vvEPA
United States
Environmental Protection
Agency
Industrial Environmental Research
Laboratory
Research Triangle Park NC 27711
Research and Development
EPA-600/S7-82-027 Sept. 1982
Project Summary
Rates and Equilibria of
Devolatilization and Trace
Element Evolution in
Coal Pyrolysis
R. M. Felder, C.-C. Kau, J. K. Ferrell, and S. Ganesan
A laminar flow furnace was used to
study the kinetics of devolatilization
and evolution of four elements —
sulfur, arsenic, lead, and mercury — in
the pyrolysis of pulverized coal in
nitrogen. The quantities varied in this
study included pyrolysis time (170
msec-2 sec), reactor temperature
(400-900°C), and parent coals (two
low-swelling subbituminous coals and
one medium-swelling bituminous
coal). Weight losses were estimated
using ash as a tracer. A Fisher Sulfur
Analyzer was used to analyze the
sulfur content of the feed coal and
spent char, and atomic absorption
spectrophotometry was used to an-
alyze for the remaining three elements.
Weight losses of the coals studied
increased significantly with time and
temperature and approached different
asymptotic values at different tem-
peratures. The devolatilization rates
and asymptotic weight losses of the
subbituminous coals were lower than
those of the bituminous coal. The
elements were released in significant
quantities during fast pyrolysis as
temperature and time increased, with
the elemental release from the bitu-
minous coal proceeding more rapidly
and to a greater extent than that from
the lower rank coals. The emissions of
arsenic and lead in subbituminous coal
and lead in bituminous coal were
found to be proportional to the total
volatile matter release, and the emis-
sions of mercury and arsenic in
bituminous coal were found to be
proportionally greater than the total
volatile matter release. The amount of
sulfur released was found to be
proportional to the amount of dry-ash-
free volatile matter released for both
coals.
Five kinetic models for devolatiliza-
tion were tested. Of these, a two-
parallel-reaction model due to Kobay-
ashi provided the best correlation of
the experimental data for the sub-
bituminous coal, and a distributed
activation energy model due to An-
thony was found best for the bitumi-
nous coal. A single first-order reaction
model for elemental release coupled
with a model for equilibrium release
derived from batch pyrolysis experi-
ments was used to correlate elemental
release data.
This Project Summary was devel-
oped by EPA's Industrial Environ-
mental Laboratory, Research Triangle
Park, NC, to announce key findings of
the research project that is fully
documented in a separate report of the
same title (see Project Report ordering
information at back).
Introduction
Coal represents about 70 percent of the
recoverable fossil fuel resources in the
U.S., and may eventually become our
-------
pri ncipal source of hydrocarbon fuel and
chemical feedstocks. National recogni-
tion of the enormous immediate and
long-range potential of coal has resulted
in substantial efforts directed toward
the development of an economically and
environmentally viable coal utilization
technology.
Coal conversion to clean gaseous and
liquid products is especially attractive
for alleviating the growing restrictions
on existing consumption patterns. The
main approaches to the conversion of
coal to gaseous fuels are steam-oxidant
gasification, hydrocarbonization, and
pyrolysis (devolatilization). In all three
processes, pyrolysis plays an important
role. When coal is heated in any
atmosphere, it releases volatile products.
In most coal conversion processes coal
is pulverized and subjected to rapid
heating which causes a significant rate
of devolatilization. The amount of
volatile matter produced depends
mainly on the type of coal, the tempera-
ture of the surrounding gas, the heating
rate, the reacting pressure, and the size
of the coal particles.
The effluent gases from a coal
conversion system contain elements in
minor or trace amounts which are toxic
to humans, animals, and plants, or
cause severe corrosion in coal-fired
boilers and gas turbines. These elements
may exist as vapor and/or in association
with particles in the effluent gases.
Studies have been published on the
occurrence and distribution of trace
elements in different coals and in
effluents from coal-fired steam plants
and coal gasification plants. The data
obtained in these studies provide a good
qualitative picture of the pathways of
trace and minor element constituents in
coal gasification; however, they do not
provide the detailed information on
reactor conditions needed to predict the
thermodynamics and kinetics oithe re-
lease of these elements.
