Reaction Kinetics of Fuel Formation

for In-Situ Combustion

Sldql A. Abu·Khamsln,· SPE, Stanford U.
William E. Brigham, SPE, Stanford U.
Henry J. Ramey Jr., SPE, Stanford U.

Summary. Chemical reactions believed to cause fuel formation for in-situ combustion have been studied and modeled. A thin,
packed bed of sand/oil mixture is heated under nitrogen flow at linearly increasing temperatures, simulating the approach of a
combustion front. Analysis of gases produced from the reaction cell revealed that pyrolysis of crude oil in porous media goes
through three overlapping stages: distillation, mild cracking (visbreaking), and severe cracking (coking). Expressions that govern
the rates of the two cracking reactions are derived, and a technique is outlined to obtain initial estimates for their parameters from
the experimental data. The parameters of a proposed distillation function, as well as refined estimates for the cracking reaction
parameters, are obtained by non-linear regression methods based on an overall kinetic model.
Successful matching of the experimental data, including the total amount of fuel deposited, was achieved with this model. It was
found that fuel formation was a result of two successive cracking reactions that the oil undergoes at temperatures above 280°C
[536°P]. Also, distillation of crude oil at temperatures below 280°C [536°F] played an important role in shaping the nature and
extent of the cracking reactions. The operating pressure and the rate of heating of the sand/oil sample were found to affect the
fuel-formation process only through the influence exerted on distillation. Clay minerals showed a catalytic effect on the cracking
reactions, especially coking. Finally, the asphaltene fraction of a crude oil was found to correlate with the fuel content of that oil.

Introduction
Fuel formation occurs in a reservoir undergoing in-situ combustion
as a result of various physical and chemical changes inflicted upon
the reservoir oil, mainly distillation and thermal cracking. An im-
portant parameter to be considered in the design of a combustion
project is the concentration of fuel deposited ahead of the combustion
front, primarily in the evaporation and cracking zones.
The nature of fuel varies widely from one reservoir to another.
It can be close to the heavy fraction of the parent oil or a solid coke-
like residue. Its apparent elemental hydrogen/carbon (H/C) ratio
could be computed from effluent gas composition with the carbon
oxides and consumed oxygen for a stoichiometric balance. Survey
of the literature revealed the following general observations re-
garding the influence of process variables on fuel properties and
concentration.
1. The fuel H/C ratio is generally lower than that of the parent
oil to an extent that depends on the operating conditions. The fuel
H/C ratio decreases with increasing combustion temperatures, 1-3
coking temperatures, 4,5 or pressure. 6,7
2. Decrease in oil API gravity or H/C ratio leads to increased
fuel deposition. Higher viscosity and higher Conradson carbon
residue gave a similar result. 2,4
3. Higher combustion-front temperature or slower front velocity
reduces fuel deposition.
4. Pressure has no general effect on fuel availability, with every
oil showing a different behavior. 6,8
5. Higher initial oil saturation was found to cause more fuel
deposition 2 ; on the other hand, it was argued that the residual oil
saturation in the steam plateau was a key factor. 6
6. The specific surface area of the reservoir rock is of particular
importance to fuel deposition, especially when clays are present.
A larger specific surface area facilitates the various heterogeneous
reactions that cause fuel formation. 9
While these observations serve as general guidelines for project
design, functional relationships between process variables and fuel
concentration are imperative for accurate prediction of in-situ com-
bustion performance. Combustion-tube tests yield fuel concentration
data, but such data have limited applicability in the light of areal
variation ofreservoir characteristics. Moreover, the saturation and
properties of the residual oil in the steam plateau are expected to
change as the combustion zone advances in the reservoir. Finally,
'Now at King Fahd U. of Petroleum & Minerals
Copyright 1988 Society of Petroleum Engineers
simulation models developed for the in-situ combustion process re-
quire accurate mathematical formulation of the kinetics of reactions
that a reservoir oil undergoes as the thermal heat wave propagates
in the reservoir.
To model fuel formation, the sequence of events that lead to it
has to be characterized and related. An idealized environment that
prevails in a reservoir volume element being approached by the
combustion front is that oflight-hydrocarbon displacement followed
by steamdrive, both reducing oil saturation to the residual with sub-
stantiallight-ends distillation. Further approach by the front causes
temperature to rise steadily with time, resulting in more distillation
and triggering mild oil pyrolysis. Finally, immediately before the
arrival of the front, severe pyrolysis of the trapped hydrocarbons
causes fuel deposition. Throughout these events, the gas phase is
composed of only nitrogen, carbon oxides, and water vapor.
Pyrolysis Reactions
Pyrolysis reactions can be represented by the following formula:
Nh(l)QH(l)+G, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
where
Nh(l) = heavy oil,
Q = heat,
H(l) = hydrocarbon derivatives, and
G = gas.
To facilitate modeling, the reactant is taken as a pseudocomponent
of the oil, usually a heavy fraction, although multi-pseudo-
component apprpaches were attempted. 10 The reaction rate ex-
pression for a pseudocomponent adopted by numerous investigators
is
de
-=-ken. ...................................... (2)
dt
The reaction rate constant is normally expressed by the Arrhenius
equation:
k=Ae- E1RT . ...................................... (3)

1308 SPE Reservoir Engineering, November 1988

 2
4

OXIDIZER

REACTOR

- FLOW LINES

Legend - - - ELECTRICAL LINES

1. Pressure Gauge 9. Pressure Regulator -. - CALIBRATION LINES

2. Drierite Bed 10. Needle Valve

3. Three-Way Valve 11. Four-Way Valve
4. 60-l-tm Filter 12. 2-l-tm Filter
5. Mass Flow Controller 13. Rotameter
6. Furnace 14. Recorder
7. Temperature Controller 15. Mass Flow Meter
8. Temperature Reader 16. Five-Way Valve

Fig. 1-Schematic of the pyrolysis apparatus.

parameters. Among the oils tested, however, a highly paraffiniC

oil was found to be the least susceptible to visbreaking.

