Thermal Transients During Nonisothermal Fluid Injection PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Journal of Petroleum Science and Engineering 42 (2004) 133 – 144

www.elsevier.com/locate/petrol

Thermal transients during nonisothermal fluid


injection into oil reservoirs
Ibrahim Kocabas *
Chemical and Petroleum Engineering Department, UAE University, Al Ain, United Arab Emirates

Abstract

During hot fluid injection into oil reservoirs, the importance of determining the temperature profile to estimate the thermal
efficiency is well known. In addition, the resultant temperature distribution due to cold water injection into a hot reservoir
during waterflooding may significantly influence the stress distribution in the reservoir. Owing to these facts, in this work we
present a new two-dimensional analytical model for analyzing the thermal transients during nonisothermal fluid injections into
oil reservoirs that may provide a better insight into the mechanisms of heat transfer in oil reservoirs.
The new model has several distinguishing aspects. Primarily, it is the first analytical solution of an unsteady state two-
dimensional heat transport process in a laterally/vertically confined layer. In addition, both finite longitudinal and transverse
heat dispersions have been accounted for in the model as well as the heat loss to the bounding layers. Thus, the model allows
one to observe the roles of both boundary conditions and fluid mechanics controls simultaneously, a feature not possessed by
the previous analytical models that assume either boundary conditions or fluid mechanics controls. Hydrodynamic heat
dispersion concept has also been incorporated in the new model in order to account for the much larger temperature transition
zones observed in the field than that would be caused by pure conduction. Finally, since the solution is free of numerical
dispersion and nonphysical oscillations, especially in two-dimensional domains, it serves as a higher stepping stone for
validation of numerical models.
D 2003 Elsevier B.V. All rights reserved.

Keywords: Thermal transients; Convection dispersion equation; Thermal dispersion; Nonisothermal flow; Heat transport

1. Introduction great importance. In addition, the temperature depen-


dence of processes such as acid reaction rate and the
All modern secondary and/or enhanced oil recov- injected material properties such as polymer rheolo-
ery methods utilize the injection of a non-isothermal gy makes an accurate estimation of temperature field
fluid into an oil reservoir. First of all, in case of a a requirement in many EOR methods. Finally and
heat carrying fluid injection for a thermal recovery most importantly, the geomechanical aspects of non-
technique, determining the temperature profile and isothermal flow is of increasing concern in the oil
consequently estimating the thermal efficiency are of industry such as the case of cold water injection into
a hot reservoir, where the thermal and fluid pressure
gradients could be used to estimate the net stress
* Tel.: +971-3-705-1546; fax: +971-3-762-4262. distribution, and hence, the possibility of fracture
E-mail address: ikobacas@uaeu.ac.ae (I. Kocabas). development.

0920-4105/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.petrol.2003.12.006
134 I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144

A comprehensive modeling of the temperature stages of numerical simulation of convective disper-


