Fuel 207 (2017) 126135

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full length article

Simulation of pyrolysis in low rank coal particle by using DAEM kinetics

model: Reaction behavior and heat transfer
Junli Wang a,b, Wenhao Lian a, Peng Li c, Zhonglin Zhang a, Jingxuan Yang a, Xiaogang Hao a,, Wei Huang c,,
Guoqing Guan d,
a
School of Chemistry and Chemical Engineering, Taiyuan University of Technology, Taiyuan 030024, China
b
Department of Chemistry and Environmental Engineering, Datong University, Datong 037009, Shanxi, China
c
Key Laboratory of Coal Science and Technology of Ministry of Education and Shanxi Province, Taiyuan University of Technology, Taiyuan 030024, Shanxi, China
d
North Japan Research Institute for Sustainable Energy (NJRISE), Hirosaki University, 2-1-3 Matsubara, Aomori 030-0813, Japan

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Kinetic parameters of coal pyrolysis

were obtained by DAEM via TG/DTG
experiments.
10 1

E0 = 186.5 kJ/mol, k0 = 3.96  10 s
and r = 39.5 kJ/mol were obtained.

A coal particle pyrolysis model
coupling with reaction and heat
transfer was proposed.

Mass fraction and temperature
profiles inside the coal particle was
well predicted.

DT > 300 K from surface to core of a
3 mm particle was predicted at
900 C pyrolysis.

a r t i c l e i n f o a b s t r a c t

Article history: A comprehensive and systematic study on the fundamental pyrolysis behaviors of a single coal particle
Received 26 March 2017 was performed in this study. The pyrolysis characteristics of coal was investigated by non-isothermal
Received in revised form 27 May 2017 thermo gravimetric analysis whereas the reaction kinetic parameters were obtained by using the dis-
Accepted 18 June 2017
tribute activation energy model (DAEM). As three heating rate profiles were applied (10, 20 and 30 C/
Available online 23 June 2017
min) in TG/DTG experiments with a final pyrolysis temperature of 900 C, the obtained kinetic parame-
ters, i.e., activation energy (E0), pre-exponential factor (k0) and standard deviation (r) were 186.5 kJ/mol,
Keywords:
3.96  1010 s1 and 39.5 kJ/mol, respectively. When these calculated kinetic parameters were used to
Low rank coal
Pyrolysis
predict devolatilization curves, the simulation results were in well agreement with the experimental
Distribute activation energy model (DAEM) data. As such, a one-dimensional, time-dependent particle pyrolysis model was proposed to characterize
Kinetics the detailed chemical and physical phenomena occurred within a pyrolyzing coal particle. It is found that
Single particle model this model successfully predicted the mass fraction residue and temperature profiles inside the coal par-
ticle. In addition, the effect of particle size on pyrolysis performance was also investigated through sim-
ulation. It is expected that such a model can be integrated with CFD simulation to provide useful insight
for the design of a practical coal pyrolysis reactor.
2017 Elsevier Ltd. All rights reserved.

Corresponding authors.
E-mail addresses: xghao@tyut.edu.cn (X. Hao), huangwei@tyut.edu.cn (W. Huang), guan@hirosaki-u.ac.jp (G. Guan).

http://dx.doi.org/10.1016/j.fuel.2017.06.078
0016-2361/ 2017 Elsevier Ltd. All rights reserved.
 J. Wang et al. / Fuel 207 (2017) 126135
Nomenclature
D particle diameter, mm
E the activation energy, kJ mol1
E0 the mean activation energies, kJ mol1
f(E) the activation energy distribution
hc the convective heat-transfer coefficient, W m2 K1
k0 the pre-exponential factor, s1
R the universal gas constant, J mol1 K1
Rv the instantaneous volatile evolution rates, kg m3 s1
T temperature, C
T0 initial temperature of coal particle, C
Ts temperature at particle surface, C
Tf reactor temperature, C
V volatile loss
V the final fractional volatile loss
1. Introduction
Coal is the mainstay of energy in China and this situation will
remain unchanged in foreseeable future. As we all known, pyroly-
sis is one of the most promising technologies for clean and effective
utilization of low-rank coal, and it is also the first step in most of
the existing multi-generation technologies [13]. However, most
of coal pyrolysis processes need pulverized coal, which suffers
from many disadvantages in their industrialized application, such
as complication and high-cost of coal pretreatment, difficult sepa-
ration of tar from pulverized coal, and easy blockage of the pipe by
the mixture of tar and ash. Hence, using medium-sized coal parti-
cles in the industrial reactor is quite necessary.
Since coal is a particularly complicated mixture, how to simu-
late the complex coal pyrolysis has been studied in the last dec-
ades. To date, the empirical models and the network models are
mainly applied. Those early empirical models, such as the first-
order reaction model [4], the multiple-step reaction model [5],
and the two-competing-rate model [6] are too simple to be applied
widely. Recently, the physicochemical structure of coal has been
well understood so that more fundamental approaches have been
used to simulate the coal pyrolysis behavior. For instances, the net-
work models, including the functional group devolatilization
vaporization cross-linking model (FG-DVC) [7], the distributed-
energy chain statistics (Flash chain) model [8,9], and the chemical
percolation devolatilization (CPD) [10,11] were proposed and
applied. However, some disadvantages, such as the accuracy of
model input parameters in terms of the coal structure and the
complexity of the model that limiting their application in combin-
ing with the simulation of reactor are still remained. Recently, Bar-
tocci et al. [12] proposed a four-parallel-reaction scheme for a
pellet made of 90% sawdust and 10% glycerol. For each pseudo-
component (cellulose, hemicellulose, lignin and glycerol), the first
order reaction was used, in which the kinetic parameters for each
component were derived using a model fitting approach. It is found
that the modeling results by using this method well matched over
the entire range of experimental data. Pitt [13] had ever proposed a
distributed activation energy model (DAEM), in which the pyroly-
sis of coal was assumed to have a large number of irreversible
independent and parallel first-order reactions with different acti-
vation energies. This model truly and effectively depicted the main
features of chemical behaviors observed in coal pyrolysis. Miura
and Maki [14] further developed a simpler and more accurate
method to estimate f(E) and k0(E) in the DAEM, in which only three
sets of thermogravimetric analysis (TG)/derivative thermogravi-
metric analysis (DTG) experimental data obtained at different
127
Greek letters
DH specific reaction heat of pyrolysis, kJ kg1
a the extent of conversion
b the heating rate, K s1
er emissivity
ks effective thermal conductivity of coal, W m1 K1
q coal bulk density, kg m3
q0 initial coal bulk density, kg m3
qCp the volumetric specific heat capacity, J m3 K1
r Stefan-Boltzman constant, W m2 K4
rE the standard deviations, kJ mol1
heating rates were needed. This model was proven to be able to
describe the pyrolysis behavior over a wider range of operating
conditions and has been successfully applied to study the pyrolysis
of complex matters, such as biomass [15,16], coal [17,18], solid
waste [19,20] and the synergistic effect of coal and biomass [21].
However, sometimes, the results obtained by isoconversional
methods including Miura-Maki method cannot well match with
the experimental data over the entire range when compared with
those obtained by the model-fit method [22]. In previous work, the
definitions of k0, f(E) and the reaction order in the model-fit
method were different. For the f(E) in the DAEM, various distribu-
tion functions including Dirac delta distribution (for SRMs), logistic
distribution [23] and Gaussian distribution [24] were used. Miura
used the Gaussian distributed function for DAEM and found that
the assumption was reasonable for 19 type coals [25]. In the opti-
mization schemes, various researchers extracted the kinetic
parameters using different methods, such as direct search method
[24], simulated annealing method [26], pattern search (PS) method
[27], multistart algorithm method [28] and differential evolution
algorithm method [29]. Among them, PS method was claimed to
be better than others [27]. In the process of fitting, due to the com-
pensation effect [30,31], multiple values of k0 and E0 fitted the
experimental data well. Thus, the suitable initial values for the
model-fit analysis are very important. However, most of studies
either pre-fixed the pre-exponential factor (randomly or from the
reported values in literature) or maintained it as the optimization
parameter based certain assumptions on the distribution of activa-
tion energy [32,33]. Therefore, it is necessary to develop a suitable
way to choose the appropriate initial values for model-fit method.
In addition, the kinetic models mentioned above can be only used
when the coal particle is very small and the temperature inside it
can be assumed as isothermal. In fact, the pyrolysis behavior of
large coal particle is very different from that of the pulverized coal
because the intraparticle mass and heat transfer noticeably affect
the coal pyrolysis progress which is very important for the design
of reactor with better performance. Furthermore, different coals
have different structures and compositions, resulting in different
pyrolysis characteristics. Besides, even for the same coal, the pyrol-
ysis processes are also different in different reactors due to the
variation in heat transfer coefficient. As such, the simulation of
kinetics of a given coal particle in a specific reactor is very neces-
sary, which is also the foundation of industrial application of the
reactor. However, most of studies on the simulation of reactor used
only the intrinsic kinetics, basically used the CFD simulation with
the single-step global model [3436]. In the coal topping process,
pyrolysis of millimeter-sized coal particles should be much
128 J. Wang et al. / Fuel 207 (2017) 126135

