PETROLEUM SOCIETY PAPER 2007-056

CANADIAN INSTITUTE OF MINING, METALLURGY & PETROLEUM

Upscaling of Kinetic Parameters for Simulation

of Reactive In Situ Bitumen Recovery
M. SADEGHI, J. ABEDI, M. POOLADI-DARVISH
University of Calgary

This paper is to be presented at the Petroleum Society’s 8th Canadian International Petroleum Conference (58th Annual Technical
Meeting), Calgary, Alberta, Canada, June 12 – 14, 2007. Discussion of this paper is invited and may be presented at the meeting if
filed in writing with the technical program chairman prior to the conclusion of the meeting. This paper and any discussion filed will
be considered for publication in Petroleum Society journals. Publication rights are reserved. This is a pre-print and subject to
correction.

EXTENDED ABSTRACT

Abstract matching of fine grid and course grid solutions, followed by a

Successful simulation of bitumen recovery processes such
as in situ combustion or reactive gravity drainage at the Alberta
Ingenuity Centre for In Situ Energy (AICISE) would require
detailed knowledge of the kinetic parameters for the chemical
reactions involved. It is known that the direct use of laboratory-
obtained kinetic data for modeling of reactive systems in
petroleum reservoirs often introduces error. It is therefore
necessary to establish the scale dependency of reaction
constants for field-scale simulation of reactive recovery
processes. Although case-specific techniques for definition of
reaction parameters in reservoir simulation applications have
been proposed, a general framework for upscaling of reaction
kinetics is not well established. We applied volume averaging
technique to establish the relationship between upscaled
reaction parameters and the grid block length scale and other
system parameters. A case study involving simulation of in situ
combustion with different grid block sizes is discussed where
the upscaled kinetic parameters are determined through direct
proper extrapolation to field-scale gird block sizes.
Introduction
Modeling of transport phenomena and chemical reaction in
porous media often requires scaling up the process from the
micro (pore) scale to the macro (continuum) scale. This
“upscaling” would make it possible to develop workable and
effective models which would not require detailed knowledge
of complex pore space and geometry within the porous medium.
Volume averaging and homogenization are the main upscaling
techniques used. The concept of upscaling discussed in this
paper is related to that implemented by the volume averaging
technique, although the goal is not to develop macroscale
transport equations. Rather, the objective is to modify the
reaction kinetics measured in the laboratory so a field-scale
reservoir simulation could be performed. This upscaling is
necessary because the direct use of laboratory-obtained
knowledge in modeling of reactive systems in natural ∂c ∂c ∂ 2c
environments often introduces errors. For example, chemical +u = D 2 − k cn
reaction rates are usually measured in well-mixed lab systems ∂t ∂x ∂x ………………………… (1)
designed to eliminate mass transfer limitations. However, in
natural porous media such as petroleum reservoirs, reactions ⎛ ∂T ∂T ⎞ ∂ 2T n
occur in individual pores with various physical and chemical ρ cp ⎜ +u ⎟ = λ 2 + k c ΔH
properties. Heterogeneities of such systems can produce mass ⎝ ∂t ∂x ⎠ ∂x ………….. (2)
transport limitations and result in spatial variations in
concentration, affecting the overall reaction rates. Equally −E
important is the question of how to select a grid block size that RT
provides reliable results when simulating chemical reactions in k = k0 e
………………………………………... (3)
petroleum reservoirs. As an example, consider a grid block of
100 m which includes a 10-meter long reaction zone. The The system is assumed to be initially at uniform temperature of
concentration changes in the reaction zone will be much larger 20°C and zero concentration of reacting component:
than that happening in the rest of grid block where only
diffusion or convection is active. For numerical solution with a c(x,0) = 0 ; T(x,0) = 20°C
finite difference scheme, the entire grid block will be assigned
an averaged value of species’ concentration. This average At (t > 0) flow of reacting component with a concentration of c0
concentration, based on which the reaction rate is calculated,
and temperature of T0 starts. Velocity is considered constant at
depends on the size of grid block which makes the simulation
u. Boundary conditions could be written as:
results highly dependant on the length scale. It is therefore
necessary to establish the scale dependency of reaction rates for
c(0,t) = c0 ; T(0,t) = T0
field-scale simulation of relevant processes in petroleum
reservoirs.
The upscaling of kinetic parameters has been studied for For the second boundary condition, the gradients of temperature
several industrial applications. Siqueira (1985) simulated the and concentration are assumed negligible at the exit of reactor
thermal recovery of bitumen for both small-scale (combustion which has a length of L:
tube) and field-scale cases and observed that for two-
dimensional field-scale simulation of in situ combustion and ∂c( L , t ) ∂T ( L , t )
=0 ; =0
steam flooding with vertical cyclic combustion, the simulator ∂x ∂x
was unable to converge to a solution. This problem was
attributed to the inadequacy of kinetic formulation, The set of nonlinear partial differential equations was
demonstrating the necessity of upscaled reaction rates. For a solved numerically using COMSOL, a finite element based
process such as in situ combustion where steep fronts occur software for multiphysics modeling. Numerical values of input
within the reservoir, a consequence of the lack of proper parameters are listed in Table 1. The resulting concentration and
upscaling of reaction kinetics is the limitations imposed on the temperature profiles are presented in Figures 1 and 2,
choice of grid block size which could have a serious respectively.
impediment in the numerical simulation. According to Chu The methodology for evaluation of upscaled reaction
(1981), to obtain meaningful results from the reaction kinetics
constant is now discussed. Since the reaction term kcn in Eqs.
measured on a small scale, the grid block size would have to be
(1) and (2) will be calculated at a single value of concentration
of the order of half a meter. Such a small grid block would
over a grid block, the intrinsic reaction constant k must be
make the field-scale numerical simulation mostly impractical.
Different methods for modifying the reaction kinetics have been modified accordingly when the concentration term is averaged
proposed (Coats, 1983; Ito and Chow, 1988; Le Thiez and over a particular length scale. This could be achieved by
Lemonnier, 1990) but no generalized framework for upscaling spatially averaging the reaction term:
of reaction constants for simulation of in situ combustion is
available. Upscaling of kinetic parameters has also studied for Δx
1
k cn = ∫ k c dx
n
applications such as simulation of gas hydrate reservoirs (Kim
et al., 1987; Hong and Pooladi-Darvish, 2005) and gel treatment Δx
0 ……………………………….. (4)
(Herbas et al., 2004a).
The examples presented in this section demonstrate that
while the kinetic parameters of industrial processes, such as in Decoupling the left hand side of Eq. (4) and incorporating the
situ combustion, has been studied extensively on small-scale, no term for average concentration will lead to:
systematic attempt to establish the dependence of reaction rate
on length scale has been made. Successful adoption of recovery n
⎡Δ x ⎤ Δx

