United States
          Environmental Protection
          Agency
Office of Air Quality
Planning and Standards
Research Triangle ParK NC 27711
EPA-450/3-91-004
December 1990
          Air

-------
                                  DISCLAIMER


This report is issued by the Emission Standards Division of the Office of Air
Quality Planning and Standards of the Environmental Protection Agency.  It
presents technical data of interest to a limited number of readers.  Copies
are available free of charge to Federal employees, current contractors and
grantees, and non-profit organizations - as supplies permit - from the Library
Services Office (MD-35), U. S. Environmental Protection Agency, Research
Triangle Park, NC 27711, phone 919-541-2777 (FTS 629-2777), or may be obtained
for a fee from the National Technical Information Service, 5285 Port Royal
Road, Springfield, VA 22161, phone 703-487-4650 (FTS 737-4650).
                       Publication No. EPA-450/3-91-004

-------
               APPENDIX A

 ANALYSIS OF THE CONTINUOUS SO? CONTROL
CAPABILITIES OF SPRAY DRYER/FABRIC FILTER
 AND  SPRAY  DRYER/ESP SYSTEMS APPLIED TO
       MUNICIPAL WASTE COMBUSTORS

-------
                                TABLE OF CONTENTS
1.0
2.0
3.0
4.0
5.0

Overview 	
1.1 Purpose 	
1.2 Overview of Data Analyzed 	
1.3 Summary of Statistical Methods 	
1.4 Conclusions 	 	
Description of Data Analyzed 	
2.1 Millbury, Massachusetts 	
2.1.1 Facility Description 	
2.1.2 Emissions Data . . . . , 	
2.2 Bridgeport, Connecticut 	
2.2.1 Facility Description 	
2.2.2 Emissions Data 	
2.3 York County, Pennsylvania 	
2.3.1 Facility Description 	
2.3.2 Emissions Data 	
2.4 Stanislaus County, California 	
2.4.1 Facility Description 	
2.4.2 Emissions Data 	
APCD Inlet Values 	
3.1 Short-Term Variability 	
3.2 Impact of Averaging Times 	
3.3 Conclusions 	
S02 Control Performance of Spray Dryer/Fabric Filter System . .
4.1 Short-Term Variability 	 	
4.2 Impact of Averaging Times 	
4.3 Conclusions 	 	
S02 Control Performance of Spray Dryer/ESP Systems 	
5.1 Short-Term Variability 	 '. 	
5.2 Impact of Averaging times 	
5.3 Conclusions 	
Page
Al-1
Al-1
Al-1
Al-1
Al-2
A2-1
A2-1
A2-1
A2-1
A2-1
A2-1
A2-3
A2-5
A2-5
A2-6
A2-6
A2-6
A2-10
A3-1
A3-1
A3 -3
A3 -9
A4-1
A4-1
A4-2
A4-7
A5-1
A5-1
A5-2
A5-4
bjw:033                                    A1

-------
                                 1.0   OVERVIEW
1.1   PURPOSE
      On December 20, 1989, the U.S.  Environmental Protection Agency (EPA)
proposed limits on emissions of sulfur dioxide ($03) from new and existing
municipal waste combustors (MWC's) (54 FR 52209 and 52251).  This report
presents the results of analyses conducted to determine achievable SO?
performance levels for spray dryer/fabric filter (SD/FF) systems and spray
dryer/electrostatic precipitator (SD/ESP) systems applied to MWC's.  The
specific objective of these analyses was to assist with selection of
appropriate S0£ emission limits and averaging times, based on data from MWC's
using these two control technologies.
1.2  OVERVIEW OF DATA ANALYZED
      Continuous~"em1ssions monitor (CEM) data for SOg were obtained and
analyzed from four different MWC's, three with SD/FF systems and one with a
SD/ESP.  The MWC's with SD/FF systems are located in Bridgeport, Connecticut;
York County, Pennsylvania; and Stanislaus County, California.  The MWC with
the SD/ESP system is located in Mi11 bury, Massachusetts.  Hourly average S02
CEM data were available at both the SD Inlet and the stack (I.e., FF or ESP
outlet) from the Bridgeport, York County, and Mill bury MWC's.  Data from
Stanislaus County were limited to hourly stack measurements.  Additional
information on the extent of data available from each of these facilities is
presented in Section 2.0.
1.3  SUMMARY OF STATISTICAL METHODS
      Several statistical analysis techniques were used to characterize the
variability of MWC S0£ emissions.  These techniques Included:

      •     time series plots to visually examine trends in S02 emissions over
            time and to identify data gaps or anomalies;

      •     normality testing and cumulative frequency distribution plots to
            evaluate the distribution of S0£ data;
                                                               •
      •     routine summary statistics .(mean, median, standard deviation,
            etc.);

      •     first-order autoregressive time-series analysis to Identify
            possible underlying time dependencies in the data that could alter
            their "randomness"; and
bjwi033
                                     Al-1

-------
      •     maximum estimated emissions (referred to as "exceedance values")
            based on the appropriate statistical means and standard deviations
            for different averaging times.
                         i                        *
      The hourly S02 data  were tested for normality using the Shapiro-Wllk
statistic.  Normality of data 1s Important because Inferential statistics,
such as probabilities used for predictions, depend on the underlying
population being normally distributed.  When departures from normality are
significant, any predictive statistics can be very unreliable.
      Based on general knowledge of environmental data, both the raw S02
measurements and the natural logarithms (In) of the raw data were examined for
normality.  The top half of Figure 1-1 shows a frequency histogram of the raw
hourly S02 inlet data from the Bridgeport HWC; the bottom half displays  the
same data on a In scale.  Clearly, the In transformation results in a much
more symmetrical distribution.  The existence of near symmetrically
distributed data is a key criteria which must be met 1f statistical data are
to be used for predictive purposes.  As a result, predictive statistics
derived from In-transformed data will be more valid than those derived from
the raw data.  Mean values calculated from In-transformed data are referred to
as geometric means, as opposed to arithmetic means, which are calculated from
the raw measurements.
1.4   CONCLUSIONS
      Based on analysis of the available data, four primary conclusions  were
drawn.  These are:

      •     SOj? emissions from MWC's are highly variable and should be
            averaged over a 24-hour or longer period.  Short-term (e.g.,
            hourly) emissions of S02 from MWC's are highly variable due to the
            heterogeneity of municipal waste.  This results in short-term S02
            emissions that are significantly greater than the long-term
            (monthly or annual) mean.  Therefore, to determine the achievable
            S02 control performance of a SD/FF or SD/ESP system, averaging of
            hourly data to reduce the Impact of short-duration "spikes"  in S02
            levels 1s necessary.  Based on the intensity and duration of these
            spikes, a 24-hour averaging period 1s beneficial.  Use of a
            shorter averaging period was not able to significantly reduce the
            Impact of many of the spikes, and expected maximum emission levels
      *A11 SO. data contained 1n this report were corrected to 7 percent 02 prior
to statistical analysis.
bjw,C33                                Al-2

-------
  400
«  300
e
••
  200
  100
       48   78  108  138  168  198  22B
 288288318348378408438488498828

Inlet 802 Concentration (ppm)
                                                                           SB9  588  819
  300
  200
  100
      3.97 4.12 4.28 4.43 4.S8 4.73 4.88 8.03 8.IB 8.33 8.48 8.83 8.78 8.83 8.08 8.23 8.38

                          Mtural Log of Inlet 802 Concentration

         Figure 1-1. Frequency Histograms for Bridgeport Inlet SO2
                                          Al-3

-------
            were much higher than the long-term mean.  Use of a longer
            averaging time had limited Impact on lowering the maximum expected
            average S02 level.

            The daily geometric means of uncontrolled S0j> levels are normally
            distributed.  For each of the three data sets with SD Inlet data,
            the dally geometric means of the uncontrolled S02 levels were
            normally distributed.  These MWC's combust different municipal
            solid wastes and represent different combustor manufacturers and
            designs.  These data were also collected at different times of the
            year.  Based on the similarity 1n emission from these different
            facilities, It Is expected that the uncontrolled S02 levels from
            other MWC's are similarly distributed and that conclusions reached
            regarding the continuous compliance performance of the APCD
            systems examined 1n this study are applicable to MWC's In general.

            SD/FF systems are capable of continuously achieving SO;? reductions
            of greater than 80 percent.  Data from the MWC 1n York County,
            Pennsylvania, demonstrate that SD/FF systems can achieve long-term
            S02 reductions of greater than 90 percent.  However, due to short-
            term variability 1n uncontrolled S02 levels, lower emission
            reductions will be frequently encountered.  Based on a 24-hour
            averaging period and use of geometric means to reduce the
            variability in Individual hourly readings, a requirement of 80
            percent S02 reduction can be continuously achieved.

            SD/ESP systems are capable of continuously achieving 70 percent
            reductions in S0£.  Data from the MWC in Millbury, Massachusetts,
            demonstrate that SD/ESP systems can achieve long-term S02
            reductions of about 80 percent.  As with SD/FF systems, however,
            normal variations in uncontrolled S02 levels will frequently
            result in lower emission reductions.  Based on a 24-hour averaging
            period and use of geometric means to reduce the variability 1n
            Individual hourly readings, a requirement of 70 percent
            reduction can be continuously achieved.
Supporting Information for the first two conclusions is presented in
Section 3.0.  Analyses supporting the last two conclusions are presented in
Sections 4.0 and 5.0, respectively.
bjwi033                                Al-4

-------
                       2.0   DESCRIPTION  OF  DATA ANALYZED

      The following four sections provide a brief description of each of the
four facilities examined in this report and a general overview of the
emissions data analyzed.
2.1   MILLBURY, MASSACHUSETTS
2.1.1  Facility Description
      The Mi 11 bury, Massachusetts facility, developed by Wheelabrator
Technologies, Inc., consists of two 750-TPD mass burn combustors.  Each of the
combustor trains consists of a Von Roll  reciprocating grate, a Babcock &
Wilcox boiler, and a SD/ESP system supplied by Wheelabrator Air Pollution
Control Systems.  The plant's permit limits S02 emissions to 0.21 Ibs/million
Btu (equal to roughly 130 ppm).  Lime slurry and dilution water are injected
into the SD through a dual-fluid nozzle.  The slurry feed rate is based on the
permitted S0£ reduction requirements.  Dilution water is added to reduce flue
gas temperature, which is normally controlled to around 255°F.  Dried SD
solids and fly ash are collected in a three-field ESP having a design SCA of
333 ft2/l»000 acfm at a flue gas flow rate of 160,000 acfm.
2.1.2  Emissions Data
      Hourly S0£ data were obtained from the Mi 11 bury MWC for a period of 63
consecutive days, from July 15 through September 15, 1988.  Figure 2-1
displays the hourly inlet data, hourly outlet data, and hourly percent
reduction data.  Some periods existed where data were missing.  Table 2-1
shows the total number of hours; the mean; and the minimum, median (i.e., 50th
percentile), and maximum values for the hourly Mi11 bury data.
      Most of the inlet data ranged from 100 to 300 ppm with a mean of roughly
180 ppm, but increased to over 500 ppm during several brief periods (a few
hours).  Outlet values were generally less than 100 ppm with a mean of less
than 40 ppm, but increased to near 200 ppm on several occasions.  Most hourly
S02 percent reductions were greater than 80 percent, but dropped to less than
50 percent during several brief periods of time.
2.2   BRIDGEPORT, CONNECTICUT
2.2.1  Facility Description   *
      The Bridgeport MWC Is located in Bridgeport, Connecticut.  The facility
consists of three 750-TPD mass burn combustor and SD/FF trains.  The
b]H:033                                A2-1

-------
                                        Inlet
                                      Reduction
0  IM100100400900OT0700M*m  IMC IIM UM UM UM UM
                    •nnntMn inmtr

  Figure 2-1.  Hourly SO2 Data for Millbury


                          A2-2

-------
        TABLE 2-1.  SUMMARY STATISTICS FOR HOURLY S02 DATA AT MILLBURY
Summary
Statistic
Number of hours
Mean
Minimum
Median
Maximum
Percent
Inlet (ppm) Outlet (ppm) Reduction (%)
1343
182.9
54.8
172.6
547.2
1343
38.0
1.9
34.9
226.7
1343
79.6
1.0
80.5
97.7
combustors and SD/FF systems were supplied by Wheelabrator Technologies, Inc.
The permit limit for S0£ 1s 0.32 Ibs/million Btu (equal to about 200 ppm)
without a stated averaging time.  There is no percent reduction requirement.
Each SD 1s equipped with a dual-fluid atomizer.  Lime slurry and dilution
water flow rates are adjusted to control SD outlet temperature, S02 percent
reduction, and outlet S02 emission rate.  The fabric filter consists of ten
compartments, each with 180 teflon-coated fiberglass bags.  Reverse air is
used for bag cleaning.  The air-to-cloth ratio with all compartments in
service is 2.28 acfm/ft2.  The baghouse can have up to two compartments off
line for maintenance, but normally operates with all compartments in service.
2.2.2  Emissions Data
      Hourly S0£ data were obtained for 122 days from the Bridgeport MVIC,
covering the period of August 1 through November 30, 1989.  Figure 2-2
displays the hourly -inlet data, hourly outlet data, and hourly percent
reduction data.  The period roughly between hourly observations 1500 and 1600
was deleted from the data file due to a spray dryer malfunction that was
caused by scaling of the slurry delivery system, low delivery pressure, and
poor atomlzatlon.  Wheelabrator Indicated that this was the first time in
approximately two years of SD operation that this problem had occurred.  The
missing data near hourly observation 1800 was caused by a maintenance outage
following startup of the repaired spray dryer.  Table 2-2 summarizes the
number of hours; the mean; and the minimum, median, and maximum values for the
hourly data.
bju:033
                                     A2-3

-------
                                  Inlet
                                    Outlet
                                  % Reduction
Figure 2-2. Hourly SO2 Data for Bridgeport
                     A2-4

-------
       TABLE 2-2.  SUMMARY STATISTICS FOR HOURLY SO  DATA AT BRIDGEPORT
Summary
Statistic
Number of hours
Mean
Minimum
Median
Maximum
Percent
Inlet (ppm) Outlet (ppm) Reduction (%)
1587
168.1
50.3
155.3
617.1
2208
38.8
10.7
32.5
286.4
1573
75.1
9.1
78.2
96.4
      The Bridgeport Inlet and outlet data'are similar to the Ml11 bury data.
Most of the Inlet data ranged from 100 to 300 ppm with a mean of roughly
170 ppm, but Increased to over 400 ppm during several brief periods and were
In excess of 600 ppm once.  Outlet values were generally less than 100 ppm
with a mean of less than 40 ppm, but Increased to near 200 ppm on several
occasions.  Most hourly $03 percent reductions were greater than 75 percent,
but dropped to less than 50 percent on several occasions.
2.3   YORK COUNTY, PENNSYLVANIA
2.3.1  Facility Description
      The York County MWC is located in Manchester Township.  The facility
consists of three 448-TPD rotary waterwall mass burn combustors supplied by
the Resource Energy Systems Division of Westinghouse Electric Corporation.
The SD/FF system for each unit was supplied by Joy Technologies, Inc.  Permit
limits for S0£ are 30 ppm at 7 percent 03 or 70 percent removal based on a
one-hour averaging period.  Each SD is equipped with a rotary atomizer and i's
designed for a flue gas floVrate of 95,144 acfm at 400°F.  During normal
operation the SD inlet temperature varies from 350 to 400°F and the outlet
temperature 1s between 260 and 300°F.  Lime slurry is fed to the atomizer head
tank and then mixed with dilution water based on the SD outlet temperature set
point.  The ratio of slurry and dilution water is based on signals from the
S02 CEM's located at the SD Inlet and the stack.  The fabric filter consists
of six compartments with pulse-jet cleaned fiberglass bags.  The design gross
air-to-cloth ratio is 2.5 acfm/ft2 at normal flue gas flow and 3.2 acfm/ft2 at
maximum flow.
bjw:033
                                     A2-5

-------
2.3.2  Emissions Data
      Hourly S02 data were obtained from all three units at York County for
two periods totaling 28 days:  March 10-30, 1990, and April 29-May 5, 1990.
However, because Unit 3 was out of service for most of this period, the
analysis of data was limited to Units 1 and 2.  Hourly inlet data, hourly
outlet data, and hourly percent reduction data for Units 1 and 2 are presented
1n Figures 2-3 and 2-4, respectively.  These plots do not Include several
periods when oil was burned to maintain steam production during waste feed
problems or during shut down and start up.  All data values recorded during
the oil burning periods were excluded from further analysis.  Table 2-3 shows
the number of hours; the mean; and the minimum, median, and maximum values for
the hourly data.
      The two York County units exhibit similar S02 levels, but are noticeably
lower at both the inlet and outlet than at Mi 11 bury and Bridgeport.  Most of
the Inlet data ranged from 50 to 250 ppm with a mean of roughly 110 ppm, but
increased to over 300 ppm during several brief periods.  Outlet values were
generally less than 100 ppm with a mean of less than 20 ppm, but increased to
near 200 ppm on several occasions.  Most hourly S02 levels were greater than
90 percent, but dropped to less than 50 percent on several occasions.
2.4   STANISLAUS COUNTY, CALIFORNIA
2.4.1  Facility Description
      The Stanislaus County MWC, located in Crows Landing, California,
consists of two 400-TPD Martin GmbH mass burn water-wall combustors.  The
overall facility was designed by Ogden Martin Systems, Inc.  Ammonia is
injected Into the upper furnace of each combustor to reduce nitrogen oxide
(NOX) emissions.  The S02 emissions from each combustor are controlled with a
Flakt SD/FF system.  Lime slurry 1s injected Into the SO through dual-fluid
nozzles, with the slurry feed rate controlled according to the stack S02
concentration and the dilution water flow adjusted based on the SO outlet
temperature.  The design flue gas flow rate exiting the SD 1s 94,000 acfm at
285°F.  The pulse-jet FF has six compartments, each with 266 teflon-coated '
fiberglass bags.  Net air-to-cloth ratio is 3.2 acfm/ft2.  At the time t'he '
data used in this analysis were collected, the facility's S02 permit was based
on an outlet S02 emission rate of 30 ppm.
bjn:033                                A2-6

