Professional Documents
Culture Documents
Cardwell Parsons k3d
Cardwell Parsons k3d
This paper was prepared for presentation at the 68th Annual Technical Conference and Exhibition of the Society of Petroleum Engineers held in Houston, Texas, 3-6 October 1993.
This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract submitted by the author(s). Contents of the paper,
as presented, have not been reviewed by the Society of Petroleum Engineers and are sUbject to correction by the author{s). The material, as presented, does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society
of Petroleum Engineers. Permission to copy is restricted to an abstract of not more than 300 words. Illustrations may not be copied. The abstract should contain conspicuous acknowledgment
of where and by whom the paper is presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3838, U.S.A. Telex, 163245 SPEUT.
Abstract
When performing fluid-flow simulations, up-scaling can be found in the literature 1,2,4,7,8,10,11 and among
techniques must be used because of the required coarse them in the case of isotropic scalar permeabilites, those
grids. For this purpose, algebraic formula, combining based on the bounds of effective permeability 2. But so
arithmetic and harmonic averagings, extension of far, to our knowledge, the influence of anisotropy was not
Cardwell and Parsons's results, derived from considered in this context. This anisotropy comes from
perturbation techniques are proposed taking into either sedimentation process or from reservoir
account the anistotropy, either natural or due to the geometry. Generally the vertical extent of reservoirs is at
reservoir geometry. These formula are valid for 2D or 3D least one order of magnitude less than the horizontal
problems and are very little computational time one.
consuming compared to direct numerical methods. We have derived from a perturbation technique an
Moreover they integrate rigorous mimima and maxima extension of the estimator proposed by Le Loc'h and
, thus showing the global uncertainty on equivalent Guerillot 5,11 :
permeability determination. They proved to work well
for large permeability contrasts and anisotropy ratios
either for simple log-normal or geostatiscally generated
media. Keq = -J K+ K- (1)
In the first application case numerical calculations were
carried out for the upscaling of isolated cells. In the Le Loc'h and Guerillot's estimator is only suitable for
other case well test simulations were performed either 2D and isotropic conditions. For this particular estimator
with the original fine grid or further coarser grids. K-, K+ are the abolute permeability minima and
maxima defined by Cardwell and Parsons 2 and
Introduction explicited later in the text.
In a 2D case our derivation takes into account
Progresses in the power of computers and in reservoir anisotropy with the following formula:
characterization allow the simulation of very high
resolution images of complex reservoirs. But for the
subsequent flow simulations, coarser grids are inevitable
Keq = (K+)9y(K-) 1-9y (2)
in most cases and one has to upscale the fluid flow
parameters (for example permeability or porosity fields). with the ponderation coefficient 9y
The objective of the present work is to present a new
algebraic formula for upscaling absolute permeability arctan -fAY .......... (3)
diagonal tensors. As opposed to more sophisticated ey 1t/2
numerical procedures, algebraic formulas are not
intented to provide exact results in the general case, but where AYis the global anisotropy coefficient:
just acceptable estimations for most of the practical
applications. Their purpose is a good compromise
kyy L'ix 2
between cost of calcul and precision. Several of them Ay=-(-) (4)
kxx L'iy
References and illustrations at end of paper where ky (resp. kx) is the Y(resp.X) direction
29
INFLUENCE OF lHE PERMEABILITY ANISOTROPY RAno ON LARGE-SCALE
2 PROPERTIES OF HETEROGENEOUS RESERVOIRS SPE 26612
with:
Considering these assumptions it is clear that the
(1-9z)ey equivalent permeability ( for X-direction) depends not
coy (6) only on the local permeabilities kxx(x,y,z) but also on
1-9y&z
AkyandAkz.
