The Prediction of Coal Char Reactivity under Combustion Conditions

9
Twenty-Fourth Symposium (International) on Combustion/The Combustion Institute, 1992/pp, 1189-1197 THE PREDICTION OF COAL CHAR REACTIVITY UNDER COMBUSTION CONDITIONS SYLVIE CHARPENAY, MICHAEL A, SERIO AND PETER R SOLOMON Advanced Fuel Research, Inc. 87 Church Street, East Hartford, CT 06108 USA The main objective of the current work was to develop a char reactivity submodel that could be used in a comprehensive code for entrained coal combustors and gasifiers. The requirement of a char reactivity model is to predict the reactivity of chars from a wide range of coals, over a wide range of temperatures, and for various degrees of burnoff. In order to predict intrinsic reactivity, correlations of reactivity with char hydrogen content, coal oxygen content and coal mineral content were used. A random pore model (high rank coals) and a volumetric model (low rank coals) were included to predict variations of intrinsic reactivity with burnoff. The correlations combined with those models gave good predictions of reac- tivity (within a factor of 2 to 4) and reactivity variations with burnoff (within 20%) for the range of chars studied. In the pore diff'lsion regime, the model uses the Thiele modulus to calculate the reaction rate as a function of the intrinsic rate (obtained using the correlations) and char properties such as porosity, tortuosity and mean pore radius, Predictions of reactivity required an es- timate of the tortuosity and the mean pore radius, The tortuosity was kept constant at a value of 2, Using a value of 6 A (corresponding to the size of micropores) for the mean pore radius led to good predictions of the onset of diHusion limitations for low heating rate, fluid chars. For high heating rate chars, pore size distribution measurements showed that values of approximately 100 A were more appropriate. The corresponding predictions using this value gave a fairly good fit of the literature data investigated. This analysis shows that it may be possible to extend a low temperature reactivity model to high temperatures. Introduction The objective of a char reactivity model is to pre- dict the reactivity of various coals and chars over a wide range of temperatures and at various degrees of burnoff. The current work focused on the de- velopment of a char reactivity sub model that could be used in a comprehensive code for entrained coal combustors and gasifiers, Intrinsic reaction rates (regime I) can be mea- sured at low temperature in a TGA apparatus. However pore diffusion effects (regime II) are im- portant under many practical combustion and gas- ification conditions and need to be accounted for. The external diffusion limitation (regime III) can also be important, but was not included in the present model since several studies l - indicate that it is not 3 usually the rate limiting step under pulverized coal combustion conditions. A model which can be used in regimes I or II must include a description of intrinsic char reactiv- ity and how it varies with temperature, coal type, and char formation conditions. In order to predict intrinsic reactivity, the measurement and evalua- tion of the number of active sites on chars surfaces have been the object of extensive work. 4,5 How- ever, the current methods (TPD, TGA, transient kinetics, etc .. ,) do not allow a direct measure- ment of the number of active sites, In the absence of such a method, our approach has been to de- velop correlations of reactivity with known param- eters of coals and chars. Correlations of reactivity with the concentrations of minerals and oxygen in the starting coal and hydrogen in the char as well as pyrolysis heating rate have been found. These correlations reflect the fact that the amount of ac- tive sites is dependent on the following: 1) The de- gree of pyrolysis, which can be characterized by the hydrogen content of the char; 2) The degree of dis- order of the char, which varies with the pyrolysis heating rate (valid for fluid coals only); 3) The coal rank, which can be characterized by the oxygen content; 4) The amount of catalytic impurities pres- ent in the starting coal, which can be characterized by the dispersed mineral content (calcium in par- ticular). For high rank coals, reactivity variations with burnoff (in the kinetic regime) have been associated with changes in internal surface area,6 Conse- quently, reactivity predictions will require a de- 1189

Transcript of The Prediction of Coal Char Reactivity under Combustion Conditions

Twenty-Fourth Symposium (International) on CombustionThe Combustion Institute 1992pp 1189-1197

THE PREDICTION OF COAL CHAR REACTIVITY UNDER COMBUSTION CONDITIONS

SYLVIE CHARPENAY MICHAEL A SERIO AND PETER R SOLOMON

Advanced Fuel Research Inc 87 Church Street East Hartford CT 06108 USA

The main objective of the current work was to develop a char reactivity submodel that could be used in a comprehensive code for entrained coal combustors and gasifiers The requirement of a char reactivity model is to predict the reactivity of chars from a wide range of coals over a wide range of temperatures and for various degrees of burnoff In order to predict intrinsic reactivity correlations of reactivity with char hydrogen content coal oxygen content and coal mineral content were used A random pore model (high rank coals) and a volumetric model (low rank coals) were included to predict variations of intrinsic reactivity with burnoff The correlations combined with those models gave good predictions of reacshytivity (within a factor of 2 to 4) and reactivity variations with burnoff (within 20) for the range of chars studied

In the pore difflsion regime the model uses the Thiele modulus to calculate the reaction rate as a function of the intrinsic rate (obtained using the correlations) and char properties such as porosity tortuosity and mean pore radius Predictions of reactivity required an esshytimate of the tortuosity and the mean pore radius The tortuosity was kept constant at a value of 2 Using a value of 6 A (corresponding to the size of micropores) for the mean pore radius led to good predictions of the onset of diHusion limitations for low heating rate fluid chars For high heating rate chars pore size distribution measurements showed that values of approximately 100 A were more appropriate The corresponding predictions using this value gave a fairly good fit of the literature data investigated This analysis shows that it may be possible to extend a low temperature reactivity model to high temperatures

Introduction

The objective of a char reactivity model is to preshydict the reactivity of various coals and chars over a wide range of temperatures and at various degrees of burnoff The current work focused on the deshyvelopment of a char reactivity sub model that could be used in a comprehensive code for entrained coal combustors and gasifiers

Intrinsic reaction rates (regime I) can be meashysured at low temperature in a TGA apparatus However pore diffusion effects (regime II) are imshyportant under many practical combustion and gasshyification conditions and need to be accounted for The external diffusion limitation (regime III) can also be important but was not included in the present model since several studies l - indicate that it is not3

usually the rate limiting step under pulverized coal combustion conditions

A model which can be used in regimes I or II must include a description of intrinsic char reactivshyity and how it varies with temperature coal type and char formation conditions In order to predict intrinsic reactivity the measurement and evaluashytion of the number of active sites on chars surfaces

have been the object of extensive work 45 Howshyever the current methods (TPD TGA transient kinetics etc ) do not allow a direct measureshyment of the number of active sites In the absence of such a method our approach has been to deshyvelop correlations of reactivity with known paramshyeters of coals and chars Correlations of reactivity with the concentrations of minerals and oxygen in the starting coal and hydrogen in the char as well as pyrolysis heating rate have been found These correlations reflect the fact that the amount of acshytive sites is dependent on the following 1) The deshygree of pyrolysis which can be characterized by the hydrogen content of the char 2) The degree of disshyorder of the char which varies with the pyrolysis heating rate (valid for fluid coals only) 3) The coal rank which can be characterized by the oxygen content 4) The amount of catalytic impurities presshyent in the starting coal which can be characterized by the dispersed mineral content (calcium in parshyticular)

For high rank coals reactivity variations with burnoff (in the kinetic regime) have been associated with changes in internal surface area6 Conseshyquently reactivity predictions will require a deshy

1189

1190 COAL COMBUSTION

scription of the char morphology For this purpose the Random Pore Model derived for the kinetic reshygime by Bhatia and Perlmutter7 and Cavalas8 was used which calculates surface area changes during reaction For low rank coals the intrinsic reactivity is primarily dependent on catalytic effects In that case the important parameter is the number of catshyalytic sites distributed through the volume of the particle and reactivity can then be represented by a volumetric model 9

In order to predict reactivity at high temperashytures (regime II) one has to account for pore difshyfusion limitations The model follows the approach described by SmithlO using the Thiele model ll to derive the value of the global reaction rate as a function of the intrinsic reaction rate and char properties such as char porosity tortuosity and mean pore radius

The transition region from kinetic to pore diffushysion control was not explicitly investigated since numerical solutions are usually necessary to calcushylate the reactive gas concentration profile in the particle and would be too detailed to be included in a comprehensive combustion code However this region represents a relatively narrow range in temshyperature and most of the available data can be conshysidered to fall in either the kinetic or pure pore diffusion regime

Experimental

Tent Measurements

Correlations have been developed using an exshytensive database of Critical Temperature (Terit)

measurements obtained in our laboratory Terit is an index of char reactivity which is related to the conshycentration of accessible active sites 12 The detershymination of Teri relies on a TCA technique in which the weight loss is measured while the sample is heated at a constant heating rate in the presence of the reactive gas The temperature (Tcrit) at which the rate of weight loss reaches a value of 0065 g g-min (on an original char mass basis) is recorded and gives an indication of the reactivity of the char The value of the reaction rate used for measureshyment of Tcrit was chosen to be high enough to be easily measurable but low enough so that it is still in the kinetic regime based on a Thiele modulus calculation 11 In practice depending on the nature of the char Teri has been observed to occur either in the kinetic regime or at the beginning of the pore diffusion regime The chars used for the developshyment of correlations were produced in several difshyferent reactors and under widely different pyrolysis conditions 13

Isothermal Reactidty Measurements

The reaction rates at low temperatures were measured isothermally using a Dupont 951 TCA All experiments were performed using an oxygen partial pressure of 172 kPa a flow rate of 240 cc min and a coal sample size of 10 mg The chars were prepared by heating at 30deg Cmin in helium up to the pyrolysis temperature followed by imshymediate cooling to the gasification temperature

Model Development for Kinetic Regime

Reactivity Model

Following the usual approach (eg see Smith l~ the reaction rate in the kinetic regime is given by

rp = S ki II MI Cs with

C = - (1)k = Aexp ( - T) p

S RT

where S = internal surface area of remaining char k = intrinsic rate coefficient II = stochiometric factor Mp = molecular weight of carbon and Cs

= oxidant concentration A reaction order of 1 was assumed although studies14 showed the reaction order to vary between 0 and 1 with the type of char and coal This approximation however typishycally leads to an error in rate of less than a factor of l5 if the true reaction order is for example 05 Future developments of the model will include a fractional reaction order

For high rank coals the value of S is given by the Random Pore Model7bull8 S = So(l - X)(l - 1Jf In(l - xW 5 where So = internal surface area of original char 1Jf = structural parameter X = char burnoff For low rank coals the volumetric model is used and S becomes S = So(1 - X)

The pre-exponential factor A of the coefficient ki includes the number of active sites The variashytions of A with coal type and pyrolysis conditions can be estimated with correlations based on Tcrit

measurements and can be written for high rank coals (HRC) as A = AHRC f(Hehar) g(Ocoal) where f(Hchar) and g(Ocoal) are char hydrogen and coal oxshyygen correlation functions respectively and AHRC

is the baseline frequency factor for high rank coals The corresponding factor for low rank coals (LRC) is given by A = ALRC f(Hchar) g(Otual) h(Clld) where h(Cad) is a correlation function based on the disshypersed calcium concentration

Development of Correlations

Correlations derived from Teri measurements can be used to calculate intrinsic reactivity The funcshy

1191 PREDICTION OF COAL CHAR REACTIVITY

00017

000165

00016

000155

00015 D

~ 000145 sshy0

00014

000135 I shy 00013

000125

00012

000115

ozap o pit Dill

00011 6 ros

000105 0 2 3 4 5 6

Hchar

FIG 1 Correlation of lT with the perccnt (daf) hydrogcn content of the char (Hch) for Zap Roseshybud Illinois No 6 and Pittsburgh No 8 coals The chars were produced in different reactors and unshyder different pyrolysis conditions 13

tions f g and h are expected to be increasing funcshytions of the char hydrogen content coal oxygen content and coal mineral content respectively Since Terit decreases with reactivity 1Tcrit was found to be a more appropriate quantity to use to develop correlations Linear correlations of 1Tclit versus the above parameters were obtained as discussed beshylow

Hchar Correlation

Good correlations of 1Tcrit vs the char hydrogen concentration (Hchar) were obtained for all coals It can be seen from Fig 1 that all the correlations have similar slopes which implies that the degree of pyrolysis of the char has a similar influence for all coals The data also seem to show a plateau for Hchar values higher than 35 The initial loss of hydrogen during pyrolysis is primarily aliphatic which does not have the annealing effect of the loss of aromatic hydrogen during the latter stages of pyshyrolysis 1213

Gcoal Correlation

For similar values of Hchan the value of 1Tcrit for different coals is expected to correlate with oxshyygen andor mineral content of the coals As shown in Fig 2 1Tcrit (using chars produced at 30deg C min up to 900deg C) and the coal oxygen content show a reasonable correlation The data points with low oxygen content which fall well below the line corshyrespond to very fluid coals In that case the low

00015 0

000145 0 0

00014

OJ000135

~ 00013

lte

0 8000125I shy 00012

000115 OArgonne coals

00011 oExxon coalsc 0 000105

0 5 10 15 20 25

oxygen (daf)

FIG 2 Correlation of lTcrit with the percent (daf) coal oxygen contcnt (O~ for the Argonne eoals and a set of Exxon coals The chars were produced at 30deg Cmin up to 9000 C and have similar hydrogen contents

heating rate used to produce the chars also has an impact on its reactivity

Mineral Correlation

Since mineral concentration is important only for low rank coals a correlation was obtained using coals of oxygen content higher than 13 Dispersed calshycium was found to be the main mineral responsible for reactivity enhancementl3 and was used to deshyvelop the correlation shown in Fig 3 For low rank coals the influence of calcium needs to be sepashy

00015

o000145 o

00014 o ~ s- 0 0

O t 000135 0 I- 0

00013

000125 lOExxon coalsl

00012 w-~~~w~w~~~--~--J

o 02 04 06 08 12

dispersed calcium

FIG 3 Correlation of lTcrit with the percent (daf) of dispersed calcium (Cad) for a set of Exxon coals The chars were produced at 30 Cmin up to 9000 C

1192 COAL COMBUSTION

rated from the effect of coal rank Since deminershyalized low rank coals show a reactivity similar to that of medium rank coals (of oxygen content apshyproximately 13) the effect of oxygen content is minor compared to that of calcium In order to acshycount for this in the model the value of 0eo l is kept constant and set equal to 13 for coals of oxshyygen content higher than 13

Other second order effects include the combined effect of fluidity and pyrolysis heating rate on char reactivity This can be important for fluid coals and leads to separate lTerit vs Hebar correlations for different heating rates At this point few measureshyments have been done on low heating rate fluid chars and this will be the object of a future study

Individual correlation plots (shown in Figs 1-3) give an uncertainty in Terit within plusmn20deg C which is approximately equivalent to a factor of 2 in rate With the Heban Oma and Cad correlations used sishymultaneously Terit can be predicted within plusmn40deg C It should be noted that the experimental uncershytainty of the measurement of Terit is plusmn10deg C

To overcome potential shortcomings of using sevshyeral correlations simultaneously it would be reashysonable to use the Hebar correlation alone associshyated with one Tnit measurement as input However for the cases presented below the three correlashytions were used

I ntnnsic Rate Predictions

At Teritgt the rate of weight loss is OOOlsec By substituting the values of lTeit and rl into equashytion (1) the functions f g and h can be calculated It is implicitly assumed that 1 - X is constant and equal to 085 (which is the experimental value usushyally observed) at the point where Teit is measured So is also kept constant and was taken as a first apshyproximation to be 300 m2g The value of the acshytivation energy E was evaluated from TGA isoshythermal data and was found to vary between 28 and 34 kcalmol for the chars studied As a first approximation a value of 30 kcalmol was used in the model

For high rank coals (Oeoal lt 13) equation (1) becomes

-15000 rl = 822 exp -- shy

l

exp(14 Hehar) exp(0263 Oml) S (2)

For low rank coals (Oenal gt 13) the equivalent of equation (1) is

-15000 rl = 686 x 102 exp -- shy

T

exp(14 Hdmr) exp(154 Cad) S (3)

The reaction rates correspond to the case where the oxygen partial pressure is 172 kPa

Predictions of reaction rates using equations (2) and (3) have been compared with data obtained in our laboratory (Zap lignite chars) as well as literashyture data from Wells et al 15 (FMC Occidental and Toscoal chars) The parameters used are indicated in Table I The parameters for the literature chars were directly taken from reference 15 The reactivshyities of the Pittsburgh No 8 chars were not preshydicted (since good correlations of low heating rate fluid coals have not been developed yet) but were fit to the data Figures 4 and 5 show the good agreement obtained between the predictions and the low temperature TGA data for the chars studshyied

Predictions of reaction rates as a function of burnoff are presented in Fig 6 for pittsburgh No 8 (high rank coal) 600deg C char and Zap lignite (low rank coal) 900deg C char using the random pore model and volumetric model respectively The predictions were done using a single adjustable parameter (without using the correlations) in order to demshyonstrate the adequacy of the burnoff models choshysen The good fit (within 20) obtained in the full range of burnoff supports the use of these models

Model Development for Pore Diffusion Regime

Calculation of Reaction Rate

In the diffusion regime the observed reaction rate can be expressed as a function of the intrinsic reshyaction rate

(4)

where YJ = effectiveness factor According to Iehta and Aris J(j the effectiveness factor can be detershymined using the relations between YJ and YJqi where cjJ is the Thiele modulus II In the case of high Thiele modulus YJcjJ2 tends to lYJ ie YJ tends to lcjJ

The Thiele modulus is given by

d ( S k )0gt

(J () IcjJ-- (5)

6 De V MMo

with d = particle diameter = apparent particle (J

density De = effective diffusivity Mo = molecular weight of oxidant From the simple pore-structure model of Wheeler17 De can be calculated

05 _ 2 T E

De - 97 x 10 rpore (

-)

