Chemical Engineering Sciwre Printed in Great Britain.
Vol. 38, No.
1I. pp.19171925.
1983
00092509/83 Pergamon
$3.00 + .OO Press Ltd.
FLUIDIZEDBED COAL COMBUSTION: INBED SORBENT SULFATION MODEL D. C. FEE,* W. IRA WILSON, Argonne National Laboratory,
K. M. MYLES and IRVING JOHNSON
Chemical Engineering Division, U.S.A.
9700 South Cass Avenue, Argonne, IL 60439,
and
LIANGSHIH F,4N The Ohio State University, Departeent of Chemical Engineering, Columbus, OH 43210, U.S.A (Received 22 January
1982; accepted 25 April 1983)
AbstractA model is developed to account for sulfur capture within the sorbent bed of a fluidizedbed coal combuslor. The model is expressed in an analytical form that contains two sorbent specific parameters. The sorbent parameters, which describe the kinetics of the sulfation reaction are obtained from thermogravimetric analysis data on the S&sorbent reaction. Various hydrodynamic and mass transfer properties in relation to the bubbling phenomena are taken into consideration in the model. The predictions of the model which are made without arbitrarv .oarameters. compare favorablv with the experimental data on sorbent performance from large 1
scale experimental Ruidizedbed coal combus&. INTRODUCTION
A key to the widespread commercialization of fluidizedbed coal combustion technology is the ability to accurately predict the amount of sulfur that will be captured by a given limestone sorbent. Because of this strong incentive, numerous methods of prediction sulfur control in fluidizedbed combustors have appeared in the literature [l24]. Much of the previous effort focused on measuring and describing the complex kinltics arising from the closely coupled reactions: (1) diffusion of gasi?ous SO* in the pores of the calcined limestone, (2) reaction of gaseous SO, with calcium oxide to form calcium sulfate and (3) diffusion SOZ through the calcium sulfate product layer. Because each of these processes may be rate controlling under the right circumstances, such a rigorous analysis of the SOZ sorbent reaction often results in mathematical expressions that are too complicated for practical engineering applications. Similar to the model described herein, some models[l, 2,1416,21,24] simplify the mathematics somewhat by using laboratory data on the overall rate of sorbentsulfur dioxide reaction (rather than the individual reactions) as the basis for predicting sorbent performance in a fluidizedbed combustor. However, the simplicity of the SO+orbent reaction is encumbered with a complex analysis of bed fluid mechanics and particle size distribution [21,24] or a much more complex mathematical expression than the moder presented here[l, 2,14161. The Westinghouse model[l, 21 based on TGA data;represents the most satisfactory previous model and is currently the basis for many of the predictions of sorbent performance found in the literature [25,26].
*Author to whom correspondence
should be addressed.
In this paper, we develop a simple analytical expression for calculating the sorbent feed rate necessary to achieve the desired sulfur capture in an atmospheric fluidizedbed coal combustor (AFBC). The equation shows explicitly the sorbent and AFBC parameters which affect sulfur capture. This is advantageous in rapidly screening and selecting candidate sorbents. MODEL
The model of sulfur capture in a fluidizedbed combustor is based on the following assumptions: (1) Laboratoryscale thermogravimetric techniques accurately simulate the ratelimiting mechanism in the sulfur capture reaction CaO(s) t SOZ(g) + 1/20&)*CaSOAs)
(1)
and, therefore, in principle, can be used to predict sorbent requirements in a fluidized bed. The thermogravimetric techniques and subsequent data reduction methods are described in the Appendix. (2) Sulfur removal by a wide particle size distribution in a fluidized bed can be predicted utilizing the sulfur sorption rate data of a narrow particle size fraction representing the average particle size in the bed. (3) The reaction of SO2 with sorbent is first order in SO,. This assumption is supported by data in this work (see Appendix) and in the literature [9, 10,27,28]. (4) The fluidized bed consists of two phases, an emulsion phase and a bubble phase. All sulfur from the coal is uniformly released in the emulsion phase as SOL. (5) The emulsion phase, which contains the solids, is well mixed. (6) Mass transfer between the bubble and emulsion phase is either (a) uniformcorresponding to a slow bubble regime or (b) nonuniformcorresponding to a 1917
