Accepted Manuscript

Title: Kinetics of the Pyrolysis of Oil Sands Based upon

Thermogravimetric Analysis

Authors: Jia Chunxia, Chen Jiajia, Bai Jingru, Yang Xun,

Song Shaobo, Wang Qing

PII: S0040-6031(18)30342-3
DOI: https://doi.org/10.1016/j.tca.2018.06.002
Reference: TCA 78015

To appear in: Thermochimica Acta

Received date: 9-3-2018

Revised date: 3-6-2018
Accepted date: 4-6-2018

Please cite this article as: Jia C, Chen J, Bai J, Yang X, Song S, Wang Q, Kinetics of
the Pyrolysis of Oil Sands Based upon Thermogravimetric Analysis, Thermochimica
Acta (2018), https://doi.org/10.1016/j.tca.2018.06.002

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Kinetics of the Pyrolysis of Oil Sands Based upon
Thermogravimetric Analysis
JIA Chunxia，CHEN Jiajia，BAI Jingru，YANG Xun，SONG Shaobo，WANG Qing
(Engineering Research Center of Ministry of Education for Comprehensive Utilization of Oil
Shale, Northeast Electric Power University, Jilin 132012, China;)

T
First author: Jia Chunxia (1981 - ), female, senior experimenter, PhD, research direction for heat conversion

IP
utilization of unconventional energy, jiachunxia_215@126.com.
Contact author: Wang Qing (1964 - ), male, Ph.D., Professor, doctoral supervisor, research direction for new

R
energy comprehensive development and utilization and clean energy technology, rlx888@126.com.

SC
Highlights
 The pyrolysis kinetics of oil sand was studied by four methods.

U
The most probable mechanism function matches General kinetic mechanism
functions.
N
 Distributed activation energy models are used to explore the pyrolysis kinetics.
 The Akaike Information Criteria method are used to determine the optimal
A
model.

M

The Weibull model is best fitted with the differential thermogravimetry curve.
ED

Abstract
In this work, the pyrolysis kinetics of oil sands were determined based upon the TG-DTG curves.
Using the single-step reaction model, the parameters of thermal decomposition kinetics of oil
PT

sands, such as the activation energy E0=186.14 kJ/mol, reaction order n=5.21 and pre-exponential
factor A0=1.56E13 s-1, were obtained. The activation energy values of 185.05 kJ/mol and 162.95
kJ/mol were received by using the Friedman and KAS method, respectively. The result of the most
E

probable mechanism function is not perfectly matched with the common kinetic model, the first
order reaction model is not suitable for the study of oil sands’ pyrolysis. Using different density
CC

probability functions, such as Logistic, Lognormal, Gaussian, Rayleigh and Weibull, the
distribution of apparent activation energy of the pyrolysis of oil sands was studied. The Akaike
Information Criteria and the calculated kinetic parameters show that the Weibull model is the most
A

suitable model for studying the pyrolysis of oil sands.

Key words: Oil sands; Pyrolysis; Kinetic analysis; DAEM model

1. Introduction
Due to depleting fossil reserves in China, oil sands, which are an unconventional energy
source, have drawn global attention because of their reserves [1]. Oil sands are usually composed of
sand, water and bitumen (heavy oil), and are also known as tar sands, heavy oil sands and
asphaltic sand. Oil sands are sedimentary rocks containing bitumen or other heavy oils [2]. Oil
sands are mainly separated using the methods of thermal caustic washing, solution extraction and
pyrolysis [3-4]. Pyrolysis technology has particularly been paid plenty of research attention due to
its small pollution, wide application range and easy industrialization.
The pyrolysis of oil sands refers to the light components of oil sands asphalt that vaporize at
high temperature. The heavy components are pyrolyzed into light components and vaporized at
high temperature, while the coke is coked on the surface of the sands, and the gas products
continue to be cleaved into small molecules [5]. After the pyrolysis, the quality of the oil is
significantly improved, due to which, the analysis of the pyrolysis features of oil sands is
particularly important. The pyrolysis of oil sands produces coke, oil and gas products in the

T
absence of oxygen. Several thermal analysis techniques, such as DSC, DTA, TGA, and EGA, can

IP
be used to acquire the experimental data. The most commonly used technique for the basic kinetic
characterization is thermogravimetric (TG) analysis. At present, the chemical kinetic model of

R
organic hydrocarbon generation [6] has the following characteristics: (1) total package first order
reaction model; (2) a limited number of parallel first-order reaction kinetic models; (3) an infinite

SC
number of parallel first-order reaction kinetic model; (4) maximum reaction rate model; (5)
tandem first-order reaction model; (6) total package n-level reaction model. Many researchers
have carried out studies on the kinetics of hydrocarbon generation. Luo Manyin [7] used C-R

U
integral method for the three kinds of oil sands, and analyzed the non-isothermal
thermogravimetric curve. Li et al. [8] calculated the kinetic parameters of Indonesian oil sands
N
based on C-R method. Young et al. [9] (2009) calculated the kinetic parameters of Alberta oil sand
bitumen using a primary package reaction model. Wang Qing [10] used asymmetric multi-peak
A
Gaussian fitting method for the analysis of pyrolysis kinetics of Indonesia’s oil sands based on
M

two-component global response model, and then, used the distribution of activation energy model
and non-linear conversion rate integral method to calculate the kinetic parameters, and achieved
good fitting results. Ma et al. [11] calculated the kinetic parameters of pyrolysis of oil sands in
ED

Indonesia using Friedman method. Guo et al. [12] used DAEM and FWO models to research the
thermal decomposition kinetics of Indonesian oil sands under different heating rates. As far as we
know, there are only a few studies focusing on the mechanism of the formation of oil sands. Pitt [13]
PT

calculates the approximate activation energy distribution of coal pyrolysis using instantaneous
reaction rate and approximate energy distribution technique. Anthony and Howard [14-15]
characterized the pyrolysis kinetics of coal by using Gaussian distribution, and studied the effects
E

of time, temperature, heating rate, pressure, hydrogen partial pressure and particle size on coal
yield. Burnham et al. [16] used Gaussian distribution of nth order reaction to explain the width
CC

asymmetry of coal pyrolysis curves, and described the differences between different DAEM
models. In order to avoid the computation difficulties of continuous PDF, Miura method [17] was
also used in the pyrolysis kinetics of oil sands.
A

In the present study, several different probability density functions, including Logistic,
Lognormal, Gaussian, Rayleigh and Weibull are used to kinetic analysis of oil sands to determine
the most suitable model for calculating the activation energy of the pyrolysis kinetic parameters
and to make a more in-depth study of the formation mechanism of oil sands. Gaussian distribution
has been studied extensively, and Weibull distribution can analyze synthetic data from nth and
Gaussian distributions. Because Weibull distribution has two additional kinetic parameters besides
A and E, a parameter mainly controls the width of distributed activation energy. TG-DTG curves
of oil sands under different heating rates were analyzed. The single-step reaction model,
iso-conversional model and several different probability density functions corresponding to
DAEM model were used to calculate the kinetics of the oil sands pyrolysis. The calculated
apparent activation energy of different models was compared to determine the most appropriate
model to calculate the pyrolysis kinetics of oil sands, thus providing the theoretical basis for a
comprehensive and efficient utilization of oil sands’ pyrolysis.

