You are on page 1of 15

Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

Contents lists available at ScienceDirect

Journal of Analytical and Applied Pyrolysis


journal homepage: www.elsevier.com/locate/jaap

Modeling biomass particle pyrolysis with temperature-dependent heat of


reactions
Y. Haseli , J.A. van Oijen, L.P.H. de Goey
Combustion Technology, Department of Mechanical Engineering, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands

a r t i c l e

i n f o

Article history:
Received 14 August 2010
Accepted 29 November 2010
Available online 10 December 2010
Keywords:
Biomass particle
Thermal degradation
Pyrolysis model
Heat of reaction
Kinetic constants

a b s t r a c t
An accurate formulation of energy conservation to model pyrolysis of a biomass particle needs to account
for variations in the heat of reaction with temperature, usually neglected in most past studies. It is shown
that by including this effect in a pyrolysis model with three parallel reactions yielding char, gas and tar, a
wide range of experimental data can be accurately predicted. In particular, through comprehensive comparisons of the simulation results with various measurements, a consistent and single value of 25 kJ/kg
is obtained for enthalpy of pyrolysis, which represents a lumped heat of volatiles and char formation at
a reference temperature. It is found that the kinetic parameters of Chan et al. [W.C. Chan, M. Kelbon, B.B.
Krieger, Fuel 64 (1985) 15051513] and Thurner and Mann [F. Thurner, U. Mann, Ind. Eng. Chem. Process
Des. Dev. 20 (1981) 482488] provide reasonable agreement between the model predictions and experiments compared to other reported kinetic constants. These comparisons also indicate that inclusion of tar
cracking reactions to yield additional light gases does not give a better prediction of the process parameters. The presented thermo-kinetic model is capable of successfully predicting various experimental
observations such as the internal temperature peak reported in some past studies. It is shown that the
sensible heat released due to the conversion of virgin biomass to the reaction products is responsible
for this phenomenon. Simulation results reveal that a temperature peak at an internal location of the
particle may occur when the corresponding local temperature reaches the particle surface temperature
while the local biomass conversion is not nalized yet.
2010 Elsevier B.V. All rights reserved.

1. Introduction
Increasing CO2 emissions and uncertainties related to the ultimate availability of fossil fuels are major concerns which push
industry and R&D sectors to seek solutions and new concepts to
overcome these issues. Fossil fuel-based power plants, in particular coal red power plants, contribute signicantly to greenhouse
gas emissions. A cost-effective CO2 reduction method is co-ring
of a high fraction of biomass in coal red power plants. Because of
the existing infrastructure of coal red power stations, the extra
investment costs for co-ring biomass takes advantage of the high
efciencies obtainable in large coal red power plants. Therefore, it
is necessary to increase our understanding and predictive capabilities of biomass combustion with respect to emissions, fuel ignition,
burnout and ash quality.
Combustion of a biomass particle is quite complex as it
undergoes various physical and chemical successive processes
including heating up, drying (in case of a wet particle), devolatilization/pyrolysis and char burnout. However, depending on the

Corresponding author.
E-mail address: y.haseli@tue.nl (Y. Haseli).
0165-2370/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jaap.2010.11.006

heating conditions, particle size and composition, moisture content, etc., some of the above mentioned processes may occur
simultaneously. As concluded by Williams et al. [1], accurate
modeling of biomass combustion with quantitative reliability still
remains a challenge.
The thermal characteristics of the pyrolysis process as one of
the unavoidable steps during thermal decomposition of a biomass
particle need to be carefully investigated at combustion conditions;
even though this phenomenon has been previously studied theoretically and experimentally by many researchers, e.g. Di Blasi [2,3],
Gronli and Melaaen [4], Bharadwaj et al. [5] and Larfeldt et al. [6].
During biomass pyrolysis, several physical and chemical processes
take place including virgin biomass heating up, moisture evaporation and transportation, kinetics involving the decomposition of
biomass to tar, char and light gases, heat and mass transfer, pressure build-up within the porous medium of the solid, convective
and diffusive gas phase ow, variation of thermo-physical properties with temperature and composition, and change in particle size,
i.e. shrinkage.
The detailed models available in the literature for biomass
pyrolysis are based on coupled time-dependent conservation equations including kinetics of the biomass decomposition. In fact, the
kinetic model directly inuences the conservation equations. A lit-

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

Nomenclature
A
cP
E
e
h
hf
h
K
k
k*
MW
n
P
Q
qx
R
Rg
r
T
t
u
YC

pre-exponential or frequency factor, 1/s


specic heat, J/kg K
activation energy, kJ/mol
surface emissivity
convective heat transfer coefcient, W/cm2 K
enthalpy of formation, kJ/kg
total enthalpy, kJ/kg
permeability, cm2
reaction rate, 1/s
effective thermal conductivity, W/cm K
molecular weight, mol/g
Shape factor
pressure, Pa
source term in energy equation, J/cm3 s
external heat ux, W/cm2
particle radius, cm
universal gas constant, kJ/mol K
radial (sphere and cylinder), axial (slab) distance, cm
temperature, K
time, s
supercial gas velocity, cm/s
fraction of char in the products

Greek letters

porosity

density, g/cm3
hBg
enthalpy of B G and B T reactions, kJ/kg
hBC enthalpy of B C reaction, kJ/kg
hTg
enthalpy of tar cracking reaction, kJ/kg

viscosity, kg/m s

StephanBoltzmann constant
Subscripts
0
initial condition

surrounding condition
B
biomass
C
char
G
light gases
g
gas phase (tar + light gases)
T
tar

erature survey indicates that several kinetic schemes have been


proposed and applied by different authors. The one-step global
model, as the simplest kinetic model, considers decomposition
of biomass into char and volatiles. This is the most frequently
applied model, see for instance Galgano and Di Blasi [7]. An
improved version of the one-step model with a single rate constant is a model according to which the main constituents of
wood (cellulose, hemicelluloses and lignin) decompose independently into char and volatiles. Another decomposition scheme is the
well-known BroidoShazadeh model for cellulose decomposition,
which assumes that the formation of an intermediate phase is followed by two competing reactions; in one reaction tar is produced,
while in the other one char and light gases are formed. The proposed mechanism of Koufopanos et al. [8] is similar to the kinetic
scheme of BroidoShazadeh, in which the virgin biomass is rst
converted into an intermediate material (reaction 1) which then
decomposes to gases and volatiles (reaction 2), and char (reaction
3). Shazadeh and Chin [9] proposed a primary wood degradation
mechanism which suggests three individual competitive reactions
forming light gases, tar and char.

