Professional Documents
Culture Documents
gas saturations (Sgc). Therefore, separate saturation The relative permeability equations developed were
normalization equations must be used for each curve as follows: compared with correlations of Honarpour et al4, Rose8 and Narr
et al9. These works either did not employ wettability preferences,
S g − S gc or did not distinguish between oil-wet and intermediate (mixed)
Sg* = (2) wettability, and between water-wet and strongly water-wet
1 − ( S gc+ Slc ) systems. Additionally, Rose's and Narr's equations are so general
that they do not specify the type of rock. The curves are in
Sg close agreement with each other in terms of normalized relative
Sl * = 1 − (3) permeability of oil with respect to water as shown in Figure 3 for
1 − Slc
a water-wet sandstone. The same is also true for a carbonate
formation. Figure 4 shows normalized water relative
Regression Analysis permeability values calculated using the various correlations.
In this study, a forward stepwise multiple linear regression The equation developed falls between Honarpour's and Rose's
technique was employed in developing relative permeability curves. Rose’s plot seems to give unrealistic prediction values
prediction equations. This technique is based on an automatic for a water-wet system because the endpoint of the krw* curve
search procedure concept which develops the best subset of (krw* at Sw = 1-Sorw) is much higher than expected for a water-
independent variables sequentially, at each regression step wet case. A criticism of Honarpour’s model is that the water
adding or deleting one independent variable at a time in an relative permeability values appeared low resulting in optimistic
attempt to get the highest possible coefficient of multiple recoveries. This criticism lead the present study to separate the
determination, R2 value. The R2 is interpreted as the proportion strongly water-wet curves from the regular water-wet curves.
of observed values of Yi that can be explained by the regression Figure 5 clearly illustrates this point. For gas-oil systems, the
model and is used to measure how well the model fits the data. equations presented have eliminated the requirement of an
The higher the value of R2, the more successful the model is in endpoint value for krg.
explaining the variation of Y. A value of one indicates that all
points lie along the true regression line, whereas a value of zero Conclusions
indicates the absence of a linear relationship between variables. Twenty four, two phase relative permeability prediction
In the latter case, the modeler has to search for an alternative equations have been developed through extensive trial and error
model such as a nonlinear model. Since observational data model building processes using linear regression analysis for
(data obtained after the experiments were completed) were used four different systems which commonly exist in the petroleum
in this analysis, an R2 value slightly higher than 60% was industry. In oil-water systems, prediction equations for three
considered highly satisfactory6. One criticism of using the R2 types of rock wettability were formed in addition to
criteria as the only indication of goodness of fit is that the R2 classification of the equations on the basis of rock type, i.e.,
value will keep increasing if more independent variables are sandstone and carbonate. Additionally, completely new
introduced into the model. To balance the use of more correlations for strongly water-wet system for both sandstone
parameters against the gain in R2, many statisticians use the and carbonate were developed. As in the oil-water systems,
adjusted R2 value (R2adj)7. Very simply, the R2adj value prediction equations according to rock type were successfully
approaching R2 indicates that excessive terms were not included developed for gas-oil systems. Completely new correlations
in the model. based on a linear regression analysis were developed for gas-
water and gas-condensate systems. Based on an extensive
Discussion and Comparisons review of existing data, modifications to wettability
The prediction equations developed in this study are listed in determination were developed.
Appendix A and defined in Table 2 which summarizes the
characteristics of the equations developed for all four fluid Nomenclature
systems. All R2 values well exceed 60% and all R2adj values are Capital Letters
within 1.5% of R2 thus indicating a reasonable fit without
excessive terms. Figures 1 and 2 graphically illustrate R2 = coefficient of multiple
normalized oil-water and gas-oil relative permeability values for determination
both sandstone and carbonate formations. Tables 3 and 4 list the R2adj = adjusted coefficient of
ranges of rock properties and fluid saturations used in multiple determination
developing prediction equations for oil-water and gas-liquid Sg = gas saturation, fraction
systems, respectively. Sgc = critical gas saturation
4 M. N. MOHAMAD IBRAHIM, L. F. KOEDERITZ SPE 65631
ka = absolute permeability, md 7. Devore, J. L. : Probability and Statistics for Engineering and the
Sciences, Duxbury Press., California, (1995), pp. 474 ff.
krcg = relative permeability of
condensate with respect to 8. Rose, W. : “Theoretical Generalizations Leading to the Evaluation
gas, fraction of Relative Permeability”, Trans. AIME, vol. 186, (1949), pp. 111 ff.
krg = relative permeability of
gas, fraction 9. Naar, J. and Henderson, J. H. : “An Imbibition Model, its
krgw = relative permeability of gas Applications to Flow Behavior and the Prediction of Oil
with respect to water, fraction Recovery”, Trans. AIME, Part II, vol. 222, (1961), p. 61.