different from that of the pulverized coal [37]. As such, the particle as the carrier gas to provide an inert atmosphere. Non-isothermal
model becomes very important for the reactor simulation as an runs were carried out at heating rates of 5, 10, 20 and 30 C/min
important sub-model. with temperatures ranging from ambient temperature to 900 C
As stated above, since DAEM has wide application range and and then held at 900 C for 20 min until the residue mass at this
can be easily coupled with the reactor simulation, in this study, temperature was unchanged. Each experiment was repeated at
it was also used for the simulation of pyrolysis in Inner Mongolia least twice for repeatability.
Xinghe coal particle. In view of the importance of the suitable ini-
tial values for the model-free analysis, a method combining the 2.3. Theoretical backgrounds of DAEM
model-fit method and model-free method was developed to obtain
the intrinsic kinetic parameters which can reflect the pyrolysis The DAEM has been confirmed to be a powerful method to sim-
characteristics of Inner Mongolia Xinghe coal more exactly. Here, ulate the complex pyrolysis process and the simulation results are
the kinetic parameters together with the activation energy distri- always good agreement with the experimental data [16,19,20]. The
butions in DAEM were obtained firstly by non-isothermal thermo DAEM model assumes that coal pyrolysis contains a large number
gravimetric analysis and then, a particle model based on the of irreversible independent parallel first-order reactions with dif-
obtained kinetic parameters was proposed to predict the tempera- ferent activation energies. The model is expressed as:
1    
ture distribution and the evolution of the volatiles during the heat- Z Z
k0 T E
ing process. After validation, the model was used to predict the aT 1  exp  exp  dT f EdE 1
mass fraction of residue, the temperature profiles inside the coal 0 b 0 RT
particle and the complete pyrolysis time. It is expected to gain a
where a is extent of conversion, T is the absolute temperature, k0 is
better understanding of the pyrolysis behaviors inside a coal parti-
the pre-exponential factor, b is the heating rate, E is the activation
cle for the design of pyrolyzer structure and optimization of reac-
energy, R is the universal gas constant, and f(E) is the activation
tion conditions.
energy distribution. To estimate the values of kinetic parameters,
the f(E) is generally assumed by a Gaussian distribution with mean
2. Materials and methods activation energy E0 and standard deviation r as shown below:
!
2.1. Materials 1 E  E0 2
fE p exp  2
r 2p 2r2
The Inner Mongolia Xinghe coal was selected as the sample. In
this work, the sample was pulverized and sieved into <0.075 mm Thus, by differentiation from the above equation, the following
(200 mesh) of particle size using stainless steel sieve and used differential equation can be obtained:
for TG/DTG experiment. Before experiment, the sample was dried Z " Z   #
daT 1 1
k0 E k0 T E E  E0 2
in an oven at 110 C for 10 h to remove moisture and then stored p exp   exp  dT  dE
in the desiccator. The proximate and ultimate analyses of the sam- dT r 2p 0 b RT b 0 RT 2r 2
ples are shown in Table 1. Here, the proximate analysis was per- 3
formed based on ASTM 3172-75. The ultimate analysis was
carried out using an elemental analyzer (Vario EL cube elemental To determine the kinetic parameters (k0, E0 and r) of non-
analyzer). Ash compositions were analyzed by an energy dispersive isothermal DAEM equations, Miura and Maki proposed a simplifi-
X-ray spectrometer (EDX-800HS, Shimadzu). The remaining mois- cation for Eq. (3), and the simplified DAEM was given as [14]:
ture was analyzed by using MX50 moisture content analyzer (AND,
   
