You are on page 1of 10

Soc:lety of Petroleum Engineers

SPE 26473

Stochastic Relative Permeabilities Usually Have Neglectable Effect


on Reservoir Performance
C.B. Tj~lsen, Norsk Hydro Research Center; Eivind Damsleth, Norsk Hydro a.s.; and Tor Bu,
Norsk Hydro Research Center
SPE Members

Copyright 1993, Society of Petroleum Engineers, Inc,

This paper was prepared for presentalion al the 68th Annual Technical Conference and Exhibition of the Society of Petroleum Engineers held in Houslon, Texas, 3-6 Oclober 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 I

as presented, have not been reviewed by the Society of Petroleum Engineers and are sUbject to correction by the author($). 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 75083-3836, U,S,A. Telex, 163245 SPEUT.

Abstract
Introduction
During the last couple of years Norsk Hydro has developed a During the recent years there has been an increased interest
3D model for simultaneous generation of stochastic absolute in stochastic modeling of heterogeneous reservoirs, and a
and relative permeabilities. number of new techniques have emerged; see Haldorsen and
Damsleth 1 for an overview.
By using core data containing relative permeability curves
measured on a large number of core plugs from one single To date, most applications of stochastic modeling have been
well in the North Sea, we have been able to model rela- limited to saturation independent properties, such as poros-
tive permeability curves (represented by endpoints and ex- ity and permeability 2-7. However, we also know that the
ponents) stochastically for four different depositional envi- relative permeabilities in a reservoir model strongly influ-
ronments ranging from highly permeable mouthbar sands to ence the production characteristics. Hence, the objective of
low permeable tidal deposits. our work is to investigate the effect of letting the relative
permeabilities vary stochastically throughout the reservoir,
We show that for all the depositional environments, stoc-
In a recent paper8 we addressed the problem of stochastic
hastic variation of the relative permeabilities have only
modeling of relative permeabilities. The paper describes a
marginal, if any, effect on the production characteristic, com-
model where both saturation independent (k H , k v , ¢J), and
pared to keeping the relative permeabilities constant at their
saturation dependent (Swi, Sor, krw(Sor) and the Corey-
mean. exponents n w and no) properties were modeled simultane-
Based on fractional flow theory, this paper summarizes the ously, with a realistic correlation structure. Furthermore,
results from a theoretical and empirical statistical analysis of the importance of taking the heterogeneity of the saturation
dependent variables into account was studied. This was done
the correlation between the water shock front velocity and
by comparing the flow behaviour of a model, where the rel-
the absolute permeability for the different depositional envi-
ative permeabilities were allowed to vary around their mean
ronments, and we show that in specific cases this correlation
value, with a simpler model where the relative permeabilities
can serve as an indication of the potential effects of stochas-
were kept constant at their mean.
tically varying relative permeability.curves.
The main conclusion from that paper was that there was
The main conclusion, which must be very' comforting to prac- hardly any difference between the models with stochastically
ticing engineers, is that in some situations stochastic model- varying relative permeabilities and the main output from the
ing of the relative permeability curves is of minor importance, models where the relative permeability curves were kept con-
However, the choice of mean relative permeabilities may be stant at their mean throughout. However, we were cautious
crucial. not to generalize the conclusion, since it was based on only

591
STOCHASTIC RELATIVE PERMEABILITIES USUALLY HAVE
2 NEGLECTABLE EFFECT ON RESERVOIR PERFORMANCE SPE26473

one set of parameters, valid for one particular depositional TABLE 1 Univariate description for each deposi-
tional environment.
environment.