krlg = relative permeability of liquid
with respect to gas, fraction General Reference
krog = relative permeability of oil Mohamad Ibrahim, M. N. :"Two-Phase Relative Permeability Prediction
Using A Linear Regression Model", Ph.D. Dissertation, University of
with respect to gas, fraction
Missouri, Rolla, (1999)
krow = relative permeability of oil
with respect to water, fraction
krw = relative permeability of
water, fraction
Greek Symbol
N = porosity, fraction
Superscript
* = normalized value
SPE 65631 TWO-PHASE RELATIVE PERMEABILITY PREDICTION USING A LINEAR REGRESSION MODEL 5
Appendix A
2 3 4
krow * = 1 − 21529
. S w * + 0.6389135 S w * + 14345325
. S w * −0.919704 S w * ( A1)
krow * = 1 − 3090996
. S w * +2.8670229 S w *1.6 − 0.768952 S w *2 ( A5)
S w *3
krw * = 0.22120304 S w *1.6 +0.24933592 S orw 2 + 21370925
. S w *2 S orw 5
φ
+ 83.491972 φ 4 S w *5 S orw1.5 − 0.4562939 S wc 3S w *4 +116107198
. (φ 2 S wc S orw S w *) 2
− 8.7866012 S orw 3S w *2.3 +0.00000578 S w *3 ( S wc ln ka )10 (1 − S orw )0.4
− 12.841061 S w *2 ((ln ka )S wc )3 φ 6 ( A6 )
krw * = 01163954
. S w *4 +2.66958338 S w *0.8 ( S wc S orw )2 + 0.47536676 S w *( S wc ln ka )2
− 0.3912824 S w *( S wc ln ka )3 + 752.014909 ( S wc S w *)2 φ 3 − 398.40214 φ 2.5 S wc 2 S w *2.2
− 152.43629 (φ S orw )3 S w *2.7 +0.22964285 S orw 0.5 S w * ( A8 )
krow * = 1 − 3254725
. S w * +38176666
. S w *2 − 1563216
. S w *3 ( A13)
krog * = 01599039
. S l * − 1045545
. Sl *2 +4.0843698 S l *3 −5.414161 S l *4
+ 3.2103149 Sl *5 ( A17)
krgw * = 13046802
. S g * −8159598
. S g *2 + 2550978
. S g *3 −3153754
. S g *4
+ 13883828
. S g *5 ( A21)
krcg * = 01194373
. S l * −0.089246 S l *2 +0.9606793 S l *3 ( A23)
krg * = 333929676
. S g *1.2 + 6.75670631 S g *0.9 S lc1.2 − 20.926791 S g * Slc 2 S gc
+ 101654474
. S g * S gc 4 − 7.3835856 Slc 0.5 S g * ( A24)
Strongly
Water-Wet: ≥ 15% ≥ 45% ≤ 0.07
Sandstone A1 16 127 87 86
Strongly A2 16 127 82 81
Water-Wet
Carbonate A3 6 49 99 99
A4 6 49 89 88
Carbonate A7 28 317 94 94
Oil-Water A8 28 317 80 80
Sandstone A9 43 396 93 93
Intermediate
Wettability A10 43 396 90 90
(Mixed-Wet)
Carbonate A11 29 278 93 93
A12 29 278 85 85
A16 19 184 86 85
A20 14 119 89 89
A22 19 166 88 87
A24 17 115 85 85
SPE 65631 TWO-PHASE RELATIVE PERMEABILITY PREDICTION USING A LINEAR REGRESSION MODEL 9
Table 3. Ranges of rock properties and fluid saturations used in developing oil-water relative permeability equations
A9 & A10 8.0 - 32.6 3.4 - 10,500 5.0 - 38.9 11.09 - 44.4
A11 & A12 5.9 - 38.3 1.08 - 4,018.7 7.0 - 41.0 13.9 - 50.0
A13 & A14 9.1 - 33.0 1 - 5,010 4.7 - 44.0 7.67 - 55.0
A15 & A16 9.8 - 35.0 1.3 - 1,420 8.0 - 53.6 9.8 - 57.0
Table 4. Ranges of rock properties and fluid saturations used in developing relative permeability equations for gas-liquid systems
A17 6.3 - 39.0 1.48 - 5580 0.6 - 25.0 3.28 - 50.0 3.5 - 48.0
A18 6.3 - 39.0 1.48 - 3650 0.6 - 25.0 3.28 - 50.0 5.0 - 48.0
A19 & A20 9.0 - 34.9 4.3 - 731 0.01 - 13.52 6.0 - 51.1 5.0 - 38.6
A21 & A22 5.0 - 25.0 0.1 - 345 3.0 - 47.9 10.0 - 61.2 Not Applicable
A23 & A24 6.0 - 26.6 Not Available 2.0 - 30.0 9.0 - 60.0
10 M. N. MOHAMAD IBRAHIM, L. F. KOEDERITZ SPE 65631
Figure 1 Figure 2
Oil-Water Relative Permeability Gas-Oil Relative Permeability
Water-Wet Rocks Sandstone & Carbonate
1
1
φ = 15%
ka = 100
0.8 Sandstone
0.8 md
Sandstone
Swc = 15%
Carbonate
Sorg =15%
0.6 0.6
Carbonate
kr *
φ = 15%
kr *
0.4 ka =100 md
Swc =20%
0.4
Sorw =20%
0.2
0.2
0
0 0.2 0.4 0.6 0.8 1
Sw (fraction) 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Sg (fraction)
Figure 3 Figure 4
Water-wet Sandstone Water-Wet Sandstone
Oil Relative Permeability to Water Water Relative Permeability
1
0.8
Equation A6
0.8
φ = 15%
ka = 100 md
Honarpour's Equation A5 Swc = 20%
0.6
0.6 Sorw = 20%
krow *
φ = 15% Honarpour's
ka = 100 md Rose's
krw *
0.2 Narr's
0.2
0
0 0.2 0.4 0.6 0.8 1 0
Sw (fraction) 0 0.2 0.4 0.6 0.8 1
Sw (fraction)
SPE 65631 TWO-PHASE RELATIVE PERMEABILITY PREDICTION USING A LINEAR REGRESSION MODEL 11
Figure 5
Water-Wet System
Carbonate
0.6
0.5 WW
φ = 10%
ka = 100 md
Swc = 20% Strongly WW
0.4
Sorw=20%
kr *
0.3 Honarpour's WW
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Sw (fraction)