2 Experimental
2.1 Materials
The oil sands used in the experiment were obtained from Quanzhou city, China. The results

T
of elemental analysis (based upon national standards) and proximate analysis are presented in

IP
Table 1.
Table 1 Results for the Quanzhou’s oil sands’ proximate analysis and ultimate analysis
Ultimate analysis Calorific value Proximate analysis

R
Sample /%(ad) /kJ·kg-1 /%(ad)
C H O N S Qnet.ar M V A FC

SC
oil
27.29 2.41 17.05 0.21 0.91 8814.24 1.42 37.79 50.71 10.08
sand
2.2 Thermogravimetric analysis

U
The pyrolysis experiments of the oil sands were conducted on TAG/DSC1
thermo-gravimetric analyzer obtained from METTLER-TOLEDO. Around 5 mg of the sample
N
was laid at the bottom of the crucible in each experiment. The heating rates were 10 °C/min,
20 °C/min, 40 °C/min, and 80 °C/min, and the final temperature was 900 °C.
A
2.3 Kinetics analysis
2.3.1.1 Single-step reaction model
M

The pyrolysis of solid materials is generally based on weight loss data, whereas the
multi-scan rate of non-isothermal method is more practical than the isothermal method [18-19]. The
ED

reaction kinetics are represented by the Arrhenius equation. The pyrolysis reaction rate of oil
sands is interrelated to the rate constant and the reaction mechanism function using Equation (1)
[21]
.
d𝛼 𝐴 𝐸
PT

= exp⁡(− )𝑓(𝛼) (1)

d𝑇 𝛽 𝑅𝑇

where k(T) is the temperature dependence of the rate constant, f(α) is the pyrolysis reaction
mechanism function of oil sands, A is the pre-exponential factor, s-1, T is the temperature, K, R is
E

the universal gas constant (8.314 x 10-3 kJ/(mol·K)), E is the apparent activation energy, kJ/mol, β
CC

𝑚−𝑚
is the heating rate, α is the conversion rate given by: 𝛼 = 𝑚−𝑚 0 , 𝑚 0 is the initial mass
∞

fraction, %, and 𝑚∞ is the remaining mass fraction,%.

A

Since the pyrolysis of oil sands follows the constitutive equation, the reaction mechanism
f(α)=(1-α)n is suitable for kinetic analysis. The integration of Eq. (4) yields Eq. (2).
𝛼 1 𝑇 𝐸 𝐴
g(𝛼) = ∫0 (1−𝛼)𝑛
d𝑇 = ∫0 exp (− 𝑅𝑇) d𝑇 = 𝛽 𝐼(𝐸, 𝑇) (2)

Malek proposed probable thermal decomposition mechanism function, which is represented

by Eq. (3).
𝑅𝑇 2 d𝛼 1
g(𝛼) = (3)
𝐸𝛽 d𝑡 𝑓(𝛼)
When α=0.5, Eq. (6) changes to Eq. (4).
2
𝑅𝑇0.5 d𝛼 1
⁡g(0.5) = ( d𝑡 ) (4)
𝐸𝛽 0.5 𝑓(0.5)

d𝛼
where T0.5 and ( d𝑡 )0.5 are the time derivatives of temperature and conversion rate for α = 0.5,

respectively.
By dividing Eq. (3) by Eq. (4), expression for y(α) is obtained, which is represented by Eq. (5).
d𝛼
𝑇 ( ) 𝑓(𝛼)g(𝛼)
y(𝛼) = (𝑇 )2 d𝑡
d𝛼 = 𝑓(0.5)g(0.5) (5)
0.5 ( )0.5
d𝑡

where y(α) is the definition function. The commonly used kinetic reaction mechanism functions

T
[f(α) and g(α)] are presented in Liu Hongpeng [20]and Vyazovkin [23].

IP
The raw data at the same α value are measured using non-isothermal DTG curves for
different heating rates. Eα and α can be obtained from the minimum value of Eq. (6).

I E ,T β

R
n n
Φ(𝐸𝛼 ) =  I E ,T β ,j j
(6)

SC
i 1 j  i
  ,j i

Vyazovkin method [21] is a common integral method, which is used to improve the accuracy.
After integration, Eq. (7) is obtained.

𝐼(𝐸𝛼 , 𝑇𝛼 ) = ∫0 𝛼 exp⁡(−
𝑇 𝐸𝛼
) d𝑇 U (7)
N
𝑅𝑇

2.3.1.2 Energy Compensation

A
The kinetic parameters corresponding to different reactions have an energy compensation
effect [22], and their mathematical expressions are written according to Eq. (8).
M

ln𝐴𝑛· = 𝜔 · 𝐸𝑛 + 𝜑 (8)
where ω and φ are linearly fitted to E and A for different values of n. The slope is ω and the
intercept is φ. The parameters of the linear fit can be replaced in Eq. (2) to obtain Eq. (9).
ED

𝛽·(d𝛼⁄d𝑇) 𝐸
ln [ (1−𝛼)𝑛
] = ln𝐴𝑛 − 𝑅𝑇𝑛 (9)

2.3.1 Iso-conversional method

PT

2.3.1.1 Friedman method

The most common method of conversion is Friedman method [23], which is obtained from Eq.
(1). The kinetic equation of pyrolysis of oil sands obtained by this method is given by Eq. (10).
E

d𝛼 𝐸
ln (𝛽 d𝑡 ) = ln[𝐴𝛼 𝑓(𝛼)] − 𝑅𝑇𝑘 (10)
CC

d𝛼 1
From⁡ln⁡(𝛽 d𝑡 ) to 𝑇, the least square method is used to fit the relationship between the rate of
A

increase in temperature and the temperature, which can effectively avoid the reaction mechanism
function. The slope can be obtained from E, while the intercept can be used to obtain the value of
A.
2.3.1.2 KAS method
Without knowing the pyrolysis reaction mechanism, the coupling of Eq. (1) and Eq. (2) is
achieved based upon the KAS [24] method.
𝛽 𝛼 𝐴 𝑅 𝐸
ln (𝑇 2) = ln (𝐺(𝛼)𝐸 ) − 𝑅𝑇𝛼 (11)
𝛼 𝛼 𝛼
 β⁡ 1
From ln⁡(𝑇 2) to𝑇 , the value of E can be obtained through linear fitting of the slope of curve
𝛼 𝛼

using the least square method. The frequency factor A can be also be obtained from the intercept
of the line. In the case of Eα, the master-plots method is used to determine the mechanism model
of f(α) or g(α). A reliable master-plots method considering comprehensively f(α) and g(α) is
reported by Gotor et al [26]. The equation of this master-plots method is given:
d𝛼 𝑄(𝑢𝛼 )𝑇𝛼
𝑓(α)g(𝛼) = (12)
d𝑡 𝑢𝛼