1918
D. C. FEE et al.
fast bubble regime. A slow bubble regime[Il,
30321 is characterized by large limestone particles (> 1000 pm) and commensurately high fluidizing velocities. A fast bubble regime [I 1;30321 is characterized by small particles and lower fluidizing velocities. For a fluidized bed with a wide range of particle sizes, a slow bubble regime is observed in the bottom of the bed near the air distributor plate and a fast bubble regime is observed in the upper portion of the bed. The transition from “slow bubble” to “fast bubble” behavior occurs at a bed height which depends on the specific operating conditions. In order to simplify the mathematical description of sulfur capture, it will be assumed that the bed operates solely in a slow bubble or in a fast bubble regime. Both of these regimes will be treated in the model because the large scatter in the experimental data currently precludes determining which assumption of bubble behavior is a more accurate description of a sulfur capture in a fluidized bed combustor. (7) The bulk of the sulfur dioxide is captured in the fluidized bed rather than in the freeboard region above the bed. There are, however, indications that the freeboard region makes a significant contribution (as high as 25%) to sulfur capture [33]. Slow bubble regime By use of these simplifying assumptions, the sorbent feed rate necessary to achieve the desired sulfur capture can readily be projected. The average conversion, E, of the sorbent in the fluidized bed may be calculated from the equation _ a= Ecx dt (2) I0 where E is the age distribution of sorbent stone in the bed and (Y is the extent of conversion of the sorbent at time, t[34]. Using assumption 5, the age distribution function is taketras
where T is the average residence time of sorbent stone in the bed (i.e. 7 is the moles of calcium in the bed divided by the calcium molar feed rate). The extent of conversion, a is taken from TGA data (see Appendix, eqn A3). Substituting eqns (3) and (A3) into eqn (2) letting b’ = b/p, C, and integrating results in (4) Now R = fractional =
retention
calcium to sulfur molar ratio of the feed. Manipulating eqn (5) gives 1 D =In Ca/S R
S
b' Cab,’
(6)
Now Cab,=AH(l)3)p,
moles
(7)
where A=area of fluidized bed (rnq; H=height of Ruidized bed (m): fl=fractional volume in the bed not occupied by sorbent particles (including the fractional volumes of the bubbles (6), emulsion voidage (e) heat exchanger tubes (&), and inert solids in the bed (Fr). Thus from eqn (6) (1+)=(1S) (le) (IFW) (lFI).
where C, and C2 ate the SOZ concentration (moles/m”) in the emulsion phase and bubble phase, respectively. For the slow bubble regime, by definition there is free exchange of the gas between the bubble and the emulsion phases so C1 and CZ are the same. Thus, the sulfur material balance in the combustor C, = CZ=w
(mole/m3)
(9)
where U is the Buidizing gas velocity (mlsec). Thus, substituting eqns (7) and (9) into eqn (8) results in
This equation can be used to predict the CalS molar ratio of the feed required to achieve the desired fractional sulfur retention, R, based on fluidized bed and sorbent parameters.
Fast bubble regime In the fast bubble regime, the SO2 concentrations are different in the bubble and emulsion phases. To calculate the difference, it is assumed that the twophase theory can be applied to simulate the performance of the reactor. Thus, the superficial velocities of gas entering the emulsion phase and the bubble phase are IJ,, and U  IJ,,, is the minimum Ruidization respectively, where U, velocity (m/se+ The material balance for the total sulfur in the bubble phase gives rise to
of SO, = (Ca/S)c (UU
(CalS)Db’(
)G=k “fdz
C) LS(C 12
(11)
(5) b,(*)+l
where Ca&,=moles of calcium in the bed; &,=molar feed rate of sulfur to the bed (moleslsec); (Ca/S)=
where kr is the interfacial mass transfer coefficient between the bubble phase and the emulsion phase, set‘, 6 is the volume fraction of the bubble phase, and t is the axial distance above the gas distributor plate.
Fluidizedbed The
boundary
conditions
coal combustion: inbed sorbent sulfation
(N,&,,
The total becomes
(&)I.
Iexp
sulfur
balance
in the
(‘3) emulsion
=0.0d134(&J”.“;
(N&,,=O.O426
Substituting
eqn (13)
I
for No_> lo5
(22)
is the Reynolds number at the minimum fluidization condition. No. is the Galileo number, defined as
phase =
4% p&g
(23)
w
H
k,pS(C,Cz)
0
dz+RSi,,.
