You are on page 1of 13

A NEW KINETIC MODEL FOR PYROLYSIS OF ATHABASCA

BITUMEN
Punitkumar R. Kapadia, Michael S. Kallos and Ian D. Gates*
Department of Chemical and Petroleum Engineering, Schulich School of Engineering, University of Calgary, 2500 University
Drive NW, Calgary, Alberta, Canada, T2N 1N4

Pyrolysis of bitumen is a key contributor to gas production in in situ combustion and in situ gasification recovery processes, yet a detailed reaction
scheme, that includes a breakdown of products into the most abundant gas components, that is simple enough to be used in detailed thermal-
reactive simulation models, does not exist. Here, we present a novel reaction scheme for pyrolysis of Athabasca bitumen that was developed and
calibrated against four experimental data sets (65 data points) over a temperature range from 130 to 422◦ C. The new model was then verified
by comparing its thermogravimetric behaviour against published literature.

Keywords: in situ gasification, pyrolysis, Athabasca bitumen, synthesis gas, oil sands reservoirs

INTRODUCTION ammonia production for agricultural fertilisers. In Alberta alone,


roughly 21 million m3 of hydrogen is consumed to upgrade about

O
ver 170 billion barrels of bitumen resource is considered
125 000 m3 of bitumen per day in the oil sands mining projects
recoverable in Western Canada (Alberta Energy, 2010).
in Northern Alberta (ACR, 2004; ERCB, 2008). Since hydrogen
This is a huge volume of petroleum, second only to Saudi
does not exist in the atmosphere in any significant quantities, this
Arabia with respect to producible oil reserves, yet current recovery
hydrogen is produced from raw feed stocks, such as natural gas.
and upgrading processes are both energy and emissions inten-
Globally, almost half of the 500 billion m3 produced every year
sive; thus, alternative methods to recover chemical energy from
come from steam reforming of natural gas, with one third from
these reservoirs are required. Currently, all commercial in situ
crude oil and the balance from coal and minor amounts from
bitumen recovery methods raise oil mobility, mainly by lower-
electrolysis (Suresh et al., 2007). With consumption at such high
ing its viscosity, and move the mobilised oil to a production well.
levels and demand expected to grow with new bitumen upgraders
Given that bitumen viscosity can drop up to 6 orders of magni-
coming online, there is a pressing need to develop more sustain-
tude by raising its temperature by 200◦ C, most in situ recovery
able hydrogen production methods.
processes use steam alone or steam and solvent to mobilise bitu-
The combined drives to improve hydrogen generation and bitu-
men. In current practice in Alberta, Canada, steam is generated by
men recovery have motivated us to explore in situ gasification of
natural gas combustion. Typically, about 300+ Sm3 (S means at
bitumen to recover its chemical energy in the form of synthesis
standard conditions) of natural gas are required (for steam gener-
gas. This means that the recovered energy vectors are methane
ation) per m3 bitumen recovered (ACR, 2004). Not only are large
and hydrogen rather than bitumen. With in situ hydrogen gen-
amounts of methane combusted yielding large amounts of emitted
eration and sufficiently high temperatures, if oil recovery occurs,
greenhouse gases, but water is consumed as well. The majority
there is also potential for in situ bitumen upgrading.
of recovered bitumen is then upgraded to convert it to a synthetic
crude oil that can be processed in conventional refineries. Dur-
ing upgrading, about 170 Sm3 of hydrogen are required per Sm3 ∗ Author to whom correspondence may be addressed.
of bitumen upgraded (ACR, 2004). Often, this hydrogen is also E-mail address: ian.gates@ucalgary.ca
sourced from natural gas. Can. J. Chem. Eng. 9999:1–13, 2012
Hydrogen is an integral part of the global petroleum energy © 2012 Canadian Society for Chemical Engineering
DOI 10.1002/cjce.21732
system. Worldwide, the petroleum industry accounts for half of Published online in Wiley Online Library
global hydrogen consumption, with most of the other half used for (wileyonlinelibrary.com).