b k0 R E
Japan). ln ln 0:6075  4
T2 E RT

2.2. Thermo gravimetric analysis The values of the activation energy, E, and the regarding pre-
exponential factor, k0, given in Eq. (4) can be estimated from the
The pyrolysis of Inner Mongolia Xinghe coal was carried out in a slope and intercept in the plot of ln (b/T2) versus 1/T. Here, the plot
thermo gravimetric analyzer (Setaram SETSYS). The thermobalance needs at least three thermogravimetric curves acquired at different
had a sensitivity of 0.001 lg. In each experiment, 10 mg of sample heating rates. Based on the obtained parameter range and by using
was loaded into the crucible of the thermal analyzer. During exper- the average value as the initial value, the kinetics parameters in the
iment, high-purity Ar gas with a flow rate of 100 cm3/min was used DAEM model can be estimated by the pattern search method. Here,
the objective function based on the conversion rate (da/dT) data is
defined according to the following equation:
Table 1 "    #2
Properties of the Inner Mongolia Xinghe coal. X
nd
da da
S  5
Proximate analysis (wt.%)a
i1
dT exp;i dT cal;i
b
Volatile matters 41.0
Ash 11.9 where nd is the number of the data points, (da/dT)exp represents the
Fixed Carbon 47.1 experimental data of conversion rate, (da/dT)cal was calculated by
Moisturec 11.4 Eq. (3) for a given set of parameters of k0, E0 and r.
Ultimate analysis (wt.%)d
C 69.1
2.4. Single particle modeling and solution derivation
H 4.6
Ob 20.6
N 1.0 As stated above, for the pulverized coal with a low heating rate,
S 4.7 the kinetics parameters of coal pyrolysis can be obtained by
a
Dry base.
assuming no thermal gradients inside particle. However, for the
b
By difference. coal with a large particle size used in an industrial reactor, due
c
Air dried basis. to the inherent heat-transfer and mass-transfer resistances, only
d
Dry ash-free base. using the kinetic model may not be adequate to predict the
 J. Wang et al. / Fuel 207 (2017) 126135 129

 
complex pyrolysis process inside it. In this work, the inside heat @qC p T 1 @ @T
2 ks r 2 q0 Rv DH 8
transfer will be considered because it is the rate-controlling factor @t r @r @r
of coal pyrolysis [38]. In this section, a comprehensive particle
T represents the local temperature at any radial position r and
model coupling with the pyrolysis kinetics and energy balance
time t. The thermal conduction of coal (ks ) was assumed to vary
was established first and then, an effective solution derivation
with the temperature, and it can be calculated using the following
method was applied to simulate the pyrolysis progress.
expression given by Badzioch [40]:

2.4.1. Model assumption 0:23 for T 6 400  C
ks 5 1:8 9
To establish a general heat transfer combined volatiles evolu- 0:23 2:24  10 T  400 for T > 400  C
tion model, some basic simplification and assumptions were made
The volumetric heat capacity (qCp) should also vary with T and
as follows: (1) the coal particle was assumed to be a porous sphere
can be estimated from [40]:
and its shape and size retained unchanged during the pyrolysis
(
process; (2) the volatiles and the remained solid at any local posi- 1:92  106 for T 6 350  C
tions inside the particle were assumed to be in a thermal equilib- qC p 6 3
1:92  10  2:92  10 T  350 for T > 350  C
rium state; (3) the pyrolysis products were divided into only
solid char and volatile; (4) the physical properties were considered 10
homogeneous and isotropic inside the single particle; (5) the mass- As described in the assumption, the single particle was spheri-
transfer and the convective heat transfer resistance of the pro- cal and symmetrical. In this case, the heat change at the outer sur-
duced gas were ignored [39]. face and the central of the particle are different from that of the
It is well known that the simple kinetics reaction models with bulk coal. Therefore, the heat transfer calculation should combine
relatively few parameters can be easily used and numerically with the following initial and boundary conditions, given in Eqs.
solved in the devolatilization submodels of CFD software. How- ((11)(13)).
ever, it is proved that these models are too simple so that they have Initial condition:
poor adaptability and cannot be applied widely. In the other
extreme, the network models which allow a better representation at t 0 for 0 6 r 6 r0 ; T T0 ; q q0 11
of the process, but they need an accuracy of model input parame- Boundary conditions:
ters in terms of the coal structure as well as significantly long com-
puting time. Despite of great improvements in the computational @T
for t > 0; at r 0; 0 12
capability, the direct coupling of detailed chemical structure with @r
CFD simulations is still difficult for a real reactor since a large

@T s
amount of computational grids are needed. Therefore, based on for t > 0; at r r0 ; ks hc T s  T f rer T 4s  T 4f 13
the above assumptions, in this work, a particle model based on @r
DAEM was developed for the pyrolysis of coal particle since the Particularly, Eq. (13) represents a hybrid boundary condition
DAEM can reflect a practical situation and easily couple with CFD combining the convective heat transfer and heat radiation. A radia-
simulation. tive effect should be included in this boundary condition because a
high pyrolysis temperature was considered in this study.
2.4.2. Mathematical model Based on Eqs. (6)(13), by coupling the DAEM model with
The instantaneous volatile evolution rates (Rv) at any radial energy equation, the temperature and mass loss of particles were
locations were calculated by the DAEM model: simulated simultaneously. The parameters used in this particle
Z    ! model are listed in Table 2.
q0 E E  E0 2
Rv p k0 exp  exp  V   V dE 6
r 2p RT 2r2 2.4.3. Solution derivation
The variables such as temperature (T) and fractional volatile
Here, the prediction of fractional volatile loss required the val- loss (V) are functions of time and radial position, and can be
ues of K0, E0, r and V which were obtained by the TG/DTG obtained by solving the partial differential equations Eqs. (6)
experiment. (13) based on the initial and boundary conditions. By using the
finite difference method, the above governing partial differential
2.4.2.1. Mass conservation equation.
@q
Rv 7 Table 2
@t The input parameters used in the particle model.