-------
1-
s
S'~
   o      toe
                     MO   /
                        tun i
                                             Inlet
                                          Outlet
                                       lioi
                                       «•     TOO
                                        % Reduction
 Figure 2-3.  Hourly SO2 Data for York County Unit 1
                    A2-7

-------
 1
 !
    o     too
Inlet
                                     Outlet
                                   lu
    0     10*     MO
                                        «**     700
Figure 2-4. Hourly SO2 Data for York County Unit 2
                      .A2-8

-------
                          TABLE 2-3.  SUMMARY STATISTICS FOR HOURLY SO, DATA AT YORK COUNTY

Summary
Statistic
0
Number of hours
Mean
Minimum
Median
Maximum
Inlet

Unit 1
563
"118.5
2.9
110.8
380.6
(ppm)

Unit' 2
539
99.6
6.5
94.0
310.9
Outlet (DDID)

Unit 1
564
15.1
1.0
10.7
213.3

Unit 2 ;
537
12.2
1.0
7.6
377.2
Percent Reduction (%)

Unit 1
562
88.7
17.4
90.6
99+

Unit 2
537
89.3
1.0
91.8
99+
I
IO

-------
2.4.2  Emissions Data
      Hourly S02 data were obtained from both units at Stanislaus County for
the period between March 16 and May 11, 1989.  As Indicated above, the
facility's permit limit during this period of time required monitoring of SOg
levels in the stack only.  Therefore, no data1were collected at the SO inlet
and percent reductions In S02 cannot be calculated.  Figure 2-5 displays the
hourly outlet data for both units.  Note that there were a significant number
of data gaps throughout this period.  Table 2-4 shows the number of hours; the
mean; and the minimum, median, and maximum values for the hourly data.
            TABLE 2-4.   SUMMARY  FROM  STATISTICS  FOR HOURLY SOp DATA
                            FROM STANISLAUS COUNTY
Summary
Statistic
Number of hours
Mean
Minimum
Median
Maximum
Unit 1
Outlet (ppm)
920
7.2
0.0
4.0
143.2
Unit 2
Outlet (ppm)
959
9.1
0.0
5.2
168.7
      The two Stanislaus County units exhibited lower outlet S02 levels than
were encountered at any of the other three MWC's, with most hourly values
below 30 ppnr and the mean from both units less than 10 ppm.  As with the other
three MWC's, however, there were several hourly SO? levels above 100 ppm.
Although these data suggest that the SD/FF systems at Stanislaus County are
capable of high levels of S0£ removal, the lack of inlet S02 measurements
meant that the Stanislaus County data could not be used to analyze the level
of S02 reduction that SD/FF systems can achieve.  However, during an earlier
compliance test at the plant, S02 reductions of greater than 90 percent were
measured.
bjw:033                                A2-10

-------
                                             Unrtl
         400    600     800     1000
               Observation Number
1200
1400
                                               Unit 2
                  i_- 	     I
                  600     800    1000
                Observation Number
   1200      1400
Figure 2-5. Outlet SO2 Data for Stanislaus County
                      A2-11

-------
                         3.0  SPRAY DRYER INLET VALUES

      This section describes the analyses of uncontrolled S02 emissions from
the Military, Bridgeport, and York County MWC's.  The objective of these
analyses was to examine statistical similarities and differences in
uncontrolled SOg emissions from these three MWC's as an aid in assessing the
applicability of conclusions reached from these specific units to MWC's in
general.
3.1  SHORT-TERM VARIABILITY
      The figures and tables presented in Sections 2.1.2, 2.2.2, and 2.3.2
demonstrate the significant degree of short-term variability in uncontrolled
S02 data.  At all three sites, maximum hourly $02 averages at the SO inlet
were more than three times greater than the mean value on the same unit.  For
example, the plot of inlet $03 levels at Bridgeport (Figure 2-2) shows several
hourly SOg spikes that were above 400 ppm and one that was above 600 ppm, even
though most of the data were less than 160 ppm.
      Note also from the summary statistics for each data set, shown in
Table 3-1, that the difference between the minimum value and the median is
significantly smaller than between the median and the maximum value.  This
skew in the data suggests that the inlet $03 levels data are not normally
distributed.
        TABLE 3-1.  SUMMARY STATISTICS FOR HOURLY INLET S02 DATA (ppm)
Summary
Statistic
Mean
Minimum
Median
Maximum
Mi 11 bury
182.9
54.8
172.6
547.2
Bridgeport
168.1
50.3
155.3
617.1
York County
Unit 1 Unit 2
118.5
2.9
110.8
380.6
99.6
6.5
94.0
310.9
bjvi033
                                     A3-1

-------
      Table 3-2 shows the Shapiro-Wilk normality statistic (U) for the hourly
SOg Inlet data and the natural log (In) of the hourly Inlet data.  An asterisk
by the W-value Indicates a normal distribution at the 95 percent confidence
level.  Frequency histograms help to visually'display the differences between
the hourly and the In hourly datasets Implied by the W's.  Figure 1-1
(presented in Section 1.3) shows the frequency histograms for the Bridgeport
hourly Inlet data (W - 0.922) and the In hourly Inlet data (W - 0.988).
  TABLE  3-2.   SHAPIRO-WILK NORMALITY TEST STATISTIC  (W) FOR HOURLY  INLET DATA
                                                            York County
   Inlet Dataset             Bridgeport    Mlllbury      Unit 1      Unit 2
Hourly
Ln hourly
0.922
0.988*
0.903
0.984
0.912
0.940
0.889
0.972
 * Data are normally distribution at the 95% confidence level


      Though only the Bridgeport In hourly data are normally distributed at
the 95 percent confidence level, the In hourly data are more normally
distributed than the hourly data at all sites, as indicated by larger W's.
Thus, for all sites, statistics generated from the In hourly data are more
valid than statistics generated from the hourly data.
      Table 3-3 shows the results of examining the In hourly Inlet data using
a first-order autoregressive, AR(1), model.  The objective of this analysis
was to evaluate underlying time series dependencies in the data.  If data are
autoregressive, it means that the hour-to-hour variations in S0g levels are
                                                                         mi

deviates from conventional statistics which assume that such dependencies do
not exist and, thus, that each observation 1s random.
      The longest consecutive segment of data was examined for each site.  As
can be seen from Table 3-3, the Bridgeport and Mi 11 bury data appear to come

bjw:033                                A3-2
not truly random and that prediction of the S0£ level in one time peribd jnust
consider the SOg level in the previous time period.  This relationship

-------
         TABLE 3-3.  TIME SERIES ANALYSIS FOR NATURAL LOG OF HOURLY
              INLET DATA (LONGEST CONSECUTIVE SEGMENT OF DATA)
-nEaB3E^BKBBB1^B— :— -aB=
Site
York County Unit 1

York County Unit 2

Bridgeport
Mill bury
\
—^=0==^
Number of
Hours

201

337
166
258
	 •
Correlation
Coefficient

0.61

0.38
0.56
0.60
Do Data Fit
AR(1) Model?

no

no
close
close
close to fitting the AR(1) model, but the correlation coefficients are  fairly
small, meaning that the Impact of autocorrelation on predicted emission levels
is small.  Hourly inlet data from York County do not fit the AR(1) model.
Because of the existence of only weak time dependencies within these three
data sets, the AR(1) model was not used in any of the subsequent inlet S02
analyses.
3.2  IMPACT OF AVERAGING TIMES
      As discussed  in Section 3.1, even though In hourly Inlet data were not
normally distributed at the 95 percent confidence level for all sites, this
data transformation resulted  in  data distributions that were closer to normal
than the original hourly  inlet data. For  lognormally distributed data, the
appropriate method  for  calculating average values is the geometric mean rather
than  the  arithmetic mean,  as  is  commonly used with normally distributed
data.2
      2The geometric mean  is defined as:
                           Geometric Mean - ( 8 X^) 'n
       H is  the  process  of  multiplying  each  nor
 W*i&^''*&}>*'9^*'*1»™  "c^be °calVulated  using  the
 transformation based on identities of natural logarithms.
                  [ in  ( ft Xj)1/"]  » exp [ 1/n In ( ft X^]  •- exp  [ 1/n 8 In X,]
 w
 bjw:
exp

                        /^
-------
      The effect of different averaging periods on the observed variability of
uncontrolled S02 levels was examined for the. In-transformed data at each of
the three MUC's.  Averaging times examined were 1-hour, 3-hour, 8-hour, and
24-hour block averages; and 1-day (equal to a 24-hour block), 3-day, and 7-day
rolling averages.  The 99th percentile inlet S02 values resulting from this
analysis are presented 1n Figure 3-1.  As can be seen, there 1s a significant
reduction in 99th percentile values between 1 and 24 hours, but a more gradual
reduction for averaging periods of greater than 24 hours.  Based on this
observation, a 24-hour block was selected as the appropriate averaging period
for further analysis.  The use of a 24-hour averaging period is also
consistent with the analysis of outlet SOg levels presented 1n Sections 4.2
and 5.2.
      The 24- hour block geometric mean was calculated for each day of record
with 18 or more hours of non-missing Inlet data as follows:

                                            Z (In
       24- hour block geometric mean  - exp[
                                                n
                i*.
where:      X^ -; hourly Inlet SOg concentration
            n  -  number of hourly SOg values for given 24-hour block (values
                  calculated only for days with 18 or more hours of data)

      Table 3-4 shows the U's for the 24-hour block means and the In 24-hour
block means of the S0£ inlet data for each site.  As with the hourly data, the
W's are larger for the In-transformed datasets than for the original datasets.
However, in contrast to the hourly data, all In-transformed datasets pass the
normality test and all original datasets but Bridgeport pass as well/  This
confirms that the In- transformed data sets are preferred for developing—-
predictive statistics.
      Figure 3-2 displays the geometric 24-hour block means for inlet- data at
Mi 11 bury, Bridgeport, and York County.  These plots show that the day-ito-jday
variability is significantly less than the hour-to-hour variability shown in
figures in Section 2.  The cumulative frequency distributions curves for the
geometric 24- hour block means at each site are shown 1n Figure 3-3.'  Though
the actual concentrations differ between the sites, the shapes of the curves
are rather similar.  Based on the fact that all of the Inlet geometric means

bju:033                                A3 -4

-------
500
400
300
200
100
      1-hr
3-day
7-day
                   3-hr       8-hr        24-hr
                      \     Averaging Period
      •  Bridgeport  +  Millbury   o  York County 1  *  York County 2

Figure 3-1.  Impact of Averaging Time on Uncontrolled SO2 Emissions
                             A3-5

-------
  400
  MO
I
41
jj aoo
  too
                                                                 Millbury
                10
                          20
                                     30         40

                                    Observation
                                                         80
                                                                    60
                                                                               70
250

240

230

220




200




160

170

160

180

140 •

130 >

120

110 •

100-
                                                                   Bridgeport
            •      • i •  • • •  i • • • '	I	 i i | i i i i  | i i  i i	, , !
            *°     20     30     40     BO     60     70     60      00     100     110

                  1                 Observation Nuatoer

             Figure 3-2.  24-Hour Geometric Mean Inlet  SO2 Data


                                       A3-6

-------
  tao

  180
  140

3 **>
«•
I
120


110


100


 90


 80


 70
        York County Unit 1
                            -I—
                             10
                                                     20
                                                                          —T
                                                                           30
                                  Observation Nurtitr
   140
   iao
   lao
   no
   100
    90
    80
    70
    80
         York County Unit 2
                              10
                                                      ao
                                                                             30
                                   Observation Nuobcr
         Figure 3-2. 24-Hour Geometric Mean Inlet SO2 Data (Continued)
                                         A3-7

-------
            TABLE 3-4.  SHAPIRO-WILK NORMALITY TEST STATISTIC  (W)
                        FOR  INLET 24-HOUR BLOCK MEANS
                                                                 York County
Inlet Dataset                       Bridgeport    Millbury    Unit 1    Unit 2
24-hour block mean
Ln 24-hour block mean
0.937
0.957*
0.969*
0.974*
0.951*
0.986*
0.936*
0.949*
* Data are normally distributed at the 95% confidence level.


are normally distributed and that the cumulative probability distributions
between sites are similar, it is expected that the Inlet S0£ 24-hour geometric
means from other MWC's are similarly distributed.  It is therefore expected
that the results of the analyses presented in Sections 4 and 5 for/FF and
SD/ESP systems will be applicable to other MWC's equipped with these APCD
systems.
3.3  CONCLUSIONS
      Based on the similarities in the inlet data between the three MWC's
examined, several general conclusions are reached:

      •     Maximum hourly S02 measurements can exceed the mean value at a
            given site by more than a factor of three.

      •     The In-transformed hourly inlet data from all three facilities
            were more normally distributed than original data for both 1-hour
            and for 24-hour blocks, suggesting that uncontrolled S02 emissions
            from MWC's in general are lognormally distributed.

      •     An AR(1) model of the hourly inlet data was not statistically
            significant and1 the correlation coefficients were relatively low.

      •     Using a 24-hour averaging period can significantly reduce the
            variability in uncontrolled S02 levels.  Use of longer averaging
            times has relatively less impact.  Use of shorter averaging times
            would result in frequent occurrences of high average S02 levels
            due to short-duration excursions in uncontrolled S02 levels.
bju:033                                A3-9

-------
       4.0  S02 CONTROL PERFORMANCE OF SPRAY DRYER/FABRIC FILTER SYSTEMS

      The performance  capability of SOg control technologies can be defined by
the percent reduction  in S02 levels between the inlet and outlet of the
control device.   For MUC's with SD/FF  systems, two sets  of inlet and outlet
CEM data were  available:  Bridgeport and York County.  The permitted S0£ limit
at Bridgeport  was based on a relatively high outlet S02  level,  and thus did
not require a  high level of continuous S02  reduction.  Therefore, York County
provided the primary data set  used to  analyze the performance capability of
SD/FF systems.  Although the primary focus  of this section is on percent
reduction, some analysis of outlet S02 concentrations is also presented.
      The Stanislaus County data,  although  having low outlet S02 levels, do
not include uncontrolled (i.e., inlet) S02  levels, and,  thus, cannot be used
to define continuously achievable  SOg  percent reductions.  However, based on
the mean measured outlet S02 levels of 7-9  ppm  (see Table 2-4)  and earlier
compliance testing of  the plant,  it appears that the average percent
reductions were about  90 percent  or greater.
4.1  SHORT-TERM VARIABILITY
      The figures and  tables presented in Section 2.3.2  demonstrate the
variability of the percent reduction  data at York County.  Although the
average percent reductions for both units are near 90 percent and the medians
.are slightly greater than 90 percent,  there are short-duration  spikes of less
than 20 percent on each unit.
      A common convention in analysis  of percent reduction data is to
calculate emissivity,  which  is defined as:
      Emissivity  - 100 - Percent  Reduction
Table 4-1 shows the Shapiro-Wilk  normality  statistic (W) for the hourly
emissivity data and the In-transformed data for York County Units 1 and 2.
Though none of the In  hourly emissivity data are normally distributed at the
95 percent confidence  level, the  In hourly  data are more normally distributed
than the  hourly data,  as indicated by larger Vl's.  Thus, statistics generated
from the  In hourly data will be more  valid  than statistics generated from the
hourly data.      *•"•
      An  AR(1) time series model  was  used to examine for time-based
autocorrelations  in the In hourly outlet data at York County.   The longest

bJH:033                                 A4-1

-------
          TABLE 4-1.  SHAPIRO-WILK NORMALITY TEST STATISTIC  (W) FOR
                    HOURLY EMISSIVITY DATA AT YORK COUNTY*

                                            .Units 1 and 2
 Inlet Dataset             Unit 1              Unit 2             Combined
Hourly
Ln Hourly
0.841
0.855
0.672
0.953
0.759
0.923
* No datasets are normally distributed at the 95% confidence level.