- (6) Mo To

where rpltre = mean pore radius Eo = porosity To = tortuosity

1193 PREDICTION OF COAL CHAR REACTIVITY

TABLE I Parameters Used in Reactivity Calculations

PIT 900deg C amp ZAP

Char 600degC 900deg C FMC OCC TOS

Parent Coal Pittsburgh Zap Pittsburgh Wyodak Utah

Rate (OCs) 05 05 103 10 10

pound0 01 03 013 055 011

10 2 2 2 2 2

dpm) 70 70 90 63 55

rpc (A) 6 6 170 100 170

So (m2g) 300 300 300 300 300

00 (gcm3) 13 13 152 086 143

01 () - 21 9 21 118

Hcha() - 074 2 21 318

M () - 1 0 11 0

Note 900deg C and 600deg C Pittsburgh No 8 chars and 900deg C Zap lignite char have been produced in our laboratory FMC Occidental and Toscoal are chars from Reference 15

estimated from measured values for similar chars or coals

The global reaction rate used in the model can then be written

05 6 ( Mgt So ki)r=-Dv--- vMC (7)

P deMo 0 p bull

The reaction rate depends on the intrinsic rate and char properties such as the tortuosity porosity and the mean pore radius The tortuosity parameter has been introduced to account for the connectivity of pores Its value has generally been taken between 15 and 10 usually equal to 2 The porosity and the mean pore radius can be measured However their determination may be difficult in the case of a non-uniform pore size distribution

Comparison of Predictions with Data

Using the physical and chemical parameters inshydicated in Table I and kinetic parameters derived earlier predictions in the pore diffusion regime for the Pittsburgh No 8 and Zap lignite chars proshy

duced at 30deg Cjmin are shown in Fig 4 For low rank coals we used a deactivation function to acshycount for the fact that catalysts have less impact on reactivity at high temperatures 1213 Following this deactivation function at high temperatures the reactivity of a low rank coal is made to be equal to that of a coal of 13 oxygen Although the number of data points in the diffusion regime is limited the transition between kinetic and diffusion control is fairly well represented by the theory for the two Pittsburgh No8 chars using a mean pore radius of 6 t which corresponds to the size of micropores Concerning the Zap char no internal diffusion limshyitations are present (since there is no change in ac~ tivation energy) but a leveling off of the rate is obshyserved This behavior is due to external mass transfer resistances in the TGA Under our conditions the maximum measurable rate in the TGA is approxishymately 1 gjg-min However under true pulvershyized coal combustion conditions external diffusion limitations are not encountered until temperatures typically higher than 1500 K Since no pore diffushysion limitations are observed for the Zap char in

1194 COAL COMBUSTION

4 4

3 -predictions OPIT 900 Cchar

CshymiddotE 2

ePIT 600 Cchar IoZAP 900 Cchar

dgt -9

~ s 0 0 15 ro (J)

gshy -1

~

OJ Q -2

-3

-4 4DE-4 6DE-4 8DE-4 1DE-3 12E-3 14E-3 16E-3

1T (11K)

FIG 4 Comparison of predicted reactivity vs TGA data for 6000 C and 9000 C Pittsburgh No 8 chars and 9000 C Zap lignite chars The chars were proshyduced at 300 Cmin up to the indicated temperashyture

the range of TGA measurements the choice of pashyrameters such as the mean pore radius and torshytuosity for that coal cannot be confirmed

The model predictions for the high heating rate chars used in Ref 15 are presented in Fig 5 (dashed lines) together with correlations based on drop tube reactor data15 (symbols) As for the chars produced in our laboratory a value of fpore = 6 Awas used In agreement with the Thiele model the apparent activation energy in the diffusion regime is well deshyscribed by half the intrinsic activation energy However predictions of the reaction rates give much lower values than the correlations based on the data For the low rank Occidental coal char this may be due to uncertainties on the deactivation process However the choice of parameters and particushylarly the value of the mean pore radius f pore also needs to be evaluated in more detail This paramshyeter represents a measure of the size of the pores which feed the reactant gas into the char Those pores can be significantly larger than the microshypores (of pores radius 6 A) especially if the char is fairly macroporous Pore size distribution measureshyments on chars similar to Occidental and FMC (produced at high heating rate) have been pershyformed by WhitelB and show a mean pore size of lOq Afor a Zap char and a bimodal distribution at 6 A and 170 A for a Pittsburgh No8 char Using ~alues of fpore of 100 Afor the Occidental and 170 A for the FMC and Toscoal predictions are shown in Fig 5 with solid lines and fit the data more accurately especially considering that the tempershy

- predictions -predictions oFMG

c- oOGG

3

2middotE TOS dgt

-9

~ 0 ~

ro -1~

o OJ -2Q

-3

-4 40E-4 6DE-4 80E-4 10E-3 12E-3 14E-3 16E-3

1T (11K)

FIG 5 Comparison of predicted reactivity (lines) vs correlations (symbols) derived from experimenshytal data from Wells et al 15 The low temperature data was measured using TGA High temperature data was obtained in a drop tube reactor The symshybols do not correspond to actual data points but serve to identifY the different correlations The dashed lines correspond to predictions using a mean pore radius of 6 A The solid lines correspond to predictions using a mean pore radius of 100 A for Occidental and 170 A for FMC and Toscoal

ature in the literature data may have been undershyestimated by as much as 2000 C 19 This underlines the importance of precisely characterizing the structure of the char in order to determine r pore

Also the fact that different values of f pore are needed in order to fit low heating rate and high heating rate char data implies a basic difference in structure between the two types of chars This is not surshyprising since low heating rate chars have been found to be effectively less porousfi probably because some reorganization of the structure is possible during the long period in which the char is fluid However pore size distribution measurements on low heating rate chars are needed in order to resolve that quesshytion Also direct diffusivity measurements20 would eliminate any amhiguity of Wheelers model in the estimation of the mean pore size and tortuosity pashyrameters This may give more reliable values of the effective diffusivity in order to calculate the reacshytivity in regime II

Conclusions

It was shown that correlations developed at low temperatures using the char hydrogen content coal

1195

i

PREDICTION OF COAL CHAR REACTIVITY

01-r-------------- a Pittsburgh 8

O~~~-r~~-+_+~~~

d

~ 1~ 005degb------------

~ Zap Lignite o ~ ~

O+-~-+~~~+_~_+~~~~

o 1Conversion

FIG 6 Comparison of TGA data (symbols) and model predictions (lines) for the reaction of a) 6000 C char from Pittsburgh No 8 at 5000 C using the ranshydom pore model with 1Jr = 50 b) 9000 C char from Zap lignite combusted at 3800 C using the volushymetric model The small fluctuations in the predicshytions correspond to slight variations in mcasured tempcrature

oxygen content and coal mineral content have given good predictions of intrinsic reactivity for different coals and chars Using these correlations combined with a random pore model for high rank coals and a volumetric model for low rank coals reactivity and reactivity variations with burnoff can be predicted

The reactivity in the pore diffusion regime was calculated using the intrinsic reactivity correlations and the Thiele model This approach necessitates the knowledge of char properties such as porosity tortuosity and mean pore radius Using a value of 6 A for the mean pore radius leads to fairly good predictions for 300 Cmin Pittsburgh No 8 chars while being inconclusive for Zap chars In the case of high he-ating rate chars pore size distribution measurements showed that values of approximately 100 Aare more appropriate and predictions using this value give a good fit of the literature data inshyvestigated Direct diffusivity measurements could also be performed and would eliminate any uncershytainty in the estimation of the mean pore size and tortuosity parameters In conclusion this analysis shows that it may be possible to extend low temshyperature reactivity models to high temperatures

Nomenclature

A = pre-exponential factor ms Cs = reactant concentration molm3

Cad = (daf) dispersed calcium in coal d = particle diameter m do = initial particle diameter m De = diffusivity m2 s E = activation energy kcajmol H char = (dar) hydrogen in char ki = reaction rate coefficient ms Mo = molecular mass of reactive gas gmol Mp = molecular mass of carbon gmol

degml (daf) oxygen in coal p = partial pressure of reactiv~ gas atm R = gas constant m3 atm(g mol K) rpore mean pore radius m

= reaction rate lsr S = internal surface area of remaining char per

mass of initial char m2 g So = internal surface area of initial char m2

g T = particle temperature K X = burnoff

Greek Symbols 1 = effectiveness factor Eo = initial porosity To = initial tortuosity (J = apparent particle density cP = Thiele modulus v = stochiometric factor 1jF = structural parameter

Acknowledgements

The authors are grateful for the support of the Morgantown Energy Technology Center of the United States Department of Energy under conshytract DE-AC21-86MC23075 for support of the study of char reactivity The authors wish to acknowledge the contributions of Hsisheng Teng and Marie DiTaranto of Advanced Fuel Research Inc to the experimental work and helpful discussions with Prof Eric Suuberg of Brown University

REFERENCES

1 FIEJIl M A Combust Flame 14 237 (1970) 2 SMITH I W Combust Flame 17 421 (1971) 3 MITCHELL R E AND McLEAN J W Nineshy

teenth Symposium (International) on Combusshytion p 1113 The Combustion Institute 1983

4 RADOVIC L R WALKER P L AND JENKIlS R G Fuel 62 849 (1983)

5 SUUBERC E M C~LO J M AND WOJTOWICZ M ACS Div of Fuel Chern Preprints 31 (3) 186 (1986)

1196 COAL COMBUSTION

6 Su J L AND PERLMUTIER D D AIChE J 14 SUUBERG E M WOJTOWIGZ M AND CALO J 31 6 973 (1985) M Twenty-Second Symposium (International)

7 BHATIA S K AND PERLMUTIER D D AlChE on Combustion p 79 The Combustion Insti shy

J 26 379 (1980) tute (1989) 15 WELLS W F KRAMER S K AND SMOOT L

D Twentieth Symposium (International) on 8 GAYALAS G R AIChE J 26 577 (1980) 9 ISHIDA 1 AJD WEN C Y Chem Eng Sci

Combustion p 1539 The Combustion Insti shy26 1031 (1971) tutc 198510 SIITH I W Nineteenth Symposium (Intershy

16 MEliTA B N AND ARIS R Chem Eng Scinational) on Combustion p 1045 Thc Comshy

26 1699 (1971)bustion Institute 1982

17 WHEELER A Adv Cata 3 249 (1951)11 THIELE E W Ind Eng Chem 31 916

18 WWTE W E M S Thesis Brigham Young(1939) University (1990)

12 SOLOMON P R SERIO M A AJD HENINGER 19 SMOOT L D Personal communication (1992)S G ACS Div of Fuel Chem Preprints 31 20 SMITH I W HARRIS D J VALIX M G AND (3) 200 (1986) TRIMM D L in Fundamental Issues in Conshy

13 SERIO M A SOLOMON P R CHARPENAY S trol of Carbon Gasification Reactivity J LashyAND SUUBERG E M to be submitted to Fuel haye and P Ehrburger (Eds) Kluwer Acashy(1992) demic Publishers p 49 1991

COMMENTS

N Y Nsakala ABB Combustion Engineering ItIC

USA The Arrhenius plot of your TGA char oxidashytion data (I believe) showed no dependence on charshytype for Illinois 6 Pittsburgh 8 zap lignite and Rosebud subbituminous coal chars Thc global acshytivation energy was -35 kcalfmole in all cases Would you expect this behaviour to hold throughshyout the whole coal rank spcctrum

Authors Reply It has been a consistent obsershyvation in our work that all the coal chars studied which have been produced under different pyrolshyysis conditions and from various coals (including chars from treated coals) showed an activation energy in the range 30 to 35 kcalmo as measured in a TGA apparatus This observation has also been made in several other studies and it is reasonable to cxpect a similar behavior for a wide range of coals

bull Robert Hurt Sandia National Laboratories USA

Your random pure model prcdictions were comshypared to data on chars that had bcen stabilized by heat treatment prior to reaction at a temperature greater than the reaction temperature In coal comshybustion chars are made in the flame-by in situ dcvolatilization-and for this situation we see very different trends Instead of large increases in area at low conversion followed by a maximum we (in our laboratory) generally observe monotonically deshycreasing surface areas Here the char formation process and thermal annealing (which tend to reshyduce surface area) occur Simultaneously with the

oxidation (which tends to increase the area inishytially) We find the net effect to be generally deshycreasing areas at the temperatures and times of inshyterest to pulverized coal combustion The existing pure models cannot predict this trend because they all consider rigid solid matrices in which pores grow by internal rcaction Have you observed such efshyfects and can you imagine how your model may acshycount for it

Authors Reply Proccsscs like thermal annealing occur at high temperature and cannot be acshycounted for in pore models in their present form Howcver the range of validity of porc models is usually at low temperature in the intrinsic regime where the reaction occurs all through the particle In our case at high temperature (ie in the pore diffusion regime) we have assumed that the parti shycles react following the shrinking core model U sshying this model the specific surface area (mg) is assumed to remain constant with burnofT Your exshyperimental observations of generally decreasing surshylace area with burnoff cannot be explained by our current model

bull Reginald E Mitchell Stanford University USA

The calculation of effectiveness factors from Thiele moduli requires knowledge of thc order of reaction with respect to the oxygen concentration Nhat values did you use for reaction orders and why were these values selected

Authors Reply A reaction order of 1 is assumed

1197 PREDICTION OF COAL CHAR REACTIVITY

in the model This choice is consistent with the apshyproach taken in the comprehensive code for enshytrained coal combustors and gasifiers for which this model is being developed (PCGC-2 from Brigham Young University) Several studies have shown that the true reaction order is probably fractional However using a reaction order of 1 typically leads to an error in rate of less than a factor of 1 5 if the true reaction order is for example 05 At this stage of our work this range of uncertainty is acceptable given the other uncertainties in the model and in the available data

bull Ian W Smith CSIRO Division of Coal amp Energy

Technology Australia The paper deals with low and high rank coals-the former coals are stated to be affected by catalysts but the high rank materials not Given the regime I nature of the measureshyments why do not the minerals present catalyse the reaction of the high rank coals

Authors Reply For low rank coals we found the best correlation of reactivity with the amount of dispersed calcium most of which is exchanged on carboxyl groups The concentration of carboxyl groups is not significant in high rank coals For this reason the concentration of well dispersed alkali metals is not high for high rank coals and there is little catalytic enhancement of the char gasification rate Unlike low rank coals demineralization of high rank coals does not significantly reduce the gasiflshy

cation rate In fact these rates have been reported to increase slightly This has been attributed to reshymoval of minerals which block pores or mild oxishydation of the coal during the demineralization proshycess which produces a less fluid coal and a more reactive char after pyrolysis

bull Fiero Salatino Universita di Napoli Italy The

Thiele model strictly applies to solids characterized by unimodal or nearly unimodal pore site distrishybution Did the materials used in your study corshyrespond to this hypothesis and if not would you comment on the effect that this simplification might have on the reliability of your diffusion-limited combustion model

Authors Reply The effect of the pore size disshytribution is folded into the value of the effective diffusivity Deff- Since it is difficult to evaluate Del we assumed the Knudsen diffusivity to hold and used a mean pore radius to account for the pore size distribution This approach is only approximate since the chars used in our study are probably not unishymodal However only a certain range of pore sizes will have an impact on diffusion limitations the mishycropore-mesopore range From our simulations for the chars where the pore size distribution is known a reasonable prediction is obtained for the reactivshyity in the pore diffusion regime Additional comshyparisons are underway in order to fully validate this approach

1190 COAL COMBUSTION

scription of the char morphology For this purpose the Random Pore Model derived for the kinetic reshygime by Bhatia and Perlmutter7 and Cavalas8 was used which calculates surface area changes during reaction For low rank coals the intrinsic reactivity is primarily dependent on catalytic effects In that case the important parameter is the number of catshyalytic sites distributed through the volume of the particle and reactivity can then be represented by a volumetric model 9

In order to predict reactivity at high temperashytures (regime II) one has to account for pore difshyfusion limitations The model follows the approach described by SmithlO using the Thiele model ll to derive the value of the global reaction rate as a function of the intrinsic reaction rate and char properties such as char porosity tortuosity and mean pore radius

The transition region from kinetic to pore diffushysion control was not explicitly investigated since numerical solutions are usually necessary to calcushylate the reactive gas concentration profile in the particle and would be too detailed to be included in a comprehensive combustion code However this region represents a relatively narrow range in temshyperature and most of the available data can be conshysidered to fall in either the kinetic or pure pore diffusion regime

Experimental

Tent Measurements

Correlations have been developed using an exshytensive database of Critical Temperature (Terit)

measurements obtained in our laboratory Terit is an index of char reactivity which is related to the conshycentration of accessible active sites 12 The detershymination of Teri relies on a TCA technique in which the weight loss is measured while the sample is heated at a constant heating rate in the presence of the reactive gas The temperature (Tcrit) at which the rate of weight loss reaches a value of 0065 g g-min (on an original char mass basis) is recorded and gives an indication of the reactivity of the char The value of the reaction rate used for measureshyment of Tcrit was chosen to be high enough to be easily measurable but low enough so that it is still in the kinetic regime based on a Thiele modulus calculation 11 In practice depending on the nature of the char Teri has been observed to occur either in the kinetic regime or at the beginning of the pore diffusion regime The chars used for the developshyment of correlations were produced in several difshyferent reactors and under widely different pyrolysis conditions 13

Isothermal Reactidty Measurements

The reaction rates at low temperatures were measured isothermally using a Dupont 951 TCA All experiments were performed using an oxygen partial pressure of 172 kPa a flow rate of 240 cc min and a coal sample size of 10 mg The chars were prepared by heating at 30deg Cmin in helium up to the pyrolysis temperature followed by imshymediate cooling to the gasification temperature

Model Development for Kinetic Regime

Reactivity Model

Following the usual approach (eg see Smith l~ the reaction rate in the kinetic regime is given by

rp = S ki II MI Cs with

C = - (1)k = Aexp ( - T) p

S RT

where S = internal surface area of remaining char k = intrinsic rate coefficient II = stochiometric factor Mp = molecular weight of carbon and Cs

= oxidant concentration A reaction order of 1 was assumed although studies14 showed the reaction order to vary between 0 and 1 with the type of char and coal This approximation however typishycally leads to an error in rate of less than a factor of l5 if the true reaction order is for example 05 Future developments of the model will include a fractional reaction order

For high rank coals the value of S is given by the Random Pore Model7bull8 S = So(l - X)(l - 1Jf In(l - xW 5 where So = internal surface area of original char 1Jf = structural parameter X = char burnoff For low rank coals the volumetric model is used and S becomes S = So(1 - X)