∫ ∫
processes currently under investigation would require reliable k ⎢ c dx ⎥ = Δ x n −1 k c n dx
tools to predict the filed-scale performance of new technologies, ⎢ ⎥
making the need for proper upscaling methods more urgent. ⎣0 ⎦ 0
…………………… (5)

k is the upscaled reaction constant. Eq. (5) could be applied to

Scale dependency of kinetic parameters the tubular reactor case by inserting the definition of intrinsic
To establish the dependency of upscaled reaction constants reaction constant from Eq. (3):
on length scale and other system parameters, a non-isothermal
tubular reactor was considered where the temperature
dependency of reaction constant provides the coupling of heat
and mass transfer according to the following equations:

2
 Δx −E blocks will result in slower progress of combustion front, i.e.
k0
k x cx =
Δx ∫ e R Tt ( x ) ct ( x ) dx the location of maximum temperature was predicted to be closer
to the air injection position when larger grid blocks were used.
0 ………………… (6) Through this analysis, model parameters with the highest
Eq. (6) establishes the dependency of k on the length scale of sensitivity to the grid block size were identified.
integration and on concentration and temperature. Because of Using the model variables thus identified as tuning
the transient nature of the solution, calculation of upscaled parameters, direct matching of coarse and fine grid solutions
reaction constant should be performed at consecutive time steps would provide the upscaled reaction constants for this case. The
as Eq. (6) only provides spatial averaging. results of this procedure will be discussed, demonstrating the
The values of k from Eq. (6) and numerical solution of effects of upscaled parameters on the numerical solution.
Applications of upscaled kinetics to filed-scale simulation of in
tubular reactor case are presented in Figure 3 as a function of
situ combustion will also be discussed.
time and length scale. Comparing the k profiles with those of
concentration and temperature, it could be concluded that the
upscaled reaction constant is initially dependant on
Conclusion
concentration and temperature variations, as well as the length 1. Scale dependency of kinetic parameters has a considerable
scale. It is interesting to note that the intrinsic reaction constant impact on the accuracy of simulation results for applications
k is only a function of temperature. As the reaction proceeds such as in situ combustion.
and concentration profiles become flat compared to those of 2. An upscaling methodology is developed which will enable
temperature, the dependence of k on concentration becomes the establishment of a functional relationship between the
upscaled reaction constants and grid block size.
negligible. The relation between k and length scale depends 3. Simulation of in situ combustion with different number of
highly on the process conditions. Initially, k is decreasing by grid blocks enabled the identification of parameters with the
increasing ∆x. This trend continues as long as large gradients of highest sensitivity to the grid block size. Upscaled reaction
c and T exist, making k a complex function of all system constants were determined through direct matching of
coarse and fine grid solutions.
parameters. As the system starts to reach the steady-state and
profiles of concentration, and to a lesser degree temperature,
become flat compared to earlier stages, k starts to increase by
NOMENCLATURE
increasing ∆x. However, as explained earlier, k has become c = concentration [kgm-3]
mainly dependant on temperature at this stage of the process cp = heat capacity [Jkg-1K-1]
and the observed increase is due to the fact that T is increasing D = diffusivity [m2s-1]
E = activation energy [Jmole-1]
in the later half of the reactor.
k = reaction constant [s-1 for n = 1]
The results of this case study demonstrate the complex
