C In Flyash
From Hoppers
Figure 3-1. Measurement points for estimating combustion efficiency.
13
-------
baghouse ash. It is assumed that a negligible amount of solids will exit
the stack; therefore, a negligible amount of combustible carbon will leave
the stack. Carbon contained in the limestone is inorganic carbon as opposed
to combustible (or organic)-carbon and hence does not affect the combustible
carbon balance.
The amount of combustible carbon entering and leaving the unit can be
determined based on the carbon content and flow rates of the inlet and
outlet streams shown in Figure 3-1. Ash analyses were performed on samples
of the baghouse ash and bed drain material to determine the total carbon and
carbonate contents. Carbonate content is required to differentiate between
organic (i.e., combustible) and inorganic carbon (unreacted CaCOj).
During the test program, the flow rates for the baghouse ash and bed
drain material were the most difficult operating variables to quantify.
This was due, in part, to the baghouse ash and bed drain material being
combined in the ash silo prior to being trucked to disposal. The bed drain
flow rate was estimated by collecting and weighing solids in drums over a
six-hour period (from 9:45 am to 3:45 pm) on March 12, 1986. The baghouse
ash flow rate was not measured for several reasons. First, the combined bed
material solids and baghouse ash contained in the ash silo could not be
completely removed prior to the test. Secondly, the baghouse ash could not
be bypassed around the ash silo to measure the flow rate of this stream
directly. Third, the costs of measuring the combined stream flow rate by
weighing the silo ash in trucks and employing a truck driver(s) to do this
job would have been very high. Finally, this method of measuring the
baghouse ash flow rate by difference had been tried in the past by plant
personnel and proved unsuccessful (i.e., due to poor material balance
closure) in accurately determining the baghouse ash flow rate.
Instead of directly measuring the baghouse ash flow rate, this flow
rate was calculated from a sulfur balance around the FBC boiler system.
Sulfur enters the system with the coal (and to a "much smaller extent with
the limestone) and leaves as S02 in the stack gas and as calcium sulfate
(CaS04) and calcium sulfide (CaS) in the bed drain material and baghouse
ash. The sulfur and total mass balances were solved to estimate the
baghouse ash flow rate (as explained in Appendix C).
3-13
-------
In addition to unburned carbon in the bed drain material and baghouse
ash, formation of CO in the flue gas represents a small source of
inefficiency in carbon use. A heating loss of 10,160 Btu per pound of
carbon in the coal results When CO is formed instead of C02. For example,
one pound of carbon will generate 14,540 Btu during complete conversion to
C02, while only 4,380 Btu is generated during conversion to CO. Therefore,
the percentage heat loss due to CO formation (% H.n) is given by the
14
following equation :
u /10,160\/ CO
Mco " I I * Cb * 100
HHV I \C02 + CO
where: HHV higher heating value of coal, Btu/lb
CO and C02 - flue gas concentration in common units
Cb fraction of combustible carbon in the coal.
By knowing the flowrate and carbon contents of the streams going into and
out of the FBC system as well as the percentage heat loss due to CO
formation, the combustion efficiency can then be calculated.
3.2.2 Combustion Efficiency Results
Because the baghouse ash samples were not taken at the same time that
the bed drain material flow rate was measured, the combustion efficiency was
estimated based on two calculation scenarios. Ash samples were taken on
March 10, 1986, two days before the bed drain material rate was measured.
Nevertheless, the operating conditions were maintained essentially constant
during those two days. In the first scenario (Scenario 1), the combustion
efficiency was estimated assuming the ash analyses results of March 10 also
applied to the ash stream on March 12, 1986. All other operating data
collected on March 12 were used in the calculations. The second scenario
(Scenario 2) assumed that the bed drain flow rate measured on March 12
applied equally to the bed drain stream on March 10. Similarly, all other
operating data from March 10 necessary to calculate the combustion
efficiency were used in this scenario.
3-14
-------
For Scenario 1, combustion efficiency was estimated to be 92.3 percent;
for Scenario 2, combustion efficiency was estimated at 92.6 percent. The
calculations to determine the combustion efficiency for both scenarios are
shown in Appendix C. The accuracy of these estimates depends upon the
accuracy of the coal, limestone, ash, and stack gas analyses as well as the
accuracy of the estimated flow rates and assumptions stated for both
scenarios. The close agreement between the results of both scenarios
indicates that the assumptions involved are reasonable and that the
combustion efficiency for boiler No. 2 during testing was approximately 92
percent.
3.3 ESTIMATION OF RECYCLE RATIO
3.3.1 Methodology and Assumptions Used in the Analysis
The recycle ratio is defined as the mass rate of entrained solids
reinjected into the boiler relative to the mass rate of coal fed. For this
analysis, the amount of solids recycled to the boiler is the amount
collected only by the multicyclone. The amount of solids recycled from the
mud drum of the convective section of the boiler to the fluidlzed bed (see
Figure 2-1) could not be determined due to a lack of process and design
information.
The recycle ratio was estimated for both scenarios based on the data
presented in Section 3.2. The approach used to estimate the recycle ratio
was to assume that the multicyclone collected ash and solids at its design
collection efficiency of 86.7 percent. Sieve analysis of solid samples
taken on March 10, 1986, for the multicyclone solids reinjected into the
boiler and for baghouse ash did not provide enough information to determine
the actual collection efficiency.
The amount of solids reinjection can be calculated using both the
assumed collection efficiency and the estimated baghouse ash flow rates
calculated in Appendix C. From this information the recycle ratio can then
be determined. It should be noted that the calculated recycle ratio can
only be considered as a rough estimate since it depends on the assumed
collection efficiency which could not be verified by direct measurement.
3-15
-------
3.3.2 Recycle Ratio Results
From information presented in Appendix C and calculations shown in
Appendix 0, the estimated recycle ratios for Scenarios 1 and 2 above were
3.9 and 4.0, respectively. It should be reiterated that additional solids
were recycled from the bottom of the mud drum but could not be measured or
estimated.
3-16
-------
4.0 EMISSION DATA VARIABILITY ANALYSIS
This chapter discusses the statistical analysis methodology used to
study the sulfur dioxide (S&2) reduction variability of data obtained from
boiler No. 2 at Prince Edward Island, Canada, and presents the results of
the analysis. The primary objective of this analysis was to investigate the
magnitude and structure of the S02 emissions data variability. A secondary
objective was to relate that variability and the mean performance results of
boiler No. 2 to a 30-day rolling average S0« reduction performance for a
fluidized bed combustion (FBC) unit with similar design and operating
characteristics.
4.1 METHODOLOGY
Sulfur dioxide emissions from a steam generating unit will vary with
time. Fluidized bed combustion boilers will exhibit a certain amount of
variability in S02 emissions due to random fluctuations in operating
parameters such as load, Ca/S molar feed ratio, bed temperature, and
fluidization velocity. Similarly, there is a certain amount of natural
variability in coal sulfur content due to the heterogeneous nature of coal,
differences in mining practices, and differences in handling and preparation
practices. Very little variation in calcium carbonate content of the
limestone was observed from the limestone analyses. There are several
factors which describe the variability characteristics of S0« data and which
influence the projected mean SO- emission levels required for compliance
with a given emission standard. These factors are the:
Standard deviation and relative standard deviation,
Autocorrelation,
Emissions distribution (normal vs.' lognormal),
Length of the averaging period and averaging method, and
Compliance policy (i.e., exceedance rate).
4-1
-------
Each of these factors and Its effect on the required mean emission level is
briefly discussed below. Appendix E contains complete details on the
computation of the percent reduction levels presented in Section 4.2.
The standard deviation-is an indicator of the spread of values that
measures the variation of measurements from the average. The relative
standard deviation (RSD) is defined as the ratio of the standard deviation
to the average or mean and is typically used to describe emissions
variability. The greater the RSD, the greater the variability in observed
S02 emission reduction. Sulfur dioxide emission datasets which have the
same mean S02 level but different SO- variability levels (standard
deviations or RSD's) will have different levels of difficulty in complying
with a given S02 reduction requirement. As increased variability is
observed, a higher mean S02 reduction efficiency will be necessary to
achieve compliance with a given SCL reduction requirement.