The pre-exponential factor A of the coefficient ki includes the number of active sites The variashytions of A with coal type and pyrolysis conditions can be estimated with correlations based on Tcrit

measurements and can be written for high rank coals (HRC) as A = AHRC f(Hehar) g(Ocoal) where f(Hchar) and g(Ocoal) are char hydrogen and coal oxshyygen correlation functions respectively and AHRC

is the baseline frequency factor for high rank coals The corresponding factor for low rank coals (LRC) is given by A = ALRC f(Hchar) g(Otual) h(Clld) where h(Cad) is a correlation function based on the disshypersed calcium concentration

Development of Correlations

Correlations derived from Teri measurements can be used to calculate intrinsic reactivity The funcshy

1191 PREDICTION OF COAL CHAR REACTIVITY

00017

000165

00016

000155

00015 D

~ 000145 sshy0

00014

000135 I shy 00013

000125

00012

000115

ozap o pit Dill

00011 6 ros

000105 0 2 3 4 5 6

Hchar

FIG 1 Correlation of lT with the perccnt (daf) hydrogcn content of the char (Hch) for Zap Roseshybud Illinois No 6 and Pittsburgh No 8 coals The chars were produced in different reactors and unshyder different pyrolysis conditions 13

tions f g and h are expected to be increasing funcshytions of the char hydrogen content coal oxygen content and coal mineral content respectively Since Terit decreases with reactivity 1Tcrit was found to be a more appropriate quantity to use to develop correlations Linear correlations of 1Tclit versus the above parameters were obtained as discussed beshylow

Hchar Correlation

Good correlations of 1Tcrit vs the char hydrogen concentration (Hchar) were obtained for all coals It can be seen from Fig 1 that all the correlations have similar slopes which implies that the degree of pyrolysis of the char has a similar influence for all coals The data also seem to show a plateau for Hchar values higher than 35 The initial loss of hydrogen during pyrolysis is primarily aliphatic which does not have the annealing effect of the loss of aromatic hydrogen during the latter stages of pyshyrolysis 1213

Gcoal Correlation

For similar values of Hchan the value of 1Tcrit for different coals is expected to correlate with oxshyygen andor mineral content of the coals As shown in Fig 2 1Tcrit (using chars produced at 30deg C min up to 900deg C) and the coal oxygen content show a reasonable correlation The data points with low oxygen content which fall well below the line corshyrespond to very fluid coals In that case the low

00015 0

000145 0 0

00014

OJ000135

~ 00013

lte

0 8000125I shy 00012

000115 OArgonne coals

00011 oExxon coalsc 0 000105

0 5 10 15 20 25

oxygen (daf)

FIG 2 Correlation of lTcrit with the percent (daf) coal oxygen contcnt (O~ for the Argonne eoals and a set of Exxon coals The chars were produced at 30deg Cmin up to 9000 C and have similar hydrogen contents

heating rate used to produce the chars also has an impact on its reactivity

Mineral Correlation

Since mineral concentration is important only for low rank coals a correlation was obtained using coals of oxygen content higher than 13 Dispersed calshycium was found to be the main mineral responsible for reactivity enhancementl3 and was used to deshyvelop the correlation shown in Fig 3 For low rank coals the influence of calcium needs to be sepashy

00015

o000145 o

00014 o ~ s- 0 0

O t 000135 0 I- 0

00013

000125 lOExxon coalsl

00012 w-~~~w~w~~~--~--J

o 02 04 06 08 12

dispersed calcium

FIG 3 Correlation of lTcrit with the percent (daf) of dispersed calcium (Cad) for a set of Exxon coals The chars were produced at 30 Cmin up to 9000 C

1192 COAL COMBUSTION

rated from the effect of coal rank Since deminershyalized low rank coals show a reactivity similar to that of medium rank coals (of oxygen content apshyproximately 13) the effect of oxygen content is minor compared to that of calcium In order to acshycount for this in the model the value of 0eo l is kept constant and set equal to 13 for coals of oxshyygen content higher than 13

Other second order effects include the combined effect of fluidity and pyrolysis heating rate on char reactivity This can be important for fluid coals and leads to separate lTerit vs Hebar correlations for different heating rates At this point few measureshyments have been done on low heating rate fluid chars and this will be the object of a future study

Individual correlation plots (shown in Figs 1-3) give an uncertainty in Terit within plusmn20deg C which is approximately equivalent to a factor of 2 in rate With the Heban Oma and Cad correlations used sishymultaneously Terit can be predicted within plusmn40deg C It should be noted that the experimental uncershytainty of the measurement of Terit is plusmn10deg C

To overcome potential shortcomings of using sevshyeral correlations simultaneously it would be reashysonable to use the Hebar correlation alone associshyated with one Tnit measurement as input However for the cases presented below the three correlashytions were used

I ntnnsic Rate Predictions

At Teritgt the rate of weight loss is OOOlsec By substituting the values of lTeit and rl into equashytion (1) the functions f g and h can be calculated It is implicitly assumed that 1 - X is constant and equal to 085 (which is the experimental value usushyally observed) at the point where Teit is measured So is also kept constant and was taken as a first apshyproximation to be 300 m2g The value of the acshytivation energy E was evaluated from TGA isoshythermal data and was found to vary between 28 and 34 kcalmol for the chars studied As a first approximation a value of 30 kcalmol was used in the model

For high rank coals (Oeoal lt 13) equation (1) becomes

-15000 rl = 822 exp -- shy

l

exp(14 Hehar) exp(0263 Oml) S (2)

For low rank coals (Oenal gt 13) the equivalent of equation (1) is

-15000 rl = 686 x 102 exp -- shy

T

exp(14 Hdmr) exp(154 Cad) S (3)

The reaction rates correspond to the case where the oxygen partial pressure is 172 kPa

Predictions of reaction rates using equations (2) and (3) have been compared with data obtained in our laboratory (Zap lignite chars) as well as literashyture data from Wells et al 15 (FMC Occidental and Toscoal chars) The parameters used are indicated in Table I The parameters for the literature chars were directly taken from reference 15 The reactivshyities of the Pittsburgh No 8 chars were not preshydicted (since good correlations of low heating rate fluid coals have not been developed yet) but were fit to the data Figures 4 and 5 show the good agreement obtained between the predictions and the low temperature TGA data for the chars studshyied

Predictions of reaction rates as a function of burnoff are presented in Fig 6 for pittsburgh No 8 (high rank coal) 600deg C char and Zap lignite (low rank coal) 900deg C char using the random pore model and volumetric model respectively The predictions were done using a single adjustable parameter (without using the correlations) in order to demshyonstrate the adequacy of the burnoff models choshysen The good fit (within 20) obtained in the full range of burnoff supports the use of these models

Model Development for Pore Diffusion Regime

Calculation of Reaction Rate

In the diffusion regime the observed reaction rate can be expressed as a function of the intrinsic reshyaction rate

(4)

where YJ = effectiveness factor According to Iehta and Aris J(j the effectiveness factor can be detershymined using the relations between YJ and YJqi where cjJ is the Thiele modulus II In the case of high Thiele modulus YJcjJ2 tends to lYJ ie YJ tends to lcjJ

The Thiele modulus is given by

d ( S k )0gt

(J () IcjJ-- (5)

6 De V MMo

with d = particle diameter = apparent particle (J

density De = effective diffusivity Mo = molecular weight of oxidant From the simple pore-structure model of Wheeler17 De can be calculated

05 _ 2 T E

De - 97 x 10 rpore (

-)

- (6) Mo To

where rpltre = mean pore radius Eo = porosity To = tortuosity

1193 PREDICTION OF COAL CHAR REACTIVITY

TABLE I Parameters Used in Reactivity Calculations

PIT 900deg C amp ZAP

Char 600degC 900deg C FMC OCC TOS

Parent Coal Pittsburgh Zap Pittsburgh Wyodak Utah

Rate (OCs) 05 05 103 10 10

pound0 01 03 013 055 011

10 2 2 2 2 2

dpm) 70 70 90 63 55

rpc (A) 6 6 170 100 170

So (m2g) 300 300 300 300 300

00 (gcm3) 13 13 152 086 143

01 () - 21 9 21 118

Hcha() - 074 2 21 318

M () - 1 0 11 0

Note 900deg C and 600deg C Pittsburgh No 8 chars and 900deg C Zap lignite char have been produced in our laboratory FMC Occidental and Toscoal are chars from Reference 15

estimated from measured values for similar chars or coals

The global reaction rate used in the model can then be written

05 6 ( Mgt So ki)r=-Dv--- vMC (7)

P deMo 0 p bull

The reaction rate depends on the intrinsic rate and char properties such as the tortuosity porosity and the mean pore radius The tortuosity parameter has been introduced to account for the connectivity of pores Its value has generally been taken between 15 and 10 usually equal to 2 The porosity and the mean pore radius can be measured However their determination may be difficult in the case of a non-uniform pore size distribution

Comparison of Predictions with Data

Using the physical and chemical parameters inshydicated in Table I and kinetic parameters derived earlier predictions in the pore diffusion regime for the Pittsburgh No 8 and Zap lignite chars proshy

duced at 30deg Cjmin are shown in Fig 4 For low rank coals we used a deactivation function to acshycount for the fact that catalysts have less impact on reactivity at high temperatures 1213 Following this deactivation function at high temperatures the reactivity of a low rank coal is made to be equal to that of a coal of 13 oxygen Although the number of data points in the diffusion regime is limited the transition between kinetic and diffusion control is fairly well represented by the theory for the two Pittsburgh No8 chars using a mean pore radius of 6 t which corresponds to the size of micropores Concerning the Zap char no internal diffusion limshyitations are present (since there is no change in ac~ tivation energy) but a leveling off of the rate is obshyserved This behavior is due to external mass transfer resistances in the TGA Under our conditions the maximum measurable rate in the TGA is approxishymately 1 gjg-min However under true pulvershyized coal combustion conditions external diffusion limitations are not encountered until temperatures typically higher than 1500 K Since no pore diffushysion limitations are observed for the Zap char in

1194 COAL COMBUSTION

4 4

3 -predictions OPIT 900 Cchar

CshymiddotE 2

ePIT 600 Cchar IoZAP 900 Cchar

dgt -9

~ s 0 0 15 ro (J)

gshy -1

~

OJ Q -2

-3

-4 4DE-4 6DE-4 8DE-4 1DE-3 12E-3 14E-3 16E-3

1T (11K)

FIG 4 Comparison of predicted reactivity vs TGA data for 6000 C and 9000 C Pittsburgh No 8 chars and 9000 C Zap lignite chars The chars were proshyduced at 300 Cmin up to the indicated temperashyture

the range of TGA measurements the choice of pashyrameters such as the mean pore radius and torshytuosity for that coal cannot be confirmed

The model predictions for the high heating rate chars used in Ref 15 are presented in Fig 5 (dashed lines) together with correlations based on drop tube reactor data15 (symbols) As for the chars produced in our laboratory a value of fpore = 6 Awas used In agreement with the Thiele model the apparent activation energy in the diffusion regime is well deshyscribed by half the intrinsic activation energy However predictions of the reaction rates give much lower values than the correlations based on the data For the low rank Occidental coal char this may be due to uncertainties on the deactivation process However the choice of parameters and particushylarly the value of the mean pore radius f pore also needs to be evaluated in more detail This paramshyeter represents a measure of the size of the pores which feed the reactant gas into the char Those pores can be significantly larger than the microshypores (of pores radius 6 A) especially if the char is fairly macroporous Pore size distribution measureshyments on chars similar to Occidental and FMC (produced at high heating rate) have been pershyformed by WhitelB and show a mean pore size of lOq Afor a Zap char and a bimodal distribution at 6 A and 170 A for a Pittsburgh No8 char Using ~alues of fpore of 100 Afor the Occidental and 170 A for the FMC and Toscoal predictions are shown in Fig 5 with solid lines and fit the data more accurately especially considering that the tempershy

- predictions -predictions oFMG

c- oOGG

3

2middotE TOS dgt

-9

~ 0 ~

ro -1~

o OJ -2Q

-3

-4 40E-4 6DE-4 80E-4 10E-3 12E-3 14E-3 16E-3

1T (11K)

FIG 5 Comparison of predicted reactivity (lines) vs correlations (symbols) derived from experimenshytal data from Wells et al 15 The low temperature data was measured using TGA High temperature data was obtained in a drop tube reactor The symshybols do not correspond to actual data points but serve to identifY the different correlations The dashed lines correspond to predictions using a mean pore radius of 6 A The solid lines correspond to predictions using a mean pore radius of 100 A for Occidental and 170 A for FMC and Toscoal

ature in the literature data may have been undershyestimated by as much as 2000 C 19 This underlines the importance of precisely characterizing the structure of the char in order to determine r pore

Also the fact that different values of f pore are needed in order to fit low heating rate and high heating rate char data implies a basic difference in structure between the two types of chars This is not surshyprising since low heating rate chars have been found to be effectively less porousfi probably because some reorganization of the structure is possible during the long period in which the char is fluid However pore size distribution measurements on low heating rate chars are needed in order to resolve that quesshytion Also direct diffusivity measurements20 would eliminate any amhiguity of Wheelers model in the estimation of the mean pore size and tortuosity pashyrameters This may give more reliable values of the effective diffusivity in order to calculate the reacshytivity in regime II

Conclusions

It was shown that correlations developed at low temperatures using the char hydrogen content coal

1195

i

PREDICTION OF COAL CHAR REACTIVITY

01-r-------------- a Pittsburgh 8

O~~~-r~~-+_+~~~

d

~ 1~ 005degb------------

~ Zap Lignite o ~ ~

O+-~-+~~~+_~_+~~~~

o 1Conversion

FIG 6 Comparison of TGA data (symbols) and model predictions (lines) for the reaction of a) 6000 C char from Pittsburgh No 8 at 5000 C using the ranshydom pore model with 1Jr = 50 b) 9000 C char from Zap lignite combusted at 3800 C using the volushymetric model The small fluctuations in the predicshytions correspond to slight variations in mcasured tempcrature

oxygen content and coal mineral content have given good predictions of intrinsic reactivity for different coals and chars Using these correlations combined with a random pore model for high rank coals and a volumetric model for low rank coals reactivity and reactivity variations with burnoff can be predicted

The reactivity in the pore diffusion regime was calculated using the intrinsic reactivity correlations and the Thiele model This approach necessitates the knowledge of char properties such as porosity tortuosity and mean pore radius Using a value of 6 A for the mean pore radius leads to fairly good predictions for 300 Cmin Pittsburgh No 8 chars while being inconclusive for Zap chars In the case of high he-ating rate chars pore size distribution measurements showed that values of approximately 100 Aare more appropriate and predictions using this value give a good fit of the literature data inshyvestigated Direct diffusivity measurements could also be performed and would eliminate any uncershytainty in the estimation of the mean pore size and tortuosity parameters In conclusion this analysis shows that it may be possible to extend low temshyperature reactivity models to high temperatures

Nomenclature

A = pre-exponential factor ms Cs = reactant concentration molm3

Cad = (daf) dispersed calcium in coal d = particle diameter m do = initial particle diameter m De = diffusivity m2 s E = activation energy kcajmol H char = (dar) hydrogen in char ki = reaction rate coefficient ms Mo = molecular mass of reactive gas gmol Mp = molecular mass of carbon gmol

degml (daf) oxygen in coal p = partial pressure of reactiv~ gas atm R = gas constant m3 atm(g mol K) rpore mean pore radius m

= reaction rate lsr S = internal surface area of remaining char per

mass of initial char m2 g So = internal surface area of initial char m2

g T = particle temperature K X = burnoff

Greek Symbols 1 = effectiveness factor Eo = initial porosity To = initial tortuosity (J = apparent particle density cP = Thiele modulus v = stochiometric factor 1jF = structural parameter

Acknowledgements

The authors are grateful for the support of the Morgantown Energy Technology Center of the United States Department of Energy under conshytract DE-AC21-86MC23075 for support of the study of char reactivity The authors wish to acknowledge the contributions of Hsisheng Teng and Marie DiTaranto of Advanced Fuel Research Inc to the experimental work and helpful discussions with Prof Eric Suuberg of Brown University

REFERENCES

1 FIEJIl M A Combust Flame 14 237 (1970) 2 SMITH I W Combust Flame 17 421 (1971) 3 MITCHELL R E AND McLEAN J W Nineshy

teenth Symposium (International) on Combusshytion p 1113 The Combustion Institute 1983

4 RADOVIC L R WALKER P L AND JENKIlS R G Fuel 62 849 (1983)

5 SUUBERC E M C~LO J M AND WOJTOWICZ M ACS Div of Fuel Chern Preprints 31 (3) 186 (1986)

1196 COAL COMBUSTION

6 Su J L AND PERLMUTIER D D AIChE J 14 SUUBERG E M WOJTOWIGZ M AND CALO J 31 6 973 (1985) M Twenty-Second Symposium (International)

7 BHATIA S K AND PERLMUTIER D D AlChE on Combustion p 79 The Combustion Insti shy

J 26 379 (1980) tute (1989) 15 WELLS W F KRAMER S K AND SMOOT L

D Twentieth Symposium (International) on 8 GAYALAS G R AIChE J 26 577 (1980) 9 ISHIDA 1 AJD WEN C Y Chem Eng Sci

Combustion p 1539 The Combustion Insti shy26 1031 (1971) tutc 198510 SIITH I W Nineteenth Symposium (Intershy

16 MEliTA B N AND ARIS R Chem Eng Scinational) on Combustion p 1045 Thc Comshy

26 1699 (1971)bustion Institute 1982

17 WHEELER A Adv Cata 3 249 (1951)11 THIELE E W Ind Eng Chem 31 916

18 WWTE W E M S Thesis Brigham Young(1939) University (1990)

12 SOLOMON P R SERIO M A AJD HENINGER 19 SMOOT L D Personal communication (1992)S G ACS Div of Fuel Chem Preprints 31 20 SMITH I W HARRIS D J VALIX M G AND (3) 200 (1986) TRIMM D L in Fundamental Issues in Conshy

13 SERIO M A SOLOMON P R CHARPENAY S trol of Carbon Gasification Reactivity J LashyAND SUUBERG E M to be submitted to Fuel haye and P Ehrburger (Eds) Kluwer Acashy(1992) demic Publishers p 49 1991

COMMENTS

N Y Nsakala ABB Combustion Engineering ItIC

USA The Arrhenius plot of your TGA char oxidashytion data (I believe) showed no dependence on charshytype for Illinois 6 Pittsburgh 8 zap lignite and Rosebud subbituminous coal chars Thc global acshytivation energy was -35 kcalfmole in all cases Would you expect this behaviour to hold throughshyout the whole coal rank spcctrum