141

Another kinetic model of biomass degradation assumes that in


addition to the primary reactions as suggested by Shazadeh and
Chin [9], tar undergoes homogeneous degradation producing additional light hydrocarbons and char. This is referred to as tar cracking
or secondary reactions. This model was applied in detailed pyrolysis simulations conducted by Gronli and Melaaen [4], Di Blasi [2,10],
Hagge and Bryden [11], Bryden and Hagge [12], Bryden et al. [13],
and Chan et al. [14]. Moreover, Koufopanos et al. [15] took into
account the nature of secondary reactions from a different viewpoint. In their model, virgin biomass undergoes primary reactions
to decompose into volatile and gases (reaction 1) and char (reaction 2). The primary pyrolysis products participate in secondary
reactions to produce also volatile, gases and char of different compositions (reaction 3). The kinetic mechanism of Koufopanos et al.
[15] has been applied, for instance, by Babu and Chaurasia [16] and
Sadhukhan [17,18] to model pyrolysis of biomass particles. In the
recent comprehensive review studies by Di Blasi [19,20], the chemical kinetics of wood and biomass pyrolysis are critically discussed.
The reader is also referred to the interesting paper by Di Blasi [21],
in which the predictions of several kinetic mechanisms, which take
into account the formation of char, tar and gas, are compared and
discussed for the primary pyrolysis of cellulose and biomass.
Literature review indicates that there are also discrepancies in
the reported kinetics and thermo-physical data applied in different theoretical investigations for predicting thermal degradation
of a biomass particle. We have found ve different data sets for
activation energies and frequency factors of the primary reactions
of the Shazadeh and Chin [9] kinetic scheme, which are usually
described as rst-order with an Arrhenius type of temperature
dependence. Moreover, the heat of these primary reactions is one of
the most important parameters inuencing the pyrolysis process,
and it has been assigned various values. Hence, it remains a question for a pyrolysis modeler what set of kinetic constants and which
value for the heat of reactions must be utilized in the simulation.
Our attempt is to nd reasonable answers for these uncertainties
in this paper.
On the other hand, discrepancies have also been found in
measurements and experimental observations reported in various
sources. For instance, the particle inner temperature showed to
exceed the surface temperature of the solid particle before reaching thermal equilibrium in the experimental studies of Koufopanos
et al. [15], Park et al. [22] and Di Blasi et al. [23], whereas this phenomenon was not observed in measurements of other workers such
as Larfeldt et al. [6], Chan et al. [14], Tan and White [24], and Lu [25].
In the current paper, assuming that the biomass decomposition
takes place according to the Shazadeh and Chin scheme, we
investigate the accuracy of the pyrolysis models when using various kinetic parameters reported in the literature. The possibility of
a tar cracking reaction to produce lighter gases is also examined.
Moreover, it is intended to highlight advantages of an accurate
formulation of the conservation of energy that allows computing
the enthalpy of pyrolysis as a function of temperature. It will be
shown that a pyrolysis model with the accurate energy equation is
capable of predicting a wide range of experimental data including
temperature peaks at internal positions observed in some past
studies; e.g. [15].
2. Literature survey on kinetic parameters
The kinetic scheme utilized in our simulations assumes that
the biomass particle decomposes to char, tar and gas according to
the model of Shazadeh and Chin [9]. Hence, the relevant kinetic
constants that have been obtained experimentally by various past
researchers are summarized. The accuracy of the particle pyrolysis
model will be examined when using different sets of kinetic parameters. Moreover, this section provides data of the kinetic parameters

142

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154


Table 4
Kinetic data of Font et al. [28].

Table 1
Kinetic data of Chan et al. [14].
Reaction

Frequency factor (1/s)

Activation energy (kJ/mole)

Test Case

Reaction

Bio Gas
Bio Tar
Bio Char

1.30 10
2.00 108
1.08 107

140
133
121

Frequency factor
(1/s)

Activation energy
(kJ/mole)

Pyroprobe 100

Bio Gas
Bio Tar
Bio Char
Bio Gas
Bio Tar
Bio Char

1.52 107
5.85 106
2.98 103
6.80 108
8.23 108
2.91 102

139.3
119
73.4
155.6
148.5
61.4

Fluidized bed reactor


Table 2
Kinetic data of Di Blasi and Branca [26].
Reaction

Frequency factor (1/s)

Activation energy (kJ/mol)

Bio Gas
Bio Tar
Bio Char

4.38 109
1.08 1010
3.27 106

152.7
148.0
111.7

of tar cracking into light gases, which were obtained by a number


of researchers. Also, a large scatter in the pyrolysis heat is reported
in the literature which will be discussed in Section 2.3.
2.1. Primary reactions
Table 1 gives the kinetic data of primary reactions as reported
by Chan et al. [14] who compared their pyrolysis model predictions with the experimental results from lodgepole pine wood
devolatilization. These parameters were employed in the pyrolysis
model of Gronli and Melaaen [4]. The kinetic mechanism applied in
their simulation of spruce wood pyrolysis was that of three primary
reactions with tar cracking producing merely additional gas.
Di Blasi and Branca [26] conducted experiments on beech wood
to determine kinetic constants of wood pyrolysis. The weight loss
of thin layers of beech wood powder (150 m) were measured
for heating rates of 1000 K/min with reaction temperatures in the
range of 587720 K. Their ndings are given in Table 2. These data
were used by Park et al. [22] for simulation of maple wood pyrolysis.
The third set of kinetic constants is of that reported by Thurner
and Mann [27] who studied the pyrolysis of oak sawdust in the range
of 573673 K at atmospheric pressure. Table 3 lists kinetic constants of the primary reactions obtained from their experiments.
Hagge and Bryden [11], Bryden and Hagge [12], and Bryden et al.
[13] used these data to validate their pyrolysis model against experiments of Tran and White [24], who measured temperature history
and char yield of redwood, southern pine, red oak and basswood at
constant radiant heat ux.
The fourth and last set of kinetic constants is the data set of Font
et al. [28], obtained based on a comprehensive experimental study
using a uidized bed reactor and a Pyroprobe 100 to investigate
the kinetics of the ash pyrolysis of almond shells and of almond
shells impregnated with CoCl2 . Experiments were conducted at
673733 K to study kinetics of the almond shells in a uidized bed.
For the kinetic study in the Pyroprobe 100 equipment, experiments
were carried out at 673878 K. Table 4 summarizes the ndings of
Font et al. [28]. Despite the apparent discrepancies in the reported
kinetic parameters obtained from two test facilities, the predicted
decomposition rate of biomass to individual products, i.e. gases, tar
and char, are approximately the same in both cases. The rst set of
kinetic constants was, for example, applied by Lu [25] for modeling
of hardwood sawdust particles pyrolysis.

Table 3
Kinetic data of Thurner and Mann [27].

Di Blasi [10] examined data of Chan et al. [14], Thurner and Mann
[27], and (the rst set of) Font et al. [28]. A good quantitative agreement was obtained between the experimental data of Lee et al.
[29] and the predicted temperature proles along the degrading
biomass particle using kinetic parameters of Thurner and Mann
[27]. However, the author believed that such an agreement was
reached mainly due to the properties of char and biomass used in
the numerical simulation, which were those measured in experiments of Lee et al. Thus, no nal conclusion was made in Di Blasis
study concerning the importance of thermo-kinetic data for quantitative predictions.
We compared the rates constants of three parallel reactions
predicted using the kinetic constants of Chan et al. [14], Di
Blasi and Branca [26], Thurner and Mann [27], and Font et al.
[28] as functions of temperature in the range of 400900 K. The
results of these observations can be classied in three temperature ranges; 400 < T < 550 K, 550 < T < 700 K, and 700 < T < 900 K, as
analysed below.
400 < T < 550 K: The rate constant of light gases calculated by
Thurner and Manns data is higher than those obtained from other
three sets. The biomass decomposition rate to produce tar is almost
the same in all cases. For the char yield, the data of Font et al. gives
higher rates compared to the other cases.
550 < T < 700 K: Di Blasi and Brancas data overestimate the production rates of light gases compared to the other data, whereas the
other three sets of kinetic constants give almost the same results.
Again, the rate of tar production computed by Di Blasi and Brancas
data are higher than those predicted by data of the other three
groups. All data sets are expected to give approximately the same
results for rate constant of biomass decomposition to char.
700 < T < 900 K: Di Blasi and Brancas data predicts faster formation of light gases compared to other three data sets. The prediction
using the Chan et al. and Font et al. data are almost in the same
range. Again, rate constants for tar production computed by Di Blasi
and Brancas data are higher than those predicted by data of other
three groups. Computed reaction rate constants of char formation
using Chan et al. and Di Blasi and Brancas data are in the same
range and higher than those computed by other two groups which
produce compatible results.
Given that several factors such as experimental set up,
experiment conditions (high heat ux, low heat ux), biomass
composition may contribute to the reasons of these discrepancies
between various kinetic data, the decision on which set of kinetic
constants must be used in a pyrolysis model, depends largely
upon how well the thermo-kinetic model will predict experimental
observations. The accuracy of the pyrolysis model using the kinetic
parameters suggested by Chan et al., Di Blasi and Branca, Thurner
and Mann, and Font et al. will be examined in Section 4.
2.2. Tar cracking reactions

Reaction

Frequency factor (1/s)

Activation energy (kJ/mol)

Bio Gas
Bio Tar
Bio Char

1.44 104
4.13 106
7.38 105

88.6
112.7
106.5

Vapor phase tar decomposition known as secondary reactions,


was investigated by a number of researchers; Liden et al. [30],
Boroson et al. [31], Kosstrin [32] and Diebold [33]. Tar cracking
was modeled as a homogenous process to give mainly light gases,

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154


Table 5
Kinetic parameters of tar cracking reported in the literature.
Frequency factor
(1/s)

Activation energy
(kJ/mol)

Temperature range
(C)

Reference

Tar light gas


4.28 106
1.00 105
3.26 104
1.00 105

107.5
93.3
72.8
87.5

460600
500800
430900
>600

[30]
[31]
[32]
[33]

an assumption which is supported by experimental results [30].


