EP A/600/A-97/Q50
REDUCING DIOXIN FORMATION THROUGH COAL CO-FIRING
K. Raghunathan
Acurex Environmental Corporation
4915 Prospectus Drive
Durham, NC 27713
Brian K. Gullett, Chun Wai Lee, James D. Kilgroe
Air Pollution Prevention and Control Division
National Risk Management Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
ABSTRACT
Past research has established through bench- and pilot-scale studies that the presence of sulfur
dioxide (SO,) inhibits formation of polychlorinated dibenzo-/?-dioxins (PCDDs) and
polychlorinated dibenzofurans (PCDFs). The inhibition mechanism appears to be a combination
of sulfur poisoning of the copper catalyst and gas-phase depletion of chlorine (Cl2) through
reaction with S02. This leads to the possibility that co-firing waste combustors with high-sulfur
coal can control PCDD and PCDF formation.
In this study, a commercial, densified, refuse-derived fuel (dRDF) was co-fired with high- sulfur
Illinois No. 6 coal in a 2-million Btu/hr (0.6 MW) stoker combustor. Additional parameters
varied in the tests include dRDF feed rate, hydrogen chloride (HC1) concentration, sorbent
injection, and flue gas quench rate. Flue gas was sampled for PCDDs and PCDFs from three
locations at nominal temperatures of 600, 300, and 150°C. Most of the PCDD and PCDF
formation took place between 600 and 300°C, within about 0.5 s. Results also show that coal
co-firing inhibits PCDD and PCDF formation substantially. Injection of hydrated lime sorbent is
also effective in reducing their formation. Combustion quality, dRDF feed rate, and HC1
concentration are some of the other parameters that influence PCDD and PCDF formation.
INTRODUCTION
Formation of polychlorinated dibenzo-p-dioxins (PCDDs) and polychlorinated dibenzofurans
(PCDFs) has been observed in waste combustors. However, PCDD and PCDF yields from coal
combustion are relatively insignificant. Previous bench- and pilot-scale research (1,2) has shown
that the presence of sulfur dioxide (S02) can inhibit PCDD and PCDF formation, and suggested
co-firing high-sulfur coal with refuse-derived fuel (RDF) to reduce the emissions. This work
describes research in a large scale combustor which shows that coal/RDF co-firing can
significantly lower PCDD and PCDF formation.
PCDD and PCDF formation mechanisms have been reviewed in detail by Addink and Olie (3).
Measured yields can be affected by the combustion quality, levels of chlorine (CI) and catalyst
species, residual carbon on the fly ash, and process parameters such as temperature and residence
1
-------
time. Two candidate mechanisms have been established for the inhibition of PCDD and PCDF
formation by sulfur species:
1) Gas-phase reaction where S02 converts molecular chlorine (Cl2) to hydrogen chloride (MCI), a
less-likely chlorinating agent (4,2):
CI, + S02 + H20 <===> 2 HC1 + SO,
Our recent work (2) appears to confirm this mechanism as a possibility.
2) Poisoning of copper catalysts by S02 (1):
CuO + S02 + Vz02 <===> CuS04
It has been shown that copper sulfate (CuS04) is a less active catalyst for the production of Cl2
through the Deacon process, as well as for the biaryl synthesis step of PCDD and PCDF
formation (1).
Lindbauer et al. (5) have reported that co-firing coal in a municipal solid waste (MSW) incinerator
leads to appreciably lower PCDD and PCDF levels. Recently, Ogawa et al. (6) compared the
effect of adding pure S02 versus generating S02 through coal addition, and found the latter to be
more effective. Thus, there may be other benefits with coal co-firing than simply being a source
of S02.
In this work, experiments were conducted in a large, pilot-scale unit which is a state-of-the-art
facility with fuel handling and combustion release rates representative of large field units. The
effect of coal co-firing was studied for a municipal-waste-based solid fuel. Tests were conducted
over a range of process variables (e.g., calcium hydroxide [Ca(OH)2] injection, HC1
concentration, flue gas temperature, quench, and residence time) so that the results could have
implications for a wide variety of waste combustors.
EXPERIMENTAL FACILITY AND PROCEDURES
The newly constructed, EPA Multi-Fuel Combustor (MFC) facility was used for conducting tests
for this project. The MFC is rated at 0.6 MW (2x106 Btu/hr) thermal output and is capable of
burning a wide variety of solid fuels including MSW, RDF, biomass, and coal. The modular
design of the facility provides flexibility for studies on various pollutant emissions and control and
for solid fuels with unknown firing and handling characteristics.