To estimate the types and quantities
of hazardous emissions released in the
pyrolysis stage of steam/oxygen gasi-
fication or hydrogasification, it is
necessary to measure and model the
rates and equilibria of both devolatiliza-
tion and evolution of specific species. In
this study, a laminar-flow furnace react-
or was used to determine devolatiliza-
tion and trace element release equilib-
ria for rapid pyrolysis at temperatures
up to 900°C. Several existing devolatil-
ization models were fit to the data by
nonlinear regression, and the results
were used to evaluate and compare the
models. In addition, trace element
evolution data were fit with a first-order
rate law, and the results were used to
draw inferences regarding the vola-
tilities of these elements.
Experimental
Preparation of Coal Samples
Experiments were performed on two
coals: North Barber No. 8 seam HVC
subbituminous coal, Navajo Mine New
Mexico (NB8); and Western Kentucky
No. 11 HVB bituminous coal (WK11).
Table 1 summarizes data on these coals.
Coal samples were crushed in a Bico
pulverizer with ceramic plates and in a
small porcelain ball mill. The pulverized
coal was then size-graded with U.S.
standard sieves and a mechanical
shaker; the 325-400 mesh size fractions
of NB8 and WK11 coals were retained
for use in the experiments.
Experimental Equipment
A laminar-flow furnace reactor used
by Agreda (1979) was also used in this
study. It was designed to:
• Feed coal particles into a reaction
zone of known length (residence
time).
• Raise the temperature of the coal
particles as rapidly as possible (i.e.,
at a rate greater than 104°C/sec)to
the predetermined pyrolysis tem-
perature.
• Quench all the pyrolysis reactions
as rapidly as possible following col-
lection to prevent conversion
through secondary reactions.
• Separate solid (i.e., char) from gas-
eous (i.e., volatile matter plus
carrier gas) products and prevent
volatile matter from depositing on
the surface of the solid.
The furnace is made of alumina; it has
an ID of 8 cm and is 1 mlong. The liner is
an alumina tube, 7 cm ID and 0.75 m
long. Heat is supplied by a three-zone,
4000-W, 230-V Thermocraft furnace,
with a heated length of 47 cm.
In operation, a small carrier flow of
cold nitrogen carries the size-graded,
finely ground coal from the feeder-
hopper into a preheated nitrogen
stream passing down through the
vertical furnace tube in laminar flow.
The furnace tube is held at the same
temperature as the preheated main gas.
The small carrier gas flow mixes rapidly
with the hot gas stream, allowing the
particles to be brought rapidly to furnace
temperature.
Coal particles travel in a narrow
laminar flow streamline along the axis
of the furnace and are aspirated into a
water-cooled collector. The collector's
tapered entry accelerates the aspirated
gases to a high-velocity turbulent flow.
The consequently high heat transfer
rates between the gas and collector wall
lead to rapid cooling of the gas and
quenching of the devolatilization reac-
tions. Residence time can be varied
from 50 to 2000 msec by adjusting the
distance between the feeder and the
collector and/or adjusting the main gas
flow rate. The temperature of the
furnace and gas stream can be adjusted
up to 1000°C. The pressure of the
furnace can be adjusted from atmos-
pheric to 0.4 atm (gauge).
A water-jacketed stainless steel
collector is used to collect the char
particles following pyrolysis. A thermo-
stated reservoir (90-100°C) supplies
cooling water to the collectors. The hot
water cooling system prevents tars from
condensing on the chars, which would
seriously affect the precision of weight
Table 1
Coal
NM
No. 8
W.K.
No. 11
NM
No. 8
W.K.
No. 11
Coal Characterization Data
Coal Proximate
Characterization Analysis
Code
NB8
WK11
C
55.18
60.07
Rank
HVC
HVB
Ultimate
Analysis (°/<
N
1.21
1.775
FSI
0.5
2.5
y
H
4.12
4.28
Moisture Ash
Volatile
10.09 18.32 33.80
6.34 15.02 34.67
Sulfur
Forms (%)
S Pyritic
0.73 0.31
4.64 2.63
Sulfatic
0.0
0.14
Fixed
37.79
43.97
Organic
0.42
1.87
-------
loss estimates. An electromagnetic
vibrator is used at the bottom of the
collector to prevent char particles from
sticking to the collector walls.
Experiments and Results
Two coals were pyrolyzed in the
laminar flow reactor (seeTable 1): North
Barber No. 8 (NB8, low-swelling New
Mexico subbituminous coal) and Western
Kentucky No. 11 (WK11, medium-
swelling bituminous coal). A total of 47
runs were carried out, the first 28 of
which were with NB8 coal at 500-
900°C.