Hence, the aim of the present study is to: Variable Dep. N Mean St. Min. Max.
env. dev.
• Utilize the model described in Tj¢lsen and Damsleth8
to study the effect of stochastically varying relative per- <I> [%] MIAd 41 21.4 2.2 15.1 24.8
meabilities for each of several different depositional en- MIAp 20 20.8 2.5 13.9 23.7
PI 9 19.8 1.4 17.5 21.5
vironments.
T 15 24.0 1.5 21.6 26.1
• Apply fractional flow theory to indicate when the use MIAd 40 5.26 0.99 2.51 7.94
of stochastically varying relative permeabilities mayor MIAp 20 7.01 1.81 2.31 9.10
may not have an additional effect on the flow perfor- PI 9 5.44 1.20 3.72 7.09
mance. T 14 4.83 0.38 4.14 5.63
Swi [frac.] MIAd 41 0.221 0.074 0.070 0.488
In this paper we mainly limit ourselves to the analysis of the
MIAp 20 0.144 0.098 0.026 0.375
break-through time as a measure of flow performance, as this
PI 9 0.230 0.101 0.075 0.363
parameter is of major importance for the project economy. In
T 15 0.180 0.040 0.129 0.271
addition the break-through time seems to be quite sensitive
SOT [frac.] MIAd 41 0.266 0.049 0.200 0.392
to variation in the reservoir properties.
MIAp 20 0.309 0.058 0.168 0.391
The resuits of our study has been reported more completely PI 9 0.284 0.048 0.205 0.351
elsewhere 9 , including many technical details which have been T 15 0.311 0.028 0.257 0.359
omitted here for brevity. loge(Corey expo MIAd 40 1.11 0.19 0.81 1.67
oil), ln o MIAp 20 1.41 0.19 1.14 1.82
PI 9 1.29 0.16 1.07 1.54
Stochastic Modeling and Simulation T 15 1.11 0.14 0.88 1.50
Experimental procedure loge(Corey expo MIAd 40 0.89 0.37 -0.37 1.91
water), ln w MIAp 20 0.26 0.37 -0.84 0.82
Based on a unique set of core data, containing relative per- PI 9 0.49 0.31 0.13 1.04
meability curves measured on 85 core plugs from one single T 15 0.77 0.22 0.18 1.02
well in the North Sea, correlations between relative permea- MIAd 41 0.39 0.15 0.06 0.69
bility curves (represented by endpoints and exponents), ab- MIAp 20 0.60 0.13 0.28 0.77
solute permeability and depositional environment have been PI 9 0.44 0.10 0.31 0.59
developed 1o . Four different depositional environments were T 15 0.37 0.06 0.28 0.49
sampled from the well:

MIAd : distal mouthbar quarter of the height in the vertical direction. Five realiza-
MIAp : proximal mouthbar tions were generated for each depositional environment.
PI : distributory channels To avoid the problem of homogenization (upscaling), we
T : tidal bay wanted to maintain the same grid for both the geological
realizations and the reservoir flow simulation model. We
Each depositional environment has its own distinct proper- therefore decided on a regular 36 x 36 x 48 (X x Y x Z) grid
with a total of 62 208 cells, with cell dimension of 20 x 20 x 2
ties in terms of means, standard deviations and correlation
meters.
matrices, as shown in Tables 1 and 2.
In this study we assume that the grid-blocks represent REVs
The described model has previously been applied to the dis- (representative elementary volumes). This is a simplification
tal mouthbar deposit 8 . In the present study we have applied which is probably not justified. However, as long as we
the stochastic model to generate four synthetic reservoirs, restrict ourselves to study the effects of stochastic relative
each representing one of the four different depositional envi- permeability, this assumption will not be important.
ronments. Each reservoir has simultaneously varying poros-
We applied a black-oil IMPES (implicit pressure and explicit
ity, permeability and relative permeability. The variables are saturation) formulated reservoir simulator l l to simulate wa-
modeled as intercorrelated (log-)Gaussian fields with a pre- terflooding in the various reservoir realizations. The geom-
scribed spatial correlation structure. The spatial correlation etry of the model is a quarter of a five spot, i.e. the injec-
was modeled using spherical variograms with range equal to tion well in one corner column and the producing well in the
the side of the reservoir in the horizontal direction and a column of the diagonally opposite corner. Only immiscible