transients in the porous media can best be achieved sive processes (Peacemean and Rachford, 1962).
by the numerical models. However, in the implemen- Later on, significant progress has been made on
tation of numerical models there are some points to be quantification (Lantz, 1971) and elimination (Leo-
understood and several critical issues to be resolved nard, 1981; Leonard et al., 1995; LeVeque, 1996) of
for the practicing engineer. First of all, gaining a good numerical dispersion. However, numerical dispersion
insight into the physical processes influencing the and nonphysical oscillations still continue to be major
transport and accurately accounting for them in the problems in numerical modeling of heat transport
numerical model is of vital importance. Analytical (Gupta, 1990; Kocabas and Margoub, 2000). Hence,
models are quite useful to provide the basic insight for quantifying the accuracy of numerical models,
into the mechanisms of heat transport in porous comparison of the numerical model results with the
media. Hence, a number of analytical models have available analytical models is of vital importance too.
been presented on the temperature distributions due to Therefore, development of more sophisticated an-
a purely convective flow in linear (Lauwerier, 1955), alytical models serves invaluably for both gaining
and radial (Malofeev, 1963) reservoirs and heat losses better insight into the transport mechanisms and
to the confining impermeable layers of infinite length. numerical model validation purposes. In addition,
Rubinshtein (1959), on the other hand, presented an the analytical models are valuable tools for studying
expression for the heating efficiency without deriving the collective role of the parameters influencing the
a temperature distribution for a convective flow in a transport (Kocabas and Islam, 2000a,b).
radial layer with both longitudinal and transverse
thermal conductivities in both the flowing and its
confining layers. Then, Avdonin provided temperature 2. Theoretical development
distribution solutions that incorporate an additional
finite longitudinal (Avdonin, 1964a) or transverse The design and evaluation of the hot fluid injection
(Avdonin, 1964b) thermal conductivity term in the into an oil reservoir depend on the ability to predict
Lauwerier and Malofeev models. Later on, Spillette the thermal transients that occurs in the field during
(1965) summarized the previous studies and presented the project life. Such a prediction requires, in addition
a numerical solution to the problem of hot water to the description of reservoir and thermal parameters,
injection into a horizontal layer, where assumptions an accurate accounting of the thermal energy utilized
made by Lauwerier and Advonin are relaxed by in the process. Therefore, an understanding of the heat
incorporating finite longitudinal and transverse ther- transfer mechanisms in porous media is essential to
mal conductivities. For both Lauwerier and Malofeev the development of the heat balance equation that
models, Spillette (1965) has also reported the formu- accurately represents the processes involved.
las for the thermal efficiency that is defined as the In this work we present a new two-dimensional
fraction of the total injected energy that remains in the analytical model for analyzing the thermal transients
productive layer. Further improvement of solutions to during nonisothermal fluid injections into oil reser-
be able to model various cases such as finite confining voirs. The new model has several distinguishing
layers separated by multiple fractures (Gringarten et aspects regarding both describing the processes and
al., 1975; Satman, 1988), modeling thermal injection as well as accurately accounting for them in modeling.
backflow tests (Kocabas and Horne, 1990), applica- The detailed features of the new model are presented
tions to naturally fractured geothermal reservoirs in the following.
(Cendejas and Rodriguez, 1994) and development of
analogies to tracer transport (Ramirez et al., 1993; 2.1. Heat transfer mechanisms and thermal disper-
Kocabas and Islam, 2000a,b) has continued up to sion concept
present.
The issues of numerical dispersion, grid size and The proposed theory in this work for heat transfer
orientation effects and other behavioral errors such as mechanisms during a hot fluid injection into an oil
nonphysical oscillations were recognized at very early reservoir has originated from the work of Sauty et al.
I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144 135

(1978). Based on this theory the heat transfer mech- 2.2. Two dimensional analytical model
anisms during non-isothermal flow in porous in media
may be divided into three major types. The analytical solutions for temperature distribu-
The first type is the energy transfer due to physical tions presented in earlier works (Lauwerier, 1955;
movement of the injected fluid. In other words, the Malofeev, 1963; Avdonin, 1964a,b; Gringarten et al.,
average bulk movement of the injected fluid in the 1975; Satman, 1988; Kocabas and Horne, 1990;
reservoir causes a forced heat convection. Ramirez et al., 1993; Cendejas and Rodriguez,
The second type of heat transfer is due to hydro- 1994; Kocabas and Islam, 2000a,b) are all one-
dynamic dispersion. The velocity variations in both dimensional solutions. In addition they can account
direction and magnitude in the porous media couple for only one type of dominant heat transfer mecha-
with the thermal conduction and give rise to a phe- nism, and hence, can be divided into two types. The
nomenon called hydrodynamic thermal dispersion. first types of solutions are those that assume the heat
The variations in the velocities are due to velocity losses to impermeable bounding layer, and hence, the
profile in a single pore, velocity differing from one boundary conditions dominate. The first and the
pore to another and tortuosity of the streamlines. most well-known of these solutions is due to Lau-
While the variations in the velocity vector cause extra werier (1955). The Lauwerier model assumes a
spreading of the transition zone between hot and cold linear, one-dimensional incompressible flow in a
fluids, the thermal conductivity makes this mixing homogeneous layer, constant physical properties
irreversible. In other words, upon the reversal of the and saturations, infinite thermal conductivity in
flow direction the transition zone does not become transverse direction and zero longitudinal thermal
smaller, on the contrary, it becomes larger. Such a conductivity in the reservoir, finite and equal trans-
hydrodynamic dispersion occurs in both longitudinal verse thermal conductivities and zero longitudinal
and transverse directions and the corresponding ther- thermal conductivities in the bounding layers, and a
mal dispersion coefficients are defined as (Sauty et al., constant injection temperature.
1978): The solution to the Lauwerier problem for the
radial flow geometry was presented by Malofeev
(1963). The two other solutions regarding the temper-
KL ¼ k þ aL qw cw uL ð1Þ ature distributions due to hot liquid injection were
presented by Avdonin (1964a,b). His first work in-
Kt ¼ k þ at qw cw ut ð2Þ cluded a constant longitudinal conductivity for both
linear and radial geometries. All other assumptions
were identical to those made by Lauwerier. Later, he
Additionally, heat is transferred by thermal con- presented solutions for a finite transverse conductivity
duction from the reservoir into the virtually imperme- in the reservoir but the longitudinal one was assumed
able confining layers. This assumption of virtual to be negligible. Such solutions are applicable to
impermeability allows us to use conductivity in the relatively thin productive layers. Then, slight modifi-
confining layers rather than dispersion. cations to Avdonin’s first model were presented by
Finally, the convective transfer between the Gringarten et al. (1975), Satman (1988), and Kocabas
injected fluid and the original reservoir fluids and and Islam (2000a,b).
solid material should be considered. Fortunately, the The second type of models assumes that the
velocities in the reservoir are low enough to justify the reservoir is thick enough to neglect the heat losses
assumption of instant thermal equilibrium. In fact, it is to the impermeable bounding layer. However, the
reported that the temperature equilibrium is reached in reservoir is assumed to consist of two superposed
1 s for 1 mm, in 1 min for 1 cm and in 2 h for 10 cm continua namely a highly permeable fracture network
diameter grains (Sauty et al., 1982). and a relatively small permeability matrix blocks.
Consequently, the concept of thermal dispersion is Such models are considered to be controlled by fluid
important to be included in the modeling, and hence, mechanics. The major heat transfer mechanism is the
it forms a distinguishing feature of the new model. one that occurs between the fractures and matrix
136 I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144