∞
∫𝑢 𝑒 −𝑢 𝑢−2 d𝑢
Q(𝑢) = 𝑒 −𝑢 𝑢−2
(13)

T
Right and left hand-side of Eq. (12) represents the theoretical and experimental combination of f(α)

IP
g(α), which is a further specification in the feature transformation α=ξ(general ξ=0.5),
f(α)g(α)/[f(α=ξ)g(α=ξ)]. The best mechanism is considered to be the best mechanism for
experimental f(α)g(α) fitting. In this way, α is calculated from the intercept using f(α) or g(α).

R
2.3.3 nth order DAEM

SC
Rather than description for many complex reaction, the DAEM can be described as a number of n
order reactions, and the activation energy is a continuous PDF. Considering the main purpose of
𝑛
∫ 𝑓(𝐸)d𝐸 = 1and𝑓(𝛼) = （1 − 𝛼） , the distributed activation energy equation for the nth

reaction [25] is given by Eq. (14). U

N
∞ 𝐴 1⁄1−𝑛
x = 1 − α = ∫0 𝑓(𝐸) · [1 + (n − 1)exp⁡(− 𝛽)Ψ(𝐸, 𝑇)] d𝐸 (14)
A
The different probability density functions used for the pyrolysis of oil sands are given in
M

Table 2. The unknown kinetic parameters for DAEM can be estimated using the least square
method [24].Nonlinear regression [27] also involves some function that minimizes the difference.
Between the calculated value and the measured value, it is usually Squares (RSS):
ED

2
RSS = min ∑𝑁
𝑗=1(𝑥𝑒,𝑗 − 𝑥𝑐,𝑗 ) (15)

Where the symbols e and c represent the experimental values and the calculated values,
PT

respectively [28-29]. A single RSS cannot obtain the best fitting results. Therefore, the F-test is
considered [30], which is given by Eq. (16).
100
ε(%) = min ∑𝑁
𝑖=1 √𝑅𝑆𝑆𝑖 ⁄(𝑁𝑑 − 𝑃) (16)
E

𝑁𝑖

Where Ni, Nd and P denote the total number of different heating rates, number of experiments, and
CC

the number of unknown parameters.

Table 2 Probability density functions of different DAEMs

A

PDF name PDF form Parameters E0

1 (𝐸 − 𝐸𝑆 )2
Gaussian 𝑓(𝐸) =
𝜎√2𝜋
exp⁡[
2𝜎 2
] ES，A，σ，n E𝑆
−2
𝜋 𝜋(𝐸 − 𝐸𝑆 ) (𝐸 − 𝐸𝑆 )𝜋
Logistic 𝑓(𝐸) =
𝜎√3
exp (−
𝜎√3
) [1 + 𝑒 xp (
𝜎√3
)] ES，A，σ，n E𝑆
1 (ln(𝐸 − 𝐸𝑆 ) − 𝑎)2
Lognormal 𝑓(𝐸) =
(𝐸 − 𝐸𝑆 )𝜎√2𝜋
exp⁡[−
2𝜎 2
] ES，A，σ，n，a E𝑆 + exp⁡(𝑎 + 𝜎 2 /2)
𝐸 − 𝐸𝑆 (𝐸 − 𝐸𝑆 )2
Rayleigh 𝑓(𝐸) =
𝜎2
exp⁡[
2𝜎 2
] ES，A，σ，n E𝑆 + σ√𝜋⁄2
𝑎 𝐸 − 𝐸𝑆 𝑎−1 𝐸 − 𝐸𝑆 𝑎
Weibull 𝑓(𝐸) = (
𝜎 𝜎
) exp [⁡(
𝜎
) ] ES，A，σ，n，a E𝑆 + σg(1 + 1⁄𝑎)

3 Results and discussion

3.1 Pyrolysis characteristics
Figure 1 shows the TG-DTG curves of oil sands for different heating rates. As can be seen in
Figure 1, the oil sands pyrolysis process is divided into three stages. And the first stage ranged
from 20 - 200 °C and belonged to the dehydration stage [12, 31]. The main reaction was the removal
of internal water, adsorbed water and adsorbed gas of the oil sands, and accounted for about 2.5%
of the weight loss. The second stage belonged to the low-temperature decomposition section from
200 - 600 °C, which was mainly due to desorption of oil and cleavage to small molecules, and was
the main phase of pyrolysis of oil sands. During the pyrolysis process, hydrocarbons are
decomposed and accompanied by the escape of gas and oil evaporation. In this phase, the weight
loss of samples occupied by about 21% of the total weight loss. According to the literature [32-36],

T
the soft asphalt and bitumen are generally considered in oil sand asphalt, and the peak weight loss

IP
at this stage is the cracking of soft pitch. At this stage, the pyrolysis of asphaltene was first carried
out mainly with the physical changes of asphaltene softening, and recombination of small

R
molecules, though without the formation of coke. Then, a significant weight loss peak was
observed in the DTG curve at 450-550 °C, which was mainly due to the volatilization [37].