592
SPE26473 C.B. TJ0LSEN, E. DAMSLETH AND T. BU 3

TABLE 2 Pearson correlation coefficients for the TABLE 3 Standard deviations for v and correlations
different environments. between v and logekH for each depositional environ-
ment.
MIAd
loge kH Swi Sor loge no logenw krw(Sor) Dep.env. St.dev.{v} corr(v,logekH}
</> 0.44 -0.42 -0.02 -0.30 0.14 0.17 M1Ad 0.42 0.13
logeklI -0.77 0.19 0.52 -0.59 0.81 M1Ap 0.97 -0.49
Swi -0.53 -0.37 0.58 -0.63 PI 0.53 0.02
Sor 0.21 -0.35 0.32
T 0.39 -0.45
logeno -0.73 0.67
logenw -0.57
oil rate and cumulative oil production versus time for the dis-
tal mouthbar deposit (M1Ad) in the case where the relative
MIAp permeabilities are constant at their mean, as well as when
loge kH 8 wi loge no logenw krw (8or ) the relative permeabilities vary stochastically, for five differ-
Sor
ent realizations. The absolute permeability and the poros-
</> 0.78 -0.64 -0.19 -0.58 0.41 0.06
ity vary stochastically in all the realizations. To compare
loge kH -0.89 0.01 -0.13 -0.01 0.50
the production characteristics of the cases with constant and
Swi -0.28 0.17 -0.07 -0.53
varying relative permeability, the absolute permeability and
Sor 0.31 -0.32 0.12
porosity fields are the same in both cases. Thus, any differ-
loge no -0.83 0.36 ence between the pair of profiles in Figure 1 is due to the
loge nw -0.25 varying relative permeabilities. Figures 2, 3 and 4 show the
same features for the proximal mouthbar deposits (M1ap),
PI the channel sands (PI) and the tidal deposits (T) respec-
tively. Neither of the figures 1--4 show any significant effect
loge kH Swi Sor loge no logenw krw(Sor)
of stochastic variation of the relative permeabilities around
</> 0.87 -0.63 0.55 0.51 -0.66 0.79 their mean.
loge kH -0.90 0.80 0.44 -0.60 0.85
Swi -0.91 -0.18 0.33 -0.77 For all these realizations, some key numbers were calculated.
Sor 0.29 -0.41 0.64 The mean time to reach 2 % and 90 % water cut, and the
loge no -0.95 0.19 cumulative oil production after 10 and 15 years. Based on
loge nw -0.39 these numbers, i-tests were performed. The i-test gives
no reason to reject the null hypothesis (based on a 5 %
significance level) that there is no difference between the
T realizations with constant and varying relative permeabilities
logekH Swi Sor loge no loge n w k rw (Sor ) for any of the four depositional environments.
</> 0.74 -0.75 0.15 -0.76 0.12 -0.16 Since stochastic variation of relative permeabilities gave little
loge kH -0.75 0.45 -0.56 -0.16 -0.06 additional effect on the production characteristic for these
Swi -0.43 0.79 -0.30 0.05 four rather different depositional environments, we would
Sor -0.19 -0.33 -0.29 like to investigate if these findings can be generalized to other
loge no -0.35 0.24 depositional environments.
loge nw 0.51
Calculation of the correlation between absolute permeability
Values above 0.7 are printed in bold face, and indicate a strong corre- and the water shock front velocity may give a clue.
lation between the variables.

Analysis of the Break-Through Time


displacement of oil by water under full pressure maintenance
is investigated. The end point mobility ratio is chosen as In one-dimensional fractional flow theory (neglecting capil-
lary forces and assuming incompressibility) the dimension-
1.0 for the mean value of k rw (Sor ). There is no capillary
less water shock front velocity, v, is given by the slope of the
pressure in the model. The density difference between water
tangent to the fractional flow curve (starting at initial water
and oil is 0.179 gjcm3 , which implies that the flow regime is saturation)12. The time to water break-through is then given
gravity dominated. See the paper by Tj0lsen and Damsleth S as ~, measured in pore volumes injected.
for further details.
The shape of the fractional flow curve, and hence the shock
front velocity, is determined by the shape of the relative
permeability curves. The obvious question is then: Is there
The results from the flow simulations for each depositional any relationship between the water shock front velocity, v,
environment are presented in Figure 1-4. Figure 1 shows the and the absolute permeability, k?