consecutive segment of data was examined for each unit.  Table 4-2 shows the
number of hours, the correlation coefficients, and a short response (yes, no,
or close) for whether the model fits.  As can be seen from Table 4-2, the
outlet S02 data from neither unit appears to fit the AR(1) model and the
correlation coefficients are small.  As discussed in Section 3.1, the inlet
SOe data at York County also did not fit an AR(1) model.  Therefore, the AR(1)
model was not used to analyze the York County emissivity data.
          TABLE 4-2.   TIME SERIES ANALYSIS FOR NATURAL LOG OF HOURLY
               OUTLET DATA (LONGEST CONSECUTIVE SEGMENT OF DATA)
Site
York County Unit 1
York County Unit 2
Number of
Hours
172
190
Correlation
Coefficient
0.45
0.27
Do Data Fit
AR(1) Model?
no
no
4.2   IMPACT OF AVERAGING TIMES
      As discussed  in  Section 4.1,  even though the  In hourly emissivity data
were  not normally distributed at  the  95 percent confidence level, this data
transformation resulted in data distributions that  were closer to normal than
the original  hourly emissivity data.   As  a  result,  the geometric mean is a
more  appropriate statistic to use for analysis of the data.
 bju:033
                                      A4-2

-------
      An examination of the effect of averaging times on the variability in
S02 emissivity was conducted similarly to that presented in Section 3.2.
Based on the lognormal distribution of the hourly emissivity data, geometric
means were used to calculate average emissivity values.  The results are shown
in Figure 4-1.  Note that the projected continuously achievable emissivity
level (based on an exceedance frequency of one per year) decreases from
roughly 35 percent (65 percent reduction) based on a 3-hour block geometric
mean to 19 percent (81 percent reduction) based on a 24-hour block geometric
mean.  If the averaging time was increased to 7 days based on a daily rolling
average, the continuously achievable emissivity 1s estimated at 13 percent.
Based on the almost 50 percent reduction in emissivity level at 24 hours
versus 3 hours, versus the smaller decrease resulting from increasing the
averaging time to 7 days, it was determined*that a 24-hour block was the most
appropriate averaging period.  At this averaging period, "high levels of
continuous S02 removal are needed, while providing allowance for the short-
term variability In MSW sulfur content and SD/FF operation that can result in
higher emissions.
      Table 4-3 shows the U's for the 24-hour arithmetic block means and the
24-hour block geometric means for the emissivity data from York County Units 1
and 2.  The 24-hour block geometric means for both units were normally
distributed at the 95 percent confidence level and the W's are greater for the
geometric means than for the arithmetic means.  This supports use of geometric
means rather than arithmetic means for predicting expected minimum daily
percent reduction levels.
           TABLE 4-3.   SHAPIRO-WILK NORMALITY TEST STATISTIC (W)  FOR
                 EMISSIVITY ^24-HOUR BLOCK MEANS AT YORK COUNTY
Inlet Dataset                           Unit 1               Unit 2
24-hour block arithmetic mean
24-hour block geometric mean
0.913*
0.918*
0.907
0.939*
* Data are normally distributed at the 95% confidence level.


bju:033                                A4-3

-------
   40
   38
   36
   34
   32
   30
   28
   26
   24
   22
   20
   18
   16
   14
   12
   10
        3-hr
8-hr          24-hr
     Averaging Period
3-day
7-day
Figure 4-1. Impact of Averaging Time on SO2 Emissivity at York County
                                A4-4

-------
       Figure  4-2 displays  the  Individual  24-hour  block geometric means of
emisslvity  for  York  County Units  1  and  2.  The missing data points  are for
days during which  hourly data  were  not  available  for  7 or more hours due to
the unit  being  off-line or other  factors.  As expected, the day-to-day
variability in  emissivity  is less than  the hour-to-hour variability shown in
Figure 2-3.   The lowest percent removal  (corresponding to the highest
emissivity) encountered during these  45 days of data  was roughly 84 percent.
       Table 4-4 presents the geometric  means, standard deviations,  and  -
calculated  once-per-year exceedance values based  on the emissivity  data from
each of the York County units  and for the two units combined.  The  mean
emissivity  for  both  York County units is near 8 percent, which translates to a
percent reduction  of near  92 percent.   The ance-per-year emissivity exceedance
values for  Units 1 and 2 are 16 and 21  percent, respectively.  Given the
limited number  of  data points  for each  unit  (22 for Unit 1 and 23 for Unit 2),
the two data  sets  were combined to  reduce the Impact  of random variation in a
small  data  set  on  calculated values.  Combining the data sets for Units 1
and 2  is  considered  valid  based on  the  fact that  both units fire wastes from
the same  MSW  collection pit (i.e.,  they have similar  variability in
uncontrolled  S02 levels) and have similar SD/FF designs.  Using the combined
data sets,  the  once-per-year maximum  24-hour geometric mean emissivity value
is 19  percent,  corresponding to a percent reduction value of 81 percent.
             TABLE 4-4.  SUMMARY STATISTICS FOR GEOMETRIC MEANS OF
                      EMISSIVITY (PERCENT) AT YORK COUNTY
Statistic
Unit 1
Unit 2
Units 1 and 2
   Combined
Number of blocks
Mean
Standard deviation
Continuous compliance level*
22
8.02
4.65
20.95
23
7.90
2.89
15.94
45
7.96
3.81
18.55
* Based on one exceedance per year (Mean + 2.777*Standard Deviation)
bjwi033
       A4-5

-------
  16
  IB-
  14-
  13-
  12-
1-
i-
»  9
41
••
a  •
I*
   e-
   a
   4'
   3'
   2
                                 Unttl
                          10
                                                                       30
                               Ob«*pv*tlon Nuabcr
  14-

  13-

  12-

  11-

  10-

   9-
   •'
   6-

   a-

   4-

   3-
                                                             Unit 2
10                     20
     Observation Niwbcr
                                                                       30
      Figure 4-2. York County 24-Hour Geometric Mean Emissivfty Values
                                    A4-6

-------
4.3  CONCLUSIONS
      Based on the York County and Stanislaus County data, the following
conclusions summarize the S02 control performance of SD/FF systems:

      •     Although average S02 reductions of 90 percent are achievable over
            an extended period, short-duration spikes in hourly percent
            reduction levels are such that these levels cannot be achieved
            continuously due to variations in inlet S0£ levels and normal
            variations in SD/FF performance.

      •     Hourly In-transformed emissivlty data approach a normal
            distribution.  This means that geometric means (rather than the
            more traditionally used arithmetic means) are appropriate for
            assessing achievable levels of S02 control.
                                            *

      •     The.variability in hourly inlet outlet data did not fit an AR(1)
            time series model and the correlation coefficients were
            small.Therefore, consideration of time dependencies was not used
            in estimating achievable performance levels.

      •     Based on analysis of the impact of averaging periods on the
            variability in S0£ emissions, it was concluded that a 24-hour
            averaging period was the most appropriate averaging period.  Use
            of a shorter averaging time results in reducing the level of
            continuously achievable reductions, while longer averaging periods
            had less of an impact.

      •     The 24-hour geometric mean emissivlty levels are normally
            distributed.  Based on the available data and resulting
            statistics, a 24-hour block geometric mean S02 percent reduction
            level of 80 percent is achievable from SD/FF systems on a
            continuous basis.
bjn:033                                A4-7

-------
            5.0  S02 CONTROL PERFORMANCE  OF SPRAY DRYER/ESP SYSTEMS

      This section describes the achievability of S0£ emission reductions at
MWC's equipped with SD/ESP systems.  This analysis Is based on the S0£ data
from the Ml11bury MWC presented In Section 2.1.  Although the primary focus of
this section Is on percent reduction, some analysis of outlet S0£ levels Is
also presented.
5.1  SHORT-TERM VARIABILITY
      The figures and tables presented in Section 2.1.2 demonstrate the
variability of hourly S02 data at Ml11 bury.  Although the average percent
reduction Is roughly 80 percent, there are short-duration spikes below 50
percent.                                     ,
      As with the York County data, percent reduction levels were converted to
emissivity levels to aid in the analysis.  Table 5-1 shows the Shapiro-Wilk
normality statistic (W) for the hourly emissivity data and the In hourly
emissivity data from Millbury.  Though the In hourly data are not normally
distributed at the 95 percent confidence level, the In hourly data are more
normally distributed than the hourly data, as indicated by a larger W.  Thus,
statistics generated from the In hourly data will be more reliable than
statistics generated from the hourly data.

       TABLE 5-1.  SHAPIRO-WILK NORMALITY TEST STATISTIC (W) FOR HOURLY
                         EMISSIVITY DATA AT MILLBURY*
       Inlet Dataset                                         Unit 1

        Hourly                                               0.846
        Ln hourly                                            0.972


* No datasets are normally distributed at the-95% confidence level.


      An AR(1) time series model was used to examine the In hourly outlet data
at Mi 11 bury for time-based autocorrelations.  Table 5-2 shows the number of
hours, correlation coefficients, and a short response (yes, no, or close) for

bj«i033                                A5-1

-------
whether the model fits.  As can be seen from this table, the Mi 11 bury outlet
data do not fit the AR(1) model and the correlation coefficient is fairly
small.  As discussed in Section 3.1, the Mi 11bury $02 inlet data approached
being AR(1), but did not satisfy the model criteria.  Therefore, an AR(1)
model was not used to analyze the Mill bury emissivity data.

          TABLE 5-2.  TIME SERIES ANALYSIS FOR NATURAL LOG OF HOURLY
               OUTLET DATA (LONGEST CONSECUTIVE SEGMENT  OF  DATA)
Site
Mi 11 bury
Number of
Hours
258
Correlation
Coefficient
0.56
Do Data Fit
AR(1) Model?
no
5.2  IMPACT OF AVERAGING TIMES
      As in Sections 3.2 and 4.2, several averaging times were examined in an
attempt to reduce the data variability.  Because of the loghormal distribution
of the hourly data, geometric means were used to calculate average
emissivities over each time period.  The results of this analysis are plotted
in Figure 5-1.  Note that the projected continuously achievable emissivity
level (based on an exceedance frequency of one per year) decreases from
roughly 45 percent (55 percent reduction) based on a 3-hour block geometric
mean to 30 percent (70 percent reduction) based on a 24-hour block geometric
mean.  If the averaging time is increased to 7 days based on a dally rolling
average, the continuously achievable emissivity 1s estimated at 25 percent.
The lower SO? removal by SD/ESP versus SD/FF systems (as discussed in Section
4.2) is caused by the reduced level of S02 control achieved by an ESP versus
the filter cake in a FF.  Based on the lower emissivity level for 24 hours
versus 3 hours, and the relatively small decrease 1n emissivity resulting from
Increasing the averaging time to 7 .days, it was determined that a 24-hour
                                                                     1   I
block was also the most appropriate averaging period for SD/ESP systems:
      Table 5-3 shows the W normality statistics for the 24-hour block
arithmetic and geometric means for the Mill bury emissivity data.  The 24-hour
block geometric mean is normally distributed at the 95 percent confidence
level and the W's are greater for the geometric means than for the arithmetic

bjwi033                                A5-2

-------
  50
  48
  46
  44
  42
  40
  38
  36
  34
  32
  30
  28
  26
  24
  22
  20
       3-hr
8-hr          24-hr
     Averaging Period
3-day
7-day
Figure 5-1.  Impact of Averaging Time on SO2 Emissivity at Millbury
                               A5-3

-------
            TABLE 5-3.  SHAPIRO-WILK NORMALITY TEST STATISTIC (W)
          FOR 24-HOUR BLOCK MEANS OF THE EMISSIVITY DATE AT MILLBURY
        Inlet Dataset                                         W

      24-hour block arithmetic mean                         0.914
                                                          •
      24-hour block geometric mean                          0.967*


* Data are normally distributed at the 95% confidence level.
means.  This supports use of geometric means rather than arithmetic means for
predicting expected exceedance levels.
      Figure 5-2 displays the Individual 24-hour block geometric means of
emissivity for Mi 11 bury.  The missing data points are for days during which
hourly data were not available for 7 or more hours due to the unit being off-
line or other factors.  As expected, the day-to-day variability in emissivity
is less than the hour-to-hour variability shown in Figure 2-1.  The lowest
percent removal (corresponding to the highest emissivity) encountered during
these 53 days of data was roughly 70 percent.
      Table 5-4 presents the geometric means, standard deviations, and
calculated once per year exceedance values based on the emissivity data.  The
mean emissivity was near 19 percent, which translates to a percent reduction
of 81 percent.  The once-per-year exceedance emissivity value 1s estimated at
30 percent, corresponding to a percent reduction of 70 percent.
5.3  CONCLUSIONS
      Based on the Mi 11 bury data, the following conclusions summarize the
achievable S02 control performance of SD/ESP systems:

      •     Although average SOg reductions of over 80 percent are achievable
            over an extended period, short-duration spikes in hourly percent
            reduction levels are such that these levels cannot be achieved
            continuously due to variations in inlet S02 levels and normal
            variations in SD/ESP performance.
                                     A5-4

-------
  40-
u
a
  10-
                                               J
10
         20
                                 30        40
                                Observation Nuabw
                                                    BO
                                                             80
                                                                       70
         Figure 5-2.  Millbury 24-Hour Geometric Mean Emissivrty Values
                               A5-5

-------
         TABLE  5-4.   SUMMARY  STATISTICS AND  EXCEEDANCES FOR GEOMETRIC
         24-HOUR BLOCK MEANS OF EMISSIVITY DATA (PERCENT) AT MILLBURY
             Statistic                                      Value


          Number of blocks                                   53
          Mean                                               19.2
          Standard deviation                                  3.9
          Continuous compliance level*                       30.1

* Based on one exceedence per year (Mean + 2.777 * Standard Deviation)
            Hourly In-transformed emissivity data approach a normal
            distribution.  This means that geometric means are appropriate for
            assessing achievable levels of SOg control.

            The variability in hourly Inlet and outlet data did not fit an
            AR(1) time series model and the correlation coefficients were
            small.  Therefore, consideration of time dependencies was not used
            in estimating achievable performance levels.

            Based on analysis of the impact of averaging periods on the
            variability in S02 emissions, it was concluded that a 24-hour
            averaging period was the most appropriate averaging period.  Use
            of a shorter averaging time results in reducing the level of
            continuously achievable reductions, while longer averaging periods
            had relatively less of an impact on achievable levels.

            The 24-hour geometric mean emissivity levels are normally
            distributed.  Based on the available data and resulting
            statistics, a 24-hour block geometric mean S02 percent reduction
            level of 70 percent is achievable from SD/ESP systems on a
            continuous basis.
bjw:033                                A5-6

-------
               APPENDIX B

ANALYSIS OF THE CONTINUOUS NOX EMISSIONS
  DATA FROM MUNICIPAL WASTE COMBUSTORS

-------
                               TABLE OF CONTENTS
1.0   Overview	     Bl-1
      1.1   Purpose	     Bl-1
      1.2   Description of Data Analyzed  	    Bl-1
      1.3   Summary  of Statistical Methods   	     Bl-2
2.0   Selective Non-Catalytic Reduction 	     B2-1
      2.1   Description of Stanislaus County MWC   	     B2-1
      2.2   Data Analysis	     B2-5
            2.2.1  Hourly Emission Data .  .	     B2-5
            2.2.2  Daily Emissions Data .  .  .'	     B2-6
      2.3   Conclusions . .  . *	     B2-10
      2.4   References	     B2-10
3.0   Rotary Waterwall Combustor   	     B3-1
      3.1   Description of York County MWC   	     B3-1
      3.2   Data Analysis	     B3-4
      3.3   Conclusions	     B3-7
bjw:033                                  81

-------
                                  1.0   OVERVIEW

 1.1   PURPOSE
       On December 20, 1989, the U.S.  Environmental Protection Agency (EPA)
 proposed limits on emissions of nitrogen oxides (NOX) from new municipal waste
 combustors (MWC's) under Section 111  of the Clean Air Act (54 FR 52209).  As
 proposed, these new source performance standards would apply to new MWC's
 having the capacity to combust more than 250 tons per day (tpd) of municipal
 solid waste (MSW).  The proposal indicated that the Agency was considering a
 NOX emission standard of 120-200 ppmv corrected to 7 percent oxygen, and that
 the final limit would be based on further analyses conducted prior to
 promulgation.  This report presents the results of statistical analyses of NOX
 data obtained from a grate-fired mass burn water-wall MWC using selective
 noncatalytic reduction (SNCR) to reduce NOX emissions and from a rotary mass
 burn waterwall MWC designed to limit NOX emissions through combustion control.
 1.2   DESCRIPTION OF DATA ANALYZED
,j.-,    The two data sets used in this analysis are from the Stanislaus County
 MWC in Crows Landing, California, and the York County MWC in Manchester
 Township,. Pennsylvania.
       The Stanislaus County MWC consists of two 400-tpd grate-fired mass burn
 waterwall combustors.  Both units are equipped with Exxon's Thermal DeNOx*
 SNCR process in which ammonia is injected to reduce NOX emissions.  Hourly NOX
 data were available from both units for two months in the Spring of 1989.
 Daily average NOX levels were also available from mid-August, 1989, to mid-
 March, 1990.
       The York County MWC consists of three Westinghouse/0'Connor rotary mass
 burn waterwall combustors, each with a capacity of 448 tpd.  Each of the York
 County units uses a water-cooled rotary combustion chamber that is designed
 and operated to control combustion conditions so as to inhibit NOX formation.
 No supplemental control technology is used to reduce NOX emissions below the
 levels formed in the combustor.  The York County NOX data were limited to one
 month of hourly NOX emissions from all three units recorded in February, 1990.
 bjw:033                                Bl-1

-------
1.3   SUMMARY OF STATISTICAL METHODS
      Several statistical analysis techniques were used to characterize MWC
NOx emissions.  These techniques  Included:

      •     time series plots to  visually examine trends 1n NOx emissions over
            time and to Identify  data gaps or anomalies;

      •     normality testing and cumulative frequency distribution plots to
            evaluate the distribution of NOx data;

      •     routine summary statistics (mean, median, standard deviation,
            etc.);

      •     maximum estimated emissions (referred to as "exceedance values"}
            based on the appropriate statistical means and standard deviations
            for different averaging times.

