Professional Documents
Culture Documents
Matrix-Fracture Transfer Shape Factors For Dual-Porosity Simulators
Matrix-Fracture Transfer Shape Factors For Dual-Porosity Simulators
SCIENCE &
ENGINEERING
ELSEVIER Journal of Petroleum Science and Engineering 13 ( 1995) I69- 178
Abstract
Matrix-fracture transfer shape factors for dual-porosity simulation of naturally-fracturedreservoirs were derived by combining
analytical solutions of pressure diffusion for various flow geometries. The resulting equations describing the matrix-fracture
flow were cast in a form similar to those proposed by Barenblatt et al. ’ and Warren and Root ‘, but without making the pseudo-
steady state assumption. Shape factors representing one, two and three parallel sets of fractures were obtained. A generalized
matrix-fracture transfer fnncticn for a non-isotropic, rectangular matrix block was also derived. Fine-grid single-porosity and
one-block dual-porosity models were used to verify the results. The results of fine-grid and one-block dual-porosity models were
in good agreement. The shape factors derived in this paper are consistent with observations made in several publications on dual-
porosity simulation studies.
1. Introduction
4N(N+ 2)
CT= (2)
The concept of treating a naturally-fractured reser- L2
voir as a dual-porosity medium was introduced by Bar- where N is the number of sets of fractures ( 1, 2 or 3).
enblatt et al. ( 1960) and later by Warren and Root For cubic matrix blocks having a fracture spacing of L,
( 1963) more than three decades ago. In both papers u has the value of 12/L2, 32/L2, 60/L2 for one, two
the transfer per unit bulk volume between the matrix and three sets of fractures, respectively.
and the fracture was assumed to take place under The application of the shape factor in numerical sim-
pseudo-steady state conditions. The resulting transfer ulation was introduced by Kazemi et al. ( 1976). Using
rate per unit bulk volume has the form: a finite-difference formulation for the flow between the
matrix and the fracture, they showed that for a three-
(1) dimensional case:
1
this manner are consistent with the observations made - (2n+ 1)%7%t
by Thomas et al. ( 1983) and Ueda et al. ( 1989). The p$?y_~ 8
I I n=,(2n+ l)%? exp [ W,L2
derivation for the case with one set of fractures is pre-
sented in the following. Derivations for systems with (5)
two and three sets of fractures follow the same principle
and are presented in the Appendix. A detailed derivation is presented in the Appendix.
In a dual-porosity simulation model, all mass flow is
from the matrix to the fracture, or vice versa. Therefore,
2. Derivation the rate of matrix-fracture transfer can be related to the
rate of mass accumulation in the matrix as follows:
In Fig. 1, a matrix block surrounded by a set of
parallel fractures is considered. The fractures have infi- 4= _p&&
at
nite lateral extent such that the matrix block resembles
a slab of thickness L, which is also the fracture spacing. Taking the partial derivative of Eq. 5 with respect to
The flow from the matrix to the fracture is one-dimen- time, t, and substituting it in Eq. 6 will result in an
sional, and perpendicular to the fracture plane. Assum- equation for the rate of matrix-fracture transfer as a
ing that the flow in the matrix obeys Darcy’s law, it can function of matrix and fracture pressures. The form of
K.T. Litn. K. Aziz /Journal of Petroleum Science and Engineering 13 (1995) 169-178 171
(8)
r? 9.87
a=,=-L (9)
L2
-
follows: matrix blocks is shown to be:
(_ 1
dkt
Prn_~-()~1 exp -
(7) (10)
Pf-Pi h-G2
This approximation is obtained by taking only the first By defining an equivalent isotropic permeability (Mus-
term in the infinite summation series in Eq. 5. Fig. 2 kat, 1937) :
shows the plots of Eq. 5 and Eq. 7. The vertical axis is
k= (kJ&) “’ (11)
a dimensionless pressure and the horizontal axis a
dimensionless time. The approximate solution is in an equivalent shape factor for an anisotropic matrix
excellent agreement with the analytical solution except block can be defined as:
for dimensionless time of less than 0.1. For typical
reservoir values, this translates to a time of less than 1
(12)
second. The departure from the origin is not important
as it is eliminated from the final form of the equation
For an isotropic (k, = kv= k,) , rectangular matrix
(see the Appendix for details). The important param-
eter is the coefficient oft in the exponent, which reflects block, the shape factor reduces to:
the rate of transient pressure change. Differentiating
Eq. 7 with respect to t and substituting into Eq. 6 ena- (13)
bles the elimination of the time parameter (see Appen-
172 K. T. Lim, K. Aziz /Journal of Petroleum Science and Engineering I3 (I 995) 169-I 78
3. Verification
I- x \L____L-__J
independently by the authors in 1993 and early 1994. It was brought
’
\
I I
to the author’s attention, after Eq. 13 was derved, that Chang ( 1993)
and Kazemi and Gilman ( 1993) have derived the same expression.