593
STOCHASTIC RELATIVE PERMEABILITIES USUALLY HAVE
4 NEGLECTABLE EFFECT ON RESERVOIR PERFORMANCE SPE26473

1200 3200 1200 3200

~E ~E
900 2400 ({)
900 2400 ({)
~ 2!S ~ 2!S
E -0 E -0
~ e ~ e
(I) 600 1600 c.. (I) 600 1600 c..
'iii 'iii
ex: 6 ex: 6
6 300 800
E
::J
6 300 800
E
::J
0 0

0 0 0 0
0 3 6 9 12 15 0 3 6 9 12 15
Years Years

1200 3200 1200 3200

900 2400
~E
900 E
....
2400 ({)
({)
~ 2!S ~ 2!S
E -0 E -0
~ e ~ e
(I) 600 1600 c.. (I) 600 1600 c..
'iii 'iii
ex: 6 ex: 6
6 300 800
E
::J
6 300 800
E
::J
0 0

0 0 0 0
0 3 6 9 12 15 0 3 6 9 12 15
Years Years

1200 3200

900 2400. J5
~ 2!S
E -0
~ e
(I) 600 1600 c..
'iii
ex: 0
6 300 800
E
::J
Constant relative permeability
0 Varying relative permeability

0 0
0 3 6 9 12 15
Years

Fig. 1 - The difference in oil rate and cumulative oil production in five realizations. Building block: MIAp, original
correlation.

594
SPE26473 C.B. TJ0LSEN, E. DAMSLETH AND T. BU 5

1200 3200 1200 3200

900 2400 "E


C/)
900 2400 "E
C/)
J2 =:. J2 =:.
E E
~
~
-0
e ~
CD 600 1600 c... CD 600 1600 c...
iii iii
a: (5 a: (5
(5
300 800
E (5
300
E
::l 800 ::l
0 0

0 0 0 0
0 3 6 9 12 15 0 3 6 9 12 15
Years Years

1200 3200 1200 3200

900 2400 "E


C/)
900 2400 "E
C/)
}2 =:. 12E =:.
E -0 -0
~ e ~ e
CD 600 1600 c... CD 600 1600 c...
iii iii
a: (5 a: (5
(5
300 800
E
::l
(5
300 800
E
::l
0 0

0 0 0 0
0 3 6 9 12 15 0 3 6 9 12 15
Years Years

1200 3200

900 2400 "E


C/)

~ =:.
-0
~ e
CD 600 1600 c...
iii
a: (5
(5 E Constant relative permeability
300 800 ::l
0 Varying relative permeability

0 0
0 3 6 9 12 15
Years

Fig. 2 - The difference in oil rate and cumulative oil production in five realizations. Building block: MIAd, original
correlation.

595
STOCHASTIC RELATIVE PERMEABILITIES USUALLY HAVE
6 NEGLECTABLE EFFECT ON RESERVOIR PERFORMANCE SPE26473

1200 3200 1200 3200

~E ~E
900 2400 (/)
900 2400 (/)
12 =. 12 =.
E E
~
(J) 600 1600 ~
0-
~
(J) 600 1600 ~
0-
iii iii
II: (5 II: (5
is 300 800
E
::l
(5
300 800
E
::l
U U

0 0 0 0
0 3 6 9 12 15 0 3 6 9 12 15
Years Years

1200 3200 1200 3200

~E ~E
900 2400 (/)
900 2400 (/)
~!!. =. !2 =.
E -0 E -0
~ e ~ e
(J) 600 1600 0- (J) 600 1600 0-
iii iii
II: (5 II: (5
(5
300 800
E
::l
(5
300 800
E
::l
U U

0 0 0 0
0 3 6 9 12 15 0 3 6 9 12 15
Years Years

1200 3200

~E
900 2400
'0 (/)
~--
E
=.
-0
~ e
(J) 600 1600 0-
iii
II: (5
(5
300 800
E
::l
Constant relative permeability
u Varying relative permeability

0 0
0 3 6 9 12 15
Years

Fig. 3 - The difference in oil rate and cumulative oil production in five realizations. Building block: PI, original correlation.