Table 5 summarizes the ndings of these researchers on the kinetic
constants of tar cracking reaction.
2.3. Heat of pyrolysis
In nearly all past modeling studies of single particle biomass
pyrolysis, decomposition of the virgin biomass to gas, tar and char
is assumed to take place through a series of endothermic reactions,
whereas tar cracking through secondary reactions for additional
production of gas and char is considered exothermic. Furthermore, it
is commonly assumed that the heat of all three primary reactions is
identical, and so is the one for the secondary tar cracking reactions.
Literature review indicates a large scatter in the reported values
for the heat of pyrolysis. A survey performed by Milosavljevic et al.
[34] revealed that it may be in the range of 2100 to 2500 kJ/kg.
Given the difculty of measuring the pyrolysis heat, most of past
researchers treated it as an adjustable parameter which would give
reasonable agreement between the results of simulation and measurements (e.g. a pyrolysis heat of 418 kJ/kg was suggested by Chan
et al. [14]; Gronli [35] assumed a heat of 150 kJ/kg for primary
reactions; Park [36] estimated an endothermic heat of 64 kJ/kg
for three parallel reactions). In contrast, there are limited studies
which have intended to experimentally (with no aid of parameter tting in the simulating model) determine the pyrolysis heat.
Havens et al. [37] showed that, based on experiments of differential
scanning calorimery (DSC), the heat of pyrolysis for pine and oak is
200 kJ/kg and 110 kJ/kg, respectively. In another effort to measure
the pyrolysis heat of Pinus Pinaster thermal degradation using DSC,
Bilbao et al. [38] reported an endothermic heat of 274 kJ/kg up to
a conversion of 60%, whereas for the remaining conversion of the
biomass, the process was observed to be exothermic with a heat of
353 kJ/kg.
An extensive investigation to determine the heat of pyrolysis of dried cylindrical maple particles with 2 cm diameter and
8 cm length was conducted by Lee et al. [29] at two heat uxes
of 30 kW/m2 and 84 kW/m2 . Their results showed that at a low
heat ux the pyrolysis layer could be divided into three zones: an

143

endothermic primary decomposition zone at temperatures up to


250 C, an exothermic partial char zone between 250 C and 340 C,
and an endothermic surface char zone at 340 C < T < 520 C. The
overall mass weighted heat of reaction was endothermic to the
extent of 610 kJ/kg. In contrast, the overall heat of reaction at the
higher heat ux was exothermic, being greater for perpendicular
heating (in the range of 1090 to 1720 kJ/kg) than for parallel
heating (in the range of 105 to 395 kJ/kg). The authors arrived at
the conclusion that the heat of pyrolysis depends upon the external heating rate, total heating time and anisotropic properties of
biomass and char relative to the internal ow of heat and gas.
Milosavljevic et al. [34] studied the thermo-chemistry of cellulose pyrolysis by a combination of DSC and thermogravimetric
analysis. They found that the main thermal degradation pathway
was endothermic in the absence of mass transfer limitations that
promoted char formation. They concluded that the endothermicity, which was estimated to be about 538 kJ/kg of volatiles evolved,
was mainly due to latent heat requirement for vaporizing the primary tar decomposition products. It was also reported that the
pyrolysis could be driven in the exothermic direction by char forming processes that would compete with tar forming processes. The
formation of char, which was favored at low heating rates, was estimated to be exothermic to the extent of 2000 kJ/kg of char formed.
The authors arrived at an interesting conclusion that the heat of
pyrolysis could be correlated with the char yield at the end of
pyrolysis, a result which was somewhat consistent with ndings
of Mok and Antal [39], who also observed a linear decrease in the
endothermic heat of pyrolysis as the char yield increased. Hence, it
was concluded that the char yield was the main factor determining whether the overall pyroysis process is endo- or exothermic.
Table 6 summaries the above discussed values of the pyrolysis heat
which have been obtained by different researchers either through
a tting procedure or based on measurements.
From these works, it can be now understood why a wide range
of pyrolysis heat has been reported in the literature. For the purpose of simulating thermal degradation of a biomass particle, it is
of importance that the pyrolysis heat must be chosen from similar
heating and comparable measurement conditions (such as similar char yield) if it is taken from a referred source. Alternatively, it
should be carefully correlated by comparing the simulation results
with measurements, otherwise the predicted parameters of the
pyrolysis model may not be reliable.
3. Pyrolysis model
The pyrolysis of biomass is a complex phenomenon. In order
to accurately describe particle conversion, a deeper understanding of the chemistry and physics of the process is necessary.

Table 6
Values of heat of pyrolysis reported in various studies.
Enthalpy of pyrolysis (kJ/kg)

Method

418
150
64
255
20
200
110
274
353
610
1090 to 1720 (prependicular heating)
105 to 395 (parallel heating)
5382000YC **

Fitting with experiments


Fitting with experiments
Fitting with experiments
Fitting with experiments
Fitting with experiments
Measured
Measured
Measured
Measured
Measured
Measured
Measured
Measured

*
**

Converion value
Final char yeild

Remarks

X* < 0.95
X > 0.95
Pine wood
Oak wood
X < 0.60
X > 0.60
Low heat ux
High heat ux
High heat ux
Cellulose

Source
Chan et al. [14]
Gronli [35]
Park [36]
Koufopanos et al. [15]
Havens et al. [37]
Bilbao et al. [38]
Lee et al. [29]

Milosavljevic et al. [34]

144

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

The one-dimensional (1D) thermo-kinetic biomass particle model


accounting for primary and secondary reactions to describe the
pyrolysis process that we employed (assuming a constant particle
volume throughout the conversion process), has previously been
presented in several past works; e.g. Di Blasi [2], Gronli and Melaaen
[4], Park et al. [22], and Lu et al. [40]. However, for the sake of completeness of our article, the basic equations of the model will be
briey described with a particular focus on accurate formulation of
the conservation of energy.
The kinetic mechanism employed in our simulation of single
particle biomass pyrolysis accounts for three parallel primary reactions with the possibility of a tar cracking reaction; that is, the virgin
biomass decomposes to gas, tar and char, and vapor tar decomposes
to yield further gas. As concluded in past experimental studies
[30,31,33], the main product of the tar cracking reaction is light
gases, and the amount of char yield is negligible.

where n denotes a shape factor (n = 0 for at, n = 1 for a cylinder,


and n = 2 for a sphere) and represents the particle porosity. During
thermal decomposition of biomass, particle porosity increases with
time and it may have different values along the spatial coordinate r.
3.2. Conservation of energy
The main assumptions for the formulation of energy conservation related to a solid biomass particle undergoing thermal
degradation are that the volume of the particle remains constant
during the process, and the solid and the gas phase are in thermal equilibrium. A proper description of the energy equation is
that it must account for (1) accumulation of energy, (2) conductive heat transfer through the particle, (3) convective heat transfer
due to ow of volatiles through pores. As the problem of study
includes the chemical structural changes through various reactions,
the enthalpy of each species must be represented as the summation
of sensible and formation enthalpies:

h i

1 n
1
(r ug h g ) = n
+ n
r r
r r

n T

r k


i = B, C, g

(6)

Despite these later studies that report possibility of cracking


tar into lighter gases at elevated temperatures, however, there is
no experimental evidence supporting that this reaction also takes
place during thermal conversion of a biomass particle through a
separate path. To our best knowledge, tar cracking reactions in the
form of secondary reactions were rst adapted by Di Blasi [2] in
a biomass pyrolysis model, who also assumed that char could be
formed as a consequence of tar cracking. Since then, some authors
have assumed the secondary reactions in their pyrolysis models,
whereas some others have argued that the possibility of this reaction directly depends on the residence time of the volatiles, so that
it does not necessarily take place as in the pyrolysis process. In a
recent study by Park et al. [22], it has been reported that accounting
for the secondary reactions is of minor importance. As a further task
of the present article, it is intended to nd out whether predictions
of the pyrolysis model with three parallel reactions are sufcient
to observe the experiments, or inclusion of a tar cracking reaction
in the model may provide better predictions.
The reaction rates are determined through an Arrhenius type
equation.