To calculate weight losses, ash was
used as a tie element between the feed
coal and spent char, with a correction
applied to account for ash losses. The
weight losses are shown in Figures 1
and 2 for NB8 and WK11 coals,
respectively, NB8 (low-swelling coal)
exhibited a much lower devolatilization
rate than WK11 (medium-swelling coal)
at a given temperature. At higher
temperatures, equilibrium was achieved
within tens of milliseconds for WK11
coal; NB8 coal required hundreds of
milliseconds.
Modeling
Experimental weight loss data ob-
tained for NB8 and WK11 coalsanddata
for Montana Rosebud subbituminous
coal obtained by Agreda et al. (1979)
were correlated with five models: a
single first-order reaction model (both
isothermal and nonisothermal reactor
operation being considered separately),
Badzioch and Hawksley's (1970) iso-
thermal model, Kobayashi's (1976)
nonisothermal two-reaction model, and
Anthony's (1974, 1975, 1976) noniso-
thermal infinite parallel reactions
model. For the last three models, the
parameters proposed by the model
developers were first used; then, for all
models, the parameters were adjusted
to obtain least-squares fits to the data.
In the latter calculations, an initial set of
model parameters was chosen, weight
losses were evaluated at a series of
times and temperatures corresponding
to experimental data points, and the
unweighted sum of squares of residuals
SSE = [V%Xp - V*modeif
(D
was calculated. Then the model param-
eters were systematically varied to
determine the set that minimized SSE.
For the isothermal first-order model, the
parameters were estimated by logarith-
mic transformation followed by linear
regression; for the others, either a
50
40
I
,30
20
10
D
D 900°C
• 800°C
O 700°C
600°C
O 500°C
200 400 600 800
Residence Time, m sec
Figure 1. Dry ash-free weight loss for NB8 coal.
1000
1200
Gauss-Marquardt nonlinear regression
or Pattern Search algorithm was used,
depending on whether or not derivatives
of the expression for V* with respect to
the model parameters could be deter-
mined analytically.
Equilibrium Devolatilization
An important parameter of the models
is the equilibrium or asymptotic weight
loss at the reaction temperature. This
quantity can be obtained either from
batch pyrolysis data or from long
residence time data in the laminar flow
reactor. Duhne (1977) has shown that
asymptotic equilibrium values for
physical and chemical processes can be
estimated from an equation of the form
V* = —+ V,*, V* X>.67V,* (2)
where V* is the weight loss on a dry ash-
free (d.a.f.) basis, V(* is the long-time
asymptotic limit of this quantity, IR is the
residence time, and d = a constant. V*,
plotted versus 1/tp, should approach a
straight line at low values of the
abscissa with the intercept equal to V(*.
Plots of this type were generated for
Montana Rosebud (MRS) (Agreda et al.,
1979) and WK11 coals. The resulting
values of Vf* are listed in Table 2, along
with batch pyrolysis equilibrium values
obtained by Agreda et al. (1979). NB8
coal did not come close enough to
equilibrium in the LFR experiments for
the given procedure to be used. The batch
values shown in Table 2 were used for
all kinetic modeling for this coal.
The fact that the asymptotic and
equilibrium weight loss estimates
shown in Table 2 differ isnotsurprising.
In the case of the MRS coal, the
equilibrium values appear to be higher
than the asymptotic values. This result
is consistent with the finding of Morton
(1979), who reviewed the fast pyrolysis
data of several researchers and noted
that a quasi-equilibrium is reached
before the particles are quenched, but
that additional devolatilization occurs
when the quenched char particles are
reheated in a batch reactor.
With WK11 coal, the equilibrium
values stay higher than the asymptotic
values at temperatures below 700°C.
This can be explained by the caking
characteristics of this coal. Caking coals
-------
D 900\°C
D SOO°C
700°C
O 600°C
D 400°C
200
400 600 800
Residence Time, m sec
1000
1200
Figure 2. Dry ash-free weight loss for WK11 coal.
often devolatilize with the formation of
bubbles and then resolidify during batch
pyrolysis, while in fast pyrolysis the
structure of the coal particles is
deformed as the volatile matter is
removed. Therefore, after caking (about
700°C), the fast pyrolysis asymptotic
weight losses should be higher than the
batch pyrolysis equilibrium values.