596
SPE26473 C.B. TJ0LSEN, E. DAMSLETH AND T. BU 7

1200 3200 1200 3200

900 2400 en
ME
900 E
2400 en
....
J2 =:. u
.;- =:.
E -0 E -0
~ e ~ e
(I) 600 1600 c.. (I) 600 1600 c..
iii iii
a: 6 a: 6
6 300 800
E
::l
6 300 800
E
::l
() ()

0 0 0 0
0 3 6 9 12 15 0 3 6 9 12 15
Years Years

1200 3200 1200 3200

ME ME
900 2400 en 900 2400 en
J2 ~
E
=:. ME =:.
-0 -0
~ e ~ e
(I) 600 1600 c.. (I) 600 1600 c..
iii iii
a: 6 a: 6
6 300 800
E
::l
6 300 800
E
::l
() ()

0 0 0 0
0 3 6 9 12 15 0 3 6 9 12 15
Years Years

1200 3200

ME
900 2400 en
U
M-
E
=:.
~
(I) 600 ~
1600 c..
iii
a: a
6 E Constant relative permeability
300 800 ::l
() Varying relative permeability

0 0
0 3 6 9 12 15
Years

Fig. 4 - The difference in oil rate and cumulative oil production in five realizations. Building block: T, original correlation.

597
STOCHASTIC RELATIVE PERMEABILITIES USUALLY HAVE
8 NEGLECTABLE EFFECT ON RESERVOIR PERFORMANCE SPE26473

If there is a strong positive correlation between v and k, one TABLE 4 Time to reach 2 % water cut
would in a "layercake" model expect an earlier break-through
Case Real. Stach. Canst. Diff.
when varying the absolute and relative permeabilities simul-
no. rel. perm. rel. perm. (St. - C.)
taneously, than in an uncorrelated situation.
Layercake, 1 1770 1933 -163
The correlations between v and k were computed for each neg. corr. 2 1598 1911 -313
of the four depositional environments, and the results are 3 1600 1704 -104
shown in Table 3.
4 1336 1517 -181
The proximal mouthbar deposit (MIAp) and the tidal bay 5 1750 1917 -167
(1') deposit both show a negative correlation between v and -------m-e-a-n------::-16""1:-:0,---- 1196 . ·---.:IS6*
10gekH, slightly below -0.5. However, the simulation results st.dev. 78 82 34
for MIAp, shown in Figure 2, does not show the expected
Layercake, 1 1770 1508 262
overall delay of water break-through, but for one realization.
pos. corr. 2 1598 1304 294
For the tidal bay deposit, Figure 4, a slight overall delay in
the water break-through can be recognized.
3 1600 1537 63
4 1336 1277 59
For the distal mouthbar deposit and the channel sand the 5 1750 1646 104
correlations between v and 10gekH are positive, though close .- --J.nea:il- .·--r6If----I454- "'I5'6"--
to zero. For both these environments, most of the realiza-
st.dev. 78 71 51
tions show an earlier break-through in the case of stochastic
varying relative permeabilities compared to the case of con- 3D model,1 1670 1562 108
stant relative permeabilities. pos. corr.
2 1639 1696 -57
3 1887 1778 109
In the more detailed study 9 , we constructed some cases where
4 1524 1325 199
the impact of the correlation between v and 10gekH could be
5 1339
demonstrated. Our findings are summarized below. ----------- _._ .._ - - _ 1228- . 111
mean 1612 1517 108'
st.dev. 51 51 45
"Layercake" reservoir
*) Based on a t-test the difference is larger or less than zero based on
To avoid some of the complexity in a realistic 3D model, a 5 % significance level.
we generated a "synthetic" reservoir of the same size and
dimension as previously, but with homogeneous layers. (This low permeable zones. Also the vertical cross flow may influ-
"layercake" model is what we would have got if the ratio ence the results. Thus, it is more difficult to predict the net
between the horizontal and the vertical spatial correlation effect of using stochastic relative permeabilities in 3D.
length, i.e. the range in the variogram, were infinity.) Suggested Technique
Using artificial (and possibly non-physical) correlation matri- Based on the above discussion, we propose the following in-
ces, we can generate realizations with strong positive or neg- formal test of whether stochastic modeling of relative per-
ative correlation between v and logek Jj. Flow simulations meability may be of importance:
were performed for five different realizations with a strong
• Estimate the correlation matrix, based on a sufficient
negative correlation and five different realizations with a pos-
number of data from the same depositional environment.
itive correlation. In all the negative correlation cases, a later
break-through was observed, while a later break-through oc- • Use the technique described by Tj¢lsen and Damsleth 9
cured when the correlation was positive. The results are to calculate an approximative correlation coefficient be-
summarized in Table 4. tween the front velocity and the absolute permeability.
• If the calculated correlation is close to ±1, and if there
in addition are strong anisotrophies in the reservoir,
Strong positive correlation for a stochastic 3D model
stochastic variation of the relative permeability curves
By using the artificial, positive correlation matrix described may affect the production profiles.
above, five stochastic 3D realizations similar to the ones de-
scribed for the MIAd building block were generated. The re- Discussion
sults, also given in Table 4, showed an earlier break-through
for four of the five realizations. t-tests confirm the signifi- The impact of a strong correlation between the permeability
cantly earlier break-through. In 3D realizations with layer- and the shock front velocity
ing trends, but no perfect layers, the water front will speed The time to water break-through depends on the water shock
up in the the high permeable zones and be retarded in the front velocity. Thus, if a positive correlation between the