| VOLUME 9999, 2012 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1 |


In situ gasification (ISG) of Athabasca bitumen has several
potential advantages over the traditional routes of steam-based
bitumen recovery and upgrading. First, water consumption will
be significantly lower if no steam generation from surface is done.
Second, if combustion is used to generate heat in the reservoir,
heat losses are reduced (below that of surface-generated steam)
and a fraction of greenhouse gases generated by combustion will
be sequestered directly within the reservoir (either dissolved in
oil or water or reacted with reservoir rock). Third, little or no
natural gas is consumed on surface and thus greenhouse gas emis-
sions due to steam generation or upgrading are reduced. Fourth,
gas emissions originating from ISG are single source rather than
several distributed sources (recovery process, upgrader and refin-
ery), as is the case with the traditional route, which makes carbon
capture easier. Fifth, if upgraded oil is realised at surface, the oil
may be sufficiently upgraded for handling in conventional refiner-
ies, thus eliminating the upgrading step and its associated energy
needs and emissions.
There are multiple reactions during combustion of Athabasca
bitumen, which compete to yield net hydrogen production. Pyrol-
ysis (or thermolysis or thermal cracking; Hayashitani et al., 1977),
aquathermolysis (Hyne et al., 1977), water–gas shift (Davis
and Jennings, 1984) and coke gasification reactions have been Figure 2. Modified pyrolysis reaction scheme for Athabasca bitumen
reported to generate hydrogen gas during in situ combustion of (present study).
heavy oil. From the results of an in situ combustion oil recov-
ery pilot operating in an oil sands reservoir, Hajdo et al. (1985) must occur outside of the combustion zone where there is no
suggested that the largest fraction of hydrogen generated is by the oxygen.
coke gasification reaction followed by the water–gas shift reaction. A key step for designing bitumen ISG processes for hydrogen
ISG requires heat, and one way to generate heat is through generation is the reaction scheme, together with associated kinetic
in situ combustion (ISC) by injecting air or oxygen into the constants for all of the reaction zones within the process. Here,
formation. ISC takes place over two temperature and oxygen we focus on the pyrolysis zone and present a novel yet practi-
consumption ranges: first, low temperature oxidation (LTO) and cal reaction scheme to describe pyrolysis of Athabasca bitumen,
second, high temperature oxidation (HTO). For bitumen, LTO which can be used in thermal-reactive reservoir simulators for ISG
takes place between 150 and 300◦ C where the range of oxy- process design.
gen consumption rates is lower, whereas HTO occurs between
380 and 800◦ C with higher oxygen consumption rates. Com-
bustion experiments reveal that ahead of the combustion zone, HYDROGEN GENERATION FROM PYROLYSIS:
conduction heats the oil sand, and because of the absence of EXPERIMENTAL DATA
oxygen, pyrolysis reactions convert maltenes to asphaltenes and There are a few key experimental lab-scale studies that have
asphaltenes to coke with generation of hydrogen, methane and been performed which can yield kinetic data for a pyrolysis
carbon oxides. With ISC-based heating, one key limitation of model. Egloff and Morrell (1927) carried out Athabasca bitumen
ISG is that generated hydrogen and injected oxygen can react thermolysis and obtained gasoline, demonstrating the capabil-
to form water. This implies that hydrogen generation reactions ity to upgrade bitumen by using pyrolysis. In their experiments,
the products of thermal cracking were distillate, noncondens-
able gas and coke. To evaluate conversion of heavy oils to
a Maltenes lower-boiling, gasoline-containing crudes under low temperature
cracking, McNab et al. (1952) investigated cracking of Athabasca

Table 1. List of components and properties (Belgrave et al., 1993;


Coke Asphaltenes Gas Perry and Green, 1997)

Molecular weight Critical temperature Critical pressure


Component Mw (kg/gmol) Tc (◦ C) Pc (kPa)
b Maltenes

Maltenes 0.4067 618.85 1478


Asphaltenes 1.0928 903.85 792
HMWG 0.04141 — —
CH4 0.01604 −82.55 4600
Coke Asphaltenes H2, CH4, CO, CO2, H2S, HMWG CO 0.02801 −140.25 3496
CO2 0.04401 31.05 7376
Figure 1. Comparison of (a) existing thermal cracking (Belgrave et al., H2 0.002016 −239.96 1315
1993) and (b) modified thermal cracking (present study) reaction H2 S 0.03408 100.0 8937
schemes (stoichiometrically unbalanced) for Athabasca bitumen. Coke 0.01313 — —

| 2 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 9999, 2012 |


Table 2. Strausz et al. (1977) experimental data for composition of gases from Athabasca oil sand pyrolysis

Temperature Time Methane CO CO2 H2 S


(◦ C) (h) (×10−8 mol/kg oil sand) (×10−8 mol/kg oil sand) (×10−8 mol/kg oil sand) (×10−8 mol/kg oil sand)

130 5.5 252.0 18 200 78 500 16.2


11.0 358.0 2598 99 600 46.7
16.5 447.5 3220 109 240 88.2
22.0 474.1 3392 111 440 97.1
27.5 513.0 3615 113 770 104.3
170 5.5 832.0 6360 229 000 2340.0
11.0 1420.0 9620 — —
16.5 2176.0 12 420 254 600 2685.0
22.0 2472.0 13 176 261 850 2846.0
27.5 2783.0 13 950 267 700 3076.0
210 5.5 2470.0 10 700 271 000 431.0
12.5 3910.0 13 300 289 600 1380.0
18.0 5610.0 14 840 301 000 2182.0

Table 3. Hayashitani et al. (1977) experimental data for thermal cracking of Athabasca bitumen

Conversion of Athabasca bitumen

Temperature (◦ C) Time (h) Maltenes (wt%) Asphaltenes (wt%) Coke (wt%) Gas (wt%)

360 1.91 84.24 14.58 0.44 0.74


4.80 83.71 13.64 1.32 1.33
7.85 83.09 11.64 3.60 1.67
11.13 81.95 12.01 4.20 1.95
13.78 80.29 9.93 7.40 2.38
17.33 79.38 9.56 8.00 3.06
19.83 78.71 9.03 9.02 3.24
23.78 78.10 8.23 9.85 3.81
397 0.79 83.88 15.26 0.19 0.67
1.01 83.14 12.83 2.57 1.46
1.64 81.22 10.68 5.70 2.40
3.00 77.91 9.14 9.36 3.59
4.29 74.48 9.06 12.34 4.12
5.81 72.82 8.36 13.64 5.19
7.82 68.58 8.18 16.86 6.38
10.35 65.20 7.74 20.28 6.78
422 1.25 75.54 8.87 11.44 4.15
1.60 71.74 8.51 14.09 5.66
2.50 65.07 7.01 19.62 8.30
3.70 57.42 5.72 26.23 10.63
5.60 52.33 2.04 33.37 12.26

Product gas composition

Temperature (◦ C) Time (h) Hydrogen (mol%) Methane (mol%) CO (mol%) CO2 (mol%) H2 S (mol%) HMWG (mol%)