ki = Ai exp

Ei
Rg T

i = 1, 2, 3, 4

(1)

where h is the total enthalpy, and k* is the effective thermal conductivity which accounts for thermal conductivity of the biomass,
char, gaseous byproducts as well as radiation heat transfer inside
the pores (using the Rossland diffusion approximation for a thick
medium). Some mathematical manipulations are required in order
to represent the energy equation in terms of temperature as
described in Appendix A, so the nal form of Eq. (6) reads
(B cPB + C cPC + g cPg )

1
T
T
+ ug cPg
= n
r r
t
r

where

r n k

T
r

Q = B (k1 + k2 )[hBg +

T k4 [hT G +

+ Q


(cB cg )dT ] + B k3 [hBC +

(cB cC )dT ]+

(8)
(cT cG )dT ]

where hBg , hBC , hTG are, respectively, enthalpies of B g,


B C and T G reactions at a reference temperature. Notice that
hBg accounts for the enthalpies of both B G and B T reactions
according to the assumed kinetic scheme. The above formulation
suggests that the heat of reactions involved in the pyrolysis process
need to be calculated as a function of temperature. However, in
many past studies, Q is dened as
Q = B (k1 + k2 + k3 )hP + T k4 hS

where A is the frequency factor, E the activation energy, Rg the


universal gas constant, and T the temperature.
3.1. Conservation of species mass

hP = (1 YC )[hBg +

B
= (k1 + k2 + k3 )B
t

(2)

hS = hT G +

C
= k3 B
t

(3)

T
1 n
+ n
(r T u) = k2 B k4 T
r r
t

(4)

g
1 n
+ n
(r g u) = (k1 + k2 )B
r r
t

(5)

(9)

where it is assumed that the heat of all three primary reactions are
identical and constant (hP ), and the heat of secondary reaction is
represented by hS .
By comparing Eq. (8) with Eq. (9), we nd

The consumption of biomass (B) and the formation of char (C)


can be described by the following equations:

Mass conservation equations of the tar (T) and total gas phase
(g) obey

(7)

(cB cg )dT ] + YC [hBC +

(cB cC )dT ]

(10)


(cT cG )dT

(11)

In Eq. (10), YC denotes the fraction of char in the pyrolysis products. Hence, in these studies the possibility of evaluating the heat
of pyrolysis at different temperatures and as a function of products
yields is taken away. The main issue with the inaccurate formulation of the source term, Eq. (9), in the energy equation is that an
endothermic heat is usually assumed for all three primary reactions as in many past studies, e.g. 150 kJ/kg [4], 418 kJ/kg [14],
and 64 kJ/kg [15]. These values have been obtained by tting the
predictions (usually mass loss or temperature histories) with the
experiments in each individual study through employing Eq. (9)

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154


Table 7
Thermo-physical properties used in biomass pyrolysis simulation.
Property

Value/correlation

Specic heat (J/g K)

cpB = 1.5 + 1.0 103 T


cpC = 0.44 + (2 103 )T 6.7 107 T2
cpT = 0.162 + (4.6 103 )T 2 106 T2
cpG = 0.761 + (7 104 )T (2 107 )T2
cpg = (T /g )cpT + (1 T /g )cpG
k* = (B /B0 )kB + (1 B /B0 )kC + kg + 13.5T3 d/

Thermal
conductivity
(W/cm K)

Porosity
Pore diameter (cm)

Permeability
(cm2 )

Gas phase viscosity


(kg/m s)

the biomass to the volatiles and char. Further demonstration of this


idea will be presented in Section 4.2.
3.3. Conservation of momentum
Darcy law is applied to describe gas phase momentum transfer within the porous media. Hence, the supercial gas velocity is
obtained from
u=

kg = 0.00026
kc = 0.001 (grain)
kc = 0.007 (radial)
= 1 (1 0 )(B + C )/B0
0 : biomass dependent
d = (B / B0 )dB + (1 B /B0 )dC
dB = 5 103
dC = 102
K = (B/ B0 )KB + (1 B/ B0 )KC
KB = 5 1012
KC = 109
g = 3 105

and without accounting for any secondary reactions. In fact, these


studies suggest that heat input is a requirement for the entire
process of biomass decomposition, which is not consistent with
experimental observations of Lee et al. [29], Bilbao et al. [37] and
Milosavljevic et al. [34] as discussed in Section 2.3. The question
that may be raised is that, would not it be better to employ the
accurate version of the source term in the energy equation as represented in Eq. (8), and try to nd enthalpy of reactions at a reference
temperature through the same method of tting of predictions with
experiments? We will further follow this idea in Section 4 to examine whether with this proposed method we may still get a variety
of reactions heats, or it may be possible to end up with a single
value that can be used in the pyrolysis model to reasonably capture
various experimental observations.
It can be inferred from Eq. (8) that the reaction heats associated with the formation of volatiles, char and possibly secondary
gases become more exothermic as the temperature increases. This
is because the specic heat of biomass is usually greater than that of
char and that of volatiles, as well as the specic heat of tar is greater
than that of light gases (see the related correlations that are given
in Table 7). Moreover, Eq. (8) shows that the heat of volatiles release
is not necessarily the same as the heat of char formation. The question that may arise is what would be the physical explanation
of

the terms associated with various specic heats, i.e. (ci cj )dT, as
appeared in Eq. (8)? This term represents the amount of sensible
heat released when species i decomposes to species j at a temperature different from the reference one. Alike any chemical reaction
in which a certain amount of energy is released/consumed at a
given temperature, it is possible to estimate the amount of energy
(enthalpy of reaction) at another temperature by simply algebraic
summation of the enthalpy of reaction at the reference temperature, and the enthalpy that is equivalent to the difference in the
sensible heats of the products and the reactants.
As a conclusion from this discussion, it is possible that the
enthalpy of pyrolysis may become exothermic at some stages of
the pyrolysis depending on the process conditions, as also conrmed in some studies. In such cases, a pyrolysis model that uses Eq.
(9) with constant endothermic heat for the primary reactions will
fail to accurately predict the temperature; and as a consequence
the chemistry of the process will be inuenced. On the contrary, a
model which employs Eq. (8) does not suffer from this issue, since
it has a potential to capture exothermicity of the pyrolysis heat by
accounting for the sensible heat released due to the conversion of

145

K p
 r

(12)

Furthermore, it is assumed that volatiles behave like a perfect


gas, so that internal pressure is determined from
p=

g Rg T
MWg

(13)

3.4. Initial and boundary conditions


The numerical solution of the transport equations described
above needs to dene ve initial and four boundary conditions.
Initial conditions (t = 0)

B = B0

C = 0
T = 0

(14)

g = inert
T = T0

Boundary conditions (t > 0)

r = 0

T

=0

r
g

=0

T = 0
r 

r = R k r =

(15)
4
h(T T ) + e(T

T 4)

or
4)
qx h(T T ) e(T 4 T

Two different types of boundary conditions may be applied to


dene the surface temperature. The rst one is used for a known
reactor temperature so that heat is transferred to the surface of the
particle through radiation and convection. The second boundary
condition is suitable when the particle is exposed to a known heating ux qx so that some heat is dissipated from the surface to the
environment via radiation and convection. Moreover, in the simulation code the velocity is initially set to zero as well as the initial
pressure and pressure at the surface are equal to the surrounding
pressure (usually atmospheric).
3.5. Simulation code
The pyrolysis model described above is implemented in
CHEM1D to study thermal conversion of a dry biomass particle.
CHEM1D [41] is a computer code developed at the Combustion
Technology Group of the Department of Mechanical Engineering
at Eindhoven University of Technology to calculate various ame
types. It is capable of solving a set of general time-dependent 1D
differential equations which includes accumulation, convection,
diffusion and source terms, on the basis of the control volume discretisation method for specied initial and boundary conditions
and known time and space domains. It uses adaptive gridding and
time stepping techniques.
Simultaneous solution of the transport and kinetic equations
requires dening the thermo-physical properties and kinetic constants discussed previously. Composition-dependence of thermal
conductivity, specic heat and solid phase permeability is taken