Emission reduction measurements taken at equal time intervals
constitute a time series in which values are not necessarily independent.
When data are collected in sequence, there can be a tendency for
observations made close together in time to be more alike than those taken
farther apart. A measure of this degree of association between observations
in a time series is termed autocorrelation and can vary from -1.0 for
inversely related observations to 1.0 for data which exhibit an extreme
degree of linear association. Sulfur dioxide emission reduction values
typically have a relatively high amount of autocorrelation (greater than
0.5), such that the value of each observation is expected to be similar to
that of the previous observation. Overall, as the autocorrelation factor
increases, a higher mean S02 reduction efficiency will be required for
compliance; this effect of autocorrelation on compliance determination is
more important for longer averaging periods. However, autocorrelation is
not as major a factor as RSD in determining S02 emission reduction
variability and predicting 30-day rolling mean Sdx removal efficiency
levels. A first order autoregressive time series model, abbreviated as
AR(1), is often used as part of the methodology to project the required
30-day rolling mean S02 removal efficiency levels. This is the most basic
autoregressive model and has been found in many cases to fit S09 reduction
15
time series measurements adequately.
4-2
-------
Observed SO- emissions data can be described by various emissions
distributions. Emissions data have generally been found to be well
represented by either the normal or lognormal distributions. The primary
difference in the two distributions is that the lognormal distribution (in
which the natural logarithms of the data are normally distributed) is
positively skewed; that is, there is an unusually large number of
observations greater than the mean. The required mean SCL reduction levels
necessary for compliance with a standard will differ depending on whether a
lognormal or normal distribution is assumed. The difference in the
projected levels is primarily a function of the length of the averaging
period and the RSD. For emission reduction data with small RSD's, the
required mean S02 reduction levels are similar for the two distributions.
However, in cases were large RSD's are expected, the difference between
projected levels is more pronounced. Similarly, for short averaging periods
the difference between the projected levels for lognormal and normal
distributions can be very pronounced, but as the length of the averaging
period increases, the differences become negligible. Because the lognormal
distribution is skewed to the right, its use results in the projection of a
slightly lower mean S02 reduction level being necessary for compliance than
if the normal distribution had been applied.
The length of the averaging period also affects the required mean
S02 removal level necessary for compliance. Longer averaging periods dampen
the effects of variability, resulting in required mean S0« reduction levels
closer to the actual reduction limit. The averaging period is defined as
the period of time (hours or days) over which emission measurements are
averaged in order to determine compliance with the S02 removal limit.
A block average (i.e., discrete average) or a rolling average may be
used to determine compliance. Discrete averages are computed on a separate,
non-overlapping basis; rolling averages are calculated by adding the most
current value and dropping the oldest value. With the use of discrete, or
block, averages, the enforceability of an emissions standard is directly
related to the averaging time. For example, a 24-hour average would be
enforced for each discrete 24-hour period and 365 such averages would be
calculated per year. A 30-day average would be enforced for each discrete
4-3
-------
30-day period and 12 such averages would be calculated per year. However, a
30-day average can be calculated and enforced on a daily basis through the
use of a rolling average. Thus, rolling average periods allow for more
frequent enforcement than do- discrete periods of the same length. For a
given averaging period, the mean SOg reduction level required for compliance
is higher for a rolling average than for a discrete average since there is a
greater potential for not meeting the emission reduction limit when using a
rolling average.
The final factor affecting the required mean S02 reduction level is the
specified compliance policy (i.e., allowed exceedance frequency). The
allowed exceedance frequency is expressed as the number or percentage of
exceedances assumed over the specified time period for a particular S02
reduction limit, and can be thought of as an assumed violation rate. Of
course, as the allowed exceedance frequency increases, the required mean SCL
reduction level moves closer to the reduction standard.
The probability of violating the limit can be calculated for a specific
averaging period and exceedance frequency; this probability is constant for
each potential period of violation over the entire term of the compliance
period. As an example, consider a 30-day rolling average method combined
with an exceedance frequency of one per 10 years. There is a potential of
having an exceedance or violation each day when using a 30-day rolling
average method, since a new 30-day average emission level is computed each
day. There are 3,650 days over a 10-year period; thus, assuming one
exceedance divided by the total number of potential exceedances (3,650)
yields 0.00027. Therefore, the probability of violating the limit each day
is a constant of 0.00027, and summing these daily probabilities over the
3,650 days results in a probability (or relative frequency) of one. The
daily exceedance probability of 0.00027 translates to a standard normal
variate Z value of 3.46.
The implication of the above example is that; in order to comply with
the emission reduction limit, the FBC boiler should operate at a target
level which is higher than the given limit by an amount equal to 3.46 times
the standard deviation of the 30-day rolling averages, before making
allowances for autocorrelation effects.
4-4
-------
4.2 RESULTS OF VARIABILITY ANALYSIS
The emission data collected at the Prince Edward Island FBC boiler
consist of hourly measurements taken over 30 consecutive days. For
variability analysis, only data from the first 7.5 days were used because
this was the longest period of continuous stable system operation during the
source testing. Data on March 27, 1986, at 8:00 a.m. were missing due to
calibration checks and were therefore deleted from the analysis. The data,
then, consisted of 179 hourly observations. The average S02 removal
efficiency during this period was 94.2 percent with a standard deviation of
2.32. This translates to an RSD of 2.46 percent.
In order to derive an estimate of the autocorrelation, SO^ reduction
values were transformed to a natural logarithmic scale and then fitted to a
first-order autoregressive time series model, or AR(1). In this analysis,
emissivity is defined as one minus the S0« removal efficiency expressed as a
fraction. Each value of the time series is modeled as a regression function
of the previous hour's value plus a random component. The transformed data
conformed to a normal distribution to a greater degree than did the original
data and, therefore, provided a better fit to the AR(1) model. The AR(2)
model, which uses the two previous hours' reduction values, was fitted to
the data to predict a given hour's SOp reduction. The coefficient of the
2-hour lag reduction value was tested for significance using the t-test
statistic. A non-significant coefficient here would support the use of the
AR(1) model.
The AR(1) model generated an estimated autocorrelation of 0.804. In
the AR(2) model, the coefficient for the second lag value was -0.12 with a
t-value of -1.61. This corresponds to a significance probability less than
0.05, meaning that this coefficient is not significantly different from
zero, at the 95 percent confidence level. Therefore, the use of the AR(1)
model is deemed appropriate for these data.
In view of the 94.2 percent average SO- reduction efficiency achieved
by the FBC unit during this period, it is natural to compare this
performance to the requirements of a 90 percent reduction standard. To meet
such a standard, the FBC unit would have to be operated to achieve a percent
reduction level above 90 percent to allow for the variability in percent
4-5
-------
reduction performance discussed above. The results of the variability
analysis described in Appendix E indicate that a mean of 91.34 percent SO2
reduction would be required to comply with a 90 percent S02 reduction limit
based on a 30-day rolling average with no more than one exceedance every 10
years. This result assumes that the FBC unit would continue to exhibit the
same variability (with respect to both amount and structure) during the
10-year period as that displayed during the 7.5 day emission test period.
The above variability analysis was also performed using the complete
dataset which contains data from February 28 to March 28, 1986. The
autocorrelation coefficient for this analysis was assumed to be the same as
the analysis indicated for the first 7.5 days, namely 0.804. This
assumption was required because a time series could not be fitted to the
complete dataset due to the deletion of data during the boiler shutdown from
March 18 to March 21. Although the variance for S02 reduction values was
greater using all 30 days of data (35.45 versus a value of 5.38 for the
first 7.5 days), the results indicated only a small difference in the
mean S02 reduction efficiency needed to comply with a 90 percent S02
reduction limit. For the complete data, the target mean SCL reduction level
is 92.55 percent, compared to the 91.34 percent S02 reduction level
determined on the basis of the first 7.5 days of operation.
The mean SOg reduction performance of the FBC boiler No. 2 at Prince
Edward Island was 94.2 percent during the first 7.5 days of stable operation
and 93.5 percent during the entire 30-day period, including process upsets
and power outages. If this performance were maintained, the FBC boiler
would be in compliance with a 90 percent SO- reduction standard using a
30-day rolling average method.