Coking Reactions. These are performed under severe thermal-

Fig. 2-Pyrolysis cell.
Pyrolysis reaction stoichiometry and hydrocarbon products vary
widely with the oil composition and reaction conditions. Most
pyrolysis reactions are endothermic, which makes temperature the
most influential parameter that dictates the severity of cracking.
Two classes of pyrolysis reactions that are based on the reaction
temperature evolved in standard processes for thermal treatment
of crude oil and its derivatives: visbreaking and coking reactions.
Visbreaking Reactions. Visbreaking reactions are performed under
mild temperatures to reduce permanently the oil's viscosity and spe-
cific gravity. This phenomenon has been observed in most thermal
recovery techniques, especially in steamdrive. 11 Henderson and
Weber l2 experimented with crude oils and Athabasca bitumen in
a batch reactor. For a first~order kinetic model, they computed ac-
tivation energies ranging from 182.4 to 260.2 MJ/kmol [78.4 x 103
to 111.9 x 103 Btullbm mol] and pre-exponential constants ranging
from 7.2 x 109 to 2.9 X 10 15 . They were unable to correlate the
effect of pressure or the presence of sand, clay, or water on these
cracking conditions to produce a solid residue with a low H/C ratio.
Literature studies dealt mostly with bitumen or shale oil. Lozano
et ai. 13 proposed a distillation-dominated step followed by a
cracking-dominated step. The cracking activation energy for
Venezuelan bitumen increased from 123.4 to 199.6 MJ/kmol
[53 x 103 to 86 X 10 3 Btu/Ibm mol] when the heating rate was in-
creased from 150 to 330 K/h [270 to 594 °F/hr]. Ritchie et al. 14
found similar results for Athabasca bitumen. They concluded that
for in-situ combustion of tar sands, distillation plays a more im-
portant role than pyrolysis. Burnham 15 proposed a kinetic model
for shale oil coking inside shale particles at temperatures between
504 and 610°C [939 and 1,130°F]. The reaction had the following
mass stoichiometry and rate constant:
N(ll ..... 0. 75C(sl +0.20CH 4 +0.05H 2 . . . . . . . . . . . . . . . . . . . (4)
and
kl =3.1 X 107 e- 35 fXXJ/RT. . .....••.•.••.............• (5)
Burnham observed a general decrease in the nitrogen content of
the cracking products as yield increased, indicating the concentration
in the coke of the aromatic nitrogen compounds.
Lin et ai. 16 presented two kinetic models for the cracking
reactions in heavy-oil pyrolysis. Their models used two or three
pseudocomponents and a temperature-dependent reaction order. In
Model A, the oil's heavy pseudocomponent cracks into coke and
a light component, which in turn converts into more coke. In Model
B, the heavy pseudocomponent cracks into coke, a medium com-

SPE Reservoir Engineering, November 1988 1309

 TABLE 1-CRUDE OIL PROPERTIES

Viscosity
Density at 65°C C H N S Atomic Asphaltenes
Field (g/cm 3 ) (Pa·s) (%) (%) (%) (%) H/C Ratio (%)
Huntington Beach 0.943 0.032 84.33 11.31 0.82 1.90 1.609 10.86
West Newport 0.961 0.201 84.87 11.41 0.76 2.21 1.613 8.45
Argentina NA >100 83 14 2 16.79

TABLE 2-INITIAL AND OPERATING CONDITIONS OF EXPERIMENTAL RUNS

Flow Rate,
10- 3 dm 3 /s Initial
Nitrogen Air Porosity Oil Saturation dTldt Gauge Pressure
Run
-- Oil (Reactor) (Oxidizer) Matrix ~ (%) (Klh) psig mPa
--
63 Huntington Beach 5.17 4.5 Clean sand 39.2 21.7 88.9 100 . 0.689
64 Huntington Beach 5.17 4.5 Clean sand 39.1 17.0 88.9 50 0.345
65 Huntington Beach 5.17 4.5 Clean sand 39.1 17.1 88.9 150 1.034
67 Huntington Beach 5.17 4.5 Clean sand 37.7 17.8 88.9 10 0.069
68 Huntington Beach 5.17 4.5 Clean sand 38.4 36.0 88.9 300 2.068
69 West Newport 5.23 4.6 Clean sand 39.2 19.2 88.9 50 0.345
70 West Newport 5.23 4.6 Clean sand 38.8 19.8 44.4 50 0.345
71 West Newport 5.23 4.6 95% sand, 5% clay 35.4 25.2 88.9 50 0.345
72 West Newport 5.23 4.6 95% sand, 5% clay 33.6 25.9 44.4 50 0.345
73 Argentina 5.23 4.6 Clean sand 39.8 7.4' 88.9 50 0.345
74 Huntington Beach 5.23 4.6 Clean sand 38.1 29.8 88.9 50 0.345
75 Huntington Beach 5.23 4.6 Clean sand 38.2 28.6 88.9 220 1.517
'Water content of the 011 resulted in an estimated water saturation of 8.8%.

