Академический Документы
Профессиональный Документы
Культура Документы
Summary. The principal objective of this study was to provide low-temperature-oxidation (LTO) reaction models that are
suitable for use in numerical simulators of in-situ combustion for bitumen and heavy-oil reservoirs. A systematic study investigated
the LTO reactions of the liquid-phase components of bitumen and heavy oils. Athabasca bitumen, free of water and minerals, was
oxidized by use of a laboratory-stirred semiflow batch reactor. Kinetic studies were carried out in the 333 to 423 K [140 to 300°F]
temperature range and at an oxygen partial pressure of SO to 2233 kPa [7.3 to 324 psi]. The total pressures applied in the reactor
ranged from 2190 to 441S kPa [318 to 640 psi]. Experimental data were collected in the kinetic subregime. Reactor product gas
was analyzed with a gas chromatograph, and the liquid-phase oxidation product was separated into six main components (lumped
or pseudocomponents): saturates, aromatics, Resins 1, Resins 2, asphaltenes, and coke (SARA).
Kinetic models are established for the liquid-phase L TO reactions. The models involve further lumping of the pseudocomponents.
Saturates and aromatics are lumped to forin the maltenes, Resins 1 and 2 to form the resins, and asphaltenes and coke to form
asphaltenes/ coke components.
On the basis of experimental kinetic data, two main types of reaction models are proposed: (1) a non-steady-state kinetic model
representing the overall rate of oxygen consumption and (2) four non-steady-state multi response kinetic models representing the
oxidation reactions of the liquid-phase components. Proposed models were found statistically adequate and suitable for use in
numerical simulators.
Introduction
Several articles describe in-situ combustion processes and give de- behavior. The SARA separation techniques 16,17 were used to iden-
tailed results of laboratory and field experiments. 1-12 The methods tify major reaction and product components in the liquid phase and
that have received extensive studies are dry forward combustion, to construct the lumped-components reaction schemes proposed in
wet forward combustion, and reverse combustion. It is documented this study.
that the performance of these processes depends on the L TO reac-
tions accompanying the in-situ combustion operations. 1.3,13,14 Fur- Experimental Program
thermore, results of published laboratory and field studies indicate Apparatus. A diagram of the experimental apparatus is shown in
that more meaningful analysis of combustion data cannot be made Fig. 1. The gas feed system consisted of nitrogen and oxygen/nitro-
until the L TO reaction kinetics are studied and the reaction mecha- gen gas cylinders. Gas feed rates are controlled by variable-area
nism elucidated. Consequently, a reliable numerical simulator for choke nozzles. Product gas from the reactor was passed through
performance prediction or evaluation of the in-situ combustion proc- a water condenser, an oil condenser, and a fiber filter. Pressure
esses must model L TO reactions adequately. The simulator should in the reactor was controlled by an automatic backpressure con-
include an LTO reaction model able to represent (1) the overall troller and the flow rate of the exit gas' was measured by a wet test
rates of oxygen consumption and (2) the major reaction compo- meter. Gas samples of the reactor exit gas were taken periodically
nents and products in the liquid phase and their individual rates of and analyzed by gas chromatography.
transformations. A magnetically coupled stirred reactor of Type 316 stainless steeel
Most studies reported in the literature considered only the overall with an ID of 3.49 cm [1.4 in.] and a length of 19.0 cm [7.S in.]
rates of oxygen consumption for the L TO reactions of crude was used for this study. An Autoclave Engineer's MagneDrive II
oils. 13-15 driven by a variable-speed DC motor allowed for high-pressure stir-
The kinetics data presented in this study were measured on bitu- ring without gas leakage problems. Two thermocouples were used
men from the Athabasca oil sands formation. Oil sands were ob- to measure the temperatures in the reactor. The reactor was equipped
tained from the Suncor mine in Fort McMurray, Alta. The bitumen with bitumen injection ports and valves. A nitrogen-driven float-
was extracted from the sand with the solvent toluene. Table 1 sum- ing piston injector was used to inject bitumen into the reactor at
marizes the properties and composition of the original bitumen the start of each run. Feed gas was injected into the liquid phase
sample. through a sintered stainless steel plate having an average pore size
Efforts were directed toward building mechanistic-type reaction of 20 Jlm. Gas was dispersed into the liquid phase by rotating a
kinetic models rather than empirical ones. The intricate chemical six-bladed standard turbine impeller. A six-bladed baffle was used
nature of the oil sands bitumen suggests immediately that the reac- in each reactor to improve the contact efficiency.