4-6
-------
5.0 REGRESSION ANALYSIS
This chapter discusses the methodology and results of the regression
analysis used to relate sulfur dioxide (S02) reduction efficiency achieved
by the fluidized bed combustion (FBC) boiler at Prince Edward Island to key
operating parameters. Regression models similar to those reported in the
literature were used to fit the observed data. The key operating parameters
for which data were expected to influence S02 reduction performance were:
the calcium-to-sulfur (Ca/S) molar ratio, bed temperature, sorbent particle
size, and operating load. The models used in the analysis were based on
those operating parameters. The impacts of these parameters on the S02
removal efficiency in FBC boilers have been discussed in previous
17 18
reports. ' The primary objectives of this analysis were to determine the
impacts of load on S02- reduction efficiency and develop a model to predict
S02 reduction efficiency at full load. A secondary objective of the
analysis was to develop a predictive model of S02 reduction performance
based on other operating parameters.
5.1 METHODOLOGY AND MODELS USED IN THE REGRESSION ANALYSIS
Several regression models were investigated in an attempt to develop a
model that predicts S02 reduction efficiency as a function of operating
parameters in the data base. If Y. represents the S02 reduction efficiency
value at hour i and X^ represents the value of an operating parameter (e.g.,
Ca/S ratio or load) at hour i, a regression equation can be expressed in the
form:
Y1 - f (X., B) + e.
where: f - a function of X.,
B - a constant coefficient, and
e^ = errors.
Assumptions implicit in this method are that the errors, e^, are independent
with mean equal to zero, and are normally distributed.
5-1
-------
An example of the above model is:
Y. - Bj exp (B2 * LOAD.) + e.
where: Y.. = S02 reduction efficiency at hour i,
LOADi = boiler load at hour i, and
Bj, B2 - coefficients estimated by least squares method.
The general forms for the nonlinear regression models used to predict
S02 reduction efficiency (EFF) in this analysis were:
EFF - 100 [1-exp (-Bj * Ca/S)]
EFF = 100 [1-expU-Bj - B2/LOAD) * Ca/S}]
EFF = 100 [1-exp {(-Bj/LOAD) * Ca/S}]
EFF = 100 [l-(Bj * exp {(-B2/LOAD) * Ca/S})]
£FF - 100 [l-exp((-Bj * Ca/S - B2) Ca/S}]
In
(1-EFF)
where:
100
Ca/S
LOAD
BEDTEMP
Model 1
Model 2
Model 3
Model 4
Model 5
Bj * Ca/S exp [-B2/(BEDTEMP + 273)] Model 6
calcium-to-sulfur ratio,
boiler load,
average bed temperature, °C, and
constant coefficients of the models.
The forms of Models 1, 2, and 3 are obtained from Reference 19.
However, Models 2 and 3 were modified by replacing the superficial
fluidization velocity term shown in Reference 19 with the operating load.
This modification was necessary since it was not possible to quantify
fluidization velocity accurately for the test period. Data for air flow
5-2
-------
rates, which are used to calculate the fluidization velocity, were
considered to be inaccurate due to improper location of the air flow
measuring devices in the ducts downstream of the forced draft fan. Instead,
it was assumed that operating load was directly proportional to superficial
fluidization velocity since the operating bed temperature and the
air-to-coal ratio remained relatively constant and the cross-sectional
area of the beds did not change. Model 4 was developed as a slight
variation to Model 3 to improve the fit of the data. The form of Model 5 is
the same as that shown in Reference 20, while Model 6 is derived from
Reference 21.
In performing the regression analyses using these six models, data
containing inaccurate coal flow rate readings from March 21 to March 25 were
deleted from the complete dataset as were data collected from a 15-hour
period on March 26 when the coarser sized limestone was used. Performing a
regression analysis which includes the effects of the coarser limestone
could not be accomplished due to the limited amount of operation (15
one-hour data points) with this limestone. Therefore, regression analyses
fff\ ))
were performed using the SAS4*'Procedure NLIN on the remaining data. This
SAS^procedure fits nonlinear regression models by least squares methods.
Since nonlinear models are more difficult to estimate than linear models,
there is no guarantee that the procedure will converge and thereby result in
estimates for the coefficients found in the model equations.
5.2 REGRESSION ANALYSIS RESULTS
Results from the nonlinear regression analyses are presented in
Table 5-1. Operating hours with Ca/S molar ratios of zero were excluded
from all analyses in order that the nonlinear procedure could converge
properly. The results indicate that Model 2 was most successful in fitting
? 7
the data with a coefficient of determination (R) of 0.67. The R is a
measure of the percent of the variability in the dependent variable (SCL
reduction efficiency) that is "accounted for" by the independent variables
(Ca/S ratio, load, bed temperature, or a combination of the three
variables). The results, then, indicate that a large percentage of the
variance of the SCL emissions remains unexplained by these models.
5-3
-------
TABLE 5-1. RESULTS OF REGRESSION MODELS
Model 1: EFF - 100 [1-exp (-Bj * Ca/S)]
0.712
0.613
Model 2: EFF - 100 [l-exp{(-B1-B2/LOAD) * Ca/S}]
2
- 1.53
B2 - -27.8
0.666
Model 3: EFF - 100 [1-exp {(-Bj/LOAD) * Ca/S}]
28.09
R - 0.529
Model 4: EFF - 100 [l-(Bj * exp {(-B2/LOAD) * Ca/S})]
» 1.12 B2 « 29.52 R - 0.533
Model 5: EFF - 100 [l-exp((-Bj * Ca/S - B2) Ca/S}]
Bj - -0.036 B2 - 0.88 R2 - 0.617
Model 6: In
d-EFF)
100
* Ca/S exp [-B2/(BEDTEMP + 273)]
- Undefined B2 Undefined R - (Undefined, SAS procedure
did not converge)
Notes:
1. EFF 1s the S02 removal efficiency 1n percent.
2. Ca/S denotes the calcium to sulfur molar ratio.
3. LOAD is boiler heat Input in 106 Btu/hour.
4. BEDTEMP is the average bed temperature in °C.
5. Bj, B2 are coefficients estimated by least square methods.
2
6. R is the coefficient of determination.
5-4
-------
Unfortunately, the results of using this best fit model (Model 2)
23 24
contradict results presented in other theoretical studies. ' This model
predicts that as the operating load (or fluidization velocity) increases,
S02 reduction efficiency increases as well. This result is counter to other
theoretical model results which predict that S02 reduction will decrease
with increasing fluidization velocity, since the residence time available
for sulfur capture reactions will be reduced. For this reason, this model
is not recommended for use in predicting S02 reduction efficiency. Instead,
the next best-fit model (Model No. 5) is recommended for use since it
explains 62 percent of the SO- reduction efficiency variability and its
structure is consistent with observed FBC boiler behavior. Using this
predictive model, a Ca/S ratio of 3.8 is required, at any operating load, to
achieve the average S02 reduction efficiency of 94.2 percent observed over
the first 7.5 days of operation of the FBC boiler at Prince Edward Island.
This predicted Ca/S ratio agrees well with the average Ca/S ratio of 3.7
determined from the data collected over the first 7.5 days of operation at
this average S02 reduction level. It should also be noted that this model
predicts that a 94.2 percent S02 reduction efficiency will be achieved at a
Ca/S ratio of 3.8 for this boiler when operating at full load.
There are two reasons for the lack of fit of the data to the above
models. First, the primary objective of the test program was to achieve an
S02 reduction efficiency of at least 93 percent for most of the testing.
This was, in fact, accomplished for a great part of the test program. As
Table 5-2 indicates, for the hours of operation that were included in the
regression analyses, only about 25 percent of the data corresponded to SO*
reduction efficiencies below 92.7 percent, with only about 10 percent of the
data below 88.9 percent. Nearly 80 percent of the S02 reduction data was
between 88.9 and 97.1 percent. Thus, the data used in the regression
analyses did not contain enough variation to predict S02 reduction efficiency
adequately as a function of the operating parameters. Some variation in the
data was included as a result of emission data collected during the power
failure periods (when the boiler was not operating at test conditions), but
these periods did not last long enough to greatly increase the total
variation in S02 reduction data.
