Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

J Therm Anal Calorim

DOI 10.1007/s10973-013-3505-5

Pyrolysis model of oil sand using thermogravimetric analysis


Qing Wang • Chunxia Jia • Qianqian Jiang •

Yin Wang • Deyin Wu

Received: 4 June 2013 / Accepted: 29 October 2013


 Akadémiai Kiadó, Budapest, Hungary 2013

Abstract In this paper, the derivative thermogravimetry with those of the separated stages, which prove the reli-
(DTG) curve displays an overlapping peak which consists ability of the two-component parallel reaction model.
of two subpeaks for the major oil-producing stage of
Indonesian oil sand. This study was conducted to supple- Keywords Oil sand  Pyrolysis 
ment the insufficiencies of the previous studies, which Thermogravimetric analysis  Overlapping peak
neglected or oversimplified the processing of the overlap-
ping peak and did not determine the kinetic mechanism.
First of all, we assume a two-component parallel reaction Introduction
model to describe the mass-loss process of the major oil-
producing stage (overall stage) to obtain the separated Thermal methods have been extensively used for the
stages. Then, we present an asymmetric multi-peak fitting recovery and upgrading processes of oil sand bitumen to
method, bi-Gauss, to separate the overlapping peak. Based obtain the main product: synthetic crude oil. These pro-
on the results of the bi-Gauss method, the kinetic mecha- cesses, including in situ combustion, surface-mining ther-
nism was obtained using the integral master plot method, mal cracking and thermal conversion (or coking), are
followed by the determination of the activation energy by aimed at heating the oil sand, to vaporize the volatile
the integral iso-conversional non-linear (NL-INT) method. compounds, pyrolytically decompose the larger non-vola-
The results show that the bi-Gauss method can produce tile molecules into lighters, refine hydrocarbons and obtain
satisfactory fitting results. Moreover, the separated stages residual carbon (char or coke) [1–3]. Therefore, it is nec-
follow single mechanism; while the overall stage does not essary to study the pyrolysis kinetics of oil sand for a more
follow a single mechanism and the mechanisms of the front comprehensive understanding of the pyrolysis process and
and back segments of the overall stage are in accordance mechanism so that the results could be used to guide the
recovery and upgrading processes, design the retorting and
combustion reactors, and improve the quality of thermal
byproducts.
Q. Wang  Q. Jiang  Y. Wang
In recent years, with the improvement of thermoanalysis
Engineering Research Centre of Oil Shale Comprehensive
Utilization, Ministry of Education, Northeast Dianli University, technique, thermal analysis of fuels has been extensively
Jilin City 132012, Jilin, China used as a means of determining the kinetic parameters of
e-mail: rlx888@126.com samples pyrolysis [4–7]. Oil sand pyrolysis has typically
been studied using thermogravimetric analysis (TG).
C. Jia (&)
North China University of Electric Power, Beijing 102206, Although there are certain differences between the pyro-
China lysis processes of oil sands from different locations, the
e-mail: jiachunxia_215@126.com process can be divided into 2 [8–10] or 3 [11–15] stages
prior to reaching 600 C. Oil sand includes a wide range of
D. Wu
Indra Boedijono, Indadi Group, Taman Jati Baru Barat 1, components; therefore, the number of parallel, competitive,
Jakarta 10150, Republic of Indonesia and consecutive reactions in the pyrolysis process is quite