Authors Reply It has been a consistent obsershyvation in our work that all the coal chars studied which have been produced under different pyrolshyysis conditions and from various coals (including chars from treated coals) showed an activation energy in the range 30 to 35 kcalmo as measured in a TGA apparatus This observation has also been made in several other studies and it is reasonable to cxpect a similar behavior for a wide range of coals

bull Robert Hurt Sandia National Laboratories USA

Your random pure model prcdictions were comshypared to data on chars that had bcen stabilized by heat treatment prior to reaction at a temperature greater than the reaction temperature In coal comshybustion chars are made in the flame-by in situ dcvolatilization-and for this situation we see very different trends Instead of large increases in area at low conversion followed by a maximum we (in our laboratory) generally observe monotonically deshycreasing surface areas Here the char formation process and thermal annealing (which tend to reshyduce surface area) occur Simultaneously with the

oxidation (which tends to increase the area inishytially) We find the net effect to be generally deshycreasing areas at the temperatures and times of inshyterest to pulverized coal combustion The existing pure models cannot predict this trend because they all consider rigid solid matrices in which pores grow by internal rcaction Have you observed such efshyfects and can you imagine how your model may acshycount for it

Authors Reply Proccsscs like thermal annealing occur at high temperature and cannot be acshycounted for in pore models in their present form Howcver the range of validity of porc models is usually at low temperature in the intrinsic regime where the reaction occurs all through the particle In our case at high temperature (ie in the pore diffusion regime) we have assumed that the parti shycles react following the shrinking core model U sshying this model the specific surface area (mg) is assumed to remain constant with burnofT Your exshyperimental observations of generally decreasing surshylace area with burnoff cannot be explained by our current model

bull Reginald E Mitchell Stanford University USA

The calculation of effectiveness factors from Thiele moduli requires knowledge of thc order of reaction with respect to the oxygen concentration Nhat values did you use for reaction orders and why were these values selected

Authors Reply A reaction order of 1 is assumed

1197 PREDICTION OF COAL CHAR REACTIVITY

in the model This choice is consistent with the apshyproach taken in the comprehensive code for enshytrained coal combustors and gasifiers for which this model is being developed (PCGC-2 from Brigham Young University) Several studies have shown that the true reaction order is probably fractional However using a reaction order of 1 typically leads to an error in rate of less than a factor of 1 5 if the true reaction order is for example 05 At this stage of our work this range of uncertainty is acceptable given the other uncertainties in the model and in the available data

bull Ian W Smith CSIRO Division of Coal amp Energy

Technology Australia The paper deals with low and high rank coals-the former coals are stated to be affected by catalysts but the high rank materials not Given the regime I nature of the measureshyments why do not the minerals present catalyse the reaction of the high rank coals

Authors Reply For low rank coals we found the best correlation of reactivity with the amount of dispersed calcium most of which is exchanged on carboxyl groups The concentration of carboxyl groups is not significant in high rank coals For this reason the concentration of well dispersed alkali metals is not high for high rank coals and there is little catalytic enhancement of the char gasification rate Unlike low rank coals demineralization of high rank coals does not significantly reduce the gasiflshy

cation rate In fact these rates have been reported to increase slightly This has been attributed to reshymoval of minerals which block pores or mild oxishydation of the coal during the demineralization proshycess which produces a less fluid coal and a more reactive char after pyrolysis

bull Fiero Salatino Universita di Napoli Italy The

Thiele model strictly applies to solids characterized by unimodal or nearly unimodal pore site distrishybution Did the materials used in your study corshyrespond to this hypothesis and if not would you comment on the effect that this simplification might have on the reliability of your diffusion-limited combustion model

Authors Reply The effect of the pore size disshytribution is folded into the value of the effective diffusivity Deff- Since it is difficult to evaluate Del we assumed the Knudsen diffusivity to hold and used a mean pore radius to account for the pore size distribution This approach is only approximate since the chars used in our study are probably not unishymodal However only a certain range of pore sizes will have an impact on diffusion limitations the mishycropore-mesopore range From our simulations for the chars where the pore size distribution is known a reasonable prediction is obtained for the reactivshyity in the pore diffusion regime Additional comshyparisons are underway in order to fully validate this approach

1191 PREDICTION OF COAL CHAR REACTIVITY

00017

000165

00016

000155

00015 D

~ 000145 sshy0

00014

000135 I shy 00013

000125

00012

000115

ozap o pit Dill

00011 6 ros

000105 0 2 3 4 5 6

Hchar

FIG 1 Correlation of lT with the perccnt (daf) hydrogcn content of the char (Hch) for Zap Roseshybud Illinois No 6 and Pittsburgh No 8 coals The chars were produced in different reactors and unshyder different pyrolysis conditions 13

tions f g and h are expected to be increasing funcshytions of the char hydrogen content coal oxygen content and coal mineral content respectively Since Terit decreases with reactivity 1Tcrit was found to be a more appropriate quantity to use to develop correlations Linear correlations of 1Tclit versus the above parameters were obtained as discussed beshylow

Hchar Correlation

Good correlations of 1Tcrit vs the char hydrogen concentration (Hchar) were obtained for all coals It can be seen from Fig 1 that all the correlations have similar slopes which implies that the degree of pyrolysis of the char has a similar influence for all coals The data also seem to show a plateau for Hchar values higher than 35 The initial loss of hydrogen during pyrolysis is primarily aliphatic which does not have the annealing effect of the loss of aromatic hydrogen during the latter stages of pyshyrolysis 1213

Gcoal Correlation

For similar values of Hchan the value of 1Tcrit for different coals is expected to correlate with oxshyygen andor mineral content of the coals As shown in Fig 2 1Tcrit (using chars produced at 30deg C min up to 900deg C) and the coal oxygen content show a reasonable correlation The data points with low oxygen content which fall well below the line corshyrespond to very fluid coals In that case the low

00015 0

000145 0 0

00014

OJ000135

~ 00013

lte

0 8000125I shy 00012

000115 OArgonne coals

00011 oExxon coalsc 0 000105

0 5 10 15 20 25

oxygen (daf)

FIG 2 Correlation of lTcrit with the percent (daf) coal oxygen contcnt (O~ for the Argonne eoals and a set of Exxon coals The chars were produced at 30deg Cmin up to 9000 C and have similar hydrogen contents

heating rate used to produce the chars also has an impact on its reactivity

Mineral Correlation

Since mineral concentration is important only for low rank coals a correlation was obtained using coals of oxygen content higher than 13 Dispersed calshycium was found to be the main mineral responsible for reactivity enhancementl3 and was used to deshyvelop the correlation shown in Fig 3 For low rank coals the influence of calcium needs to be sepashy

00015

o000145 o

00014 o ~ s- 0 0

O t 000135 0 I- 0

00013

000125 lOExxon coalsl

00012 w-~~~w~w~~~--~--J

o 02 04 06 08 12

dispersed calcium

FIG 3 Correlation of lTcrit with the percent (daf) of dispersed calcium (Cad) for a set of Exxon coals The chars were produced at 30 Cmin up to 9000 C

1192 COAL COMBUSTION

rated from the effect of coal rank Since deminershyalized low rank coals show a reactivity similar to that of medium rank coals (of oxygen content apshyproximately 13) the effect of oxygen content is minor compared to that of calcium In order to acshycount for this in the model the value of 0eo l is kept constant and set equal to 13 for coals of oxshyygen content higher than 13

Other second order effects include the combined effect of fluidity and pyrolysis heating rate on char reactivity This can be important for fluid coals and leads to separate lTerit vs Hebar correlations for different heating rates At this point few measureshyments have been done on low heating rate fluid chars and this will be the object of a future study

Individual correlation plots (shown in Figs 1-3) give an uncertainty in Terit within plusmn20deg C which is approximately equivalent to a factor of 2 in rate With the Heban Oma and Cad correlations used sishymultaneously Terit can be predicted within plusmn40deg C It should be noted that the experimental uncershytainty of the measurement of Terit is plusmn10deg C

To overcome potential shortcomings of using sevshyeral correlations simultaneously it would be reashysonable to use the Hebar correlation alone associshyated with one Tnit measurement as input However for the cases presented below the three correlashytions were used

I ntnnsic Rate Predictions

At Teritgt the rate of weight loss is OOOlsec By substituting the values of lTeit and rl into equashytion (1) the functions f g and h can be calculated It is implicitly assumed that 1 - X is constant and equal to 085 (which is the experimental value usushyally observed) at the point where Teit is measured So is also kept constant and was taken as a first apshyproximation to be 300 m2g The value of the acshytivation energy E was evaluated from TGA isoshythermal data and was found to vary between 28 and 34 kcalmol for the chars studied As a first approximation a value of 30 kcalmol was used in the model

For high rank coals (Oeoal lt 13) equation (1) becomes

-15000 rl = 822 exp -- shy

l

exp(14 Hehar) exp(0263 Oml) S (2)

For low rank coals (Oenal gt 13) the equivalent of equation (1) is

-15000 rl = 686 x 102 exp -- shy

T

exp(14 Hdmr) exp(154 Cad) S (3)

The reaction rates correspond to the case where the oxygen partial pressure is 172 kPa

Predictions of reaction rates using equations (2) and (3) have been compared with data obtained in our laboratory (Zap lignite chars) as well as literashyture data from Wells et al 15 (FMC Occidental and Toscoal chars) The parameters used are indicated in Table I The parameters for the literature chars were directly taken from reference 15 The reactivshyities of the Pittsburgh No 8 chars were not preshydicted (since good correlations of low heating rate fluid coals have not been developed yet) but were fit to the data Figures 4 and 5 show the good agreement obtained between the predictions and the low temperature TGA data for the chars studshyied

Predictions of reaction rates as a function of burnoff are presented in Fig 6 for pittsburgh No 8 (high rank coal) 600deg C char and Zap lignite (low rank coal) 900deg C char using the random pore model and volumetric model respectively The predictions were done using a single adjustable parameter (without using the correlations) in order to demshyonstrate the adequacy of the burnoff models choshysen The good fit (within 20) obtained in the full range of burnoff supports the use of these models

Model Development for Pore Diffusion Regime

Calculation of Reaction Rate

In the diffusion regime the observed reaction rate can be expressed as a function of the intrinsic reshyaction rate

(4)

where YJ = effectiveness factor According to Iehta and Aris J(j the effectiveness factor can be detershymined using the relations between YJ and YJqi where cjJ is the Thiele modulus II In the case of high Thiele modulus YJcjJ2 tends to lYJ ie YJ tends to lcjJ

The Thiele modulus is given by

d ( S k )0gt

(J () IcjJ-- (5)

6 De V MMo

with d = particle diameter = apparent particle (J

density De = effective diffusivity Mo = molecular weight of oxidant From the simple pore-structure model of Wheeler17 De can be calculated

05 _ 2 T E

De - 97 x 10 rpore (

-)

- (6) Mo To

where rpltre = mean pore radius Eo = porosity To = tortuosity

1193 PREDICTION OF COAL CHAR REACTIVITY

TABLE I Parameters Used in Reactivity Calculations

PIT 900deg C amp ZAP

Char 600degC 900deg C FMC OCC TOS

Parent Coal Pittsburgh Zap Pittsburgh Wyodak Utah

Rate (OCs) 05 05 103 10 10

pound0 01 03 013 055 011

10 2 2 2 2 2

dpm) 70 70 90 63 55

rpc (A) 6 6 170 100 170

So (m2g) 300 300 300 300 300

00 (gcm3) 13 13 152 086 143

01 () - 21 9 21 118

Hcha() - 074 2 21 318

M () - 1 0 11 0

Note 900deg C and 600deg C Pittsburgh No 8 chars and 900deg C Zap lignite char have been produced in our laboratory FMC Occidental and Toscoal are chars from Reference 15

estimated from measured values for similar chars or coals

The global reaction rate used in the model can then be written

05 6 ( Mgt So ki)r=-Dv--- vMC (7)

P deMo 0 p bull

The reaction rate depends on the intrinsic rate and char properties such as the tortuosity porosity and the mean pore radius The tortuosity parameter has been introduced to account for the connectivity of pores Its value has generally been taken between 15 and 10 usually equal to 2 The porosity and the mean pore radius can be measured However their determination may be difficult in the case of a non-uniform pore size distribution

Comparison of Predictions with Data

Using the physical and chemical parameters inshydicated in Table I and kinetic parameters derived earlier predictions in the pore diffusion regime for the Pittsburgh No 8 and Zap lignite chars proshy

duced at 30deg Cjmin are shown in Fig 4 For low rank coals we used a deactivation function to acshycount for the fact that catalysts have less impact on reactivity at high temperatures 1213 Following this deactivation function at high temperatures the reactivity of a low rank coal is made to be equal to that of a coal of 13 oxygen Although the number of data points in the diffusion regime is limited the transition between kinetic and diffusion control is fairly well represented by the theory for the two Pittsburgh No8 chars using a mean pore radius of 6 t which corresponds to the size of micropores Concerning the Zap char no internal diffusion limshyitations are present (since there is no change in ac~ tivation energy) but a leveling off of the rate is obshyserved This behavior is due to external mass transfer resistances in the TGA Under our conditions the maximum measurable rate in the TGA is approxishymately 1 gjg-min However under true pulvershyized coal combustion conditions external diffusion limitations are not encountered until temperatures typically higher than 1500 K Since no pore diffushysion limitations are observed for the Zap char in

1194 COAL COMBUSTION

4 4

3 -predictions OPIT 900 Cchar

CshymiddotE 2

ePIT 600 Cchar IoZAP 900 Cchar

dgt -9

~ s 0 0 15 ro (J)

gshy -1

~

OJ Q -2

-3

-4 4DE-4 6DE-4 8DE-4 1DE-3 12E-3 14E-3 16E-3

1T (11K)

FIG 4 Comparison of predicted reactivity vs TGA data for 6000 C and 9000 C Pittsburgh No 8 chars and 9000 C Zap lignite chars The chars were proshyduced at 300 Cmin up to the indicated temperashyture

the range of TGA measurements the choice of pashyrameters such as the mean pore radius and torshytuosity for that coal cannot be confirmed

The model predictions for the high heating rate chars used in Ref 15 are presented in Fig 5 (dashed lines) together with correlations based on drop tube reactor data15 (symbols) As for the chars produced in our laboratory a value of fpore = 6 Awas used In agreement with the Thiele model the apparent activation energy in the diffusion regime is well deshyscribed by half the intrinsic activation energy However predictions of the reaction rates give much lower values than the correlations based on the data For the low rank Occidental coal char this may be due to uncertainties on the deactivation process However the choice of parameters and particushylarly the value of the mean pore radius f pore also needs to be evaluated in more detail This paramshyeter represents a measure of the size of the pores which feed the reactant gas into the char Those pores can be significantly larger than the microshypores (of pores radius 6 A) especially if the char is fairly macroporous Pore size distribution measureshyments on chars similar to Occidental and FMC (produced at high heating rate) have been pershyformed by WhitelB and show a mean pore size of lOq Afor a Zap char and a bimodal distribution at 6 A and 170 A for a Pittsburgh No8 char Using ~alues of fpore of 100 Afor the Occidental and 170 A for the FMC and Toscoal predictions are shown in Fig 5 with solid lines and fit the data more accurately especially considering that the tempershy

- predictions -predictions oFMG

c- oOGG

3

2middotE TOS dgt

-9

~ 0 ~

ro -1~

o OJ -2Q

-3

-4 40E-4 6DE-4 80E-4 10E-3 12E-3 14E-3 16E-3

1T (11K)

FIG 5 Comparison of predicted reactivity (lines) vs correlations (symbols) derived from experimenshytal data from Wells et al 15 The low temperature data was measured using TGA High temperature data was obtained in a drop tube reactor The symshybols do not correspond to actual data points but serve to identifY the different correlations The dashed lines correspond to predictions using a mean pore radius of 6 A The solid lines correspond to predictions using a mean pore radius of 100 A for Occidental and 170 A for FMC and Toscoal

ature in the literature data may have been undershyestimated by as much as 2000 C 19 This underlines the importance of precisely characterizing the structure of the char in order to determine r pore

Also the fact that different values of f pore are needed in order to fit low heating rate and high heating rate char data implies a basic difference in structure between the two types of chars This is not surshyprising since low heating rate chars have been found to be effectively less porousfi probably because some reorganization of the structure is possible during the long period in which the char is fluid However pore size distribution measurements on low heating rate chars are needed in order to resolve that quesshytion Also direct diffusivity measurements20 would eliminate any amhiguity of Wheelers model in the estimation of the mean pore size and tortuosity pashyrameters This may give more reliable values of the effective diffusivity in order to calculate the reacshytivity in regime II

Conclusions

It was shown that correlations developed at low temperatures using the char hydrogen content coal

1195

i

PREDICTION OF COAL CHAR REACTIVITY

01-r-------------- a Pittsburgh 8

O~~~-r~~-+_+~~~

d

~ 1~ 005degb------------

~ Zap Lignite o ~ ~

O+-~-+~~~+_~_+~~~~

o 1Conversion

FIG 6 Comparison of TGA data (symbols) and model predictions (lines) for the reaction of a) 6000 C char from Pittsburgh No 8 at 5000 C using the ranshydom pore model with 1Jr = 50 b) 9000 C char from Zap lignite combusted at 3800 C using the volushymetric model The small fluctuations in the predicshytions correspond to slight variations in mcasured tempcrature

oxygen content and coal mineral content have given good predictions of intrinsic reactivity for different coals and chars Using these correlations combined with a random pore model for high rank coals and a volumetric model for low rank coals reactivity and reactivity variations with burnoff can be predicted

The reactivity in the pore diffusion regime was calculated using the intrinsic reactivity correlations and the Thiele model This approach necessitates the knowledge of char properties such as porosity tortuosity and mean pore radius Using a value of 6 A for the mean pore radius leads to fairly good predictions for 300 Cmin Pittsburgh No 8 chars while being inconclusive for Zap chars In the case of high he-ating rate chars pore size distribution measurements showed that values of approximately 100 Aare more appropriate and predictions using this value give a good fit of the literature data inshyvestigated Direct diffusivity measurements could also be performed and would eliminate any uncershytainty in the estimation of the mean pore size and tortuosity parameters In conclusion this analysis shows that it may be possible to extend low temshyperature reactivity models to high temperatures