The values of Vf* determined above
(from the LFR measurements for MRS
and WK1 1 and from batch measure-
ments for NB8) were fitted versus T with
a fourth-order polynomial:
VfV100 = Bo
B4T'4
B3T'3
(3)
where T' = T(°C)/100. The coefficients
of this polynomial for the three coals are
shown in Table 3.
Devolatilization Rate Models
The devolatilization rate data collected
for all three coals studied were fitted
with the five cited models. Some of the
models presume a fixed (temperature-
independent) equilibrium volatiles yield.
Fits were obtained for these models,
both with and without this assumption;
for the latter, equilibrium yields (deter-
mined as described above) were substi-
tuted. For isothermal models, the reactor
temperature was assumed constant
but the residence time was corrected to
account for the time required to heat the
particles following their entry into the
reactor. For nonisothermal models, the
true temperature-time history of the
particles in the reactor was substituted
into the model equations. For the
Badzioch-Hawksley, Kobayashi, and
Anthony models, the fits obtained by
regression were compared with the fits
provided by the model parameters
suggested by the proponents.
The following cases were examined:
• Single first-order reaction — iso-
thermal, fitted parameters.
• Single first-order reaction — non-
isothermal, fitted parameters.
• Badzioch-Hawksley model — iso-
thermal, variable equilibrium yield,
fitted and original parameters.
• Kobayashi model — nonisothermal,
fixed and variable equilibrium
yield, original and fitted param-
eters.
• Anthony model — nonisothermal,
fixed and variable equilibrium
yield, original and fitted param-
eters.
Complete descriptions of the models
and the modeling procedures used, and
tabulations and plots of the results, are
given in the full report. The paragraphs
that follow discuss the principal findings
and summarize the conclusions.
Comparative listings of the model
parameters (both original and fitted) and
the sum of squares of residuals for all
five models and all three coals are given
in Tables 4-6. As indicated by the
smaller values of SSE, the fitted
isothermal models (single first-order
reaction model and Badzioch-Hawksley
model) provide reasonable correlations
of the experimental data. However, the
arbitrary specification of an isothermal
reaction time limits the flexibility
required to describe nonisothermal
pyrolysis in any reactor, and leads to
anomalously low activation energies.
Therefore, these two models were
eliminated from further consideration.
The results obtained for the models of
Anthony (1974) and Kobayashi (1976)
are best understood in light of the
differences between these two models:
• Anthony postulates an infinite
number of parallel first-order
reactions, with a common fre-
quency factor for all rate laws and
a Gaussian distribution of activa-
tion energies. Kobayashi post-
ulates two parallel first-order
reactions, with different stoichio-
metric extents, frequency factors,
and activation energies for each
reaction.
• Anthony's model has three adjust-
able parameters: the frequency
factor and the mean and standard
deviation of the Gaussian distribu-
tion of activation energies. How-
ever, indications are that the
model is relatively insensitive to
the value of the frequency factor,
so that this parameter may be set
to an arbitrary value,andthe fitting
performed by adjusting only two
variables. Kobayashi's model has
six adjustable parameters: two
stoichiometric extents, two fre-
quency factors, and two activation
energies.
• The data on which Anthony's
model was based were obtained in
an electrically heated wire grid
reactor; experiments by Suuberg
et al (1978) that appeared to con-
firm the validity of the model were
4
-------
Table2. Equilibrium and Asymptotic D.A.F. Weight Losses for WK11 and MRS
Coals
Asymptotic(O) Weight Loss, %
Coal
WK11
WK11
WK11
WK11
MRS
MRS
MRS
MRS
MRS
NB8
NB8
NB8
NB8
NB8
Table 3.
Coal
MRS
WK11
NB8
Temperature, °C
400
600
800
900
300
400
600
800
900
300
400
600
800
900
Equilibrium(X)
0
X
0
X
0
X
0
X
0
X
0
X
0
X
0
X
0
X
X
X
X
X
X
Coefficients of the Fourth-Order Polynomial for Asymptotic
Bo 81
-1.783 1.1411
-4.291 2.691
1.052 -0.7150
Bz B3
-0.2590 0.02479
-0.5927 0.05477
0.1615 -0.01367
V,
2.94
12.94
17.65
38.76
49.90
44.67
52.90
46.50
0.85
3.27
6.76
11.63
13.44
33.42
37.40
41.73
39.27
46.85
3.27
11.63
33.42
41.73
44.23
Weight Loss
B*
-0.0008233
-0.001775
0.0004044
performed on the same type of ap-
paratus. The difficulties associated
with this approach are uncertainty
of heating time, inability to quench
the pyrolysis rapidly, and resolid-
ification of volatile matter during
cooling. On the other hand, Ko-
bayashi's experiments were per-
formed in a laminar flow reactor,
similar to the one used in the
present study. Numerous prob-
lems arise in experiments of this
type, such as ash loss errors and
tar recondensation on the collect-
ed char. The principal advantages
offered by the LFR are precise con-
trol over residence time and ex-
tremely low heating and quench
times compared to those char-
acteristic of the heated grid.