146

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

Table 8
Comparison of the predicted and measured biomass conversion and char yield, with experimental data taken from Gronli and Melaaen [4] (T&M: Thurnner and Mann; C:
Chan et al.; D&B: Di Blasi and Branca; F: Font et al.).
Duration

Kinetic data

5 min
10 min
5 min
10 min
5 min
10 min
5 min
10 min

T&M

Measured

C
D&B
F

Char yield (kg/m3 )

Converted biomass (%wt)

25.7
45.5
25.7
45.5
25.7
45.5
25.7
45.5

1.63
3.61
1.63
3.61
1.63
3.61
1.63
3.61

Predicted

%

Measured

Predicted

%

25.1
41.7
24.3
39.9
25.6
41.8
26
44.7

2.3
8.4
5.4
12.3
0.4
8.1
1.2
1.8

30.4
58.8
30.4
58.8
30.4
58.8
30.4
58.8

33.4
56
29.5
50.4
19
35.3
62.7
120.5

9.9
4.8
3.0
14.3
37.5
40.0
106.3
104.9

into consideration. A survey on thermo-physical properties allowed


us to take a xed value/correlation for most of the parameters,
except the thermal conductivity, density and initial porosity of
biomass, and the convective heat transfer coefcient h, which may
vary depending upon the type of biomass and process conditions.
Table 7 lists the required data and some auxiliary equations which
are included in the simulation code. The presented relationships
for the specic heats of char, tar and gas are based on correlated
data of Raznjevic [42].
4. Results and discussion
The accuracy of the developed pyrolysis model and the reliability of the four different kinetic data are examined by comparing the
predictions with experimental data taken from Gronli and Melaaen
[4], Koufopanos et al. [15] and Rath et al. [43]. The idea is to nd
out which set of kinetic parameters given in Table 1 through Table 4
may provide a better insight into the thermal degradation of a single
biomass particle. Notice that an efcient model must be capable of
reasonably predicting both heat transfer and kinetics of the process;
i.e. mere validation of heat transfer or mass transport parameters is
not sufcient. This has been taken into consideration in only a few
studies [4,13,15,18,44]. The second idea is to nd what value can be
obtained for hBg and hBC (assuming hBg = hBC ) at various experimental conditions when Eq. (8) is used as the source term
in the energy equation. Also, we intend to examine whether the
kinetic scheme with three parallel reactions is sufcient enough to
capture the experimental observations, or inclusion of a tar cracking reaction to yield further light gases in the model could provide
better predictions of the thermo-kinetics of the pyrolysis process.
4.1. Comparison with data of Gronli and Melaaen
In the experimental work of Gronli and Melaaen [4], birch,
pine, and spruce particles were one-dimensionally heated in a bellshaped glass reactor using a xenon arc lamp as a radiant heat source.
The total times of exposure (heating times) were 5 and 10 min. For
validation of their pyrolysis model, which did not account for particle shrinkage, experimental results of spruce heated parallel with
the grain were chosen which had shown the lowest axial shrinkage at both low and high heat uxes compared to pine and birch
particles.
Fig. 1 shows a comparison between the predicted (obtained from
our simulation code) and measured temperature proles at ve
axial positions 0.6, 1.8, 2.2, 2.6 and 3 cm for an external heat ux of
80 kW/m2 . A good agreement between the predicted and the measured temperatures can be seen using all sets of kinetic parameters.
This agreement was obtained by tting the predictions with experimental data for a weighted endothermic reaction heat of 25 kJ/kg;
i.e. hBg = hBC = 25 kJ/kg, without accounting for tar cracking
reaction; i.e. k4 = 0. When a possibility of secondary reaction of
tar was taken into consideration, no reasonable match between

the measured and the predicted proles was observed, given that
variety of heat of primary reactions was tested.
As mentioned previously, mere validation against heat transfer
parameters is not sufcient since the process also involves kinetics
and mass transport of various species. The biomass conversion fraction and char yield were also calculated using different kinetic data.
In Table 8, the predicted and measured values of these parameters
are compared for 5 min and 10 min as the duration of pyrolysis. The conversion of biomass is well-predicted using all kinetic
data sets at 5 min heating duration. The converted biomass is still
reasonably predicted for the 10 min heating condition with most
under-prediction attributed to the data of Chan et al. Nevertheless,
looking at the predicted char yields, it is seen that the predictions
using the data of Thurner and Mann and the data of Chan et al. are
much better than those obtained using the data of Font et al. and
the data of Di Blasi and Branca. Kinetic constants of Font et al. give a
signicant over-prediction of the char yields to the extent of 106%.
On the other hand, data of Di Blasi and Branca leads to a notable
under-prediction of char yields to the extent of 40%.
Figs. 2 and 3 depict typical simulated time and space evolution of various parameters related to the experimental condition
of Gronli and Melaaen [4]. As the particle with initial temperature
of 300 K is exposed to a high heating ux, the surface temperature begins to rise while some heat is dissipated from the surface
to the surrounding due to radiation and convection heat transfer
mechanisms. The heat received by the surface is transfered into
the particle through conduction heat transfer. As the temperature
of the particle increases, the primary reactions are activated so that
the virgin biomass begins to slightly decompose into three main
groups of byproducts at low temperatures according to the kinetic
model employed. The rates of conversion become higher as the
temperature increases. As the gaseous byproducts are generated by
continuous decomposition of biomass, internal non-uniform pressure slightly greater than atmospheric pressure is built up. This
causes a ow of volatiles through the pores of the particle mainly
towards the surface of the particle where they escape; but there
also exists a ow of gaseous species in the opposite direction which
increases the pressure at the center of the particle. During the initial stages of the process, the pressure peak takes place near the
surface; however, as the process continues the location of the maximum pressure is shifted to the internal positions. This is attributed
to the higher rate of volatiles generated compared to their local
velocity. The ux of gaseous ow is dependent on the local permeability so that the volatiles mass ux is higher in the charred part
of the particle than in the virgin part since the permeability of char
is higher than that of biomass.
As long as decomposition of the particle surface has not been
nalized, the particle can be divided into two regions: a partial char
region where still there is some biomass to decompose, and a virgin
biomass region. In this stage of the process, maximum formation
of volatiles (including tar) and char takes place at the surface. Soon
after all virgin material at the surface has converted to byproducts

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

147

Fig. 1. Comparison of predicted (solid lines) and measured (areas between broken lines) temperature proles at ve axial positions using kinetic parameters of (a) Thurner
and Mann [27], (b) Di Blasi and Branca [26], (c) Chan et al. [14], (d) Font et al. [28]. Experimental data are taken from Gronli and Melaaen [4].

a thin layer of char is formed at the surface. As the process of degradation proceeds, the thickness of the charred region increases and
its front moves towards the particle center. By increasing the thickness of the char layer, the location of maximum tar yields is shifted
to the inside of the particle and its magnitude increases. In other
words, the local density of tar increases continuously along the axis
of the particle up to a point where formation of tar stops. Beyond
this position, the density of tar decreases due to the ow of volatiles
towards the surface of the particle.
4.2. Comparison with data of Koufopanos et al.
An isothermal mass-change determination technique was
employed in the work of Koufopanos et al. [15] to measure the
pyrolysis rate of dried wood cylinders initially at room temperature
at reactor temperatures in the range of 573873 K. Temperature
variations at two positions inside the particles were also measured during the pyrolysis process. An interesting observation in
these experiments was that the temperature inside the particle
exceeded the reactor temperature, and a peak was observed; with
a higher magnitude at a lower reactor temperature. To capture this
phenomenon, Koufopanos et al. [15] proposed that certain stages
of the conversion take place endothermically, but the rest of the
process proceeds exothermically. Through tting the model predictions with the experiments, they found an endothermic heat of
255 kJ/kg (up to a conversion of 95%), as well as an exothermic heat
of 20 kJ/kg (for the remaining stages of the particle conversion).
Saduhkhan et al. [17] were also able to model the internal temperature peaks observed in the experiments of Koufopanos et al.