ponent, and a light component. Additional coke is produced from Experimental

only the medium component. While both models successfully
matched the isothermal batch reactor data, the experiments do not
fully simulate a real reservoir cracking environment. In this envi-
ronment, any light-oil fraction will be removed from the cracking
zone immediately by sweeping gases. Thus, the second cracking
step in Model A is unrealistic. Model B is more suitable except
for the continued presence of the light fraction, whose influence
on the overall cracking process is unknown. The depth-of-cracking
effect, which necessitates a decreasing reaction rate order, is a phe-
nomenon that may not take place in a dynamic reservoir envi-
ronment.
In this study, reaction kinetics of heavy-oil pyrolysis in porous
media were investigated under simulated reservoir conditions where
an inert gas was constantly sweeping the oil/sand mixture, whose
temperature was raised linearly with time. The experiments were
terminated only when the hydrocarbon gas production rate from
the reactor became negligible, which usually coincided with tem-
perature levels normally encountered in combustion zones.
Fig. 1 shows schematically the experimental setup. The central units
of the system are the reactor and the oxidizer. The other compo-
nents provide the necessary gas-metering and control, g~-analysis,
pressure-regulation, and data-gathering systems. The reactor
consists of a controlled temperature oven, which houses the pyrolysis
cell (Fig. 2).
After the sand/oil mixture is charged to the pyrolysis cell, the
reactor temperature is raised at a constant rate while nitrogen is
continuously sweeping the sample. The reactor effluent is mixed
with air and heated to 677°C [1,251 oF] in the oxidizer. The purpose
of the oxidizer is to convert all distillation and r~~ction gaseous
products into carbon oxides and water for ease of analysis. Com-
plete oxidation is ensured by the oxidizer's temperature, excess air
supply, and residence time (20 seconds). A run is terminated when
hydrocarbon production from the reactor drops to negligible levels,
signaling complete coking. At this point, air is supplied to the
pyrolysis cell to consume its coke content and hence to determine
the coke's mass and H/C ratio.

if TEMPERATURE .. ,if.", .. 6.----::::::~------------..,

68

I.
-t HIe RATIO " 0
+ 0 PRODUCTION RATE ••• 715
+ )( CUM. MASS, ~. • ... 700
o
615
<----': /"'.. •........ -----7
o
63
o
~ o~ 0.,..__ ?o 600 :w::

i\! ...... ~'b lOoo 64

o
'A....,..i..
: .... ..........,
V 0

~ ......... ;

~
.....
Q.
1:
4
oE---,: .... ...........++ ~~o +

;,0.... 011.'> ~ 0
.~. .I~··I. +

o
G/ / . ,
0 .. , ••••••
;., '"
,.",
300
3

2 L-____________L -____________ ~ ______ ~

TIME. h
1.8 2.0 2.2
Fig. 3-Experlmental data for Run 67: Huntington Beach 011 , I T xl 0 3 I 1/1<
at 10 pslg.
Fig. 4-Varlatlon of distillation peak temperature with pres-
sure for the Huntington Beach.

1310 SPE Reservoir Engineering, November 1988

 TABLE 3-AMOUNT AND HIC RATIO OF COKE DEPOSITED
Pressure Heat Rate Coke Coke
Run Oil psig MPa (K/h) Matrix (gIg oil) HIC Ratio
-
67 Huntington Beach 10
--
0.069 88.9 Clean sand 0.1239 1.45
64 Huntington Beach 50 0.345 88.9 Clean sand 0.1153 0.96
74 Huntington Beach 50 0.345 88.9 Clean sand 0.1284 0.87
63 Huntington Beach 100 0.689 88.9 Clean sand 0.1348 1.12
65 Huntington Beach 150 1.034 88.9 Clean sand 0.1455 0.66
75 Huntington Beach 220 1.517 88.9 Clean sand 0.1682 1.23
68 Huntington Beach 300 2.068 88.9 Clean sand 0.1614 0.99
69 West Newport 50 0.345 88.9 Clean sand 0.1180 1.30
70 West Newport 50 0.345 44.4 Clean sand 0.1213 1.48
71 West Newport 50 0.345 88.9 Sand + Clay 0.1633 0.90
72 West Newport 50 0.345 44.4 Sand + Clay 0.1435 1.01
73 Argentina 50 0.345 88.9 Clean sand 0.1586 0.45