blocks. Such types of models are based on tracer Based on the assumptions of the conceptual model
transport models (Ramirez et al., 1993) or geothermal described in the previous section, the governing
models (Cendejas and Rodriguez, 1994) and are equations become:
applicable to relatively thick and, also, densely frac-
tured reservoirs. BT BT B2 T B2 T
qr cr þ /uqf cf  KL 2  Kt 2 ¼ 0; ð8Þ
In the new analytical model presented here, the Bt Bx Bx Bz
assumptions in Lauwerier’s model have been relaxed
to include both a finite longitudinal dispersion (usu- BTm B2 Tm
ally neglected in earlier models) and a transverse qm cm  km ¼ 0: ð9Þ
Bt Bz2
thermal dispersion in the productive layer, making
the solution two-dimensional. In order to build a
The initial and boundary conditions are specified
conceptual model, first we assume a linear flowing
as follows:
unit confined by two impermeable layers. We also
assume an incompressible fluid flow with constant
T ¼ Tm ¼ T0 at t¼0 ð10Þ
linear flow velocities in x and z directions leading to
constant longitudinal and transverse thermal disper-
sion coefficients. In addition, the assumption that the T ¼ Ti at x¼0 ð11Þ
velocity in z direction is negligible compared to that in
x direction allows us to neglect the convective trans-
port in z direction and yet preserve the thermal Both mediums are assumed to be semi-infinite:
transverse dispersion term in the governing equations. T !0 as xD ! l ð12Þ
Finally, we assume constant and equal heat conduc-
tion coefficients in the bounding layers. Tm ! 0 as zD ! l ð13Þ

2.3. Governing equations and development of solution Then the two equations are coupled through the
equality of temperatures and fluxes at the boundary.
The two-dimensional governing differential equa-
tions of the heat transport during nonisothermal fluid T ¼ Tm at z¼b ð14Þ
injection into an oil layer where no phase change and
instantaneous equilibrium of injected and resident
liquids are assumed are as follows: BT BTm
Kt ¼ km at z¼b ð15Þ
Bz BZ
Bðqr cr T Þ
þ j  ðqf cf ðq=AÞT Þ  j  KjT ¼ 0 ð3Þ
Bt The mathematical formulation is simplified using
following dimensionless variables:
and
T0  T T0  Tm
Bðqm cm Tm Þ TD ¼ and TmD ¼ ð16Þ
 j  km jTm ¼ 0; ð4Þ T0  Ti T0  Ti
Bt

where /qf cf ux ð/qf cf uÞ2 t


xD ¼ ; tD ¼ ; and
KL qr cr KL
q=A ¼ /u and qf cf ¼ qw cw Sw þ qo co So ð5Þ /qf cf uz
zD ¼ pffiffiffiffiffiffiffiffiffiffi ð17Þ
KL Kt
qr cr ¼ /qf cf þ ð1  /Þqs cs ð6Þ
qm cm Kt km pffiffiffiffiffi
h1 ¼ ; h2 ¼ ; and h ¼ h2 h1 ð18Þ
qm cm ¼ /m qw cw þ ð1  /m Þqs cs ð7Þ qr cr km Kt
I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144 137