[15] by assuming an exothermic heat of 245 kJ/kg for the entire


of the process in their model. They also employed the same kinetic
scheme of Koufopanos et al. [15]. Interesting to note is that, in a
subsequent publication, Saduhkhan et al. [18] used a different value
for the pyrolysis heat; i.e. 220 kJ/kg, in order to capture their own
experimental data which also included a temperature peak.
In Fig. 4, the predicted center temperature and mass loss are
compared with experiments of Koufopanos et al. at reactor temperatures of 623 K (top graphs) and 773 K (bottom graphs). The
normalized temperature is dened as (T Tr )/(T0 Tr ) with T0 and
Tr representing the initial and reactor temperatures. Simulations
have been performed using all sets of kinetic parameters; however, the results obtained using the data of Font et al. are excluded
in Fig. 4 since the pyrolysis rate was signicantly underestimated;
even though the predicted temperature proles were satisfactory.
In each case, the best match between the predicted temperature
and mass loss proles with the measured data was achieved with
a weighted endothermic heat of 25 kJ/kg; the same value found in
the previous sub-section. These satisfactory agreements between
the simulation results and the experimental data were obtained
without accounting for tar cracking reactions, because inclusion
of additional gas formation through secondary reactions did not
give reasonable results given that various values of heat of pyrolysis were examined. The temperature rise at the early stages of the
process is predicted to be faster than the measured data. One possible reason for this is that the value assumed for the convective heat
transfer rate in our simulations may be different from the real one
which was not reported in Ref. [15]. Furthermore, thermal conductivity of the virgin wood can also contribute to the gap between the

148

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

Fig. 2. Time and space evolution of temperature, volatiles mass ux, internal pressure, biomass density, tar density and char density. Simulated time: 10 min; different lines
correspond to different times; external heat ux: 80 kW/m2 ; horizontal axis: half thickness of particle (3 cm); kinetic data: Thurner and Mann.

predictions and the measurements at the early stages of the process


as a single constant value was chosen in the simulation that may be
higher than the actual value of the biomass thermal conductivity.
Interesting to note is that the presented thermo-kinetic model
is capable of capturing the temperature peak as observed in the
experiments of Koufopanos et al., which means that the three parallel reactions yielding gas, tar and char and the enhanced version
of the heat transfer equation are sufcient enough to predict the
temperature peak at the center of the particle. Critics may argue
how an internal temperature can exceed the surface temperature
while the source of thermal energy is external. As discussed in Section 3.2, the local temperature can be affected by heat released due
to conversion of biomass to products.

Imagine the moment during the pyrolysis process that the center temperature reaches the surface temperature while thermal
degradation of the biomass at the center of the particle has not
yet been nalized. Hence, there is no heat/energy transfer to the
center due to conduction. At this time, the center of the temperature is at some enthalpy level. Since the center temperature is
high enough to cause decomposition of the remaining biomass to
char and volatiles, there is a possibility of a local temperature rise
because of the reduction of the local heat capacity in order to satisfy
the local conservation of energy. In fact, this may occur when the
rate of particle decomposition is higher than the rate of heat transfer. The competition between these two processes continues until
the heat transfer rate becomes dominant and therefore the center

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

149

Fig. 3. Time and space evolution of temperature, volatiles mass ux, internal pressure, biomass density, tar density and char density. Simulated time: 1 min; different lines
correspond to different times; applied external heat ux: 80 kW/m2 ; horizontal axis: 0.5 cm inside the particle; kinetic data: Thurner and Mann.

temperature begins to drop (mainly due to conduction) and reaches


thermal equilibrium; the surface temperature. This hypothesis is no
longer valid if the particle decomposition has already been nalized
before the center temperature (or any internal location temperature) reaches the surface temperature. This will be demonstrated
in Section 4.3.
The above hypothesis is conrmed by the simulated temperature and biomass density proles in the center and at the surface
of the particle as depicted in Fig. 5. It can be seen that the center temperature reaches the surface temperature at around 630 s
after initiation of the heating process. At this moment, over 50%
of the biomass in the center still remains to be decomposed. As
the corresponding local temperature is within the range of primary
reactions, decomposition of the remaining biomass continues with
a rate higher than the local heat transfer rate, which results in a
local temperature rise due to the release of sensible heat by conver-

sion of the virgin material to the byproducts. A careful observation


of the graphs shown in Fig. 5 indicates that the temperature peak
takes place at the moment when the rate of biomass decomposition
becomes slower and it is close to nalize. It seems that when there
is only a small amount of virgin material left, the heat transfer rate
becomes dominant compared to the rate of the biomass decomposition so that the local temperature drops as the remaining biomass
undergoes its nal stage of conversion.
The temperature-peak phenomenon may occur not only at the
center of a pyrolyzing biomass particle but also at any other internal position; for example at the location of half the radius. Fig. 6
illustrates an additional comparison between the predicted and the
measured temperature proles at the location of half the radius.
A good agreement is observed between the experiments and our
simulation results, especially when using the Thurner and Mann or
Chan et al. kinetic data, which further demonstrates the capability

150

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

Thurner and Mann give the best agreement between the experiments and the predictions. Using their data, the effects of the heat
released due to the conversion of biomass to products which are
represented in terms of differences between the specic heats of
virgin material and those of char and volatiles in Eq. (8) is demonstrated through generating the center temperature and mass loss
proles without accounting for this term, at various heats of reactions. The results are depicted in Fig. 7 for a reactor temperature of
623 K. Evident from this gure is that by neglecting the amount
of heat released due to change of nature of the virgin wood to
other materials, and by assuming an endothermic heat for the reactions involved, the temperature peak observed in the experiments
will not be predicted. However, this phenomenon can be captured
by feeding an exothermic heat of reactions into the model. Moreover, the mass loss is remarkably under-predicted should we only
consider an endothermic heat for these reactions.
The phenomena of center temperature peak was also observed
in the work of Park et al. [22] who conducted pyrolysis experiments on moisture free maple wood particles heated in a vertical
tube furnace at temperatures ranging from 638 to 879 K. However, their pyrolysis model with three parallel primary reactions
did not allow observation of this phenomenon. This is because
they did not account for the sensible heat released due to the
chemical decomposition of the virgin particle, and an endothermic
heat of 64 kJ/kg was assumed for all primary reactions. Therefore, they proposed a kinetic model according to which the virgin
biomass is decomposed to gas, tar and an intermediate solid as primary reactions, with secondary reactions including conversion of
the intermediate solid to char and cracking of tar to form additional gas and char. Despite that this model was not supported
by any experimental observations, however, as an endothermic
heat for primary gas and tar formation was used while an exothermic heat for char generation was considered, they were on a right
track since these considerations were consistent with ndings of
Milosavljevic et al. [34].
4.3. Comparison with data of Rath et al.

Fig. 4. Comparison of predicted mass loss and center temperature (lines) with
measurements of Koufopanos et al. [15] (symbols) at two reactor temperatures:
623 K (top), 773 K (bottom). () Measured temperatures; () measured mass losses;
) kinetic data of Chan et al.; (
) kinetic data of Thurner and Mann; (
)
(
kinetic data of Di Blasi and Branca.

of the presented thermo-kinetic model in capturing the behavior of


the pyrolyzing particle. Different magnitudes of temperature peaks
are expected at various positions with the highest one taking place
in the center. A possible reason for this is that the rate of heat
transfer at the surface of the particle is the highest. As the ow
of heat penetrates inside the particle the local heat transfer rate
decreases as the thermal resistance increases. Thus, in the competition between heat transfer rate and biomass decomposition rate,
the later one is dominant for a longer time at positions closer to the
center. This reasoning can be extended to explain why the magnitude (and duration) of the temperature peak is lower (and shorter)
at a higher reactor temperature as observed in the experiments of
Koufopanos et al. [15] and Park et al. [22]: at higher reactor temperatures the rate of heat transfer becomes higher and more dominant
in the above competition so that there is less opportunity for the
chemical decomposition rate to exceed the heat transfer rate.
To provide a better comprehension of the predictability of the
model using various kinetic constants, the average and standard
deviation of the relative errors corresponding to those shown in
Figs. 4 and 6 are reported in Table 9. From this table, the data of