Two heavy California crudes and an Argentinian paraffinic oil for coking. The rates of the two reactions are given by
were used. Table 1 lists their properties and elemental analysis.
The matrix was 20/30-mesh Ottawa sand, which was sometimes 171 = -klmom= -Ale-EjIRTmom . .................... (8)
mixed with kaolinite to investigate the catalytic effect of clay.
A total of 12 runs was performed at different pressures and rates and
of heating. Table 2 summarizes their initial and operating conditions.
Reactor data from a typical run are plotted iri Fig. 3. These data
include the cumulative mass of produced hydrocarbons, the
hydrocarbon production rate (HPR) normalized to 10 g of initial In Eq. 8, the rate of consumption of oil is that caused by visbreaking
oil, the produced hydrocarbons HIC ratio, and the reactor's tem- only.
perature. Note that pyrolysis ended at 6.05 hours and coke was Modeling these two reactions requires that the total reactor gas
burned from 6.05 to 6.6 hours. be broken down into its constituents: G 1 , G2 , and distillation gas.
Because all these phenomena occur simultaneously, it becomes nec-
Kinetic Model essary to locate those parts of the HPR curve where one phenomenon
Pyrolysis of crude oil is assumed to take place in a chain reaction dominates the other two and gas production can be attributed solely
as follows: to this phenomenon.
heat
Distillation dominates the low-temperature segment of the HPR
crude oil ---+- visbroken oil + gas curve, and the first peak is clearly a distillation peak. At high tem-
peratures, distillation has ceased and complete visbreaking of the
and oil has taken place. Therefore, the high-temperature peak is at-
heat tributed to the coking reaction. At medium temperatures, two
visbroken oil ---+- coke + gas. reactions influence the mass of visbroken oil, mo: generation by
visbreaking and consumption by coking. Thus, mo goes through
A basic assumption here is that no crude oil cracks directly into a maximum at the second peak, the visbreaking peak.
coke and gas; it must go through the intermediate step of visbreaking
first. The two reactions are irreversible and have the following mass Distillation. A normalized distillation curve for an oil can be de-
stoichiometry : fined by some function of temperature, f( T). Vossoughi et at. 17
have shown that f( T) resembles the normal distribution function:
N u ->(3N v +(I-{3)G 1 • •.•.....•..•....•..••••••••... (6)
f(T)=oe)..(T-T)2, ................................. (10)
for visbreaking and
where 0, A, and T are influenced by the operating conditions of dis-
N v->aC+(I-a)G2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7) tillation. The gas production rate in the low-temperature regime

1.5
0
0.2r---------------------------------,
0

2 0
~ 1.0 0
<[ 0
0
a: 0
'0 0
...
I 0
"
"...
0.15 :I: 0

w 0.5
~
ki
~
0
0
0
0
0
o

0.1
0 50 100 150 200 250 300
PRESSURE, psig
PRESSURE, pslg
Fig. 6-H/C ratio of coke vs. pressure for the Huntington
Fig. 5-Effect of pressure on coke deposition for the Hunt- Beach runs.
Ington Beach runs.

SPE Reservoir Engineering, November 1988 1311

 0.16

0.1~
0.6

0.' .
0 EXPERIMENTAL
INITIAL MATCH
DATA ...
~
- FINAL MATCH

0.14 "0 0.4

"0
...
I ..;
~
...
"- 0.13 0 '" 0.'
\.\
Ii
0
...;
0
II: \ .-
'\"'/
0.'

~ 0.12
o
0.1

10 II 12 13 14 1~ 16 17

ASPHALTENE CONTENT. % .00 300 400 SOO 600 700 800

TEMPERATURE, J(

Fig. 7-Varlatlon of coke deposition with the 011 asphaltene

content. Fig. a-experimental HPR and Its regression matches: Run 67.

of pyrolysis is thus given by To refine the kinetic and distillation parameter estimates, Eq. 14
can be fitted to the experimental HPR curve with nonlinear
dmg dT regression on these parameters. However, the variation of mo and
-=mof(T)-· ............................... (11) ma with time should be known before regression. This can be
dt dt achieved by solving the following system of ordinary differential
The parameters off(t) can be derived from the low-temperature equations:
segment of the HPR curve up to the distillation peale

Coking. Near and beyond the coking peak, the rate of hydrocarbon
production is given by
and
dmg
- - =(I-a)k2m an . .............................. (12)
dt dm o dT
-=-mof(T)--klmom , ...................... (16)
Visbreaking. Around the visbreaking peak, the rate of hydrocarbon dt dt
production is given by
with the initial conditions mo(O)=moi and ma(O) =0, and the in-
dmg itial parameter estimates. Eq. 15 is Eq. A-tO, and Eq. 16 is de-
-=(I-{3)k 1m am. ............................. (13)
dt rived from Eqs. 11 and 13. Once regression converges on a new
set of parameter values, the mo and ma curves are recomputed and
Rough estimates of the pyrolysis kinetic parameters and distil- the regression process repeated. This cycle can go on until a good
lation parameters can be extracted from the experimental data by match of the experimental HPR curve is obtained.
application of Eqs. 11 through 13 to their respective temperature
regimes. A complete description of the mathematical procedure and Results and Discussions
assumptions are found in Ref. 18; a summary of this procedure General observations regarding the influence of various operating
is presented in the Appendix. parameters on pyrolysis are discussed in this section.
At any temperature, the HPR is the summation of all contributions
described by Eqs. 11 through 13:

:g =m ooe}..(T-T)2: +(I-{3)Ale-ElIRTmom
0.7

0.6 0

01

III
:.: o.!! 0
0
U
• L&..

..
0 0.4
0 UNREACTED OIL

~
VIS.ROKEN OIL
III
III
I.. COKE ct
:i 0.3
0

,.j 0
III
"
U
0
1
~
U 0.2
II:
0 0
III

.
-' \0000 Q:
0 Q.
0 ... 0.1

.
0.'

.
'"'",. .;:,
...
°0 ; " . ,

o~ .. +t-++++++

0
.... ~'-+..++"'.. o.!! 0.6 0.7
.00 .00 400 800 800 700 800
MEASURED MASS OF COKE. 9
TEMPERATURE I K

Fig. 9-Computed masses of unreacted oil. vlsbroken 011. and Fig. 10-Correlatlon between measured and predicted mass
coke: Run 67. of coke.