In terms of the dimensionless variables the govern- TD ¼ TmD at zD ¼ zDb ð25Þ


ing equations and initial and boundary conditions
reduce to:
BTD BTmD
¼ h2 at zD ¼ zDb ð26Þ
BzD BzD
BTD BTD B2 TD B2 TD
þ   ¼ 0; ð19Þ
BtD BxD Bx2D Bz2D
In order to derive the solution, we have first
performed a dependent variable transformation for
BTm B2 Tm dimensionless temperatures in both regions:
h1  ¼ 0; ð20Þ
BtD Bz2D
TD ¼ vexpðxD =2Þ; ð27Þ
TD ¼ TmD ¼ 0 at tD ¼ 0 ð21Þ
TmD ¼ vm expðxD =2Þ: ð28Þ
TD ¼ 1 at xD ¼ 0 ð22Þ
Then we have applied the Laplace and Fourier sine
transforms to the equations of v and vm. Finally,
TD ! 0 as xD ! l ð23Þ
applying the initial and boundary conditions we have
obtained the following solution in the Laplace-Fourier
TmD ! 0 as zD ! l ð24Þ space (Kocabas, 2001).

" pffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi   pffiffiffiffiffiffiffiffiffiffiffi  #


2x hs exp s þ k zD þ exp  s þ k zD
v̈ ¼ 1  pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffi  pffiffiffiffiffiffiffiffiffiffiffi  ð29Þ
psðs þ kÞ s þ k þ hs exp s þ k zDb  s þ k  hs exp  s þ k zDb

where k = x2 + a.

To simplify the derivations we have assumed that space solution for the dimensionless temperature in
the parameter h has a unit value. This assumption is the oil layer:

made in order to obtain a real space solution. After 1 xD  tD


manipulating Eq. (29) considerably and using tables TD ¼ erfc pffiffiffiffi
2 2 tD
of Laplace transforms together with the convolution

xD þ t D
theorem, we have obtained the following Fourier sine þ expðxD Þerfc pffiffiffiffi  TDsinv2 ð30Þ
2 tD

where
l Z
X tD
expððð2n þ 1ÞzDb  zD Þ2 =4sÞ þ expððð2n þ 1ÞzDb þ zD Þ2 =4sÞ
TDsinv2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Fsinv ds ð31Þ
n¼0 0 p sðtD  sÞ

and
Z l 2

1 x þa x þa
Fsinv ¼ exp  ðtD  sÞ I nþ1=2 ðt D  sÞ sin xxdx: ð32Þ
0 ðx2 þ aÞ1=2 2 2
138 I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144

To carry out the above inverse Fourier sine trans- Then we have performed a term-by-term inversion
form (Eq. (32)), we have utilized the following using the tables of Fourier sine and cosine integral
relation which is selected as the most convenient transforms, and the theory of convolution. For exam-
one among the available Modified Bessel Function ple when n = 0, the first approximation to the term
In + 1/2 representations: TDinv2 in the solution becomes:
rffiffiffiffiffi Z
2y 1 0 expððzDb  zD Þ2 =4sÞ þ expððzDb þ zD Þ2 =4sÞ
tD
Inþ1=2 ðyÞ ¼ TDinv2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p 2y 0 p sðtD  sÞ
8 X 9 8

9
>
>
n
ðn þ mÞ! m >
> > erfc s pxffiffiffiD  expðx Þerfc s þpxffiffiffiD  >
> >
> ð2yÞ expðyÞ > 1< =
D
< m!Cðn  m þ 1Þ = 2 s 2 s
m¼0

ds
Xn : > erfc D pffiffiffiffiffiD þ expðx Þerfc D pffiffiffiffiffiD >
2> t  x t þ x >
>
> ðn þ mÞ! > : ;
>
: ð1Þ
n
ð2yÞm expðyÞ >
>
; 2 tD
D
2 tD
m¼0
m!Cðn  m þ 1Þ
ð33Þ ð34Þ

For n = 1, the second order approximation to the term TDinv2 in the solution is obtained as:

Z
1 0 expðð3zDb  zD Þ2 =4sÞ þ expðð3zDb þ zD Þ2 =4sÞ
tD
TDsinv2 ¼ TDinv2 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0 p sðtD  sÞ
8 
 