tion mechanisms in L TO reactions are many and complex. Because The reactor was heated by use of a fluid bed and was cooled by
the major objective of this study was to build an adequate kinetic a cold-water bath, both mounted on a swivel platform. The tem-
model for use in numerical simulators, the model should be as simple perature of the fluid-bed sand heater was controlled by a thermo-
and as practical as possible. This implies that kinetic models be electric heater controller. "The platform level was controlled by two
developed that use a lumped-component reaction scheme. Lumping hydraulic pistons.
is basically a grouping of chemical species having similar kinetic
'Now with Esso Resources Canada Ltd. Procedure. The experimental procedure consisted of preheating
, 'Now with Hycal Energy Research Laboratories Ltd. the reactor to the desired test temperature, establishing the desired
Copyright 1987 Society of Petroleum Engineers gas flow rate and stirrer speed, and rapidly injecting 0.06 kg [0.l3
SPE Reservoir Engineering, November 1987 573
BH 1 = ice water bath H = fluidized bed heater with QV = quick mechanism valve
BH2 = dry-ice/methanol bath controller R = rotameter
CRT = pressure controller MSC = motor speed control RT=reactor
CV = check valve NV = needle valve RCV = research control valve
D/P = differential pressure cell NRV = no return valve SV = fluidized bed sand valve
F = fiber filter P = pressure gauge TACH = tachometer
FSB = fluidized sand bed PIS = pressure switch TC = thermocouple
V = regulating valve
(j)
C/}
«
A~A
40.0
~C
::::iE
~
ID A ___
I-
Z
W
o
L-_"7_----, 4
10
II: "7 NOTATION
w
X SATURATES
~30.0 "7
/
0 'AROMATICS
C/)
A OILS
!r
o RESINS I
() RESINS II
"7 ASPHAL TENES
• COKE
a "7
0 20 .0 , 0
1100~~~--"":'----:----------"':~~~..o: 10
o () ___
Z ----.() 423
a
o TEMPERATURE ( K)
TIME (h)
REACTOR CONDITIONS
104
Fig. 3-Conversion of bitumen at 480 K. PRESSURE 4390 k Pa 098
I RUN NO
TEMP
I (K)
I,GAS IRUN
COMP TIME! h)
1.0 • I 511 I 393 I AIR I 8.0 097
• I 92 I 378
I AIR I 4.0
Once the operating conditions were established, a series of isother-
mal runs was carried out at 423, 408, 393, 378, and 333 K [300,
275, 248, 221, and 140°F]. The inlet oxygen partial pressure ranged 1.02
·· ORIGINAL BITUMEN
OILS
96
from 40 to 2233 kPa [5.8 to 324 psi]. To ensure operation of the
reactor in the kinetic subregime, all kinetic data used for model
building were collected at gas fluxes above 750 cm 3 /(cm 2 • min) ':'0';; 1.01 095
'"'2
.
(standard conditions) and impeller speeds of more than 3,000
",-
,
2 '",2
rev/min. ...'" 1.00 094
~
A typical set of curves of concentration vs. reaction time obtained ...en>- ...en
>-
from these runs is present in Fig. 3 of kinetic reactions at 408 K zt.J z
t.J
[275°F]. These data indicate that the LTO reactions of bitumen con- 0 0
0.99 093
sist of a series of consecutive steps resulting in the formation of
the asphaltenes/coke component from the oils, with the resins as
the intermediate components. Similar behavior was observed for 0.98 0.92
the kinetic reactions carried out at the other temperatures. Note the
deviation of a single data point from the Resins 1 curve. This oc-
curred for the I-hour reaction-time run. Similar deviations were
091
observed for data collected at 423, 393, and 378 K [300, 275, and
248°F]. Insufficient data were collected during the early reaction
time to allow reasonable explanations of the deviations. Further
studies are needed to understand them. 353°90
The following are significant observations arising from the ex- TEMPERATURE IK)
perimental work. (Details are presented in Ref. 14.)
1. Marked changes occurred in the properties of natural bitumen Fig. 5-Effect of LTO on the density of Athabasca bitumen.
When subjected to oxidation. Both the density and apparent viscosity
increased with severity of oxidation. Oxidized bitumen viscosities
were orders of magnitude higher than those of the original materi-
al. Figs. 4 and 5 show the effect of oxidation on the viscosity and
density of original bitumen, respectively.