123
Q. Wang et al.

numerous. It is impractical to develop a chemical reaction [32] functions. However, DTG curves of the major oil-
model that covers all the components in the pyrolysis producing stage in this paper cannot be fitted using sym-
process comprehensively. Thus, the pseudo-components metric functions, i.e., neither Gaussian nor Lorenzian
that efficiently simplify the real pyrolysis process have function, because of the asymmetric properties observed.
been widely accepted. Barbour et al. [16] proposed the Fortunately, the theory [23, 26, 27] and empirical results
following global reaction to describe the oil sand pyrolysis [33–36] of bi-Gauss method have been suitably applied for
process: Bitumen ! Oil þ Gas þ Coke. This reaction is the asymmetric peak in analytical chemistry; however,
also known as an overall single-component reaction model. little application for TG has been reported. Therefore, we
Other reactions belong to the multi-component reaction consider the simple asymmetry function, bi-Gauss, which
model. For example, Rosenvold et al. [11] and Lin et al. is a modification of the Gaussian function to try to separate
[13] realized that the light and heavy organics that the overlapping peak. Then, we calculate the activation
decompose one after the other represent the reaction of energy for the separated stages and the overall stage by the
maltene and asphaltene. Hayashitani [17] assumed that oil integral iso-conversional non-linear (NL-INT) method and
sand bitumen consists of asphaltene and maltene only and the two-component parallel reaction model, respectively.
used a four pseudo-component reaction model to describe Finally, we use the integral master plot method to select the
the oil sand pyrolysis process. Phillips et al. [18] used a six most probable mechanism function for the separated stages
pseudo-component reaction model to describe the solid and the overall stage. The procedures and methods pre-
phase reaction of oil sand below 693 K. Seader and sented in this paper could be applicable to a much wider
coauthors [19] presented a more complex reaction model range of similar kinetic problems.
by combining the solid phase and gas phase reactions of oil
sand with the results of Lin and Phillips. Starting from its
original overall first-order kinetic [15, 16, 20] based on the Experimental
overall single-component reaction model, many optimized
methods [8–10, 12–14, 20–22] have been developed later Materials
to estimate the kinetic parameters for the entire pyrolysis
process (or certain stage) based on the multi-component Oil sand samples, labeled KA BUNGKA (S1) and LAW-
reaction model. ELE ST3 (S2) for this study, were obtained from Indone-
The derivative thermogravimetry (DTG) curves of oil sia. Prior to use, the samples were crushed and screened to
sand samples in this paper contain two clear subpeaks at the desired particle sizes. The proximate analysis, ultimate
the major oil-producing stage. The two subpeaks are so analysis, heating value, and Fischer assay were conducted
close that they have partly overlapped. After reviewing the
DTG curves of other oil sands, only the Alberta oil sand at
Table 1 Properties of Indonesian oil sands
350–600 C showed a similar behavior [15]; however, the
author made no explanation or treatment to those subpeaks. Properties S1 S2
In general, each peak represents a pyrolysis reaction of the Proximate analysis (mass%, ad)
corresponding pseudo-component. Therefore, it seems Moisture/% 1.30 1.06
unreasonable to describe the two subpeaks using an overall Ash/% 46.68 48.03
single-component reaction model. What’s more, another Volatile/% 52.02 50.91
deficiency is the lack of the determination of kinetic Fixed carbon/% a
0.00 0.00
mechanism according to plenty of literatures. Thus, the Ultimate analysis (mass%, daf)
purpose of this paper is to separate the overlapping peak
Carbon/% 55.59 55.43
based on a more suitable two-component parallel reaction
Hydrogen/% 4.54 5.48
model, and then try to determine the kinetic mechanism of
Oxygen/%a 36.95 36.12
the pyrolysis of Indonesian oil sands.
Nitrogen/% 0.79 0.75
In general, there are some kinds of mathematical sepa-
Sulfur/% 2.73 2.22
ration methods of overlapping signals in analytical chem-
Fischer assay data (mass %, ad)
istry such as the multi-peak fitting (deconvolution),
Oil/% 20.41 18.91
Fourier-transform and wavelet transformation, etc. Various
Water/% 1.40 1.50
methods are based on the selection of elementary functions
Char/% 74.07 76.24
[23–27]. More recently, the multi-peak fitting as the most
Gas/%a 4.12 3.35
popular method has been successfully applied in the field
Heating value/kJ kg-1 (ar) 9,798.50 9,494.42
of biomass pyrolysis to separate the overlapping peak of
a
the DTG curve by the Gaussian [28–31] and Lorenzian By difference

123
Pyrolysis model of oil sand

according to the ASTM standard. The properties of the separate overlapping peaks in the DTG curve. The basic
samples are listed in Table 1. idea of this method is as follows: first, choose an appro-
priate function to describe the DTG curve of the subpeak
Thermogravimetric analysis and select three initial parameters: peak height, peak
In this research, thermogravimetry (TG:DTG) experiments position, and full width at half maximum (FWHM) to form
were conducted with a SDT-Q600 thermogravimetric a three-parameter matrix P. Second, apply the Levenberg–
analyzer at heating rates of 5, 15, and 25 C min-1 in a Marquardt algorithm to obtain the matrix P with repeated
range of ambient temperature to 550 C under a steady iterations until the sum of squares of the differences
flow of nitrogen (80 L min-1). In all cases, the average between the experimental and fitted data SS (Eq. 1) meets
initial sample mass was about 10 mg, and the particle size the minimum value.
was 0–0.2 mm. Xn

Three oil sands samples exhibit similar patterns of SS ¼ wi ½yi  fi ðxi ; PÞ2 ð1Þ
i¼1
thermal decomposition as pyrolysis temperature was
increased. The analysis of these curves demonstrates that where wi is the weighted factor and n is the number of
there are three stages of mass-loss profile best illustrated iterations.
in the derivative curves. Although many chemical and In the above non-linear least square fitted method, the
physical processes contribute to the observed mass loss fitted result would not converge if inappropriate initial
of oil sand, three basic stages encompass the conversion: parameters were selected. Therefore, another evaluation
the lower temperature region, from ambient up to function v2 (reduced Chi squared) should be used to
approximately 130 C, produced mass loss which has ascertain the number of iterations. In this way, the calcu-
been attributed to the loss of moisture, which accounts lations would be performed with the actual number of
for almost 2 % of sample mass. The subsequent stage iterations when the tolerance (the relative variation of twice
took place at 150–520 C, wherein represents a slow iterations) is achieved before the original number of iter-
mass loss at 380–500 C and accounts for almost 8 % of ations, which was arbitrarily defined. Here, v2 have various
sample mass, can be assumed as the decomposition of forms with their own wi , as for the no weighting form, each
light hydrocarbons and desorption of organic gases; point has the same weighted factor (wi ¼ 1) and the v2 can
however, a major portion of mass loss with a maximum be defined as:
rate at about 450 C is due to the thermal decomposition Xn
of indigenous bitumen. In the third stage, between 580 v2 ¼ ðyi  fi Þ2 ð2Þ
and 750 C, the mass loss is attributed to the decom- i¼1

position of carbonate minerals. At the same time, the Bi-Gauss multi-peak fitting method
presence of CO2 (which evolved as a product of car-
bonate decomposition) and which at high temperature In theory, the thermal-effect peak of a simple reaction
reacted with residual char to form CO also caused some should be a single peak with a symmetrical shape. The
mass loss. Gaussian or the Lorentzian function is commonly used to
From the DTG curves, it can be seen that heating rate solve this kind of peak. In practice, the real peak is likely
also affects the mass loss at a certain temperature. For not strictly symmetrical, which needs a modified function
example, at the same temperature, the lower the heating to correctly fit the peak. We suggest an asymmetric
rate, the more the volatile material evolved. The effect of Gaussian function bi-Gauss, to describe the overlapping
heating rate to overall mass loss is not clear, but it obvi- peak in the DTG curve. The bi-Gauss function can be
ously shortens the time of decomposition, advanced the expressed by
rate of mass loss. 8 h i
< H exp ðxi  xc Þ2 =2r21 ; xi \xc
fi ¼ h i ð3Þ
Theoretical : H exp ðxi  xc Þ2 =2r2 ; xi [ xc
2