360 13.78 19.3 43.7 5.3 2.2 1.4 28.1


17.33 14.0 45.8 3.2 1.1 2.6 33.3
19.83 18.2 38.8 3.0 3.9 4.6 31.5
23.78 13.3 28.2 4.6 3.1 5.4 45.4
397 0.79 0.0 58.4 4.1 9.2 1.7 26.6
1.01 0.0 45.2 5.0 5.9 2.1 41.8
1.64 0.0 44.9 2.6 1.0 1.7 49.8
3.00 0.0 47.9 1.1 2.6 1.8 46.6
4.29 0.6 50.2 0.7 2.3 1.3 44.9
7.82 0.5 46.5 0.0 1.7 1.0 50.3
10.35 0.5 53.0 1.0 1.5 0.9 43.1
422 1.25 4.4 37.5 0.0 0.9 7.3 49.9
1.60 2.9 41.2 0.5 0.5 5.9 49.0
2.50 1.8 40.6 0.2 0.4 6.2 50.8
3.70 2.6 41.7 0.0 0.2 5.8 49.7
5.60 2.7 42.0 0.0 0.2 5.4 49.7

| VOLUME 9999, 2012 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 3 |


tar sand in a series of experiments and found kinetic parameters
Table 4. Phillips et al. (1985) experimental data for thermal cracking
for the overall thermal cracking reaction of Athabasca bitumen.
of Athabasca oil sand
Henderson and Weber (1965) conducted mild thermal cracking
experiments on seven different crude oils—six from convention- Temperature Time Maltenes Asphaltenes Coke Gas
ally produced oil reservoirs and one from Athabasca tar sands—to (◦ C) (h) (wt%) (wt%) (wt%) (wt%)
study the effect of cracking on viscosity and specific gravity. They
estimated a time–temperature dependent viscosity relationship 360 4 66.93 25.82 5.64 1.61
to predict the degree of viscosity reduction when crude oils are 8 64.48 22.59 8.89 4.04
12 60.42 21.19 13.43 4.96
subject to thermal methods. Speight (1970) described structural
16 60.11 16.50 17.92 5.47
changes in Athabasca bitumen during cracking in which destruc-
20 58.09 15.88 20.03 5.99
tive distillation of bitumen yielded light oil or distillate, resins, 24 53.79 13.48 24.02 8.71
coke and gas. Bunger et al. (1976) pyrolised three Utah and one 400 2 69.80 13.00 13.77 3.43
Athabasca bitumens at a heating rate of about 5◦ C/min to an end 4 58.47 11.12 25.05 5.35
point of 625◦ C and found overall reaction parameters for bitumen 6 56.32 8.48 27.70 7.49
cracking. The products from these experiments were gases (C5 8 49.38 6.59 35.95 8.08
and lighter), liquid condensate (C6 to 535◦ C boiling fraction) and 10 46.01 6.26 37.67 10.06
coke. Barbour et al. (1976) performed experiments where they 420 1 66.09 13.29 15.53 5.09
thermally cracked samples of four Utah tar sands and a Cana- 2 57.68 8.19 25.44 8.69
3 51.29 6.61 30.60 11.51
dian oil sand. The products of the reactions were upgraded oil,
4 44.44 6.17 35.80 13.58
gas and coke. Strausz et al. (1977) carried out low temperature
5 36.93 6.13 41.60 15.34
thermolysis of oil sand, asphaltene and maltene and analysed the
composition of produced gases. They found that activation ener-
gies of the product formed had low values, suggesting catalytic reaction model for mathematical modelling of ISC of Athabasca
effects of the mineral matter present. bitumen. Yang and Gates (2009) extended Belgrave’s kinetic
Hayashitani et al. (1977, 1978) conducted an experimental oxidation model to include gas combustion reactions and re-
study of Athabasca bitumen thermal cracking to develop a reac- calibrated the kinetic parameters against a single Athabasca
tion model. The products of cracking were separated into six bitumen combustion tube experiment. To validate the general
pseudo-components: coke, asphaltenes, heavy oils, middle oils, application of the kinetics, they used the fitted parameters to
light oils and gases. Phillips et al. (1985) used this pseudo- successfully predict the behaviour of four other independent
component approach for cracking Athabasca bitumen in the Athabasca bitumen experiments. Kapadia et al. (2011) extended
presence of a sand matrix to determine its catalytic effect on Yang and Gates’ model to include hydrogen generation and
thermal cracking. Millour et al. (1985) used Hayashitani’s data consumption reactions. The Belgrave et al. (1993) model is a
to develop thermal cracking kinetic model for Athabasca bitu- good compromise between the number of pseudo-components
men which incorporated coke, asphaltenes, maltenes and gas as and the complexity of the reaction scheme. It contains four oil
pseudo-components. Lin et al. (1987) examined extent-of-reaction pseudo-components (maltenes, asphaltenes, coke and gas) and
effects, that is how the composition of pseudo-components change six reactions for thermal cracking, LTO and HTO of Athabasca
during the reaction process, and determined reaction models bitumen. Although the maltenes and asphaltenes are historically
with associated kinetic parameters. Mazza and Cormack (1988) based on solubility classification, here they are treated as two oil
subjected SARA (Saturates, Aromatics, Resins, Asphaltenes) sep- pseudo-components. The maltenes, asphaltenes and coke pseudo-
arated fractions of Athabasca bitumen to liquid phase thermal components, as described in the literature, are based on solubility
cracking and developed kinetic models to describe the behaviour classes. The maltenes pseudo-component is obtained from solu-
of these fractions during the reactions. Despite these studies, at bility in pentane, whereas the separation of coke and asphaltenes
this point, there are no pyrolysis reaction schemes that handle
generic Athabasca bitumen that are practical enough to be used
in a thermal-reactive reservoir simulator for in situ process design. Table 5. Millour et al. (1985) experimental data for thermal cracking
of Athabasca oil sand