(1-ey)9z
roz (7) Examples are displayed on figures I and 2 for 2D and
1-9y9z 3D conditions. Keq ANISO and Keq ISO are the
calculated equivalent permeability respectively for the
and: anisotropic and isotropic cases. They correspond to the
arctan ....["A;; solution of the finite difference sytem, obtained for 10xl0
9z (8) <figure I ) and 6x6x6 grids (figure 2), with a log-normal
1t/2
distributed permeability for different values of cr2 ( In k
where: variance ). The anisotropy effects may lead to 100%
under or overestimation of the corresponding equivalent
permeability.
kzz ~ 2
Az=kxx (~) (9) Another way to show the importance of anisotropy is to
come back to the equation corresponding to an
incompressible monophasic flow and to consider the
with 9yand Ay defined as before. K3 and K4 like K+3D grid block size influence over the flow (figure 3).For sake
and K-3D, result from proper combinations of of simplification we consider the case where the grid size
arithmetic and harmonic means, as will be later in the Y direction is two times bigger than that of the X-
explicited. direction The general equation is:
If anisotropy is uniform, exponents have to be
computed only once, and this approach means V.(K V(P» = 0 (11)
negligible additive computational cost.
The above expressions stand for the X-direction. In case of a 2D problem we have:
Similar relationships for the other directions are easily
obtained with a simple rotation of the indexes. ap ap
a (Kx ~) a (Kyay)
Theory ax + dy 0 (12)
The fundamental aspects of this work derives from an If we change the variables:
analytical development of the equivalent permeability
valid at the second order, in the domain of short x'= x
variances. andy' =2y
The case when the local permeabilities are diagonal
terms will be solely considered. The anisotropy global this equation yields:
ratios are supposed to be constant in the whole space.
Then the permeabilities to be homogeneized of the ap ~ap
grids can be written: a (Kx ax-) a(4 at)
ax' + dy' =0 (13)
where:
In the case of anisotropic permeability this equation
30
SPE26612 I-PATRICK DUQUERROIX P. LEMOUZY B. N<ETINGER R.K. ROMEU
We remark that tJ.x,tJ.y and tJ.z which apparently do not For a 20 grid (supposed geometrically isotropic) with k
recall any physical meaning are indeed generally taken log-normal and no correlated, without any grid refining
proportional to the overall reservoir geometrical and with the direct calculation for transmissivities,
proportions or to the correlation lengths of the reservoir Ncetinger and Romeu derived 9 :
in the three directions. In this sense, the formula
proposed by Ababou et al. 1, where the equivalent _ 02 1 _
Keq - Kg e Cj 52)· (18)
permeability is evaluated as a function of the ratios
among lx,ly,lz can be seen as an estimator for media
having anisotropy in permeability. where:
Cardwell and Parsons have defined the two limits K+ This equation is only valid up to the second order series
and K-. expansion in the parameter 0 2 ,Kg are the geometric
K+ and K- are combinations of arithmetic and mean and 0 2 .the variance of In k xx.
harmonic means in X and Y directions.
With the help of an electrical analogy (where
52 converges very quickly for increasing values of N.
31
INFLUENCE OF THE PERMEABILITY ANISOlROPY RATIO ON LARGE-SCALE
4 PROPERTIES OF HETEROGENEOUS RESERVOIRS SPE26612
K+3D = ax(ay(az»
K3 = ayChx(az»
1-12
arctan ..fAY (21)
A small variance series-expansion of the equivalent
1C/2 permeability by Nretinger and Romeu corresponds to:
From: 1
Keq=Kgeo2(2-53) (28)
with
and by analogy with the log-normal case where
K- = Kh = Kge-a2 (22)
lt2
8 . 2
13= lim S3 =- sm a dadlXiy
and thus we propose the following ponderation: n__ 1C 2[
0
[
0 1.
[
sin2 a+Ay sin 2 J3+Az sinZy
By
arctan ..fAY By means of geometrical analogy with 2D results, we
have found the following algebraic formula:
1C/2
Ay = 1. (l-By)9z
CIlZ
l-ayaz
3D Anisotropy and:
roy and CIlZ are function of Ay and Az. Applications and results.
32
SPE26612 J-PATRICK DUQUERROIX P. LEMOUZY B. NCETINGER R.K. ROMEU
corresponding to the validation of the upscaling formula locally refined grid around the central well (grid q.
that we propose. And finally a reservoir interpretation
with realistic data. The different grids are represented in figure 8. For the
fine grid, all the cells have the same dimensions:
Individual heterogeneous blocks.
Ax=Ay=20m, Az = 1m,
2D conditions
For the coarse grid we have:
A reference case obtained with finite differences, with
periodic boundary conditions and a direct calculation of (Ax (i), i =1 to 5) = 60,100,100,100,60 m
transmissivities is used as reference. The kxx values are (Ay (i), i =1 to 5) = 60,100,100,100,60 m
randomly generated with a no correlated lognormal law.