(14)
eqn (14) yields:
into
(*)I
Sim=*C,U,,+Cl[ lexp
for 18< 10’
(N,&“6m;
where WR.)~, defined as d&,,fp,.p,
Noa
Sin= AC, U,,+
by
(12)
From eqns (11) and (12), it can be shown that
c&q
1919
calculated from the empirical correlations proposed Wen and Fan[35]. The equations have the form:
for eqn (11) are
r=O, c2=0
model
(UU~f)+RSim~
where dp is the particle diameter (m), p8 is the gas density (kg/ma), p,, is the particle density (kg/m?, 1~ is the viscosity of the gas (kg/m set) and g is the acceleration of gravity. On rearranging eqns (22) and(23). U,,,, is given by
Rearranging eqn (15) leads to for 18
SdlRI
C,= AU,&A[lexp Rearranging
(e)](U
U,,)’
= 0.0426~ d,ps
(1’)
eqn (16) gives (17)
dp3(pbp,Jpsg
P
z
o/602
)
for Na,> 10’. The gas exchange coefficient, k,, between the bubble phase and emulsion phase (set‘), is estimated from Kobayashi et a[.[361 as
where
y=[ c_L,+[1erg
($+J](UvI)]&.
(18)
Generally, y is kss than unity. However, when k, is infinitely large, y becomes equal to unity and there is uniform gas mixing between bubble and emulsion phases. Substituting eqns (7) and (17) into eqn (8) yields
where b,(m) is the average bubble diameter in the bed. The correlation equations of Mori and Wen[37] are used to calculate for the average bubble diameter at the location of H/2. The equations have the form: &=DB,,,(DhD&exp(0.305)
(26)
L
WS=(;  ub(l_$&) >
(19)
where Z= HI2 and DBm( m) is the maximum diameter which can be expressed by
The SO, concentration in the flue gas above the bed, C., can be calculated by
(20)
De,,, = 1.64[A( U U,,,f)]“s
WI
or DBm = bed diameter (DR, m), whichever
Substituting
bubble
is smaller
eqns (13) and (16) into eqn (20) gives c
e
=%(1R) AU
(21)
as would be expected from sulfur material balance considerations. The three additional parameters, U,,,,, k, and 6 used in describing the fast bubble regime are estimated as follows. U,,. the minimum fluidization velocity (m/set) is CES Vol. 38. No. IIJ
DrsO is the initial bubble expressed by
diameter
Cm) which
; for a performated or bubble cap ; for a porous plate
can be
plate (28)
where ND is the total number of orifice openings on the distributor. The bed average bubble diameter has also
1920
D. C.
been assumed to occur at the location of H&2 where the bed height (m) at minimum HINf is fluidization [38,39]. In the fluid&d bed operation, the height of bed expansion, H, is normally a known parameter, which is taken as the height of the outlet for the spent sorbent. Assuming that the heat transfer tubes are uniformly distributed in the bed, the height of the bed at the minimum fluidization conditions, H,,,f, can, thus, relate to H from the mass balance of the gas in the bubble phase [37] as
where Ub(m/sec) is the linear bubble velocity which can be estimated by the equation of Davidson and Hahison [32] ub= uu,,+o.711(g~s)0~5.
(30)
FEE et al.
For the fast bubble regime, exchange of gas between bubble and emulsion is less than infinitely rapid, y is less than unity. As a result, the concentration of SO2 in the emulsion phase is higher than for the slow bubble case (see eqn 17). This increases the amount of sulfur captured because the rate of sulfur capture increases linearly with increasing SO2 pressure. As a consequence, lower sorbent requirements to achieve the same fractional sulfur retention are predicted for the fast bubble regime compared to the slow bubble regime. Use of the fast bubble approach (7~1) is slightly more difficult than use of the slow bubble approach (y = 1) because three additional parameters, k,_, 6, and U,,,,, must be calculated. Further, the accuracy of these calculated parameters cannot be easily verified experimentally. However, the projected sorbent requirements using the twophase treatment are not extremely sensitive to the k~, S and U,,+ values. So errors in calculating these parameters do not detract from the general trend (discussed above) of lowering the predicted sorbent feed required for a desired sulfur retention.
Thus Hm,=
(
0.711 (g&,“.’
u  urn* + 0.7 11 (g&l )“.’)
H.
The final parameter 6, the volume fraction bubble phase in the bed, is obtained by
gJHmdLf _H
Uh
.
(31) of the
(32)
DISCUSSION
The equations which can be used to predict sorbent performance in the slow bubble regime (eqn 10) and the fast bubble regime (eqn 19) differ only in that y, the gas exchange parameter, equals unity for the slow bubble regime, and is less than unity for fast bubble regime. In both cases the equation shows explicitly the sorbent and FBC parameters which affect sulfur capture. Three conclusions are immediately obvious. First, the maximum fractional sorbent utilization parameter, D, is extremely important. If the sorbent reacts infinitely rapid with SO, (i.e. b++, the U/Hb(lR)(lp) term becomes zero and the required (CalS) ratio is R/D. This is the minimum Ca/S ratio for the sorbent in question to achieve the desired sulfur retention, R. Further, a plot of retention versus (Ca/S) ratio has a slope of D. For finite values of b, the $JHb(lR)(lfi) term is nonnegligible and (Ca/S)>R/D. The second conclusion is that the gas residence time in the bed, H/U, is important. Increasing the gas residence time decreases the Ca/S ratio required (provided, of course, that these changes in the H/U ratio are over an interval small enough so that p is constant). The final c‘onclusion is the intuitively obvious onenamely, for two stones with the same maximum sulfation (D) values and the same combustor operating conditions, the more reactive stone with the larger b value will require a smaller (CalS) ratio in the feed to achieve the same sulfur capture, R. This concept is developed in more detail in the Appendix.