Separation of overlapping peak where H is the peak height, r1 is the left-FWHM, r2 is the
right-FWHM, and r1 6¼ r2 . Different values of FWHM r
Basic idea of multi-peak fitting method indicate that each branch of asymmetric peak can be
described separately by the function with the own value of
The multi-peak fitting method, which increases the signal width parameter [26, 27]. In addition, the parameter values
resolution by a mathematical algorithm based on the can be simply determined from the experimental curve
appropriate function as the peak fitting function, is valid to [33].

123
Q. Wang et al.

Multi-component parallel reaction model The integral form of the kinetic equation in the non-
isothermal pyrolysis is given by
The multi-component parallel reaction model regards the Z T
overall stage as a result of the combination of several GðaÞ  A=b expðE=RT ÞdT ¼ ðA=bÞ  IðE; TÞ ð9Þ
0
independent pyrolysis reactions by neglecting the interac-
tions among pseudo-components; i.e., if we assume that a where GðaÞ is the integral form of the kinetic mechanism
subpeak represents an independent reaction of pseudo- and IðE; TÞ is the temperature integral.
component, the overall stage kinetic parameters can be The kinetic mechanism GðaÞ of the substance is the
obtained by weighting and superimposing the subpeak same at the selected a for different b. In this case, it leads
kinetic parameters. The weighted factor is the ratio of the to the following relationship:
pseudo-component mass loss to the total mass loss. ðA=b1 ÞIðEa ; T1 Þ ¼ ðA=b2 Þ  IðEa ; T2 Þ ¼   
In non-isothermal pyrolysis, the differential form of the ¼ ðA=bn Þ  IðEa ; Tn Þ ð10Þ
kinetic equation of component i is given by
dai =dT ¼ Ai =b expðEi =RT Þfi ðaÞ ð4Þ where n is the number of heating rates. When i 6¼ j,
rearranging Eq. (9) gives
where ai is the conversion, Ai is the apparent pre-expo-   
bj  IðEa ; Ti Þ bi  IðEa ; Tj Þ ¼ 1 ð11Þ
nential factor, b is the heating rate, Ei is the apparent
activation energy, R is the universal gas constant, T is the Superposition of the expansion terms of the Eq. (10), the
absolute temperature, and fi ðai Þ is the differential form of problem is converted into one that determines the most
kinetic mechanism. probable activation energy Ea to meet the following equation:
The weighted factor of component i can be calculated 
X n X n   
using 
XðEa Þ ¼ min b  IðEa ; Ti Þ bi  IðEa ; Tj Þ
 i¼1 j6¼i j
ri ¼ ðMi0  Mif Þ=ðM0  Mf Þ ð5Þ 

X
m
 nðn  1Þ
ri ¼ 1 ð6Þ
i¼1
ð12Þ
where Mi0 and Mif are the initial and final mass of com-
Using the secondary rational approximation of Senum-
ponent i, respectively, M0 and Mf are the initial and final
Yang (Eq. 12) to simply IðEa ; TÞ, an optimization
mass of sample, respectively; m is the number of pseudo-
algorithm can be programed by a FORTRAN software.
component.   
The conversion and activation energy of the overall IðEa ; TÞ ¼ Teu ðu þ 4Þ u2 þ 6u þ 6 ð13Þ
stage can be calculated by
Kinetic mechanism
Xm
a ¼ ðM0  MT Þ=ðM0  Mf Þ ¼ ðri ai Þ ð7Þ
i¼1 The most probable kinetic mechanism can be determined
X
m by the integral master plot method proposed by Gotor [39]
E¼ ðri Ei Þ ð8Þ with the premise of accurate E value. Therefore, the
i¼1 E value obtained by the NL-INT method can be used here.
The integral form of kinetic equation Eq. (8) can also be
where MT is the sample mass at temperature T.
written as
GðaÞ  ðA=bÞ  IðE; TÞ ¼ ðAE=bRÞ  PðuÞ ð14Þ
Calculation of kinetic parameters
Using a reference point at a ¼ 0:5, the following
Activation energy equation can be easily derived from Eq. (13):
GðaÞ=Gð0:5Þ ¼ PðuÞ=Pð0:5Þ ð15Þ
One of the model-free methods, integral iso-conversional
non-linear (NL-INT) method which was developed by where the temperature integral PðuÞ can be expressed by
Vyazovkin [37], is chosen to determine the activation the secondary rational approximation of Senum-Yang:
energy in this paper. This method produces an insignificant   
PðuÞ ¼ eu =u ðu þ 4Þ u2 þ 6u þ 6 ð16Þ
amount of error because it does not use the oversimplified
temperature integral approximation and has no limit to the Substituting the average E value of the overall stage and
validity of u ¼ E=RT, while the linear iso-conversional the corresponding temperature T at the selected a under
method requires that u  13 [38]. different b into the right side of Eq. (15) allows the