9
> 1 s  xD s  xD s þ xD 1 s þ xD >
> þ
< 2 t s erfc p ffiffi
ffi  þ expðx D Þerfc p ffiffi
ffi >
=
D 2 s tD  s 2 2 s
 
 
ds ð35Þ
>
> t  xD 1 t D  xD t D þ xD 1 t D þ xD >
: D þ erfc pffiffiffiffi þ þ expðxD Þerfc pffiffiffiffi >
;
tD  s 2 2 tD tD  s 2 2 tD

The equations for n = 2 are quite complex and do In the two dimensional temperature distributions,
not lend themselves to a practical solution easily. the following features are of great importance,
However, as discussed in the average temperatures namely, the extent of the transition region in both
section, the accuracy of this second order approxima- lateral/vertical and longitudinal directions and mag-
tion is quite adequate for the computational purposes, nitude of gradients in each direction. These features
and for most of the time we have utilized this form to may be inferred from the number of grids with
compute the temperature profiles. similar shape, size and orientation. The square
shaped small grids indicate regions of constant
temperatures at or close to the injection temperature.
3. Two-dimensional temperature profiles in the oil The rectangular shaped grids indicate regions of
layer temperature variations, the larger the aspect ratio of
a rectangle the greater the temperature gradient.
The two-dimensional temperature distributions Finally, the orientation of rectangles coincides with
give us a general idea about the temperature surface the direction of the gradient.
in the oil layer. Such information may be used to Fig. 1 shows that the temperature surface for a large
design and interpret the fluid injection operations into Peclet number and a small L/b ratio consists of three
oil reservoirs. It may also be used to justify/discard a distinct regions. The small size square grids occupy the
one-dimensional model. Figs. 1, 2 and 3 show tem- largest area indicting that most of the reservoir is
perature gradients developed along the transport di- heated to the injection fluid temperature. Then, there
rection as well as across half of the injection plane of exists a narrow region of high lateral/vertical temper-
the reservoir, where the other half would be the exact ature gradients next to the confining layer boundary.
mirror image. Finally, there is a region of high temperature gradients
I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144 139

Fig. 1. Dimensionless temperature distributions across half of the injection plane, Pe = 100, KL/Kt = 10, L/b = 10.

along the flow direction. From these observations, we zone gets larger the temperature gradients become less
expect a high heating efficiency. pronounced.
Fig. 2 shows the effect of increasing dispersivity Fig. 3 shows the effect of a large L/b ratio
(smaller Peclet numbers) on the temperature distribu- representing a long thin reservoir. In this case, the
tion. There are also three distinguishable regions in transition zone develops only along the flow direction
Fig. 2. Here the increasing dispersivity reduces the extending over the whole flow distance. The
portion of the reservoir heated close to the injection smoothed transverse gradients indicate the total dom-
fluid temperature. The two transition regions near the inance of heat losses through the confining layer
confining layer boundary and along the flow direction boundary leading to a low heating efficiency.
occupy relatively larger portions than that in Fig. 1. More importantly, however, is that two-dimension-
These regions are characterized by large rectangles al temperature distribution surfaces serve a higher
with small aspect ratios. Thus, while the transition stepping stone for validating numerical models. Nu-

Fig. 2. Dimensionless temperature distribution across half of the injection plane, Pe = 100, KL/Kt = 10, L/b = 10.
140 I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144

Fig. 3. Dimensionless temperature distributionsacross half of the injection plane, Pe = 100, KL/Kt = 10, L/b = 100.

merical models are known to suffer at varying degrees Applying this relation to the solution, one obtains:
from errors such as numerical dispersion, and grid size

orientation effects and other behavioral errors such as 1 xD  t D


T̄D ¼ erfc pffiffiffiffi
nonphysical oscillations. In fact these errors may 2 2 tD


become quite difficult to evaluate for multidimension- xD þ tD
al models (Kocabas and Margoub, 2000). Therefore, þ expðxD Þerfc pffiffiffiffi  T̄Dsinv2 ð37Þ
2 tD
this is the first time a two-dimensional analytical
model is presented for confined reservoirs, and hence, where
it gives us the opportunity to match two dimensional l Z
error free temperature surfaces serving to better eval- 1 X tD
T̄Dsin v2 ¼
uate some of the errors involved in numerical models. zDb n¼0 0