As stated above, the shape and size of coal particle were Parameters Values Data
assumed to remain unchanged during the pyrolysis, and in this source
case, the variation in particle bulk density should be mainly caused E0(kJ/mol), k0(s1) and r(kJ/mol) 186.5, 3.96  1010 Present
by the devolatilization, which can be calculated by Eq. (7). Here, and 39.5 work
the volume of solid was also assumed to be constant while its den- Diameter of coal particle, D(mm) 3, 0.3
Specific reaction heat of pyrolysis, DH 300 [27]
sity (q) decreased in proportion to the total volatile matter loss (kJ/kg)
during the pyrolysis process. Initial temperature, T0 (C) 25
Bulk temperature, Tf (C) 900
2.4.2.2. Heat conservation equation. Also, based on the above Effective thermal conductivity of coal, ks Eq. (9) [26]
(W m1 K1)
assumption, the volatiles and the remained solid at any local posi- The volumetric specific heat capacity, Eq. (10) [26]
tions inside the particle were considered to be in a thermal equilib- qCp (J m3 K1)
rium state. The heat change inside the coal particle was mainly Emissivity, er 0.9 [28]
caused by the thermal conduct and the endothermic reaction dur- Stefan-Boltzman constant, r (W m2 K4) 5.67  108
The final fractional volatile loss, V* 0.38 Present
ing pyrolysis. As such, the temperature variation can be calculated
work
by Eq. (8).
130 J. Wang et al. / Fuel 207 (2017) 126135
equations were transformed into the discretized nonlinear alge-
braic system of equations which were solved by using the MATLAB
code which was developed by us to solve the model equations.
3. Results and discussion
3.1. Thermo gravimetric analysis
The mass loss (TG) and differential thermo gravimetric (DTG)
curves of Inner Mongolia Xinghe coal pyrolysis obtained at four dif-
ferent heating rates (5, 10, 20 and 30 C/min), from 30 C to 900 C
under Ar atmosphere, are depicted in Fig. 1a and b. It can be seen
that both TG and DTG curves showed similar trends for all heating
rates. From DTG curves in Fig. 1b, at the temperature below 200 C,
the first mass loss stage was observed, which should correspond to
the removal of physically adsorbed water and gases. In the temper-
ature range of 200600 C in DTG curve, a strong peak was
observed, which should be the main pyrolysis stage, where large
amounts of tar and light gases were given off so that the mass loss
of coal varied dramatically in this region. A shoulder peak between
200 and 300 C appeared, indicating the release of bonded water
and decomposition of carboxylic acid. Thereafter, in the tempera-
ture range between 300 and 600 C, the thermal energy could
result in the breakage of chemical bonds between Cal or Car and
O, S and N, and SnS so that a large amount of tar was produced.
Finally, after the pyrolysis temperature was increased above
700 C, only a slight weight loss was observed, which should be
the result of decomposition of carbonates in coals to generate
CO2 as well as the condensations of aromatic rings in biochar to
release H2 [41].
However, as shown in Fig. 1, both TG and DTG curves existed
some differences with the increase in the heating rate. It can be
observed in Fig. 1a that the mass loss of coal increased slightly with
the increase in the heating rate. Here, when the heating rate was
increased, the thermal shock to the coal structure should be more
intense, leading to the thermal decomposition of more species with
higher molecular weight which cannot be broken at lower heating
rate. Moreover, as seen from DTG in Fig. 1b, the heating rate
affected the pyrolysis rate obviously and the decomposition tem-
perature was delayed as the heating rate was increased. This phe-
nomenon should be due to that more thermal energy was provided
for the pyrolysis at higher heating rate. In other words, the high
heating rate could promote the heat transfer between surrounding
and interior of sample.
The generation mechanism of the main products during coal
pyrolysis has been widely studied [4245]. Mae et al. [43] found
that the formation of H2O, CO and CO2 was significantly dependent
Fig. 1. (a) TG and (b) DTG of Inner Mongolia Xinghe coal at different heating rate.
on the amount of oxygen functional groups in the coal. Even at rel-
atively low temperature, the hydrogen bonds could be decom-
posed to form the inorganic gases, resulting in the production of
cross-linked chemicals such as anhydrides, ether, and ester, which
could be further decomposed into smaller molecules at a temper-
ature above 520 C. Porada [44] proposed the generation mecha-
nism of H2 and pointed out that at the early stage of coal
pyrolysis, H2 was mainly formed from the dehydrogenation of
hydroaromatic structures; and at higher temperatures, H2 was
mainly from the dehydrogenation of aromatic structures and the
condensation of the aromatic nuclei into coal char. Based on the
pyrolysis behaviors of various coals and the relevant model sub-
stances, van Heek and Hodek [45] found that three typical reaction
steps were existed in the coal pyrolysis process: in the first pyrol-
ysis step, CH4 was from the cleavage of aryl-methyl ether bonds,
and simultaneously, some of the remaining phenoxy radicals were
decomposed so that CO was generated; in the second pyrolysis
step, the cleavage of strong bonds, in particular biaryl ethers and
methylene bridges resulted in the formation of CH4, CO and tar,
and the tar formation was mainly from the cleavage of the methy-
lene bridges between the aromatic units; in the third step, mainly
the cleavage of aromatic heterocyclic structures in the char
occurred. These studies on the mechanism of volatile formation
help us deeply understand the coal pyrolysis process.
3.2. Kinetic analysis using DAEM model
In this study, the DAEM model was used to evaluate the activa-
tion energies and pre-exponential factors for the coal pyrolysis. At
different heating rates, i.e., from 10 to 30 C/min, eighteen levels of
conversions varying from 0.10 to 0.95 were used to determine the
variation of kinetic parameters during the pyrolysis progress. Using
Eq. (4), a series of E and k0 were calculated from the slope and
intercept of all lines from three different heating rates. Fig. 2 pre-
sents the plots of ln (b/T2) vs 1/T at various conversions ranging
from 0.10 to 0.95. The fitted equations, correlation coefficients
(R2) and kinetics parameters are summarized in Table 3.
As seen in Fig. 2 and Table 3, for all heating rates, R2 values of all
regressing lines, except for 0.95, were higher than 0.90. Thus, it can
be concluded that the DAEM model was reliable to determine the
kinetic parameters of coal pyrolysis. On the other hand, it can be
seen that the activation energy (E) and pre-exponential factor
(k0) were not constant but varied with the progress in the conver-
sion, which proved that there were more than one single reaction
mechanism in the coal pyrolysis. The activation energies were also
found to be in the range of 73.98224.38 kJ/mol, and the pre-
exponential factors for the corresponding activation energy were
 J. Wang et al. / Fuel 207 (2017) 126135 131