SC
Carbonaceous rocks (calcite, dolomite and iron-rich dolomite) were mainly pyrolyzed after 600 °C,
while CO2 was generated from the carbon contained in oil sands to produce CO at high
temperature that also exacerbated the weight loss of oil sands. The weight loss of oil sands

100 U
samples accounted for about 46% of the total weight loss in high temperature section.
0
N
A
80
DTG/%· min-1

-20
M
TG/%

-1
1 10℃ min
●

60 2 20℃ min-1
●
ED

3 40℃ min-1
●

-40
4 80℃ min-1
●

40
200 400 600 800
PT

T/℃

Figure 1 TG and DTG curves of oil sands pyrolysis at the heating rates of 10 °C/min, 20 °C/min, 40 °C/min, and

80 °C/min,
E

Fig. 1 shows that, with the increasing of heating rate, the TG curve of oil sands moved to the
CC

right, while the temperature corresponding to the highest rate of weight loss and the termination
temperature of the oil sands pyrolysis increased. At the volatilization stage of oil sands’ pyrolysis,
at the same temperature, for higher heating rate, less pyrolysis and small hydrocarbon molecules,
the corresponding weight loss would be higher, while the small hydrocarbon precipitation time
A

will be shortened. It can be seen from the pyrolysis curves that the conversion rate of oil sand
decreases in turn with the increase of heating rate, and the DTG curves with higher heating rate
move to the higher temperature side as a whole, that is, the pyrolysis temperature corresponding to
the maximum weight loss rate and the maximum weight loss rate all shift to the high temperature
side. It is known from the heat transfer theory, the change in TG-DTG curve may be caused by the
"thermal hysteresis", which was caused by the heat transfer effect at different heating rates. It may
also control the chemical kinetics during the pyrolysis process. According to the test of Figure 9 in
Lyon [38], the thermal resistance can be reduced to the negligible level due to the error of heat
diffusion of 5% when the maximum heating rate is 80℃ /min. The result is consistent with the
result of Burnham [39] in Figure 4.2, so the effect of temperature rising rate on the pyrolysis of oil
sand is mainly influenced by chemical kinetics.
3.2 Single-step reaction model analysis
The single-step reaction model can be achieved through MATLAB. The reaction order and
pre-exponential factor of the single-step reaction of oil sands’ pyrolysis are obtained from the
kinetic parameters of oil sands’ pyrolysis under single heating rate and the average activation
energy calculated by Vyazovkin method. The average activation energy of oil sands’ pyrolysis can
be obtained using Vyazovkin [40-41] method Eq. (17).

T
𝐸 𝐸
𝐼(𝐸, 𝑇) = 𝑇exp (− ) − 𝑅 𝐸𝑖 (𝑥)⁡⁡ (17)

IP
𝐸𝑅

∞ 𝑒 −𝑥𝑢
𝐸𝑖 = ∫1 𝑢
(18)

R
Eq. (17) and (18) are imported into the MATLAB for calculations with the residual of 1e-8. Eq. (6)

SC
is solved for the minimum value using Nelder-Mead simplex direct search algorithm, while the
range of conversion rate α is [0.1, 0.9].


（ a）
280

 E 30
60
50
40  E  30 %
（ b）
10 n=8

U
α=0.95
N
E0=186.14KJ·mol-1
20
240 10 5
ln(β(dα/dT)/(1-α)n)

α=0.05
A
0
-1
Eα/KJ·mol

0.0 0.2 0.4 0.6 0.8 1.0

α
200 0
M

-1
E =186.14KJ·mol
0
160 -5
ED

n=0

120 -10
1.3 1.4 1.5 1.6 1.7 1.8 1.9
0.0 0.2 0.4 0.6 0.8 1.0
1/T/10-3
α
PT

（ c） 50 （ d） 10℃/min
2.0
lnA0=30.38/lns-1 20℃/min
-4
40
lnAn=2.10×10 En-7.38 40℃/min
E

f(α)=(1-α)5..21 80℃/min
1.5
30 1
CC lnAn/lns-1

y(α)

20 1.0

10 15
0.5
A

-10 0.0
0 50 100 150 200 250 300
0.2 0.4 0.6 0.8 1.0
En/KJ·mol-1 α

Figure 2 Calculations for the kinetics of oil sands’ pyrolysis using single-step reaction model:(a)Change in
apparent activation energy with conversion rate;(b)Arrhenius point from Eq.(9);(c)Energy compensation of n-order
reaction series; (d)Correlation curve between y(α)-α test and theoretical results
 Fig. 2(a) exhibits the change in apparent activation energy of the pyrolysis of oil sands as
calculated by Eq. (6) with the conversion rate. The E of the oil sands’ pyrolysis was not constant,
and changed with the conversion rate, while the average value of the apparent activation energy E0
was 186.14 kJ/mol. When α was within the range of [0-0.2], the activation energy showed a
downward trend, indicating that small hydrocarbon compounds overflowed with the increase in
temperature during the pyrolysis process, which resulted in the reduction of activation energy.
When α was within the range of [0.2-0.5], the activation energy of the pyrolysis of oil sands
monotonically grown in pace with the conversion rate, and the slope of the tangent decreased with
the increase in α value. The activation energy of the pyrolysis of oil sands increased monotonically
with α when α lied within the range of [0.5-0.9]. However, in this case, the tangent of the slope

T
decreased with the increase in α value. As can be seen in Figure 1, the pyrolysis of oil sands did

IP
not show any peak between 200 and 600 °C, indicating that the pyrolysis of the oil sands was not a
single reaction mechanism, and that the activation energy did not increase monotonically.
Figure 2(b) shows the corresponding date point of the apparent activation energy of oil sands’

R
pyrolysis at 20 °C/min for the conversion rate α of 0.05-0.95. The corresponding slopes and

SC
intercepts were obtained using linear fitting of different n-value data. The E and A factor of the
pyrolysis of oil sands with 17 different reaction orders (ranging between 0-8) were calculated
using the slope and the intercept. When n = 0 and 0.5, the effect of linear fitting was not ideal.

U
Based on a single thermal curve, due to the different reaction mechanism functions, there was a
huge error in the kinetic parameters. Due to this, it can be said that the kinetic compensation effect
N
existed in the pyrolysis of oil sands.
Figure 2 (c) shows the kinetic compensation of lnAn and En by fitting different reaction order.
A
Fig. 2 (a) obtains the mean value of apparent activation energy E0= 186.14 kJ/mol, and the linear
M

regression results indicate that the average pre-exponential factor lnA0 is 30.38lns-1,reaction order
n is 5.21, which means that, the result of dynamic compensation is given by Eq. (19).
d𝛼 1.86×105
= (1.56 × 1015 /𝛽)exp⁡(− )(1 − 𝛼)5.21 (19)
ED

d𝑇 𝑅𝑇

As considered by Vyazovkin et al [23], the deviation between the maximum or minimum of Eα

is greater than 20 - 30% E0, then the variation of Eα with a means that the pyrolysis process is a
PT

multistep reaction rather than a single reaction mechanism. As can be seen in Figure 2 (a), when
the conversion rate is greater than 0.8, the absolute deviation (εE, %) of Eα and E0 is higher than
30%. The deviation of more than 30% may be caused by the secondary reaction of volatile matter
E

(gas) and coke. Such a side reaction is in a heterogeneous reaction system. Although the
temperature is very high, the reaction rate has a very low response to the DTG curve.
CC