123
Pyrolysis model of oil sand

Fig. 1 DTG curves of S1 A 0.05


(a) and S2 (b) oil sands at 100

Mass loss rate/% K–1


various heating rates 0.00

Mass loss/%
–0.05
90

–0.10
5K min–1
80
–0.15
15K min–1
a: 5/K min–1
b: 15/K min–1 25K min–1
–0.20
70 c: 25/K min–1
–0.25
200 300 400 500 600 700 800 900 300 400 500 600 700 800 900
Temperature/K Temperature/K
S1 (A)–TG S1 (A)–DTG

B 100 0.05
0.00

Mass loss rate/% K–1


90 –0.05
Mass loss/%

–0.10
80
–0.15

70 5/K min–1 –0.20


15/K min–1 5K min–1
25/K min–1 –0.25 15K min–1
60 25K min–1
–0.30
50 –0.35
200 300 400 500 600 700 800 900 300 400 500 600 700 800 900
Temperature/K Temperature/K
S2 (B)–TG S2 (B)–DTG

experimental master plot of PðuÞ=Pð0:5Þ versus a to be evaporation of moisture. The mass loss in this stage is
drawn; then, arbitrary a values (e.g., 0.05, 0.10…0.95, 1) can approximately 2 % of the sample mass, slightly larger than the
be substituted into appropriate integral kinetic mechanisms moisture contents according to the Fischer assay (c.f. Table 1).
to produce the standard master plot of GðaÞ=Gð0:5Þ versus a. • In the temperature range of 160–350 C, S2 has a very
If the experimental plot coincides with (or approximates to) small but much wider peak at 200–300 C while the
the standard plot, the chosen GðaÞ is correct. peak of S1 drops continuously and joins the next stage.
We assume that the oil sand bitumen consists of
Pre-exponential factor asphaltene and maltene only, and the mass loss of this
stage attributes to the decomposition of maltene. As
To determine the pre-exponential factor, first substitute the summarized in reaction formula (a), no coke is formed
GðaÞ and E into Eq. (13); then, take the natural logarithm [40]. It is a low-degree decomposition stage, and
of both sides of the equation. After transposing and rear- approximately 8 % sample mass has been decomposed.
ranging, the equation is given by Maltene ! Oil þ Gas ðaÞ
ln PðuÞ þ lnðE=bRÞ ¼ ln½GðaÞ  ln A ð17Þ
• In the temperature range of 350–520 C, two
By taking the linear regression of ½ln PðuÞ þ lnðE=bRÞ pronounced peaks at 400 and 460 C at 5 C min-1
versus ln½GðaÞ, one can obtain the pre-exponential factor could be observed. The mass loss is approximately
from the straight line of intercept  ln A. 20 % of the sample mass, which is in accordance with
the oil contents of the Fischer assay. Hence, it is
Results and discussion considered a major oil-producing stage. In addition, we
attributed the first peak to the decomposition of
Analysis of the pyrolysis process asphaltene (primary reaction b) and the second peak
to the secondary decomposition of the product of the
The pyrolysis process can be described by the DTG curve, primary reaction (secondary reaction c).
as shown in Fig. 1. As the two samples display similar Asphaltene ! Heavy oil þ Coke þ Gas ðbÞ
pyrolysis characteristics both the mass-loss processes can
be divided into three stages: Heavy oil ! Light oil þ Coke þ Gas ðcÞ
• In the temperature range of 80–160 C, one obtains a It is worth noting that the second peak at 25 C min-1 of
pronounced mass-loss peak that corresponds to the S1 (c.f. Fig. 1a) is a very weak peak due to short exposure

123
Q. Wang et al.

Fig. 2 Results of bi-Gauss A 12 B 18


multi-peak fitting for S1 (a) and experimental experimental

–1
–1
S2 (b) at different heating rates: 10 fitted 16 fitted

Conversion rate *10 /K


Conversion rate *10 /K
peak 1
a 5 K min-1, b 25 K min-1,

3
3
14 peak1
peak 2
and c 25 K min-1 8 12 peak2
10
6
8
6
4
4
2 2
0
600 620 640 660 680 700 720 740 760 600 620 640 660 680 700 720 740 760
Temperature/K Temperature/K
(a) 5 K min–1
12

–1
–1

experiment 14

Conversion rate *10 /K


Conversion rate *10 /K

10 fitted experiment
12

3
3

peak1 fitted
8 peak2 10 peak1
peak2
8
6
6 ⽿2
⽿1
4 4
2
2
0
0 600 620 640 660 680 700 720 740 760 780
620 640 660 680 700 720 740 760 780 800
Temperature/K Temperature/K
(b) 25 K min–1