      The hourly and dally NOX data were tested for normality using the
Shapiro-Wilk statistic.  These analyses were conducted using both the actual
NOX data (after correction to 12  percent C0£ or 7 percent 03) and the natural
logarithms (In) of the actual data.
bjw:033                                Bl-2

-------
                    2.0  SELECTIVE NON-CATALYTIC REDUCTION

      The assessment of the NOX emissions control performance of SNCR
technology was based on data obtained from the Stanislaus County MWC.  Section
2.1 describes the facility and the NOX emissions data analyzed.  Section 2.2
summarizes the results of the data analysis.  Section 2.3 presents the
conclusions regarding the continuous compliance capabilities of SNCR.
2.1   DESCRIPTION OF STANISLAUS COUNTY MWC
      The Stanislaus County MWC, developed by Ogden Martin Systems, Inc.,
consists of two 400-TPD Martin GmbH grate-fired mass burn water-wall
combustors.  A Thermal DeNOx* system, Installed under license from Exxon
Research & Engineering Company, Injects ammoftia into the upper furnace of each
combustor to reduce NOX emissions.  Each unit is also equipped with a Flakt
spray dryer and fabric filter system to reduce acid gas, particulate, trace
metals, and organic emissions.  At the time the data was collected, the
facility's permit allowed a maximum NOX emission level of 165 ppm over a 24-
hour averaging period and 175 ppm over a 3-hour averaging period, both of
which are corrected to 12 percent C02>
      Plant construction was completed during the late Summer of 1988.
Initial compliance testing was conducted between November 29 and December 10,
1988.  The hourly NOX data used in this analysis was collected between March
16 and May 14, 1989.  The total number of hourly observations collected during
this period was 920 hours for Unit 1 and 959 hours for Unit 2.  Daily
arithmetic averages were also obtained for the period between August 17, 1989,
and March 12, 1990.  The total number of daily observations, including the
hourly data collected during the Spring of 1989, was 223 days for Unit 1 and
228 days for Unit 2.  Data on the flue gas Og content were not available from
the plant.  Therefore, the NOX data were corrected to 12 percent C02> which is
roughly equivalent to correction to 7 percent 03.
      The hourly and daily average NOX emissions from both units are shown in
Figures 2-1 and 2-2, respectively.  The daily averages between March 16 and
May 14, 1989, are plotted only for those days in which 18 or more hours of
data were available.  The data capture criteria associated with the other
bjw:033                                B2-1

-------
                                                                        Unrtl
o-
0    100   ZOO   300  400   900
                                BOO   700   BOO   900

                                  Observation Ifciabsr
                                                          i     i      i     i      i     r
                                                        1000  1100  taOO  1300  1400  1900
                                                                         Unit 2
       100    ZOO    300
                              900   600
                                          TOO
                                                         —I-	1	1	1	T
                                                          looo   1100   laoo   1300  1400
                                   OMcrvatlon Nuae«r

           Rgure 2-1. Stanislaus County Hourly NOx Concentrations
                                        B2-2

-------
  1M
  1BO
  140
  130
  tao
  too
   90
   M
   170
   160
                                                                            Unfti
                                 100
                                                              200
                                                                                          300
                                        Observation
   ISO
1
140
   120
   110
   100
   90
                                                                            Unit 2
                                                                MwditMO
                                  100
                                                           200
300
                                         ObMrvatlon
          figure 2-2.  Stanislaus County Daily Average NOx Concentrations
                                              B2-3

-------
dally averages are not known and the average NOX values are presented as
obtained from the plant.
      Review of the dally average NOX values, shown in Figure 2-2, suggests
three distinct trends in the data.  During the first of these periods, from
March 16 to May 14, 1989, there is more day-to-day variability in the data
than during subsequent time periods and the daily average NOX levels exhibit
an apparent downward trend over this time period, especially on Unit 2.
During the second of these time segments, extending from August 17 to October
30, 1989, there is less day-to-day variability in the data compared to the
first time segment, but there is an upward trend In the dally average values.
During the third time segment, the dally averages also exhibit relatively low
variability and have relatively stable values over time.
      There are several possible explanations for these variations.  First,
the apparent trends in daily average NOX levels may be due to differences in
MSW composition.  Specifically, the high nitrogen content of vegetative wastes
present in MSW can affect NOX emissions.  Based on a previous study,* however,
it would be expected that the NOX levels would be higher in the summer and
lower in the winter, which 1s the opposite of the trends observed in the
Stanislaus County data.  This may be due to the dry summers and wet winters in
the San Joaquin Valley of California, where the plant 1s located, that result
in seasonal variations in waste composition that are different from many other
areas of the country.  The yard waste composting program implemented within
the plant's waste collection service area was also reported to have changed
during the time period covered by the data.2
      Second, the larger apparent variability in daily NOX levels between
March 16 and May 14, 1989, may reflect the plant personnel's lesser
familiarity with how to best operate the combustor and ammonia Injection
system to minimize NOX emissions during the first few months of the plant
operation.  The Initial Thermal DeNox* process control system was also being
evaluated by the plant during this time period based on actual operating
experience.2  The reduced variability of NOX emissions during later periods
may reflect Increased knowledge of how to best operate the plant and
modifications in combustor or ammonia Injection system design.  This Increased
knowledge of how to operate the combustor and ammonia Injection system to

bjw:033                                B2-4

-------
control NOX may also account for the small variability 1n dally NOX levels
during the third time segment.
      Third, the differences 1n daily NOX levels may reflect differences in
data averaging methods used by Ogden during the last two time segments versus
those used by EPA during the first time segment.
2.2   DATA ANALYSIS
      Two sets of statistical analyses were conducted—one with the hourly
data and the second with the daily data.  The objectives of these analyses
were to determine the statistical distribution of the data (I.e., normal
versus lognormal), the affect of averaging period on the variability in NOX
emissions, and the selection of a continuously achievable NOX emission level.
2.2.1  Hourly Emissions Data
      The hourly data consisted of the period between March 16 and May 14,
1989.  Summary statistics for this period for Units 1 and 2 are presented 1n
Table 2-1.  Note that the mean (average) and median (50th percentile) values
on both units are between 120 and 130 ppm and that the mean and median are
both relatively close to the midpoint of the minimum and maximum recorded
values.  This suggests that both data sets are normally, rather than
lognormally, distributed.  As shown in Table 2-2, this conclusion is supported
by the Shapiro-Wilk normality test that found the measured data from both
units to be normally distributed at the 95 percent confidence level, while the
In-transformed data were not normally distributed at the 95 percent confidence
level.  This finding supports the use of the raw data, rather than""
In-transformed data, for estimating potential emission exceedance levels.
              TABLE  2-1.  SUMMARY  STATISTICS  FOR HOURLY NOX DATA
                          FROM  STANISLAUS. COUNTY (ppm @ 12% C02)
Summary
Statistic
Mean
Minimum
Median
Maximum
Unit 1
124.1
20.2
126.4
218.4
Unit 2
126.3
31.9
127.9
245.6
 bjw:033                                 B2-5

-------
          TABLE 2-2.   SHAPIRO-UILK NORMALITY  TEST  STATISTIC  (W)  FOR
                      HOURLY NOX DATA FROM STANISLAUS COUNTY
Dataset                               Unit 1                  Unit 2

Hourly0.98*0.99*
Ln Hourly                              0.92                    0.94
^Indicates a normal data distribution at the 95% confidence level.
      To assess the Impact of different averaging times on the variability in
NOX emission levels, six different averaging periods were examined:  1-hour,
3-hour, 8-hour, and 24-hour block averages, and 3-day and 7-day rolling
averages.  As shown In Figure 2-3, Increasing the averaging time from 1 hour
up to 8 hours resulted In about a 10 percent decrease in estimated 99th
percent!le NOX emission level.  Increasing the averaging time to 24 hours
decreased the 99th percentlie value by roughly another 15 percent.  Further
increases in the averaging time to 3 days and 7 days resulted in a further
reduction in the estimated NOX emission level, but the change was smaller than
that achieved between 8 and 24 hours.  As a result, a 24-hour block (i.e.,
daily) averaging period was selected for further analyses.
      The small change in estimated NOX levels between 1 and 8 hours suggests
that hourly NOx levels are autocorrelated, meaning that the NOX level in one
hour is a function of the NOX level during the previous hour.  As a result,
averaging periods of 8 hours and less have limited impact on reducing the
variability in NOX emissions.  By increasing the averaging period to 24 hours,
however, it Is possible to reduce the variability in estimated NOX emissions.
      The 24-hour block averages calculated with the Harch through Nay, 1989,
data were used to assess the continuously achievable NOX emission level.  As
shown in Table 2-3, this analysis found that a NOX emission limit of 165-180
ppm could be continuously achieved at an exceedance frequency of one per year.
2.2.2   Daily Emissions Data
      The other two periods of daily NOX data (August to October 1989 and
November 1989 to March 1990) were also examined to assess continuously

bjw:033                                B2-6

-------
   I   *»
   i
    w
    *>


   UJ
   
-------
             TABLE 2-3.   SUMMARY STATISTICS  FOR  DAILY NOX AVERAGES
                         FROM STANISLAUS COUNTY (ppm 9 12% C02)


Statistic
Number of Blocks
Mean
Standard Deviation
Continuous
Compliance Level*
ii
March-May 1989
Unit 1
34
121.2
13.6

163.5


Unit 2
38
125.3
17.4

179.5
*Based on one exceedance per year (Mean + 2.777 * Standard Deviation)
achievable NOX levels and the possible Impact of seasonal variations In waste
composition on NOX levels.  As discussed in Section 2.1, however, the average
NOX emission level in the March through May, 1989, data set appears to
decrease over time, and the average NOX emission level 1n the August through
October, 1989, data set appears to Increase over time.  This variation Is
between a low of about 105 ppm 1n August and a high of 130 ppm in late
October, or a difference of roughly 20 percent.  In theory, some of the
variability could be statistically compensated for by using a time series
model.  However, because of the limited number of dally observations within
the March-May and August-October time blocks and the significant number of
missing daily averages, use of a time series model was not possible.
      As a result, the additional analysis of continuously achievable NOX
levels focused on the data collected between November 1989 and March 1990.  As
shown in Figure 2-2, NOX emissions during this period varied from day to day,
but did not exhibit any general upward or downward trend.  An analysis of the
statistical distribution of the data on both units for this period, shown in
Table 2-4, found that neither the raw data or the In-transformed data were
normally distributed at the 95 percent confidence level.  The raw data on both
units do have higher Shapiro-Milk normality statistics, however, Indicating
that predictive statistics based on the raw data will be more valid than

bjw:033                                B2-8

-------
          TABLE 2-4.  SHAPIRO-WILK NORMALITY TEST STATISTIC (W) FOR
                       DAILY NOX DATA FROM STANISLAUS  COUNTY*

 DatasetUnit 1                   Unit  2

 Hourly                                 (T93                    0.92
 Ln  Hourly                              0.91                    0.90
 *for November  1989  to  March  1990.
 NOTE:   None  of the  data  sets were  normally  distributed  at  the  95% confidence
        1evel.
statistics  based  on  the  In-transformed data.   As  shown  in Table  2-5,
thepredicted continuous  compliance  levels  based on the daily  average NOX
emission levels for  November  1989 through  March 1990  are 148  ppm  for Unit  1
and  150 ppm for Unit 2.
             TABLE 2-5.  SUMMARY STATISTICS FOR DAILY NOX AVERAGES
                          FROM STANISLAUS COUNTY  (ppm 0  12% C02)

                                       "Nov.  1989-March  1990
Statistic                          Unit 1                     Unit 2

Number of Blocks124122
Mean                                 128.9                     131.5
Standard Deviation                     7.0                       6.6
Continuous
  Compliance Level*                  148.3                     149.8
*Based on one exceedance per year (Mean + 2.777 * Standard Deviation)
bjw:033                                B2-9

-------
 2.3    CONCLUSIONS
       Based on  the Stanislaus  County data,  the  following conclusions are
 reached  regarding the continuously  achievable NOX emission level  associated
 with  SNCR:

       •      Long-term average  NOX emissions of  less than 130 ppm  at 12 percent
             C02 are achievable,  but short-term  variations in NOX  emissions due
             to  changes in HSM  composition,  combustion conditions, and SNCR
             operation,  influence the NOX emission rate that can be
             continuously achieved.

       •      The hourly NOX data  appear to be normally distributed, thus
             supporting the use of the raw (I.e., untransformed) data to
             estimate continuously achievable emission levels.

       •      An  averaging period  of  24 hours 1s  beneficial in reducing the
             variability in average  NOX levels.  Averaging periods of 8 hours
             and less  have relatively little impact on lowering the variability
             In  average  NOX emissions.  Longer averaging times also have a
             lower  impact than  the difference between 8 and 24 hours.

       •      The daily average  NOX emission  rates suggest differences in
             achievable  NOX levels that are  different for the three periods.
             These  variations may be due to  seasonal variations in MSW
             composition, the familiarity of plant operating personnel with
             operation  of the combustion and SNCR systems, or other factors.
             The Stanislaus County data contained three such time blocks.

       •      The time  block with  the greatest variability 1n NOX emissions at
             Stanislaus  County  occurred between mid-March and mid-Nay, 1989.
             Based  on  the data  from this time period, a NOX emission level of
             180 ppm at  12 percent C0£ with  24 hour-block averaging period can
             be  continuously achieved.
2.4   REFERENCES
1.    Hahn, J. L., and D. S. Sofaer, Variability of NOX Emissions from Modern
      Mass-fired Resource Recovery Facilities.  Paper presented at 81st Annual
      Meeting of Air Pollution Control Association, Dallas, Texas.  June 19-
      24, 1988.
bjw:033                                B2-10

-------
 2.     Telecon.  Hahn, J. Ogden Martin Systems, Inc., and D. White,  Radian
       Corporation.  November 19, 1990.  Stanislaus County NOX data.
bjw:033                                B2-11

-------
                        3.0  ROTARY WATERWALL COMBUSTOR

      The assessment of the NOX emissions control in a Westinghouse/0'Connor
rotary mass burn waterwall combustor was based on data obtained from the York
County MWC.  Section 3.1 describes the facility and the NOX emissions data
analyzed.  Section 3.2 summarizes the results of the data analysis.  Section
3.3 presents the conclusions reached regarding continuous compliance
capabilities of the Westinghouse/0'Connor combustor.
3.1   DESCRIPTION OF YORK COUNTY MWC
      The York County MWC consists of three 448-TPD Westinghouse/0'Connor
water-cooled rotary mass burn waterwall combustors supplied by the Resource
                                             r i
Energy Systems Division of Westinghouse Electric Corporation.  The
Westinghouse/0'Connor combustors at York County are designed to inhibit
formation of both thermal and fuel-related NOX by controlling combustion
temperatures and fuel-air mixing 1n the rotary chamber.  No supplemental
control technology is used to reduce NOX emissions.  Each unit is also
equipped with a Joy Technologies, Inc., spray dryer and fabric filter system
to reduce add gas, particulate, trace metals, and organics emissions.  The
facility does not have a permitted NOX emission limit.
      Plant construction was completed in late 1989 with start-up of the unit
in November 1989.  The hourly NOX emissions data used in this analysis were
collected between February 1 and February 24, 1990, and were corrected to 7
percent 02 by the plant.  The total  number of hourly observations were 538
hours for Unit 1, 558 hours for Unit 2, and 547 hours for Unit 3. The hourly
and dally average NOX emissions from all three units are shown in Figures 3-1
and 3-2, respectively.
      Note that the hourly NOX emission levels from all three units are
between 60 and 200 ppm, except for a two-hour period on February 20 from
Unit 3.  During this time period the unit was in the process of stopping waste
feed and starting oil firing.  The flue gas Og level during this transition
period was high and resulted in a large adjustment in NOX emission level when
corrected to 7 percent 0£.  Although these two data points could have been
deleted based on atypical operation during shutdown of the unit, they were
left in the data set that was used for analysis.

bjw:033                                B3-1

-------
r
  110




  100
                                                 Unitl
            too
                                                 Unit 2
 1-
 *
                                                    Unit3
                               100
                          OOMTVItlMI I
    Rgura 3-1. York County Hourly NOx Concentrations