k0 = constant in Eq. (3) [s-1]
relationship between the upscaled reaction constant, system
n = reaction order
parameters (concentration and temporaries) and the length
R = gas constant [Jmole-1K-1]
scale. More studies are currently underway to correlate the
t = time [s]
dependency of k on time and length scale observed here with T = temperature [K]
those relevant to field-scale simulation of reactive recovery u = velocity [ms-1]
processes. x = distance [m]
∆H = enthalpy of reaction [Jkg-1]
λ = thermal conductivity [Wm-1K-1]
Simulation of in situ combustion ρ = density [kgm-3]
The main purpose of the analysis presented in the previous
section is to identify the main parameters affecting the upscaled REFERENCES
reaction constant and to quantify the said relationship between 1. CHU, C., The Curse of Grid Size; AOSTRA Workshop
the relevant variables. While the development of upscaling on Computer Modeling (1981).
methodology is still ongoing, a preliminary attempt to apply the 2. COATS, K.H., Some Observations on Filed-Scale
technique to thermal recovery of bitumen by in situ combustion Simulation of the In Situ Combustion Process; SPE
was made. We simulated the combustion of an oil phase 12247 (1983).
consisting of three components (heavy oil, medium oil and light 3. HERBAS, J., MORENO, R., ROMERO, M.F.,
oil) in a combustion tube 1.7526 m long. The system parameters COOMBE, D. and SERNA, A., Gel Performance
and operating conditions were adopted from experimental Simulations and Laboratory/Field Studies to Design
results. Simulations were performed using 10, 40 and 200 grid Water Conformance Treatments in Eastern Venezuelan
blocks. HPHT Reservoirs; SPE 89398 (2004).
Comparing the simulation results for different number of 4. HONG, H. and POOLADI-DARVISH, M., Simulation
grid blocks demonstrated the high sensitivity of cumulative of Depressurization for Gas Production from Gas
medium oil and heavy oil production to the grid block size Hydrate Reservoirs; JCPT, 44(11), pp. 39-46 (2005).
while the cumulative production of light oil was not affected to 5. ITO, Y. and CHOW, A.K.Y., A Filed-Scale In Situ
the same degree. In addition, cumulative production of oxygen Combustion Simulator with Channeling Considerations;
in gas phase was highly dependant on the grid block size but SPE 13220 (1988).
this sensitivity was not observed for cumulative production of 6. KIM, H.C., BISHNOI, P.R., HEIDEMANN, R.A. and
water. Other system parameters were also affected by the grid RIZVI, S.S.H., Kinetics of Methane Hydrate
block size. For example, we observed that the use of larger grid

3
 Decomposition; Chemical Engineering Science, 42(7), 8. SIQUEIRA, R., Numerical Simulation of In Situ
pp. 1645-1653 (1987). Combustion; M.Sc. Thesis, University of Calgary
7. LE THIEZ, P.A. and LEMONNIER, P.A., An In Situ (1985).
Combustion Reservoir Simulator with a New
Representation of Chemical Reactions; SPE Reservoir
Engineering, pp. 285-292 (1990).

Table 1: Numerical values of tubular reactor parameters.

Parameter Value

T0 300°C
c0 1 kg/m3
u 3.6×10-5 m/s
L 1m
k0 144 s-1
D 1.2×10-6 m2/s
E 83870 J/mole
∆H 3.13×106 J/kg
λ 0.01 W/(mK)
ρ 20 kg/m3
cP 420 J/(kgK)

0.9

0.8

0.7
2.3 hr
0.6 4.6 hr
c (kg /m3)

9.3 hr
0.5 11.6 hr
13.9 hr
0.4 18.5 hr
27.8 hr
0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x (m)

Figure 1: Concentration profiles for tubular reactor.

4
 700

650

600

550

500 2.3 hr
T (K)

4.6 hr
450 9.3 hr
11.6 hr
13.9 hr
400
18.5 hr
27.8 hr
350

300

250
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x (m)

Figure 2: Temperature profiles for tubular reactor.

10.0

4.6 hr
9.0
9.3 hr
11.6 hr
8.0
13.9 hr
18.5 hr
7.0
27.8 hr

6.0

5.0

4.0

3.0

2.0

1.0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Δx (m)
Figure 3: Upscaled reaction constant as a function of length scale and time.