14

–1
12
–1

experiment

Conversion rate *10 /K


Conversion rate *10 /K

fitted 12 experiment
3
3

10 fitted
peak1 10
peak2 peak1
8 8 peak2
⽿1
6 6
⽿1 ⽿2
⽿2 4
4
2
2
0
0 620 640 660 680 700 720 740 760 780 800
620 640 660 680 700 720 740 760 780 800
Temperature/K Temperature/K
(c) 25 K min–1

time to a certain temperature at the higher heating rate. It is temperature, which should volatilize at a lower
known that fragmentary peaks may lead to inaccurate temperature, so that the DTG curve shifts to a higher
results (e.g., inaccurate multi-peak fitting results, activation temperature region.
energy results by the model-free method). Thus, the
appropriate choice of heating rate for the Indonesian oil Separation of overlapping peak for the major oil-
sand is important. The original purpose of this study is to producing stage
apply the kinetic parameters calculated by TG for the
design of oil sand retorting reactor; therefore, the heating As can be seen from Fig. 1, the existence of two peaks
rates of 5, 15, and 25 C min-1 have been chosen based on overlapped in the major oil-producing stage in the tem-
experience. Because of the unique characteristics of perature range of 350–520 C can be described by the two-
Indonesian oil sand, a comprehensive study on the low- component parallel reaction model. The multi-peak fitting
or medium-level heating rate is necessary to achieve more method is applied to separate the overlapping peak.
accurate kinetic results. With an increase of the heating
rate, the peak on the DTG curve shifts to a higher Bi-Gauss multi-peak fitting results
temperature region. The mass-loss process happens in a
shorter time at a higher heating rate, which makes the part To use the bi-Gauss multi-peak fitting method, the rela-
of a relatively light organic substance volatilize at a higher tionship of da=dT versus T for the major oil-producing

123
Pyrolysis model of oil sand

Table 2 Characteristic parameters of the bi-Gauss multi-peak fitting method


Sample b Peak1 Peak2 R
a b c c
Trange /K Tpeak /K r1 Trange /K Tpeak /K r2

S1 5 600–760 673 0.5447 665–760 730 0.4553 0.9984


15 620–780 711 0.5474 710–780 757 0.4526 0.9992
25 640–790 723 0.6263 720–790 765 0.3737 0.9993
S2 5 600–760 678 0.4177 670–760 730 0.5823 0.9991
15 620–780 708 0.5545 705–780 755 0.4455 0.9985
25 630–790 730 0.5756 720–790 767 0.4244 0.9996
a
Temperature range
b
Peak temperature
c
Weighted factor of component i, i ¼ 1; 2

stage is plotted firstly. Each step of peak fitting procedures


240
can be easily processed in the peak fitting module (PFM)
which is built in an OriginPro 7.5 software as given in the 220

literature [41]. Due to space limitations, we only show the 200 S1-peak2
fitting results at 5 C min-1.
E/kJ mol–1
180
As shown in Fig. 2, the separated stages (peak1 and S2-peak2
160
peak2) as well as the fitted curve are obtained. The fitting
curves of bi-Gauss method almost coincide with the 140 S1-peak1

experimental curves. The fitting correlation coefficient R 120 S2-peak1


reaches to 0.9984 and 0.9991, respectively.
100

80
Pyrolysis characteristics of the major oil-producing stage 0.2 0.3 0.4 0.5 0.6 0.7 0.8
α
The characteristic parameters of pyrolysis for the separated
Fig. 3 E-a curves of NL-INT method
stages obtained by the bi-Gauss method are listed in
Table 2. The table displays the pyrolysis characteristics
with the increasing of the heating rate as follows: First, the From the E–a curves in Fig. 3, it illustrates that the dis-
subpeaks shifted to higher temperatures and became shar- tribution of E fluctuates slightly within the main range;
per; furthermore, the temperature difference between two therefore, the pyrolysis reaction in the separated stages fol-
subpeaks decreased (the original temperature difference is lows single kinetic mechanism. In addition, the trends of the
57 and 52 C for S1 and S2 at 5 C min-1); finally, the E–a curves for peak1 or peak2 of both samples exhibit almost
proportion of the area of peak2 decreased, which may be the same trends, suggesting that the kinetic mechanisms for
interpreted as the delay in the secondary decomposition peak1 or peak2 of both samples should be similar. It also can
reaction of the product of asphaltene due to incomplete be observed that, for each sample, the average E value for
heating absorption. These results, which are in accordance peak1 (as listed in Table 4) is close to each other, but that of
with the academic analysis, demonstrate the credibility of the S1 for peak2 is apparently smaller than that of S2. This is
the fitting results by the bi-Gauss method [33–35]. to say that the pyrolysis reaction in the temperature range of
peak2 for S2 occurs faster. Furthermore, the E value of peak2
Calculation of kinetic parameters for the major oil- is higher than that of peak1 in any samples, which confirms
producing stage the academic analysis that the occurrence of the secondary
decomposition reaction is more slowly than that of the pri-
Kinetic parameters of separated stages mary reaction [20].
Once the average value of E is known, the value can
The NL-INT methods are introduced to calculate the dis- be used to determine the kinetic mechanism using the
tribution of activation energy curve E–a at the main con- integral master plot method. In this paper, we only
version range of 0.2–0.8. The average value of E can be investigate the kinetic mechanism at 5 C min-1. This is
calculated to determine the kinetic mechanism using the because the pyrolysis process of the oil sand represents
integral master plot method. the intrinsic mechanism at the thermal equilibrium state

123
Q. Wang et al.

Table 3 Commonly used mechanisms for the pyrolysis of solid state material
Mechanism Symbol f ðaÞ GðaÞ

Two-dimensional diffusion (Valensi Eq.) D1 ½ lnð1  aÞ1 a þ ð1  aÞ lnð1  aÞ


h i
1 1
h i
1 2
Two-dimensional diffusion (Jander Eq., n ¼ 2) D2 1
ð1  aÞ2 1  ð1  aÞ2 1  ð1  aÞ2
h i
1 1
h i
1 2
Three-dimensional diffusion (Jander Eq., n ¼ 3) D3 3 2