Z
Also, results consistent with Eq. 13 were obtained by Barkve ( 1992)
and Zimmerman et al. ( 1993). Although the assumptions and
approach as used in arriving at the shape factors are slightly different Fig. 3. Schematic of single-porosity models. Shaded areas represent
in this and other studies (Lim, 1995), it is reassuring that all the fracture planes. Fewer than actual number of gridblocks am shown
results are consistent. for clarity.
K.T. Lim, K. Aziz / Journal of Petroleum Science and Engineering 13 (1995) 169-178 173
Table 2
Data for single-porosity and dual-porosity models
Fine grid, N = 1
Grid dimensions 11X1X1
Grid spacing Ax=2.445, 1.28, 0.64, 0.32, 0.16,0.08, 0.04, 0.02, 0.01, 2x0.005 ft
Ay=Az= loft
Fine grid, N = 2
Grid dimensions 11x1x1
Grid spacing Ax = same as Ax for N = 1
Ay=2.445,6.17, 8.09, 9.05, 9.53.9.77, 9.89, 9.95, 9.98, 2x9.995 ft
AZ= 10 ft
Fine grid, N = 3
Grid dimensions 6X6X6
Grid spacing A~=A~=Az=0.01,0.02,0.08, 0.32, 1.28, 3.3Oft
Dual porosity
Grid dimensions 1X1X1
Grid spacing Ax=Ay=Az=lOft
Common data for all models
Matrix porosity 0.05
Matrix permeability 0.001 mD
Initial pressure 1000 psia
Initial temperature 600°F
Fracture pressure 500 psia for t > 0
Fluid properties are correlated from steam table
60
this paper are for single-phase superheated steam flow.
The single-phase compressed water flow cases were
also analyzed, with essentially the same observations.
In general, the results of dual-porosity models with
shape factors derived in this paper are in good agree-
60
60 ,
Eq. 8 is capable of describing the transient behavior
I
of the matrix fracture flow because the matrix pressure,
Pm, and the fracture pressure, pr are both time depend-
ent. Proper selection of time step size during numerical
simulation should capture the transient behavior.
The shape factors derived in this paper are consistent
with the observations made by authors cited in the
introduction. The factor of 25/L* used by Thomas et
al. ( 1983) for the three-dimensional model is in close
agreement with the factors of 25.67/L2 and 3$lL2
derived in this paper. Ueda et al. ( 1989) stated that for
one-dimensional flow the Kazemi et al. shape factor
3D Fine Grid, 216 blocks
1 block DP, Warren-Root, a=60/L*
(4/L2) needs to be multiplied by a factor of at least 2.
1 block DP, This work, 0=3&L’ This observation is consistent with the factor of d/L2
1 block DP, This work. a=2567iL’
1 block DP. Coats, a=24/L2 for one set of fractures. For a system with two set of
1 block DP, Kazemi, 0=12/L* ~
fractures, Ueda et al. argued hueristically that the
4 6 Kazemi et al. shape factor ( 8/L2) needs to be modified
Time (day)
by a factor of 3. The analysis in this study showed that
it should be a factor of about 2.5.