2. The oxidation of maltenes (asphaltenes/coke-free component)
resulted in the formation of the asphaltenes/coke components. Ox-
SPE Reservoir Engineering, November 1987 575
NOTATIONS GAS OUT ....- - - . k----IMPELLER
.. OXYGEN
OILS
"
o RESINS I
:c-
<) RESINS II Z
fii
.• 10.0 W
en 0 RESINS ~
< ASPHALTENES ::::>
~ I-
>- 100.0
COKE iii
!D TEMP. - 423 K 0,
I-
Z
W
RUN TIME - 1.6 h ~
(5
U
~
80.0 6.0
II:
W
eo. ~
60.0 b ....---+- LIQUID PHASE
en 4.0
I-
Z X {]
W Z
Z
"---
0
Q.
~
50.0
.. 2.0 Q
Ii: t
~
0 ::::> GAS IN
U 40.0 "- en
u.
0
, ::::J 0.0 z
0 Fig. 7-Semiflow batch reactor system: well-mixed-gas/well-
~.-
U
Z u. mixed-batch-liquid configuration.
0 30.0 0
t=
< "".'0 W
II:
I- ----0_
~
Z
W
20.0
Q
=l .) II:
..J
..J
Apply Henry's law to relate the concentration Cj of dissolved gas,
Component j, to its partial pressure Pj at equilibrium .
U 3 ___ <
Z <) II:
0 0
W
U 10.0 >
Yj*=-=--C
Pj K He *. ................................ (2)
<)
0 0
j
• Pt Pt
0.0 100.0 200.0 300.0 400.0 700.0
Liquid Phase. The transfer rate of Componentj from the gas into
the liquid phase minus the rate of chemical reaction of Component
idation of the asphaltenes (pentane-insoluble fraction) resulted in
j in the liquid phase equals the rate of accumulation of Component
a high yield of toluene-insoluble material (coke).
j in the liquid phase.
3. No significant effect of total pressure on reaction rates was
observed.
4. Oxygen partial pressure had a relatively strong influence on
the overall rate of oxygen consumption and a lesser effect on the
oxidized oil composition. Fig. 6 summarizes the effect of oxygen
partial pressure at 423 K [300°F]. Similar behavior was observed For chemical reactions occurring in the kinetic subregime, the over-
at other temperatures. The overall rates of oxygen consumption all driving force is almost entirely used up by the reaction, so that
reported in Fig. 6 were estimated from the difference in the molar in practice the liquid phase is saturated with the absorbing gas; i.e.,
flow rates of oxygen measured at the reactor inlet and outlet. Yj=Yj*. Therefore,
5. Analysis of the reactor exit gas showed that carbon oxides were
the only reaction products in the gas phase. Carbon oxides were KgPt(Yj -C/KHeIPt)
present only in the 418 and 423 K [293 and 300°F] tests.
6. Overall rates of oxygen uptake increased rapidly during the is negligible. Therefore, Eq. 4 becomes
early stages of the reaction, reached a peak, and then declined gradu-
ally. Steady· state compositions were not attained over the reaction d
periods considered. -rjV/=-(V/Cj ). . ............................... (5)
dt
Design Equations for the Semiflow Batch Reactor Interphase Material Balance for Oxygen. The inlet rate of oxygen
The reactor configuration used in this study is essentially that of into the reactor minus the outlet rate of oxygen from the reactor
well-mixed-gas/well-mixed-batch-liquid and is illustrated in Fig. plus the trahsfer rate of oxygen from the gas into the liquid phase
7. Both phases are intimately mixed by mechanical agitation and minus the rate of chemical reaction of oxygen in the liquid phase
the action of the gas on the liquid. The concentrations in both phases equals the rate of accumulation of oxygen in the liquid phase.
are assumed to be time-dependent but spatially invariant. The fol-
lowing are the general non-steady-state component material balances
for the reactor.
d
Gas Phase. The inlet tate of Component j into the reactor minus =-(V/C02 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6)
the outlet rate of Component j from the reactor minus the transfer dt
rate of Component j from the gas into the liquid phase equals the Assuming that the reaction takes place in the kinetic subregime,
rate of accumulation of Component j in the gas phase. we can assume that equilibrium concentration of oxygen exists in
the liquid phase. The overall driving force is thus entirely used up
in the chemical reaction; Le.,
C-4A
~
As
(asphaltenes/coke)
C-4B A k 23 ,E 23 A k34,E34 A k 45 ,E 4S A
2 »3 >4 »S
(oils) (Resins 1) (Resins 2) (asphalteneS/coke)
and
TABLE 3-RATE DATA FOR MODEL C-1
K gPt(Y0 2 -Y0 2*) Reaction Model
is negligible.
dy FHeqg
Eq. 6 can be written - = - - ( Y o . - y)exp( - L IRT)
dt VIP! 2' v
where r02 {C02*} denotes the rate of consumption of oxygen in the E' =E+(1 +m 1 )L v
liquid phase as a function of CO2 .