COMPONENTS AND REACTIONS Temperature Time Maltenes Asphaltenes Coke Gas


(◦ C) (h) (wt%) (wt%) (wt%) (wt%)
The representation of bitumen by pseudo-components is neces-
sary given available data but it hides complex chemistry. For 360 2 76.88 13.99 4.56 4.56
example, bitumen as a single pseudo-component represents many 4 76.72 12.74 5.15 5.39
individual components, which react at elevated temperatures to 6 76.00 12.57 5.71 5.71
yield different products. Also, the relative amounts of the products 8 75.96 12.56 5.86 5.62
16 75.78 11.86 6.54 5.81
formed by the reactions change with temperature. This means that
397 1 77.26 12.41 5.76 4.57
each reaction for a bitumen pseudo-component to each product
2 75.78 10.86 7.52 5.85
must be treated independently, each with its own reaction rate 4 73.16 9.50 9.50 7.84
law. 6 70.17 9.08 10.73 10.02
Reaction parameters, fluid and thermal properties (including 8 67.77 8.06 12.32 11.85
molecular weight, critical properties, thermal properties and vis- 420 1 74.12 7.67 11.26 6.96
cosity vs. temperature correlations) of the pseudo-components 2 64.84 8.15 16.01 11.01
are uncertain, so there is a balance to be set by the number of 4 56.00 6.54 22.55 14.91
the pseudo-components and the complexity of the reaction sys- 6 51.17 3.74 26.50 18.59
tem. Belgrave et al. (1993) described a unified pseudo-mechanistic 8 48.73 3.06 28.65 19.56

| 4 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 9999, 2012 |


Figure 3. Arrhenius plots for Reactions 1–4.

Figure 4. Arrhenius plots for Reactions 5–8.

| VOLUME 9999, 2012 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 5 |


Table 6. Global kinetic parameters for thermal cracking of Athabasca
bitumen

Reaction Frequency factor (A; h−1 ) Activation energy (E; J/gmol)

1 4.891 × 1016 2.358 × 105


2 1.296 × 1014 1.897 × 105
3 9.281 × 103 9.963 × 104
4 1.069 × 107 1.122 × 105
5 2.650 4.892 × 104
6 7.809 × 10−2 2.313 × 104
7 1.411 × 105 9.721 × 104
8 2.572 × 1014 2.013 × 105
The values for Reactions 1–8 are the values determined after matching the
experimental data listed in Tables 2–5.

pseudo-components are based on the solubility in benzene or


Figure 5. Arrhenius plot for the generation of pseudo-component ‘‘gas’’
toluene (Strausz et al., 1977; Behar and Pelet, 1984; Millour et al., from asphaltenes as per Figure 1b.
1985). This characterisation scheme is similar to previous stud-
ies of pyrolysis of asphaltenes and vacuum residue (Yasar et al.,
2001). listed in Figure 2 occur in parallel, but each one has its own kinetic
parameters. Each individual reaction does not conserve elemen-
tal moles, but collectively Reactions 2–8 conserve the moles of
PYROLYSIS REACTIONS all elements involved in the reaction. The properties of the com-
Belgrave et al. (1993) and Yang and Gates (2009) used a ponents and pseudo-components are listed in Table 1. Properties
thermal cracking reaction scheme for Athabasca bitumen as of all but HMWG pseudo-components were taken from Belgrave
described by Figure 1a. According to their reaction scheme, pyrol- et al. (1993), whereas properties of components were taken from
ysis is described by three reactions: conversion of maltenes to Chapter 2 of Perry and Green (1997). The molecular weight of
asphaltenes, asphaltenes to coke and asphaltenes to a single gas the HMWG pseudo-component was estimated by using composi-
component. Their scheme is limited, since it does not account for tion of combustible C2+ gas components evolved during thermal
individual gas components, such as hydrogen, methane, carbon cracking experiments of Hayashitani et al. (1977).
oxides and hydrogen sulphide, but considers them collectively
as a single ‘‘gas’’ pseudo-component. However, many studies KINETIC MODEL
have shown that these gas components result from thermal
For the thermal cracking reaction scheme depicted in Figure 2,
cracking experiments (McNab et al., 1952; Hayashitani et al.,
the rate of either consumption or production for each component
1977; Strausz et al., 1977; Phillips et al., 1985). Hence, the
was modelled by using simple first order rate law. This results in
new pyrolysis reaction scheme developed here separates the
the following set of first order differential equations:
pseudo-component ‘‘gas’’ into six gas components: hydrogen,
methane, carbon monoxide, carbon dioxide, hydrogen sulphide
d[Maltenes]
and HMWG (Heavy Molecular Weight Gas). HMWG is a new = −k1 [Maltenes] (1)
dt
pseudo-component in the reaction scheme that represents all com-
bustible gases with greater than two carbon atoms, that is C2+ . d[Asphaltenes]
Figure 1b describes this modified thermal cracking scheme for = k1 [Maltenes] − k2 [Coke] − k3 [H2 ] − k4 [CH4 ]
dt
Athabasca bitumen. In total, in the reaction scheme proposed
− k5 [CO] − k6 [CO2 ] − k7 [H2 S] − k8 [HMWG] (2)
here, there are eight reactions and nine components, displayed
graphically in Figure 2. Here, Athabasca bitumen consists of the d[Coke]
two pseudo-components Maltenes and Asphaltenes. Given that = k2 [Asphaltenes] (3)
dt
Asphaltenes is a pseudo-component representing many individ-
ual components, when it reacts, it can be converted into different d[H2 ]
products through a parallel reaction system. Reactions 2–8 as = k3 [Asphaltenes] (4)
dt