Comparisons with plant operating data The proof of any model is the accuracy of its predictions. To facilitate a comparison of experiment and prediction, an extensive review of the literature on sorbent performance in atmospheric fluidizedbed combustors has been undertaken; a compilation of the data is provided in an Argonne National Laboratory report[40]. Selected combustor performance data have been extracted from the data base and compared with predicted sorbent performance. The comparisons are made for larger scale combustors (to avoid the problems associated with channeling and slugging found in small combustors) although several smaller scale combustors are included in the literature review. The comparisons are shown on a plot of sulfur retention versus calcium to sulfur molar ratio in the feed. For each combustor, the combustor data and the model predictions are shown for the same set of operating conditions in Figs. 1 and 2. The fluidized bed operating conditions listed in Table 1 and the sorbent data are shown in Fig. A4. The operating conditions listed represent the midpoint of narrow ranges of the actual operating conditions; for example, a temperature of 850°C represents a range of 84086O’C, approximately. The model predictions for both slow and fast bubble regimes are shown in Figs. 1 and 2. The data for the Babcock and Wilcox (B&w) 61~x6 combustor covers series 115 in which the cyclone catch was not recycled to the bed. Similarly, the B&W 3’X 3’, combustor data is for a oncethrough sorbent usage. Thus, the 6’X6’AFBC data (Fig. 1) and the 3’~3’ AFBC data (Fig. 2) are for conditions similar to those used for previously published comparisons of experimental data with the prediction of the Westinghouse model[2]. In both figures, for the same Ca/S ratio, greater sulfur retention is predicted for the fast bubble regime than the slow bubble regime because of the higher SO2 concentration (and consequent higher sulfation rate) in the emulsion phase in the fast bubble regime. The difference
Fluidizedbed coal combustion: inbed sorbent sulfation model
1921
Table 1. Fluidized bed operating conditio& Cambustor B&W 5&Y 6x6ft 3x3ft Fig. 1 Fig. 2
Fluidizing
Velocity,
Red
Heiqht,
Red
Area.
Coal
Avg.
m2
Feed
sulfur
m/%x
II
Rate,
Content
2.54
1.20
0.42
3.34
0.98
kg/h in
Coal.
%
900
221
3.6
3.0
Particle
Sorbent
Diameter
2.44
in
bed,
rm
o.gd
1 .oc
Keference
33.4143
aTemperature  SSO’C, presswe  1 atm. sorbent limestone (93 wt X CaC03; 2 wt X HgC03; kg/d), coal size  less than 6.35 mn diameter.

calcine
bAdditiona1 E = O.4534, og = 0.33
data
used
FI
= 0.05.
kg/m3,
and
in
the
comparison
FH = 0.04. ND/A
= 1070
we
as
II = 4.5x105
4648
Lwellville density

1700
follows: kg/ms,
openings/w?
CThe average sarbent particle diameter in the bed for an individual experlnent in Fig. 1 ranged from 0.8 to 1.2 run. The average sorbent particle diameter in the bed for the group of experiments shown in Fig. 1 was 1.0 + 0.1 rrml (one std. dev.). dThe average sorbent particle diameter in the bed for experiment in Fig. 2 ranged from 0.6 to 1.3 rum. The ticle diameter in the bed for the group of experiments Fig. 2 was 0.9 t 0.2 rm (one std. dew.).
4 I 2 3 7, ‘ . IOO~,.~~,.“ ‘,~.~~,“ ‘~~‘ ~~’
an individual average parshown in
Jo0
so
I30
60
FAST BUBBLE SLOWBUBBLE * COMBUSTOR
60
20

20DATA
6
FAST BUBBLE 20 SLOW BUBBLE _ COMBUSTOR DA?\_
Ol,,l,l,,,.rl.,.l,.,,1,.,.1...,1 Co/S
MOLAR
RATIO
0
I
2 Co/S
0 3 MOLAR
4
5
6
RATIO
Fig. I.