A comparison of the predictions of
both models is shown in Figure 3. A coal
with the properties of Montana Rosebud
subbituminous is used as the basis of
comparison. Devolatilization curves
predicted by both models are shown for
600°C, corresponding to a relatively low
pyrolysis temperature, and 800°C, a
moderate-to-high temperature. The
model parameters are those suggested
by Anthony (1974) and Kobayashi
(1976). The predictions of the two
models are seen to be quite different:
Anthony's model predicts that equilib-
rium is reached in a few hundred milli-
seconds, while Kobayashi's model pre-
dicts an increasing trend even at long
residence times. At all temperatures,
Anthony's model predictions are con-
sistently above Kobayashi's model pre-
dictions.
The observed differences are consis-
tent with the previous observations
regarding the two models. Anthony let
his coal samples remain in the heated
grid reactor for a few seconds after the
heater was turned off. Therefore, his
weight loss values should reach equilib-
rium during this period of time and stay
higher than those of Kobayashi.
Results obtained for the parameters
determined by fitting Anthony's model
to our data have been compared with
results based on Anthony's original
model parameters. The fitted parameters
for the medium-swelling high rank coal
(WK11) are close to Anthony's values,
although the spread of the fitted
activation energy distribution is narrow-
er than Anthony's. On the other hand,
contradicting Anthony's prediction, the
fitted parameters for NB8 and MRS
coals suggest that a single reaction with
a low activation energy dominates the
decomposition of low-swelling low-
rank coals. However, this result does
not eliminate the possibility that the
pyrolysis may be better described as a
finite set of decomposition reactions with
different activation energies.
Kobayashi's model has this possibility.
However, for this model a* well, the
values of the fitted parameters for all
three coals imply that a low activation
energy first-order reaction dominates
the overall pyrolysis between 400 and
900°C. The rate constants for the first
reaction (Ki) at 900°C obtained from the
parameters in Tables 4-6 are one to two
orders of magnitude greater than those
for the second reaction (K2), with the
disparity increasing as the temperature
decreases and the coal rank increases.
Thus, adding a second reaction with a
high activation energy makes little
difference for a medium-swelling high-
rank coal like WK11 and essentially no
difference for low-swelling low-rank
coals like MRS and NB8.
The conclusion is that at below
1000°C, pyrolysis of MRS and NB8
coals (and presumably of other low-
swelling low-rank coals) is well described
by a single first-order decomposition
rate law; replacing this law by a discrete
set of reactions with higher activation
energies does not improve the fit
significantly, and replacing it with a
continuous distribution of high activation
energies makes the fit worse. This
conclusion basically supports the
findings of Kobayashi and is inconsistent
with those of Anthony and Suuberg. On
the other hand, Anthony's model is
quite consistent with our results for the
pyrolysis of a medium-swelling high-
rank coal, although the standard devia-
tion of our activation energy distribution
-------
i
I
Ci
•o
-------
Table 4.