                            B3-2

-------
                                             Unttl
                                              Unit 2
1-
 1U
                                               Untt3
                     M
     Rgure 3-2. York County Daily Average NOx Concentrations
                         B3-3

-------
3.2   DATA ANALYSIS
      The objectives of the statistical analysis conducted with the York
County data were:  to determine the distribution of the data, to assess the
Impact of averaging time on variability 1n NOX emissions, and to select a
continuously achievable NOX emissions level.  Summary statistics for all three
units are presented In Table 3-1.  Note that the mean and median values on all
three units are between 108 and 115 ppm.  These values are roughly at the
midpoint of the minimum and maximum NOX readings, except for the previously
noted high NOX levels recorded on Unit 3.  This suggests that the data are
normally distributed.  This 1s supported by the results of the Shap1ro-Wilk
normality test, shown 1n Table 3-2, that found the hourly NOX levels on Units
1 and 2 to be normally distributed at the 95 percent confidence level.  The
In-transformed NOX levels for Units 2 and 3 are also normally distributed at
the 95 percent confidence level.  The higher normality test statistic for
Unit 3 reflects the Impact of the few hours of high NOX levels recorded on
that unit.  If these data were deleted from the data set, the remaining data
would appear to be more normally distributed.
              TABLE 3-1.  SUMMARY STATISTICS FOR HOURLY NOX DATA
                          FROM YORK COUNTY (ppm 9 7% C02)
Summary
Statistic
Mean
Minimum
Median
Maximum

Unit 1
109.1
62.8
110.3
159.6

Unit 2
114.8
61.7
114.6
177.2

Unit 3
110.1
51.4
107.7
283.8
      To assess the Impact of averaging time on NOX emissions variability,
block averaging periods of 1-hour, 3-hours, 8-hours, and 24-hours and rolling
averages of 3-days and 7 days were analyzed using the raw NOX data.  The
Impact of different averaging times on each of the three units Is shown 1n

bjw:033                                B3-4

-------
            TABLE  3-2.   SHAPIRO-MILK NORMALITY TEST STATISTIC (W)
                        FOR HOURLY NOX DATA FROM YORK COUNTY
Dataset
Hourly
Ln Hourly
Unit 1
0.98*
0.96
Unit 2
0.98*
0.98*
Unit 3
0.94
0.99*
*Indicates a normal data distribution at the 95% confidence level
Figure 3-3.  The higher estimated 99th percentile NOX emission levels for
Unit 3 for periods of 24 hours or less are caused by the two high NOX levels
measured on February 20, and are not considered representative of normal
operating conditions.  Selection of an averaging time based on the other two
units 1s less clear than with the Stanislaus County data.  For consistency
with the Stanislaus County analysis, however, a 24-hour averaging period was
used.
      As shown in Table 3-3, the 24-hour block means from all three units are
normally distributed at the 95 percent confidence level, as are the In-
transformed averages for Units 2 and 3.  Further, the normality test
statistics on Units 1 and 3 using the natural data are higher than for the In-
transformed data.  As a.result, the continuously achievable emission rates
were estimated based on the assumption that the dally average NOX levels are
normally distributed.
            TABLE 3-3.  SHAPIRO-WILK NORMALITY TEST STATISTIC (W)
                        FOR DAILY NOX DATA FROM YORK COUNTY
Dataset
Hourly
Ln Hourly
Unit 1
0.92*
0.90
Unit 2
0.97*
0.98*
Unit 3
0.97
0.96*
*Ind1cates a normal data distribution at the 95* confidence level

bjw:033                                B3-5

-------
         180
         170
I   160
3


I   150



    140



    130



    120



    110
    ,§
              1-hr
                     3-hr
                           Unit!
8-hr       24-hr


Averaging Period

    +  Unit 2
                                                          3-day
                                                        Units
                                                                 7-day
Figure 3-3.   Impact of Averaging Time on NOY Emissions from York County
                                       B3-6

-------
      Based on the one month of available data,  it appears that a daily
average NOX emission limit of 140 ppm can be continuously achieved by Units 1
and 2, as shown in Table 3-4.  The estimated limit of 150 ppm on Unit 3 is
caused by the higher standard deviation in the data resulting from the two
hours of high NOX readings on February 20.  However, all  of these data were
collected in February when less vegetative waste Is likely to be present in
the data.  As a result, the impact of seasonal variations on achievable NOX
levels 1s uncertain.
             TABLE 3-4.   SUMMARY  STATISTICS  FOR  DAILY NOX AVERAGES
                             FROM YORK COUNTY'(ppm 0 7%
Statistic
Number of Blocks
Mean
Standard Deviation
Continuous
Compliance Level*
Unit 1
22
108.8
8.3

131.9
Unit 2
24
115.0
7.1

134.7
Unit 3
23
100.6
14.0

149.4
'Based on one exceedance per year (Mean + 2.777 * Standard Deviation)
3.3   CONCLUSIONS
      Based on the York County data, the following conclusions are reached
regarding the continuously achievable NOX emission level associated with
Westinghouse/0'Connor combustion systems:

      •     Long-term average NOX emissions of less than 115 ppm at 7 percent
            02 are achievable, but short-term variations in operating
            conditions and fuel composition will Influence the NOX emission
            level that can be continuously achieved.

      •     The hourly NOX data appear to be normally distributed, thus
            supporting the use of untransformed data to estimate continuously
            achievable emission levels.
bjw:033
                                     B3-7

-------
            An  averaging  period  of 24  hours  1s beneficial  in reducing the
            variability 1n  average NOX levels.
                                             i '
            The one month of available data  1s Insufficient to conclude what
            level of NOX  emissions can be achieved during  all times of the
            year.  Based  on the  available data, however, a NOX level of 140-
            150 ppm over  a  24-hour averaging period can be continuously
            achieved.
bjw:033                               . B3-8

-------
         APPENDIX C
                  t

 ADDITIONAL MUNICIPAL WASTE

COMBUSTOR EMISSIONS TEST DATA

-------
                              TABLE OF CONTENTS

                                                                         Page

1.0   Background and Objective  ...................     Cl-1


2.0   Babylon ............................     C2-1


3.0   SEMASS ............................     C3-1


4.0   Vancouver ...........................     C4-1


5.0   Indianapolis   ............ '.' ............     C5-1


6.0   Summary of Performance  ....................     C6-1

      6.1   Spray Dryer/Fabric Filter Performance ..........     C6-1

            6.1.1  Acid Gas .....................     C6-1
            6.1.2  Particulate Matter ;  ...............     C6-2
            6.1.3  Metals  ......................     C6-2
            6.1.4  Dloxins/Furans  ........ ..........     C6-2

      6.2   Spray Dryer/Electrostatic Precipitator Performance   .  .  .     C6-3

            6.2.1  Acid Gas .....................     C6-3
            6.2.2  Partlculate Matter ................     C6-3
            6.2.3  Metals  ......................     C6-3
            6.2.4  Dioxins/Furans  ..................     C6-4

      6.3   Dry Sorbent Injection/Fabric Filter  Performance  .....     C6-4

            6.3.1  Acid Gas .....................     C6-4
            6.3.2  Partlculate Matter ................     C6-4
            6.3.3  Metals  ......................     C6-4
            6.3.4  Dioxins/Furans  ..................     C6-5
7.0    References
bjw:033

-------
                         1.0   BACKGROUND AND OBJECTIVE

     Prior to the proposal of performance standards and emission guidelines
for municipal waste combustors (MWC's), all available emissions test data
collected at a number of MWC's were reviewed to assess the performance of
different air pollution control technologies.*  This appendix reviews data
obtained subsequent to the proposal which was collected during testing at four
MWC's:  Babylon, New York; Rochester, Massachusetts (commonly referred to as
SEMASS); Vancouver, British Columbia; and Indianapolis, Indiana.  The Babylon
and Indianapolis facilities were selected for'review because total
dioxin/furan (CDD/CDF) emissions were reported to be higher than for similar
MWC's that had been previously examined.  In the case of SEMASS and Vancouver,
there were limited data on the emissions control performance of the air
pollution control technologies used at these two MWC's.
pw/137
90-1.com

-------
                                2.0   BABYLON2.3

     The Babylon Resource Recovery facility in West Babylon, New York,
consists of two identical mass burn waterwall  combustors, each designed to
combust 375 tons per day (tpd) of municipal solid waste (MSW).  The MSW is
combusted in a Martin stoker combustion system equipped with a Zurn waterwall
boiler.  Each boiler normally produces about 88,000 pounds per hour (Ibs/hr)
of steam at 680°F and 640 pounds per square inch gauge (psig).
     The combustion gases exiting each combustor enter an air pollution
control device (APCD) system supplied by Belco Corporation.  The APCD system
consists of a Deutsche-Babcock-Anlagen (DBA) spray dryer (SD) followed by an
American Air Filer pulse-jet fabric filter (FF).  The DBA SD uses dual-fluid
nozzles and an up-flow reactor.  The FF consists of six compartments and uses
acid-resistant fiberglass bags.  The design air-to-cloth ratio is 2.5 gross
and 3.1 net, with one compartment out of service.
     In January and February of 1989, emissions testing was performed by Ogden
Projects, Inc., to demonstrate compliance with permit conditions.  Acid gas,
particulate matter (PM), and metals data were collected simultaneously.  A
continuous emissions monitoring (CEM) system was used to monitor nitrogen
oxides (NOX), carbon dioxide (C02), and carbon monoxide (CO) at the SD/FF
outlet during these runs.  Emissions of CDD/CDF were measured separately.
peration of the APCD system during compliance testing was in accordance with
directions supplied by DBA based on their experience with units in Europe.
The FF inlet temperature for various full-load test runs ranged from 316 to
326°F.  The FF inlet temperatures were estimated by adding 15°F to the stack
temperatures since the temperature at the Inlet to the FF was not monitored.
These temperatures are higher than for other SD/FF systems previously
examined, where inlet FF temperatures were generally less than 290°F.l  The
somewhat higher operating temperature of the FF during compliance testing at
Babylon reflects European vendor concern about wetting of the filter bags by
damp filtercake, and the desire to minimize the potential'for bag blinding.
Operating at these higher temperatures will generally reduce sorbent
utilization and S02 removal.  The impact on HC1 is generally small.
                                        ,
                                       -1
90-1.com

-------
     Flue gas at the SD/FF Inlet and outlet of both units was sampled for acid
gases, and the data are presented in Table 2-1.  Hydrogen chloride (HC1)
wasanalyzed by manual methods and sulfur dioxide (SO?) was measured by a CEM
system.  Triplicate runs were made on each unit at normal operating (full-
load) conditions.  Triplicate runs were also made on Unit 1 at reduced-load
conditions (approximately 70 percent.of full-load conditions).
     Outlet S02 concentrations for the six full-load runs on both units ranged
from 16 to 40 parts per million (ppm) at 7 percent oxygen (02), with an
average estimated FF Inlet temperature of 318°F.  The percent reduction of S02
during the full-load runs on Units 1 and 2 ranged from 64 to 87 percent.  The
lowest percent reduction of S02 occurred during the first run on Unit 1 (64
percent) and coincided with the lowest S02 inlet concentration of any of the
runs.  The percent reduction in S02 during the other five runs all  exceeded 82
percent.  The S02 concentration at the FF outlet during the reduced-load runs
on Unit 1 ranged from 6 to 30 ppm at 7 percent 02, with an average estimated
FF Inlet temperature of 293°F.  The percent reduction of S02 during the
reduced-load runs ranged from 87 to 96 percent, with an average percent
reduction of 92 percent.
     Outlet HC1 measurements for the full-load tests ranged from 19 to 22 ppm
at 7 percent 02 on Unit 1 and from 40 to 61 ppm on Unit 2.  Reduced-load
measurements at the Unit 1 outlet found HC1 concentrations of 21 to 26 ppm at
7 percent Q£.  During each of the nine test runs, HC1 reduction efficiencies
ranged from 93 to 98 percent.  The higher HC1 outlet concentrations during the
full-load runs on Unit 2 correspond to higher HC1 concentrations at the SO
inlet and were not.due to significantly decreased HC1 removal efficiencies.
Stoichiometric ratios could not be calculated because the lime feed rates were
not in the available data.
     The PM data for Babylon are presented in Table 2-2.  Sampling for PM was
performed at the SD/FF outlet of each unit simultaneously with the acid gas
sampling.  Outlet PM concentrations for the six full-load runs on both units
ranged from 0.0008 to 0.0022 grains per dry standard cubic feet (gr/dscf) at
12 percent C02-  The average outlet PM concentration was 0.0017 gr/dscf for
Unit 1 and 0.0012 gr/dscf for Unit 2.  The outlet PM concentrations for the
three reduced-load runs on Unit 1 ranged from 0.0022 to 0.0036 gr/dscf with an
                                     f9-
90-1.con                               C2-2

-------
                                                               TABLE 2-1.  ACID GAS DATA FOR BABYLON
o
ro
 i
Test
Condition
Coabustor « Nonaald
SD/FF - Nonsal
Average Unit 1
Conbustor • Normald
SD/FF » Normal
Average Unit 2
Conbustor « Reduced Load*
SD/FF = Normal

Average Unit 1
Run
Nuaber*
1-1
1-2
1-3

2-1
2-2
2-3

1-R1
1-R2
1-R3


FF Inlet -
Temperature*
316
316
321
318
319
318
316
318
292
293
294

293
Stoichiometric
Ratio6
NA
NA
NA

NA
NA
NA

NA
NA
NA


Acid Gas Concentration
(ppmv, dry at 7X 02>
Inlet Outlet
S02
98
251
186
178
130
125
169
141
188
149
237

191
KCl
573
785
794
717
886
1.149
1.017
1,017
857
706
723

762
SOz
36
40
34
37
23
16
24
21
12
6
30

16
HCl
20
19
22
20
61
47
40
49
26
21
25

24
Acid Gas
Removal
Efficiency (X)
502 HCl
64
84
82
77
82
87
85
85
94
96
87

92
96
98
97
97
93
96
96
95
97
97
97

97
              •Run nuaber consists of unit nuaber followed by the run nuaber for that  unit.   "R"  indicates a run at reduced-load conditions.



              bEstiMted by adding 15°F to average stack temperature.


              cLia» feed rate data not available, therefore, stolchioawtric ratio cannot be  calculated.



              dFull-load conditions = 375 tpd.


              •70 percent of full-load conditions « 263 tpd.
              pnw/137

              90-2-1.tbl

-------
                        TABLE  2-2.   PM DATA FOR BABYLON
Condition
Combustor • Normal c
SD/FF - Normal
Average Unit 1
Combustor - Normal c
SD/FF - Normal
Average Unit 2
Combustor - Reduced
1oad°
SO/FF - Normal
Average Unit 1
Run
Number3
1-1
1-2
1-3

2-1
2-2
2-3

1-R1
1-R2
1-R2

FF Inlet
Temperature0
(OF)
316
316
321
318
319
318
316
318
292
293
294
293
Flue Gas
Flow (acfm)
58,500
59,300
59,300
59,000
61,000
60,500
56,200
59,200
38,300
39,400
39,900
39,200
Outlet PN
Concentration
(gr/dscf
at 12% C02)
0.0016
0.0022
0.0012
0.0017
0.0013
0.0008
0.0016
0.0012
0.0022
0.0036
0.0029
0.0029
aRun number consists  of unit number followed by the run number for that  unit.
 "R" Indicates a  run  at reduced-load conditions.
^Estimated by adding  15°F to average stack temperature.
CFull-load conditions - 375 tpd.
d70 percent of full-load conditions » 263 tpd.
pnrw/117
M»-2-2.tbl

-------
average of 0.0029 gr/dscf.  Removal efficiencies for PM could not be
determined because PM concentrations at the SD/FF inlet were not measured.
     Table 2-3 presents the metals emissions data.  Triplicate runs for metals
concentrations were performed at the SD/FF outlet of Unit 2 only.  These
samples were analyzed for beryllium, lead, mercury, arsenic, cadmium,
chromium, nickel, antimony, cobalt, copper, manganese, selenium, vanadium, and
zinc.  Outlet lead concentrations ranged from 0.8 to 2.5 micrograms per dry
standard cubic meter  (ug/dscm)  at  7 percent QZ with an average  concentration
of 1.4 |ig/dscm.   Average  mercury emissions were  451 u.g/dscm,  ranging from 190
to 620 ug/dscm.   Arsenic,  cadmium,  chromium, and nickel  emission levels were
below detection  limits.   Metals concentrations at the SO inlet  were  not
measured.  Based on typical  inlet  concentrations of lead and mercury from
other mass burn  MWC's,1 lead  reductions exceeded 99.9 percent while  mercury
reductions appeared negligible.
     A separate  set of three  runs  were performed for CDD/CDF  and polycyclic
aromatic hydrocarbons/polychlorinated biphenyls  (PAH/PCB)  at  the SD/FF outlet
of Unit  1.   The  CDD/CDF data  are presented  in Table 2-4.   Outlet
concentrations ranged from 12.6 to 27.2 nanograms  per dry  standard cubic  meter
(ng/dscm)  at 7 percent Q£ over three runs and averaged  21.9 ng/dscm.  Inlet
CDD/CDF  concentrations were not measured  and removal   efficiencies could  not
be calculated.   Temperatures at the FF  inlet during these  runs  ranged from 314
to 326°F.
 pan/137                                £2-5
 90-1.com                               u* 3

-------
                                                               TABLE 2-3.  METALS EMISSIONS DATA FOR BABYLON
r>
ro
o>

Teit
Condition
Combustor • Nonul6
SD/FF • NorMl

Average Unit 2

Run
Hurter*
2-1
2-2
2-3

FF Inlet
TeMperatureb
(°F)
319
318
316
318
Hi
Concentration
(gr/dscf at
12X C02)
0.0013
0.0008
0.0016
0.0012
Outlet Concentration
(Ua/dscm at 71 0,)
As
N0d
NO
NO

Cd
NO
NO
ND

Cr
ND
ND
NO

Pb
0.8
2.5
1.0
1.4
Hg
544
620
190
451
Nt
ND
ND
ND

'Run lumber consists of unit number followed by the  run number for that unit.
Estimated by adding 1S°F to average stack temperature.
GFull-load conditions - 375 tpd.
dND • Not detected.   Considered as  zero when calculating averages.

