Relative permeability

Consider flow of three phases (water, oil, gas) through a core:

kkrw A dPw
qw  
 w dx
kkro A dPo
qo  
o dx
kkrg A dPg
qg  
g dx

kr is a dimensionless quantity called the relative permeability that describes the

flow of each phase according to the pressure gradient of each phase.
Drainage Imbibition
 Salient points

1. Nonlinear functions of saturation (but may be linear at

large phase saturation)
2. Sum of relative permeabilities is always < 1 except when
Sw = 1 or 0
3. At Swirr, krw goes to 0 and krnw goes to some maximum
endpoint
4. At Snwr, krnw goes to 0 and krw goes to some maximum
endpoint
5. krnw at Swirr > krw at Snwr during imbibition (why?)
(and why is this not true for drainage?)
Determining relative permeability from capillary drainage curves

In a naïve sense, relative permeability should be proportional to the cross-sectional

area available for flow of each phase. This gives

Sw Sw
dS w
0  r  S w   dS w 0  Pc  S w   2
2

k rw  1  1
dS w
0  r  S w   dS w 0  Pc  S w   2
2

dS w
1

S  Pc  S w   2
k rnw  w

dS w
1

0  Pc  S w   2
 This naïve picture is not quite correct, however, since it does not allow for nonzero
residual saturations and always results in krw + krnw = 1. Burdine (1953) came up with
a better set of equations.

Recall that in a saturated medium, τ = Fφ = 1/φ assuming m = 2. We can generalize this

to saturations less than 1 as τ(Sw) = 1/φSw.

Ignoring constants that do not depend on porosity or saturation,

r2
k  2  k  r 2 2 S w2


By specifying that tortuosity goes to infinite when Sw goes to Swirr, this can be modified as

k  r 2 2 S w*2
 Applying this proportionality to the equations we came up with previously yields
S w* S w*
dS w*
 r  S    S   P  S  
2
* 2 *2
dS w*
k rw 
 
k rw S *
w
 0
w w

 S w*2
0 c w
* 2


k rw S w  1  dS 
   
* 1 1 *
2

0
r S w*  2 dS w*   P  S  
0
w
* 2
c w

Note that Sw* doesn’t appear in the denominator because Sw* = 1 in this case.
dS w*
1

k rnw  S 
  P  S  
* 2
 w
*
* 2 S w*
k rnw  w
 1  S c w

k rnw  S w*  0 1
dS  *

  P  S  
0
w
* 2
c w

(we need to specify this because krnw at Sw* = 0 is usually not 1)

For the wetting phase, k rw  k rw but for the nonwetting phase we need to know krnw

at Sw = Swirr to convert k rnw to krnw.

Using the Brooks-Corey Pc model:

1
Pc  Pe  S 
* 

w

S w* S w* 2
1 * 1 * *
 Pc2
dS w  Pe2
S w dS w
krw  S w*    S w
* 2 0
1
S 
* 2 0
w 1 2
1 * 1 * *
0 Pc2 w
dS 0 Pe2 Sw dSw
S w*
 2
1
S  * 

2
w 2 2  3
3
 S w* 2 0
1
S *
w S * 
w
 2
1
S  * 

2
w
0
Similar analysis for nonwetting phase:

2 
 *  
krnw   1  S 
* 2
w 1  S w 
 
The true relative permeabilities are given by

2  3
 S w  S wirr  
krw   
 1  S wirr 

2 
2 
 S  S wirr    S w  S wirr  

krnw  krnw  S wirr  1  w  1    
 S m  S wirr    1  S wirr  
 

Sm = wetting phase saturation at which krnw goes to 0. Often = 1, but not always.
Berea sandstone data from Brauns and Holland (1995)

Relative permeability

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Water saturation

Swirr = 0.25 Water Oil

Sm = 1
kro(Swirr) = 1
 1

0.1
Relative permeability

0.01

0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Water saturation

Fit with λ = 2.2

Permeability jail

If there is a range of Sw within which both krw and krnw < 0.02, you won’t be able to get
either phase to flow at a reasonable rate.

Typically happens in tight gas sandstones with absolute k < 0.05 mD.
Cluff and Byrnes, SPWLA, 2010
Hysteresis in relative permeability
Assume each pore is
connected to another
pore body
Pc1
kro = 0
Pc2
kro = 4
Pc3
kro = 8
Pc4
kro = 12
Pc3’
kro = 8
Pc2’
kro = 4
Pc1’
kro = 0
 Results

Drainage Drainage
Imbibition Imbibition

Pc kro

Sw Sw
 Real data: Berea sandstone

Drainage Imbibition
10000

1000

100
Pc (psi)

10

0.1
Braun and Holland, 1995
0.01
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Sw
Using fractal descriptions to determine wetting-phaserelative permeability

Fractal scaling: the number of pores with size larger than r is proportional to r -D,
where D is the fractal dimension.

Pore-solid fractal model (Perrier et al. 1999; Bird et al., 2000):

Consider a domain with dimension (side length) L in a space of Euclidean dimension d.
The domain is separated into N equal parts with linear size L/n. Of these N equal parts, a
fraction p is assigned as pore space, s assigned as solid, and f assigned as fractal (where the
pattern is to be repeated). p + s + f = 1.
The fractal dimension is given by d + log(1-(p+s))/log(n).
Iteration 1 (initial) Iteration 2 Iteration 3

p s p p s p p s p

s p f s p s p

s s s
f f

L = 3, N = 9, n = 3, d = 2, p = 1/3, s = 1/3, f = 1/3, D = 1.