Model
Single
First -Order
Badzioch
and
Hawk si 'ey
Comparison of Model Parameters and SSE Values (MRS)
Isothermal Equivalent Yield Parameters
Yes No Fixed Variable Fixed Fitted
X XX
XXX
X XX
X XX
y
s\
y
A
x
X X
Parameter
Values
Bo = 93. 5(1 '/sec)
Eo = 8. 7(kcal/mole)
B0= 7677(17 sec)
Fo= 18(kcal/mole)
K-< = 0.001031
K2 = 589
Q = 1.8
A =8.36* 104(l/sec)
B - 8.9(kcal/mole)
K, = 0.001031
Kz = 589
Q = 1.8
A =906(l/sec)
P = 6 37(kcal/mole)
32 = 1.0
5, = 2 x I0s(l/sec)
82=1.3* 107(l/sec)
F, = 25(kca//mole)
£2 = 40(kcal/mole)
SSE
413
242
2184
194
270
590
Kobayashi
32 = 1.0
5, = 6030(l/sec)
F, - 18(kcal/mole>
B2=6.6* 106(l/sec)
E2 - 40(kcal/mole)
230
X
X
Anthony
X
Ko= /.Ox 10'°(//sec)
V* = 47%
Eo = 48.7fkcal/mole)
s = 9.36(kca//mo/e)
Ko = /.Ox W'Wsec)
£o = 48.7(kca//mo/e)
s = 9.36(kcal/mo/e)
K0= /.Ox W13(l/sec)
£0= 18.9(kca//mo/e)
s = 1.0 x JO~3(kcal/mole)
1079
588
227
ments. The solid lines shown in Figures
4 and 5 are the predictions of this model.
No fitting was done for the other
elements in NB8 coal or for any
elements in WK11 coal because of the
high degree of scatter of the data.
Details of the calculations and results
are given in the full report.
Conclusions
• Several models can be used to fit
most of the devolatilization rate
data obtained in this study with
more or less equivalent degrees of
success; in fact, a model consisting
of a single first-order decomposi-
tion reaction is as good as any of
the more complex models for the
two low-rankcoalsstudiedand not
too much worse than the best of
the models for the bituminous
coal. Moreover, the single reaction
model parameter values for both
low-rank coals are quite similar:
the pre-exponential factor is roughly
7750(s~1), and the corresponding
activation energy is 18 kcal/mole.
All fitting studies support the
notion that the kinetics of low-rank
coal pyrolysis are dominated by a
single first-order reaction with a
low activation energy. This contra-
dicts the findings of Anthony
(1974) and Suuberg et al. (1978),
whose results support the con-
clusion that pyrolysis is in reality a
series of parallel reactions with
activation energies distributed
over a much higher range. On the
other hand, the data obtained in
this study for the bituminous coal
are quite consistent with Anthony's
picture, albeit the spread of the
activation energy distribution is
narrower than that proposed by
Anthony.
• Kobayashi's model with its original
parameters does a reasonable job
of representing the pyrolysis of the
two low-rank coals studied. Im-
provements can be made by ad-
-------
Table 5. Comparison of Model Parameters and SSE Values (NB8)a
Isothermal Equivalent Yield Parameters
Model Yes No Fixed Variable Fixed Fitted
X XX
Single
First -Order XXX
X XX
Badzioch
and
Hawksley
X XX
X X
X
X X
Parameter
Values
Bo = 353 (I/ sec)
E0 = 11.3(kcal/mole)
B0= 7853(1 '/sec)
E0= 18 (kcal/mole)
K, = 255* 10~3
K2=471
Q = 1.2
A =8.36* 10" (//sec)
B - 8.9(kcal/mo/e)
/C, = 2.55x 70~3
K2=471
Q = 1.2
A =4335 (//sec)
B = 84. (kcat/mo/e)
a2 = 1.0
Bi = 2.0* 10s (//sec)
£, = 25 (kcal/mole)
B2= 7.3x JO7 (I/ sec)
£2 = 40 (kcal/mole)
SSE
397
154
11281
134
722
679
Kobayashi
az = 1.0
B, = 7530 (//sec)
E,= 18 (kcal/mole)
B2= 1.44* JO6 (I/sec)
Ez=40(kcal/mole)
160
Anthony
K0= /.Ox J0'°(l/sec)
l/,*= 43%
£o = 49.38 (kcal/mole) 1341
s =0.36 (kcal/mole)
K0= 1.0 x 10 (I/sec)
£o = 48.7 (kcal/mole) 953
s =9.36 (kcal/mole)
K0= /.Ox JO'3 (I/sec)
£0= 18.9 (kcal/mole) 246
s =O.16 (kcal/mole)
"Batch equilibrium yields used for V*
justing the parameter values, but
not to the extent that the original
values are discredited. Anthony's
model appears to be superior for
describing the pyrolysis of high-
rank coals.
• Anthony's model with its original
parameters fails to fit the data for
any of the coals studied. Consider-
able improvements are obtained
for bituminous coal by allowing the
asymptotic weight loss to vary with
temperature, and decreasing the
standard deviation of the activa-
tion energy distribution. However,
for low-rank coals, the model is
consistently outperformed by a
single first-order decomposition
reaction model with a low activa-
tion energy.