Figure 5 shows the excellent agreement for a small and:
variance between the reference case and the anisotropic
response. In fact , we are in the conditions when this
Az = 6 m for all blocks.
estimator can be fruitfully used. And the results are thus
very precise.
Even for significant values of variance, again the 3D reservoir simulations
anisotropic estimator yields satisfactory results. In
figure 6 is plotted the ratio of the estimation of the Homogeneous and heterogeneous runs were
equivalent permeability over the reference permeability performed. For heterogeneous conditions the different
versus the global anisotropy factor. Results from five upscaling formulas were tested.
different realizations of the medium are shown.
For the extreme values of the global anisotropy factor, Homogeneous validation
the error is nul. When Ay = 1, the anisotropic estimator
degenerates into the isotropic estimator and thus the Firstly two runs were performed in homogeneous
errors are of the same order. For intermediate values conditions to check the influence of grid size. The
there is a systematic error: underestimations (resp. permeability chosen was 100mD. On figure 9 we have
overestimations) if Ay < 1 (resp. > 1). plotted the evolution of bottom hole pressure versus
time. There is only an influence of grid refining near the
3D conditions producing well due to the numerical productivity index.
For longer times the two responses are eqUivalent.
In figure 7, we make a comparison between isotropic But if we consider the well test interpretation the two
and anisotropic responses. The grid is 6 x 6 x6 with again equivalent permeabilities are significantly
2 different(figure 10). On this figure OP or oP' (pressure
an uncorrelated lognormal law and 0" = 4. The results
are very demonstrative. Ratios between the estimation derivative responses) are plotted versus time. For the
of equivalent permeability and the calculation are fine grid we find again 100mD while in the other case we
varying between 0.3 and 10. when the estimator is find out only 80mD. This is due to the use of a too much
isotropic. coarse grid, which enhanced boubary effects. We had
With the anisotropic estimator the results are fairly not enough information in this peculiar case to perform
well in the global range of Ay and Az. simulations ·with a greater number of grids, so we have
developed pseudo fine grid simulations.
A more complete reservoir simulation.
Heterogeneous runs
Presentation of the reservoir simulations
For this series of heterogeneous runs we have firstly
Firstly a geostatistical simulation generated from real compared bottom hole pressure responses versus time
(figure 11) between refined grid (A) and coarse grid (B)
geological data 3 has been performed on a fine grid ( 21 using for each coarse grid three upscaling formulas:
x 21 x 30 cells ), then we have performed a well-test
simulation with 3 different grids:
,. the square root upscaling ..J K+ K- ,
- with a coarse cartesian grid (5 x 5 x 5 cells) (grid B), 6~----
- with a fine grid, taking as effective permeabilities for ,. the sixth root upscaling:;,j K+2K3K4K-~
the small grids, those of corresponding coarser grids
(grid A), This isotropic 3D formula has been proposed by
- with a fine grid, taking as effective permeabilities for Lemouzy .
the small grids, those of correponding coarser grids,
except for the grid containing the central well. This is a
33
INFLUENCE OF THE PERMEABILITY ANISOlROPY RATIO ON LARGE-SCALE
6 PROPERTIES OF HETEROGENEOUS RESERVOIRS SPE26612
34
SPE26612 I-PATRICK DUQUERROIX P. LEMOUZY B. NCETINGER R.K. ROMEU
media, to be published.
CT
Q)
:::.:::
10---- W""" ,., : ... .J
" -6- u' = 0.5
,_A ~"
9 Ncetinger B. ,Romeu R.," About the influence of 0.4 -' -e- u' = 1
-
the transmissivity evaluation rule upon the results of -h- u' = 2
numerical simulations of fluid-flows in heterogeneous -~- u' = 4
porous media" submitted to Water Resources
Research 0.2 I I
10 Warren J.E. and H.S. Price, "Flow in heterogeneous 10'3 10· 10' 10' 10 3
media", Soc. Petrol. Eng. J., 1, 153-169, Sept. 1961.
11 Guerillot et aI., "An integrated model for computed A y
aided reservoir description, from outcrop study to fluid Fig.l Anisotropy influence on Keq (20 conditions)
(I Ox 10 k log·nonnal no correlated)
flow simulations", 5th European Symposium for
Improver Oil Recovery, Budapest, 1989.