-------
                                           TABLE 2-4.   CDO/COF DATA FOR BABYLON
Test
Condition
Ccabustor - NonMlc
SO/FF - Nonul
Average Unit 1
Run
Nwfcer*
1-1
1-2
1-3

FF Inlet
Tenperatureb
. (°FJ
317
314
326
319
Inlet COO/CDF
Concentration
(ng/dicm at 7X 02)
HHd
NN
NN

Outlet COO/CDF
Concentration
(ng/dsm at 7X 02>
12.6
25.9
27.2
21.9
'Run number consists of unit nuober foilowed by the run nuri>er for that unit.
Estimated by adding 15°F to average stack tenperature.
cFull-load conditions « 375 tpd.
•to • Not measured.
pmw/137
90-2-4.tbl

-------
                                 3.0  SEMASS*

     The SEMASS Waste-to-Energy Resource Recovery facility in Rochester,
Massachusetts, consists of two identical refuse-derived fuel (RDF) fired
MWC's, each rated 900 tpd of RDF.  The units are identical semi-suspension
fired Riley boilers which are capable of firing both fuel oil and RDF.  Fuel
oil is burned during start-up, shutdown, and process upset conditions to
maintain a minimum furnace temperature of 1,800°F-  Combustion gases from each
boiler pass through a spray dryer and a 5-field electrostatic precipitator
(SD/ESP) supplied by Joy Manufacturing.
     Emission testing was conducted from February 15 through April 18,  1989,
to determine compliance with the emission standards set by the Massachusetts
Department of Environmental Quality and Engineering and the Southeastern
Massachusetts Air Pollution Control District.  The facility was tested  while
firing No. 2 fuel ojl and while firing RDF.  Only the results of the RDF
firing are summarized in this report.  Testing was conducted while the
facility was operating at or near its rated capacity.  During these tests,
only four of the five ESP fields were in operation.
     Triplicate runs were made to measure PM and S02 at the inlet and outlet
of the SD/ESP of each unit.  During these tests, hydrogen fluoride (HF), CO,
and NOX were also measured at the SD/ESP outlet.  A separate set of triplicate
runs was performed at the SD/ESP outlet of both units to determine the
concentrations of trace metals including lead, mercury, arsenic, cadmium,
chromium, and nickel.  A third set of three runs was made at the SD/ESP outlet
to determine concentrations of CDD/CDF.
     Acid gas data for SEMASS are presented in Table 3-1.  Outlet
concentrations of SOg averaged 67 ppm at 7 percent Oa over three test runs  for
Unit 1 and 55 ppm for Unit 2, with average estimated ESP  inlet temperatures of
294 and 300°F, respectively.  The outlet S0£ concentrations on Unit 1 varied
from 51 to 75 ppm and on Unit 2  from 9  to 96 ppm.  Average removal
efficiencies were 56 percent  for Unit 1 and 65 percent for Unit 2.  Available
data do not explain the variability in  SOg emissions from Unit 2.  Data for
HC1 were not reported.
piw/137
90-1.com

-------
                                                     TABLE 3-1.   ACID  GAS DATA  FOR SEHASS
Test
Condition
Combustor * Normal6
SO/ESP = Normal
Average Unit 1
Combustor = Normal6
SO/ESP » Normal
Average Unit 2
u> aRun number consists of
i
Run
Number"
1-1
1-2
1-3

2-1
2-2
2-3

unit number
ESP Inlet
Temperature Stoichiometric
(Of)b Ratio0
295
294
294
294
300
301
299
300
followed by run number
NA
NA
NA
NA
NA
NA
NA
NA
for that unit.
Acid Gas Concent rat'i on
(pprov. dry at 7X 02)
Inlet Outlet
S°2
154
157
150
154
155
142
190
162

HCl"
NA
NA
NA
NA
NA
NA
NA
NA

S02
75
74
51
67
9
96
60
55

HCl°
NA
NA
NA
NA
NA
NA
NA
NA

Acid Gas
Removal
Efficiency (X)
S02 HCld
51
53
65
56
94
32
68
65

NA
NA
NA
NA
NA
NA
NA
NA

Estimated by adding 15°F to average stack temperature.
cstoichiometric ratio cannot be determined since HCl data were not  reported.
Information on the HCl tests is not available.
eFull-load conditions "900 tpd.

-------
     The PM data for SEMASS are presented 1n Table 3-2.   Outlet PM
concentrations on Unit 1 averaged 0.0080 gr/dscf at 12 percent C02 over three
runs with a removal efficiency of 99.8 percent (with 4 ESP fields operating).
Unit 2 exhibited an average outlet PM concentration of 0.0120 gr/dscf at a
removal efficiency of 99.7 percent (with 4 ESP fields operating).
     Table 3-3 presents the metal emissions data for the triplicate runs
conducted at the SO/ESP outlet of both units.  Average mercury emissions were
59 |ig/dscm at 7 percent 0? for Unit 1 and 105 ug/dscm for Unit  2.
Concentrations of arsenic, cadmium, and lead were relatively consistent for
all runs and averaged 0.7, 9.6, and 300 ug/dscm, respectively,  for Unit 1,
and 1.5, 6.8, and 235 ug/dscm, respectively, for Unit 2.  Concentrations of
nickel and chromium were relatively consistent for all runs on Unit 1,
averaging 6.8 and 6.5 ug/dscm, respectively.  These levels are  similar to the
nickel and chromium concentrations measured during Run 2 on Unit 2.  During
Runs 1 and 3 on Unit 2, however, the nickel and chromium concentrations were
an order-of-magnitude higher.  There is no discussion in the test report
Indicating whether these elevated levels of chromium  and nickel may have been
caused by a sampling or analytical error.
     Average metals emission  levels were calculated using all of the data.
Based on typical Inlet concentrations of these metals at other MWC's,1 removal
efficiencies were greater than 75 percent for mercury and greater than
99 percent for arsenic, cadmium, chromium,  and lead.  Estimated removal
efficiencies for nickel were  99  percent for Unit  1  and 97 percent for Unit 2.
     The CDD/CDF data for the SD/ESP outlet of Units  1 and 2 are presented in
Table 3-4.  Outlet concentrations  for Unit  1 ranged from 5.1 to  13.6 ng/dscm
at 7 percent 02  for three test runs and averaged  9.3  ng/dscm.   The CDD/CDF
levels during the  three test  runs  on Unit 2 were  18.0, 6.6, and  907 ng/dscm.
The high reading during the third  run was attributed  in the compliance test
report to unsteady MWC  operating conditions.  The average CDD/CDF
concentration for  Unit  2,  excluding the high value, was 12.3 ng/dscm.  No
measurements of  uncontrolled  CDD/CDF emissions were made.
 pw/137
 90-1.com

-------
                                                             TABLE  3-2.   PM DATA FOR SENASS

Condition
Combustor = Normald
SO/ESP « Normal

Average Unit 1
Combustor - Normald
SO/ESP = Normal

Average Unit 2

Run
Hunter*
1r1
1-2
1-3

2-1
2-2
2-3

ESP inlet
Temperature
<°F)b
295
294
294
294
300
301
299
300

Flue Cas
Flow (acfm)
180,000
162,000
176,000
179.000
213,000
218.000
209,000
213.000
Inlet PM
Concentration
(gr/dscf at 12X COj)
4.09
4.10
4.65
4.28
3.76
4.31
3.50
3.86
Outlet PM
Concentration6
(gr/dscf at 12X COj)
0.0090
0.0080
0.0080
0.0080
0.0170
0.0120
0.0070
0.0120
Removal
Efficiency
(X)
99.7
99.8
99.8
99.8
99.5
99.7
99.8
99.7
"Run number consists of unit number followed by the run number for that test.
^Estimated by adding 15°F to average stack temperature.
C4 of 5 ESP fields operating.
dFull-load conditions = 900 tpd.

-------
                                                          TABLE 3-3.   METALS EMISSIONS DATA FOR SEMASS
Test
Condition
Combustor • Normal0
SO/ESP •

Average
> Normal

Unit 1
Combustor - Norms lc
SO/ESP i

Average
• Normal

Unit 2
2 "Run number consists of
i
U1 BEBtinal

ted hu aHHina «<
Run
Number"
1-1
1-2
1-3

2-2«
2-3
2-4

unit number

'C tA •u«f»aru
Particulate
ESP Inlet Concentration
Temperature6 (gr/dscf at
<°F> 12X COj)
287
288
287
287 0.0080d
293
292
293
293 0.0120°*
followed by the run number for that test.

k •>•>•*• If +^riMP\*f»ek+llf»*k
Outlet Concentration
tua/dscm at 7X AS)
As
0.2
1.8
0.2
0.7
1.9
2.5
0.2
1.5



Cd
5.2
6.5
17.0
9.6
6.5
6.3
7.5
6.8



Cr
4.0
8.3
7.3
6.5
26.6
5.7
14.4
15.6



Pb "
177
213
509
300
220
267
219
235



Hg
70
53
55
59
141
100
75
105



Ni
4.6
10.7
5.1
6.8
62.2
6.3
28.8
32.4



         cFull-load condition*  = 900  tpd.

         dAverage of results  fro* separate PM testing.

         •Run maters consistent with those used  in test report.  No information IMS provided in test report regarding missing runs.
pmw/137
90-3-3.tbl

-------
                      TABLE 3-4.  CDD/CDF DATA FOR SEMASS
Unit
Combustor * Normal0
SD/ESP - Normal
Average Unit 1
Combustor • Normal c
SO/ESP - Normal
Average Unit 1
Run
Number9
l-2d
1-3
1-5

2-1
2-2
2-5

ESP Inlet
Temperature
(0F)b
298
300
295
298
298
295
300
298
Outlet CDO/COF
Concentration
(ng/dscm at 7% 03)
5.1
13.6
9.2
9.3
18.0
6.6
9076
12.3
aRun number consists of unit number followed by the  run  number for that test.
^Estimated by adding 15°F to average stack temperature.
cFull-load conditions « 900 tpd.
     numbers consistent with those used  1n test  report.   No Information was
 provided in test report regarding missing runs.
6H1gh value suspected to have been caused by  unsteady conditions during
 sampling.  Number not Included  in average.
pmw/137
90-3-».tbl                                 C3-6

-------
                            4.0  VANCOUVER*.6,7,8

     The Greater  Vancouver  Regional District MWC in Burnaby, British Columbia,
Canada, consists  of  three Martin mass burn waterwall combustors each with  a
capacity of  265 tpd  of MSW.  The combustion gases from each unit enter  an  APCD
system which consists of a  dry sorbent  injection system followed by a fabric
filter (DSI/FF).  The DSI system was designed by Flakt and consists of  a
quench chamber, into which  water is sprayed to cool the gases, followed by a
reaction chamber  where dry  hydrated lime  is injected to remove acid gases.
Sodium sulfide is injected  through a separate spray system at the entrance to
the quench chamber to enhance the removal of mercury from the flue gas.
     In November  of  1988, emissions tests were conducted to demonstrate
compliance with Environment Canada emission regulations.  Particulate matter
emissions were measured in  accordance with the British Columbia Ministry of
Environment's (BCMOE) Source Testing Code using a U. S. Environmental
Protection Agency (EPA) Method 5 sampling train.  Triplicate runs were
conducted at the  DSI/FF outlet (stack)  of each unit.  Concurrent with each of
the PM emissions  tests, flue gas was sampled for S02, HC1, and HF (Unit 3
only) by manual methods.  Triplicate runs were made for SOj according to BCMOE
Method 6 at  both  the DSI/FF system inlet and outlet.  A separate EPA Method 5
sampling train was used to  measure HC1  emissions.  In addition, OEM's were
used to measure S02, HC1, CO, and NOX at the exit to the DSI/FF.  On Unit  3,
an EPA Modified Method 5 sampling train was also used to sample metal
emissions (arsenic,  cadmium, chromium,  lead, mercury, and nickel) at the
DSI/FF outlet.  Sampling was conducted  according to the British Columbia
Source Testing Code  for the Measurement of Emissions of Particulates from
Stationary Sources.  In addition, flue  gas at the DSI/FF inlet and outlet  of
Unit 3 was sampled for CDD/CDF emissions using a California Air Resources
Board Semi-VOST sample train.  The CDD/CDF runs were conducted separately  and
do not correspond to the run numbers for metals, particulate, and acid  gas.
     Additional testing of  mercury emissions was performed between March and
December 1989 to  study the  effect of sodium sulfide injection on mercury
emissions.
paw/137
90-1. com                               C4-1

-------
     Emissions of S02 and HC1 are summarized 1n Table 4-1.  The unbracketed
values are S02 and HC1 emissions taken by manual methods.  The values shown In
parentheses Indicate averaged data taken from the CEM system at the exit of
the DSI/FF-  Removal efficiencies in parentheses are values calculated from
manual data at the DSI/FF Inlet and CEM data at the DSI/FF outlet.  Inlet
concentrations of SOg averaged 139 ppm at 7 percent 03 over three runs for
Unit 1, 157 ppm for Unit 2, and 161 ppm for Unit 3i  Outlet S02 concentrations
based on the manual sampling method averaged 15 ppm on Unit 1, 31 ppm on
Unit 2, and 18 ppm on Unit 3.  This corresponds to average S0£ removal
efficiencies of 90 percent, 79 percent, and 89 percent, respectively.
     Average inlet concentrations of HC1 were 194 ppm at 7 percent 02 over
three runs for Unit 1, 238 ppm for Unit 2, and 270 ppm for'Unit 3.  Outlet HC1
concentrations for these units (based on the manual sampling method) averaged
9 ppm, 17 ppm, and 17 ppm, respectively.  High wind conditions during the
testing of Unit 2 resulted in outlet HC1 data for only two runs.  The
resulting average HC1 removal efficiencies were 95 percent for Unit 1, and 93
percent for Units 2 and 3.
     Data of the PM emissions for Vancouver are presented in Table 4-2.
Average PM concentrations at the DSI/FF outlet averaged 0.0079 gr/dscf at
7 percent 03 over three runs on Unit 1, 0.0143 gr/dscf on Unit 2, and
0.0044 gr/dscf on Unit 3.  The Unit 2 average includes only two runs due to
cancellation of one run because of high winds.
     Metals emission data from the DSI/FF inlet and outlet of Unit 3 are
presented in Table 4-3.  Mercury emissions ranged from approximately 300 to
750 |ig/dscm over three runs with average emissions of 485 ug/dscm,  suggesting
little or no mercury removal based on an average measured mercury
concentration at the DSI/FF Inlet of 525 |ig/dscm.  Outlet arsenic, cadmium,
and lead emissions were fairly uniform between runs and averaged 1.6, 3.7, and
78 ug/dscm at 7 percent 03, respectively.  Outlet emissions of chromium  and
nickel were highly variable during the three runs.  The variability in outlet
levels of chromium and nickel during Run 3 versus Runs 1 and 2 may be
attributable to sampling problems.  When the high values from Run 3-3 are
excluded, the average chromium concentration 1s 41 ug/dscm at 7  percent  03
pw/137
90-1 .co.                               C4-2

-------
                                                                TABLE 4-1.  AGIO GAS DATA FOR VANCOUVER
o
•fh
Test
Condition
Combustor - Normal8
DSI/FF - Normal

Average Unit 1
Combustor • Normal8
DSI/FF > Normal


Average Unit 2
Combustor « Normal6
DSI/FF » Normal

Average Unit 3
Run
Number9
1-1
1-2
1-3

2-1
2-2
2-3
2-4

3-1
3-2
3-3

FF Outlet
Temperature"
<°F)
295
295
295
295
295
295
295
" 295
295
295
295
295
295
Stoichiometric
; Ratio6
HA
MA
NA

NA
NA
NA
NA

NA
NA
NA

Acid Gas Concentration4*
(ppmv, dry at 7X fe)
Inlet Outlet
S°2
162
128
128
139
136
189
145
MM
157
195
142
145
161
HCl
184
183
214
194
NM<
225
265
225
238
299
294
216
270
$02
31(55)
9(6)
6(1)
15(21)
27(14)
28(22)
39(41)
NM(NN)
31(26)
25(68)
6(39)
22(4)
18(37)
^ ^
HCl
19(21)
3(4)
4(3)
9(9)
NN(NN)
NN(NN)
21(24)
13(15)
17(20)
15(11)
14(17)
23(6)
17(11)
Acid Gas
Removal
Efficiency (X)
SOj HCl
81(66)
93(96)
95(99)
90(87)
80(69)
65(88)
73(71)
NA(NA)
79(83)
87(65)
96(73)
85(97)
89(78)
90(89)
98(99)
98(98)
95(95)
NA(NA)
NA(NA)
92(91)
94(93)
93(92)
95(96)
95(94)
89(97)
93(96)
            •Run number consists of unit number followed by the run number for that test.
            ^Estimated based on stack temperature data,  which were reported to be 12.5 to  30°F higher  than temperatures measured by the CEHs at the FF
             outlet.   (Plant personnel indicated this increase was not unusual.)   The plant also  reports normal operating temperatures at the FF outlet
             between  284 and 302°F.
            cLime feedrate data not available,  therefore, Stoichiometric ratio cannot be calculated.
            ^Quantities in parentheses denote values obtained from average of CEN data analysis at  the DSI/FF outlet (stack).
            •Full-load conditions - 265 tpd.
            *NM * not measured.  High winds resulted in  the cancellation of HCl measurement at stack during Run 2-2.  Run 2-4 conducted for HCl only.
           pmw/137
           90-4-1.tbl

-------
                       TABLE 4-2.   PM DATA FOR VANCOUVER
Condition
Combustor = Normal d
DSI/FF - Normal
Average Unit 1
Combustor = Normal d
DSI/FF = Normal
Average Unit 2
Combustor - Normal d
DSI/FF = Normal
Average Unit 3
Run
Number3
1-1
1-2
1-3

2-1
2-2
2-3

3-1
3-2
3-3

FF Outlet
Temperature^
(°F)
295
295
295
295
295
295
295
295
295
295
295
295
Flue Gas
Flow (acfm)
32,400
30,900
28,100
30,500
NMe
31,500
30,200
30,900
30,700
30,500
30,700
30,600
Outlet PM
Concentration
(gr/dscf
at 7% Og )c
0.0144
0.0050
0.0042
0.0079
NM6
0.0117
0.0169
0.0143
0.0043
0.0040
0.0048
0.0044
aRun number consists of unit number followed by the run number  for  that  test.