A schematic of the MFC facility is shown in Figure 1. The MFC consists of a waste feeding
system, a coal feeder, a lower combustion chamber containing a stoker, a radiant section, a
convective flue gas passage, a baghouse, and a scrubber system. In addition, there is a separate
fuel preparation system for shredding, screening, and mixing of the fuel. A large loading hopper
conveys the processed fuel to the fuel silo.
As indicated in Figure 1, there are several ports for flue gas sampling and temperature
2
-------
measurements along the convective section and the duct. The convective section is equipped with
cooling coils with high-pressure water circulation. The cooling coil originally present in location
B was removed for this project, to accommodate PCDD/PCDF sampling, A typical temperature
drop across the convective section is from 600 to 150°C which includes the PCDD/PCDF
formation temperature "window" (200 to 500°C). Residence time/quench across this window is
known to be an important parameter, and it can be varied in our tests by changing the temperature
set point of the cooling water. From the convective section, flue gas was sampled through
continuous emission monitors (CEMs) for recording the oxygen (02), carbon dioxide (C02),
carbon monoxide (CO), HC1, and S02 concentrations.
A commercial, densified refuse-derived fuel (dRDF), from municipal waste, was the main fuel
used. The coal used was an Illinois #6 coal, donated by Monterey Coal Company, Carlinville, IL,
for this project. The coal was ground and classified to an average size of about 1 mm. The coal
was fed using a screw feeder and entered the burner at the same location as the dRDF. Results of
analyses of the dRDF and coal are shown in Table 1. For runs with sorbent injection, a
commercial hydrated lime was injected as a slurry.
The experimental parameters varied in the tests were:
dRDF feed rate - LOW or HIGH
coal feed - ON or OFF
quench - LOW or HIGH
sorbent injection - ON or OFF
HC1 addition - ON or OFF
The experiments were structured to provide information on all two-factor interactions among the
above parameters for statistical modeling. A total of 30 tests, including blanks, were run for
various parameter settings.
During each test, the fuel feed rate was adjusted, whenever necessary, to maintain a constant flue
gas temperature. Nearly the same firing rate was maintained between tests, again, by matching
the flue gas temperature reading. For low dRDF feed runs, natural gas was co-fired to
compensate for the decrease in heat release and maintain similar temperature ranges between runs.
Quench was varied by varying the temperature set point of the cooling water used in the
convective section of the furnace. For runs with HC1 addition, there was about a 150 ppm
increase in HC1 concentration. The sorbent feed rate was about 2.2 kg/hr and the injection
temperature was approximately 700°C.
For most of the runs, flue gas was sampled from two locations (A and B in Figure 1) for
PCDD/PCDF according to EPA Method 23 (7), and the sampling duration was 2 hours. For
selected runs, dioxin samples were drawn from an additional sampling port, just before the
convective section (location D), Before and after running Method 23 trains, a velocity traverse of
the duct was carried out to measure the flue gas flow rate. The samples were analyzed in EPA's
in-house Organics Support Laboratory (OSL), via procedures described elsewhere (2). Recently,
the OSL has expanded its capability to include quantification of mono-tri PCDD and PCDF
congener class mass as well. Thus, the results obtained are levels of each mono-octa PCDD and
3
-------
PCDF congener class in the sample. However, since mono-tri PCDD and PCDF are not subject
to regulation, most of the following discussions are based on the tetra-octa PCDD/PCDF yields.
DATA ANALYSIS
For each test, temperature and flue gas composition data were stored in the data acquisition
system. Run averages of these data were calculated for the duration of PCDD/PCDF sampling so
that these values correspond to the measured yields. With the total flue gas flow rate known from
the velocity traverse data, the average flue gas temperature profile data were used to determine
the flue gas residence times at various locations in the MFC, In the temperature profiles, the
residence time variable, tR, is set to zero at 650°C. The choice of 650°C stems from the fact that
most of PCDD/PCDF formation is known to take place below this temperature.