Further experimental validation of the pyrolysis model is carried


out by comparing the simulation outcome with experimental data
of Rath et al. [43], who measured center and surface temperatures
and mass loss proles of beech wood particles heated in a mufe
furnace maintained at 850 C. The predicted center temperature
and mass loss proles obtained from various kinetic parameters
are compared with the measurements of Rath et al. in Fig. 8. The
best match between the predicted and the measured data related
to the center temperature is achieved using the kinetic constants of
Thurner and Mann as well as the data of Font et al. This agreement
is obtained without accounting for tar cracking reaction and with
the same weighted reactions heat of 25 kJ/kg for the primary reactions. Likewise, when the possibility of the secondary reaction was
considered the results were not as satisfactory as those obtained
without tar cracking reaction.
On the other hand, the mass loss over the duration of the pyrolysis process is overestimated with the data of Thurner and Mann
and Font et al. From Fig. 8, the best match between the predictions and the experiments is obtained by employing Chan et al.
kinetic data, given that the nal char yield is overestimated. Using
the data of Di Blasi and Branca leads to underestimate the mass
loss of the pyrolyzing particle, but the nal char yield is slightly
closer to the experimental value than that computed using the Chan
et al. data. Table 10 gives the average and the standard deviation
of the relative errors corresponding to the data shown in Fig. 8.
From Table 10 and Fig. 8, it can be inferred that the mass loss history is best predicted using the kinetic data of Di Blasi and Branca,
which may be attributed to the fact that the kinetic constants of Di

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

151

Table 9
Validation of the model with experiments of Koufopanos et al. [15] (AVE: average relative error; SD standard deviation of relative error).
Thurner and Mann

Mass loss errors


Center temperature errors
Mid temperature errors

Chan et al.

Di Blasi and Branca

Tr [K]

AVE

SD

AVE

SD

AVE

SD

623
773
623
773
673

0.204
0.070
0.561
0.183
0.255

0.218
0.149
0.257
0.260
0.400

0.357
0.034
0.642
0.247
0.255

0.344
0.169
0.281
0.408
0.315

0.161
0.217
0.440
0.332
0.612

0.187
0.251
0.348
0.642
0.487

Fig. 5. Predicted temperature and biomass density proles at the center and the surface of the particle related to the experimental conditions of Koufopanos et al. [15] at
reactor temperature of 623 K.

152

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

Table 10
Validation of the model with experiments of Rath et al. [43] (AVE: average relative error; SD: standard deviation of relative error).
Thurner and Mann

AVE
SD

Chan et al.

Di Blasi and Branca

Font et al.

Mass loss

Tr=0 [K]

Mass loss

Tr=0 [K]

Mass loss

Tr=0 [K]

Mass loss

Tr=0 [K]

0.383
0.374

0.027
0.19

0.205
0.243

0.017
0.202

0.224
0.147

0.006
0.251

0.62
0.555

0.052
0.172

Fig. 6. Comparison of the predicted temperature (lines) with measurements of


Koufopanos et al. [15] (symbols) at the location of half radius for a reactor tem) kinetic data of Chan et al.; (
) kinetic data of Thurner
perature of 673 K; (
) kinetic data of Di Blasi and Branca.
and Mann; (

Blasi and Branca were obtained at fast pyrolysis conditions (with


a heating rate of 1000 K/min) noting that the reactor temperature
in the experiments of Rath et al. was relatively high; i.e. 850 C.
Overall, among various groups of kinetic data, the kinetic parameters of Chan et al. provide the most satisfactory agreement with
the measurements of Rath et al.
Interesting to note with respect to the experiments of Rath et al.
is that unlike the experiments of Koufopanos et al. [15] and Park
et al. [22], the particle center temperature reached the surface temperature and it remained in thermal equilibrium without any peak.
This can be explained with the same reasoning discussed previously. The temperature peak did not occur because the biomass

Fig. 8. Comparison of the predicted center temperature and mass loss (lines) with
measurements of Rath et al. [43] (symbols); (
) kinetic data of Chan et al.; (
)
) kinetic data of Di Blasi and Branca; (
)
kinetic data of Thurner and Mann; (
kinetic data of Font et al.

Fig. 7. Comparison of the predicted mass loss and center temperature (lines) with
measurements of Koufopanos et al. [15] (symbols) at a reactor temperatures of 623 K
at three different values for the heat of reaction without accounting for the sensible
heat released due to the conversion of the virgin material to char and volatiles.

in the center of the particle had completely decomposed before


its temperature reached the surface temperature. This is demonstrated by the numerical results obtained from the pyrolysis model
as shown in Fig. 9. As seen from this gure, the decomposition process at both center and surface of the particle takes place very
rapidly, which can be attributed to the high heating rate. After
110 s from initiation of the pyrolysis process, conversion of biomass
is nalized, while there is still a considerable difference between
the center and the surface temperatures. Thus, beyond this point
the charred particle undergoes only the heating process due to the
conduction mechanism since there is no other heat source as the
sensible heat related to the conversion of virgin biomass to char
and volatiles has already been released.

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

153

heat, in a more accurate simulation, the endothermic heat of


volatiles formation and the exothermic heat of char generation
must be treated separately according to Milosavljevic et al. [34].
Nevertheless, as a consistent value of reactions heat allowed us to
accurately predict thermal degradation of various biomass particles at different experimental conditions, this can be an essential
step towards better understanding of the pyrolysis heat, noting that
a large scatter of values has been reported in the literature. The
results of this study indicate that it is necessary to account for the
release of sensible heat due to the conversion of virgin biomass
to products. It is demonstrated that a release of this energy, usually overlooked in past works, is responsible for the temperature
peak observed in some past studies. This phenomenon may occur
when the local temperature reaches the surface temperature while
the decomposition of biomass at the corresponding position in not
completely nalized; unless otherwise the internal local temperature remains in thermal equilibrium after it reaches the surface
temperature. Finally, mere validation of a pyrolysis model against
heat transfer or kinetic experiments is not sufcient, because
an accurate model should reasonably capture both groups of
parameters.
Acknowledgment
The nancial support provided by the Technology Foundation
STW, the Netherlands, through project No. BioOxyFuel 10416 is
gratefully acknowledged.
Appendix A. Derivation of heat transfer equation
Conservation of energy for a biomass particle undergoing thermal degradation assuming a constant volume during the process,
and thermal equilibrium between solid and gas phases can be
expressed as below in terms of total enthalpy which accounts for
enthalpy of formation and sensible enthalpy.

t
Fig. 9. Predicted temperature and biomass density proles at the center and the
surface of the particle related to the experimental conditions of Rath et al. [43].

5. Conclusions
Several conclusions can be drawn from this study. Comparison
of the simulation results with experimental data of large biomass
particles obtained at different process conditions reveals that the
thermo-kinetic model presented in this work provides satisfactory
predictions when the kinetic parameters of Chan et al. or Thurner
and Mann are employed. The pyrolysis model with three parallel reactions yielding tar, gas and char together with an extended
version of energy equation sufciently captures the experimental
observations of pyrolyzing biomass particles, such as the center
temperature peak reported in some past studies. These experimental validations clearly demonstrate the signicant inuence of the
amount of heat released during the conversion of virgin biomass
on the thermal degradation of the particle. Furthermore, the kinetic
model with three parallel reactions is sufcient to achieve a reasonable agreement between the predictions and experiments so that
any inclusion of secondary reactions is not needed.
A single value for the heat of primary reactions; i.e. 25 kJ/kg,
has been obtained in this work by assuming hBg = hBC and tting the predictions to the measurements. In fact, it represents a
lumped heat of volatiles and char formation at a reference temperature. Given that a satisfactory agreement between the simulations
and experimental data is achieved with the above endothermic

pi h i

1 n
1
(r ug h g ) = n
r n r
r r

r n k

T
r


i = B, C, g

(A.1)

The derivation is presented assuming that decomposition of the


biomass particle takes place according to only three primary reactions.
The rst term in Eq. (A.1) is expanded as follows.

all



h

[B h B + C h C + g h g ]
t

(B h B )
h B
B
= B
+ h B
t
t
t
(C h C )
h C
C
= C
+ h C
t
t
t
(g h g )
h g
g

= g
+ hg
t
t
t
Hence,

all



h


=

B

+
+

h B
B
+ h B
t
t

g

h g
t

+ h g

 

g
t

C

h C
C
+ h C
t
t

 
=

B cB

g
B
C
h B
+ h C
+ h g
t
t
t

T
T
T
+C cC
+g cg
t
t
t

 (A.2)

154

Y. Haseli et al. / Journal of Analytical and Applied Pyrolysis 90 (2011) 140154

We also expand the second term in Eq. (A.1).