Nomenclature

A = pre-exponential factor ms Cs = reactant concentration molm3

Cad = (daf) dispersed calcium in coal d = particle diameter m do = initial particle diameter m De = diffusivity m2 s E = activation energy kcajmol H char = (dar) hydrogen in char ki = reaction rate coefficient ms Mo = molecular mass of reactive gas gmol Mp = molecular mass of carbon gmol

degml (daf) oxygen in coal p = partial pressure of reactiv~ gas atm R = gas constant m3 atm(g mol K) rpore mean pore radius m

= reaction rate lsr S = internal surface area of remaining char per

mass of initial char m2 g So = internal surface area of initial char m2

g T = particle temperature K X = burnoff

Greek Symbols 1 = effectiveness factor Eo = initial porosity To = initial tortuosity (J = apparent particle density cP = Thiele modulus v = stochiometric factor 1jF = structural parameter

Acknowledgements

The authors are grateful for the support of the Morgantown Energy Technology Center of the United States Department of Energy under conshytract DE-AC21-86MC23075 for support of the study of char reactivity The authors wish to acknowledge the contributions of Hsisheng Teng and Marie DiTaranto of Advanced Fuel Research Inc to the experimental work and helpful discussions with Prof Eric Suuberg of Brown University

REFERENCES

1 FIEJIl M A Combust Flame 14 237 (1970) 2 SMITH I W Combust Flame 17 421 (1971) 3 MITCHELL R E AND McLEAN J W Nineshy

teenth Symposium (International) on Combusshytion p 1113 The Combustion Institute 1983

4 RADOVIC L R WALKER P L AND JENKIlS R G Fuel 62 849 (1983)

5 SUUBERC E M C~LO J M AND WOJTOWICZ M ACS Div of Fuel Chern Preprints 31 (3) 186 (1986)

1196 COAL COMBUSTION

6 Su J L AND PERLMUTIER D D AIChE J 14 SUUBERG E M WOJTOWIGZ M AND CALO J 31 6 973 (1985) M Twenty-Second Symposium (International)

7 BHATIA S K AND PERLMUTIER D D AlChE on Combustion p 79 The Combustion Insti shy

J 26 379 (1980) tute (1989) 15 WELLS W F KRAMER S K AND SMOOT L

D Twentieth Symposium (International) on 8 GAYALAS G R AIChE J 26 577 (1980) 9 ISHIDA 1 AJD WEN C Y Chem Eng Sci

Combustion p 1539 The Combustion Insti shy26 1031 (1971) tutc 198510 SIITH I W Nineteenth Symposium (Intershy

16 MEliTA B N AND ARIS R Chem Eng Scinational) on Combustion p 1045 Thc Comshy

26 1699 (1971)bustion Institute 1982

17 WHEELER A Adv Cata 3 249 (1951)11 THIELE E W Ind Eng Chem 31 916

18 WWTE W E M S Thesis Brigham Young(1939) University (1990)

12 SOLOMON P R SERIO M A AJD HENINGER 19 SMOOT L D Personal communication (1992)S G ACS Div of Fuel Chem Preprints 31 20 SMITH I W HARRIS D J VALIX M G AND (3) 200 (1986) TRIMM D L in Fundamental Issues in Conshy

13 SERIO M A SOLOMON P R CHARPENAY S trol of Carbon Gasification Reactivity J LashyAND SUUBERG E M to be submitted to Fuel haye and P Ehrburger (Eds) Kluwer Acashy(1992) demic Publishers p 49 1991

COMMENTS

N Y Nsakala ABB Combustion Engineering ItIC

USA The Arrhenius plot of your TGA char oxidashytion data (I believe) showed no dependence on charshytype for Illinois 6 Pittsburgh 8 zap lignite and Rosebud subbituminous coal chars Thc global acshytivation energy was -35 kcalfmole in all cases Would you expect this behaviour to hold throughshyout the whole coal rank spcctrum

Authors Reply It has been a consistent obsershyvation in our work that all the coal chars studied which have been produced under different pyrolshyysis conditions and from various coals (including chars from treated coals) showed an activation energy in the range 30 to 35 kcalmo as measured in a TGA apparatus This observation has also been made in several other studies and it is reasonable to cxpect a similar behavior for a wide range of coals

bull Robert Hurt Sandia National Laboratories USA

Your random pure model prcdictions were comshypared to data on chars that had bcen stabilized by heat treatment prior to reaction at a temperature greater than the reaction temperature In coal comshybustion chars are made in the flame-by in situ dcvolatilization-and for this situation we see very different trends Instead of large increases in area at low conversion followed by a maximum we (in our laboratory) generally observe monotonically deshycreasing surface areas Here the char formation process and thermal annealing (which tend to reshyduce surface area) occur Simultaneously with the

oxidation (which tends to increase the area inishytially) We find the net effect to be generally deshycreasing areas at the temperatures and times of inshyterest to pulverized coal combustion The existing pure models cannot predict this trend because they all consider rigid solid matrices in which pores grow by internal rcaction Have you observed such efshyfects and can you imagine how your model may acshycount for it

Authors Reply Proccsscs like thermal annealing occur at high temperature and cannot be acshycounted for in pore models in their present form Howcver the range of validity of porc models is usually at low temperature in the intrinsic regime where the reaction occurs all through the particle In our case at high temperature (ie in the pore diffusion regime) we have assumed that the parti shycles react following the shrinking core model U sshying this model the specific surface area (mg) is assumed to remain constant with burnofT Your exshyperimental observations of generally decreasing surshylace area with burnoff cannot be explained by our current model

bull Reginald E Mitchell Stanford University USA

The calculation of effectiveness factors from Thiele moduli requires knowledge of thc order of reaction with respect to the oxygen concentration Nhat values did you use for reaction orders and why were these values selected

Authors Reply A reaction order of 1 is assumed

1197 PREDICTION OF COAL CHAR REACTIVITY

in the model This choice is consistent with the apshyproach taken in the comprehensive code for enshytrained coal combustors and gasifiers for which this model is being developed (PCGC-2 from Brigham Young University) Several studies have shown that the true reaction order is probably fractional However using a reaction order of 1 typically leads to an error in rate of less than a factor of 1 5 if the true reaction order is for example 05 At this stage of our work this range of uncertainty is acceptable given the other uncertainties in the model and in the available data

bull Ian W Smith CSIRO Division of Coal amp Energy

Technology Australia The paper deals with low and high rank coals-the former coals are stated to be affected by catalysts but the high rank materials not Given the regime I nature of the measureshyments why do not the minerals present catalyse the reaction of the high rank coals

Authors Reply For low rank coals we found the best correlation of reactivity with the amount of dispersed calcium most of which is exchanged on carboxyl groups The concentration of carboxyl groups is not significant in high rank coals For this reason the concentration of well dispersed alkali metals is not high for high rank coals and there is little catalytic enhancement of the char gasification rate Unlike low rank coals demineralization of high rank coals does not significantly reduce the gasiflshy

cation rate In fact these rates have been reported to increase slightly This has been attributed to reshymoval of minerals which block pores or mild oxishydation of the coal during the demineralization proshycess which produces a less fluid coal and a more reactive char after pyrolysis

bull Fiero Salatino Universita di Napoli Italy The

Thiele model strictly applies to solids characterized by unimodal or nearly unimodal pore site distrishybution Did the materials used in your study corshyrespond to this hypothesis and if not would you comment on the effect that this simplification might have on the reliability of your diffusion-limited combustion model

Authors Reply The effect of the pore size disshytribution is folded into the value of the effective diffusivity Deff- Since it is difficult to evaluate Del we assumed the Knudsen diffusivity to hold and used a mean pore radius to account for the pore size distribution This approach is only approximate since the chars used in our study are probably not unishymodal However only a certain range of pore sizes will have an impact on diffusion limitations the mishycropore-mesopore range From our simulations for the chars where the pore size distribution is known a reasonable prediction is obtained for the reactivshyity in the pore diffusion regime Additional comshyparisons are underway in order to fully validate this approach

1192 COAL COMBUSTION

rated from the effect of coal rank Since deminershyalized low rank coals show a reactivity similar to that of medium rank coals (of oxygen content apshyproximately 13) the effect of oxygen content is minor compared to that of calcium In order to acshycount for this in the model the value of 0eo l is kept constant and set equal to 13 for coals of oxshyygen content higher than 13

Other second order effects include the combined effect of fluidity and pyrolysis heating rate on char reactivity This can be important for fluid coals and leads to separate lTerit vs Hebar correlations for different heating rates At this point few measureshyments have been done on low heating rate fluid chars and this will be the object of a future study

Individual correlation plots (shown in Figs 1-3) give an uncertainty in Terit within plusmn20deg C which is approximately equivalent to a factor of 2 in rate With the Heban Oma and Cad correlations used sishymultaneously Terit can be predicted within plusmn40deg C It should be noted that the experimental uncershytainty of the measurement of Terit is plusmn10deg C

To overcome potential shortcomings of using sevshyeral correlations simultaneously it would be reashysonable to use the Hebar correlation alone associshyated with one Tnit measurement as input However for the cases presented below the three correlashytions were used

I ntnnsic Rate Predictions

At Teritgt the rate of weight loss is OOOlsec By substituting the values of lTeit and rl into equashytion (1) the functions f g and h can be calculated It is implicitly assumed that 1 - X is constant and equal to 085 (which is the experimental value usushyally observed) at the point where Teit is measured So is also kept constant and was taken as a first apshyproximation to be 300 m2g The value of the acshytivation energy E was evaluated from TGA isoshythermal data and was found to vary between 28 and 34 kcalmol for the chars studied As a first approximation a value of 30 kcalmol was used in the model

For high rank coals (Oeoal lt 13) equation (1) becomes

-15000 rl = 822 exp -- shy

l

exp(14 Hehar) exp(0263 Oml) S (2)

For low rank coals (Oenal gt 13) the equivalent of equation (1) is

-15000 rl = 686 x 102 exp -- shy

T

exp(14 Hdmr) exp(154 Cad) S (3)

The reaction rates correspond to the case where the oxygen partial pressure is 172 kPa

Predictions of reaction rates using equations (2) and (3) have been compared with data obtained in our laboratory (Zap lignite chars) as well as literashyture data from Wells et al 15 (FMC Occidental and Toscoal chars) The parameters used are indicated in Table I The parameters for the literature chars were directly taken from reference 15 The reactivshyities of the Pittsburgh No 8 chars were not preshydicted (since good correlations of low heating rate fluid coals have not been developed yet) but were fit to the data Figures 4 and 5 show the good agreement obtained between the predictions and the low temperature TGA data for the chars studshyied

Predictions of reaction rates as a function of burnoff are presented in Fig 6 for pittsburgh No 8 (high rank coal) 600deg C char and Zap lignite (low rank coal) 900deg C char using the random pore model and volumetric model respectively The predictions were done using a single adjustable parameter (without using the correlations) in order to demshyonstrate the adequacy of the burnoff models choshysen The good fit (within 20) obtained in the full range of burnoff supports the use of these models

Model Development for Pore Diffusion Regime

Calculation of Reaction Rate

In the diffusion regime the observed reaction rate can be expressed as a function of the intrinsic reshyaction rate

(4)

where YJ = effectiveness factor According to Iehta and Aris J(j the effectiveness factor can be detershymined using the relations between YJ and YJqi where cjJ is the Thiele modulus II In the case of high Thiele modulus YJcjJ2 tends to lYJ ie YJ tends to lcjJ

The Thiele modulus is given by

d ( S k )0gt

(J () IcjJ-- (5)

6 De V MMo

with d = particle diameter = apparent particle (J

density De = effective diffusivity Mo = molecular weight of oxidant From the simple pore-structure model of Wheeler17 De can be calculated

05 _ 2 T E

De - 97 x 10 rpore (

-)

- (6) Mo To

where rpltre = mean pore radius Eo = porosity To = tortuosity

1193 PREDICTION OF COAL CHAR REACTIVITY

TABLE I Parameters Used in Reactivity Calculations

PIT 900deg C amp ZAP

Char 600degC 900deg C FMC OCC TOS

Parent Coal Pittsburgh Zap Pittsburgh Wyodak Utah

Rate (OCs) 05 05 103 10 10

pound0 01 03 013 055 011

10 2 2 2 2 2

dpm) 70 70 90 63 55

rpc (A) 6 6 170 100 170

So (m2g) 300 300 300 300 300

00 (gcm3) 13 13 152 086 143

01 () - 21 9 21 118

Hcha() - 074 2 21 318

M () - 1 0 11 0

Note 900deg C and 600deg C Pittsburgh No 8 chars and 900deg C Zap lignite char have been produced in our laboratory FMC Occidental and Toscoal are chars from Reference 15

estimated from measured values for similar chars or coals

The global reaction rate used in the model can then be written

05 6 ( Mgt So ki)r=-Dv--- vMC (7)

P deMo 0 p bull

The reaction rate depends on the intrinsic rate and char properties such as the tortuosity porosity and the mean pore radius The tortuosity parameter has been introduced to account for the connectivity of pores Its value has generally been taken between 15 and 10 usually equal to 2 The porosity and the mean pore radius can be measured However their determination may be difficult in the case of a non-uniform pore size distribution

Comparison of Predictions with Data

Using the physical and chemical parameters inshydicated in Table I and kinetic parameters derived earlier predictions in the pore diffusion regime for the Pittsburgh No 8 and Zap lignite chars proshy

duced at 30deg Cjmin are shown in Fig 4 For low rank coals we used a deactivation function to acshycount for the fact that catalysts have less impact on reactivity at high temperatures 1213 Following this deactivation function at high temperatures the reactivity of a low rank coal is made to be equal to that of a coal of 13 oxygen Although the number of data points in the diffusion regime is limited the transition between kinetic and diffusion control is fairly well represented by the theory for the two Pittsburgh No8 chars using a mean pore radius of 6 t which corresponds to the size of micropores Concerning the Zap char no internal diffusion limshyitations are present (since there is no change in ac~ tivation energy) but a leveling off of the rate is obshyserved This behavior is due to external mass transfer resistances in the TGA Under our conditions the maximum measurable rate in the TGA is approxishymately 1 gjg-min However under true pulvershyized coal combustion conditions external diffusion limitations are not encountered until temperatures typically higher than 1500 K Since no pore diffushysion limitations are observed for the Zap char in

1194 COAL COMBUSTION

4 4

3 -predictions OPIT 900 Cchar

CshymiddotE 2

ePIT 600 Cchar IoZAP 900 Cchar

dgt -9

~ s 0 0 15 ro (J)

gshy -1

~

OJ Q -2

-3

-4 4DE-4 6DE-4 8DE-4 1DE-3 12E-3 14E-3 16E-3

1T (11K)

FIG 4 Comparison of predicted reactivity vs TGA data for 6000 C and 9000 C Pittsburgh No 8 chars and 9000 C Zap lignite chars The chars were proshyduced at 300 Cmin up to the indicated temperashyture

the range of TGA measurements the choice of pashyrameters such as the mean pore radius and torshytuosity for that coal cannot be confirmed

The model predictions for the high heating rate chars used in Ref 15 are presented in Fig 5 (dashed lines) together with correlations based on drop tube reactor data15 (symbols) As for the chars produced in our laboratory a value of fpore = 6 Awas used In agreement with the Thiele model the apparent activation energy in the diffusion regime is well deshyscribed by half the intrinsic activation energy However predictions of the reaction rates give much lower values than the correlations based on the data For the low rank Occidental coal char this may be due to uncertainties on the deactivation process However the choice of parameters and particushylarly the value of the mean pore radius f pore also needs to be evaluated in more detail This paramshyeter represents a measure of the size of the pores which feed the reactant gas into the char Those pores can be significantly larger than the microshypores (of pores radius 6 A) especially if the char is fairly macroporous Pore size distribution measureshyments on chars similar to Occidental and FMC (produced at high heating rate) have been pershyformed by WhitelB and show a mean pore size of lOq Afor a Zap char and a bimodal distribution at 6 A and 170 A for a Pittsburgh No8 char Using ~alues of fpore of 100 Afor the Occidental and 170 A for the FMC and Toscoal predictions are shown in Fig 5 with solid lines and fit the data more accurately especially considering that the tempershy

- predictions -predictions oFMG

c- oOGG

3

2middotE TOS dgt

-9

~ 0 ~

ro -1~

o OJ -2Q

-3

-4 40E-4 6DE-4 80E-4 10E-3 12E-3 14E-3 16E-3

1T (11K)

FIG 5 Comparison of predicted reactivity (lines) vs correlations (symbols) derived from experimenshytal data from Wells et al 15 The low temperature data was measured using TGA High temperature data was obtained in a drop tube reactor The symshybols do not correspond to actual data points but serve to identifY the different correlations The dashed lines correspond to predictions using a mean pore radius of 6 A The solid lines correspond to predictions using a mean pore radius of 100 A for Occidental and 170 A for FMC and Toscoal

ature in the literature data may have been undershyestimated by as much as 2000 C 19 This underlines the importance of precisely characterizing the structure of the char in order to determine r pore

Also the fact that different values of f pore are needed in order to fit low heating rate and high heating rate char data implies a basic difference in structure between the two types of chars This is not surshyprising since low heating rate chars have been found to be effectively less porousfi probably because some reorganization of the structure is possible during the long period in which the char is fluid However pore size distribution measurements on low heating rate chars are needed in order to resolve that quesshytion Also direct diffusivity measurements20 would eliminate any amhiguity of Wheelers model in the estimation of the mean pore size and tortuosity pashyrameters This may give more reliable values of the effective diffusivity in order to calculate the reacshytivity in regime II

Conclusions

It was shown that correlations developed at low temperatures using the char hydrogen content coal

1195

i

PREDICTION OF COAL CHAR REACTIVITY

01-r-------------- a Pittsburgh 8

O~~~-r~~-+_+~~~

d

~ 1~ 005degb------------

~ Zap Lignite o ~ ~

O+-~-+~~~+_~_+~~~~

o 1Conversion

FIG 6 Comparison of TGA data (symbols) and model predictions (lines) for the reaction of a) 6000 C char from Pittsburgh No 8 at 5000 C using the ranshydom pore model with 1Jr = 50 b) 9000 C char from Zap lignite combusted at 3800 C using the volushymetric model The small fluctuations in the predicshytions correspond to slight variations in mcasured tempcrature

oxygen content and coal mineral content have given good predictions of intrinsic reactivity for different coals and chars Using these correlations combined with a random pore model for high rank coals and a volumetric model for low rank coals reactivity and reactivity variations with burnoff can be predicted