Fig. 2. Plots for determination of activated energy at different conversion.
observed to be in the range of 2.88  1041.71  1011 s1 for Inner
Mongolia Xinghe coal. Moreover, it is found that the compensatory
effect existed in the pyrolysis of Xinghe coal. That is, with the
increase in the activation energy, the pre-exponential factor also
increased. Cai et al. [31] reported that the activation energy values
for solid-state reactions should be in the range of 50350 kJ/mol
[46], indicating that the activation energy obtained in this study
was reasonable.
The validity of the above method was clarified by Maki using 19
different coals [47]. However, it should be noted that this isocon-
versional method did not result in a good match with the experi-
mental data over the entire range when compared with the
model-fit method [22]. In addition, the forms of f(E) and k0(E)
could be unfavorable to couple with the particle model. The other
approach based on the above method was also developed to
describe the coal pyrolysis. Ulloa et al. [48] investigated the rapid
pyrolysis of coal blends in a drop-tube reactor at high tempera-
tures (10001300 C) with low residence times (60320 ms) using
the DAEM, in which a defined Pyrolysis Yield Index (PYI) that
combined the maceral composition and ash content seemed to
be correlated well with the parameter of the kinetic model at high
temperatures. However, the secondary cracking and re-
polymerization reactions of the carbonaceous matrix are predom-
inant at the high temperatures (over 1000 C). Thus, their parame-
ters were not suitable for this study since the coal pyrolysis
occurred at low temperature in this study and the pyrolysis mech-
anism at high temperature is different from that at low tempera-
ture. Gunes et al. [49] proposed a simple direct search method
for the determination of DAEM kinetic parameters from the TGA
data of coals. Cai et al. [27] considered that the pattern search
(PS) method should be better than the simple direct search method
since the minimum value can be achieved in the least number of
iterations [22]. Hence, the PS method was also used in this study.
Moreover, recently, Caprariis et al. [50] developed a two-
Gaussian DAEM (2-DAEM) and compared with the results obtained
by the standard DAEM. It is found that the 2-DAEM was more suit-
able to represent the experimental data than the standard one.
However, as most other studies, they pre-fixed the pre-
exponential factor (k0) from the reported values in literature. In
this case, five parameters should be confirmed and thus, when cou-
pling with CFD simulation, a long computation time is necessary.
Therefore, the standard DAEM, which can balance the description
of intrinsically complex nature of coal pyrolysis and the computa-
tional capability, was used in this study.
To obtain more reasonable kinetics parameters, a new method
which combined model-fit method and model-free method was
developed. Based on the above obtained E0 and k0 value ranges,
and using the average value as the initial value, the kinetic param-
eters of DAEM can be estimated by the pattern search method. As a
result, the kinetic parameters E0, k0 and r in this study were
186.5 kJ/mol, 3.96  1010 s1 and 39.5 kJ/mol, respectively. As seen
from Fig. 3, the simulation results using the obtained parameters
were in good agreement with the experiment data, and therefore,
it can be used further in the following particle model. These results
indicate that the DAEM can truly and effectively depict the main
features of chemical behaviors observed in coal pyrolysis for the
coal conversion fractions varying from 0.10 to 0.90.
3.3. Modeling of large coal particle pyrolysis
3.3.1. Model validation
As stated above, the particle model was set up based on some
simplifications and assumptions, and therefore, it is necessary to
validate it firstly. Herein, the supposed model was used to predict
the heating history in the center of coal particle with different par-
ticle size in a convective environment as experimentally studied by
Adesanya [40]. The modeling results compared with the experi-
mental findings are given in Fig. 4.

Table 3
Calculated activation energy (E0) and the pre-exponential factor (k0) of Inner Mongolia Xinghe coal.

Weight loss (V/V*) Slope Intercept R2 E0 (kJ/mol) k0 (s1)