1 n
1
(r ug h g ) = n
r n r
r

h g

T
1 n
r ug
+ h g (r n ug ) = ug cg
+ h g n
(r ug )(A.3)
r r
r
r
r
n

Combination of Eqs. (A.2) and (A.3) gives

all


1 n
h + n
(r ug h g )
r r


=

B cB

T
T
T
+ C cC
+ g cg
t
t
t

 
+

g
B
C
h B
+ h C
+ h g
t
t
t

T
T
B
C
= (B cB + C cC + g cg )
+ ug cg
+ h B
+ h C
+ h g
t
r
t
t


+ ug cg

T
1 n
+ h g n
(r ug )
r r
r

(A.4)

g
1 n
+ n
(r ug )
r r
t

With the aid of the mass conservation equations of different species,


i.e. Eqs. (2), (3) and (5) introduced in Section 2, the last three terms
in Eq. (A.4) are expressed as
B
C
h B
+ h C
+ h g
t
t

g
1 n
+ n
(r ug ) = h B (k1 + k2 + k3 )B + h C (k3 B ) + h g (k1 + k2 )B (A.5)
r r
t

Inserting Eq. (A.5) into Eq. (A.4), and then substituting it into Eq.
(A.1), we get
(B cB + C cC + g cg )

T
T
1
+ ug cg
= n
r r
t
r

r n k

T
r

+ Q

(7)

[16]
[17]
[18]

where
Q

= h B )(k1 + k2 + k3 )B h C (k3 B ) h g (k1 + k2 )B


= h B (k1 + k2 ) B + h B k3 B h C k3 B h g (k1 + k2 )B
= (h B h g )(k1 + k2 )B + (h B h C )k3 B

Equation (A.6) can be reshaped using h = hf +

Q = (k1 + k2 )B [hBg +

(cB cg )dT ] + k3 B [hBC +

(A.6)

(A.7)

If a tar cracking reaction was also taken into account, Eq. (A.7) would
be as follows.


Q = B (k1 + k2 )[hBg +

(cB cg )dT ] + B k3 [hBC

(cT cG )dT ]

[26]
[27]
[28]

[30]
[31]

(cB cC )dT ] + T k4 [hT G

[32]

[24]
[25]

[29]

[19]
[20]
[21]
[22]
[23]

cp dT to read
(cB cC )dT ]

[11]
[12]
[13]
[14]
[15]

(8)

[33]
[34]
[35]

References
[1] A. Williams, M. Pourkashanian, J.M. Jones, Prog. Energy Combust. Sci. 27 (2001)
587610.
[2] C. Di Blasi, Combust. Sci. Technol. 90 (1993) 315340.
[3] C. Di Blasi, Ind. Eng. Chem. Res. 35 (1996) 3746.
[4] M.G. Gronli, M.C. Melaaen, Energy Fuels 14 (2000) 791800.
[5] A. Bharadwaj, L.L. Baxter, A.L. Robinson, Energy Fuels 18 (2004) 10211031.
[6] J. Larfeldt, B. Leckner, M.C. Melaaen, Fuel 79 (2000) 16371643.
[7] A. Galgano, C. Di Blasi, Ind. Eng. Chem. Res. 42 (2003) 21012111.
[8] C.A. Koufopanos, G. Maschio, A. Lucchesi, Can. J. Chem. Eng. 67 (1989) 7584.
[9] F. Shazadeh, P.S. Chin, Chem. A: ACS Symp. Ser. 43 (1977) 5781.
[10] C. Di Blasi, Chem. Eng. Sci. 51 (1996) 11211132.

[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]

M.J Hagge, K.M. Bryden, Chem. Eng. Sci. 57 (2002) 28112823.


K.M. Bryden, M.J. Hagge, Fuel 82 (2003) 16331644.
K.M. Bryden, K.W. Ragland, C.J. Rutland, Biomass Bioenergy 22 (2002) 4153.
W.C. Chan, M. Kelbon, B.B. Krieger, Fuel 64 (1985) 15051513.
C.A. Koufopanos, N. Papayannakos, G. Maschio, A. Lucchesi, Can. J. Chem. Eng.
69 (1991) 907915.
B.V. Babu, A.S. Chaurasia, Chem. Eng. Sci. 59 (2004) 19992012.
A.K. Sadhukhan, P. Gupta, R.K. Saha, J. Anal. Appl. Pyrol. 81 (2008) 183192.
A.K. Sadhukhan, P. Gupta, R.K. Saha, Bioresour. Technol. 100 (2009) 3134
3139.
C. Di Blasi, Kinetics and modeling of biomass pyrolysis, in: A.V. Bridgwater (Ed.),
Fast Pyrolysis of Biomass: A Handbook, vol. 3, CPL Press, 2005, pp. 121146.
C. Di Blasi, Prog. Energy Combust. Sci. 34 (2008) 4790.
C. Di Blasi, J. Anal. Appl. Pyrol. 47 (1998) 4364.
W.C. Park, A. Atreya, H.R. Baum, Combust. Flame 157 (2010) 481494.
C. Di Blasi, C. Branca, A. Santoro, E.G. Hernandez, Combust. Flame 124 (2001)
165177.
H.C. Tran, R.H. White, Fire Mater. 16 (1992) 197206.
H. Lu, Experimental and modeling investigations of biomass particle combustion, Ph.D. Thesis, Brigham Young University, 2006.
C. Di Blasi, C. Branca, Ind. Eng. Chem. Res. 40 (2001) 55475556.
F. Thurner, U. Mann, Ind. Eng. Chem. Process Des. Dev. 20 (1981) 482488.
R. Font, A. Marcilla, E. Verdu, J. Devesa, Ind. Eng. Chem. Res. 29 (1990)
18461855.
C.K. Lee, R.F. Chaiken, J.M. Singer, 16th International Symposium on Combustion, The Combustion Institute, Pittsburgh, 1977, pp. 14591470.
A.G. Liden, F. Berruti, D.S. Scott, Chem. Eng. Commun. 65 (1988) 207221.
M.L. Boroson, J.B. Howard, J.P. Longwell, W.A. Peters, AIChE J. 35 (1989)
120128.
H. Kossitrin, Proc. Spec. Workshop on Fast Pyrolysis of Biomass, Copper Mountain, Co., 1980, pp. 105121.
J.P. Diebold, The cracking of depolymerized biomass vapors in a continuous,
tabular reactor, M.Sc. Thesis, Colorado School of Mines, Golden, Co., 1985.
I. Milosavljevic, V. Oja, E.M. Suuberg, Ind. Eng. Chem. Res. 35 (1996) 653662.
M.G. Gronli, A theoretical and experimental study of the thermal degradation of
biomass, Ph.D. Thesis, Norwegian University of Science and Technology, 1996.
W.C. Park, A study of pyrolysis of charring materials and its application to re
safety and biomass utilization, Ph.D. Thesis, The University of Michigan, 2008.
J.A. Havens, J.R. Welker, C.M. Sliepcevich, J. Fire Flammability 2 (1971) 321333.
R. Bilbao, J.F. Mastral, J. Ceamanos, M.E. Aldea, J. Anal. Appl. Pyrol. 36 (1996)
8197.
W.S.L. Mok, M.J. Antal Jr., Thermochim. Acta 68 (1983) 165186.
H. Lu, E. Ip, J. Scott, P. Foster, M. Vickers, L.L. Baxter, Fuel 89 (2010) 11561168.
CHEM1D, www.combustion.tue.nl.
K. Raznjevic, Handbook of Thermodynamic Tables and Charts, Hemisphere
Publishing Corporation, McGraw-Hill, 1976.
J. Rath, G. Steiner, M.G. Wolnger, G. Staudinger, J. Anay. Appl. Pyrol. 62 (2002)
8392.
E.J. Kansa, H.E. Perlee, R.F. Chaiken, Combust. Flame 29 (1977) 311324.