Fig. 2.
between fast and slow bubble predictions is greater in the 3x3 ft combustor (Fig. 2) because there is less emulsion to bubble exchange (smaller y) due to lower bed height (compared to the 6X6ft combustor). The predicted curve of retention versus CalS ratio for the slow bubble regime is similar in shape to that of the Westinghouse model, but slightly higher in value. This similarity is expected because both the slow bubble and Westinghouse predictions assume a uniform SOz mixing between bubble and emulsion phases.
For the 6X6ft combustor (Fig. 1) the ANL model appears to slightly underpredict the sulfur retention for a given CajS ratio in the feed. Several possible explanations are as follows: (1) Freeboard sulfationthe models deal solely with sulfur capture within the Auidizedbed. However, sulfur capture above the bed, in the freeboard region, has been reported to be as high as 25% of the total[33]; (2) Attritionthe abrasive action of particleparticle contact in a fluidizedbed tends to remove the outer calcium sulfate layer on tbe sorbent
D. C. FEE er al.
1922
particle and expose the unreacted core of calcium oxide to SOZ. The attrition process thus rejuvenates the reactive surface area of the sorbent and increases the overall sulfur dioxide adsorption capacity. A similar mechanism does not exist in the standard TGA test where the sorbent particles are not in motion; (3) Inadequate Standard TGA TestThe model predictions are based on thermogravimetric analyzer (TGA) data. However, even though great care is taken in an attempt to make the calcinationlsulfation conditions identical in terms of temperature, C02, SO=, and N, pressure, etc. the calcinationlsulfation conditions are apparently different in the TGA and in the fluidizedbed combustors. Part of the difference between TGA and FBC appears in the utilization of the calcium in the sorbent in that the maximum calcium utilization achieved in the combustor is much higher than in the TGA test. For example, Lotiellville limestone achieves a maximum utilization of only 32% of the calcium in the TGA test while utilizations of 36% were reported in the B & W 6’x6’ combustor at steadystate conditions[44]. A difference in the maximum utilization of this magnitude would increase the predicted retention, at a Ca/S feed ratio of 3.0, of 54% of the sulfur fed to the bed in Fig. 1 and yield retentions that fall closer within the scatter of the B&W FBC data. A similar difference in maximum calcium utilization between a small laboratory fixed bed reactor and combustor has been noted previously[21]. The apparent difference between sorbent behavior in a TGA versus a combustor currently preclude determining whether a slow or a fast bubble regime is a better representation of a fluidizedbed coal combustor. The ANL model shows explicitly the sorbent and FBC parameters which affect sorbent performance. The ANL model proposed here is a simple analytical expression which is adequatk for ranking the suitability of limestones and dolomites as sorbents in fluidizedbed combustors. AcknowledgementsThe authors express their appreciation to the U.S. Department of Energy, Morgantown Energy Technology Center, for their support and encouragement, The administrative assistance of Mr. A. A. Jonke and Mr. L. Burris, Argonne National Laboratory, is gratefully acknowledged.
NOTATION
crosssectional area of the bed, m2 reaction rate parameter, see’ SO2 concentration in the emulsion phase, mole/m3 in the bubble phase, SO, concentration mole/m3 moles of calcium in the bed, mole SO, concentration in the flue gas above the bed, mole/m3 maximum fractional conversion conversion of the sorbent parameter, particle diameter, m average bubble diameter in the bed, m maximum bubble diameter in the bed, m initial bubble diameter, m bed diameter,
m
E
F, FH
exit age distribution function, set’ volume fraction of inert particles (char, ash, rocks, etc.) in the bed volume fraction of heat transfer surface in bed acceleration due to gravity, 9.80 m/set’ gas exchange coefficient between the bubble phase and emulsion phase, set’ height of expanded bed, m bed height at the minimum fluidization condition, m Reynolds number at the minimum fluidization condition, Galileo number total number of orifice openings on the distributor, fractional sulfur retention in the bed, 
k”, H
H mf (Ndmf N Ga ND
time molar
dow
rate
of
sulfur
at
the
bedinlet,