The reactivity in the pore diffusion regime was calculated using the intrinsic reactivity correlations and the Thiele model This approach necessitates the knowledge of char properties such as porosity tortuosity and mean pore radius Using a value of 6 A for the mean pore radius leads to fairly good predictions for 300 Cmin Pittsburgh No 8 chars while being inconclusive for Zap chars In the case of high he-ating rate chars pore size distribution measurements showed that values of approximately 100 Aare more appropriate and predictions using this value give a good fit of the literature data inshyvestigated Direct diffusivity measurements could also be performed and would eliminate any uncershytainty in the estimation of the mean pore size and tortuosity parameters In conclusion this analysis shows that it may be possible to extend low temshyperature reactivity models to high temperatures

Nomenclature

A = pre-exponential factor ms Cs = reactant concentration molm3

Cad = (daf) dispersed calcium in coal d = particle diameter m do = initial particle diameter m De = diffusivity m2 s E = activation energy kcajmol H char = (dar) hydrogen in char ki = reaction rate coefficient ms Mo = molecular mass of reactive gas gmol Mp = molecular mass of carbon gmol

degml (daf) oxygen in coal p = partial pressure of reactiv~ gas atm R = gas constant m3 atm(g mol K) rpore mean pore radius m

= reaction rate lsr S = internal surface area of remaining char per

mass of initial char m2 g So = internal surface area of initial char m2

g T = particle temperature K X = burnoff

Greek Symbols 1 = effectiveness factor Eo = initial porosity To = initial tortuosity (J = apparent particle density cP = Thiele modulus v = stochiometric factor 1jF = structural parameter

Acknowledgements

The authors are grateful for the support of the Morgantown Energy Technology Center of the United States Department of Energy under conshytract DE-AC21-86MC23075 for support of the study of char reactivity The authors wish to acknowledge the contributions of Hsisheng Teng and Marie DiTaranto of Advanced Fuel Research Inc to the experimental work and helpful discussions with Prof Eric Suuberg of Brown University

REFERENCES

1 FIEJIl M A Combust Flame 14 237 (1970) 2 SMITH I W Combust Flame 17 421 (1971) 3 MITCHELL R E AND McLEAN J W Nineshy

teenth Symposium (International) on Combusshytion p 1113 The Combustion Institute 1983

4 RADOVIC L R WALKER P L AND JENKIlS R G Fuel 62 849 (1983)

5 SUUBERC E M C~LO J M AND WOJTOWICZ M ACS Div of Fuel Chern Preprints 31 (3) 186 (1986)

1196 COAL COMBUSTION

6 Su J L AND PERLMUTIER D D AIChE J 14 SUUBERG E M WOJTOWIGZ M AND CALO J 31 6 973 (1985) M Twenty-Second Symposium (International)

7 BHATIA S K AND PERLMUTIER D D AlChE on Combustion p 79 The Combustion Insti shy

J 26 379 (1980) tute (1989) 15 WELLS W F KRAMER S K AND SMOOT L

D Twentieth Symposium (International) on 8 GAYALAS G R AIChE J 26 577 (1980) 9 ISHIDA 1 AJD WEN C Y Chem Eng Sci

Combustion p 1539 The Combustion Insti shy26 1031 (1971) tutc 198510 SIITH I W Nineteenth Symposium (Intershy

16 MEliTA B N AND ARIS R Chem Eng Scinational) on Combustion p 1045 Thc Comshy

26 1699 (1971)bustion Institute 1982

17 WHEELER A Adv Cata 3 249 (1951)11 THIELE E W Ind Eng Chem 31 916

18 WWTE W E M S Thesis Brigham Young(1939) University (1990)

12 SOLOMON P R SERIO M A AJD HENINGER 19 SMOOT L D Personal communication (1992)S G ACS Div of Fuel Chem Preprints 31 20 SMITH I W HARRIS D J VALIX M G AND (3) 200 (1986) TRIMM D L in Fundamental Issues in Conshy

13 SERIO M A SOLOMON P R CHARPENAY S trol of Carbon Gasification Reactivity J LashyAND SUUBERG E M to be submitted to Fuel haye and P Ehrburger (Eds) Kluwer Acashy(1992) demic Publishers p 49 1991

COMMENTS

N Y Nsakala ABB Combustion Engineering ItIC

USA The Arrhenius plot of your TGA char oxidashytion data (I believe) showed no dependence on charshytype for Illinois 6 Pittsburgh 8 zap lignite and Rosebud subbituminous coal chars Thc global acshytivation energy was -35 kcalfmole in all cases Would you expect this behaviour to hold throughshyout the whole coal rank spcctrum

Authors Reply It has been a consistent obsershyvation in our work that all the coal chars studied which have been produced under different pyrolshyysis conditions and from various coals (including chars from treated coals) showed an activation energy in the range 30 to 35 kcalmo as measured in a TGA apparatus This observation has also been made in several other studies and it is reasonable to cxpect a similar behavior for a wide range of coals

bull Robert Hurt Sandia National Laboratories USA

Your random pure model prcdictions were comshypared to data on chars that had bcen stabilized by heat treatment prior to reaction at a temperature greater than the reaction temperature In coal comshybustion chars are made in the flame-by in situ dcvolatilization-and for this situation we see very different trends Instead of large increases in area at low conversion followed by a maximum we (in our laboratory) generally observe monotonically deshycreasing surface areas Here the char formation process and thermal annealing (which tend to reshyduce surface area) occur Simultaneously with the

oxidation (which tends to increase the area inishytially) We find the net effect to be generally deshycreasing areas at the temperatures and times of inshyterest to pulverized coal combustion The existing pure models cannot predict this trend because they all consider rigid solid matrices in which pores grow by internal rcaction Have you observed such efshyfects and can you imagine how your model may acshycount for it

Authors Reply Proccsscs like thermal annealing occur at high temperature and cannot be acshycounted for in pore models in their present form Howcver the range of validity of porc models is usually at low temperature in the intrinsic regime where the reaction occurs all through the particle In our case at high temperature (ie in the pore diffusion regime) we have assumed that the parti shycles react following the shrinking core model U sshying this model the specific surface area (mg) is assumed to remain constant with burnofT Your exshyperimental observations of generally decreasing surshylace area with burnoff cannot be explained by our current model

bull Reginald E Mitchell Stanford University USA

The calculation of effectiveness factors from Thiele moduli requires knowledge of thc order of reaction with respect to the oxygen concentration Nhat values did you use for reaction orders and why were these values selected

Authors Reply A reaction order of 1 is assumed

1197 PREDICTION OF COAL CHAR REACTIVITY

in the model This choice is consistent with the apshyproach taken in the comprehensive code for enshytrained coal combustors and gasifiers for which this model is being developed (PCGC-2 from Brigham Young University) Several studies have shown that the true reaction order is probably fractional However using a reaction order of 1 typically leads to an error in rate of less than a factor of 1 5 if the true reaction order is for example 05 At this stage of our work this range of uncertainty is acceptable given the other uncertainties in the model and in the available data

bull Ian W Smith CSIRO Division of Coal amp Energy

Technology Australia The paper deals with low and high rank coals-the former coals are stated to be affected by catalysts but the high rank materials not Given the regime I nature of the measureshyments why do not the minerals present catalyse the reaction of the high rank coals

Authors Reply For low rank coals we found the best correlation of reactivity with the amount of dispersed calcium most of which is exchanged on carboxyl groups The concentration of carboxyl groups is not significant in high rank coals For this reason the concentration of well dispersed alkali metals is not high for high rank coals and there is little catalytic enhancement of the char gasification rate Unlike low rank coals demineralization of high rank coals does not significantly reduce the gasiflshy

cation rate In fact these rates have been reported to increase slightly This has been attributed to reshymoval of minerals which block pores or mild oxishydation of the coal during the demineralization proshycess which produces a less fluid coal and a more reactive char after pyrolysis

bull Fiero Salatino Universita di Napoli Italy The

Thiele model strictly applies to solids characterized by unimodal or nearly unimodal pore site distrishybution Did the materials used in your study corshyrespond to this hypothesis and if not would you comment on the effect that this simplification might have on the reliability of your diffusion-limited combustion model

Authors Reply The effect of the pore size disshytribution is folded into the value of the effective diffusivity Deff- Since it is difficult to evaluate Del we assumed the Knudsen diffusivity to hold and used a mean pore radius to account for the pore size distribution This approach is only approximate since the chars used in our study are probably not unishymodal However only a certain range of pore sizes will have an impact on diffusion limitations the mishycropore-mesopore range From our simulations for the chars where the pore size distribution is known a reasonable prediction is obtained for the reactivshyity in the pore diffusion regime Additional comshyparisons are underway in order to fully validate this approach

1193 PREDICTION OF COAL CHAR REACTIVITY

TABLE I Parameters Used in Reactivity Calculations

PIT 900deg C amp ZAP

Char 600degC 900deg C FMC OCC TOS

Parent Coal Pittsburgh Zap Pittsburgh Wyodak Utah

Rate (OCs) 05 05 103 10 10

pound0 01 03 013 055 011

10 2 2 2 2 2

dpm) 70 70 90 63 55

rpc (A) 6 6 170 100 170

So (m2g) 300 300 300 300 300

00 (gcm3) 13 13 152 086 143

01 () - 21 9 21 118

Hcha() - 074 2 21 318

M () - 1 0 11 0

Note 900deg C and 600deg C Pittsburgh No 8 chars and 900deg C Zap lignite char have been produced in our laboratory FMC Occidental and Toscoal are chars from Reference 15

estimated from measured values for similar chars or coals

The global reaction rate used in the model can then be written

05 6 ( Mgt So ki)r=-Dv--- vMC (7)

P deMo 0 p bull

The reaction rate depends on the intrinsic rate and char properties such as the tortuosity porosity and the mean pore radius The tortuosity parameter has been introduced to account for the connectivity of pores Its value has generally been taken between 15 and 10 usually equal to 2 The porosity and the mean pore radius can be measured However their determination may be difficult in the case of a non-uniform pore size distribution

Comparison of Predictions with Data

Using the physical and chemical parameters inshydicated in Table I and kinetic parameters derived earlier predictions in the pore diffusion regime for the Pittsburgh No 8 and Zap lignite chars proshy

duced at 30deg Cjmin are shown in Fig 4 For low rank coals we used a deactivation function to acshycount for the fact that catalysts have less impact on reactivity at high temperatures 1213 Following this deactivation function at high temperatures the reactivity of a low rank coal is made to be equal to that of a coal of 13 oxygen Although the number of data points in the diffusion regime is limited the transition between kinetic and diffusion control is fairly well represented by the theory for the two Pittsburgh No8 chars using a mean pore radius of 6 t which corresponds to the size of micropores Concerning the Zap char no internal diffusion limshyitations are present (since there is no change in ac~ tivation energy) but a leveling off of the rate is obshyserved This behavior is due to external mass transfer resistances in the TGA Under our conditions the maximum measurable rate in the TGA is approxishymately 1 gjg-min However under true pulvershyized coal combustion conditions external diffusion limitations are not encountered until temperatures typically higher than 1500 K Since no pore diffushysion limitations are observed for the Zap char in

1194 COAL COMBUSTION

4 4

3 -predictions OPIT 900 Cchar

CshymiddotE 2

ePIT 600 Cchar IoZAP 900 Cchar

dgt -9

~ s 0 0 15 ro (J)

gshy -1

~

OJ Q -2

-3

-4 4DE-4 6DE-4 8DE-4 1DE-3 12E-3 14E-3 16E-3

1T (11K)

FIG 4 Comparison of predicted reactivity vs TGA data for 6000 C and 9000 C Pittsburgh No 8 chars and 9000 C Zap lignite chars The chars were proshyduced at 300 Cmin up to the indicated temperashyture

the range of TGA measurements the choice of pashyrameters such as the mean pore radius and torshytuosity for that coal cannot be confirmed

The model predictions for the high heating rate chars used in Ref 15 are presented in Fig 5 (dashed lines) together with correlations based on drop tube reactor data15 (symbols) As for the chars produced in our laboratory a value of fpore = 6 Awas used In agreement with the Thiele model the apparent activation energy in the diffusion regime is well deshyscribed by half the intrinsic activation energy However predictions of the reaction rates give much lower values than the correlations based on the data For the low rank Occidental coal char this may be due to uncertainties on the deactivation process However the choice of parameters and particushylarly the value of the mean pore radius f pore also needs to be evaluated in more detail This paramshyeter represents a measure of the size of the pores which feed the reactant gas into the char Those pores can be significantly larger than the microshypores (of pores radius 6 A) especially if the char is fairly macroporous Pore size distribution measureshyments on chars similar to Occidental and FMC (produced at high heating rate) have been pershyformed by WhitelB and show a mean pore size of lOq Afor a Zap char and a bimodal distribution at 6 A and 170 A for a Pittsburgh No8 char Using ~alues of fpore of 100 Afor the Occidental and 170 A for the FMC and Toscoal predictions are shown in Fig 5 with solid lines and fit the data more accurately especially considering that the tempershy

- predictions -predictions oFMG

c- oOGG

3

2middotE TOS dgt

-9

~ 0 ~

ro -1~

o OJ -2Q

-3

-4 40E-4 6DE-4 80E-4 10E-3 12E-3 14E-3 16E-3

1T (11K)

FIG 5 Comparison of predicted reactivity (lines) vs correlations (symbols) derived from experimenshytal data from Wells et al 15 The low temperature data was measured using TGA High temperature data was obtained in a drop tube reactor The symshybols do not correspond to actual data points but serve to identifY the different correlations The dashed lines correspond to predictions using a mean pore radius of 6 A The solid lines correspond to predictions using a mean pore radius of 100 A for Occidental and 170 A for FMC and Toscoal

ature in the literature data may have been undershyestimated by as much as 2000 C 19 This underlines the importance of precisely characterizing the structure of the char in order to determine r pore

Also the fact that different values of f pore are needed in order to fit low heating rate and high heating rate char data implies a basic difference in structure between the two types of chars This is not surshyprising since low heating rate chars have been found to be effectively less porousfi probably because some reorganization of the structure is possible during the long period in which the char is fluid However pore size distribution measurements on low heating rate chars are needed in order to resolve that quesshytion Also direct diffusivity measurements20 would eliminate any amhiguity of Wheelers model in the estimation of the mean pore size and tortuosity pashyrameters This may give more reliable values of the effective diffusivity in order to calculate the reacshytivity in regime II

Conclusions

It was shown that correlations developed at low temperatures using the char hydrogen content coal

1195

i

PREDICTION OF COAL CHAR REACTIVITY

01-r-------------- a Pittsburgh 8

O~~~-r~~-+_+~~~

d

~ 1~ 005degb------------

~ Zap Lignite o ~ ~

O+-~-+~~~+_~_+~~~~

o 1Conversion

FIG 6 Comparison of TGA data (symbols) and model predictions (lines) for the reaction of a) 6000 C char from Pittsburgh No 8 at 5000 C using the ranshydom pore model with 1Jr = 50 b) 9000 C char from Zap lignite combusted at 3800 C using the volushymetric model The small fluctuations in the predicshytions correspond to slight variations in mcasured tempcrature

oxygen content and coal mineral content have given good predictions of intrinsic reactivity for different coals and chars Using these correlations combined with a random pore model for high rank coals and a volumetric model for low rank coals reactivity and reactivity variations with burnoff can be predicted

The reactivity in the pore diffusion regime was calculated using the intrinsic reactivity correlations and the Thiele model This approach necessitates the knowledge of char properties such as porosity tortuosity and mean pore radius Using a value of 6 A for the mean pore radius leads to fairly good predictions for 300 Cmin Pittsburgh No 8 chars while being inconclusive for Zap chars In the case of high he-ating rate chars pore size distribution measurements showed that values of approximately 100 Aare more appropriate and predictions using this value give a good fit of the literature data inshyvestigated Direct diffusivity measurements could also be performed and would eliminate any uncershytainty in the estimation of the mean pore size and tortuosity parameters In conclusion this analysis shows that it may be possible to extend low temshyperature reactivity models to high temperatures

Nomenclature

A = pre-exponential factor ms Cs = reactant concentration molm3

Cad = (daf) dispersed calcium in coal d = particle diameter m do = initial particle diameter m De = diffusivity m2 s E = activation energy kcajmol H char = (dar) hydrogen in char ki = reaction rate coefficient ms Mo = molecular mass of reactive gas gmol Mp = molecular mass of carbon gmol

degml (daf) oxygen in coal p = partial pressure of reactiv~ gas atm R = gas constant m3 atm(g mol K) rpore mean pore radius m

= reaction rate lsr S = internal surface area of remaining char per

mass of initial char m2 g So = internal surface area of initial char m2

g T = particle temperature K X = burnoff

Greek Symbols 1 = effectiveness factor Eo = initial porosity To = initial tortuosity (J = apparent particle density cP = Thiele modulus v = stochiometric factor 1jF = structural parameter

Acknowledgements

The authors are grateful for the support of the Morgantown Energy Technology Center of the United States Department of Energy under conshytract DE-AC21-86MC23075 for support of the study of char reactivity The authors wish to acknowledge the contributions of Hsisheng Teng and Marie DiTaranto of Advanced Fuel Research Inc to the experimental work and helpful discussions with Prof Eric Suuberg of Brown University

REFERENCES

1 FIEJIl M A Combust Flame 14 237 (1970) 2 SMITH I W Combust Flame 17 421 (1971) 3 MITCHELL R E AND McLEAN J W Nineshy

teenth Symposium (International) on Combusshytion p 1113 The Combustion Institute 1983

4 RADOVIC L R WALKER P L AND JENKIlS R G Fuel 62 849 (1983)

5 SUUBERC E M C~LO J M AND WOJTOWICZ M ACS Div of Fuel Chern Preprints 31 (3) 186 (1986)

1196 COAL COMBUSTION

6 Su J L AND PERLMUTIER D D AIChE J 14 SUUBERG E M WOJTOWIGZ M AND CALO J 31 6 973 (1985) M Twenty-Second Symposium (International)