Frequency Factors
Formulation of the Kinetics Models Arrhenius' law constant, F Ko'
cm 3 ·mol- 1 ·h- 1 8.066 x 10 29
Basically, two main types of models were proposed: one to account Henry's law constants, F He ,
for the overall rate of oxygen consumption (Model C-l) and the kPa·cm 3 ·mol- 1 3.678 x 10 2
other (Models C-2 to C-4) to represent the reaction rates of the bi-
tumen pseudocomponents. Four reaction schemes, comprising the Activation Energies, J/mol
bitumen pseudocomponents, were derived by use of the product E 238.422
distribution curves obtained for the series of isothermal reactions Lv 6.774
carried out. Fig. 3 represents typical product distribution curves
obtained for the LTO reactions. Reaction schemes were developed Rate Constants
on the basis of two (Model C-2), three (Model C-3), and four Temperature k K He
(Models C-4A and C-4B) pseudocomponents. Table 2 summarizes (K) (cm3.mol-1.h -1) (kPa ·em 3 . mol-I)
the reaction schemes. As shown by the reaction models, the pseudo- 423 4.34 1.036 x 10 2
components represent lumped fractions from the SARA analysis. 408 1.279 6.853 x 10
393 1.439 x 10- 2 9.816
Overall Rate of Oxygen Consumption (Model C-l). An unsteady- 378 9.281 x 10- 4 3.748
state model representing the overall rate of oxygen consumption 333 1.850 x 10 -10 1.835x10- 2
was derived from the interphase material balance for oxygen (Eq.
8). When the exponential form 14,18 of Henry's law is applied to Order of reaction with respect to oxygen partial pressure: m 1 = 2.0.
..................................... (9)
where
10'
• 05 X '0"~-------------------,5.0 X 10'
o
-2
10
I-
I- Z
Z w
u.. u
U
MODEL C-1 u:: tEw
u.
w a
NOTATION a u
u -3 10-
4
~
k [kPa.mol- ' .Cm- 3 .h- ' ] w'O r~------------~ <{
~ MODEL C·3
eK He [mol- ' .kPa.cm- 3 ] a: • a:
• k' [h-'.kPa-0 50 J
65
-8 o k' 26 I h-'.kP"o"J
10
-5
-9 o '0
10
-'0
10 L-____~~----~----~~----~~----~~~ 3.0 X 1O-Sll::.9--,,!2.""0----2"L.2,.----;2f;.4,-----;;2!;;.6---;!-o-----,;'n~
2.0 2.2 2.4 2.6 2.8 3.0
RECIPROCAL TEMPERATURE
RECIPROCAL TEMPERATURE (1fT) X 103 [K- ' ]
Fig. 9-Model C-3: reaction rate coefficients vs. inverse ab-
Fig_ a-Rate coefficients vs_ increase in absolute temperature. solute temperature.
for a constant molal flow process, mation techniques, which involve the removal of singularities from
the multiresponse kinetic data and the minimization of a determinant
AYo 2 =Y02 i -y, .................................. (11) criterion derived from the Bayes theorem. 14,19,20 Parameter esti-
mates of Models C-3 and C-l are presented in Tables 3 and 4, re-
spectively.
Yo 2 P=Y ........................................ (12)
Note that Model C-l contains both the Arrhenius and the Hen-
ry's law parameters and that these parameters were estimated from
for a mixed reactor, the oxygen consumption data. Model C-l also predicts that the or-
der of the overall oxygen consumption reaction is second order with
Fk~=FkoFHe I-ml, ............................... (13) respect to oxygen partial pressure.
Fig. 8 shows the Arrhenius plot for Model C-1 over the temper-
and ature range 333 to 423 K [140 to 300°F). A good correlation is
observed over the total temperature range. The plot of Henry's law
E'=E+(l-ml)L.. . .............................. (14) coefficients vs. inverse absolute temperature isalso given in Fig.