1312 SPE Reservoir Engineering, November 1988

 TABLE 4-FINAL PARAMETER OUTPUT FROM THE REGRESSION ROUTINE

0 A- T E11R E21R
(10- 4 I<) L n
Run
- ~ (I<)
--
m Al ~ -ex- - A2 ~
67 0.0025 -1.136 447 0.588 1.456 5.244 x 10 5 6800 0.536 1.49 3.107 x 10 14 21650
64 0.0033 -1.359 472 0.674 0.806 2.191x10 6 8010 0.408 1.64 4.910x10 15 23430
74 0.0030 -1.338 482 0.821 0.352 1.277x 10 6 7160 0.714 0.77 2.563x10 15 22780
63 0.0028 -1.490 478 0.850 0.567 3.420 x 10 4 5350 0.859 1.82 1.101x10 17 24560
65 0.0027 -1.541 480 0.686 0.505 2.111 x10 5 6760 0.487 1.78 1.547x 10 17 25480
75 0.0024 -1.577 511 0.700 0.371 4.537 x 10 3 4520 0.346 1.93 5.386 x 10 14 22530
68 0.0015 -1.503 520 0.878 0.319 1.978x103 3400 0.703 1.39 2.544 x 10 14 21580
69 0.0041 -1.527 480 0.630 0.660 7.332 x 10 3 5000 0.454 1.70 6.642 x 10 18 28410
70 0.0040 -1.326 458 0.782 0.992 2.928 x 10 6 7850 0.591 1.14 4.290 x 10 13 20650
71 0.0038 -1.427 480 0.830 0.877 1.233 x 10 8 9780 0.625 1.49 2.654 x 10 17 26000
72 0.0039 -1.290 460 0.994 0.691 8.932 x 10 4 4410 0.654 1.20 2.158 x 10 12 18160
73 0.0045 -1.511 590 0.891 0.993 1.685 x 10 7 9330 0.728 1.94 6.465 x 10 13 19860

DistUlation. As expected, the distillation peak was delayed until
higher temperatures as pressure was increased. This is shown by
Fig. 4 for Huntington Beach oil. Reduced heating rates had the same
effect as reduced pressures. The cumulative fraction of oil vapo-
rized at the distillation peak, however, was the same regardless of
pressure or heating rate. Adding clays to the matrix did not change
the distillation characteristics, even though clays provide a large
surface area. Phase equilibrium between oil and gas was therefore
always achieved. It should be noted that the heating rates used in
this study were much higher than those encountered in field appli-
cations. Farouq Ali 19 estimated field heating rates to be between
11 and 28 Klh [20 and 50°F/hr].
Visbreaking. The visbreaking peak occurred at 304 °C [579°F] for
all pressures investigated except for the 69-kPa [1O-psig] run, where
it occurred at 267°C [513°F]. This apparent insensitivity to pressure
was not the case with the normalized HPR at the visbreaking peak.
A slight general increase in HPR with pressure was observed for
the runs at 69,345,690, and 1034 kPa [10, 50, 100, and 150 psig]
followed by a speculated, though not experimentally investigated,
decrease at higher pressures until the visbreaking peak disappeared
for 1517- and 2069-kPa [220-and 300-psig] runs. Two opposite
factors are at play here. As pressure is elevated, distillation is sup-
pressed, causing more reactant to be available for visbreaking at
higher temperatures. On the other hand, this reactant will be progres-
sively lighter, thus more resistant to visbreaking. The combined
effect will be as observed experimentally. To confirm the latter
factor, the rate of visbreaking was computed at a typical visbreaking
temperature of 309°C [588°F] for 10 g of oil. It was found to
decrease from 76 g/h at 69 kPa [10 psig] to <7 glh at 2069 kPa
[300 psig].
H/C ratio estimates of the oil remaining in the pack at the vis-
breaking peak revealed a heavier oil when clays were added or when
the heating rate was reduced. This explains the earlier occurrence
of the visbreaking peak for these two conditions, which was facili-
tated by more efficient distillation. The Argentinian paraffinic oil
did not show a visbreaking peak.
Coking. The coking peak occurred at around 41rC [774°F] with
a maximum deviation of 8°C [14 OF] for all pressures investigated.
The HPR at the coking peak also increased with pressure. Both
observations are attributed to larger and lighter oil fractions being
left in the pack at coking temperatures as pressure is increased.
Table 3 lists the mass fraction and H/C ratio of the coke left in
the pack at the termination of every pyrolysis run. Table 3 data,
plotted in Fig. 5, show a general increase in coke deposition with
increase in pressure, while Fig. 6 shows a slight decrease in its
H/C ratio. Table 3 also reveals that a heavier and larger coke deposit
is produced when clays are used or when slower heating is applied.
While clays facilitate coking by catalysis, slower heating leads to
smaller weight loss upon visbreaking, leaving a larger oil fraction
to coke. Fig. 7 plots coke deposition vs. the oil's asphaltene content
for the three oils used, all at 345 kPa [50 psig] in clean sand heated
at a rate of 88.9 K/h [l60°F/hr].
Non-Linear Regression. Fig. 8 shows the HPR curve for Run 67
(69 kPa [10 psig]) predicted by the kinetic model of this study. The
curve labeled "initial match" was predicted with values for the
kinetic parameters estimated directly from the experimental data.
The curve labeled "final match" was generated by the regression
routine after refined estimates of these parameters were computed.
The same quality match was obtained with the other runs.
Fig. 9 shows the masses of the oil, visbroken oil, and coke pre-
dicted by thp kinetic model. The predicted vs. measured coke mass
is plotted in Fig. 10 for all runs. The positions of the data points
relative to the 45° line indicate good matching.
The parameters of the distillation function are listed in Table 4
and are plotted vs. pressure for the Huntington Beach oil in Figs.
11 through 13. Both distillation-peak temperature and distillation-
peak spread parameters increase with pressure as shown in Figs.
11 and 12, respectively. These trends result from prolonged distil-
lation phases. The distillation-peak magnitude, on the other hand,
decreased with pressure (Fig. 13). This is necessitated by material-
balance considerations if the oil fraction distilled at the distillation
peak is to remain constant, as shown earlier.