^Estimated based on stack temperature data, which were reported to  be  12.5  to
 30°F higher than temperatures measured by the CEMs at the  FF outlet.   (Plant
 personnel Indicated this Increase was not unusual.)  The plant also reports
 normal operating temperatures at the FF outlet between 284 and 302°F.

CC02 data not available, therefore, corrections were made to 7  percent 02.

^Full-load conditions = 265 tpd.

eNM = not measured; cancelled because of high winds.
pmw/137
90.4-2.tbl
C4-4

-------
                                                                 TABLE 4-3.  METALS EMISSIONS DATA FOR VANCOUVER
Test
Condition
Combustor - Normald
DS1/FF - Normal
Average Unit 3
Outlet PI
Concent rat i<
Run FF Outlet (gr/dscf at
NuBfaera Temp.(oF)b 7X02)
3-1
3-2
3-3

295
295
295
295
O.OM3
0.0040
0.0048
0.0044
1
XI
t
As
1080
1090
100
760
Inlet Concentration
(ug/dsca at 7X 02)c
Cd
1490
1140
840
1160
Cr
475
380
485
445
Pb
38200
33100
19800
30400
Hg
795
595
190
525
Hi
3300
2500
615
2140
As
1.7
1.6
1.4
1.6
Outlet Concentration
(ug/dscm at 7X 02)c
Cd
4.3
3.9
2.9
3.7
Cr
7.4
74
490e
41
Pb
85
75
75
78
Hg
300
750
400
485
Ni
7.6
15.9
436e
11.8
As
99.8
99.9
98.6
99.4
Removal Efficiency (X)
Cd
99.7
99.7
99.7
99.7
Cr
98.4
80.4
-I.Oe
89.4
Pb
99.8
99.8
99.6
99.7
Hg
62.3
-26,1
•111.0
-25
Ni
99.8
99.4
29.1e
99.6
    a Run nunber consists of unit number followed by the run nuaber for that test.
    b Estimated based on stack teaperatura data,  which were reported to be 12.5 to 30°F higher than temperatures mfeasured by the CEMs at the FF outlet.  (Plant personnel
      indicated this increase was not unusual.)  The plant also reports normal  operating temperatures at  the  FF outlet between 284 and 302°F.
    c CO, data not available,  therefore, corrections were made to 7 percent 0,.
S  d Full-load conditions * 265 tpd.
 i                                 "^
01  e Suspiciously high numbers may be due to sampling problem.   These numbers  are not  included in  averages.

-------
and the average nickel concentration Is 11.8 iig/dscm.  Removal efficiencies
were greater than 99 percent for arsenic, cadmium, lead, and nickel, and
approximately 90 percent for chromium.
     Due to the relatively high mercury emissions measured during the
compliance testing in November 1988, tests were run to Investigate the effect
of injecting a sodium sulfide solution at the quench chamber inlet on mercury
removal efficiency.  Table 4-4 presents the results of these tests for all
three units.
     In March 1989, three runs were made at the OSI/FF inlet and outlet of
Unit 1.  During the tests, a 10 percent solution of sodium sulfide was sprayed
into the quench chamber inlet at rates ranging from 1 to 2 kilograms per hour
(kg/hr) of sodium sulfide.  Outlet mercury concentration averaged
456 ng/dscm.  The removal efficiency for the three runs ranged from 59 to 66
percent, averaging 62 percent.  Increasing the spray rates did not appear to
enhance mercury removal.
     Testing was repeated in April 1989, this time a with solution
concentration of 15 percent sodium sulfide.  Three sampling runs were
conducted at the DSI/FF inlet and outlet on Unit 1 to detect mercury
emissions.  The sodium sulfide injection rate was 3 kg/hr.  The average inlet
concentration was 1,357 ug/dscm at 7 percent 02 and the average outlet
concentration was 632 ug/dscm.  This resulted in an average removal
efficiency of 53 percent.
     In July 1989, more testing was performed utilizing a 1 percent solution
of sodium sulfide at liquid flowrates of 200 to 500 liters per hour (1/hr).
Results of the tests were not found in the available data.  In August 1989,
five runs were conducted.  Measurement of mercury concentrations were made at
the DSI/FF inlet and outlet on Unit 2.  Sodium sulfide solution ranged from 2
to 4 percent, and flow rates ranging from 2 to 6 kg/hr were tested.  The
average inlet concentration was 661 (ig/dscm at 7 percent Q£ and the average
outlet concentration was 95 u.g/dscm over five test runs.  The mercury removal
efficiency during each run ranged from 76 to 88 percent and did not show any
trend with sodium sulfide feed rate.
     In December 1989, the Burnaby MWC started continuous injection of sodium
sulfide on all three units to reduce mercury emission levels.  Tests on

p»/137
90-1 .cen                               C4-6

-------
TABLE 4-4.  HERCURY DATA FOR VANCOUVER
Date
11/88

03/89

04/89

08/89

12/89





pmw/137
90-4-4. tbl
Test
Conditions
Combustor = Normal6
DSI/FF = Normal
Average
Combustor » Normal6
DSI/FF = Normal
Average
Combustor = Normal6
DSI/FF = Normal
Average
Combustor = Normal6
DSI/FF = Normal-
Average
Combustor « Normal6
DSI/FF « Normal
Average

Average

Average

Run
Number*
3-1
3-2
3-3

a
^
3d

1-1
1-2 ~~
1-3

2-1
2-2
2-3
2-4
2-5

1-1
1-2
1-3

2-1
2-2
2-3

3-1
3-2


FF Outlet
Temperature
<°F>
29Sc
295c
29Sc
295c
NAe
NAe
NAe

NA*
NA*
NA*
295C
NA«
NA«
NA«
NAe
NAe

NA«
NAe
NA*

NAe
NAe
NAe

NAe
NAe


Na2S Inlet Hercury Outlet Mercury Removal
Flowrate Concentration Concentration Efficiency
(kg/hr) (fig/dscn at 7X 02> (pg/dscm at 7X 02) (X)
0
0
0

1.0
2.0
2.0
1.7
3.0
3.0
3.0
3.0
2.5
6.0
2.0
3.0
6.0
3.9
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.0
Continued
795
597
188
527
1.465
993
1.151
1,203
1.423
1.443
1.205
1.357
406
775
670
793
661
661
NMfl
tuft
NH9

NH9
MH9
MH9

NN9
NH9


303
752
400
485
570
406
393
456
670
750
473
632
98
91
84
101
103
95
138
67
146
117
149
115
118
127
152
159
156

62
-26
-113
-26
61
59
66
62
53
48
61
53
76
88
88
87
64
85
--
--
--
--
--
-•


-------
                                                                TABLE  4-4.   FOOTNOTES  (CONCLUDED)
CO
        *Run number consists of unit followed by the run number for that test.
        bFul(-load conditions = 265 tpd.
        cEstinated based on stack temperature data, which were reported to be 12.5 to 30°F higher than temperatures Measured by the CEHs at  the  FF
         outlet.  (Plant personnel indicated this increase was not unusual.)  The plant also reports normal operating temperatures at the FF outlet
         between 284 and 302°F.
              not known.
        'Temperature data not available.  Plant normally operates between 264 and 302°F at the FF outlet.
        'only the average temperature for three tests was reported.
        "Not measured.

-------
Units  1,  2,  and  3  were  conducted  with  a  2-percent  sodium sulfide solution and
a  4  kg/hr feed rate.  Average  outlet concentrations  were 117  jig/dscm at 7
percent 02 over  three test  runs on  Unit  1  and  127  ug/dscm at  Unit 2.   The
average outlet level for  Unit  3 was 156  jig/dscm over two test runs.   Although
inlet  mercury concentrations were not  reported for tests at any of the units,
an 80-percent removal efficiency  is suggested,  based on  the average inlet
levels during the  August  1989  tests.
     Inlet and outlet CDD/CDF  data  for Unit  3  are  shown  in Table 4-5.   Only
two  runs  were reported  due  to  the discovery  that portions of  the inlet and
outlet samples from Run 3-3 had been mixed together  inadvertently.   Run 1
exhibited an inlet CDD/CDF  concentration of  101 ng/dscm  at 7  percent 02 and an
outlet concentration of 0.03 ng/dscm,  resulting in a removal  efficiency of
99.9 percent.  The inlet  concentration for Run 2 was 55  ng/dscm at 7 percent
02 and the outlet  concentration was 9.25 ng/dscm,  resulting in a CDD/CDF
removal efficiency of 83.3  percent. The average outlet  CDD/CDF concentration
based  on  these two runs was 4.6 ng/dscm, corresponding to a removal  efficiency
of 92  percent.
pnw/137
90-1.com                               C4-9

-------
                                                   TABLE 4-5.  COD/COF DATA FOR VANCOUVER


Test
Condition
Coefcustor - NorMlb
DS1/FF • Nornal
Average Unit 3


Run
Nusber*
3-1
3-2
3-3



FF Outlet
Temperature
NAC
NAC
NAC


Inlet COO/CDF
Concentration
(ng/dsoi at 7X Oj)
101
NAd
78

Outlet COD/COF
Concentration
(ng/dscs) at 7X 02)
0.03
'•?
NAd
4.64
CDO/COF
Resnval
Efficiency
(X)
99.9
B3.3
MAd
92

•Run nuster consists of unit nusber followed _, 		
 do not correspond to the run nusbers for the awtals,  participate,  and acid gas  tests.


^Full-load conditions « 265 tpd.

cTeaperatures data not available.  Plant nonMlly operates between 284 and 302°F at  the  FF exit.

        was not analyzed after it was discovered that portions of the inlet and  outlet saisples had been accidentally aiixed.

-------
                              5.0  INDIANAPOLIS9

     The Indianapolis Resource Recovery facility in Indianapolis, Indiana,
consists of three Martin grates equipped with waterwall boilers.  Each unit
has a capacity to process 787 tpd of MSW and is rated at 167,000 Ib/hr of
steam.  Each boiler train is equipped with a SD/FF APCD system supplied by
Environmental Elements.  Each SD has three rotary atomizers to introduce lime
slurry Into the combustion gas stream.  Following each SD is a pulse-jet FF
which has 10 modules and uses acid-resistant fiberglass bags.  The design
air-to-cloth ratio is 3.0 gross and 3.3 net,-with one compartment out of
service.
     From June 27 through July 1, 1989, Ogden Projects, Inc., conducted tests
to demonstrate compliance with permit conditions for all three units.
Simultaneous PM and HC1 samples were collected from each unit using an EPA
Method 5 sampling train.  Separate runs were conducted to measure metals  (lead
and mercury) and CDD/CDF from Unit 1 only.  Lead and mercury sampling was
conducted using a combined Method 12/101A train.  Sampling for CDD/CDF was
done with a Modified Method 5 train.  Sampling was limited to triplicate
samples collected at the stack for all pollutants.
     Acid gas (HC1) data are presented in Table 5-1.  Outlet HC1
concentrations ranged from 4 to 35 ppm at 7 percent Q£ over three test runs on
Unit 1, from 0.1 to 0.5 ppm on Unit 2, and from 0.2 to 0.7 ppm for Unit 3.  No
explanation 1s available for the variation in HC1 measured on Unit 1 versus
Units 2 and 3.  The estimated FF Inlet temperatures averaged 307°F, 309°F and
310°F for each unit respectively.  Data for HC1 inlet concentrations were not
reported, therefore, removal efficiencies could not be determined.  Data for
S02 were not reported.
     The PM data for Indianapolis are presented in Table 5-2.  Outlet PM
concentrations for triplicate runs on Unit 1 averaged .0040 gr/dscf at
12 percent C02-  For Unit 2, the average was .0041 gr/dscf.  Unit 3 had an
average of .0026 gr/dscf.  Removal efficiencies could not be determined since
Inlet PM values were not measured.
     Table 5-3 lists the test results for lead and mercury emissions from
Unit 1.  Outlet concentrations for lead, based on three runs, averaged

pro/137                                r. .
90-1.co«                               C5-1

-------
                                                     TABLE 5-1.  ACID GAS DATA FOR INDIANAPOLIS
Unit
Combustor » Normal d
SD/FF = Normal
Average Unit 1
Combustor = Normal d
SD/FF • Normal
Average Unit 2
Combustor - Normal d
SD/FF = Normal
o Average Unit 3
tn
Run
Number8
1-1
1-2
1-3

2-1
2-2
2-3

3-1
3-2
3-3


FF Inlet
Temperature
(•F)b
306
307
307
307
310
308
310
309
307
311
311
310

Stolchlometrlc
Rat1oc
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA

Acid Gas Concentration
(ppmv, dry at 7X 0?)
Inlet 'Outlet
S02
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA

HC1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA

S02
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA

HC1
3S.O
11.1
4.4
16.8
0.5
0.1
0.1
0.2
0.7
0.5
0.2
0.5

Acid Gas
Removal
Efflclencv (X)
S02
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA

HC1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA

"Run number consists of unit number followed by run number for that unit.
Estimated by adding 15*F to average stack temperature.
cstolchlometrtc ratio cannot be determined since lime feedrate data Is unavailable  for Unit  1.  Stolchlometrlc data for Unit 2 will  be Included  In
 the final memorandum.
dFull-load conditions = 787 tpd.

-------
                      TABLE 5-2.  PM DATA FOR INDIANAPOLIS
Unit
Combustor • Normal c
SD/FF - Normal
Average Unit 1
Combustor - Normal0
SD/FF - Normal
Average Unit 2
Combustor - Normal0
SD/FF -Normal
Average Unit 3
Run
Number*
1-1
1-2
1-3

2-1
2-2
2-3

3-1
3-2
3-3

FF Inlet .
Temperature
CF)
306
307
307
307
310
308 .
310
309
307
311
311
310
Flue Gas
Flow (acfm)
171,000
171,000
177,000
173,000
187,000
171,000
175,000
178,000
169,000
170,000
173,000
171,000
Outlet PM
Concentration
(gr/dscf
at 12% C02)
.0064
.0021
.0036
.0040
.0038
.0050
.0034
.0041
.0031
.0011
.0035
.0026
aRun number  consists of unit number followed by the  run  number for that unit.
''Estimated by adding 15*F to average stack-temperature.
cFull-load conditions - 787 tpd.
pmw/137
90.5-2.tab
                                   C5-3

-------
                                         TABLE 5-3.  METALS EMISSIONS DATA FOR INDIANAPOLIS
O
Ul
 i



Unit
Coatustor • Normal0
SO/ff » Normal

Average Unit 1


Run
Number'
1-1
1-2
1-3

Particulate
FF Inlet Concentration
Temperature (gr/dscf
CF)b at 12X 0>2>
309
308
304
. 307 0.00359**


Outlet Concentration
f pq/dso at
Pb
4.22
4.16
4.39
4.26
TXfe)
ft
354
210
28S
283
•Run number conaiatt of unit  muter followed by the run nunber for that test.