Table 7. Comparison of global kinetic parameters (present study) with other studies in available literature

Refs. Reaction 1 Reaction 2

Present study Frequency factor (h−1 ) 4.891 × 1016 1.296 × 1014


Activation energy (J/gmol) 2.358 × 105 1.897 × 105
Hayashitani et al. (1978) Frequency factor (h−1 ) — 5.076 × 1014
Activation energy (J/gmol) — 1.946 × 105
Phillips et al. (1985) Frequency factor (h−1 ) — 4.176 × 1016
Activation energy (J/gmol) — 1.743 × 105
Belgrave et al. (1993) Frequency factor (h−1 ) 3.273 × 1016 1.463 × 1013
Activation energy (J/gmol) 2.347 × 105 1.772 × 105

| 6 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 9999, 2012 |


d[CH4 ]
= k4 [Asphaltenes] (5)
RESULTS AND DISCUSSION
dt
Model and Parameter Estimation
d[CO]
= k5 [Asphaltenes] (6) To determine the kinetic parameters for Reactions 1–8, the new
dt reaction model was fitted against 65 data points from four
Athabasca bitumen pyrolysis experiments listed in Tables 2–5.
d[CO2 ] Then, these parameters were used to compare experimental data
= k6 [Asphaltenes] (7)
dt from thermogravimetric analysis of Athabasca bitumen from
published literature. In this work, the experimental data from
d[H2 S] Hayashitani et al. (1977), Phillips et al. (1985) and Millour
= k7 [Asphaltenes] (8)
dt et al. (1985) on thermal cracking of Athabasca bitumen were
used to estimate parameters of Reactions 1 and 2. The kinetic
d[HMWG] parameters for Reactions 3–8 were estimated by using thermal
= k8 [Asphaltenes] (9) cracking of Athabasca bitumen experimental data from Strausz
dt
et al. (1977) and Hayashitani et al. (1977). One key assumption
where k1 –k8 are the forward reaction rate constants for the Reac- that has been made is that the composition of the Athabasca
tions 1–8, respectively. It was assumed that rate constant values bitumen samples used in all the experiments was the same.
are always positive and non-zero. The rate constants depend This is a limitation of the study, but it must be made, given
on temperature, according to the Arrhenius relationship (Fogler, that there is no additional information on the composition of
2006): the bitumens used. For the set of ordinary differential equations
given by Equations (1)–(9), the initial mass fractions of maltenes,
asphaltenes and coke were equal to 0.8004, 0.1994 and 0.0000,
k = A e−E/RT (10) respectively, for Hayashitani et al. (1977); 0.8293, 0.1707 and
0.0000, respectively, for Strausz et al. (1977); 0.7780, 0.2220 and
where A is the pre-exponential factor, E is the activation energy 0.0000, respectively, for Phillips et al. (1985); and 0.78, 0.18
and R is the universal gas constant. and 0.04, respectively, for Millour et al. (1985). Initial values

Figure 6. Comparison of bitumen pyrolysis experimental data (Strausz et al., 1977) (data points) and predictions from the new kinetic model (lines).

| VOLUME 9999, 2012 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 7 |


of the gas fractions for all gas components were set equal to plots for Reactions 1 and 2 also reveal that with increasing tem-
zero. perature, the kinetic rate constants for Reactions 1 and 2 from
The procedure used to estimate kinetic parameters was as different thermal cracking experiments merge with each other.
follows: This phenomenon was earlier observed for low temperature oxi-
dation of Athabasca bitumen (Millour et al., 1987) but is also
(1) Estimate the initial value for reaction rate constants at a given true for the thermal cracking of Athabasca bitumen. As shown in
temperature. For the first run, values of all the rate constants Figure 5, this phenomenon was also observed for the reaction, rep-
were set equal to 0.02 h−1 , although other values were tested resenting conversion of asphaltenes to pseudo-component ‘‘gas,’’
to ensure the algorithm converged to a global solution. as shown in Figure 1a.
(2) Estimate the conversion of components and pseudo- The Arrhenius plot for Reaction 3 (hydrogen generating reac-
components at times during which experimental data are tion) shows scatter, which arises from scatter observed in
available by using Equations (1)–(9). At each time, calculate actual hydrogen generation experimental data, listed in Table 3
the deviation between the model predictions and experimen- (Hayashitani et al., 1977). Though the reaction describing hydro-
tal data. gen generation through pyrolysis constitutes a key step for in situ
(3) Calculate the global residual (equal to the sum of the square gasification, it has already been established that most of the hydro-
of all of the deviations at all times for all experiments) by gen produced at pilot scale is due to coke gasification followed by
using Equation (11): the water-gas shift reaction (Hajdo et al., 1985). Arrhenius plots
 for Reactions 4–7 showed some degree of linearity, even though
GR = (yi − Yi )2 (11) they were obtained from different sets of experimental data. The
i

where GR is global residual, yi are calculated concentration


data and Yi are experimental concentration data.
(4) Minimise the global residual by using the lsqcurvefit function
within the MATLAB (2009) software package. This numer-
ical algorithm is a subspace trust-region method based on
the interior-reflective Newton method for parameter estima-
tion. In each Newton iteration, the linear system of equations
is solved by using the preconditioned conjugate gradient
method (Coleman and Li, 1996). To ensure that the fitting
algorithm was converging to the global solution, several runs
were conducted with different initial values of the rate con-
stants. In all cases, the algorithm converged to the same
values.
(5) Repeat Steps 1 and 4 for different temperatures.
(6) The reaction constants obtained from Steps 1–5 at different
temperatures are then fitted to the Arrhenius relationship
given by Equation 10 to calculate the pre-exponential factor
and activation energy.