The contradiction between the re-
sults obtained here for low-rank
coals and those of Anthony (1974)
and Suuberg, et al. (1978) has not
yet been resolved, although it may
be attributable to the difference be-
tween the operating character-
istics of laminar flow reactors
(such as those used in this study
and by Kobayashi) of electrically
heated grid reactors (such as those
used by Anthony and Suuberg).
Sulfur, mercury, lead, and arsenic
are released in significant quan-
tities during fast pyrolysis; the re-
lease is completed in tenths of
seconds. The release of these ele-
ments from low-swelling coal is
quite different from that in medi-
um-swelling coal, with elements
being released much more rapidly
and to greater extents from the lat-
ter.
The emissions of arsenic and lead
in NB8 coal and lead in WK11 coal
are proportional to the total volatile
matter release, and the emissions
of mercury and arsenic in WK11
8
-------
Table 6.
Model
Single
First-Order
Badzioch
and
Hawksley
Comparison of Model Parameters and SSE Values (WK1 1)
Isothermal Equilibrium Yield Parameters
Yes No Fixed Variable Fixed Fitted
X XX
XXX
X XX
X XX
Parameter
Values
B0=6320(l/sec)
E0= 14.2 (kcal/mole)
B0= 16035(1 /sec)
fo= 19. 12( kcal/mole)
K,= 1.930* 10'3
K2=676
Q = 1.8
A = 6. 1 x W7 (I/ sec)
B = 16.3 (kcal/mole)
K^ = 7.338 x W3
K2=676
Q = 1.8
A = 6. 1 x 107 (I/ sec)
B = 16.3 (kcal/mole)
SSE
597
439
1143
513
X
X
X
X
X
a2= 1.0
fl, = 2.0 x 10s (I/sec)
B2= /.3x W7 (I/ sec)
Et = 25 (kcal/mole)
E2= 40 (kcal/mole)
7679
4681
Kobayashi
a2 = 1.0
fl, = 4.8 x 104(l/sec)
E1 = 18 (kcal/mole)
82= 5.0 x 106 (I/sec)
E2=40 (kcal/mole)
394
Anthony
K0= /.Ox W3 (I/sec)
l/*,= 48%
E0= 54.8 (kcal/mole)
s = 17.2 (kcal/mole)
K0= 1.67* 10'3 (I/sec)
f0= 54.8 (kcal/mole)
s = 17.2 (kcal/mole)
E0= 55.2 (kcal/mole)
s =6.18 (kcal/mole)
2842
670
243
coal are proportionally greater
than the total volatile matter re-
lease. The amount of sulfur re-
lease was nearly directly propor-
tional to the amount of dry-ash-
free volatile matter released for
NB8 and WK11 coals. This con-
firms the findings of Agreda et al.
(1979).
The kinetics of sulfur and lead re-
lease for NB8 coal can be modeled
using a first-order reaction model
coupled with a model for equilib-
rium release derived from batch
pyrolysis experiments. The lack of
precision in the retentions of other
volatile trace elements prevented
the meaningful determination of
kinetic parameters for the release
of these elements.
Recommendations
The most important addition to the
work performed in this study is the
sampling and chromatographicanalysis
of the gases produced in both the batch
and laminar flow reactors. Gas samples
should be analyzed for major compo-
nents by thermal conductivity detection,
for trace hydrocarbons by flame ioniza-
tion detection, and for trace sulfur gases
by flame photometry. The results will
elucidate the mechanism of pyrolysis,
and will shed light on the presently
unresolved contradiction between the
results favoring the single first-order
decomposition rate law and the distrib-
uted activation energy model.
The procedures for trace element
sampling and analysis should be refined
to afford a greater degree of experimental
precision, and the modeling efforts
commenced in this study should be
continued. The list of elements studied
should be extended to cover a wider
range of species that are both volatile
and potentially hazardous, such as
selenium, beryllium, antimony, and
vanadium.
References
1. Agreda, V. H., "Devolatilization
Kinetics and Elemental Release in
the Pyrolysis of Pulverized Coal,"
-------
.§
100
90
75
60
45
30
90
75
60
45
30
900°C
WK11
D
D
l
es, Plenum Press, New York, NY,
1979.
9. Kau, C.C., "Devolatilization Kinetics
and Elemental Release in the
Pyrolysis of Pulverized Coal," M. S.