molelsec superficial gas velocity, m/set minimum fluidization velocity, m/set linear bubble velocity, m/set axial distance, m Greek
letters fractional extent of sorbent conversion at time t, _ fractional volume in the bed not occupied by sorbent particles, gas exchange parameter, defined by eqn (18) volume fraction of the bubble phase in the bed, emulsion phase fractional voidage, molar density of calcium in the sorbent, mole/m3 average sorbent conversion at the bedoutlet, density of the sorbent, kg/m’ density of the gas, kg/m3 average residence time of the sorbent in the bed, set viscosity of the gas, kg/m set REZERENCES
[I] Keairns D. L., EPA 6501275027C, Westinghouse Research Laboratory, Sept. 1975. [Zl Newby R. A.. Ulerich N. H. and Keaims D. L.. Atechnique to of Ruidizedbed project the sulfur removal performance combustors. 6th Int. Conf. FluidizedBed Combustion, Atlanta, Georgin, Aptil 911, 1980, Conf. 800428, p. 803. [3j Hartman M. and Coughlin R. W., Ind. Engng Chem. Proc. Des. Dev. 1974 13 248 (1974). [4] Hartman M. Int. Chem. Engng 1976 16 86. 151 Hartman M. and Coughlin R. W., A.1.Ch.E.L 197762 490. 161 Hartman M., Pata J. and Coughlin R. W., fnd. Engng Chem. Proc. Des. Deu. 1978 17 411. [7] Hartman M., Hejna 5. and Beran Z.. Chem. Engng Sci. 1979 34 475. [8] Hartman M. and Trnka O., Chem. Engng Sci. 1980 35 1189. [91 Borgwardt R. H., Environ. Sci. Technol. 1970 4 59. [iO] Borgwardt R. H., Environ. Sci. Technol. 1972 6 350. [ii] Tung S. E., Goldman J. and Louis J. F., Proc. 5th Znt. Conf. FfuidizedBed Combustion, Washington. D.C., December 1214, 1977, Vol. III, p. 406. Mitre Corp. 1978. [12] Chrostowski J. W. and Georgakis C., Am. Chem. Sot. Symp. Series 1978 65 225.
[I31 Georgakis C., Chang C. W. and Szekely Sci. 1979 34 1072.
J., Chem. Engng
Fluid&dbed
coal combustion:
[14] Lee D. C., Hodges J. L. and Georgakis C., Chem. Engng Sci. 1980 35 302. [151 Lee D. C. and Georgakis C., J. A.1.Ch.E. 1981 27 472. [16] Zheng J., Yates J. G. and Rowe P. N.. Chem. Engng Sci. 1982 37 167. 1171 Wen C. Y. and Ishida M.. Environ. Sci. Tech. 1973 7 703. [18] Horio M. and Wen C. Y.. in Fluidization Tech. (Edited by Keaims D.L.). Vol. 2, p, 289, 1976. [I91 Rajan R. R. and Wen C. Y., A.I.Ch.E.3. 1980 26 642. [20] Pigford R. L. and Sliger G.. Ind. Engng Chem. Proc. Des. Deu. 1973 12 85. 1211 Bethel1 F. W., Gill D. W. and Morgan 6. B.. Fuel 1973 52 121. [22] Chen T. P. and Saxena S. C., Fuel 1977 56 401. [23] Ramchandran P. A. and Smith .I. M., A.LCh.E.J. 1977 23 353. [24] Daniel K. I. and Finnigan S. D., Proc. 6fh Int. Conf. FluidizedBed Combustion, Atlanta, Georgia. April 911, CONF800428, p. 1l%, 1980. 1251 Henschel D. B., Proc. 6th Int. Conf. F’luidiredBed Combustion, Atlanta. Georgia, April 9l 1, CONF800428, p. 50 1980. Ml Young C. W., Robinson 3. M., Thunem C. B. and Fennelly P. F., Technology Assessment Repoti for Industrial Boiler Applications: fluidizedBed Combustion EPA6001779178e. 1979. [271Yang R. T.. Cunningham P. T.. Wilson W. I. and Johnson S. A.. Advan. Chem. 1975 139 149. 1281James N. J. and Hughes R., 2nd Int. Confi Control of Gaseous Sulfur and Nitrogen Compound Emission (Vol. I), Salford, United Kingdom, 68 April 1976. [29] O’Neill E. P.. Ulerich N. H., Newby R. A. and Keairns D. L., Electric Power Institute Rep. EPRI FP1307, 1979. I301 Park D., Levenspiel 0. and Fitzgerald T. I., Chem. Engng Sci. 1980 35 295. 1311 BarCohen A., Glickman L. and Hughes R., Proc. 5th Int. Symp. on FluidizedBed Combustion, Washington, DC., December 1214, 1979, Vol. III, p. 158, Mitre Corp. 1978. I321 Davidson 1. F. and Harrison D., Ruidized Particles. Cambridge University Press 1963. bed combustion facility and r331 LR : 78: 4775 :03 Fluid&d commercial utility AFBC design assessment. Babcock & Wilcox Annual Rep. 1978, 1979. Engineering, Chap. 13411 Kunii D. and Levenspiel 0.. Huidization 15. Wiley, New York 1969. [35] Wen C. Y. and Fan L.S., Ind. Engng Chem. Proc. Des. & Dee. 1974 13 194. [36] Kobayashi H.. Arai F. and Sunagawa, T., Chem. Engrs (Japan) 1967 31 239. [37] Mori S., and Wen C. Y., A.1.Ch.E.J. 1975 21 109. [3R] Kato K. and Wen C. Y., Chem. Engng Sci. I%9 24 1351. 