7 BHATIA S K AND PERLMUTIER D D AlChE on Combustion p 79 The Combustion Insti shy

J 26 379 (1980) tute (1989) 15 WELLS W F KRAMER S K AND SMOOT L

D Twentieth Symposium (International) on 8 GAYALAS G R AIChE J 26 577 (1980) 9 ISHIDA 1 AJD WEN C Y Chem Eng Sci

Combustion p 1539 The Combustion Insti shy26 1031 (1971) tutc 198510 SIITH I W Nineteenth Symposium (Intershy

16 MEliTA B N AND ARIS R Chem Eng Scinational) on Combustion p 1045 Thc Comshy

26 1699 (1971)bustion Institute 1982

17 WHEELER A Adv Cata 3 249 (1951)11 THIELE E W Ind Eng Chem 31 916

18 WWTE W E M S Thesis Brigham Young(1939) University (1990)

12 SOLOMON P R SERIO M A AJD HENINGER 19 SMOOT L D Personal communication (1992)S G ACS Div of Fuel Chem Preprints 31 20 SMITH I W HARRIS D J VALIX M G AND (3) 200 (1986) TRIMM D L in Fundamental Issues in Conshy

13 SERIO M A SOLOMON P R CHARPENAY S trol of Carbon Gasification Reactivity J LashyAND SUUBERG E M to be submitted to Fuel haye and P Ehrburger (Eds) Kluwer Acashy(1992) demic Publishers p 49 1991

COMMENTS

N Y Nsakala ABB Combustion Engineering ItIC

USA The Arrhenius plot of your TGA char oxidashytion data (I believe) showed no dependence on charshytype for Illinois 6 Pittsburgh 8 zap lignite and Rosebud subbituminous coal chars Thc global acshytivation energy was -35 kcalfmole in all cases Would you expect this behaviour to hold throughshyout the whole coal rank spcctrum

Authors Reply It has been a consistent obsershyvation in our work that all the coal chars studied which have been produced under different pyrolshyysis conditions and from various coals (including chars from treated coals) showed an activation energy in the range 30 to 35 kcalmo as measured in a TGA apparatus This observation has also been made in several other studies and it is reasonable to cxpect a similar behavior for a wide range of coals

bull Robert Hurt Sandia National Laboratories USA

Your random pure model prcdictions were comshypared to data on chars that had bcen stabilized by heat treatment prior to reaction at a temperature greater than the reaction temperature In coal comshybustion chars are made in the flame-by in situ dcvolatilization-and for this situation we see very different trends Instead of large increases in area at low conversion followed by a maximum we (in our laboratory) generally observe monotonically deshycreasing surface areas Here the char formation process and thermal annealing (which tend to reshyduce surface area) occur Simultaneously with the

oxidation (which tends to increase the area inishytially) We find the net effect to be generally deshycreasing areas at the temperatures and times of inshyterest to pulverized coal combustion The existing pure models cannot predict this trend because they all consider rigid solid matrices in which pores grow by internal rcaction Have you observed such efshyfects and can you imagine how your model may acshycount for it

Authors Reply Proccsscs like thermal annealing occur at high temperature and cannot be acshycounted for in pore models in their present form Howcver the range of validity of porc models is usually at low temperature in the intrinsic regime where the reaction occurs all through the particle In our case at high temperature (ie in the pore diffusion regime) we have assumed that the parti shycles react following the shrinking core model U sshying this model the specific surface area (mg) is assumed to remain constant with burnofT Your exshyperimental observations of generally decreasing surshylace area with burnoff cannot be explained by our current model

bull Reginald E Mitchell Stanford University USA

The calculation of effectiveness factors from Thiele moduli requires knowledge of thc order of reaction with respect to the oxygen concentration Nhat values did you use for reaction orders and why were these values selected

Authors Reply A reaction order of 1 is assumed

1197 PREDICTION OF COAL CHAR REACTIVITY

in the model This choice is consistent with the apshyproach taken in the comprehensive code for enshytrained coal combustors and gasifiers for which this model is being developed (PCGC-2 from Brigham Young University) Several studies have shown that the true reaction order is probably fractional However using a reaction order of 1 typically leads to an error in rate of less than a factor of 1 5 if the true reaction order is for example 05 At this stage of our work this range of uncertainty is acceptable given the other uncertainties in the model and in the available data

bull Ian W Smith CSIRO Division of Coal amp Energy

Technology Australia The paper deals with low and high rank coals-the former coals are stated to be affected by catalysts but the high rank materials not Given the regime I nature of the measureshyments why do not the minerals present catalyse the reaction of the high rank coals

Authors Reply For low rank coals we found the best correlation of reactivity with the amount of dispersed calcium most of which is exchanged on carboxyl groups The concentration of carboxyl groups is not significant in high rank coals For this reason the concentration of well dispersed alkali metals is not high for high rank coals and there is little catalytic enhancement of the char gasification rate Unlike low rank coals demineralization of high rank coals does not significantly reduce the gasiflshy

cation rate In fact these rates have been reported to increase slightly This has been attributed to reshymoval of minerals which block pores or mild oxishydation of the coal during the demineralization proshycess which produces a less fluid coal and a more reactive char after pyrolysis

bull Fiero Salatino Universita di Napoli Italy The

Thiele model strictly applies to solids characterized by unimodal or nearly unimodal pore site distrishybution Did the materials used in your study corshyrespond to this hypothesis and if not would you comment on the effect that this simplification might have on the reliability of your diffusion-limited combustion model

Authors Reply The effect of the pore size disshytribution is folded into the value of the effective diffusivity Deff- Since it is difficult to evaluate Del we assumed the Knudsen diffusivity to hold and used a mean pore radius to account for the pore size distribution This approach is only approximate since the chars used in our study are probably not unishymodal However only a certain range of pore sizes will have an impact on diffusion limitations the mishycropore-mesopore range From our simulations for the chars where the pore size distribution is known a reasonable prediction is obtained for the reactivshyity in the pore diffusion regime Additional comshyparisons are underway in order to fully validate this approach

1194 COAL COMBUSTION

4 4

3 -predictions OPIT 900 Cchar

CshymiddotE 2

ePIT 600 Cchar IoZAP 900 Cchar

dgt -9

~ s 0 0 15 ro (J)

gshy -1

~

OJ Q -2

-3

-4 4DE-4 6DE-4 8DE-4 1DE-3 12E-3 14E-3 16E-3

1T (11K)

FIG 4 Comparison of predicted reactivity vs TGA data for 6000 C and 9000 C Pittsburgh No 8 chars and 9000 C Zap lignite chars The chars were proshyduced at 300 Cmin up to the indicated temperashyture

the range of TGA measurements the choice of pashyrameters such as the mean pore radius and torshytuosity for that coal cannot be confirmed

The model predictions for the high heating rate chars used in Ref 15 are presented in Fig 5 (dashed lines) together with correlations based on drop tube reactor data15 (symbols) As for the chars produced in our laboratory a value of fpore = 6 Awas used In agreement with the Thiele model the apparent activation energy in the diffusion regime is well deshyscribed by half the intrinsic activation energy However predictions of the reaction rates give much lower values than the correlations based on the data For the low rank Occidental coal char this may be due to uncertainties on the deactivation process However the choice of parameters and particushylarly the value of the mean pore radius f pore also needs to be evaluated in more detail This paramshyeter represents a measure of the size of the pores which feed the reactant gas into the char Those pores can be significantly larger than the microshypores (of pores radius 6 A) especially if the char is fairly macroporous Pore size distribution measureshyments on chars similar to Occidental and FMC (produced at high heating rate) have been pershyformed by WhitelB and show a mean pore size of lOq Afor a Zap char and a bimodal distribution at 6 A and 170 A for a Pittsburgh No8 char Using ~alues of fpore of 100 Afor the Occidental and 170 A for the FMC and Toscoal predictions are shown in Fig 5 with solid lines and fit the data more accurately especially considering that the tempershy

- predictions -predictions oFMG

c- oOGG

3

2middotE TOS dgt

-9

~ 0 ~

ro -1~

o OJ -2Q

-3

-4 40E-4 6DE-4 80E-4 10E-3 12E-3 14E-3 16E-3

1T (11K)

FIG 5 Comparison of predicted reactivity (lines) vs correlations (symbols) derived from experimenshytal data from Wells et al 15 The low temperature data was measured using TGA High temperature data was obtained in a drop tube reactor The symshybols do not correspond to actual data points but serve to identifY the different correlations The dashed lines correspond to predictions using a mean pore radius of 6 A The solid lines correspond to predictions using a mean pore radius of 100 A for Occidental and 170 A for FMC and Toscoal

ature in the literature data may have been undershyestimated by as much as 2000 C 19 This underlines the importance of precisely characterizing the structure of the char in order to determine r pore

Also the fact that different values of f pore are needed in order to fit low heating rate and high heating rate char data implies a basic difference in structure between the two types of chars This is not surshyprising since low heating rate chars have been found to be effectively less porousfi probably because some reorganization of the structure is possible during the long period in which the char is fluid However pore size distribution measurements on low heating rate chars are needed in order to resolve that quesshytion Also direct diffusivity measurements20 would eliminate any amhiguity of Wheelers model in the estimation of the mean pore size and tortuosity pashyrameters This may give more reliable values of the effective diffusivity in order to calculate the reacshytivity in regime II

Conclusions

It was shown that correlations developed at low temperatures using the char hydrogen content coal

1195

i

PREDICTION OF COAL CHAR REACTIVITY

01-r-------------- a Pittsburgh 8

O~~~-r~~-+_+~~~

d

~ 1~ 005degb------------

~ Zap Lignite o ~ ~

O+-~-+~~~+_~_+~~~~

o 1Conversion

FIG 6 Comparison of TGA data (symbols) and model predictions (lines) for the reaction of a) 6000 C char from Pittsburgh No 8 at 5000 C using the ranshydom pore model with 1Jr = 50 b) 9000 C char from Zap lignite combusted at 3800 C using the volushymetric model The small fluctuations in the predicshytions correspond to slight variations in mcasured tempcrature

oxygen content and coal mineral content have given good predictions of intrinsic reactivity for different coals and chars Using these correlations combined with a random pore model for high rank coals and a volumetric model for low rank coals reactivity and reactivity variations with burnoff can be predicted

The reactivity in the pore diffusion regime was calculated using the intrinsic reactivity correlations and the Thiele model This approach necessitates the knowledge of char properties such as porosity tortuosity and mean pore radius Using a value of 6 A for the mean pore radius leads to fairly good predictions for 300 Cmin Pittsburgh No 8 chars while being inconclusive for Zap chars In the case of high he-ating rate chars pore size distribution measurements showed that values of approximately 100 Aare more appropriate and predictions using this value give a good fit of the literature data inshyvestigated Direct diffusivity measurements could also be performed and would eliminate any uncershytainty in the estimation of the mean pore size and tortuosity parameters In conclusion this analysis shows that it may be possible to extend low temshyperature reactivity models to high temperatures

Nomenclature

A = pre-exponential factor ms Cs = reactant concentration molm3

Cad = (daf) dispersed calcium in coal d = particle diameter m do = initial particle diameter m De = diffusivity m2 s E = activation energy kcajmol H char = (dar) hydrogen in char ki = reaction rate coefficient ms Mo = molecular mass of reactive gas gmol Mp = molecular mass of carbon gmol

degml (daf) oxygen in coal p = partial pressure of reactiv~ gas atm R = gas constant m3 atm(g mol K) rpore mean pore radius m

= reaction rate lsr S = internal surface area of remaining char per

mass of initial char m2 g So = internal surface area of initial char m2

g T = particle temperature K X = burnoff

Greek Symbols 1 = effectiveness factor Eo = initial porosity To = initial tortuosity (J = apparent particle density cP = Thiele modulus v = stochiometric factor 1jF = structural parameter

Acknowledgements

The authors are grateful for the support of the Morgantown Energy Technology Center of the United States Department of Energy under conshytract DE-AC21-86MC23075 for support of the study of char reactivity The authors wish to acknowledge the contributions of Hsisheng Teng and Marie DiTaranto of Advanced Fuel Research Inc to the experimental work and helpful discussions with Prof Eric Suuberg of Brown University

REFERENCES

1 FIEJIl M A Combust Flame 14 237 (1970) 2 SMITH I W Combust Flame 17 421 (1971) 3 MITCHELL R E AND McLEAN J W Nineshy

teenth Symposium (International) on Combusshytion p 1113 The Combustion Institute 1983

4 RADOVIC L R WALKER P L AND JENKIlS R G Fuel 62 849 (1983)

5 SUUBERC E M C~LO J M AND WOJTOWICZ M ACS Div of Fuel Chern Preprints 31 (3) 186 (1986)

1196 COAL COMBUSTION

6 Su J L AND PERLMUTIER D D AIChE J 14 SUUBERG E M WOJTOWIGZ M AND CALO J 31 6 973 (1985) M Twenty-Second Symposium (International)

7 BHATIA S K AND PERLMUTIER D D AlChE on Combustion p 79 The Combustion Insti shy

J 26 379 (1980) tute (1989) 15 WELLS W F KRAMER S K AND SMOOT L

D Twentieth Symposium (International) on 8 GAYALAS G R AIChE J 26 577 (1980) 9 ISHIDA 1 AJD WEN C Y Chem Eng Sci

Combustion p 1539 The Combustion Insti shy26 1031 (1971) tutc 198510 SIITH I W Nineteenth Symposium (Intershy

16 MEliTA B N AND ARIS R Chem Eng Scinational) on Combustion p 1045 Thc Comshy

26 1699 (1971)bustion Institute 1982

17 WHEELER A Adv Cata 3 249 (1951)11 THIELE E W Ind Eng Chem 31 916

18 WWTE W E M S Thesis Brigham Young(1939) University (1990)

12 SOLOMON P R SERIO M A AJD HENINGER 19 SMOOT L D Personal communication (1992)S G ACS Div of Fuel Chem Preprints 31 20 SMITH I W HARRIS D J VALIX M G AND (3) 200 (1986) TRIMM D L in Fundamental Issues in Conshy

13 SERIO M A SOLOMON P R CHARPENAY S trol of Carbon Gasification Reactivity J LashyAND SUUBERG E M to be submitted to Fuel haye and P Ehrburger (Eds) Kluwer Acashy(1992) demic Publishers p 49 1991

COMMENTS

N Y Nsakala ABB Combustion Engineering ItIC

USA The Arrhenius plot of your TGA char oxidashytion data (I believe) showed no dependence on charshytype for Illinois 6 Pittsburgh 8 zap lignite and Rosebud subbituminous coal chars Thc global acshytivation energy was -35 kcalfmole in all cases Would you expect this behaviour to hold throughshyout the whole coal rank spcctrum

Authors Reply It has been a consistent obsershyvation in our work that all the coal chars studied which have been produced under different pyrolshyysis conditions and from various coals (including chars from treated coals) showed an activation energy in the range 30 to 35 kcalmo as measured in a TGA apparatus This observation has also been made in several other studies and it is reasonable to cxpect a similar behavior for a wide range of coals

bull Robert Hurt Sandia National Laboratories USA

Your random pure model prcdictions were comshypared to data on chars that had bcen stabilized by heat treatment prior to reaction at a temperature greater than the reaction temperature In coal comshybustion chars are made in the flame-by in situ dcvolatilization-and for this situation we see very different trends Instead of large increases in area at low conversion followed by a maximum we (in our laboratory) generally observe monotonically deshycreasing surface areas Here the char formation process and thermal annealing (which tend to reshyduce surface area) occur Simultaneously with the

oxidation (which tends to increase the area inishytially) We find the net effect to be generally deshycreasing areas at the temperatures and times of inshyterest to pulverized coal combustion The existing pure models cannot predict this trend because they all consider rigid solid matrices in which pores grow by internal rcaction Have you observed such efshyfects and can you imagine how your model may acshycount for it

Authors Reply Proccsscs like thermal annealing occur at high temperature and cannot be acshycounted for in pore models in their present form Howcver the range of validity of porc models is usually at low temperature in the intrinsic regime where the reaction occurs all through the particle In our case at high temperature (ie in the pore diffusion regime) we have assumed that the parti shycles react following the shrinking core model U sshying this model the specific surface area (mg) is assumed to remain constant with burnofT Your exshyperimental observations of generally decreasing surshylace area with burnoff cannot be explained by our current model

bull Reginald E Mitchell Stanford University USA

The calculation of effectiveness factors from Thiele moduli requires knowledge of thc order of reaction with respect to the oxygen concentration Nhat values did you use for reaction orders and why were these values selected

Authors Reply A reaction order of 1 is assumed

1197 PREDICTION OF COAL CHAR REACTIVITY

in the model This choice is consistent with the apshyproach taken in the comprehensive code for enshytrained coal combustors and gasifiers for which this model is being developed (PCGC-2 from Brigham Young University) Several studies have shown that the true reaction order is probably fractional However using a reaction order of 1 typically leads to an error in rate of less than a factor of 1 5 if the true reaction order is for example 05 At this stage of our work this range of uncertainty is acceptable given the other uncertainties in the model and in the available data

bull Ian W Smith CSIRO Division of Coal amp Energy

Technology Australia The paper deals with low and high rank coals-the former coals are stated to be affected by catalysts but the high rank materials not Given the regime I nature of the measureshyments why do not the minerals present catalyse the reaction of the high rank coals

Authors Reply For low rank coals we found the best correlation of reactivity with the amount of dispersed calcium most of which is exchanged on carboxyl groups The concentration of carboxyl groups is not significant in high rank coals For this reason the concentration of well dispersed alkali metals is not high for high rank coals and there is little catalytic enhancement of the char gasification rate Unlike low rank coals demineralization of high rank coals does not significantly reduce the gasiflshy

cation rate In fact these rates have been reported to increase slightly This has been attributed to reshymoval of minerals which block pores or mild oxishydation of the coal during the demineralization proshycess which produces a less fluid coal and a more reactive char after pyrolysis

bull Fiero Salatino Universita di Napoli Italy The

Thiele model strictly applies to solids characterized by unimodal or nearly unimodal pore site distrishybution Did the materials used in your study corshyrespond to this hypothesis and if not would you comment on the effect that this simplification might have on the reliability of your diffusion-limited combustion model