598
SPE26473 C.B. TJ0LSEN, E. DAMSLETH AND T. BD 9

permeability and the shock front velocity exists, an earlier


water break-through is expected, for the case of varying
1200 3200
the relative permeabilities stochastically around their mean,
compared to use one constant relative permeability curve. ... ..... ,..
900 E
2400 en
This is validated for a "layercake" model. However, for a '0
;;- 6
3D model, with no perfect layers, the water front will speed E
up in the the high permeable zones and be retarded in the
~
(I) 600 1600 ~0-
iii
low permeable zones. The effect of the shock front velocity 0: 0
will thus tend to be averaged out, so to speak, except in the 0 300 800 E
~
extreme cases. ()

The observed correlations between v and 10gekH, Table 3, are


0 0
rather weak, and although the spatial model is anisotrophic
0 3 6 9 12 15
with fairly long correlation lengths compared to the inter-
Years
well distance; it is not by any means a "layercake". Thus,
it is not surprising that we do not observe any significant
Fig. 5 - Varying permeability, constant relative permeabil-
or systematic differences for neither of the four depositional
ity, compared to the base case (all constant). M1Ad.
environments, as shown in Figures 1-4.

The effect of stochastic absolute and relative permeability 1200 3200


Our previous work 6 ,9 as well as other contributions 2 ,7 have
shown that by introducing stochastic variations in the per-
,..
900 E
2400 en
'0
meability, while the relative permeability curves are kept M-
E
6
constant, we obtain a production profiles which vary around ~
-0
some mean values. The spread indicates the uncertainty in (I) 600 1600 e
0-
iii
the profile. Furthermore, when stochastic variations of per- 0: 0
meability is taken into account, the profiles tend to shift 0 300 800 E
~
systematically, compared with the profiles based on smooth ()

or constant permeability maps. Whether an earlier or later


break-through will result must be examined in each case. 0 0
Figure 5 gives one example of this effect in the present set- 0 3 6 9 12 15
ting. Years

Figure 6, on the other hand, shows the production profiles Fig. 6 - Constant permeability, varying relative permeabil-
obtained by keeping the absolute permeability constant at ity, compared to the base case (all constant). M1Ad.
its mean while the relative permeabilities vary stochastically.
The base case, where all the variables are kept constant, is Thus, there is much more numerical experimentation to be
also shown. Here the base case seems to represent a mean done before this field is fully understood. All the limitations
curve for the five realizations. Thus, these results indicate above could be relaxed, the effect of gravity is an important
that letting the relative permeabilities vary stochastically question, other well patterns could be examined, etc.
around their mean gives a somewhat larger spread between
the different realizations, while the mean characteristics are Conclusions
conserved.
In the four depositional environments we have studied and
under the given injection/production scheme, the introduc-
Limitations. reservations and future work tion of stochastic relative permeability curves had no signif-
icant effect on the mean production performance, compared
The present study covers a wide range of depositional envi- with constant relative permeabilities throughout the reser-
ronments. However, we have limited ourselves to study the VOlr.
effect of stochastic relative permeability curves within a wa-
ter injection production scheme with a unit mobility ratio, The stochastic relative permeability curves introduce an ad-
and we have ignored the effects of capillary forces. We have ditional variability between production profiles from different
not looked at other production mechanisms, and it may well realizations.
be that the effect of stochastic relative permeability is more Based on a simple linearization scheme, it is possible to cal-
important in more complex lOR/EaR schemes. further- culate an approximative correlation between absolute per-
more, our analysis has focused on the break-through time. meability and the front velocity. Extreme values of this