Figure 2(d) shows the common kinetic reaction mechanism curve, whereas the step length of
α between the range of 0.1-0.9 was 0.1. The experimental data was compared with the modeled
curve. The mechanism function of the oil sands’ pyrolysis was matched with the kinetic model of
A

the reaction series n=2 when α was within the range of 0.1-0.5. When α was within the range of
0.5-0.8, the pyrolysis reaction mechanism of oil sands matched with the 3-D diffusion model. It
means that as the temperature (conversion) increased, reaction rate is mainly controlled by
diffusion. When α=0.8-0.95, the pyrolysis process is affected by the secondary reaction from
volatiles (gas) and coke. There is no fixed model to match the reaction mechanism of oil sand
pyrolysis. Therefore, the pyrolysis of oil sands is not a single reaction mechanism, while there is
an error in solving the kinetic parameters of the pyrolysis of oil sands using single-step reaction
model.
3.3 Iso-conversional method
The kinetic parameters of the oil sands’ pyrolysis was solved using the iso-conversional
method. From the TG curve of oil sands’ pyrolysis, the data points of oil sands’ pyrolysis at
different heating rates were selected using the least squares linear fitting for the conversion rate of
0.05-0.95 and the step size of 0.05.
-4 -12.0
（ a） 10℃ /min
10℃/min （ b） 20℃ /min
40℃ /min
-12.5
-5 20℃/min 80℃ /min

40℃/min
-13.0

T
80℃/min
-6

ln(dα/dt)
-13.5
ln(dα/dt)

IP
0.70
-7 0.75 -14.0
0.80 0.55
0.50
0.85 0.45
0.40
0.90

R
0.35 0.15 0.10
0.30 -14.5
-8 0.25 0.25 0.20
0.20 0.30
0.95 0.15 0.45 0.35
0.10 0.05 0.65 0.55
-15.0 0.80 0.70

SC
0.95 0.90 0.85

-9
1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0
1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0
1/T/10 ·K -3 -1 1/T/10-3·K-1

（ c） 400 （ d）
40
U
N
Friedman Friedman

KAS 35 KAS methed

350
A
30 lnAα=1.35×10-4Eα+0.60
300 r=99.89
Eα/KJ·mol-1

lnAα/lns-1

25
M

250

200 20
-4
lnA =1.26×10 E -6.42
α α
ED

150 15 r=99.80

100 10
0.0 0.2 0.4 0.6 0.8 1.0 150 200 250 300
α Eα
PT

Figure 3 Calculation results of iso-conversional method for:(a)Friedman method; (b)KAS method; (c)Apparent
activation energy obtained by Friedman and KAS methods; (d)Energy compensation effect of different
E

iso-conversional method
Figure 3(a) shows the results for the solution to the pyrolysis kinetic parameters of oil sands
CC

using Friedman method. The relationship between ln(d𝛼/d𝑡) and 1/𝑇 was calculated between
the α values of 0.05 and 0.95 based on the pyrolysis TG-DTG curve under different heating rates.
The linear fitting of the least square method was used to get the pyrolysis kinetic parameters of the
A

oil sands. As can be seen in Fig. 3(a), the linearity is good from Friedman method, then the
average value of the kinetic parameter of 185.05 kJ/mol is obtained.
Figure 3(b) demonstrates the results for the solution of pyrolysis kinetic parameters of oil
sands using KAS method. The relationship between ln(𝛽/𝑇 2 ) and 1/T was calculated in the
range of α = 0.05-0.95. Therefore, the average value of KAS for solving the apparent activation
energy E0 of oil sands is 162.95 kJ/mol. Eα values obtained from Friedman method were higher
than that from KAS method. This being Eα differences in two iso-conversional method mainly
were generated from using of temperature integral I(E,T) and data-noises once using numerical
differentiation technique to acquire dα/dT data [42].
Figure 3(c) shows the relationship between the apparent activation energy Eα and the
conversion rate α calculated using different iso-conversional methods. The dependence of
activation energy versus conversion is similar but noticeably different for the KAS and Friedman
methods, were found to be alike to the curves obtained using the single-step reaction. The
activation energy showed a downward trend between α=0-0.2. The activation energy of oil sands
monotonically increased with the increase in the conversion, while the slope of the curve
decreased with the increase in α value between 0.2 and 0.5. The activation energy of oil sands also
monotonically increased with the increase in conversion rate, though the slope of the tangent

T
decreased with the increase in α value between 0.5 and 0.9. From Figure 3 (c), it is known that the

IP
activation energy is not a fixed value, its fluctuation range is very large, and it does not change
monotonously with the reaction. The activation energy of the oil sand decreases first and then

R
slowly rises with the pyrolysis reaction. The weak downward trend is attributable to the increase
of the rate of volatilization and escape with the increase of temperature. Chemical energy shows a

SC
downward trend. The slow rise is due to the increasing dominance of pyrolysis compared to
evaporation of pre-existing hydrocarbons. From the analysis of the kinetic parameters of Meng
Meng [43] in different oil sand, it is found that the high degree of thermal cracking reaction is

U
basically finished when the activation energy of the oil sand tends to 220kJ/mol, so the final
activation energy should be greatly affected by the secondary reaction of the volatiles (gas) and
N
coke. In addition, baseline issues cannot be discounted as providing some of that increase. The
reference temperature for α =1 shifts up with heating rate just as the midpoint of conversion shifts
A
up. As shown in Figure 3, as the heating rate increased, the higher the reference temperature of
M

baseline.
Figure 3(d) demonstrates the effect of energy compensation on the pyrolysis kinetic
parameters of the oil sands for different iso-conversional methods. The E and A were satisfactorily
ED

represented by ln 𝐴𝛼 = 𝑎𝐸𝛼 + 𝑏 (see Figure 3(a)). Therefore, there is an energy compensation

effect. The fitting results for the Friedman method produced the relationship to be⁡ln 𝐴𝛼 = 1.35 ×
10−4 𝐸𝛼 + 0.60. In this case, the correlation coefficient was 99.89. Additionally, the fitting results
for the KAS method produced the correlation of⁡ln 𝐴𝛼 = 1.26 × 10−4 𝐸𝛼 − 6.42, for which, the
PT

correlation coefficient had the value of 99.80. Figures 3(a) and 3(b) show that the mean value
apparent activation energy of different iso-conversional methods were E0=185.05 kJ/mol and
E

E0=162.95 kJ/mol. The pre-exponential factors obtained by the Friedman method and KAS
method were 1.29e+11 s-1 and 1.34e+6 s-1, respectively. The kinetic parameters calculated by the
CC

two methods are similar to those calculated by Meng Meng [44].