III 5!1O -1.3....------------------,

It:
:;) o

~
III
0
«0
III o
0.. 0
2 9!'t
III 500 -1.4
~ ~o
«-
~~
~ .
~~
z .... o
0 i= -
'<It< -1.5 o
~ oJ
oJ
oJ o
oJ i=
CI)
i= is
CI) o
is

PRESSURE. pilI! PRESSURE. psll!

Fig. 11-Dlstlllatlon-peak temperature vs. pressure. Fig. 12-Dlstlllatlon-peak spread vs. pressure.

SPE Reservoir Engineering, November 1988 1313

 15 2150
11.1
0
:::> 4
!::: 200 o
z
C)
0
o
«o, 0

:lEO 3 )-
C)
:,.:-
«>C II:
w 1150
I.I.ht
a..~
z-
W O ( 0 VISBREAKING)
2
z ,
QCO 0
...E
z'
(0 COKING)

!q 0'"
i=:lE
100
..J
..J ~LLi
i= i= 0
!!!
0 0 ~ 150
0
o
0 150 100 1150 200 2150 300

PRESSURE, psig
0
0 50 100 1150 200 2150 300
Fig. 13-Dlstlllatlon-peak magnitude vs. pressure.
PRESSURE, psig

Fig. 15-Activation energy vs. pressure.

40 D
0

a::: 0 2.0
0

~
u 0
0
c
0 0

rt It:
1.6
B
0

...J 30 ILl [
c( C
It:
i= 0 1.2
z
W Z
VISBREAKING)
~
Q (0
( 0 VISBREAKING) ~ 0
D... U 0.8 ( 0 COKING)

~c( (0 COKING) ~
ILl 0
I It: 0
We 20
a:::- 0.4 0
D...
IL
0 o ____ ____ ____ L -_ _ _ _L -_ _ _ _L -_ _
~
~
~
~
~ 0 o 50 100 150 200 250 300
J: 0
~ 0 PRESSURE, psig
a:
c( 0 Fig. 16-Reactlon order vs. pressure.
(!)
to
0
9 c
1.0

c
It: e
o L -__ ~ 0.8
~ ____ ~ __ ~ ____ ~ __ ~ ____ ~

u
~
0 [
0
o 50 100 150 200 250 300 0

PRESSURE, psiO u 0.6 0

~ 0
~
0
Fig. 14-Pre-exponentlal factor vs. pressure. ILl
~
Q 0.4 0

::r: 0
u
Also listed in Table 4 are the kinetic model parameters. Figs.
14 through 17 show plots of the pre-exponential factor, activation
energy, reaction order, and stoichiometric factor vs. pressure, re-
~ 0.2
( 0

(0
VISIREAKING )