0.10 8898.73 1.78 0.97 73.98 2.88  104
0.15 10625.8 3.54 0.95 88.34 1.99  105
0.20 12011 4.86 0.92 99.86 8.48  105
0.25 13264.6 6.08 0.90 110.28 3.15  106
0.30 14811.4 7.76 0.91 123.14 1.89  107
0.35 16193.6 9.18 0.92 134.63 8.60  107
0.40 17707.8 10.77 0.90 147.22 4.58  108
0.45 19230 12.28 0.91 159.88 2.25  109
0.50 20611.7 13.50 0.92 171.37 8.17  109
0.55 21901.7 14.48 0.90 182.09 2.32  1010
0.60 23385.2 15.56 0.91 194.42 7.32  1010
0.65 24145 15.57 0.96 200.74 7.60  1010
0.70 25582 16.27 0.96 212.69 1.62  1011
0.75 26561 16.29 0.97 220.83 1.71  1011
0.80 26987.2 15.53 0.97 224.37 8.14  1010
0.85 26988.1 14.20 0.98 224.38 2.17  1010
0.90 26552.8 12.26 0.97 220.76 3.06  1010
0.95 24420.1 8.71 0.88 203.03 8.08  107
132 J. Wang et al. / Fuel 207 (2017) 126135
Fig. 3. Model prediction compared with TG experimental data.
It can be seen that the temperature rising process in the center
of particle had three different stages. In the first stage, the temper-
ature increased monotonically with time owing to only thermal
conductivity occurred in the coal particle. In the second stage,
the pyrolysis reaction happened concurrently with heat-transfer
progress, during which a large amount of volatiles were released.
Because the pyrolysis reaction is endothermic below 600 C, a slow
heating platform was formed in the curve. As such, the endother-
mic effect of pyrolysis should be taken into account in the model
to properly predict the transient temperature distribution inside
the coal particle during heating. In the third stage, it was the end
of pyrolysis reaction, where the residual char was heated up to
the environmental temperature with a rapid heating process.
These results indicate that the predictions using this model fitted
in well with the experimental data and thus, the particle model
can be used for the following further simulation study.
3.3.2. Modeling of a coal particle in bubbling fluidized reactor
Fluidized bed can provided uniform temperature distribution,
high specific capacity, fast heat-up and are suitable for a large scale
continuous production. It has been widely used in various pro-
cesses [51,52]. Particularly, by using a fluidized bed reactor, high
oil yield can be achieved from the coal pyrolysis since high heating
rate, short vapor residence time, and rapid cooling of the liquid
product are easily realized in it [51]. In this section, the above
model was used to predict the pyrolysis characteristics of a single
Fig. 4. The comparison of model prediction with experimental data in center
temperature.
Inner Mongolia Xinghe coal particle with a particle size of 3 mm in
diameter in the bubbling fluidized reactor [37]. The temperature
history, the time for coal complete pyrolysis and the volatile evolu-
tion of the coal particle in a bubbling fluidized reactor were inves-
tigated. Fig. 5a and b presented the temperature history and
density change of the coal particle at different time as well as dif-
ferent positions. In order to understand the physicochemical pro-
cess in a heated coal particle more clearly, here, the temperature
variation and the volatile yield in three different regions, i.e., the
core r = 0, the middle layer r = r0/2 and the surface r = r0, with the
pyrolysis time are given in Fig. 5c and d.
As we known, the ideal mixing of inert heat carriers and raw
coal in the bubbling fluidized bed can be easily achieved, and a
heat transfer efficiency as high as 356 W/(m2.s) can be reached.
As seen in Fig. 5c, there was an obvious temperature gradient along
the radial direction in the particle, and only within 10 s, the tem-
perature of the entire area of coal particle reached the environmen-
tal temperature while the pyrolysis reaction finished within 6.5 s
as shown in Fig. 5d. From Fig. 5c, it can be seen that the surface
temperature of coal particle increased greatly at the beginning
and then increased gradually. In this case, the heating rate reached
525 K.s1. It should be noted that the middle layer as well as the
core did not begin to be heated up until 3 s later. The maximum
difference in temperature between the surface and the core of coal
particle reached 423 K. The temperature rising rates in the middle
and core increased at the beginning, however, after the maximum
heating rate values of 142 K.s1 and 125 K.s1 were achieved,
respectively, the temperature rising rates began to decrease grad-
ually due to the endothermic pyrolysis reaction. Thereafter, the
other peak appeared again. This implied that it is unreasonable
to consider the coal particle to have a uniform temperature distri-
bution during the pyrolysis process. The different temperature ris-
ing rate between the surface and the inner should be due to the
fact that they were under completely different environments. The
surface of coal contacted directly with the high surrounding tem-
perature when it entered the fluidized bed. The coal particle sur-
face was rapidly heated by the convective radiation heat and
then, the heat was conducted into the core of the particle. The
heating rate decreased as the temperature difference between
the surrounding and surface decreased. On the other hand, the
heating up of inner region of coal particle was only depended on
the thermal conduction and the heat capacity.
Because of the temperature gradient in the radial direction of
coal particle, the pyrolysis reaction in the particle should not be
uniform and occurred at different time in different local positions,
which can be seen from Fig. 5d. The pyrolysis reaction was com-
pleted within only 2 s on the surface while that in the middle layer
 J. Wang et al. / Fuel 207 (2017) 126135
Fig. 5. The simulation results for a coal particle with a particle size of 3 mm in diameter in the bubbling fluidized reactor: (a) Time(t)-position(r)-temperature(T); (b) Time(t)-
position(r)-density(q); (c) Local transient temperature and the rate of temperature rose at core, middle layer and surface; (d) Local transient volatile yields and the rate of
volatile released at core, middle layer and surface.
and the core began to occur until over 2 s and 3 s, respectively. The
rate of volatile evolution was found to have an upward-convex
shape and the peak positions for the local pyrolysis appeared at dif-
ferent time, indicating that the complete pyrolysis of different
regions occurred at different time. It is found that the pyrolysis
reaction completed layer by layer from the surface to the core with
different pyrolysis rate and time and as such, the reaction front
existed inside coal particle, which can be described by using a
shrinking-core model, seen in Scheme 1. In addition, as seen from
Fig. 5d, the complete pyrolysis time for the Inner Mongolia Xinghe