8. This shows the expected linear relationships.
Among the LTO lumped-component models, Model C-4B is the
Details of derivations of Eq. 9 are presented in the Appendix. best model according to analysis for lack of fit. Only Model C-3
is described in detail in this paper because the model is simpler
Bitumen Lumped-Component Models (Models C-2 to C-4). The than the four-component models and yields excellent predictions
saturate and aromatic components were not represented individually of the LTO reactions. Detailed presentations for all the models are
in the reaction schemes. They were lumped together and represented provided in previous publications. 14,21
by the oil component. Similarly, the asphaltenes and the coke com- Fig. 9 shows the Arrhenius plots for two reaction rate coeffi-
ponents were lumped together. cients contained in Model C-3. It is observed that the Arrhenius
On the basis of the time-dependent lumped-component distribu- plots are linear only in the temperature range 378 to 423 K [220
tion curves, nonsteady-state kinetic rate expressions were developed to 300°F). This implies that the model does not adequately repre-
for the four reaction models with Eq. 5. Table 3 presents the rate sent the oxidation reactions at 333 K [140°F). The final parameter
expressions for Model C-3. The Appendix presents details of deri- estimates presented in Table 4 were determined with kinetics data
vations of the rate expressions in Table 3. obtained in the temperature range 378 to 423 K [220 to 300°F).
This same behavior was noted for the two- and four-component
Summary of Results models, and further experimental work is required to explain the
Mathematical expressions for the kinetic models are obtained by change in reaction kinetics.
incorporation of the kinetic rate equations into the appropriate reac- Fig. 10 reproduces curves for concentration vs. reaction time at
tor design equations summarized in Eqs. 3, 5, and 8. The invari- 408 K [275°F) for Model C-3. This figure shows that the experimen-
ant kinetic parameters in the proposed multiresponse reaction models tal curves of concentration vs. reaction time are well reproduced
were estimated by use of the maximum-likelihood parameter esti- by the model. Similarly, good predictions of the experimental data
578 SPE Reservoir Engineering, November 1987
TABLE 4-RATE DATA FOR MODEL C-3
z-
00.60 MODEL C-3 Reaction Scheme
t« TEMPERATURE 408 K
.OILS
0:
u.. j RESINS A 2 k 26 ,E 2s" As k 65 ,E ss,. A 5
en 0.50 • ASPHALTENES-COKE (oils) (resins) (asphaltenes/coke)
en EXPERIMENTAL
« •••• j
~0.40 dX m
Z -A2=-[Fico X A exp(-E2s IRT»p o 5
w dt 26 2 2
Z
~ dX m
-As=Fico XA Po 5 exp(-E2sIRT)
~0.30 dt 26 6 2
U
u..
o
ZO.20·
o dX m
-A5=[Fico X A exp(-E Ss IRT»p o 6
~
0:
dt 65 6 2
~
~ 0.10
U
Z
o
U Frequency Factors, h -1 .kPa om;
o 1.0 2.0 3.0 4.0 5.0
F
ko ,h -1 .kPa -0.41
F' 26 h -1 • kPa - 0.50
2.638 x 10 7
7.688 X 10 8
TIME (h) k0 65 '
KHe=FHe exp(-LpIRT), ......................... (A-4) Liquid-Phase Lumped-Components Models. The saturates and
aromatic components are not represented individually in the reaction
scheme. They are lumped together and represented by the oil com-
where FHe is the integration constant. This form of Henry's law ponent. Similarly, the asphaltenes and the coke components are
is most suitable for use in developing the rate equations and is ap- lumped together.
plied in this study. Development of the rate equations are illustrated for the three-
component Model C-3. The reaction rietwork for this model is
Model c-l: Overall Rate of Oxygen Consumption. An unsteady- shown in Table 4. The rate expressions for this scheme can be de-
state model representing the overall rate of oxygen consumption veloped from the uhsteady-state material balance of Eq. S.
can be derived from the interphase material balance for oxygen
(Eq.8).
d
-rjV[= -(V[9 . .............................. (A-14)
dt
dCA2 m
dCA
Two further assumptions are made to simplify the solution of Eq. __6 -k C em 5 -kc65 CA6 Cm6
- C26 A2 02 ° (A - 17)
A-6: (1) the total concentration of the hydrocarbons, Ch , remains 2 , •.............
dt
essentially unchanged by chemical reaction and (2) the change in
density of the liquid phase resulting from chemical reactions is in- and
significant.
Eq. A-6 can be written as
dCA5
- - = kc CA C06 • •..••.••••••.••••.•..••.•• (A-I8)
dt 65 6 2
The rate constahts, kqj , and the Henry's law constant, K He , can
be expressed as a function of temperature, T, with the Arrhenius-
Dividing Eq. A-7 by total pressure, Pt, and rearranging, one Substitution of Eqs. A-19 and A-20 into Eqs. A-16 and A-18 and
obtains
application of Eq. 2 yield the following:
where
where and