5-5
-------
TABLE 5-2. SOg REDUCTION EFFICIENCY STATISTICS3
Percent
Mean 93.4
Standard Deviation 5.92
Maximum . 100
90% Quantile 97.1
75% Quantile 95.8
50% Quantile (Median) 94.5
25% Quantile 92.7
10% Quantile 88.9
Minimum 44.1
aNumber of Data Points Used in the Analysis: 554
5-6
-------
The second reason for the the lack of fit of the data 1s also related
to the test program objective. Only two short-term parametric tests (each
less than one day) were performed. A better fit of the data would have been
realized 1f more parametric'tests had been performed or if these tests were
of longer duration. Figures 5-1, 5-2, and 5-3 show the lack of variation of
the S02 reduction data when plotted against Ca/S ratio, load, and bed
temperature, respectively. These figures also show that the data were
scattered about a small band of SOg reduction efficiency values and that an
essentially horizontal line can be fitted to this data. In fact, Figure 5-2
shows graphically no relationship between SO* reduction efficiency and load.
This conclusion 1s supported by the results using the above regression
models in that operating load had little or no effect on SO* reduction. The
lack of variation 1n the S02 reduction and operating data as shown in the
three figures may also have contributed to the contradictory results
discussed above for Model 2.
5-7
-------
en
i
1.0
o.v
0.8
0.7
£ 0.4
O
ID
£ O.S
«
I
o
-o
oc
o"
to
«
0.3
0.2
0.1
0.0 *
-*-
0
-*-
2
*MM*M*MM* *
t*M**MMIM* »
M MMMM* M * t *
MM * * ** M*
» » »
* «« *
*
-+-
S
-+-
8
NOTEI
Ca/S Ratio
71 018 HAD HISSINO VALUES OR WERE OUT OF RANGE 364 DBS HIDDEN
+
10
Figure 5-1. Plot of S02 emission reduction versus Ca/S ratio.
-------
01
I
10
1.0 t
O.f
0.8
0.7
c
o
U 0.4
to
o
r-
M
O
3
o
Si
oo
o.s
0.4
t
»
*
t * *
* * *««« M***MM**«* * **
M I M M MMtMMMMMMMMM ** *
* * »»» * $ M* MMMMMt
« MM* * * **** ** * *
* ** *»**
* *
* «
0.2
O.I
0.0 »
-4 4 4 4 4 4 4 4 4 1 4 + * 4 + + 1 » + + +
20 21 22 23 24 29 26 27 28 29 30 31 32 33 34 33 36 37 38 39 40
Load, Ib million Btu/hr
NOTEI 70 0§8 HAD HISSING VALUES OR HERE OUT OF RANOE 328 018 HIDDEN
Figure 5-2. Plot of S02 emission reduction versus load.
-------
t.o *
0.9
O.B
C °'7
O
1
+J
o
a 0.6
S-
A
C 0.9
0
M
O
cn .§ o.4
o
cvj
0 0.3
to
0.2
0.1
0.0 <
********************* **
** ***********************>******«* * *
** ***** ****>*** * **»* **
* * * * ** ********* ** *
* *** * *** * ***** *** **
* *** * * * *
** *
*
*
*
*
* * *
'
790 760 770 780 790 BOO BIO 820 830 840 890 860 B70 880 890 900 910 920 930 940 930
Bed Temperature, °C
NOTE! 70 018 HAD HI8SIHO VALUES OR MERE OUT OF RANGE 299 0*8 HIDDEN
Figure 5-3. Plot of SC^ emission reduction versus bed temperature.
-------
6.0 REFERENCES
1. Alliance Technologies Corporation. Fluldlzed Bed Boiler Emission Test
Report. Canadian Forces Base, Summerside, Prince Edward Island, Canada.
Prepared for the U.S. Environmental Protection Agency. Research Triangle
Park, NC. EMB No. 86-SPB-2. March 1986.
2. Taylor, M.E.D. and F.D. Frledrich. The CFB Summerslde Project Canadian
State-of-the-Art 1n ABC Boilers. Canada Centre for Mineral and Energy
Technology. Ottawa, Canada. Report No. ERP/ERL 82-10(TR). April 1982.
p. 46.
3. Reference 2, p. 44.
4. Reference 2, p. 42.
5. SAS Institute Incorporated. SAS User's Guide: Basics, 1982 Edition.
Gary, NC. SAS Institute Incorporated. 1982. 921 p.
6. U.S. Environmental Protection Agency. Stack Sampling Technical
Information: A Collection of Monographs and Papers - Volume I (Summary
of F-Factor Method for Determining Emissions from Combustion Sources).
Research Triangle Park, NC. Publication No. EPA-450/2-78-042b.
October 1978. pp. 29 to 43.
7. Reference 1, Appendix D.
8. Newby, R.A., et al. A Technique to Project the Sulfur Removal
Performance of Flu1d1zed-Bed Combustors. In: The Proceedings of the
Sixth International Conference on Fluldlzed Bed Combustion,
Volume Ill-Technical Sessions, U.S. Department of Energy (ed.).
Washington, DC. August 1980. pp. 803 to 814.
9. Reference 1, Appendix D.
10. Letter and attachments from Newby, R.A., Westlnghouse Electric
Corporation, to Schlff, H., GCA Corporation. May 21, 1986. p. 6.
Results of the sorbent tests on the limestone at Prince Edward Island.
11. Reference 1, Appendix D.
12. Reference 10.
13. GCA Corporation. Monitoring of A1r Emissions from the Georgetown
University Fluldlzed Bed Boiler. Prepared for the U.S. Environmental
Protection Agency. Research Triangle Park, NC. EPA Contract
No. 68-02-2687. September 1982. p. 33.
6-1
-------
14. Reference 7, p. 34.
15. DuBose, D.A., et al. (Radian Corporation). Statistical Analysis of
Wet Flue Gas Desulfurization Systems and Coal Sulfur Content.
Volume I: Statistical. Analysis. Prepared for the U.S. Environmental
Protection Agency. Research Triangle Park, NC. EPA Contract No.
68-02-3816. August 18, 1983. p. 7.
16. Technical note from Giguere, G.C., et al., Radian Corporation, to
Stevenson, W.H., EPArSDB. March 1985. pp. 2 to 8. Determination of
mean SO, emission levels required to meet a 1.2 ID/million Btu emission
standard for various averaging times and compliance policies.
17. Young, C.W., et al. (GCA Corporation). Technology Assessment Report
for Industrial Boiler Applications: Fluidized-Bed Combustion.
Prepared for U.S. Environmental Protection Agency. Research Triangle
Park, NC. Publication No. EPA-600/7-79-178e. November 1979.
pp. 64 to 84.
18. Aul, E.F., et al. (Radian Corporation). Fluidized Bed Combustion:
Effectiveness of an S02 Control Technology for Industrial Boilers.
Prepared for U.S. Environmental Protection Agency. Research Triangle
Park, NC. Publication No. EPA-450/3-85-010. September 1984. pp. 3-11
to 3-15.
19. Anthony, E.J., et al. The Fluidized Bed Combustion of a High-Sulphur
Maritime Coal in a Pilot-Scale Rig and Industrial FBC Boiler. In:
Proceedings of the Eighth International Conference on Fluidized-Bed
Combustion, Volume I, U.S. Department of Energy. Morgantown, WV. July
1985. pp. 248 to 249.
20. Razbin, V.V., et al. Fluidized Bed Combustion of High-Sulphur Maritime
Coal. Canada Centre for Mineral and Energy Technology. Ottawa,
Canada. Report No. 85444 (OPJ). December 1984. p. 8.
21. Wheeldon, J.M., et al. Experimental Results from the Grimethorpe PFBC
Facility. In: Proceedings of the Eighth International Conference on
Fluidized-Bed Combustion, Volume I, U.S. Department of Energy.
Morgantown, WV. July 1985. p. 324.
22. SAS Institute Incorporated. SAS User's Guide: Statistics, Version 5
Edition. Cary, NC. SAS Institute Incorporated. 1985. pp. 575 to
606.