nzDb ðn þ 1ÞzDb
erfc pffiffiffi  erfc pffiffiffi
s s
4. Average temperature profile in the oil layer pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p sðtD  sÞ
Obtaining an expression of average temperature Fsinv ds: ð38Þ
profile is important for determining the heating effi-
ciency, which is the most important parameter of Similar to the derivation of the two dimensional
thermal recovery methods. Particularly for the present temperatures, the following approximate solutions are
model, the average temperature serves as an accurate derived for the average temperatures along the trans-
one-dimensional solution. In addition, we can infer port direction. For n = 0, the first approximation to the
the collective roles of parameters and geometric term T̄Dinv2 in the solution is:
dimensions on the heating efficiency. The average

zDb
Z erf pffiffiffi
temperature across the half width of the reservoir 0 1 tD
s
T̄Dsinv2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(which is also equal to the average temperature zDb 0 p sðtD  sÞ
8

9
distribution across the whole width) is found using > s  xD s þ xD
>
> erfc pffiffiffi  expðxD Þerfc pffiffiffi  >>
>
the following relation: 1 < 2 s 2 s =


ds
2>> t x t þx >
>
Z : erfc D pffiffiffiffiffiD þ expðxD Þerfc D pffiffiffiffiffiD >
> ;
ZDb
1 2 tD 2 tD
T̄D ¼ TD ðzD Þ dzD ð36Þ
zDb 0 ð39Þ
I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144 141

Fig. 6. Comparison of Lauwerier, Avdonin and 2D models for KL/


Fig. 4. The new solution for L/b = 20 and various KL/Kt values.
Kt = 1 and various L/b values.

For n = 1, the second order approximation to the solution is:



zDb 2zDb
Z tD erfc pffiffiffi  erfc pffiffiffi
1 0 1 s s
T̄Dsin v2 ¼ T̄Dsinv2 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
zDb 0 pðtD  sÞ
8 
 
9
> 1 s  xD s  xD s þ xD 1 s þ xD >
> p ffiffi

< 2 þ t  s erfc 2 s  t  s þ 2 expðxD Þerfc 2 s p ffiffi
ffi >
=
D D
 
 
ds ð40Þ
>
> 1 t  xD t D  xD tD þ xD 1 t D þ xD >
>
: þ D erfc pffiffiffiffi þ þ expðxD Þerfc pffi ;
2 tD  s 2 tD tD  s 2 2 tD

The average temperature equations are first used to rate. The lowest pairs of curves in both figures
investigate the accuracy of the first and second order correspond to the combinations of a large L/b with a
approximations of the full solution (n = 0 and n = 1, low KL/Kt. In the case of a large L/b, however, one
respectively). Figs. 4 and 5 show the temperature would expect that the Lauwerier and Avdonin
profiles for a range of parameters for the cases of assumptions to hold, and hence, those solutions to
n = 0 and n = 1 where curves of the first and second approach the full solution. In fact, Figs. 6 and 7 show
approximations coincide except the lowest pairs of the comparison of the second order approximation
curves in both figures. The concurrence of the first with the Lauwerier and Avdonin solutions. The lower
and second order approximations indicates that the pair of curves in Figs. 6 and 7 corresponding to L/
approximate solution for those cases is highly accu-

Fig. 7. Comparison of Lauwerier, Avdonin and 2D models for L/


Fig. 5. The new solution for KL/Kt = 1 and various L/b values. b = 1 and various KL/Kt values.
142 I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144