3.4 DAEM kinetics
During the process of calculating the kinetic parameters of oil sands’ pyrolysis, different
A

probability density functions correspond to different DAEMs. Different density functions are
imported into MATLAB software to carry out the Nelder-Mead simplex direct search algorithm.
The number of iterations was 1500, while the maximum residual of the objective function was
1e-8. The initial values of apparent activation energy E, pre-exponential factor A, reaction order n
and standard deviation δ were assumed using obtained from iso-conversional method of oil sands’
pyrolysis. The initial value was close to the actual value. The higher the parameter, better was the
iteration. Furthermore, F-test was used to examine the rigor of the calculation results. The fitting
results r> 0.99 were considered to give a good fit between the experimental and calculated results.
1.0 1.0
（ a） experiments calculations （ b） experiments calculations
20℃/min 20℃/min 20℃/min 20℃/min
0.8 40℃/min 40℃/min 40℃/min 40℃/min
60℃/min
0.8
60℃/min 60℃/min 60℃/min
80℃/min 80℃/min 80℃/min 80℃/min

0.6 0.6
x=1-α

x=1-α
Gaussian DAEM Logistic DAEM
0.4 0.4

0.2 0.2

T
0.0 0.0
400 500 600 700 800 900 400 500 600 700 800 900

IP
T/K T/K

R
1.0 1.0

SC
（ c） experiments calculations
（ d）
experiments calculations
20℃/min 20℃ /min 20℃/min 20℃ /min
40℃/min 40℃ /min 40℃/min 40℃ /min
0.8 60℃/min
0.8 60℃/min 60℃ /min
60℃ /min
80℃/min 80℃/min 80℃ /min
80℃ /min

0.6 0.6

U
x=1-α

x=1-α

Rayleigh DAEM
N
0.4 0.4 Log-normal DAEM
A
0.2 0.2
M

0.0 0.0
400 500 600 700 800 900 400 500 600 700 800 900
T/K T/K
ED

1.0 0.16

（ e）
experiments calculations （ f）
20℃ /min 20℃ /min 0.14
40℃ /min 40℃ /min Rayleigh PDF
0.8 60℃ /min 60℃ /min
0.12 Log-normal PDF
80℃ /min 80℃ /min
PT

Logistic PDF
0.10 Gaussian PDF
f(E)/mol·KJ-1

0.6 Weibull PDF

Weibull DAEM
x=1-α

0.08

0.4
E

0.06

0.04
CC

0.2
0.02

0.0 0.00
0 100 200 300
400 500 600 700 800 900
E/KJ·mol-1
T/K
A

Fig.4 Comparison of experimental results and calculated results of different density probability functions:
(a)Gaussian PDF;(b)Logistic PDF;(c)Rayleigh PDF;(d)Log-normal PDF;(e)Weibull PDF; (f) Different DAEM
activation energy distribution
Figure 4 shows the pyrolysis experiment results and calculation results of oil sands under
different PDF corresponding to DAEMs. The kinetic parameters were calculated using different
DAEM models, and the results are presented in Table 4. The models of Logistic, Lognormal,
 Gaussian, Rayleigh and Weibull were used to calculate the thermogravimetric data. The results of
different density probability functions’ fitting produced r > 0.99, while the standard residuals were
ε<0.41. Therefore, the kinetic parameters of oil sands’ pyrolysis can be calculated using Logistic,
Lognormal, Gaussian, Rayleigh, and Weibull DAEMs. From the results of Table 3, it can be
shown that E0(Gaussian)> E0(Logistic)> E0(Weibull)> E0(Rayleigh)> E0(Lognormal). The kinetic
parameters of oil sands’ pyrolysis derived from different activation energy models shows that
kinetic parameters of Weibull model is the closest to that of the single-step reaction and
iso-conversional method. The Weibull distribution is more flexible, which can change the shape
and width, as described by Burnham and Braun [45]. With the increase of Weibull distribution
parameter a, Weibull distribution matches Gaussian distribution or σ and n of nth-order reaction

T
model.

IP
Table 3 Different DAEM calculation kinetic parameters of oil sands’ pyrolysis
DAEMs A（s-1） E（KJ·mol-1） σ（KJ·mol1） a n E0（KJ·mol-1） Pearson’s Mean RSS ε（%）

R
s

Gaussian 2.55E+15 226.95 25.04 _ 1.001 226.95 0.9941 0．93E-3 0.38

SC
Logistic 5.66E+14 218.84 25.57 _ 1.001 218.84 0.9951 0．78E-3 0.36
Lognormal 270.69 47.43 97.49 0.169 1.367 64.81 0.9946 0．93E-3 0.36
Rayleigh 1176.53 61.56 5.87 _ 1.088 68.92 0.9954 0．84E-3 0.41
Weibull 1.2E+11 98.11 79.89 4.708 1.33

U 171.21 0.9953 0.77E-3 0.33

N
Figure 4(f) demonstrates the probability density function of different DAEMs. The
logarithms of Logistics and Lognormal were small, though the dispersion values were higher than
A
other probability density functions. It is reasonable that all the ε (%) values obtained by F-test
were lower than 0.41. The reaction process⁡𝑘(𝑇) = 𝐴exp(−𝐸/𝑅𝑇)⁡was calculated using different
M

DAEM kinetic parameters, and the results for the logarithm of the same on both sides of the
equation are shown in Figure 5(a). The slope of the Weibull model was parallel to the slope of the
ED

single-step reaction model and iso-conversional method, computed using the Friedman method
and KAS method. It demonstrates that the apparent activation energy of the pyrolysis of oil sands,
as obtained by the Weibull model, was the closest to the single-step reaction model, the Friedman
method and the KAS method, while the energy compensation effect of kinetic parameters obtained
PT

from different DAEMs is shown in Fig. 5(b). In this case, the correlation was as high as 0.999.
40
（ a）
E

Gaussian DAEM
（ b） 35
Logistic DAEM
Log-normal DAEM
30 Rayleigh DAEM
CC

30
Weibull DAEM
one-step model
20 Friedman method 25 lnA=1.7902×10-4E-5.08313
KAS method
ln(A)/ln(s-1)

Pearson's r=0.99924
lnk

20 Gaussian DAEM
Logistic DAEM
A

10
Log-normal DAEM
15 Rayleigh DAEM
Weibull DAEM
0
10

-10 5
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 60 80 100 120 140 160 180 200 220 240
-3 -1
1/T/10 K E/KJ·mol-1