23. Reference 16.
24. Reference 17, p. 71.
25. Reference 16, pp. B-l to B-3.
6-2
-------
APPENDIX A
SAMPLE PROCESS DATA SHEET AND LIST OF NOMENCLATURE
-------
Attached are sample process data sheets collected by computer from
boiler No. 2 at Prince Edward Island. A description of each parameter
listed in the header of each sheet is presented in the nomenclature. For
example, stack SO* emissions were monitored by Channel number 309 on the 0
to 1,000 ppm scale.
A-l
-------
TABLE A-l. SAMPLE PROCESS DATA SHEET
t\i
C.F.B. SUHMERSIDE FLUID1ZED 6.ED-6QILER KO.l OR ND.2- TEST MiTfi
Of*^ 02 03 04 05 06 07 08 09 10 II 12 13 14 15 i-.
2-STH FL DR« PR ST.H.PR STK TEMP E M IN E « OUT DW1 LEU FU ORFT FD *ISC TOT AIR IP SUC ID DIS OXVGEH SG2 SMOKE Ft) PS
IE G IH E G GUT 86 Cl'TP E G PIH E G POUT 02 C02 CO S02 S02 HO 2E G IH E G OUT 66 Ol'TF E G PIH E G FQl-
,«i2A- AIR PLH PR TOT COflL LIME FD TOT LEV PAR LEU STK T24 STK T12 CTR T24 CTft T12 IGH T24 IGH TI2 ft'.' t T F£l- Tl FEl- T2 fit' FED
26- AIR PLH PR TOT COM. LIRE Ft> TOT LEU PAR LEU STK T24 STK TI2 CTfc T24 CTR T12 IGH T24 IGH T12 AU 6 T FBl< Tl FBC> T> nl> FBI-
' 20*45:00
! ?8^
) tOO
' 700
1
; 21:00:00
' 300
; 600
,' 700
i
' 21:15:00
: 300
; ffi
I 700
; 21:30:00
JQJ»
i tOu
1 700
21:45:00
300
600
700
22:00:00
3.3 lSt'6
-67.3' -169.)
494. 4C SOI
459. 9C 494
2.3 187t
-69.2 -173.)
499. 3C 507
457.4C 453
2.3 l*»8
-69.9 -176.1
516. 1C 522
456. 7( 492
2.4 IS45
-69.4 -I?*/*
5! 9. it 529
456. 7C 492
2.4 1££3
-62.6 -171.6
504. 1C SOS
452. ?( K-u
-------
TABLE A-l. SAMPLE PROCESS DATA SHEET (CONTINUED)
CO
" *'V *-^*^ C'F-B- SUMHERS» FLUIMZEt- BED-BOILER HO. 1 OP
f Oi^
02
2-STH FL DRH'PR
^**IE G IH E G OUT
»**2fl- RIB
28- AIR
ie!3o:oo 16333
300 25.0
600 19922
700 30879
18:45:00 16507
300 24.7
600 20343
700 30329
19:00:00 17198
300 24.6
600 22312
700 30623
19: 15: 00 17023
300 24.4
600 22480
700 30905
19:30:00 16731
300 14.2
600 22119
700 30774
19:45:00 16569
300 24.1
600 22321
700 29776
20:00:00 17290
300 24.0
600 22133
700 30415
20M5:00 16879
eOO 22*32
700 30625
PLH PR
PLH PR
839
19.3
1052.9
1333.1
851
19.2
994.3
1348.2
842
19.1
1098.1
1341.1
862
19.0
1102.3
1345.1
884
19.9
1114.9
1351.2
915
18.8
1169.8
1370.3
911
19.9
1117.9
1353.1
887
18.3
1092.4
1359.0
03
ST.H.PR
88 OUTP
04
STK TEMP
E 6 PIH 1
TOT CORL L1HE Ft-
TOT COflL
406
9.0
2322
2522
406
9.0
2595
2658
406
9.0
2436 '
2428
406
9.0
2520
2513
406
9.9
2515
2463
406
3.9
2662
2614
406
9.0
2572 .
2551
406
8.9
2370
2357
LIRE FD
171.9
10.3
886
-480
172.8
10.3
891
-480
173.5
10.3
889
-480
175.!
10.3
870
-481
175.9
10.3
346
-481
176.4
10.3
857
-491
177.3
10.3
343
-481
177.4
10.3
346
-482
05
E y IH
E G POUT
TOT LEV
TOT LEV
101.7
-4.8
407
370
101.8
-4.9
363
,346
101.7
-4.9
359
339
101.7
-4.9
385
351
101.8
-5.0
415
352
101.. 6
-5.0
417
368
102.0
-5.0
382
364
101.9
-5.0
410
374
06 07
E U OUT DRH LEV
02 C02
PRR LEV STK T24
PflR LEV STK T24
129.9 3.3
10.32 10.61
34 843. 9C
68 897. 9C
129.4 4.6
10.51 10.21
33 8M.IC
67 92I.3C
131.3 6.4
10.49 10.29
30 871. 3C
67 916.0C
130.6 2.0
10.32 10.39
31 885. 4C
66 911. 4C
130.4 6.7
10.56 10.10
33 871. 7C
67 908. 3C
130.7 6.7
10.37 10.30
32 864.SC
63 911. SC
130.2 6.5
10.12 10.42
34 907. 1C
63 908. OC
130.4 7.6
10.71 9.86
34 872. 6C
69 909. 8C
08 09
IIO.j- (til I'HTh
10 11
FU DRFT FD DISC TOT BIR II- SUC
CO S02 502 HO
12
13
11 US OXYGEN
2E G IH E G OUT
STt TI2 CTR T24 CTR T12 IGH T24 IGH TI2
STK TI2 CTR T24 CTR TI2 IGH T24
-8.3 1926
442.6 177.9
842. 7C 834. 6C
783. 2C 886. 3C
-11.5 1921
408.2 IbO.O
660. 3C 851. OC
811. SC 907. 8C
-13.3 1930
359.7 157.5
667.4C 657.7C
805. 1C 903. 8C
-10.6 1936
360.8 181.8
881. 7C 872. 1C
800. 4C 900. 3C
-14.5 1894
347.8 106.3
868. 7C 859. 3C
793. 3C 899. OC
-11.0 1958
339.3 107.5
661. 4C 652.61
798. 6C 901. 4C
-15.3 1921
331.4 216.8
903.21 892. 3C
795. 7C 899. 3C
-10.8 1914
336.3 109.5
868. 8C ftfcl.lt
793. 4C 899.6C
43060 -269. 1
406 236.9
838. 3C 840.3C
763. 9C -I.2C
42629 -282.5
334 234.2
855.6C 857.7C
737. 8C 22.8C
44509 -312.8
369 237.2
861. 6C 864. SC
734. 8C 10.5C
44720 -306. b
383 .234.5
876.9C .878.9C
790. 9C -6.2C
44187 -304.8
320 242.8
864. OC 865.8C
730. 5C 1.3C
44627 -307.0
306 241.2
e56.ec 656.9C
773. 1C -3.3C
43982 -3H.O
411 231.3
898. 5C 900. 2C
775.5C -15.9C
44648 -306. fc
303 241.3
864. 6C 6bt.7C
775. 3C -8.3C
IGH T12
17.3
246.3
5.0C
869. 7C
16.9
243.7
5.0C
383. 7C
17.6
251.6
-10. 1C
896. 3C
19.8
253.2
-0.1C
833. 6C
19.6
253.3
1.1C
883. 9C .