coal particle with a diameter of D = 3 mm in the bubbling fluidized
reactor with a surrounding temperature of 900 C was about 6.2 s.
It should be noted that the pyrolysis time is a very important
parameter for the design of reactor and the optimization of operat-
ing conditions.
The shrinking-core model:
pyrolysis
pyrolysis pyrolysis
coal Coal+Char Coal+Char Char
The volume reaction model:
pyrolysiss pyrolysiss
coal Coal+Char Char
Scheme 1. The pyrolysis reaction mechanism for a heating coal particle.
133
3.3.3. Effect of particle size
Fig. 6 presents the surface and center temperature profiles and
the pyrolysis time for the particle with D = 0.3 mm at 900 C. It can
be found that the temperature difference between the surface and
the core and the time to reach the surrounding temperature obvi-
ously significantly decreased as the particle size decreased. For the
particle with a diameter of 3 mm, it usually took about 10 s to
reach 900 C and an obvious temperature difference more than
300 K existed between the surface and the core (Fig. 5c). In con-
trast, for the pulverized coal with a diameter of 0.3 mm, the tem-
perature difference sharply decreased down to zero even within
a time-interval of 0.1 s. Meanwhile, it only took 0.8 s to reach the
environmental temperature but the pyrolysis reaction should fol-
low the volume reaction mechanism (seen in Scheme 1). Therefore,
although it is acceptable to assume the pulverized coal as an
isothermal particle, it is unreasonable for a large coal particle.
As seen from Fig. 6b, for coals with different particle sizes, the
time for complete pyrolysis were also different. It is found that
the pyrolysis of pulverized coal with a diameter of 0.3 mm was
completed within 0.35 s, and with the increase in the particle size,
the temperature difference between the surface and the core
became larger and larger while the complete pyrolysis time was
substantially increased. As indicated above, the complete pyrolysis
time for a coal particle with a diameter of 3 mm needed 6.2 s. In a
word, the smaller the particle, the faster the heating rate, and the
shorter the pyrolysis reaction time. It indicates that larger coal par-
ticles needed longer time to be heated up, resulting in larger
energy consumption. However, as stated above, the pulverized coal
suffered from many disadvantages in their industrialized applica-
tion due to the complication and high-cost of coal pretreatment,
134 J. Wang et al. / Fuel 207 (2017) 126135
1200
a b
1000
800
The core
The surface
T/K
600
400
200
0.0 0.2 0.4 0.6
t/s
Fig. 6. (a) The surface and center temperature profiles for particle size with D = 0.3 mm; (b) The time for coal pyrolysis with different particle size at 900 C.
difficult separation of tar from the pulverized coal, and the easy
blockage of the pipe by the mixture of tar and ash. Therefore, it
is of great importance for a special reactor to use coal with a suit-
able particle size.
4. Conclusions
A thorough and fundamental study on the pyrolysis character-
istics of Inner Mongolia Xinghe coal was conducted by using exper-
imental as well as mathematic modeling method. The kinetic
parameters in DAEM were obtained by model fitting simulations
and the isoconversional method was employed to provide the ini-
tial values. The obtained kinetic parameters E0, k0 and r were
186.5 kJ/mol, 3.96  1010 s1 and 39.5 kJ/mol, respectively. It is
found that the DAEM truly and effectively depicted the main fea-
tures in coal pyrolysis. Based on the obtained intrinsic kinetic
parameters, a comprehensive single particle model coupling with
heat transfer and reaction was established to characterize the
detailed chemical and physical phenomena occurred within a coal
particle in a fluidized bed. The simulation results indicated that the
heat transfer and pyrolysis reaction were complicatedly and
strongly coupled. The inherent heat-transfer resistance weakened
the heating-up rate, resulting in a long heating-up time and a long
devolatilization time for a large particle. For a coal particle with
D = 3 mm, more than 300 K temperature difference was found in
the particle and the reaction followed the shrinking core mecha-
nism, resulting in a complete pyrolysis time as long as 6.2 s, which
was far longer than the pyrolysis time (0.35 s) of coal particle with
D = 0.3 mm following the volume reaction mechanism. Such a par-
ticle model should be very helpful for the industrial reactor
simulation.
Acknowledgements
This work was supported by the International Joint Research Pro-
ject of Shanxi Province (Nos. 2015081051 and 2015081052) and
Special Talents Project of Shanxi Province (No. 201605D211005).
References
[1] Guan G, Fushimi C, Tsutsumi A. Prediction of flow behavior of the riser in a
novel high solids flux circulating fluidized bed for steam gasification of coal or
biomass. Chem Eng J 2010;164:2219.
[2] Qu X, Liang P, Wang Z, Zhang R, Sun D, Gong X, et al. Pilot development of
polygeneration process of circulating fluidized bed combustion combined with
coal pyrolysis. Chem Eng Technol 2011;34:618.
0.4
0.3mm
0.3 3mm
0.4
volitle yield
D=0.3 mm
0.2 0.3
volitle yield
0.2
0.1 0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5
t/s
0.0
0.8 1.0 0 2 4 6 8 10
t/s
[3] Zhang J, Wang Y, Dong L, Gao S, Xu G. Decoupling gasification: approach
principle and technology justification. Energy Fuel 2010;24:622332.
[4] Maffei T, Frassoldati A, Cuoci A, Ranzi E, Faravelli T. Predictive one step kinetic
model of coal pyrolysis for CFD applications. Proc Combust Inst
2013;34:240110.
[5] Sommariva S, Maffei T, Migliavacca G, Faravelli T, Ranzi E. A predictive
multistep kinetic model of coal devolatilization. Fuel 2010;89:31828.
[6] Kobayashi H, Howard JB, Sarofim AF. Coal devolatilization at high
temperatures. Symp Combust 1977;16:41125.
[7] Solomon PR, Hamblen DG, Carangelo RM, Serio MA, Deshpande GV. General
model of coal devolatilization. Energy Fuel 1988;2:40522.
[8] Niksa S, Kerstein AR. Flashchain theory for rapid coal devolatilization kinetics.
1. Formulation. Energy Fuel 1991;5:64765.
[9] Niksa S. Flash chain theory for rapid coal devolatilization kinetics. 7. Predicting
the release of oxygen species from various coals. Energy Fuel 1996;10:17387.
[10] Grant DM, Pugmire RJ, Fletcher TH, Kerstein AR. Chemical model of coal
devolatilization using percolation lattice statistics. Energy Fuel
1989;3:17586.
[11] Jupudi RS, Zamansky V, Fletcher TH. Prediction of light gas composition in coal
devolatilization. Energy Fuel 2009;23:30637.
[12] Bartocci P, Anca-Couce A, Slopiecka K, Nefkens S, Evic N, Retschitzegger S, et al.
Pyrolysis of pellets made with biomass and glycerol: kinetic analysis and
evolved gas analysis. Biomass Bioenergy 2017;97:119.