Conventionally, for comparative analysis, emission data are normalized to a specific 02
concentration (e.g., 7%) to decouple possible differences in flue gas dilution between tests. In
this work, however, for tests with LOW dRJDF feed rate, the co-fired natural gas (and its
stoichiometric combustion air) effectively dilutes the flue gas, but without the equivalent increase
in 02. Thus, normalizing the PCDD/PCDF data with 7% O, would underestimate the
PCDD/PCDF yields attributed to the LOW dRDF feed rate. Therefore, the PCDD/PCDF data
are normalized with the dRDF feed rate for each run. Experimental data on fuel feed rates are not
available and had to be estimated. Mass balance equations for 02, C02, and total flue gas flow
rates were set up and (from the data on the fuel composition, combustion stoichiometry, and flue
gas concentration and flow rate) the individual fuel feed rates were estimated.
RESULTS AND DISCUSSION
For the tests performed in this work, measured/estimated parameter values are given in Table 2,
For most of the runs, the CO levels from dRDF burning were low, indicating good combustion
quality. Due to fluctuations in the dRDF feed rate and the possible inherent variations in the
dRDF composition, the flue gas composition exhibited fluctuations throughout a run, and there
were some variations in the average compositions from run to run.
In this project, for three runs, PCDD/PCDF levels were measured simultaneously from three
locations, including the high-temperature (~ 600°C) dioxin sampling port (Port D). For one of
these runs, which was a dRDF baseline test (without coal or sorbent), the PCDD/PCDF levels
measured at these three ports are shown in Figure 2. The figure shows the total PCDD and
PCDF yields for tetra-octa as well as for mono-octa congener classes. The PCDD/PCDF levels
measured at the high-temperature port are relatively low, and the majority of the formation takes
place between Ports D and B, in less than 0.5 sec. An examination of the temperature profile,
shown in the same figure, indicates that in this region, the temperature decreases sharply. A more
gradual temperature decline would result in larger residence times in this temperature window,
possibly allowing more PCDD and PCDF formation. After Port B, the change in PCDD and
PCDF levels is small. The figure also shows that the yields of mono-tetra PCDD and PCDF are
significant, and it is possible that under different process conditions some of them might shift to
the toxic, higher chlorinated congeners.
4
-------
Figure 3 shows a comparison of the PCDD and PCDF yields, expressed in nanograms per dry
standard cubic meter of the flue gas, between the LOW and HIGH dRDF feed settings, and each
bar represents the average from several runs. The average dRDF feed rates for these settings
were 28 and 80 kg/hr, respectively. As expected, yields are higher for the higher dRDF feed rate.
The figure also shows that the yields at the downstream sampling port are lower. Since the
cooling coils occupy part of the cross sectional area of the convective section, a significant fly ash
deposition was noted on the surface of these coils between ports A and B. With such system-
specific limitations, mechanistic explanations for the lower yields measured at the downstream
port, such as the occurrence of destruction reactions after Port B, should not be inferred.
If the amount of PCDD and PCDF generated varies linearly with the dRDF feed, then the yields
normalized with the dRDF feed rate should remain relatively unchanged. With the estimated
dRDF feed rates, the PCDD and PCDF data for the two sampling ports were normalized and the
average yields for the LOW and HIGH dRDF feed settings are shown in Figure 4. Comparable
normalized yields between the two cases, especially for the data from Port B, show the strong
correlation with the dRDF feed rate. With this normalization, effects of other variables can be
assessed, in spite of the run-to-run variations in the dRDF feed rate.
An average analysis of all the experimental data has been performed, combining the data from
both sampling ports. The data are divided into three groups, dRDF alone, dRDF with coal co-
firing, and dRDF with sorbent injection, although within a group other parameters such as HC1
concentration may vary. For each group, average normalized PCDD and PCDF yield is computed
for each congener class. Results are plotted for tetra-octa PCDD and PCDF congener classes in
Figure 5. Both coal co-firing and sorbent injection decrease PCDD and PCDF formation
significantly; the congener class pattern is similar to that of the base dRDF case and, therefore, the
reduction is not congener-class-specific. The total tetra-octa PCDD and PCDF yields for the
three groups are plotted in Figure 6. The results clearly indicate that coal/dRDF co-firing as well
as calcium-based sorbent injection reduce PCDD and PCDF formation. The average flue gas
sulfiir-to-chlorine ratio for the coal co-fired tests was about 1.5.