l?.9
254. '5
0.6C
836. 5C
20.5
257.6
-3.4C
834. 6C
19.5
254.6
2. 1C
335. 1C
fil' 6 T
fll* B T
8.9
182.9
829
899
9.1
183.7
846
921
9.2
185.9
652
916
8.8
187.2
666
912
9.1
187.3
853
910
8.9
187.7
64t
912
8.7
189.0
8t5
909
is?:l
653
910
14
SOI
6B OUTP
FB(< Tl
FBD Tl
279.8
-63.0
506. 4C
546. 9C
276.0
-59:3
506. 1C
5*53. 3C
276.5
-70.5
5I7.8C
550. SC
271.9
-67.1
525. 5C
552. 3C
267.5
-66.6
522. 3C
S53.5C
J71.3
-87.4
524. 4(
553. 3C
274.8
-67.6
539. 1C
551. 9C
273.3
-68.5
520. 3C
554. OC
15 16
SHDKE FV PRS
E G PIH E G POUT
FBD T2 RV FBD 1
FBD T2 RM FBD T
2.5 IB89
-63.1 -161.3
508. OC 504
476. 1C 310
2.5 I860
-61.3 -162.2
501. 6C 501
487. OC 519
2.4 1881
-71.3 -173.3
511. 6C 512
479.5C 514
2.4 1867
-69.8 -171.7
509. 7C 515
478. 6C 314
2.4 1871
-70.3 -173.1
516.6C 516
479. 9C 515
2.4 1B75
-70.2 -172.8
518.8C 518
482. 1C 516
2.4 1843
-71.2 -174.8
529. OC 531
480.7C 314
-7?.-24 -m
521. 4C 517
482.6C 516
-------
TABLE A-2. PROCESS DATA NOMENCLATURE
GENERAL BOILER PARAMETERS
Channel
Number
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
306
307
308
309
Variable
Name
STM FL
DRM PR
ST.H.PR
STK TEMP
E W IN
E W OUT
DRM LEV
FU DRFT
FD DISC
TOT AIR
ID SUC
ID DIS
Oxygen
so2
Smoke
FW PRS
°2
co2
CO
so2
Description
Steam Flow Rate
Steam Drum Pressure
Steam Header Pressure
Stack Temperature
Economizer Water Inlet Temperature
Economizer Water Outlet Temperature
Steam Drum Level
Furnace Draft Pressure
Forced Draft (FD) Discharge
Total Air Flow Rate
Induced Draft Suction Pressure
Induced Draft Discharge Pressure
Oxygen Content (Economizer inlet)
S02 Content
Smoke
Feedwater Pressure
Stack 02 (dry)
Stack C02 (dry)
Stack CO (dry)
Stack S02 (dry)
Range
0-25,000
0-14,000
0-14,000
35-260
0-150
0-260
+250
+ 80
0-2,500
0-65,000
-800-0
-250-0
0-19.9
0-2,000
0-100
0-225
0-25
0-20
0-5,000
0-1,000
Units
kg/hr
kPa
kPa
°C
°C
°C
mm.HpO
mm.H20
mm.hLO
Ib/hr
mm.HnO
mm.H20
vol. %
ppm
%
kPa
vol. %
vol. %
ppm
ppm
A-4
-------
TABLE A-2. PROCESS DATA NOMENCLATURE
GENERAL BOILER PARAMETERS (CONTINUED)
Channel
Number
310a
311
312
313
314
315
316
Variable
Name
so2
NO
2E G IN
E G OUT
BB OUTP
E G PIN
E G POUT
Description
Stack S02 (dry)
Stack NOY (dry)
A
Economizer Gas Inlet
Temperature
Economizer Gas Outlet
Temperature
Boiler Bank Outlet
Pressure
Economizer Gas Inlet
Pressure
Economizer Gas Outlet
Pressure
Range
0-5,000
0-1,000
35-315
35-315
+100
+100
+200
Units
ppm
ppm
°C
°C
mm.H20
mm.H20
mm.hLO
aThe signal from this channel on the plant S0~ analyzer was not calibrated
and was not used during the testing.
A-5
-------
TABLE A-3. PROCESS DATA NOMENCLATURE
BED "A" PARAMETERS
Channel
Number
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
Variable
Name
2A-AIR
PLN PR
TOT COAL
LIME FD
TOT LEV
PAR LEV
STK T24
STK T12
CTR T24
CTR T12
IGN T24
IGN T12
AV B T
FBD Tl
FBD T2
AV FBD T
Description
Air Flow to Bed A
Plenum Pressure
Total Coal Flow
Limestone Feed Flow
("Bed A")
Total Bed Level
(pressure differential)
Partial Bed Level
(pressure diffential)
Bed Temperature Near the
Stoker End (24 in high)
Bed Temperature Near the
Stoker End (12 in. high)
Bed Temperature at the
Center of Bed (24 1n. high)
Bed Temperature at the
Center of Bed (12 in. high)
Bed Temperature Near The
Burner End (24 in. high)
Bed Temperature Near the Burner
End (12 in. high)
Average Bed Temperature
Freeboard Temperature
at Location 1
Freeboard Temperature
at Location 2
Average Freeboard Temperature
Range
0-35,000
0-2000
0-5077
0-3090
0-1000
0-700
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
Units
Ib/hr
mm.H20
Ib/hr
Ib/hr
mm.H20
mm. FLO
°C
°C
°c
°c
°c
°c
°c
°c
°c
°c
A-6
-------
TABLE A-4. PROCESS DATA NOMENCLATURE
BED "B" PARAMETERS
Channel
Number
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
Variable
Name
2B-AIR
PLN PR
TOT COAL
LIME FD
TOT LEV
PAR LEV
STK T24
STK T12
CTR T24
CTR T12
IGN T24
I6N T12
AV B T
FBD Tl
FBD T2
AV FBD T
Description
Air Flow to Bed B
Plenum Pressure
Total Coal Flow
Limestone Feed Flow ("Bed B")
Total Bed Level (pressure
differential)
Partial Bed Level (pressure
differential)
Bed Temperature Near the
Stoker End (24 in. high)
Bed Temperature Near the
Stoker End (12 in. high)
Bed Temperature at the Center
of Bed (24 in. high)
Bed Temperature at the Center
of Bed (12 in. high)
Bed Temperature Near the
Burner End (24 in. high)
Bed Temperature Near the
Burner End (12 in. high)
Average Bed Temperature
Freeboard Temperature
at Location 1
Freeboard Temperature
at Location 2
Average Freeboard Temperature
Range
0-35,000
0-2000
0-5077
0-3090
0-1000
0-700
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
300-1100
Units
Ib/hr
mm.HLO
Ib/hr
Ib/hr
mm.HpO
mm.H-O
°C
°C
°c
°c
°c
°c
°c
°c
°c
°c
A-7
-------
APPENDIX B
DESCRIPTION AND RESOLUTION OF COAL FEED TOTALIZER MALFUNCTION
FROM MARCH 21 TO MARCH 25, 1986
-------
Totalizers were used to calculate the average material feed rates to
the boiler. Each weigh feeder was equipped with a totalizer which
maintained a running total of the weight passing over the weigh feeder
belts. There were three totalizers needed: one to record the total weight
of the coal fed to the unit and one each to record the total weights of
limestone fed to Bed "A" and Bed "B". Since the totalizer readings were not
indicated on the Digistrip (computer) printout, Radian personnel maintained
a daily log sheet and recorded the instantaneous totalizer readings once
each hour as indicated in Table B-l. From these readings, the average coal
and limestone rates for the hour just past were computed. Also, the plant
operators recorded the totalizer readings each day at midnight in order to
complete the plant daily summary sheet. The midnight readings were used by
Radian personnel to determine the average flow rates during the periods in
which totalizer readings were not recorded (e.g., normally from the hours of
1700 to 2400 and from 0000 to 0900). The totalizer readings were the only
accurate way to determine the feed rates to the unit.
It was noticed during the period the boiler was being restarted on
March 21, 1986, that the totalizer indicated coal feed rates were
artificially low compared to the coal feed controller setting, steam
production, and S02 reduction performance at the calculated Ca/S ratios.
For this reason, Radian personnel requested the plant instrument man to
check the coal totalizer. The instrument man replaced the electronic
circuit board in the back of the totalizer at 1345 on March 25, 1986, and
also recalibrated the weigh feeder at 1545 on March 26, 1986, to ensure the
accuracy of the coal rate for the remainder of the test.