The results of kinetic parameter estimation for Reactions 1–8 (as


shown in Figure 2) are shown in Figures 3 and 4. In each plot, the
slope of the least square fitted straight line is equal to E/R, and the
intercept is the value of the logarithm of the pre-exponential factor.
The values of the reaction constants for Reactions 1–8 obtained
from the least squares fit are listed in Table 6. From Table 7, kinetic
parameters, as obtained by global match, compare well with those
available in the literature for Reactions 1 and 2 (data for Reactions
3–8 are not available in the literature). The Arrhenius plots for
Reactions 1 and 2 reveal that there is a considerable variability in
the rate constants obtained by different experiments, even though
the experiments were carried out at nearly the same tempera-
ture. There are several reasons for this: first, the bitumen samples
used in different experiments could be from different wells and/or
areas within Athabasca, thus their compositions may be different.
Second, some thermal cracking experiments were carried out in
the presence of the sand matrix, whereas the others were not.
Third, the experimental reactor wall could have played a catalytic
role altering the thermal cracking reaction rate either as a whole
or selectively on only some reactions. Fourth, the different ther-
mal cracking conditions in which all the experiments were carried Figure 7. Comparison of bitumen pyrolysis experimental data
out, for example initial mass of bitumen or oil sands, heating rate, (Hayashitani et al., 1977) (data points) and predictions from the new
product removal methods and age of the samples. The Arrhenius kinetic model (lines).

| 8 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 9999, 2012 |


Arrhenius plot for Reaction 8 showed a good match for all exper- energies obtained from different sources of experimental data
imentally obtained rate constants. The key reason for this is the for Reactions 4–7, as shown in Figure 3. Figures 7 and 8 com-
fact that they were obtained from a single experiment, eliminating pare Hayashitani et al.’s thermal cracking experimental data for
the factors contributing to variability. pseudo-components maltenes, asphaltenes, coke and gas com-
ponents hydrogen, methane, carbon monoxide, carbon dioxide,
Matching of Experimental Data hydrogen sulphide and HMWG with the predictions from the
Comparisons were made between experimentally determined kinetic model. The match has been carried out for experimental
yields of several species through time and predictions of the fit- data at 360, 397 and 422◦ C. Figure 8 shows prediction of carbon
ted kinetic model, shown in Figures 6–10 (65 data points used monoxide composition for 360◦ C only because of high scatter in
to fit the Arrhenius parameters). Overall, the majority of matches experimental data at the higher temperatures and the simulation
are good, although some are better than others. Figure 6 com- would appear to have a high degree of error. The matches between
pares Strausz et al.’s experimental data for the gas analysis during the experimental data and the model are reasonably close. Figure 9
pyrolysis of Athabasca bitumen with the predictions from the cur- compares Phillips et al.’s experimentally determined thermal
rent kinetic model for methane, carbon monoxide, carbon dioxide cracking data for pseudo-components maltenes, asphaltenes, coke
and hydrogen sulphide at 130, 170 and 210◦ C. There is reason- and gas products at 360, 400 and 420◦ C with current model predic-
able agreement between model predictions and experimental data, tions. Here, the gas has been treated as a single pseudo-component
and any differences may be due to slightly different activation because of unavailability of gas composition data (the amounts of

Figure 8. Comparison of bitumen pyrolysis experimental data on gas composition (Hayashitani et al., 1977) (data points) and predictions from the
new kinetic model (lines).

| VOLUME 9999, 2012 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 9 |


individual gas components accounted for in the model proposed that observed during thermogravimetric analysis on Athabasca
here were summed). The comparisons reveal that the matches bitumen (Ritchie et al., 1979; Murugan et al., 2012). The deriva-
are reasonable. Similarly, Figure 10 compares Millour et al.’s ther- tive weight loss of the bitumen resulted initially because of
mal cracking experimental data for pseudo-components maltenes, maltenes conversion to asphaltenes (<450◦ C). Subsequently,
asphaltenes, coke and gas at 360, 400 and 420◦ C with predictions when asphaltenes also started decomposing at higher tempera-
from the model. The results of the model appear to provide a tures, the derivative weight loss increased considerably after going
reasonable match with the experimental data. through a peak temperature of around 450◦ C. Ritchie et al. (1979)
observed the derivative weight loss peak around 425◦ C for the
Chemical Reaction Pathways Athabasca bitumen.
Figure 12 describes the simulation results of rate of evolu-
The ability of the reaction model to match most of the experimen-
tion of different components (Table 1) during thermogravimetric
tal data suggests that parallel reactions occur among components
analysis of Athabasca bitumen (ramp rate equal to 5◦ C/min).
of the bitumen and that these components react under differ-
The gas component’s evolution during thermogravimetric anal-
ent reaction rate laws. Figure 11 describes the simulation results
ysis of Athabasca asphaltenes (Ritchie et al., 1985) is very
of ramped temperature pyrolysis of Athabasca bitumen at the
similar to those obtained here. Initially, maltenes started convert-
rate of 5◦ C/min (initial conditions are 0◦ C and 101.325 kPa, ini-
ing to asphaltenes as per Reaction 1, whereas the asphaltenes
tial mass of bitumen is equal to 1 kg). The trend is typical of

Figure 9. Comparison of bitumen pyrolysis experimental data (Phillips Figure 10. Comparison of bitumen pyrolysis experimental data (Millour
et al., 1985) (data points) and predictions from the new kinetic model et al., 1985) (data points) and predictions from the new kinetic model
(lines). (lines).

| 10 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 9999, 2012 |


Figure 11. Ramped temperature pyrolysis (5◦ C/min) results for Athabasca bitumen percent weight loss and weight loss derivative.