2 ð1  aÞ 1  ð1  aÞ 1  ð1  aÞ3
3 3

h i1
Three-dimensional diffusion (G-B Eq.a) D4 3 13 1  23 a  ð1  aÞ3
2

2 ð1  aÞ 1
h i1 h i2
Three-dimensional diffusion (Z-L–T Eq.b) D5 13
ð1  aÞ3 1
4 1
3
2 ð1  aÞ ð1  aÞ 1
3

Nucleation and nuclei growth (A-E Eq.c, n ¼ 1) A1 1a  lnð1  aÞ


 aÞ½ lnð1  aÞ2
1 3
Nucleation and nuclei growth (A-E Eq., n ¼ 1:5) A1.5 2
3 ð1
2
½ lnð1  aÞ
aÞ½ lnð1  aÞ1
2
Nucleation and nuclei growth (A-E Eq., n ¼ 2) A2 1
2 ð1  ½ lnð1  aÞ
1 1
Exponential nucleation (Mample Eq., n ¼ 2) P2 2a a2
Order type chemical reaction OC ð1  aÞ2 ð1  aÞ1 1
a
Ginstling–Brounstein equation
b
Zhuralev–Lesokin–Tempelman equation
c
Avrami–Erofeev equation

Fig. 4 Experimental and A 3 3


B 3 3
D3 D2
standard master plots of D1 D3
D2 D5
separated stages of S1-peak1 P2 D1
experimental master plot experimental master plot A2 D4
(a), S1-peak2 (b), S2-peak1 (c), 2 2 P(u)/P(0.5)
2 P2 2

G(α )/G(0.5)
G(α )/G(0.5)

standard master plot standard master plot

P(u)/P(0.5)
and S2-peak2 (d) A1 OC

1 1 1 1

0 0 0 0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7 0.8
α α

C 3
D2
3
D 3
A1.5
3
OC D2
D3 D1
D3 D1
experimental master plot P2
P2 2 experimental master plot 2
2
G(α )/G(0.5)
P(u)/P(0.5)
G(α )/G(0.5)

P(u)/P(0.5)
standard master plot standard master plot

1 1 1 1

0 0 0 0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7 0.8
α α

Table 4 Kinetic parameters of separated stages


at low-level heating rates. The process does not show
effects, such as diffusion, that are caused by thermal Sample Peak E/kJ mol-1 Mechanism A/s-1 R
hysteresis. In addition, the effects of various heating rates
S1 Peak1 145 ± 27 OC 4.48E?10 0.99997
on the characteristics of the components differ from each
Peak2 188 ± 12 D3 2.03E?12 0.99980
other due to the extremely complex composition of oil
S2 Peak1 146 ± 27 OC 4.59E?10 0.99976
sand. We realize that the multiple heating rates may lead
Peak2 159 ± 13 D2 3.66E?11 0.99999
to various kinetic mechanisms. This could serve as future
research direction.
The significant master plots, as listed in Table 3 [42], are the determination of E and A are listed in Table 4. The results
given in Fig. 4 with experimental plot. The kinetic param- show that both peak1 of S1 and S2 follow a chemical reaction
eters and the linear correlation coefficient R generated during (OC) mechanism, while peak2 of S1 follows the three-

123
Pyrolysis model of oil sand

Fig. 5 Experimental and A3 3


B 3 3
D3 D1
standard master plots of the A2
D3
A2 D1
major oil-producing stage of S1 experimental master plot A1.5 D2 A
P2 1.5 2

G(α )/G(0.5)
(a) and S2 (b) 2 2 experimental master plot

G(α )/G(0.5)

P(u)/P(0.5)
2

P(u)/P(0.5)
standard master plot P2
standard master plot
OC

1 1 1 1

OC
OC

0 0 0 0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.2 0.3 0.4 0.5 0.6 0.7 0.8
α α