In order to determine the accuracy of the suspect coal totalizer
readings during the period in question, system performance indicators were
plotted comparing the performance of the unit during this period to the
performance of the unit after the totalizer was worked on and also to a
smooth period of operation from March 11 through'March 13, 1986. Figure B-l
shows the steam production rate as a function of coal rate for the three
time periods. Only the hourly average coal rates during the period from the
hours of 1000 to 1800 for each day were used to compute the steam unit
B-l
-------
TABLE B-l. SAMPLE DAILY LOG OF TOTALIZER READINGS AND OTHER PROCESS READINGS
00
ro
Coal
Time
0000
0900
1000
1100
1200
1300
1415
1515
1615
2400
Totalizer
Reading
(kg)
93226692
93239645
93241087
93242529
93243965
93245411
93247209
93248660
93250087
93261232
Coal Rate"
(tonnes/day)
-
_
-
34.61
34.46
34.70
34.52
34.82
34.25
34.51
Bed "A"
Totalizer
(kg)
2096009
2100755
2101297
2101834
2102367
2102792
2103309
2103715
2104097
2106469
L Imestone
Bed "B"
Totalizer
(kg)
254035
259256
259835
260389
260937
261485
262136
262643
263135
267099
Temperatures (°C)
L Imestone8
(tonnes/day)
-
_
-
26.18
25.94
23.35
22.43
21.91
20.98
19.62
Economizer
Gas
Outlet
-
189
188
190
191
191
191
189
189
-
Air Fan
Outlet
_
51
52
53
53
53
55
55
55
-
Air Fan
Inlet
_
22
23
24
24
24
25
26
26
-
Stack
so2
(ppm)
_
69.7b
55.3
46.2
75.1
96.4
89.2
149.1
-
Analysis
°2
(Percent)
_
10.625
-
10.32
10.53
10.48
10.43
10.58
10.55
-
aTonnes 1s metric tons. .
Values of these variables 60900 were taken as equivalent to 0845 recording because analyzers had been calibrated at 0900.
-------
e-a
Steam Rate/Coal Rate
Ibs. of steam produced/lb of coal
_iO
ro
GJ
cn
en
-n
^.
10
c
o>
00
1 >
00
0
^.
n>
-$
n>
-*>
-h
o
^j.
fD
S O
s ^
S n
o
O)
-s
^.
to
o
3
1 «
<£>
00
O\
W i
5]
OJ J
I J
*M
^H
4
"~
ro -
u :
I j
> ~
-
-
ro -
tn j
I j
Z. j
_
ro -
vl J
I j
>
^ "
1800(3/27)*.:
1000(3/38)^
s n "
1000(3/11) , D _
-1800(3/11) ° D
^1000(3/12) I^j
D
«-1800(3/12) n
N 000(3/1 3) 0
0° °
j-1800(3/13) Boiler down due to clinker formation ^
3/18 @ 1000 to 3/21 @ 1300
NoOO(3/22) n
n
Q
a
*-1800(3/22)
S 000(3/23) n C3
D
D
a
«-1 800(3/23)
"k nLJ
^ 1000(3/24) D D
LJL
a
-1800(3/24) D
L pM ^J
1000(3/25) Totalizer Circuit Board Replaced nn
[JI " 3/25 @ 1345
-1800(3/25) n
^1000(3/26) n ,_,
0°
Recallbratlon
3/26 @ 1545
r 1800(3/26)
1000(3/27) 0.
n°
n
-------
ratios for the purpose of the system performance comparison. During these
periods, Radian personnel were present to record totalizer readings each
hour. As can be seen from the figure, the steam production unit ratio for
the period from March 22, 1986, to the time when the totalizer circuit board
was replaced appears to be out of line with the other two time periods of
data. If the totalizer had been indicating artificially low coal feeds
during this period, with no major changes in other operating variables
compared to the other periods, this would explain the shift in the data.
The steam production during the period in question was within the range
maintained during the majority of the length of the test.
Calcium use was also compared for the three periods and is shown in
Figure B-2. The SO- captured, in moles/hour, was computed by subtracting
the outlet S02 flow rate from the inlet S02 flow rate. The inlet S02 flow
rate was calculated using the coal feed rate determined from totalizer
readings and the sulfur content (as fired) in the coal measured by an
on-site sulfur analyzer. All sulfur in the coal was assumed to be liberated
as S09. The outlet SO- flow rate was computed by adjusting the outlet S00
6
concentration from a ppm basis to a lb/10 Btu basis using the F-factor
method and then multiplying this emission rate by the heat input to the
boiler (10 Btu/hr). The heat input rate was determined from the coal feed
rate and assumed heating value. A coal heating value of 12,500 Btu/lb on a
dry basis was assumed based on previous coal analysis data provided by the
plant. This value was adjusted to the actual (wet basis) heating value
using the measured coal moisture content from coal samples which correspond
to the coal feed at the time.
As can be seen from Figure B-2, the S02 capture rate during the period
from March 11 to 13, 1986, was greater than the capture rate during the
period in question even though the calcium feed rates were in the same
range. Again, this discrepancy would be explained if the indicated coal
feed rate during the period in question was artificially low. The data
shown for the period after recalibration were recorded at lower coal feed
rates while operating at very high sorbent ratios in order to obtain 99+
percent S02 removal. For this reason, the S02 capture rate is lower due to
the reduced S02 inlet feed rate.
B-4
-------
SO2 CAPTURE VS. SORBENT FEED
CD
I
cn
ft:
i
\
3
o
2
uT
ft:
t
g
o
(/>
wJ -
4.5 -
4 -
3.5 -
3 -
2.5 -
2 -
1.5 -
u__ n
f^u n ^^
n
0 AS *
° o
0 A A .
A A
A
A A
0 ^A ^ A
0 0
0 ° °
$ ^
O
0 -March 11 -13, 1988
O March 22, 1986 Recallbratlon
A Recallbratlon Test End
' I i i i i i i i i i i i i
14 18 22 26 3O 34 38
SORBENT FEED, MOLES Ca/HR
Figure B-2. Calcium utilization comparison.
-------
Examination of the steam production and calcium utilization
relationships, both being functions of the Indicated coal feed rate, show
the period from March 21 to 25, 1986, to be inconsistent with the period of
smooth operation from March 11 to 13 and with the period after the totalizer
circuit board was replaced from March 25 to the end of the test on
March 28, 1986. For this reason, the data from March 21 to 25 were excluded
from the regression analyses.
B-6
-------
APPENDIX C
COMBUSTION EFFICIENCY CALCULATIONS
-------
DETERMINATION OF BAGHOUSE ASH FLOW RATE BY SULFUR BALANCE
Scenario #1 - Assume that ash analysis on March 10
1s the same as that on March 12.
Limestone
Coal
Date: March 12, 1986
Time: 9:45 A.M. to 3:45 P.M.
Coal Data
Coal Feed Rate - 3,168 Ib/hr
Percent Sulfur in Coal - 5.91
Baahouse Ash Data
Percent Sulfur in Ash = 6.91
1
^Baghouse
Solids
FBC
SYSTEM
Bed Material
Solids
Limestone Data
Limestone Feed Rate = 2,147 Ib/hr
Percent Sulfur in Limestone = 0.02
Bed Material Solids Data
Bed Material Solids Flow Rate =
616 Ib/hr (measured during this
time period)
Percent Sulfur in Bed Material
Solids >7.68
Other Data
SOg Removal Efficiency -97.4 percent
SULFUR BALANCE: AMOUNT THAT WILL BE RETAINED, IN
AMOUNT RETAINED, OUT
(SO, removed) (Coal Rate) * S 1n Coa1 ) + (Limestone Rate) * s 1n
(BEDRATE)
100
S 1n BEI
100
+ (BAGRATE)
100
100
(0.974)(3,168)(0.0591) + (2,147)(0.0002) = (616) (0.0768) + (BAGRATE) (0.0691)
Solving this equation for baghouse ash rate gives:
BAGRATE - 1,961 Ib/hr
C-l
-------
DETERMINATION OF BAGHOUSE ASH FLOW RATE BY SULFUR BALANCE
Scenario #2 - Assume Bed Material Solids Flow Rate Measured
on March 12 is the same as that on March 10.
-^ Baghouse
Solids
Limestone
Coal
FBC
SYSTEM
Date: March 10, 1986
Time: 5:00 P.M.
Coal Data
Coal Feed Rate - 3,135 Ib/hr
Percent Sulfur in Coal - 5.88
Baahouse Ash Data
Percent Sulfur in Ash - 6.91
Bed Material
Solids
Limestone Data
Limestone Feed Rate = 2,435 Ib/hr
Percent Sulfur in Limestone = 0
Bed Material Solids Data
Bed Material Solids Flow Rate =
616 Ib/hr (measured on March 12
from 9:45 A.M. to 3:45 P.M.)