[13] Pitt GJ. The kinetics of the evolution of volatile products from coal. Fuel
1962;41:26774.
[14] Miura K. A new and simple method to estimate f(E) and k0(E) in the distributed
activation energy model from three sets of experimental data. Energy Fuel
1995;9:3027.
[15] Sonobe T, Worasuwannarak N. Kinetic analyses of biomass pyrolysis using the
distributed activation energy model. Fuel 2008;87:41421.
[16] Soria-Verdugo A, Garcia-Gutierrez LM, Blanco-Cano L, Garcia-Hernando N,
Ruiz-Rivas U. Evaluating the accuracy of the distributed activation energy
model for biomass devolatilization curves obtained at high heating rates.
Energy Convers Manage 2014;86:10459.
[17] Liu X, Li B, Miura K. Analysis of pyrolysis and gasification reactions of
hydrothermally and supercritically upgraded low-rank coal by using a new
distributed activation energy model. Fuel Process Technol 2001;69:112.
[18] Wang J, Li P, Liang L, Yang J, Hao X, Guan G, et al. Kinetics modeling of low-rank
coal pyrolysis based on a three-gaussian distributed activation energy model
(daem) reaction model. Energy Fuel 2016;30:9693702.
[19] Bhavanam A, Sastry RC. Kinetic study of solid waste pyrolysis using distributed
activation energy model. Bioresour Technol 2015;178:12631.
[20] Wu W, Cai J, Liu R. Isoconversional kinetic analysis of distributed activation
energy model processes for pyrolysis of solid fuels. Ind Eng Chem Res
2013;52:1437683.
[21] Filippis P, Caprariis B, Scarsella M, Verdone N. Double distribution activation
energy model as suitable tool in explaining biomass and coal pyrolysis
behavior. Energies 2015;8:173044.
[22] Jain AA, Mehra A, Ranade VV. Processing of TGA data: analysis of
isoconversional and model fitting methods. Fuel 2016;165:4908.
[23] Cai J, Jin C, Yang S, Chen Y. Logistic distributed activation energy model-Part 1:
derivation and numerical parametric study. Bioresour Technol
2011;102:155661.
[24] Gnes M, Gnes S. A direct search method for determination of DAEM kinetic
parameters from nonisothermal TGA data (note). Appl Math Comput
2002;130:61928.
[25] Miura K, Maki T. A simple method for estimating f(E) and k0(E) in the
distributed activation energy model. Energy Fuel 1998;12:8649.
 J. Wang et al. / Fuel 207 (2017) 126135
[26] Mani T, Murugan P, Mahinpey N. Determination of distributed activation
energy model kinetic parameters using simulated annealing optimization
method for nonisothermal pyrolysis of lignin. Ind Eng Chem Res
2009;48:14647.
[27] Cai JM, Ji LQ. Pattern search method for determination of DAEM kinetic
parameters from nonisothermal TGA data of biomass. J Math Chem
2006;42:54753.
[28] Kirtania K, Bhattacharya S. Application of the distributed activation energy
model to the kinetic study of pyrolysis of the fresh water algae Chlorococcum
humicola. Bioresour Technol 2012;107:47681.
[29] Santos KG, Lobato FS, Lira TS, Murata VV, Barrozo MAS. Sensitivity analysis
applied to independent parallel reaction model for pyrolysis of bagasse. Chem
Eng Res Des 2012;90:198996.
[30] Czajka K, Kisiela A, Moron W, Ferens W, Rybak W. Pyrolysis of solid fuels:
thermochemical behaviour, kinetics and compensation effect. Fuel Process
Technol 2016;142:4253.
[31] Holstein A, Bassilakis R, Wjtowicz MA, Serio MA. Kinetics of methane and tar
evolution during coal pyrolysis. Proc Combust Inst 2005;30:217785.
[32] Cai J, Liu R. New distributed activation energy model: numerical solution and
application to pyrolysis kinetics of some types of biomass. Bioresour Technol
2008;99:27959.
[33] Fiori L, Valbusa M, Lorenzi D, Fambri L. Modeling of the devolatilization kinetics
during pyrolysis of grape residues. Bioresour Technol 2012;103:38997.
[34] Sharma R, May J, Alobaid F, Ohlemller P, Strhle J, Epple B. Euler-Euler CFD
simulation of the fuel reactor of a 1 MWth chemical-looping pilot plant:
influence of the drag models and specularity coefficient. Fuel 2017;200:43546.
[35] Klimanek A, Adamczyk W, Katelbach-Wozniak A, Wecel G, Szlek A. Towards a
hybrid Eulerian-Lagrangian CFD modeling of coal gasification in a circulating
fluidized bed reactor. Fuel 2015;152:1317.
[36] Shu Z, Fan C, Li S, Wang J. Multifluid modeling of coal pyrolysis in a downer
reactor. Ind Eng Chem Res 2016;55:263445.
[37] Yang C, Li S, Song W, Lin W. Pyrolysis behavior of large coal particles in a lab-
scale bubbling fluidized bed. Energy Fuel 2013;27:12632.
[38] Chern JS, Hayhurst AN. A model for the devolatilization of a coal particle
sufficiently large to be controlled by heat transfer. Combust Flame
2006;146:55371.
135
[39] Liu X, Wang G, Pan G, Wen Z. Numerical analysis of heat transfer and volatile
evolution of coal particle. Fuel 2013;106:66773.
[40] Adesanya BA, Pham HN. Mathematical modelling of devolatilization of large
coal particles in a convective environment. Fuel 1995;74:896902.
[41] Shi L, Liu Q, Guo X, Wu W, Liu Z. Pyrolysis behavior and bonding information of
coal a TGA study. Fuel Process Technol 2013;108:12532.
[42] Butala SJM, Medina JC, Taylor TQ, Bartholomew CH, Lee ML. Mechanisms and
kinetics of reactions leading to natural gas formation during coal maturation.
Energy Fuel 2000;14:23559.
[43] Mae K, Maki T, Miura K. A new method for estimating the cross-linking
reaction during the pyrolysis of brown coal. J Chem Eng Jap 2002;35:77885.
[44] Porada S. The reactions of formation of selected gas products during coal
pyrolysis. Fuel 2004;83:11916.
[45] van Heek KH, Hodek W. Structure and pyrolysis behaviour of different coals
and relevant model substances. Fuel 1994;73:88696.
[46] Varhegyi G, Bobly B, Jakab E, Chen H. Thermogravimetric study of biomass
pyrolysis kinetics: a distributed activation energy model with prediction tests.
Energy Fuel 2011;25:2432.
[47] Maki T, Takatsuno A, Miura K. Analysis of pyrolysis reactions of various coals
including argonne premium coals using a new distributed activation energy
model. Energy Fuel 1997;11:9727.
[48] Ulloa C, Gordon AL, Garcia X. Distribution of activation energy model applied
to the rapid pyrolysis of coal blends. J Anal Appl Pyrolysis 2004;71:46583.
[49] Gunes M, Gunes S. A direct search method for determination of DAEM kinetic
parameters from nonisothermal TGA data (note). Appl Math Comput
2002;130:61928.
[50] De Caprariis B, De Filippis P, Herce C, Verdone N. Double-gaussian distributed
activation energy model for coal devolatilization. Energy Fuel
2012;26:61539.
[51] Warnecke R. Gasification of biomass: comparison of fixed bed and fluidized
bed gasifier. Biomass Bioenergy 2000;18:48997.
[52] Pielsticker S, Gvert B, Kreitzberg T, Habermehl M, Hatzfeld O, Kneer R.
Simultaneous investigation into the yields of 22 pyrolysis gases from coal and
biomass in a small-scale fluidized bed reactor. Fuel 2017;190:42034.