Figure 5 (a) Arrhenius plots of different models; (b) Different DAEM model kinetic compensation
3.5 Optimum model
Based upon above results, it cannot be done well to determine which PDF corresponds to the
DAEM most efficiently for calculating the pyrolysis kinetics of oil sands. Therefore, the Akaike
Information Criteria was proposed to compare the kinetic parameters of different models [45], which
is given by Eq. (20).
AIC = 2 · P + N𝑑 · ln⁡(RSS/N𝑑 ) (20)
△𝑤 = 𝐴𝐼𝐶𝑤 − 𝐴𝐼𝐶𝑚𝑖𝑛 (21)
Different DAEMs have different P and Nd values, due to which, the best kinetic model for
calculating the pyrolysis of oil sands was determined using Eq. (24). The⁡△𝑤 >10 was considered

T
to represent an inappropriate kinetic model, while the calculation of oil sands’ pyrolysis kinetic

IP
model was determined using the most appropriate average conversion rate obtained from the
apparent activation energy and △𝑤 .
Table 4 Akaike Information Criteria (AIC) value for different DAEMs

R
Models Mean RSS AIC △𝑤 AIC

SC
Gaussian DAEM 0.93E-3 -689.2 18.31
Logistic DAEM 0.78E-3 -707.5 0
Lognormal DAEM 0.93E-3 -687.2 20.29
Rayleigh DAEM
Weibull DAEM
0.84E-3
0.77E-3
U
-700.0
-706.4
7.52
1.17
N
From Table 4, the △𝑤 AIC values of Logistic, Rayleigh and Weibull DAEM models were
A
found to be less than 10. However, in the distribution of reactivity, it is not the use of the logistic
distribution function to fit the shape of the reaction profile, as Cai and Liu [46] have studied.
M

Compared with the kinetic parameters of oil sands presented in Table 4, the apparent activation
energy (E0=171.21 kJ/mol) calculated using the Weibull model, was the closest compared to the
ED

activation energy of E0=186.14 kJ/mol that was calculated using the single-step reaction, the
activation energy E0=186.14 kJ/mol calculated using the Friedman method and the activation
energy of E0=162.95 kJ/mol calculated using the KAS method. Therefore, Weibull distributed
activation energy is the most suitable for the study of oil sands’ pyrolysis kinetics.
PT

4 Conclusions
E

The pyrolysis kinetic reaction mechanism of oil sands with single-step reaction was found to
be f(α)=(1-α)5.21. The apparent activation energy was E0 =186.14kJ/mol, while the pre-exponential
CC

factor was A0 =1.56E13 s-1. There is an energy compensation effect between the E and A. However,
the apparent activation energy is not constant, and varies continuously in pace with the change in
conversion rate α. Iso-conversional method was used to compute the kinetic parameters of oil
A

sands. The Friedman method was used to compute the activation energy E0 = 185.05 kJ/mol, while
the corresponding pre-exponential factor was 1.29e +11 s-1. The KAS method was used to
compute the activation energy E0 = 162.95kJ/mol, while the corresponding pre-exponential factor
was 1.34e+6 s-1. The kinetic parameters exhibited energy compensation effect. The kinetic
parameters of the iso-conversional method were almost the same as those of the single-step
reaction, though the most probable mechanism was not the single-step reaction mechanism. The
pyrolysis kinetic parameters of oil sands were computed using DAEMs with Logistic, Lognormal,
Gaussian, Rayleigh and Weibull probability density functions, and the energy compensation
effects were obtained for the calculated kinetic parameters of oil sands’ pyrolysis. The Weibull
DAEM calculated the apparent activation energy of E0=171.21 kJ/mol, which was the closest to
that of the kinetic parameters computed using the single-step reaction, the Friedman method and
the KAS method. It is found that the Weibull DAEM is the most suitable choice for calculating the
pyrolysis kinetics of oil sands based upon the comprehensive F-test and Akaike Information
Criteria (AIC) test.

T
Acknowledgements: The authors are grateful to the Natural Science Foundation of China (No. 51676032);
The science and technology innovation and development projects of Jilin city (20166006).

IP
References

R
[1] Zhang Chunhe，Zheng Xiufen．New economic growth propellant for oil sands，J．China Mining，16（2007）

SC
37-39．
[2] Xu Xiuqiang，Wang Hongyan，Zhang Dewen，Ge Chi-xin，Zhang Fan-ju，Cao Zu-bin．Research progress
on the development and utilization of oil sands，J．Liao Ning Chemical，（2008）268-271．
[3]

U
Yang Hongqiang，Ding Mingshan，Hu Bin，Ren Suli．Extraction of Asphalt from Oil Sand with
Water-Assisted Solvent Method，J．Progress in Chemical Industry，33（2014）2495-2500 + 2516．
N
[4] Zhang Jianqiang， Li Xingang，Sui Hong．Ionic Liquid Promoted Solvent Extraction of Oil Sand Bitumen，
J．Progress in Chemistry，33（2014）1986-1991．
A
[5] Ma Xiaolong，Zhang Zisheng，Gao Xin，Li Xingang．Present status and review on research and development
M

of oil sands pyrolysis，J．Progress in Chemical Industry，35（2016）484-3490．

[6] Fu Jiamo，Qin Kuang Zong．Geochemical Oil Shale，J．Guangdong Science and Technology Press，1995，
541-551．
ED

[7] Luo Man-yin，Xia De-hong，Liu Weidong，et al．Proceedings of the Boiler，27（2007）73-76．

[8] Li Shuyuan，Wang Jianqiu， Qian Jialin.Research，Development and Utilization of Oil Sands in the World，
J．Chinese & Foreign Energy，5（2011）10-23．
[9] Young Cheol Park，Jin-Young Paek，Dal-Hee Bae et al．Study of Pyrolysis Kinetics of Alberta Oil Sand by
PT

Thermogravimetric Analysis，J．Korean Journal of Chemical Engineering，26（2009）1608-1612．

[10] Wang Qing，Jia Chunxia，et al．Study on combustion characteristics of three oil sands in Indonesia，
E

J．Proceedings of the CSEE，32（2012）23-30．

[11] Ma Yue，Li Shuyuan．Studyof the characteristics and kinetics of oil sand pyrolysis，J．Energy ＆Fuels，
CC

24（2010）1844—1847．
[12] Guo Xiuying，Wang Qing，Jiang Qianqian，et al．Study on pyrolysis characteristics and kinetic models of
oil sands in Indonesia，J．Journal of Northeast Electric Power，32（2012）26-32．
A