1391 Fan, L. T. and Fan LS., Chem. Engng Sci. 1979 34 171. [40] Fee D. C.. Wilson W. I., Shearer J. A., Smith G. W.. Lent J. F., Fan L.S., Mytes K. M. and Johnson I., Sorbenf Utilizntion Prediction Methodology: Sulfur Control in HuidizedBed Combustors, ANL/CEN/FE8010, 1980. [41] LR:79:4775:01, Fluidized bed combustion facility and commercial utility AFBC design assessment. Babcock & Wilcox Quarterly Rep. JanMar 1979. [42] LR: 79 :4775 : 02 Fluidized bed combustion development facility and commercial utility AFBC design assessment. Babcock & Wilcox Quarterly Rep. AprilJune 1979. [431 LR: 79 :4775 : WA, Fluidized bed combustion developmenl facility and commercial utility AFBC design assessment. Babcock & Wilcox Quarterly Rep. JulySept 1979. [44] LR: 80: 4775 : 03 Fluidized bed combustion development facility and commercial utility AFBC design assessment. Babcock & Wilcox Quarterly Rep. JanMarch 1980. [45] LR: 80: 4775 :04 Vol. I. Fluid&cd bed combustion development facility and commercial utility AFBC design assessment. Babcock & Wilcox Quarterly Rep. AprilJune 1980, 1980. (461 LR: 77: 4737: 1, SOz absorption in fluid&d bed combustion of coaleffect of limestone particle size (Task I). Babcock Bi Wilxoc Rep. 1977.
inbed sorbent
sulfation
model
1923
[473 LR :77 :4737 : 2 S& absorption in Ruidized bed combustion of coaleffect of limestone particle size (Tasks 1 gL 2). Babcock & Wilcox Rep. Nov 1976Feb 1977, 1977. [48] EPRI FB667, SO2 absorption in tluidized bed combustion of coaleffect of limestone particle size (Tasks 1, 2 & 3). Babcock & Wilcox Final Rep. Electric Power Research 1nstitr;te 1978. [49] Snyder R. B., Wilson W. I., Johnson 1. and Jonke A. A., Air J., Poll&on Control Association 1977 27 975. [50] Snyder R. B., Wilson W. I. and Johnson I., Thermochima Acto 1978 26 257.
AFTENDIX
Themogravimetric
analysis
The reactivity and the capacity of a sorbent for sulfur dioxide are etermined in a thermogravimetric analyzer in a manner similar to that previously described[49,50]. For both calcination and sulfation reactions in the thermogravimetric analyzer, the sample’s placed in a wire mesh platinum basket (0.5 cm is ID, 1.8 cm height) suspended in a quartz reactor lube (2.5 cm ID) from one arm of an Ainsworth Model RVA recording microbalance. The reactor tube is heated by a platinumwound furnace which is controlled to within 5°C at temperatures up to 12OO’C. Two Pt. Pt10% % Rh thermocouples are used for temperature monitoring and control. The gas mixture passes upward through the heated reaction tube and reacts with the samples, then exits throfigh a condenser and series of scrubbers. A nitrogen purge gas stream that flows countercurrently through the microbalance bell jar keeps it free of corrosive reactant gases. The gas mixtures are prepared by blending streams of the individual gaseous constituents. Mass spectrometric analyses of samples of the feed gas mixtures are made to verify their composition; the measured values of the SO2 and & concentrations arc within 5% of the values computed from the flow rates. In the calcination step, a sample of the stone is initially screened to the desired particle size. A sample of the stone (3oo4omg) is calcined at 8SOT in 20~01% C&Nz. After calcination is completed. the sample is exposed to a gaseous mixture consisting of the desired So2 level. 5 ~01% 02, 20 ~01% CO2, n N2 at 8JO”C and a total flow rate of 2Umin. The SO2 levels were fixed for each experiment at either 0.05, 0.1, 0.2, 0.3, 0.5 or I.0 ~01%. The TGA unit continuously records the weight change of the sample during its reaction with a gas mixture. At the completion of TGA run, chemical analyses of the sample are performed to help determine and quantify chemical changes which have occurred. Data reduction
The observed weight gains from the TGA are converted, using the sample weight and chemical composition, into the fractions of the total original calcium that reacted to yield calcium sulfations (dtilizations). The fractional calcium utilization versus time data, a, are least squares curve fitted by the following empirical quation: ,=a[lexp(y)]
(AU
where t is time, C, is the SO2 concentration (moleslms), p= is the calcium concentration in the sorbent (moles/m’) and O, k, and n are fitting constants. The constant (1 equals the maximum fractional utilization a sorbent can achieve in infinite time under the standard test conditions. A typical plot is shown in Fig. Al. The appropriateness of the use of eqn (Al) to represent the utilization is clearly indicated by the [email protected] of the fit of the experimental values in Fig. Al, especially over the sensitive early periods of the test. The rate of calcium utilization (set‘) is obtained by differentiating eqn (Al):