^Estiauted by adding 15*F to  average stack temperature.


cFull-load conditions = 787 tpd.


^Average of results fron separate PN testing.

-------
4.26 ng/dscm at 7 percent 02-  The outlet values for mercury ranged from 210
to 354 ug/dscm, and averaged 283 ug/dscm.  The  average estimated  FF Inlet
temperature was 307°F.  Based on typical uncontrolled lead and mercury levels
from other MWC's,1 the estimated reduction Is greater than 99 percent for
lead, but relatively low for mercury.
     Outlet concentrations of CDD/CDF for Unit 2 are shown in Table 5-4.  The
three CDD/CDF test runs gave values ranging from 6.1 to 15 ng/dscm at
7 percent 62, with an average value of 11.3 ng/dscm.  The average estimated FF
inlet temperature was 311°F-  Inlet concentrations were not reported,
therefore, removal efficiencies could not be calculated.
pnw/137                                rc r
90-1.com                                Cb'5

-------
                                                               TABLE  5-4.  COD/COF DATA FOR  INDIANAPOLIS
r>
en
 i
o»
Unit
Confcustor = Normal0
SO/FF * Normal
Average Unit 2
Run
Number8
2-1
2-2
2-3

FF Inlet
Temperature"
310
311
312
311
Inlet COD/CDF
Concentration
(ng/dscn at 7X 02)
NM*1
NH
NH

Outlet COD/COF
Concentration
(ng/dscn at 7X
12.7
15.0
6.1
11.3
02>


                     •Run nunfcer consists of unit nunber followed by the run number  for that  unit.



                     ^Estimated by adding 15°F to average stack temperature.



                     cFull-load conditions = 787 tpd.



                     *NN = not measured.

-------
                          6.0   SUMMARY OF  PERFORMANCE

     Three types of APCD systems are reviewed 1n this appendix:  SD/FF
(Babylon and Indianapolis), SD/ESP (SEMASS), and DSI/FF (Vancouver).
Section 6.1 evaluates SD/FF performance, Section 6.2 evaluates SD/ESP
performance, and Section 6.3 evaluates DSI/FF performance.  This data is
similar to those found in the background information document (BID) entitled
"Municipal Waste Combustors - Background Information for Proposed Standards:
Post-Combustion Technology Performance".!
6.1  SPRAY DRYER/FABRIC FILTER PERFORMANCE (BABYLON AND INDIANAPOLIS)
6.1.1  Acid Gas
     Sulfur dioxide data were reported for Babylon only.  For the full-load
tests, S0£ removal was between 82 and 87 percent on five of six runs.  During
the sixth run, the inlet SO? level (95 ppm) was lower than during the other
runs and the S0£ reduction was 64 percent.  For the reduced-load tests, the
SO? removal efficiency averaged 92 percent over 3 runs.
     The BID concluded that a 90 percent S02 removal efficiency is achievable
for SD/FF systems with increased stoichlometric ratios and low FF inlet
temperatures (<300°F).  The relationship between stoichiometric ratio and S0£
removal efficiency cannot be determined at Babylon since lime feedrate data
were not available.  Jhe FF inlet temperatures at Babylon were estimated to be
between 315 to 320°F during the full-load tests and 290 to 295°F during the
reduced-load tests.  Since the S0£ removal efficiency during the reduced-load
tests exceeded 90 percent, the results appear to support the importance of
temperature on S02 removal .efficiency.  However, the higher removal
efficiencies could also be the result of  increased stoichiometric ratios and
longer flue gas residence times in the SD and FF during the reduced-load
tests.
     The HC1 data at Babylon show removal efficiencies of approximately 95 to
97 percent for full-load tests.  Average outlet levels ranged from
approximately 20 to 50 ppm at 7 percent Og.  Reduced-load tests averaged
97 percent removal, with outlet levels of 21 to 26 ppm.  At Indianapolis, HC1
outlet levels averaged less than 20 ppm on Unit and 1 and less than 1 ppm on
Units 2 and 3, which are among the lowest HC1 levels measured.  These results

pnw/137                                r- .
90-1.com                               CO-1

-------
are consistent with the BID'S conclusion that 97 percent removal of HC1 1s
achievable for SD/FF systems opertlng at low Inlet temperatures and high
stolchlometrlc ratios.
6.1.2  Particulate Matter
     The average PH outlet levels were less than 0.003 gr/dscf at Babylon, and
less than 0.004 gr/dscf at Indianapolis.  These levels are similar to those in
the BID for SD/FF systems.
6.1.3  Metals
     At Indianapolis and Babylon, measurable levels of metals were recorded
for lead and mercury only.  Data on other metals were not measured at
Indianapolis and emissions of arsenic, cadmium, chromium, and nickel were
below detection levels at Babylon.  Although inlet levels of lead and mercury
were not measured at either facility, removal efficiencies for lead exceeded
99 percent based on typical inlet levels at other MWC's, which supports the
99 percent removal efficiency cited in the BID.
     Indianapolis had an average outlet mercury level of 283 ug/dscm over
three runs, and Babylon averaged 451 ug/dscm over 3 .runs.  Although inlet
levels were, not measured at either MWC, the outlet levels suggest little
mercury reduction was achieved.  In the BID it was noted that the level of
mercury removal with the SD/FF system varied from site to site, but stated
that an outlet level of 300 jig/dscm is achievable with SD/FF systems.  This is
not supported by the Babylon data.
6.1.4  Dioxins/Furans
     The average CDD/CDF outlet levels at Babylon and Indianapolis were
21.9 ng/dscm and 11.3 ng/dscm, respectively.  These levels are higher than the
CDD/CDF level of less than 10 ng/dscm concluded in the BID to be achievable by
SD/FF systems operating at FF inlet temperatures of 300°F or less.  The higher
CDD/CDF emissions during the tests conducted at Babylon and Indianapolis may
have been due to the FF inlet temperatures experienced during these tests,
which were greater than 300°F.  However, the effect of temperature 1s not
conclusive, and other process conditions may also play a role in CDD/CDF
control by SD/FF systems.
pmw/137
90-1.com                               C6-2

-------
6.2  SPRAY DRYER/ELECTROSTATIC PRECIPITATOR PERFORMANCE (SEMASS)
6.2.1  Add Gas
     Only S02 data were reported at SEMASS.  Removal efficiencies were
generally between 50 and 70 percent for six runs.  This is lower than the
reported 75 percent removal efficiency for other SD/ESP systems operating at
similar ESP inlet temperature levels  (<300°F).1  The average outlet levels of
60 ppm are consistent with emission limits achievable with a SD/ESP system,
however.  Because HC1 and lime feed rates were not reported at SEMASS, the
stoichiometric ratio cannot be determined and the relationship between the
stoichiometric ratio and removal efficiencies cannot be established.
6.2.2  Particulate Matter
     The PM results at SEMASS show three run-average removal efficiencies of
99.7 for Unit 1 and 99.8 percent for  Unit 2, corresponding to 0.012 and
0.008 gr/dscf, respectively.  These results are similar to the 0.01 gr/dscf PM
levels reported to be achievable by a SD/ESP system.1
6.2.3  Metals
     Although uncontrolled metals concentrations were not measured at SEMASS,
estimated removal efficiencies for arsenic, cadmium, chromium, and lead were
greater than 99 percent.  Estimated removal efficiencies for nickel were
99 percent for Unit 1 and 97 percent  for Unit 2.  These removal efficiencies
are comparable to those reported in the BID, which  indicated that SD/ESP
systems are capable of reducing arsenic, cadmium, lead, and nickel by 98
percent and chromium by 95 percent.
     The average mercury levels from  SEMASS were 60 and 105 iig/dscm over three
runs for Unit 1 and Unit 2, respectively.  These outlet data indicate an
estimated 75 percent removal efficiency, based on typical mercury inlet
levels.1  The estimated 75 percent reduction in mercury disagrees with the
earlier conclusion reached in the BID that mercury  is not effectively removed
by a SD/ESP system.  As discussed in  the BID, it has been theorized that
mercury removal 1s enhanced by carbon in the flyash, which provides adsorption
sites for the mercury.  The observed  mercury control at SEMASS may be related
to the RDF semi-suspension firing method used, which may result in increased
flyash carryover, and thus, Increased mercury adsorption onto flyash.
pmw/137
90-1.com                               C5-3

-------
6.2.4  Dioxlns/Furans
     The outlet CDD/CDF levels at SEMASS during five of the six runs were
between 5 and 20 ng/dscm.  These levels are lower than those reported for
SD/ESP equipped facilities, which ranged from 40 to over 250 ng/dscm.1  The
CDD/CDF emission rates from SEMASS are similar to those at Babylon and
Indianapolis, but are somewhat higher than those recorded at most other SD/FF-
equipped MWC's.  The SEMASS data, along with other data from SD/ESP-equipped
systems, suggest that SD/ESP systems are less efficient at control of CDD/CDF
emissions than SD/FF systems.
     The CDD/CDF level was 907 ng/dscm during the sixth run at SEMASS.  The
test report stated that plant operating records indicated that several periods
of unsteady combustion conditions occurred during this run.  Similar increases
in CDD/CDF emissions from SD/FF systems have not been reported.  This suggests
that SD/ESP systems may be more sensitive to combustor operating conditions
than SD/FF systems.
6.3  DRY SORBENT INJECTION/FABRIC FILTER PERFORMANCE (VANCOUVER)
6.3.1  Acid Gas
     The S02 data at Vancouver indicate removal efficiencies of 80 to 90
percent.  Hydrogen chloride data at Vancouver indicate removal efficiencies of
93 to 95 percent.  Flue gas temperatures at the FF outlet were an estimated
295°F, but lime feed rates were not reported.  These results are generally
lower than the 90 percent S0£ reduction and 95 percent HC1 reduction which
were determined to be achievable by DSI/FF systems in the BID.
6.3.2  Particulate Matter
     The average outlet PM concentrations from Units 1 and 3 at Vancouver were
0.0079 gr/dscf and 0.0044 gr/dscf, respectively.  The PM data collected for
two runs on Unit 2 were slightly higher, averaging 0.0143 gr/d'scf.  These
results therefore, support an achievable PM level of 0.015 gr/dscf, rather
than the average PM outlet level of 0.01 gr/dscf determined in the BID to be
achievable by DSI/FF systems.
6.3.3  Metals
     Average removal efficiencies for arsenic, cadmium, lead, and nickel at
Vancouver were greater than 99 percent.  The average removal efficiency for
chromium was 90 percent.  A 99 percent reduction of arsenic, cadmium, and lead
and a 96 percent removal of nickel and chromium was determined to be
pw/137
90-i.com                               C6-4

-------
 achievable for DSI/FF systems from previous  data.   With  the exception of
 chromium,  the results from Vancouver for these metals  support  the data in the
 BID.   As Indicated above,  the large variations in  the  chromium (and nickel)
 data  may have been due to  sampling problems.
      Mercury data indicated a 70 percent removal efficiency for DSI systems
 operating  at FF inlet temperatures below 300°F.  At Vancouver,  mercury removal
 without sodium sulfide injection was minimal.   With sodium sulfide injection,
 actual  removal efficiencies ranged from 60 to  85 percent.   The levels of
 mercury removal achieved during the August 1989 tests  were 75  to 85 percent.
 These higher levels,  compared to the earlier tests, are  believed to be the
 result of  improved atomization and mixing when feeding higher  volumes of lower
 concentration solution versus lower volumes  of high concentration solutions.
 6.3.4  Dioxins/Furans
      The CDD/CDF emissions at Vancouver averaged 4.6 ng/dscm over two runs,
 with  a removal efficiency  of 92 percent.  These results  are consistent with
 previous data, which  indicate that a level of  less than  10 ng/dscm is
 achievable at FF inlet temperatures less than  300°F.  The  FF inlet temperature
 was not available at  Vancouver, but was believed to have been  relatively
 constant at around 295°F.
pw/137
90-1.com                               C6-5

-------
                                7.0   REFERENCES


1.   U. S. Environmental Protection Agency.  Municipal Waste Combustors -
     Background Information for Proposed Standards:  Post-Combustion
     Technology Performance.  EPA-450/3-89-27c.  August 1989.
                                                      •
2.   Ogden Projects, Inc.  Environmental Test Report, Babylon Resource
     Recovery Test Facility, Units 1 and 2.  (Prepared for Ogden Martin
     Systems of Babylon, Inc.)  March 1989.

3.   Telecon.  Bahor, B., Ogden-Martln Systems, with White, D., Radian
     Corporation.  April 20, 1990.  Description of Babylon, NY, MWC APCD
     system.
                                           • *
4.   Eastmount Engineering, Inc.  Final Report, Waste-to-Energy Resource
     Recovery Facility, Compliance Test Program, Volumes II - V.   (Prepared
     for SEMASS Partnership.)  March 1990.

5.   B. H. Levelton & Associates, Ltd.  A Report on  Plant Emissions from the
     Burnaby Incinerator Compliance Monitoring.  (Prepared for Greater
     Vancouver Regional District  [GVRD].)  February  1989.

6.   Trip Report.  Burnaby MWC, British Columbia, Canada.  White,  D., Radian
     Corporation.  May  1990.

7.   Telefax.  Frame, G., Flakt,  Canada, to White, D., Radian Corporation.
     Summary from GVRD.  January  1990.

8.   Telefax.  Knizek,  0.  GVRD to Romesberg,  L., Radian Corporation.
     Information on Burnaby MWC.  February 1990.

9.   Ogden Project, Inc.  Environmental Test Report,  Indianapolis  Resource
     Recovery Facility, Appendix  A and Appendix B, Volume  I.   (Prepared for
     Ogden Martin Systems of  Indianapolis, Inc.)  August 1989.
pw/137                                r,  ,
90-1.com                               C7-1

-------
                                  TECHNICAL REPORT DATA
                           (Please read Instructions on the reverie before completing)
I. REPORT NO.
    EPA-450/3-91-004
                            2.
            3. RECIPIENT'S ACCESSION NO.
«. TITLE AND SUBTITLE
 Municipal Waste Combustion:  Background Information
 for Promulgated Standards and Guidelines - Summary of
 Public Comments and  Responses Appendices A to  C	
            5. REPORT DATE
            8. PERFORMING ORGANIZATION CODE
r. AUTHOR(S)
                                                         8. PERFORMING ORGANIZATION REPORT NO.
I. PERFORMING ORGANIZATION NAME AND ADDRESS

  Office of Air Quality Planning and Standards
  U.S. Environmental  Protection Agency
  Research Triangle Park,  NC  27711
            1O. PROGRAM ELEMENT NO.
             11. CONTRACT/GRANT NO.
                68-02-4378
12. SPONSORING AGENCY NAME AND ADDRESS
  DAA for Air Quality Planning and Standards
  Office of Air and  Radiation
  U.S. Environmental Protection Agency
  Research TTManglo  Pa-rV.  Mr  ?7711	
                OTE!
                                                          13. TYPE OF REPORT AND PERIOD COVERED
                                                             Final
             14. SPONSORING AGENCY CODE

                200/04
15. SUPPLEMENTARY NOTES
16. ABSTRACT
   Appendices A to  C  to the "Municipal Waste  Combustion: Background Information
   for Promulgated  Standards and Guidelines - Summary of -Public Comments  and
   Responses" (EPA-450/3-91-004), address key technical issues related  to the
   promulgated rules  for municipal waste combustors (MWC's).
   Appendix A provides analysis of the continuous  S02 control capabilities of
   spray dryer/fabric filter (SD/FF) and spray dryer/electrostatic precipitator
   (SD/ESP) control systems for MWC's.  Achievable 862 performance levels for
   these systems are  determined based on the  analysis.
   Similarly, Appendix B provides analysis of continuous NOx emissions  data from
   MWC's.  Results  are presented for the statistical analysis of NOx data obtained
   from a grate-fired mass burn waterwall MWC using selective noncatalytic reduction
   (SNCR) to reduce NOx emissions and from a  rotary mass burn waterwall MWC designed
   to limit NOx emissions through combustion  control.
   Appendix C provides additional MWC emissions test data which became  available
   following proposal of the standards and guidelines for MWC's on December 20,  1989.
   The appendix reviews data at four MWC's with either unique air pollution control
   technologies or  emissions which are higher than for similarly controlled MWC's
   that had been previously
17.
                               KEY WORDS AND DOCUMENT ANALYSIS
                  DESCRIPTORS
                                             b.lDENTIFIERS/OPEN ENDED TERMS  C. COSATI Field/Group
   Air Pollution
   Municipal Waste Combustors
   Incineration
   Pollution Control
   Costs
  Air Pollution  Control
13B
18. DISTRIBUTION STATEMENT
19. SECURITY CLASS (Tilts Repor
   Unclassified	
20. SECURITY CLASS (This pageI
                                                                           118
                                                                       22. PRICE
                                                Unclassified
EPA Form 2220-1 (R.w. 4-77)   PREVIOUS EDITION is OBSOLETE

-------