dimensional diffusion (D3) mechanism, and peak2 of S2 The high level of similarity in the kinetic mecha-
follows the two-dimensional diffusion (D2) mechanism. nism between separated stages and the overall stage
demonstrates the reliability of the kinetics analysis,
Kinetic parameters of the major oil-producing stage especially for the two-component parallel reaction
model.
The E values of the separated stages were substituted into
Eq. (7) to obtain the E value of the overall stage at
5 C min-1. The results of S1 and S2 are 165 and Conclusions
154 kJ mol-1, respectively. Studies on the high-degree
decomposition of different oil sands at low heating rates (5 In this paper, we focus on the problem of the overlapping
or 10 C min-1) show varying E values. For example, the peak in the DTG curve for the pyrolysis of Indonesian oil
E value of Tumuji oil sand [9] calculated by Coats-Redfern sand. We realize that the two-component parallel reaction
method is 161 kJ mol-1, that of the Asphalt Ridge oil sand model may be more appropriate to describe the major oil-
[12] calculated by the Anthony-Howard method is producing stage, in which the DTG curve displays two
251 kJ mol-1, that of the Whiterocks oil sand [13] pronounced and asymmetric subpeaks. Based on the sat-
obtained by the modified Anthony-Howard method is isfying results of the bi-Gauss method, we determined the
121 kJ mol-1, that of the Alberta oil sand [15] by the kinetic mechanism by the NL-INT method with the integral
Coats-Redfern method is 246 kJ mol-1, and that of the master plot method. The results of the kinetic parameters
Karamay oil sand [22] obtained by a combination of the show that the E values of the overall stage are 165 and
Freeman and Coats-Redfern method is 56.7 kJ mol-1. 154 kJ mol-1 for S1 and S2, respectively. The separated
Relative to the oil sands all over the world, the activation stages corresponding to the pseudo-component follow
energy of Indonesian oil sand has a moderate value. single kinetic mechanism due to the small range variation
Based on the result of the E value, the kinetic mecha- of Evalue, while the mechanism of the overall stage cer-
nism for the major oil-producing stage for S1 and S2 tainly does not follow a single mechanism as expected. If
could be determined using the integral master plot the overall stage was divided using the reference point at
method. As shown in Fig. 5, the mechanism of the overall a ¼ 0:5, both of the front segments of S1 and S2 follow the
stage certainly does not follow a single mechanism, which chemical reaction mechanism. The back segment of S1
is consistent with expectations. If the master plots are follows the three-dimensional diffusion mechanism (Jander
divided using the reference point at a ¼ 0:5, the overall Eq.), and the back segment of S2 follows the two-dimen-
stage mechanisms of the front and back segment are sional diffusion mechanism (Jander Eq.), which are in
identical with those of the separated stages. In addition, accordance with the mechanisms of the separated stages.
the weighted factors of peak1 and peak2 are located The unity of the kinetic mechanisms of the overall stage
approximately at ri ¼ 0:5 (c.f. Table 2). Once again, the and the subpeaks demonstrates the reliability of the two-
front segments of S1 and S2 follow the OC mechanism, component parallel reaction model. Furthermore, the bi-
while the back segment of S1 follows the D3 mechanism Gauss multi-peak fitting method could be applicable to a
and the back segment of S2 follows the D2 mechanism. much wider range.