Authors Reply The effect of the pore size disshytribution is folded into the value of the effective diffusivity Deff- Since it is difficult to evaluate Del we assumed the Knudsen diffusivity to hold and used a mean pore radius to account for the pore size distribution This approach is only approximate since the chars used in our study are probably not unishymodal However only a certain range of pore sizes will have an impact on diffusion limitations the mishycropore-mesopore range From our simulations for the chars where the pore size distribution is known a reasonable prediction is obtained for the reactivshyity in the pore diffusion regime Additional comshyparisons are underway in order to fully validate this approach

1195

i

PREDICTION OF COAL CHAR REACTIVITY

01-r-------------- a Pittsburgh 8

O~~~-r~~-+_+~~~

d

~ 1~ 005degb------------

~ Zap Lignite o ~ ~

O+-~-+~~~+_~_+~~~~

o 1Conversion

FIG 6 Comparison of TGA data (symbols) and model predictions (lines) for the reaction of a) 6000 C char from Pittsburgh No 8 at 5000 C using the ranshydom pore model with 1Jr = 50 b) 9000 C char from Zap lignite combusted at 3800 C using the volushymetric model The small fluctuations in the predicshytions correspond to slight variations in mcasured tempcrature

oxygen content and coal mineral content have given good predictions of intrinsic reactivity for different coals and chars Using these correlations combined with a random pore model for high rank coals and a volumetric model for low rank coals reactivity and reactivity variations with burnoff can be predicted

The reactivity in the pore diffusion regime was calculated using the intrinsic reactivity correlations and the Thiele model This approach necessitates the knowledge of char properties such as porosity tortuosity and mean pore radius Using a value of 6 A for the mean pore radius leads to fairly good predictions for 300 Cmin Pittsburgh No 8 chars while being inconclusive for Zap chars In the case of high he-ating rate chars pore size distribution measurements showed that values of approximately 100 Aare more appropriate and predictions using this value give a good fit of the literature data inshyvestigated Direct diffusivity measurements could also be performed and would eliminate any uncershytainty in the estimation of the mean pore size and tortuosity parameters In conclusion this analysis shows that it may be possible to extend low temshyperature reactivity models to high temperatures

Nomenclature

A = pre-exponential factor ms Cs = reactant concentration molm3

Cad = (daf) dispersed calcium in coal d = particle diameter m do = initial particle diameter m De = diffusivity m2 s E = activation energy kcajmol H char = (dar) hydrogen in char ki = reaction rate coefficient ms Mo = molecular mass of reactive gas gmol Mp = molecular mass of carbon gmol

degml (daf) oxygen in coal p = partial pressure of reactiv~ gas atm R = gas constant m3 atm(g mol K) rpore mean pore radius m

= reaction rate lsr S = internal surface area of remaining char per

mass of initial char m2 g So = internal surface area of initial char m2

g T = particle temperature K X = burnoff

Greek Symbols 1 = effectiveness factor Eo = initial porosity To = initial tortuosity (J = apparent particle density cP = Thiele modulus v = stochiometric factor 1jF = structural parameter

Acknowledgements

The authors are grateful for the support of the Morgantown Energy Technology Center of the United States Department of Energy under conshytract DE-AC21-86MC23075 for support of the study of char reactivity The authors wish to acknowledge the contributions of Hsisheng Teng and Marie DiTaranto of Advanced Fuel Research Inc to the experimental work and helpful discussions with Prof Eric Suuberg of Brown University

REFERENCES

1 FIEJIl M A Combust Flame 14 237 (1970) 2 SMITH I W Combust Flame 17 421 (1971) 3 MITCHELL R E AND McLEAN J W Nineshy

teenth Symposium (International) on Combusshytion p 1113 The Combustion Institute 1983

4 RADOVIC L R WALKER P L AND JENKIlS R G Fuel 62 849 (1983)

5 SUUBERC E M C~LO J M AND WOJTOWICZ M ACS Div of Fuel Chern Preprints 31 (3) 186 (1986)

1196 COAL COMBUSTION

6 Su J L AND PERLMUTIER D D AIChE J 14 SUUBERG E M WOJTOWIGZ M AND CALO J 31 6 973 (1985) M Twenty-Second Symposium (International)

7 BHATIA S K AND PERLMUTIER D D AlChE on Combustion p 79 The Combustion Insti shy

J 26 379 (1980) tute (1989) 15 WELLS W F KRAMER S K AND SMOOT L

D Twentieth Symposium (International) on 8 GAYALAS G R AIChE J 26 577 (1980) 9 ISHIDA 1 AJD WEN C Y Chem Eng Sci

Combustion p 1539 The Combustion Insti shy26 1031 (1971) tutc 198510 SIITH I W Nineteenth Symposium (Intershy

16 MEliTA B N AND ARIS R Chem Eng Scinational) on Combustion p 1045 Thc Comshy

26 1699 (1971)bustion Institute 1982

17 WHEELER A Adv Cata 3 249 (1951)11 THIELE E W Ind Eng Chem 31 916

18 WWTE W E M S Thesis Brigham Young(1939) University (1990)

12 SOLOMON P R SERIO M A AJD HENINGER 19 SMOOT L D Personal communication (1992)S G ACS Div of Fuel Chem Preprints 31 20 SMITH I W HARRIS D J VALIX M G AND (3) 200 (1986) TRIMM D L in Fundamental Issues in Conshy

13 SERIO M A SOLOMON P R CHARPENAY S trol of Carbon Gasification Reactivity J LashyAND SUUBERG E M to be submitted to Fuel haye and P Ehrburger (Eds) Kluwer Acashy(1992) demic Publishers p 49 1991

COMMENTS

N Y Nsakala ABB Combustion Engineering ItIC

USA The Arrhenius plot of your TGA char oxidashytion data (I believe) showed no dependence on charshytype for Illinois 6 Pittsburgh 8 zap lignite and Rosebud subbituminous coal chars Thc global acshytivation energy was -35 kcalfmole in all cases Would you expect this behaviour to hold throughshyout the whole coal rank spcctrum

Authors Reply It has been a consistent obsershyvation in our work that all the coal chars studied which have been produced under different pyrolshyysis conditions and from various coals (including chars from treated coals) showed an activation energy in the range 30 to 35 kcalmo as measured in a TGA apparatus This observation has also been made in several other studies and it is reasonable to cxpect a similar behavior for a wide range of coals

bull Robert Hurt Sandia National Laboratories USA

Your random pure model prcdictions were comshypared to data on chars that had bcen stabilized by heat treatment prior to reaction at a temperature greater than the reaction temperature In coal comshybustion chars are made in the flame-by in situ dcvolatilization-and for this situation we see very different trends Instead of large increases in area at low conversion followed by a maximum we (in our laboratory) generally observe monotonically deshycreasing surface areas Here the char formation process and thermal annealing (which tend to reshyduce surface area) occur Simultaneously with the

oxidation (which tends to increase the area inishytially) We find the net effect to be generally deshycreasing areas at the temperatures and times of inshyterest to pulverized coal combustion The existing pure models cannot predict this trend because they all consider rigid solid matrices in which pores grow by internal rcaction Have you observed such efshyfects and can you imagine how your model may acshycount for it

Authors Reply Proccsscs like thermal annealing occur at high temperature and cannot be acshycounted for in pore models in their present form Howcver the range of validity of porc models is usually at low temperature in the intrinsic regime where the reaction occurs all through the particle In our case at high temperature (ie in the pore diffusion regime) we have assumed that the parti shycles react following the shrinking core model U sshying this model the specific surface area (mg) is assumed to remain constant with burnofT Your exshyperimental observations of generally decreasing surshylace area with burnoff cannot be explained by our current model

bull Reginald E Mitchell Stanford University USA

The calculation of effectiveness factors from Thiele moduli requires knowledge of thc order of reaction with respect to the oxygen concentration Nhat values did you use for reaction orders and why were these values selected

Authors Reply A reaction order of 1 is assumed

1197 PREDICTION OF COAL CHAR REACTIVITY

in the model This choice is consistent with the apshyproach taken in the comprehensive code for enshytrained coal combustors and gasifiers for which this model is being developed (PCGC-2 from Brigham Young University) Several studies have shown that the true reaction order is probably fractional However using a reaction order of 1 typically leads to an error in rate of less than a factor of 1 5 if the true reaction order is for example 05 At this stage of our work this range of uncertainty is acceptable given the other uncertainties in the model and in the available data

bull Ian W Smith CSIRO Division of Coal amp Energy

Technology Australia The paper deals with low and high rank coals-the former coals are stated to be affected by catalysts but the high rank materials not Given the regime I nature of the measureshyments why do not the minerals present catalyse the reaction of the high rank coals

Authors Reply For low rank coals we found the best correlation of reactivity with the amount of dispersed calcium most of which is exchanged on carboxyl groups The concentration of carboxyl groups is not significant in high rank coals For this reason the concentration of well dispersed alkali metals is not high for high rank coals and there is little catalytic enhancement of the char gasification rate Unlike low rank coals demineralization of high rank coals does not significantly reduce the gasiflshy

cation rate In fact these rates have been reported to increase slightly This has been attributed to reshymoval of minerals which block pores or mild oxishydation of the coal during the demineralization proshycess which produces a less fluid coal and a more reactive char after pyrolysis

bull Fiero Salatino Universita di Napoli Italy The

Thiele model strictly applies to solids characterized by unimodal or nearly unimodal pore site distrishybution Did the materials used in your study corshyrespond to this hypothesis and if not would you comment on the effect that this simplification might have on the reliability of your diffusion-limited combustion model

Authors Reply The effect of the pore size disshytribution is folded into the value of the effective diffusivity Deff- Since it is difficult to evaluate Del we assumed the Knudsen diffusivity to hold and used a mean pore radius to account for the pore size distribution This approach is only approximate since the chars used in our study are probably not unishymodal However only a certain range of pore sizes will have an impact on diffusion limitations the mishycropore-mesopore range From our simulations for the chars where the pore size distribution is known a reasonable prediction is obtained for the reactivshyity in the pore diffusion regime Additional comshyparisons are underway in order to fully validate this approach

1196 COAL COMBUSTION

6 Su J L AND PERLMUTIER D D AIChE J 14 SUUBERG E M WOJTOWIGZ M AND CALO J 31 6 973 (1985) M Twenty-Second Symposium (International)

7 BHATIA S K AND PERLMUTIER D D AlChE on Combustion p 79 The Combustion Insti shy

J 26 379 (1980) tute (1989) 15 WELLS W F KRAMER S K AND SMOOT L

D Twentieth Symposium (International) on 8 GAYALAS G R AIChE J 26 577 (1980) 9 ISHIDA 1 AJD WEN C Y Chem Eng Sci

Combustion p 1539 The Combustion Insti shy26 1031 (1971) tutc 198510 SIITH I W Nineteenth Symposium (Intershy

16 MEliTA B N AND ARIS R Chem Eng Scinational) on Combustion p 1045 Thc Comshy

26 1699 (1971)bustion Institute 1982

17 WHEELER A Adv Cata 3 249 (1951)11 THIELE E W Ind Eng Chem 31 916

18 WWTE W E M S Thesis Brigham Young(1939) University (1990)

12 SOLOMON P R SERIO M A AJD HENINGER 19 SMOOT L D Personal communication (1992)S G ACS Div of Fuel Chem Preprints 31 20 SMITH I W HARRIS D J VALIX M G AND (3) 200 (1986) TRIMM D L in Fundamental Issues in Conshy

13 SERIO M A SOLOMON P R CHARPENAY S trol of Carbon Gasification Reactivity J LashyAND SUUBERG E M to be submitted to Fuel haye and P Ehrburger (Eds) Kluwer Acashy(1992) demic Publishers p 49 1991

COMMENTS

N Y Nsakala ABB Combustion Engineering ItIC

USA The Arrhenius plot of your TGA char oxidashytion data (I believe) showed no dependence on charshytype for Illinois 6 Pittsburgh 8 zap lignite and Rosebud subbituminous coal chars Thc global acshytivation energy was -35 kcalfmole in all cases Would you expect this behaviour to hold throughshyout the whole coal rank spcctrum

Authors Reply It has been a consistent obsershyvation in our work that all the coal chars studied which have been produced under different pyrolshyysis conditions and from various coals (including chars from treated coals) showed an activation energy in the range 30 to 35 kcalmo as measured in a TGA apparatus This observation has also been made in several other studies and it is reasonable to cxpect a similar behavior for a wide range of coals

bull Robert Hurt Sandia National Laboratories USA

Your random pure model prcdictions were comshypared to data on chars that had bcen stabilized by heat treatment prior to reaction at a temperature greater than the reaction temperature In coal comshybustion chars are made in the flame-by in situ dcvolatilization-and for this situation we see very different trends Instead of large increases in area at low conversion followed by a maximum we (in our laboratory) generally observe monotonically deshycreasing surface areas Here the char formation process and thermal annealing (which tend to reshyduce surface area) occur Simultaneously with the

oxidation (which tends to increase the area inishytially) We find the net effect to be generally deshycreasing areas at the temperatures and times of inshyterest to pulverized coal combustion The existing pure models cannot predict this trend because they all consider rigid solid matrices in which pores grow by internal rcaction Have you observed such efshyfects and can you imagine how your model may acshycount for it

Authors Reply Proccsscs like thermal annealing occur at high temperature and cannot be acshycounted for in pore models in their present form Howcver the range of validity of porc models is usually at low temperature in the intrinsic regime where the reaction occurs all through the particle In our case at high temperature (ie in the pore diffusion regime) we have assumed that the parti shycles react following the shrinking core model U sshying this model the specific surface area (mg) is assumed to remain constant with burnofT Your exshyperimental observations of generally decreasing surshylace area with burnoff cannot be explained by our current model

bull Reginald E Mitchell Stanford University USA

The calculation of effectiveness factors from Thiele moduli requires knowledge of thc order of reaction with respect to the oxygen concentration Nhat values did you use for reaction orders and why were these values selected

Authors Reply A reaction order of 1 is assumed

1197 PREDICTION OF COAL CHAR REACTIVITY

in the model This choice is consistent with the apshyproach taken in the comprehensive code for enshytrained coal combustors and gasifiers for which this model is being developed (PCGC-2 from Brigham Young University) Several studies have shown that the true reaction order is probably fractional However using a reaction order of 1 typically leads to an error in rate of less than a factor of 1 5 if the true reaction order is for example 05 At this stage of our work this range of uncertainty is acceptable given the other uncertainties in the model and in the available data

bull Ian W Smith CSIRO Division of Coal amp Energy

Technology Australia The paper deals with low and high rank coals-the former coals are stated to be affected by catalysts but the high rank materials not Given the regime I nature of the measureshyments why do not the minerals present catalyse the reaction of the high rank coals

Authors Reply For low rank coals we found the best correlation of reactivity with the amount of dispersed calcium most of which is exchanged on carboxyl groups The concentration of carboxyl groups is not significant in high rank coals For this reason the concentration of well dispersed alkali metals is not high for high rank coals and there is little catalytic enhancement of the char gasification rate Unlike low rank coals demineralization of high rank coals does not significantly reduce the gasiflshy

cation rate In fact these rates have been reported to increase slightly This has been attributed to reshymoval of minerals which block pores or mild oxishydation of the coal during the demineralization proshycess which produces a less fluid coal and a more reactive char after pyrolysis

bull Fiero Salatino Universita di Napoli Italy The

Thiele model strictly applies to solids characterized by unimodal or nearly unimodal pore site distrishybution Did the materials used in your study corshyrespond to this hypothesis and if not would you comment on the effect that this simplification might have on the reliability of your diffusion-limited combustion model

Authors Reply The effect of the pore size disshytribution is folded into the value of the effective diffusivity Deff- Since it is difficult to evaluate Del we assumed the Knudsen diffusivity to hold and used a mean pore radius to account for the pore size distribution This approach is only approximate since the chars used in our study are probably not unishymodal However only a certain range of pore sizes will have an impact on diffusion limitations the mishycropore-mesopore range From our simulations for the chars where the pore size distribution is known a reasonable prediction is obtained for the reactivshyity in the pore diffusion regime Additional comshyparisons are underway in order to fully validate this approach

1197 PREDICTION OF COAL CHAR REACTIVITY

in the model This choice is consistent with the apshyproach taken in the comprehensive code for enshytrained coal combustors and gasifiers for which this model is being developed (PCGC-2 from Brigham Young University) Several studies have shown that the true reaction order is probably fractional However using a reaction order of 1 typically leads to an error in rate of less than a factor of 1 5 if the true reaction order is for example 05 At this stage of our work this range of uncertainty is acceptable given the other uncertainties in the model and in the available data

bull Ian W Smith CSIRO Division of Coal amp Energy

Technology Australia The paper deals with low and high rank coals-the former coals are stated to be affected by catalysts but the high rank materials not Given the regime I nature of the measureshyments why do not the minerals present catalyse the reaction of the high rank coals

Authors Reply For low rank coals we found the best correlation of reactivity with the amount of dispersed calcium most of which is exchanged on carboxyl groups The concentration of carboxyl groups is not significant in high rank coals For this reason the concentration of well dispersed alkali metals is not high for high rank coals and there is little catalytic enhancement of the char gasification rate Unlike low rank coals demineralization of high rank coals does not significantly reduce the gasiflshy

cation rate In fact these rates have been reported to increase slightly This has been attributed to reshymoval of minerals which block pores or mild oxishydation of the coal during the demineralization proshycess which produces a less fluid coal and a more reactive char after pyrolysis

bull Fiero Salatino Universita di Napoli Italy The

Thiele model strictly applies to solids characterized by unimodal or nearly unimodal pore site distrishybution Did the materials used in your study corshyrespond to this hypothesis and if not would you comment on the effect that this simplification might have on the reliability of your diffusion-limited combustion model

Authors Reply The effect of the pore size disshytribution is folded into the value of the effective diffusivity Deff- Since it is difficult to evaluate Del we assumed the Knudsen diffusivity to hold and used a mean pore radius to account for the pore size distribution This approach is only approximate since the chars used in our study are probably not unishymodal However only a certain range of pore sizes will have an impact on diffusion limitations the mishycropore-mesopore range From our simulations for the chars where the pore size distribution is known a reasonable prediction is obtained for the reactivshyity in the pore diffusion regime Additional comshyparisons are underway in order to fully validate this approach