D. C. FEE et al.
1924
SULFATION 0.40
4000 SO2
I
6000 PRESSURE,
I
.
.
5 ,
.
I
IO ,
I
I
I
I
IS (
I
15
8000 pprn
a(FRACflONAL
UTILIZATION)
Fig. A3.
Fig. AZ.
Equation (A2) implies that the sulfation reaction shows a first order dependence on sulfur dioxide concentration. A 6rst order dependence is shown in Fig, A2 and is consistent with data in the literature[9,10,30,31]. If n is equal to one, the rate of calcium utilization decreases linearly as the extent of calcium utilization, a, increases; for values of n less than one, the rate will also decrease as the calcium utilization increases, but not linearly. For various sorbents which have been studied[40] the average value of n was 0.76 with a range from 0.6 to 1.1. The dashed curve in Fig. Al represents the utilization reaction rate (set‘) as a function of the extent of calcium utilization as defined by eqn W). Unfortunately, the form of eqns (Al) and (AZ) does not yield closed athematical expressions for the average sorbent utilization (see eqn 2) in the combustor. Accordingly, simpler forms were derived in which n is set equal to one, As will be demonstrated forthwith, this simplification does not affect the accuracy of the relationships over the limited range defined by practical FBC experiences. The new equations are: a=D[lexp(y)]
I
RATE .$IO5s’
(A3)
where b is defined as the reaction rate parameter (set‘) and D is defined as the maximum fractional utilization parameter. The b and D parameters are derived from TGA data in the following manner. Differentiating eqn (A4) with respect to Q and rearranging gives j&($)1=b.
Thus, a plot of [p&(da/dt)] versus II is a straight line, with a slope of b and an intercept of D on the rr axis. This is evident in Fig. A3 using the data from Fig. Al. Use of the more rigorous eqn (A2) shows that the actual plot in Fig. A3 is a curve functionally dependent on t”‘. The close.correlation (k3% = 1 standard deviation) between two plots clearly indicates that the use of the simpler linear relationship does not introduce an appreciable error to the analysis of the TGA data over the range of interest. The critical range is defined from AFBC experiende
Fluidizedbed
coal combustion:
inbed sorbent sulfation
model
1925
4r
3i
.a0 l
\ d
200
b
SOLID
SYMBOLS
OPEN
SYMBOLS
 ANL WESTINGHOUSE
li IO
05 dp(PAATICLE
0 1.5
DIAMETER).mm
Fig. A4.
as corresponding to values of [pJCstdoldt)l ranging from approximately IO50 set’ with a ratio of bed height, (H, m), to fluidizing velocity (0, mlsec) ranging from 0.3 to 1.2 s in the AFBC. This sorbent reactivity range will result in achieving the mandatory New Source Performance Standards of 90% sulfur capture. Lower sorbent reactivity [p./CJda/dt)l values than 10 see’ result in insufficient sulfur capture to be of interest. Sorbent reactivities in the bed higher than 50 set’ are generally achieved by only excessive sorbent feed rates which are too high to be of economic interest. (The reaction rate data shown in Fig. A2 were selected to lie within this range of interest for AFBC systems.) The parameters D and b which are derived from TGA data in
this manner are shown in Fig. A4. As the pat’ticle size decreases, the maximum utilization parameter, D, increases while the reaction rate parameter, b, decreases. However, since the change in D with particle size dominates in the rate expression, eqn A4, the rate of sulfation increases as the particle size decreases. Smaller sorbent particles react more rapidly and to a greater extent, i.e. achieve higher calcium utilizations than larger particles. The trends of increasing maximum sutfation parameter and decreasing reaction rate parameter with decreasing particle size are consistent with previous results in the literatureI4, 9, 10, 291. Also shown in Fig. A4 are the D and b values of a sample of limestone that was supplied by Westinghouse and was used to verify the accuracy of the TGA results.