35
INFLUENCE OF THE PERMEABILITY ANISOlROPY RATIO ON LARGE-SCALE
PROPERTIES OF HETEROGENEOUS RESERVOIRS SPE26612
Influence of gridding anisotropy over fluid-flow
K K
liY 1,2 ,
22
'lliY
K K 2 ,1
liY ,
11
'lli Y
• Equation: • Equation:
V.( K V(P) )=0 V.( K V(P) ) = 0
A, Ay
10" 10-' 10'
1.025 10' la'
10.0
" 10.0 u
-J
-J
CT C5
Q)
Y:
1.0 J ~
cr
~ 1.0
~
anisotropic estimator
I Isotropic estimator I I
0.95+=--+-i I I I ~
10- 3 10-' 10-' 10° 10' 10' 10 3 anisotropic estimator
A
Fig. 5 Comparative resulJ'between isotropic and anisotropic A. y 'v' la'
N=6
= .4
estimators ( 0 2 = 0.1, 6Ox60. k log-normal no correlated)
4.0 Ay
~
Iit_.
-..:;;:.::: .... .. , ,
................ 10" 10-' 10' la'
u
j ......:::~ ,, , ,
2.0
... '
" 10.0
10.0
(J
- J-
CT -J
Q) ''''
,~'!!- , «
r!f
Y: ,~"
,
, anisotropic
~ ~
" 1.0 ...;,......
..... estimator
f-
~
CT
0.7
t----
-.... pt~,
.~,. .... .... ""
~
.... '"Jlf-
, -- ~
~~
~
Q) ........
y: 0.4~- 0'2 = 4 ,
N = 10
.. isotropic
estimator
~
I ,isotropic estimator
0.2 .
10- 3 10-' 10- 1 10° 10' 10' 10 3 a' = 4
A. y N=6
A y
Fig. 6 Comparative results between isotropic and.anisotropic
estimalors (,,2 =4, 10.10, k log-nonnal no correlated)
Fig. 7 Comparative results between isotropic and anisotropic
estimators ( 0'2 =4, 6x6x6, k log-normal no correlated)
100
99
A
98
97
I 96
f*
~
I
i
95 • ANEGRID
o COARSE GRID
i 94
I;JI;JI;JI;JI;JI;JI;JI;JI;JI;JOOOO 0 0 0 0 0
93
B
92
91
• 90 I I
o.oOE+OO 2.00E-ol
I
4.00E-Ol 6.00E-ol
TIme (clays)
I I
8.00E-ol
I
1.00E+OO
Fine grid
98
. ._grid
I
L. .
r ··I······~
.1CJI8f81'OO1~
L 1001 i . / [' I .. ,
1
'
'0000000000000 0 0 0
... IbdhI'OOlI4)lCllling
~
o .... 1IIPft*h up8CIIng
1
10- I
2
,.
.'
I
I
I
I
I i
I
I 8B I I
.,2
I
0," 0,6
I I
0,8
I
4 3 2 1 B 1 "-(_I
U,-5 18- 18- It, dew UJ- 111I- UI 10
~
I I
I . 1i
98
..' 6P 1
i/ I .
~ lla"
1·-'" .,
"
~
..
OP' I
•
I ..
1
.
••
- a•••••••
.
** .. I:~==I
:l;
10
-1
i .2
:1 , , , ...,
••2 0,4 0.6
111M I...,.)
Fig.lO Pressure derivative responses for homogeneous Fig. 12 Influence of upscaling on bottom well pressure
conditions. (after redistribution)
SPE26612 I-PATRICK DUQUERROIX P. LEMOUZY B. NCETINGER R.K. ROMEU
100
98
96
~
e :!? • refined grid
~
~
2i. 94
•••• ...... ••••• - square root upscaling
before
•
E
••• • square root upscaling after
~
92
88 +----t-----i-----t---+---i
98
96
j
~ - new approach upscaling
tl2i. 94 after modified
1 - refined grid
E
2
S
92
90
88 +---+---+---+---+-----1
o 0.2 0.4 0.6 0.8
TIme I...,.'
Fig. 14 Influence of upscaling on bottom well pressure
(after new redistribution)
40