COKING )
spectively, for both reactions. Consistent trends are depicted in these
figures for the visbreaking parameters. The computed visbreaking 0
rate constants at a typical visbreaking-peak temperature increased 0 50 100 150 200 250 300
steadily with pressure. Because of decreasing reaction order, m, PRESSURE, psig
however, the specific rate of reaction decreased with pressure, as
indicated before. The visbreaking activation energy ranges between Fig. 17-Stolchlometrlc factor vs. pressure.
28.3 and 81.3 MJ/kmol [12.2xW 3 and 35xW3 Btu/Ibm mol].
These values differ from those reported elsewhere lJ ,12 because of
different oils and reaction conditions and because different kinetic merely an artifact of curve fitting without more data. The coking
models were used to fit the experimental data. activation energy computed by the regression routine, however,
The coking parameter data lack any consistent trend. Aside from was found to be insensitive to the initial value entered into the
the scatter in the stoichiometric factor data, an apparent reversal routine. For all Huntington Beach runs, variation of the initial E2
in trend is observed at 1034 kPa [150 psig] for the other three pa- value by as much as 50 % resulted in a maximum of 2.3 % change
rameters. It is difficult to ascertain whether this reversal is real or in the computed value.
1314 SPE Reservoir Engineering, November 1988
 The catalytic effect of clays is very distinct in the data of Table
4. This effect is demonstrated by reduction of coking activation
energy (Runs 69 and 70 and Runs 70 and 72), and increased con-
version to coke, as reflected by the stoichiometric factor. Slower
heating rates gave effects similar to those of clay.
Conclusions
1. An experimental apparatus was designed to investigate the fuel
formation process for in-situ combustion.
2. A chemical reaction mechanism was proposed for the cracking
reactions that take place in crude oil pyrolysis.
3. A mathematical model based on distillation and the reaction
mechanism was developed in which kinetic expressions were pro-
posed for the rates of the reactions. Graphical and analytical schemes
were also proposed whereby estimates for the parameters used in
the kinetic expressions can be obtained from the experimental data.
4. A nonlinear regression technique was used to refine the es-
timates for the kinetic and distillation parameters and to match the
experimental data.
S. The regression technique also allows prediction of the total
fuel mass deposited during oil pyrolysis. The prediction method
is believed to be more appropriate than the conventional approach
used in simulation models.
6. Pyrolysis of crude oil in porous media goes through three over-
lapping stages: distillation, visbreaking, and coking. During dis-
tillation, the oil loses most of its light-gravity and part of its
medium-gravity fractions. During visbreaking, the oil mildly cracks
into a slightly different product. At high temperatures, the oil re-
maining in the porous medium cracks into a semisolid residue that
is rich in carbon. Both cracking reactions produce gas.
7. Distillation of crude oil at low temperatures plays an important
role in shaping the nature and extent of the cracking reactions. Both
high operating pressures and fast heating rates suppress distillation.
8. The operating pressure and heating rate affect the fuel-
formation process largely through the influence they exert on dis-
tillation, as described above. High pressures lead to the formation
of more fuel that is leaner in hydrogen. This is consistent with liter-
ature findings.
9. Clay minerals have a catalytic effect on the cracking reactions,
especially on coking.
10. The asphaltene content of a crude oil appears to be a good
correlating parameter to determine the fuel content of that oil.
11. The paraffinic oil studied was capable of producing fuel for
in-situ combustion. This fuel appeared to be derived from both
paraffinic and asphaltic fractions of the oil; however, this oil resisted
visbreaking. Presumably, this type of result would be seen with
other paraffinic oils.
Nomenclature
A = pre-exponential factor of reaction rate constant
c = concentration
C = coke
E = activation energy of reaction, MJ/kmol
[Btu/Ibm mol]
f = distillation function
G = gas produced by chemical reaction
H = hydrocarbon derivatives
k = reaction rate constant, g(l- reaction order) /h
m = visbreaking reaction order (unsubscripted)
m = mass at any time (subscripted), g
n = coking reaction order
N = oil
N h = heavy oil
N u = uncracked oil
N v = visbroken oil
Q = heat
R = universal gas constant, J/(Kmol' K) [Btu/(lbm
mol-OF)]
t = time, hours
T = temperature, °C or K [OF]
whn = normalized hydrocarbon mass production rate,
g/(g oil'h)
SPE Reservoir Engineering, NovelJ'her 1988
ex = coking stoichiometric factor
f3 = visbreaking stoichiometric factor
8 = distillation function parameter, K-l [OF-I]
11 = rate of reaction, g/h
A = distillation function parameter, K- 2 [oF-2]
T = dlstillation function parameter, K [oF]
Subscripts
c = coke
e = termination point
g = gas
i = initial
o = uncracked oil
o = visbroken oil
op = visbroken oil at the coking peak
pc = at the coking peak
1 = visbreaking reaction
2 = coking reaction
References
1. Martin, W.L., Alexander, J.D., and Dew, J.N.: "Process Variables
of In-Situ Combusion," Trans., AlME (1958), 213, 28-31.
2. Alexander, J.D., Martin, W.L., and Dew, LN.: "Factors Affecting
Fuel Availability and Composition During In-Situ Combusion," JPT
(Oct. 1962) 1154-64; Trans., AIME, 225.
3. Dew, J.N. and Martin, W.L.: "Air Requirement for Forward Com-
bustion, Part I," Pet. Eng. (Dec. 1964) 82.
4. Bousaid,I.S. and Ramey, H.J. Jr.: "Oxidation of Crude Oil in Porous
Media," SPEl (June 1968) 137-48; Trans., AIME, 243.
5. Fassihi, M.R., Brigham, W.E., and Ramey, H.J. Jr.: "Reaction Ki-
netics of In-Situ Combustion: Part I-Observations," SPEl (Aug. 1984)
399-408.
6. Wilson, L.A. et aI.: "Some Effects of Pressure on Forward and Reverse
Combustion," SPEl (June 1963) 127-35; Trans., AlME, 228.
7. Fassihi, M.R.: "Analysis of Fuel Oxidation in In-Situ Combustion Oil
Recovery," PhD dissertation, Stanford U., Stanford, CA (May 1981).
8. Bae, J.H.: "Characterization of Crude Oil for Fireflooding Using
Thermal Analysis Methods," SPEl (June 1977) 211-18.
9. Poettmann, F.H., Schilson, R.E., and Surkalo, H.: "Philosophy and
Technology of In-Situ Combustion in Light Oil Reservoirs," World
Oil (1967) 165, 487.
10. Hayashitani, M. et al.: "Thermal Cracking Models for Athabasca Oil
Sands Oil," paper SPE 7549 presented at the 1978 SPE Annual Tech-
nical Conference and Exhibition, Houston, Oct. 1-3.
11. Shu, W.R. and Hartman, K.J.: "Thermal Visbreaking of Heavy Oil
During Steam Recovery Processes," SPERE (Sept. 1986) 474-82.
12. Henderson, J.H. and Weber, L.: "Physical Upgrading of Heavy Crude
Oils by the Application of Heat, " J. Cdn. Pet. Tech. (Oct.-Dec. 1965)
206.
13. Lozano, J.A., Gonzalez, A.R., and Fernandez, C.M.: "Thermal Degra-
dation of Bitumen from the Faja of the Orinoco, " Ind. & Eng. Chern.,
Prod. Res. Dev. (1978) 17, No. I, 71.
14. Ritchie, R.G.S., Roche, R.S., and Steedman, W.: "A Pyrolysis-Gas
Chromatographic Analysis of Athabasca Bitumen," lnil. & Eng. Chem.,
Prod. Res. Dev. (1978) 17, No.4, 370.
15. Burnahm, A.K.: "Chemistry of Shale Oil Cracking," Oil Shale, Tar
Sands, and Related Materials, American Chemical Soc. Symposium
Series, Reprint No. 163 (1981).
16. Lin, C.Y., Chen, W.H., and Culham, W.E.: "New Kinetic Models
for Thermal Cracking of Crude Oils in In-Situ Combustion Processes, "
SPERE (Feb. 1987) 54-66.
17. Vossoughi, S., Bartlett, G.W., and Willhite, G.P.: "Development of
a Kinetic Model for In-Situ Combustion and Prediction of the Process
Variables Using TGAIDSC Techniques," paper SPE 11073 presented
at the 1982 SPE Annual Technical Conference and Exhibition, New
Orleans, Sept. 26-29.
18. Abu-Khamsin, S.A.: "The Reaction Kinetics of Fuel Formation for
In-Situ Combustion," PhD dissertation, Stanford U., Stanford, CA (Oct.
1984).
19. Farouq Ali, S.M.: "A Current Appraisal ofIn-Situ Combustion Field
Tests," JPT (April 1972) 477-81.
Appendix-Estimation of Kinetic Parameters
In the following discussion, the analytical and graphical techniques
used to estimate the kinetic parameters from the experimental data
are presented.
1315
 Assuming that only visbroken oil is left in the pack at the time Therefore,
corresponding to the coking peak, tpc, the mass of this oil at the
coking peak, mop, is estimated by: dm g
- - =(I-{3)k l mg'+(I-a)k 2mijn . ................. (A-9)
dt
.......................... (A-I)
The mass of visbroken oil changes with time according to