b = 30 and 20 respectively and KL/Kt = 1, show that


the differences between the Avdonin and the two
dimensional models are negligible. Thus, as Avdonin
solution approaches the full solution, its proximity to
the second order approximation as shown in Figs. 6
and 7, further builds the confidence on the accuracy of
the second order approximation. Therefore, consider-
ing that curves in Figs. 4, 5 and 6 correspond to a
wide range of practical values, we can be confident
that the second order approximation is sufficiently
accurate for computational purposes.
In Figs. 6 and 7, a comparison of the Luwerier and Fig. 9. Effect of Peclet number on average temperature distribution,
Avdonin models with the two dimensional model for PV = 0.9, K/Kt = 10 and L/b = 20.
indicates that for a small KL/Kt value, as the aspect
ratio (i.e. large L/b) increases the Lauwerier and sivity will increase the requirement of L/b ratio or PV
Avdonin solutions approximate the two dimensional injection.
solution fairly closely. On the other hand, as aspect Fig. 9 shows the effect of dispersivity on the
ratio gets smaller or KL/Kt increases even in the case average temperature distributions for an L/b ratio of
of fairly large L/b (i.e. 20) values, use of the two 10 and indicates that as the dispersivity increases the
dimensional solution becomes a necessity for accurate transition zone will get larger, reducing the effective-
temperature profile estimations. ness of the reservoir heating.
The average temperature distributions for a wide Finally, one can infer that as L/b ratio becomes
range of values of the aspect ratio, L/b and the Peclet greater or Peclet number gets smaller, the transition
number, Pe, are given in Figs. 8 and 9, respectively. zone along the flow direction increases. However, as
Fig. 8 shows the effect of L/b ratio on the average for their collective role on increasing the transition
temperature distribution for a large Peclet number. zone along the flow direction, L/b ratio dominates,
Qualitatively, heating efficiency is proportional to the smoothing the temperature gradient across the reser-
area under the curve of average temperature. Thus, voir width and spreading the transition zone over the
one can see that as L/b ratio increases, heating whole flow distance.
efficiency decreases rapidly. If one were to define A major advantage of the present solution is that
the effective heating of the reservoir as increasing the it provides a more accurate thermal efficiency esti-
reservoir temperature to 0.5 TD with 2PV injection, mate than the one-dimensional models that use
Fig. 8 shows that even for a small dispersivity case, L/ infinite transverse conductivity. However, neither
b should be at least 20. Naturally, increasing disper- the development of an analytical expression nor a
numerical integration of the average temperature of
the new model profile for the thermal efficiency
calculations was attempted here and should be dealt
with elsewhere.

5. Conclusions

A new two-dimensional analytical solution of the


convection – dispersion equation based on a constant
linear velocity and constant longitudinal and trans-
verse dispersion coefficients is presented. The thermal
Fig. 8. Effect of L/b ratio on average temperature distribution, for transients during nonisothermal fluid injection into a
PV = 0.9, Pe = 100, and K/Kt = 10. linear confined oil reservoir have been studied
I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144 143

through these solutions. The model presented for C Gamma function


studying the thermal transients has several significant q density
features such as ability to generate error free two- k parameter equal to (x2 + a)
dimensional temperature surfaces and employing the x Fourier sine transform variable
thermal dispersion concept. Furthermore, it serves to j gradient operator
observe the roles of both fluid mechanics through
dispersion parameters and boundary conditions via Subscripts
heat loss terms at the confining boundaries. Thus, it b boundary
has the potential to be used in all modern nonisother- D dimensionless
mal liquid injection processes for design and interpre- i inlet
tation purposes as well as identifying the numerical L longitudinal
dispersion errors involved in two dimensions for m confining layer
comprehensive numerical simulators. 0 initial
o oil
Nomenclature r reservoir
a a constant equal to 0.25 t transverse
A area open to flow s solid grains
c heat capacity sinv inverse Fourier sine
k thermal conductivity w water
K thermal dispersion
In + 1/2 Modified Bessel Function of order (n + 1/2)
L length of the reservoir Acknowledgements
m dummy variable (integer)
n dummy variable (integer) Funding of this research was provided by the Abu
Pe longitudinal Peclet number, Pe=xDxf Dhabi National Oil Company (ADNOC). Many
PV pore volume of water injected, PV=ut/L=tD/Pe fruitful discussions with Dr. Hassan Hejase of UAE
q injection rate into the reservoir University have significantly facilitated our work.
s Laplace transform variable Both contributions are gratefully acknowledged.
S saturation
t time variable of the transport equations
T temperature References
T̄D dimensionless average temperature
u interstitial flow velocity Avdonin, N.A., 1964a. Some formulas for calculating the temper-
x space variable along the flow direction ature field of a stratum subject to thermal injection. Neft Gaz 7
y dummy variable (3), 37 – 43 (in Russian).
Avdonin, N.A., 1964b. On the different methods of calculating the
z space variable perpendicular to the flow temperature fields of a stratum subject during thermal injection.
direction Neft Gaz 7 (8), 39 – 44 (in Russian).
a dispersivity Cendejas, F.A., Rodriguez, H.R., 1994. Heat transfer processes
/ porosity during low or high enthalpy fluid injection into naturally frac-
v transformed temperature variable tured reservoirs. Proc. Nineteenth Workshop on Geothermal En-
gineering. Stanford University, Stanford, CA.
v¨ Laplace and Fourier Sine transformed v Gringarten, A.C., Witherspoon, P.A., Ohnishi, Y., 1975. Theory of
v˙ Laplace transformed v heat extraction from hot dry rock. J. Geophys. Res. 80 (8),
g space convolution variable 1120 – 1124.
s time convolution variable Gupta, A.D., 1990. Accurate resolution of physical dispersion in the
pffiffiffiffiffi multidimensional numerical modeling of miscible and chemical
h dimensionless parameter equal to h2 h1
displacement. SPE Reserv. Eng., Nov., 581 – 588.
h1 dimensionless parameter equal to (qmcmKt)/ Kocabas, I., 2001. Modeling of heat/tracer transport in oil reservoirs.
(qr cr km) Externally Funded Project Report IV submitted to Exploration
h2 dimensionless parameter equal to km/Kt and Production Division ADNOC, UAE Univ., Al Ain, UAE.
144 I. Kocabas / Journal of Petroleum Science and Engineering 42 (2004) 133–144