599
STOCHASTIC RELATIVE PERMEABILITIES USUALLY HAVE
10 NEGLECTABLE EFFECT ON RESERVOIR PERFORMANCE SPE26473

correlation (close to ±1) combined with strong anisotrophy 6. Damsleth, E., Tjll5lsen, C.B., Omre, H. and Haldorsen,
in the reservoir affects the time to water break-through. H.H.: "A Two-Stage Stochastic Model Applied to a North
Strong positive correlation gives an earlier break-through Sea Reservoir," JPT (April 1992) 402··408.
while strong negative correlation delays the break-through, 7. Waggoner, J.R., Castillo, J.L. and Lake, L.W.: "Simula-
when compared to the constant relative permeability situa- tion of EaR Processes in Stochastically Generated Per-
tion. In our real data sets these correlations are fairly small, meable Media," SPEFE (June 1992) 173-180.
however, and the effect is neglectable.
8. Tjll5lsen, C. B. and Damsleth, E.: "A Model for the Simul-
The main conclusion is that in some situations the stochastic taneous Generation of Core-Controlled Stochastic Abso-
modeling of relative permeability curves is of minor impor- lute and Relative Permeability Fields," paper SPE 22691
tance. However, the choice of mean relative permeabilities presented at the 1991 SPE Annual Technical Conference
may be crucial. and Exhibition, Dallas, Oct. 6-9.
9. Tjll5lsen, C. B. and Damsleth, E.: "The effect of Stochastic
Nomenclature Relative Permeabilities in Reservoir Simulation," Submit-
ted to Journal of Petroleum Science (3 and Engineering.
tjJ porosity, fraction 10. Tjll5lsen, C.B., Scheie, A. and Damsleth, E.: "A Study of
loge natural (base e) logarithm the Correlation between Relative Permeability, Air Per-
kH Horizontal Permeability, md meability and Depositional Environment on the Core-
kv vertical Permeability, md Plug Scale," Advances in Core Evaluation II, Reservoir
k rw relative permeability for water, dimensionless Appraisal, P.F. Worthington and D. Longeron (eds.), Gor-
krw(Sor)= relative permeability for water at Sor don & Breach, London (1991),169-183.
Sor residual oil saturation, fraction
initial water saturation, fraction 11. Young, L.C. and Hemanth-Kumar, K.: "High Perfor-
Swi
mance Black Oil Computations," paper SPE 21215 pre-
Corey-exponent for the oil relative
sented at the 1991 SPE Symposium on Reservoir Simula-
permeability curve, dimensionless
tion, Anaheim, California, Feb. 17-20.
Corey-exponent for the water relative
permeability curve, dimensionless 12. Welge, H.J.: "A Simplified Method for Computing Oil
v water shock front velocity, dimensionless Recover by Gas or Water Drive," Trans., AIME, 195
(1952) 91-98.

Acknowledgements
Thanks are extended to Norsk Hydro A.S for permission to
publish the paper.

References
1. Haldorsen, H.H. and Damsleth, E.: "Stochastic Model-
ing," JPT (April 1990) 404-412.
2. Omre, H., Halvorsen, K.B., Holden, L. and HlI5iberg, J.:
"Reservoir Heterogeneity, Geological Description and Ef-
fects on Fluid Flow - Models for Heterogeneity," Proc.,
SPaR-Seminar, Stavanger (Sept. 20-211988) 296-308.
3. Rudkiewicz, Guerillot,D. and Galli, A., 1990. "High Def-
inition Reservoir Lithology and Property Prediction with
an Integrated Software. Application to the Yorkshire
Middle Jurassic Formation," North Sea Oil (3 Gas Reser-
voirs II, A.T. Buller et al.(eds.), Graham & Trotman Ltd.,
London (1990) 399-406.
4. Hewett, T. and Behrens, R.A.: "Conditional Simulations
of Reservoir Heterogeneity With Fractals," SPEFE (Sept.
1990),217-225.
5. Journel, A.G. and Alabert, F.G.: "New Method for Reser-
voir Mapping," JPT (Feb. 1990) 212-218.

600

You might also like