123
Q. Wang et al.

Acknowledgements The authors would like to thank Mr. Deyin Wu 19. Oblad AG, Bunger JM, Dahlstrom DA, Deo MD, Fletcher JV,
from Indonesia for the supply of oil sand samples. Natural Science Hanson FV, et al. The extraction of bitumen from Western oil
Foundation of China (51276034). sands. Final report. 1989 July–1993 Sep. Salt Lake City: Uni-
versity of Utah. Department of Chemical and Fuels Engineering
and Department of Metallurgical Engineering. 1994 Mar. Report
No.: DOE/MC/26268-3822. Contract No: DE-FC21-
References 89MC26268. Sponsored by the U.S. Department of Energy,
Office of Fossil Energy. Morgantown Energy Technology
1. McLendon TR, Bartke TC. Tar sand. U.S. Department of Energy, Center.
Morgantown Energy Technology Center, Morgantown. 1990 Jan. 20. Lin LC. The kinetics of the pyrolysis of tar sands and of the
Report No.: DOE/METC-91/0274. Sponsored by the U.S. combustion of coked sands (Dissertation). Salt Lake City: Uni-
Department of Energy, Washington, DC. versity of Utah; 1988.
2. Wilcott HR. A chemical reaction kinetic study into the low 21. Meng M. Extraction of organic substance from Tumuji oil sand
severity pyrolysis of Athabasca bitumen (dissertation). Calgary: (dissertation). Dalian: Dalian University of Technology; 2007.
University of Calgary; 1998. 22. Luo MY, Xia DH, Liu WD, Liu XG. A study on the thermo-
3. Oil Sands Discovery Centre. Facts about Alberta’s oil sands and gravimetric characters of oil sands and its application to thermal
its industry (online). Edmonton: Government of Alberta. Culture oil-extracting system. Boil Manuf. 2007;1:73–6.
and Community Services; 2002 (cited 2009 Nov 3). http://history. 23. Grushka E, Myers MN, Giddings JC. Moments analysis for the
alberta.ca/oilsands/resources/docs/facts_sheets09.pdf. discernment of overlapping chromatographic peaks. Anal Chem.
4. Liu XW, Feng YL, Li HR, Zhang P, Wang P. Thermal decom- 1970;42(1):21–6.
position kinetics of magnesite from thermogravimetric data. 24. Li JW. Development and evaluation of flexible empirical peak
J Therm Anal Calorim. 2012;107(1):407–12. functions for processing chromatographic peaks. Anal Chem.
5. Han B, Chen Y, Wu Yl, Hua Dr, Chen Z, Feng W, Yang MD, Xie QH. 1997;69(21):4452–62.
Co-pyrolysis behaviors and kinetics of plastics–biomass blends 25. Jung KH, Shin YH. Deconvolution studies of multicomponent
through thermogravimetric analysis. J Therm Anal Calorim. 2013;1–9. systems on chromatography. Bull Korean Chem Soc.
6. Oliveira LE, Giordani DS, Paiva EM, Castro HF, Silva MP. Kinetic 1986;7(5):403–5.
and thermodynamic parameters of volatilization of biodiesel from 26. Romanenko SV, Stromberg AG. Classification of mathematical
babassu, palm oil and mineral diesel by thermogravimetric analysis models of peak-shaped analytical signals. J Anal Chem.
(TG). J Therm Anal Calorim. 2013;111(1):155–60. 2000;55(11):1024–8.
7. Sasmal S, Goud V, Mohanty K. Determination of salutary 27. Romanenko SV, Stromberg AG. Modelling of analytical peaks:
parameters to facilitate bio-energy production from three peaks modifications. Anal Chim Acta. 2007;581(2):343–54.
uncommon biomasses using thermogravimetric analysis. J Therm 28. Trick KA, Saliba TE, Sandhu SS. A kinetic model of the pyro-
Anal Calorim. 2013;111(3):1649–55. lysis of phenolic resin in a carbon/phenolic composite. Carbon.
8. Li SY, Wang JQ, Tan HP, Wu ZL. Study of extraction and 1997;35(3):393–401.
pyrolysis of Chinese oil sands. Fuel. 1995;74(8):1191–3. 29. Ma W, Wang S, Cui JP, Zhang ST, Fan BC, He YZ. Thermal
9. Meng M, Hu HQ, Zhang QM, Li X, Wu B. Pyrolysis behaviors of decomposition kinetic model of phenolic resin. Acta Phys Chim
Tumuji oil sand by thermogravimetry (TG) and in a fixed bed Sin. 2008;24(6):1090–4.
reactor. Energy Fuels. 2007;21(4):2245–9. 30. Li R, Jin BS, Zhong ZP, Fu XF, Jia XR. Mechanism research on
10. Ma Y, Li SY. Study of the characteristics and kinetics of oil sand rice husk bio-oil combustion based on parallel reaction model.
pyrolysis. Energy Fuels. 2010;24(3):1844–7. Proc CSEE. 2009;29(29):89–95.
11. Rosenvold RJ, DuBow JB, Rajeshwar K. Thermophysical char- 31. Li R, Jin BS, Zhong ZP, Fu XF. Research on biomass pyrolysis
acterization of oil sands. 4. Thermal analyses. Thermochim Acta. three-pseudocomponent model by Gaussian multi-peaks fitting.
1982;58(3):325–31. Acta Energy Sol Sin. 2010;31(7):806–10.
12. Turner TF, Nicherson LG. Pyrolysis of asphalt ridge tar sand. 32. Katsikas L, Popović IG. Improvement to the Flynn-Wall method
Laramie: Western Research Institute. 1986 Aug. Report No.: of determining apparent activation energies of the thermal
DOE/FE/60177-2315. Contract No.: FC21-83FE60177. Spon- degradation of polymers. J Phys Chem B. 2003;107(30):
sored by the U.S. Department of Energy, Office of Fossil Energy. 7522–3.
Morgantown Energy Technology Center, Laramie Project Office. 33. Buys TS, De Clerk K. Bi-Gaussian fitting of skewed peaks. Anal
13. Lin LC, Deo MD, Hanson FV, Oblad AG. Kinetics of tar sand Chem. 1972;44(7):1273–5.
pyrolysis using a distribution of activation energy model. AIChE 34. Dacheux N, Aupiais J, Courson O, Aubert C. Comparison and
J. 1990;36(10):1585–8. improvement of the determinations of actinide low activities
14. Sun N, Zhang QM, Guan J, He DM. Pyrolysis of Zhalaiteqi oil using several a liquid scintillation spectrometers. Anal Chem.
sands under nitrogen atmosphere. J Fuel Chem Technol. 2000;72(14):3150–7.
2007;35(2):241–4. 35. Tw Yu, Peng HS. Quantification and deconvolution of asym-
15. Park YC, Paek JY, Bae DH, Shun D. Study of pyrolysis kinetics metric LC–MS peaks using the bi-Gaussian mixture model and
of Alberta oil sand by thermogravimetric analysis. Korean J statistical model selection. BMC Bioinformatics. 2010;11(1):
Chem Eng. 2009;26(6):1608–12. 559–68.
16. Barbour RV, Dorrence SM, Vollmer TL, Harris JD. Pyrolysis of 36. Nebelung C, Baraniak L. Simultaneous determination of 226Ra,
233
Utah tar sands: products and kinetics. Am Chem Soc Div Fuel U and 237Np by liquid scintillation spectrometry. Appl Radiat
Chem. 1976;21(6):278–89. Isot. 2007;65(2):209–17.
17. Hayashitani M. Thermal cracking of Athabasca bitumen (dis- 37. Vyazovkin S. Modification of the integral isoconversional
sertation). Calgary: University of Calgary; 1978. method to account for variation in the activation energy. J Com-
18. Phillips CR, Haidar NI, Poon YC. Kinetic models for the thermal put Chem. 2001;22(2):178–83.
cracking of Athabasca bitumen: the effect of the sand matrix. 38. Hu RZ, Gao SL, Zhao FQ, Shi QZ, Zhang TL, Zhang JJ. Thermal
Fuel. 1985;64(5):678–91. analysis kinetics. 2nd ed. Beijing: Science Press; 2008.

123
Pyrolysis model of oil sand

39. Gotor FJ, Criado JM, Melek J, Koga N. Kinetic analysis of solid- 41. Mydlová J, Krupčı́k J, Korytár P, Sandra P. On the use of com-
state reactions: the universality of master plots for analyzing puter assisted resolution of non-separable peaks in a congener
isothermal and nonisothermal experiments. J Phys Chem A. specific polybrominated diphenyl ether capillary gas chromato-
2000;104(46):10777–82. graphic analysis. J Chromatogr A. 2007;1147(1):95–104.
40. Strausz OP, Jha KN, Montgomery DS. Chemical composition of 42. Hu RZ, Shi QZ. Dynamic thermal analysis. Beijing: Science
gases in Athabasca bitumen and in low-temperature thermolysis Press; 2001. p. 125–36.
of oil sand, asphaltene and maltene. Fuel. 1977;56(2):114–20.

123

You might also like