Kocabas, I., Horne, R.N., 1990. A new method of forecasting the advection in incompressible flow. SIAM J. Numer. Anal. 33 (2),
thermal breakthrough time during reinjection in geothermal res- 627 – 665.
ervoirs. Proc. Sixteenth Workshop on Geothermal Engineering. Malofeev, G.E., 1963. Calculation of the temperature distribution in
Stanford University, Stanford, CA. a formation when pumping hot fluid into a well. Neft Gaz 9 (2),
Kocabas, I., Islam, M.R., 2000a. Concentration and temperature 31 – 45 (in Russian).
transients in heterogeneous porous media, Part I: Linear trans- Peacemean, D.W., Rachford, H.H., 1962. Numerical calculation of
port. J. Pet. Sci. Eng. 26, 211 – 220. multidimensional miscible displacement. Soc. Pet. Eng. J.,
Kocabas, I., Islam, M.R., 2000b. Concentration and temperature 327 – 338.
transients in heterogeneous porous media: Part I. Radial trans- Ramirez, J., Samaniego, F.V., Rivera, J.R, Rodriguez, F., 1993. Tracer
port. J. Pet. Sci. Eng. 26, 221 – 233. flow in naturally fractured reservoirs. SPE Paper #25900, Pro-
Kocabas, I., Margoub, A., 2000. Improved numerical schemes for ceedings, Rocky Mountain Regional/Low Permeability Reser-
transport equations in oil reservoirs. Externally Funded Project voirs Symposium held in Denver, CO, USA, April 12 – 14.
Report III submitted to Exploration and Production Division Rubinshtein, L.I., 1959. The total heat losses in injection of a hot
ADNOC, UAE Univ., Al Ain, UAE. liquid into a stratum. Neft Gaz 2 (9), 110 – 116 (in Russian).
Lantz, R.B., 1971. Quantitative evaluation of numerical diffusion Satman, A., 1988. Solutions of heat- and fluid flow problems in
(truncation error). Soc. Pet. Eng. J., 315 – 320. naturally fractured reservoirs: Part 1-Heat flow problems. SPE
Lauwerier, H.A., 1955. The transport of heat in an oil layer caused Prod. Eng., 463 – 466 (Nov.).
by the injection of hot fluid. Appl. Sci. Res. 5 (2 – 3), 145 – 150. Sauty, J.P, Gringarten, A., Landel, P.A., 1978. The effect of thermal
Leonard, B.P., 1981. A survey of finite differences with upwinding dispersion on injection of hot water in aquifers. Proc. The Sec-
for numerical modelling of the incompressible convective dif- ond Invitational Well Testing Symposium. Div. of Geothermal
fusion equation. In: Taylor, C., Morgan, K. (Eds.), Computa- Energy, U.S. Dept. of Energy, Berkeley, CA, pp. 122 – 131.
tional Techniques in Transient and Turbulent Flow, vol. 2. Sauty, J.P, Gringarten, A.C, Menjoz, A., Landel, P.A., 1982. Sen-
Pineridge, Swansea, pp. 1 – 35. sible energy storage in aquifers: 1. Theoretical study. Water
Leonard, B.P., MacVean, M.K., Lock, A.P., 1995. The flux integral Resour. Res. 18 (2), 245 – 252.
method for multidimensional convection and diffusion. Appl. Spillette, A.G., 1965. Heat transfer during hot fluid injection into an
Math. Model. 19, 333 – 342. oil reservoir. In: Dietz, D.N. (Ed.), Thermal Recovery Techni-
LeVeque, R.J., 1996. High-resolution conservative algorithms for ques. SPE Reprint Series, vol. 10, pp. 21 – 26.

You might also like