[13] Pitt，G. J．The kinetics of the evolution of volatile products from coal，J．Fuel，41（1962）267-274．
[14] Anthony，D. B．Coal Devolatilization and Hydrogasification，J．AICHE J，22（1976）625-656．
[15] Anthony D. B.， Howard J. B．Apparatus for determining high pressure coal-hydrogen reaction kinetics
under rapid heating conditions，J．Review of Scientific Instruments，45（1974）992-995．
[16] R. L. Braun，A. K. Burnham．Analysis of Chemical Reaction Kinetics Using a Distribution of Activation
Energies and Simpler Models，J．Energy & Fuels，1（1987）153-161．
[17] Miura K，Maki T．A simple method for estimating f (E) and k 0 (E) in the distributed activation energy model，
J．Energy Fuel，12（1998）864–9.
[18] Zhang Baosheng，Liu Jianzhong，Zhou Junhu，et al．A new method for solving coal combustion reaction
parameters based on multi-scan rate method，J．Proceeding of the CSEE，29（2009）45-50．
[19] Hu Rongzu，Shi Qizhen，Thermal analysis kinetics．Science Press，Beijing，2001．
[20] Hongpeng Liu．Study on Thermal Kinetic Mechanism of Oil Shale，C．2010 Asia-Pacific Power and Energy
Engineering Conference，2010．
[21] Sergey Vyazovkin，Konstantinos Chrissafis，Maria Laura Di Lorenzo c，et al．ICTAC Kinetics Committee
recommendations for collecting experimental thermal analysis data for kinetic computation ，
J．Thermochimica Acta，590（2014）1–23．
[22] B. Roduit，P. Folly，B. Berger，et al．Evaluating Sadt by Advanced Kinetics-based Simulation Approach，
J．Journal of Thermal Analysisand Calorimetry，Vol．93（2008）153–161．

T
[23] Sergey Vyazovkin，Alan K，Burnham，José M，et al．ICTAC Kinetics Committee recommendations for

IP
performing kinetic computations on thermal analysis data，J．Thermochimica Acta，520（2011）1–19．
[24] Tingting Xu，Feng Xu，Zhiquan Hu．Non-isothermal kinetics of biomass-pyrolysis-derived-tar

R
(BPDT) thermal decomposition via thermogravimetric analysis，J．Energy Conversion and Management，
138（2017）452–460．

SC
[25] R. L. Braun，A. K. Burnham．Analysis of Chemical Reaction Kinetics Using a Distribution of Activation
Energies and Simpler Models，J．Energy & Fuels，1（1987）153-161．
[26] Gotor F，Criado J，Malek J，et al．Kinetic analysis of solid-state reactions: the universality of master plots

U
for analyzing isothermal and non-isothermal experiments．Phys. Chem. A 104（2000）10777–10782．
[27] Non-linear Least Squares，http://en.wikipedia.org/wiki/Non-linear-least-squares．Accessed 7 Aug 2016．
N
[28] Zhihua Chen，Jianshen Lei，Yunbei Li，Xianfa Su，Zhiquan Hu，Dabin Guo．Studies on thermokinetic of
Chlorella pyrenoidosa devolatilization via different models，J．Bioresource Technology，224（2017）320–
A
327．
M

[29] Soria Verdugo A，Garcia Hernando N， Garcia Gutierrez L，Ruiz-Rivas U．Analysis of biomass and sewage
sludge devolatilization using the distributed activation energy model，J．Energy Convers Manage，65（2013）
239–44．
ED

[30] Cai J，Wu W，Liu R．An overview of distributed activation energy model and its application in the pyrolysis
of lignocellulosic biomass，J．Renew Sustain Energy Rev，36（2014）236–46．
[31] Soria-Verdugo A，Goos E，Arrieta Sanagustin J，Garcia Hernando N．Modeling of the pyrolysis of biomass
PT

under parabolic and exponential temperature increases using the distributed activation energy model，
J．Energy Convers Manage，118（2016）223–30．
[32] Colin R. hillips，Nagib I. Haidar，Yiu C. Poon．Kinetic models for the thermal cracking of Athabasca
E

bitumen: The effect of the sand matrix，J．Fuel，（1985）64，678–692．

[33] Lin. Liang C.，De. Milind D.，Hanson. Francis V. Oblad，Alex G.．Kinetics of tar sand pyrolysis using a
CC

distribution of activation energy model，J．American Institute of Chemical Engineers，36（1990）1585–1588．

[34] Seader J. D，Coronella C. J，Bhadkamkar A. S，et al．Extraction of bitumen from Western oil sands by an
energy-efficient thermal method ， R ． Technical Final Report for U.S. Dept ． Energy ，
A

No．DE-FC21-93MC30256-99，（1997）229-236．
[35] Hayashitani M，Bennion D. W，Donnelly J. K，et al．Thermal cracking of Athabasca bitumen，R．AGARD
conference proceedings in oil sands of Canada-Venezuela symposium，（1997），233-247．
（1989）
[36] Strausz O P．Bitumen and Heavy Oil Chemistry，M．AOSTRA Technical Handbook on Oil Sands，
33-73．
[37] WANG Qing，JIANG Qian-Qian，JIA Chun-Xia， WU De-Yin．New Exploration on Pyrolysis Kinetics of
Oil Sands in Indonesia，J．Proceeding of the CSEE，32（2012）127-132．
[38] R.E. Lyon ，N. Safronova，J. Senese， S. Stoliarov．Thermokinetic model of sample response in
nonisothermal analysis，J．Thermochim Acta，545（2012）82–89．
[39] Alan K. Burnham ． Global Chemical Kinetics of Fossil Fuels: How to Model Maturation and
Pyrolysis．Springer International Publishing，2017．
[40] Vyazovkin S．Evaluation of the activation energy of thermally stimulated solidstate reactions under an
arbitrary variation of the temperature，J．Comput Chem，18（1997）393–402．
[41] Vyazovkin S，Dollimore D．Linear and nonlinear procedures in iso-conversional computations of the
activation energy of thermally induced reactions in solids，J．Chem Inf Comput Sci，36（1996）42-5．
[42] Meng Meng．Extraction of organic matter from Tumuji sands，D．Doctoral Dissertation in chemical
technology，Dalian University of Technology，2007．

T
[43] Meng Meng ， Haoquan Hu ， Qiumin Zhang ， et al ． Pyrolysis Behaviors of Tumuji Oil Sand by

IP
Thermogravimetry (TG)and in a Fixed Bed Reactor[J]．Energy & Fuels，21（2007）2245-2249．
[44] A.K. Burnham，R.L. Braun．Global kinetic analysis of complex materials，J．Energy Fuels, (1999)13，1–

R
22．
[45] Burnham K，Anderson D．Multimodel inference understanding AIC and BIC in model selection，J．Sociol

SC
Methods Res，32（2004）261–304．
[46] J. Cai，R. Liu．Weibull mixture model for modeling non-isothermal kinetics of thermally stimulated
solid-state reactions: application to simulated and real kinetic conversion data，J．Phys. Chem. B，111（2007）
10681–10686．
U
N
A
M
ED
E PT
CC
A