Fig. 6. Results for system with three sets of fractures.
1
- (2n+ l)*dkt
Appendix 1 4/&*
P(P) =p”u +c(P-P”)l (AZ) where pi is the average pressure in the matrix at time
176 K.T. Lbn, K. Aziz /Journal of Petroleum Science and Engineering I3 (1995) 169-l 78
-dkt
‘$=l-0.81exp
1 i
-
&-%L2 1 (AlO)
-
respect to t yields:
1
1 a& r?k dkt
(All)
pf-pi ~=~‘81 exp [ ~~C,L2
(Al21
0.2 )
using Eq. AlO. Simplification of the equation yields:
a% ,+,“:t-
-=. - Pm-,,f) (Al3)
0.0 Ii]
at CI
2
0.0 0.2 0.4 0.6 0.6 1.0
WwRZ
Substituting Eq. Al3 into Eq. 6 results in the following
matrix-fracture transfer function: Fig. 7. Analytical solution of diffusion in a cylinder and its approx-
imation.
(A14) Using arguments similar to those for the case with one
set of fractures, the analytical solution can be expressed
as follows (Crank, 1975):
1.1.2. Two sets offractures
(Al9)
In this case the matrix block is surrounded by two
sets of perpendicular fractures, with both sets of frac- where the rr,‘s are the roots of:
tures having a spacing L. Pressure diffusion from the
J,(Ra,) =0 (A20)
matrix to the fracture can be closely approximated by
that of a cylindrical matrix block. The equivalent where J,,(x) is the Bessel function of the first kind of
radius, R, of the cylinder is the radius that would yield order zero. The analytical solution is approximated by
-
the same volume as a bar with a L XL square-shaped taking only the first summation term in Eq. A19:
1 [ 1
cross-section. The volume is used as a basis of equating
5.78kt
the two geometries due to the considerations given ‘s=l-0.69exp - (‘421)
when deriving Eq. A8. In this case: dwR=
The analytical solution for cylindrical diffusion and its
R = 0.564L (A15)
approximations are plotted in Fig. 7. Differentiating
The pressure diffusion equation (Eq. Al) for radial Eq. A21 with respect to t, and substituting in Eq. 6
geometry is: yields:
(‘416)
,2$~(,, -p,) +z+-m_pf) (A=)
R = 0.62OL (~23)
(~24)
0.2
The initial and boundary conditions are:
r
o Analytical Solution
p=pi, O<r<R t=O (A25)
- Fvst Term only
0.0 0 ___
p=pr, r=R, t>O (A261 0 1 2 3 4 5
k&/@pqR’
The analytical solution can be expressed as follows
(Crank, 1975; Zimmerman et al., 1992) : Fig. 8. Analytical solution of diffusion in a sphere and its approxi-
mation.
~427)
Pi -Pi _
The analytical solution is again approximated by taking Pf_Pi
only the first summation term in Eq. A27:
_
1-+2_y l
a=op=oy=o (2a+ 1)2(2p+ 1)2(2y+ 1)2
‘s=l-0.61exp (A281
I
Consider a rectangular
matrix block of dimensions
Lx, Lx and L, with corresponding directional permea- (A31)
bilities k,, ky and k,. If the system is initially at pressure Differentiation of Eq. A3 1 with respect to t yields:
pi and is exposed to a constant fracture pressure pf, the
analytical solution can be derived by taking the New- (~32)
man product of the dimensionless solution for diffusion
in a plane (Eq. A7) as follows (Holman, 1990; Abba- Substituting Eq. A32 into Eq. 6 yields the final form of
zadeh, pers. commun., 1994) : the matrix-fracture transfer function:
178 K.T. Lim, K. Aziz / Journal of Petroleum Science and Engineering 13 (1995) 169-I 78