Percent Sulfur in Bed Material
Solids - 7.68
Other Data
S02 Removal Efficiency - 96.4 percent
SULFUR BALANCE: AMOUNT RETAINED, IN -
AMOUNT RETAINED, OUT
(SO, removed) (Coal Rate) f * s 1n Coa1 | + (Limestone Rate) I % S 1n L1me )
* y 100 f \ 100 J
(BEDRATE)
S 1n BEI
(BAGRATE)
1n BAG
100 i 100
(0.964) (3, 135) (0.0588) + (2,435)(0) - (616)(0.0768) + (BAGRATE) (0.0691)
Solving this equation for baghouse ash rate gives:
BAGRATE = 1,886 Ib/hr
C-2
-------
CALCULATION OF COMBUSTION EFFICIENCY BASED ON SCENARIO #1
Combustion Efficiency - 100,*
[(Coal Rate)(% C)-(BEDRATE)(% C-% CC02)BE[)-(BAGRATE)(%C-% CC02)BA(,] _ %H
(Coal Rate) (% C) " co
where % C - Total carbon content, wt. percent
% CC02 = Carbon content present in carbonate form, wt. percent
% HCQ - Percentage heat loss due to CO formation
Bed Materials
Coal Data Baohouse Ash Data Solids Data
Coal Rate - 3,168 Ib/hr BAGRATE - 1,961 Ib/hr BEDRATE = 616 Ib/hr
% C - 64.9 % C - 9.38 % C = 1.96
HHV - 11,860 Btu/lb % CC02 - 1.58 % CC02 = 230
Other Data
CO = 401 ppm - 0.0401 percent
C02 » 10.27 percent
Note: The total carbon content for the Bed Materials Solids is less than
carbon content present in carbonate form. Therefore, assume that
the total carbon content is the same as carbon content present in
carbonate form. In other words, it is expected that a negligible
amount of combustible carbon will be present in the bed material
solids.
10.160 CO rh * inn 10.160 / 0.0401
co \ HHV / \C0 + CQ LD 1UU "11,860 10.27 + 0.0401
/1
\
(0.649) * 100 = 0.22
Combustion (3168)(64.9) - 1,961 (9.38 - 1.58) - 616 (2.30-2.30) n
EffiClenC^ " - 3168(64.9). -
92.3 Percent
C-3
-------
CALCULATION OF COMBUSTION EFFICIENCY BASED ON SCENARIO #2
Combustion Efficiency = 100..*
[(Coal Rate)(% C)-(BEDRATE)(% C-% CC02)BEp-(BAGRATE)(%C-% CC02)BAG] _ %H
(Coal Rate) (% C) ' co
where % C - Total carbon content, wt. percent
% CCO* - Carbon content present in carbonate form, wt. percent
% H - Percentage heat loss due to CO formation
Bed Materials
Coal Data Baohouse Ash Data Solids Data
Coal Rate - 3,135 Ib/hr BAGRATE - 1,886 Ib/hr BEDRATE - 616 Ib/hr
% C - 66.67 % C - 9.38 % C - 1.96
HHV - 11,660 Btu/lb % CC02 -1.58 % CC02 - 230
Other Data
CO « 544 ppm * 0.0544 percent
C02 * 10.06 percent
Note: The total carbon content for the Bed Materials Solids is less than
carbon content present in carbonate form. Therefore, assume that
the total carbon content is the same as carbon content present in
carbonate form. In other words, it is expected that a negligible
amount of combustible carbon will be present in the bed material
solids.
60\ / 0.0544
10.06 + 0.0544)
Combustion (3,135)(66.67) - 1,886 (9.38 - 1.58) - 616 (2.30-2.30) n
Eff1ClenC*" (3,135) (66.67)..
92.6 Percent
C-4
-------
APPENDIX D
ESTIMATION OF RECYCLE RATIO
-------
Definition: Recycle Ratio = Flow Rate of Recycled solid/Coal Feed Rate
Assumptions: (1) Collection efficiency of multicyclone is at the design
level of 86.7 percent
(2) Collection efficiency of baghouse is 100 percent.
Equation Used:
(1) Solid Flow to Multicyclone, F - Recycle Solid Flow, R
+ Baghouse Ash Flow, B
(2) Collection Efficiency, Percent - {f x 100 - ^ x 100
(3) Solving Equation 2 for R: (Collection Efficiency) B
R
(100 - Collection Efficiency)
I. For Scenario I, B - 1,961 Ib/hr and Coal Rate - 3,168 Ib/hr
P f86.7Ul.961) 19 7fl, lh/.
R « mn OK. 7 * 12,783 lD/hr
1UU - OO./
Recycle Ratio - 12,783/3,168 -
II. For Scenario 2, B - 1,886 Ib/hr and Coal Rate - 3,135 Ib/hr
o r86.7Hl.886) ,
R " 100 - 86.7 " 1
,.,.
lb/hr
Recycle Ratio = 12,294/3,135 -
D-l
-------
APPENDIX E
PROCEDURES USED IN VARIABILITY ANALYSIS25
-------
Let X represent mean SCL emissivity for the given averaging period.
The SO2 emissivity In any period Is equal to 1.0 minus the SO* reduction
percentage In that period, expressed as a fraction. Thus, emissivity
represents the fraction of the Inlet or uncontrolled SO* emissions that are
emitted. Similarly, a percent reduction requirement can be expressed as an
emissivity limit (e.g., a 90 percent S0« reduction requirement corresponds
to an emissivity limit of 10 percent SOg, or 0.10).
Assume that X is normally distributed and let L be the mean of all X.
This projected long-term mean emissivity level, or L, is also the target
level required to meet a given emissivity limit. Let C be the allowance for
random variation in emissivity or the difference between the target level
and the emissivity limit. Thus, the emissivity limit equals (L + C). The
formula for computing C is derived as follows:
P (X > L + C) = 0.00027, where
X is normally distributed with mean y and variance a .
*» X
0.00027
Z > *- \ - 0.00027, where Z is a standard normal variate
P ( Z > 3.46 ) - 0.00027
- 3.46 or C = 3.46 a^
a- X
It is assumed that hourly S02 emissivity values follow an AR(1) time
series process. Therefore, in computing the standard deviation of the mean
E-l
-------
it is necessary to adjust for the covariance structure associated with a
first-order time series process. The standard error is calculated as
where pis the lag 1 or first-order autocorrelation coefficient and CTh is the
hourly standard deviation. Since
the standard error has increased relative to the case where independence is
assumed. The equation for f_(p) is derived using the theoretical
I
autocorrelation function of <|>kfor an AR(1) process, where is the
autocorrelation coefficient and k equals the lag. The covariance at lag k
equals * multiplied by the variance. After applying this result to the
standard statistical formula for the variance of a sum, extensive algebraic
manipulation leads to the following equation for fn(p):
'n (P)
- P
1 + P -
fi - P"V
- P.
The final step in deriving the formula used to compute the target level
is to develop an equation for o_. Recall that
where h - hourly emissivity values
n - number of hours in averaging period
where d - daily emissivity measurement
- (RSDd)(L)/Vf24(P),
E-2
-------
where RSD^ = relative standard deviation on a daily basis.
Also recall that C = (Za) a?,
^
where Z is the critical standard normal variate value corresponding to the
probability (e.g., 0.00027 above) for a given exceedance frequency and
averaging period. Substitution yields
C = (Za)(L)(RSDd) 'fn(p)/f24(p)
Target Level = L = - , for n > 25
r+ (Za)(RSDd) 'fn(p-)/f24(p)
and L = - jL- - - , for n < 25
1 + (Za)(RSDd) /f^pT
where (L + C) equals the given emissivity limit.
As an example, a 90 percent S02 reduction standard corresponds to an
emissivity limit of 0.10. If C, the allowance for variability, is
determined for a given situation to be 0.03, then L, the long-term or target
level mean emissivity, would be 0.10 minus 0.03, or 0.07. Thus, the
long-term mean S02 reduction required for compliance would be 100 minus 7,
or 93 percent.
E-3
------- |