where te is the termination time. Eq. A-I uses coking stoi- dm o

chiometry and is based on the assumption that distillation and vis- --={3k l mg'-k2m on . ......................... (A-to)
breaking are completed before tpc' With the cumulative mass of dt
coke deposited at tpc approximated to be half of that deposited at
te' material balance dictates that Combining Eqs. A-9 and A-tO yields

where moi and mc are the initial mass of oil and the cumulative
mass of coke, respectively. Eqs. A-I and A-2 are used to compute
mllJ! and a.
The normalized HPR at time t is defined by
1 dm g
Whn(t)=- - .............................. (A-3)
moi dt
At any time t> tpc, Whn(t) is given by
dm o (3 dm g l-a{3
- - = - - - - - - - k2mon • ................ (A-ll)
dt 1-{3 dt 1-{3
Eq. A-ll can be numerically integrated to solve for mo(t) with an
assumed value for (3. The starting point is t=tpc' where mo=mop-
Integration is performed back in time until mo becomes negligible.
The rate of consumption of uncracked oil is given by
dm o
--=-klmg' . ................................ (A-12)
dt

Combining Eqs. A-9 and A-12 yields

dm o I-a 1 dm g
- - = - - k2 mon - - - - - . ................. (A-I3)
Differentiating Eq. A-4 with respect to time and noting that dt 1-{3 1-{3 dt

Knowing mo(tpc ) ==0, Eq. A-13 can be integrated to yield the

dWhn
- - =0 at tpc, mass of uncracked oil at any time t$,tpc:
dt

we derive for t =tpc m -

mg(tpc)-mgCt) I-a I pck
t
n dt .........
o(t)- --- 2 mo . (A-14)
1-{3 1-{3t
E2 dT
- - - = - - - - - ; ......................... (A-5)
R'1c dt mop dt
The integral in Eq. A-14 can be evaluated numerically. Eq. A-
12 is rearranged into
but at tpc,

dm o
- - = -k2m3p . ................................ (A-6)
In kl =In( - :0 )-m In mo . .................... (A-I5)

dt
Once a value is assumed for m, In kl can be computed at any
Substituting Eq. A-6 into Eq. A-5, combining with Eq. A-4, and time or, In effect, for any temperature below the coking-peak tem-
rearranging, we obtain perature, Tpc. Plotting In k I vs. liT should yield a straight line if
the true values of (3 and m have been used. The slope of the graph
is EllR and the intercept is In A I. Deviation from linearity should
E2
R
-d-T--:I----. ........................
n T~cmoiWhn(tpc)

- (1-a)mij
(A-7)
occur gradually as the temperature gets farther below Tpc and as
distillation becomes significant.

dt 'P SI Metric Conversion Factors

tpc
°API 141.5/(131.5 + °API) g/cm 3
Also from Eqs. A-4 through A-6, bbl x 1.589 873 E-Ol m3
cp x 1.0* E-03 Pa's
OF (OF +459.67)/1.8 K
................. (A-8) in. x 2.54* E+OO cm
psi x 6.894 757 E+OO kPa
All quantities on the right sides of Eqs. A-7 and A-8 are known
except n, which can be approximated by unity. • Conversion factor is exact. SPERE
Assuming that distillation ends before the visbreaking peak, gas
Original SPE manuscript received for review March 7, 1987. Paper accepted for publication
production during the time period between the visbreaking and May 17,1988. Revised manuscript received May 5,1988. Paper (SPE 15736) first presented
coking peaks is caused by visbreaking and coking reactions only. at the 1987 SPE Middle East Oil Show held in Bahrain, March 7-10.

1316 SPE Reservoir Engineering, November 1988