Thesis, Department of Chemical
Engineering, North Carolina State
University, Raleigh, NC, 1980.
10. Kobayashi, H., Sc.D. Thesis, De-
partment of Mechanical Engineer-
ing, Massachusetts Institute of
Technology, Cambridge, MA, 1976.
11. Suuberg, M., W. A. Peters, and J. B.
Howard, "Product Composition and
Kinetics of Lignite Pyrolysis," Ind.
Eng. Chem. Process Design Devel-
op., 17, p. 1 1978.
300 600 900
Residence Time, m sec
1200
Figure 4. Sulfur retention in LFR chars.
Ph.D. Thesis, North Carolina State
University, Raleigh, NC, 1979.
2. Agreda,V. H.,R. M. Felder,and J. K.
Ferrell, "Devolatilization Kinetics
and Elemental Release in the
Pyrolysis of Pulverized Coal," U.S.
Environmental Protection Agency
Report EPA-600/7-79-241 (NTIS
PB 80-130222), November 1979.
[Abbreviated version in Coal Pro-
cessing Technology, Vol. VI, CEP
Technical Manual, p. 87 (1980).]
3. Anthony, D.B., Sc.D Thesis, Massa-
chusetts Institute of Technology,
Cambridge, MA 1974,
4. Anthony, D. B., J. B. Howard, H. C.
Hottell, and H. P. Meissner, "Rapid
Devolatilization of Pulverized Coal,"
Fifteenth International Symposium
on Combustion, p. 1313, The Com-
bustion Institute, Pittsburgh, PA,
1975.
5. Anthony, D. B., J. B. Howard, H. C.
Hottell, and H. P. Meissner, "Rapid
Devolatilization and Hydrogasifica-
tion of Bituminous Coal," Fuel, 55,
p. 121 1976.
6. Badzioch, S.,andP G.W. Hawksley,
"Kinetics of Thermal Decomposition
of Pulverized Coal Particles," Ind.
Eng. Chem. Process Design Devel-
op., 9, p. 521 (1970).
7. Duhne, C. R., "Calculating the
Approach to Equilibrium," Chem.
Eng., 84, No. 18, August 29, 1977.
8. Horton, M. D. "Fast Pyrolysis," in L.
D. Smoot and D. T. Pratt (Ed.),
Pulverized-Coal Combustion and
Gasification: Theory and Applica-
tions for Continuous Flow Process-
10
-------
110
100
90
80
70
60
c
.o
I 50
Q)
9)
100
90
80
70
60
SO
40
500°C
900°C
30\
O
o
Q
Q
D
Q
D
I
WK11
O
200 400 600 800
Residence Time, m sec
1000
1200
Figure 5. Lead retention in LFR chars.
-------
100
60
20
100
60
20
0,
B
D
D D
Q ^
D
Q
D
NB8
O
O
,Q
n i
O
Q
WK11
I
700 300 500 700
Residence Time, m sec
Figure 6. Mercury retention in LFR chars.
900
1100
72
-------
700
80
60
40
- wo
a
4)
ec
80
6'0
40
20
D
D
D
D
O
O
D
NB8
D 900°C
d800°C
O 600°C
0 500°C
t> 400°C
WK11
O
O
O
O
D
O
O
D
100 300 500 700 900
Residence Time, m sec
1100 130O
Figure 7. Arsenic retention in LFR chars.
ft. M. Felder, C.-C. Kau. J. K. Ferrell. and S. Ganesan are with North Carolina
State University, Raleigh, NC 27650.
N. Dean Smith is the EPA Project Officer (see below).
The complete report, entitled "Rates and Equilibria ofDevolatilization and Trace
Element Evolution in CoalPyrolysis. "(Order No. PB 82-260 944; Cost: $12.00.
subject to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
OUSGPO: 1982 — 559-092/0523
-------
X c
(/>
fi
Is
CD
c
(/>
CD
o en c o
3P CO» — •
-i. o^ co -••
o • -»»
oi CO -h
to o m o
O C 3 -J
«. r+ < 0.
r- oo -*•
— ' 3 (/>
Q» 3 — •
3 — •
O CO O)
-p» ft — •
n>
(0
o
c*
O
Q>
(£1
(D
85
@j
O
5'
o
5
3
0)
O
I
en
K)
O)
oo
p- Va*^
m
m > T) m
-o
------- |