Figure 7 shows the effect of added HC1 on the yields. In computing the average yields, runs with
sorbent or coal addition were not included. PCDD and PCDF formation increases with the flue
gas HC1 concentration. However, unlike with the dRDF feed rate, the variation does not appear
to be linear — the yields increase by a factor of 2 for a 4-fold increase in the HC1 concentration.
The effect of combustion quality was examined for two runs at a HIGH dRDF feed rate (no coal),
in which average flue gas parameters were similar except for the CO. Due to inconsistencies in
the dRDF feed, spikes in the CO concentration were observed during these runs, but at varying
levels as shown in Figure 8. For each test, the average and the standard deviation in the CO
concentration, and the corresponding PCDD and PCDF yields are provided in the figure. For the
run with a higher CO spike, yields are higher, underlining the importance of burn quality in
reducing PCDD and PCDF formation. Suppression of PCDD and PCDF formation during coal
co-fired runs, however, could not be attributed to improvements in burn quality because, on
average, the coal co-fired runs had higher CO values than RDF-only runs.
As discussed earlier, coal co-firing, on an average, can reduce PCDD and PCDF formation. The
5
-------
coal co-fired results shown in Figure 6 come from runs where the "memory effects" were
minimized by flushing the facility with a high flow of air before each run. As an additional check,
a pair of consecutive tests were run to determine if the PCDD and PCDF formation increases as
soon as the coal feed is stopped. Thus, immediately after a coal co-fired test, the coal feed was
turned off and sampling for the latter run was performed as soon as the flue gas composition
became steady. Results from these two tests are shown in Figure 9. When the coal feed is
stopped, the yields increase immediately, retaining, especially for the PCDF, the same congener
distribution.
CONCLUSIONS AND RECOMMENDATIONS
For commercial dRDF combusted in the 0.6 MW stoker combustor, a majority of the PCDD and
PCDF formation took place between 600 and 300°C within about 0.5 sec.
Co-firing Illinois #6 coal with dRDF reduced PCDD and PCDF formation substantially. This
reduction appears to be uniform across the entire congener range. Injection of Ca(OH)2 sorbent
was nearly as effective as coal co-firing.
Co-firing coal with waste-derived fuel is a candidate technology for reducing PCDD and PCDF
emissions in commercial waste burning facilities. Further development on a full-scale facility is
needed to demonstrate field applicability.
ACKNOWLEDGMENTS
This work was also co-sponsored by Illinois Clean Coal Institute (ICCI) (Project Manager: Ken
Ho) and the National Renewable Energy Laboratory (NREL) (Project Monitor: Phil Shepherd).
Logistical support from Richard Valentine (U.S. EPA/APPCD) is greatly appreciated. Suh Lee,
Joey Valenti, Russell Logan, Scott Moore, John Foley, Dennis Tabor, and Ann Preston (Acurex
Environmental Corporation), and Jeff Ryan (U.S. EPA/APPCD) provided extensive technical,
sampling, and analytical assistance.
DISCLAIMER STATEMENT
Any opinions, findings, conclusions, or recommendations expressed herein do not necessarily
reflect the view of the ICCI, who sponsored this work in part. A detailed disclaimer from ICCI is
available upon request.
REFERENCES
1. B.K. Gullett, K.R. Bruce and L.O. Beach, "Effect of Sulfur Dioxide on the Formation
Mechanism ofPolychlorinated Dibenzodioxin and Dibenzofuran in Municipal Waste
Combustors," Environ. Sei. Technol.. 26(10): 1938 (1992).
2. K. Raghunathan and B.K. Gullett, "Role of Sulfur in Reducing PCDD and PCDF Formation,"
Environ. Sci. Technol.. 30(6): 1827 (1996).
6
-------
3. R. Addink and K. Olie, "Mechanisms of Formation and Destruction of Polychlorinated
Dibenzo-/?-dioxins and Dibenzofurans in Heterogeneous Systems," Environ. Sci. TechnoL 29(6):
1425 (1995),
4. R.D. Griffin, "A New Theory of Dioxin Formation in Municipal Solid Waste Combustion,"
Chemosphere. 15, 1987 (1986).
5. R.L. Lindbauer, F. Wurst and T. Prey, "Combustion Dioxin Suppression in Municipal Solid
Waste Incineration with Sulfur Additives," Chemosphere. 25, 1409 (1992).