present in original bitumen started converting mostly to carbon experimental data sets (consisting of 65 data points in total) was
oxides because of their low activation energies as compared to attempted to determine kinetic parameters for a set of new gas gen-
the other components. During this time, asphaltenes’ consump- eration reactions from bitumen over temperatures ranging from
tion rate was higher than its production rate from maltenes, 130 to 420◦ C. All experimental data sets were treated equally. To
which resulted in net consumption of asphaltenes. Subsequently, estimate kinetic parameters, the final composition of the produced
when the maltenes decomposition rate increased leading to species of multiple experiments was plotted versus temperature
more asphaltenes being produced, the asphaltenes net production for each reaction. To verify the model, its predictions were then
increased with temperature. When components other than carbon compared to the complete composition versus time behaviour of
oxides started forming due to their higher activation energies, the each experiment. The results reveal that the global match between
net production rate of asphaltenes also started reducing. Finally, the experimental results and model exhibits both examples of
when temperature reached a high enough value, the rates of pro- excellent matches and examples of poorer matches. The poorer
duction of all the components from asphaltenes also increased matches suggest that the reaction scheme is not ‘‘rich’’ enough
considerably. Since there was not enough conversion of maltenes to fully represent the complexity of the kinetics. On the other
to asphaltenes, asphaltenes started reducing after reaching the hand, the variability of the feed bitumen in the experiments may
peak temperature. be an additional factor leading to poorer matches. However, the
Though the proposed reaction scheme captures the experimen- new kinetic model demonstrates reasonable capability to predict
tal observations qualitatively better than quantitatively, it is also product gas volumes and composition when compared to indepen-
required to incorporate the effect of bitumen composition along dent experimental data with a practical reaction model that can be
with sand mineral catalytic role into the reaction scheme. A first implemented in a thermal-reactive reservoir simulator. Together
step would be to vary the ratio of maltenes to asphaltenes in with oxidation and aquathermolysis kinetics, the model can be
the bitumen. Also, as shown in Figures 3–5, all reactions do not used to design in situ processes for in situ hydrogen generation
demonstrate linear behaviour on an Arrhenius plot, suggesting from bitumen.
that a distribution of activation energy should be invoked for the
reaction system.
NOMENCLATURE
[x] concentration of x component, weight fraction
CONCLUSIONS A frequency factor (h−1 )
A new model for pyrolysis of Athabasca bitumen has been devel- E activation energy (J/gmol)
oped which is not only capable of predicting pseudo-components, n order of reaction
but also gas composition evolved during thermal cracking of bitu- Mw molecular weight (kg/gmol)
men. This is an essential element for eventual modelling and Tc critical temperature (◦ C)
design of in situ bitumen gasification processes for hydrogen gen- Pc critical pressure (kPa)
eration. A global match against four Athabasca bitumen pyrolysis R universal gas constant (Pa m3 /gmol K)

| VOLUME 9999, 2012 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 11 |


Figure 12. Ramped temperature pyrolysis (5◦ C/min) results for the evolution of (a) maltenes and asphaltenes, (b) coke, (c) H2 , (d) CH4 , (e) CO,
(f) CO2 , (g) H2 S and (h) HMWG.

REFERENCES Barbour, R. V., S. M. Dorrence, T. L. Vollmer and J. D. Harris,


Alberta Chamber of Resources. Oil Sands Technology Roadmap ‘‘Pyrolysis of Utah Tar Sands—Products and Kinetics,’’
Unlocking the Potential, www.acr-alberta.com/LinkClick. Preprints of Papers—ACS, Division of Fuel Chemistry; 21 (6),
aspx?fileticket=48xNO8LRbKk%3d&tabid=205 (2004). 278–289 (1976).
Alberta Energy Resources Conservation Board, Alberta’s Energy Behar, F. and R. Pelet, ‘‘Characterization of Asphaltenes by
Industry. An Overview, www.energy.gov.ab.ca/Org/pdfs/ Pyrolysis and Chromatography,’’ J. Anal. Appl. Pyrolysis
Alberta Energy Overview.pdf (2010). 7(1–2), 121–135 (1984).

| 12 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 9999, 2012 |


Belgrave, J. D. M., R. G. Moore, M. G. Ursenbach and D. W. Low-Temperature Oxidation of Athabasca Bitumen,’’ J. Can.
Bennion, ‘‘A Comprehensive Approach to In-Situ Combustion Petrol. Technol. 26(3), 24–32 (1987).
Modeling,’’ SPE Adv. Technol. Ser. 1(1), 98–107 (1993). Murugan, P., T. Mani, N. Mahinpey and M. Dong, ‘‘Pyrolysis
Bunger, J. W., S. Mori and A. G. Oblad, ‘‘Processing of Tar Sand Kinetics of Athabasca Bitumen Using a TGA Under the
Bitumens. Part I—Thermal Cracking of Utah and Athabasca Influence of Reservoir Sand,’’ Can. J. Chem. Eng. 90(2),
Tar Sand Bitumens,’’ Preprints of Papers—American Chemical 315–319 (2012).
Society, Division of Fuel Chemistry, 21(6), 147–158 (1976). Perry, R. H. and D. W. Green, ‘‘Perry’s Chemical Engineers’
Coleman, T. F. and Y. Li, ‘‘An Interior, Trust Region Approach for Handbook,’’ 7th ed., McGraw-Hill, Toronto (1997).
Nonlinear Minimization Subject to Bounds,’’ SIAM J. Optim. Phillips, C. R., N. I. Haidar and Y. C. Poon, ‘‘Kinetic Models for
6(2), 418–445 (1996). the Thermal Cracking of Athabasca Bitumen,’’ Fuel 64(5),
Davis, B. E. and J. W. Jennings, ‘‘State-of-the-Art Summary for 678–691 (1985).
Underground Coal Gasification,’’ J. Pet. Technol. 36(1), 15–21 Ritchie, R. G. S., R. S. Roche and W. Steedman, ‘‘Non-isothermla
(1984). Programmed Pyrolysis Studies of Oil Sand Bitumens and
Egloff, G. and J. C. Morrell, ‘‘Cracking of Bitumen Derived From Bitumen Fractions 1. Athabasca Asphaltene,’’ Fuel 64(3),
Alberta Tar Sands,’’ Can. Chem. Metall. 11(2), 33 (1927). 391–399 (1985).
Energy Resources Conservation Board. ST98-2008: Alberta’s Ritchie, R. G. S., R. S. Roche and W. Steedman, ‘‘Pyrolysis of
Energy Reserves 2007 and Supply/Demand Outlook Athabasca Tar Sands: Analysis of the Condensible Products
2008–2017, www.ercb.ca/sts/ST98/st98-2008.pdf (2008). From Asphaltene,’’ Fuel 58(7), 523–530 (1979).
Fogler, H. S., ‘‘Elements of Chemical Reaction Engineering,’’ Speight, J. G., ‘‘Thermal Cracking of Athabasca Bitumen,
4th ed., Prentice Hall International Series in the Physical and Athabasca Asphaltenes, and Athabasca Deasphlated Heavy
Chemical Engineering Sciences, NJ, USA (2006). Oil,’’ Fuel 49(2), 134–145 (1970).
Hajdo, L. E., R. J. Hallam and L. D. L. Vorndran, ‘‘Hydrogen Strausz, O. P., K. N. Jha and D. S. Montgomery, ‘‘Chemical
generation during in situ combustion,’’ SPE California Composition of Gases in Athabasca Bitumen and in
Regional Meeting, Bakersfield, California, SPE 1366 1-MS, Low-temperature Thermolysis of Oil Sand, Asphaltene and
27–29 (March 1985). Maltene,’’ Fuel 56(2), 114–120 (1977).
Hayashitani, M., D. W. Bennion, J. K. Donnelly and R. G. Moore, Suresh, B., M. Yoneyama and S. Schlag, ‘‘Chemical Economics
‘‘Thermal Cracking of Athabasca Bitumen,’’ The Future of Handbook Report: Hydrogen,’’ SRI Consulting, Menlo Park,
Heavy Crude and Tar Sands, Second International Unitar California (2007).
Conference, Venezuela, pp. 233–247 (1977). Yang, X. and I. D. Gates, ‘‘Combustion Kinetics of Athabasca
Hayashitani, M., D. W. Bennion, J. K. Donnelly and R. G. Moore, Bitumen From 1D Tube Experiments,’’ Nat. Resour. Res.
‘‘Thermal Cracking Models for Athabasca Oil Sands Oil,’’ 18(3), 193–211 (2009).
53rd Annual Fall Technical Conference and Exhibition of the Yasar, M., D. M. Trauth and M. T. Klein, ‘‘Asphaltene and Resid
Society of Petroleum Engineers of AIME, SPE 7549. Houston, Pyrolysis. 2. The Effect of Reaction Environment on Pathways
Texas, pp. 1–16 (1–3 October, 1978). and Selectivities,’’ Energy Fuels 15(3), 504–509 (2001).
Henderson, J. H. and L. Weber, ‘‘Physical Upgrading of Heavy
Crude Oils by the Application of Heat,’’ J. Can. Petrol.
Technol. 4(4), 206–212 (1965). Manuscript received February 15, 2012; revised manuscript
Hyne, J. B., J. W. Greidanus, J. D. Tyrer, D. Verona, C. Rizek, P. received May 9, 2012; accepted for publication May 11, 2012.
D. Clark, R. A. Clark and J. Koo, ‘‘Aquathermolysis of Heavy
Oils,’’ The Future of Heavy Crude and Tar Sands, Second
International Unitar Conference, Venezuela, pp. 404–411
(1977).
Kapadia, P. R., M. S. Kallos and I. D. Gates, ‘‘Potential for
Hydrogen Generation From In Situ Combustion of Athabasca
Bitumen,’’ Fuel 90(6), 2254–2265 (2011).
Lin, C. Y., W. H. Chen and W. E. Culham, ‘‘New Kinetic Models
for Thermal Cracking of Crude Oils in In-Situ Combustion
Processes,’’ SPE Reservoir Eng. 2(1), 54–66 (1987).
MATLAB Optimization ToolboxTM User’s Guide, The Mathworks,
Inc., MA, USA (2009).
Mazza, A. G. and D. E. Cormack, ‘‘Thermal Cracking of the
Major Chemical Fractions of Athabasca Bitumen,’’ AOSTRA J.
Res. 4(3), 193–208 (1988).
McNab, J. G., P. V. Smith Jr. and R. L. Betts, ‘‘The Evolution of
Petroleum,’’ Ind. Eng. Chem. 44(11), 2556–2563 (1952).
Millour, J. P., R. G. Moore, D. W. Bennion, M. G. Ursenbach and
D. N. Gie, ‘‘A Simple Implicit Model for Thermal Cracking of
Crude Oils,’’ SPE Annual Technical Conference and
Exhibition, Las Vegas, SPE 14226, pp. 1–15 (September
1985).
Millour, J. P., R. G. Moore, D. W. Bennion, M. G. Ursenbach and
D. N. Gie, ‘‘An Expanded Compositional Model for

| VOLUME 9999, 2012 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 13 |

You might also like