6. H. Ogawa, N. Orita, M. Horaguchi, T. Suzuki, M. Okada and S. Yasuda, "Dioxin Reduction
by Sulfur Component Addition," Chemosphere. 32, 151 (1996).
7. Method 23, Title 40 Code of Federal Regulations. Part 60, Appendix A, U.S. Government
Printing Office, Washington, DC (1991).
7
-------
Table I. Analyses of the densified refuse-derived fuel (dRDF) and Illinois #6 coal
dRDF
Coal
Proximate analysis (%)
Moisture
7,33
14.96
Ash
8.46
8.32
Volatile matter
71.20
33.51
Fixed Carbon
12.94
43.21
Ultimate analysis (%)
Moisture
7.33
14.96
Carbon
41.74
60.25
Hydrogen
5.56
4.22
Nitrogen
0.58
1.09
Sulfur
0.09
3.36
Ash
8.46
8.32
Oxygen (by difference)
36.20
7.80
Chlorine
0.22
0.09
Calcium
1.35
n.m.
Copper
0.002
n.m.
kJ/kg
16,408
25,257
(Btu/lb)
(7,059)
(10,866)
n.m. not measured
8
-------
Tabic 2. Average test conditions in the MFC tests
Parameter Average
02 (%) 12.9
C02 (%) 6.6
CO(ppm) 12
H20 (%) 8.8
HC1 (ppm)
without HC1 doping 58
with doping 206
S02 (ppm)
dRDF runs 6
dRDF-coal runs 175
Flue gas flow rate (std m3/min) 16.3
Residence time (sec)
Port B (convective section) 0.91
Port A (pre-baghouse) 2.42
Sampling temperature (°C)
Port B 341
Port A 171
Note: All concentrations are dry, not corrected to 7% 02.
Residence time based on t=0 at 650°C.
9
-------
COAL
FEEDER
SORBENT
RADIANT
FURNACE
ONVECTlyE
ECTIO
STOKER
PORT "B"
PORT "A"
STACK
COOLING BAGHOUSE SCRUBBER
WATER
UNDERFIRE AIR
-------
400
E
Q
W
"O
£
300
¦o
©
LL
Q
O
D_
200
+
a
a
o
CL
cd
o
100
o
CO
o
0
Port B CD Port A
LOW
dRDF feed rate
Figure 3. Effect of dRDF feed rate on PCDD+PCDF yield.
HIGH
LL
O
cr
"O
O)
~CT>
c
4,000
3,000 -
*>>
U_
O
2 2,000
Q
Q
O
CL
o
o
-------
dRDF only
dRDF & Ca
'dRDF & coal
TeCDD PeCDD HxCDD HpCDD OCDD
PCDD congeners
U- 500
a
cc
~o
en 400
o 200
o>
cd
i_
CD
£ 100
dRDF only
dRDF & Ca
7/dRDF & coal
TeCDF PeCDF HxCDF HpCDF OCDF
PCDF congeners
Figure 5. Effects of coal co-firing and sorbent injection on PCDD and PCDF congener yield.
-------
2,000
IX.
Q
CC
"a
05
~Q)
c
a>
cn
CO
Cti
1,500
1,000
500
0
s;
fiX^-X
~££L4X
tetra-octa PCDD
tetra-octa PCDF
dRDF only
dRDF & coa!
dRDF & Ca
Figure 6. Effects of coal co-firing and sorbent injection on total PCDD and PCDF yield.
6000
lT
Q
5000
D5
O)
c
^ 4000
Q
">>
LL
D
O
Q_
+
Q
g 2000
CL
-------
CO cc
1,200
1,000
800
600
400
200
0
1,200
1,000
800
600
400
200
0
(ppm)
CO:
avg
7
std dev:
14
PCDD+PCDF
(ng/kg dRDF):
Port B = 2,306
Port A = 2,048
CO:
avg = 28
std dev = 71
PCDD+PCDF
(ng/kg dRDF):
Port B = 21,607
Port A = 15,488
40 60
time (minutes)
100
12
igure 8. Comparison of CO profiles between two HIGH dRDF feed runs.
-------
dRDF only
dRDF & coal
TeCDD PeCDD HxCDD HpCDD OCDD
PCDD congeners
dRDF only
dRDF & coal
TeCDF PeCDF HxCDF HpCDF OCDF
PCDF congeners
Figure 9. Effect of stopping the coal feed on PCDD and PCDF congener yield.
------- |