Analysis of Ghanbarian-Alavijeh and Hunt, Geoderma, 2012:
The incremental probability distribution function for pore size R in the PSF model is

1 p 3  D 2 D  3  D 2 D
f  R  R  R
 p  s rmax
3 D
 rmax
3 D

From critical path analysis, the critical pore size rc is defined as

rmax

 f  R  dR  p
rc
c

3 D rmax
So,
 3 D
rmax
 R     rc 
3 D

 R dR     1      pc
2 D

 rmax
3 D
  rmax     rmax  
rc
rc  
1
 p  3 D
and rc  rmax 1  c 
  
To introduce a nonwetting phase, consider the case of primary drainage. The nonwetting
phase in this case will occupy the largest pores. Let r(S w) be the smallest pore that is
occupied by the nonwetting phase (or the largest pore that is occupied only by the
wetting phase) at water saturation Sw. We can now write
3 D r  S w 
  r S w  
 
r  Sw  r  Sw  3 D 3 D
 3 D   R  r 
S w   f  R  dR  R
2 D
dR         min  
 rmax
3 D
  rmax    rmax   rmax  
rmin rmin
rmin 
1
S r 
3 D
 3 D

and r  S w   rmax  w   min  

   rmax  

In order for the wetting phase to flow, it must overcome the percolation threshold. We
can define another critical pore size rc(Sw) such that the fraction of the pore volume
between rc(Sw) and r(Sw) is equal to pc:
3 D r  S w 
  r S w  
  rc  S w   
r  Sw  3 D 3 D
 3 D  R 
           p c
2 D
R dR 
 rmax rc  S w 
3 D
  rmax    rmax   rmax  
rc  S w  
1
1
 r  S   3 D
p c 
3 D
 p 3 D  3 D
rc  S w   rmax  w
    r 3 D  S w   c rmax 
so r
 max      

 1
 S r 
3 D
 3 D
Recalling that r  S w   rmax  w   min  
   rmax  
1 1
   rmin 
3 D
 p  3 D
S  rmin 
3 D
p c 
3 D
3 D  S w  c r 
we get rc  S w     

 rmax  
3 D
 rmax  w
 
    r    max     rmax   
  max
  

Since the integral of f(R) over the entire pore size range is equal to 1, we have
3 D rmax
   rmin 
rmax rmax 3 D
 3 D  R  
1   f  R  dR  3 D 
R 2 D
dR     1    
 rmax rmin   rmax     rmax  
rmin
rmin 

3 D 1
 rmin   S  p  3 D
so    1

and rc  S w   rmax  w  1   c 
 rmax      
One consequence of fractal descriptions of porous media is that pore size
is proportional to pore length. From the Poiseuille equation, recall that

r 4 dP r4
q q
8 dl l

But if r is proportional to l, we can express this as

q  r3
and we can go a little further and say that
k q rc3  S w 
 k rw   3
k sat qsat rc  S w  1

Thus,
3 3
3  S w  p  1 
3 D
          S w  pc   
3 D
rmax   1   c  3

             S w  pc   3 D
k rw     
3
 1    pc
3  pc  D
3 
     p  
 
rmax  1     c

  
This result can be improved a bit. The probability distribution for the conductances, f(g), is
related to f(R) by f(g)dg = f(R)dR. Therefore,
dR
f  g   f  R
1
dg
From our previous arguments, R  g and f  R   R
3 2 D
so
2 2 D 2 D
  
f  g  R 2 D
g 3
g 3
g 3
g 3

Now let’s take the average of the conductances along the percolating pathways (i.e., all
flow paths with g ≥ gc). To do this, we will average the resistances (reciprocal of
conductance) and then take the inverse of that average.
1 1 1
 g max
  g max
1
D    D3  
D D

g g f  g  dg 
1 1

1
g  g   g 3
dg    g max  g c 3  g c
3

 c   gc   

The last step arises from the fact that rc  rmax so g c  g max
D D
k g c3  S w  rcD  S w         S w  pc   3 D
Thus,  kr   D k rw  
k sat D
rc  S w  1 and 
g c
3
 S w  1    pc 
 Berea sandstone example

1 1

0.8 0.8

Mercury saturation
0.6 0.6
V(> r)

0.4 0.4

0.2 0.2

0 0
0 0 0.01 0.1 1 10 100 1000 0 5 10 15 20 25 30 35 40 45 50

Pore radius (microns) PHg (psi)

from micro-CT Model

rmax = 40 μm pc = 0.30
rmin = 0.5 nm
D = 2.55
β = 0.255
φ = 0.21
 Comparison to data of Braun and Holland (1995)
1

0.1

0.01
krw

0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Sw

Data Model
 What is there are 3 phases present?
Should we even care?

Leverett and Lewis, 1941

 If it turns out we should care, then…

krw

kro
krg

Leverett and Lewis, 1941

Commonly used models for 3-phase relative permeability
Stone I and II
Saturation-weighted (Baker)
Delshad and Pope
 Stone I: (Stone, JPT, 1970)
water and gas block the flow of the oil phase
S oe k ro  w  k ro  g 
k ro 
k ro  wi  1  S we  1  S ge 
S o  S om S w  S wi Sg
S oe  S we  S ge 
1  S wi  S om 1  S wi  S om 1  S wi  S om

kro(w) = kro at So = 1 – Sw
kro(g) = kro at So = 1 – Sg – Swi
Som = 3-phase residual oil saturation
Sg
S om  S orw  1    S org   1
1  S wi  S org
(Fayers and Matthew, SPE J, 1984)

Sorw = residual oil saturation in the presence of water

Sorg = residual oil saturation in the presence of gas and
irreducible water
Stone, 1970
 Stone II: (Stone, J. Can. Pet. Tech., 1973)
segregated flow of all three phases; residual oil saturation does
not need to be defined

k ro   k ro  w   k ro  wi  k rw o    k ro  g   k ro  wi  k rg  o    k ro  wi   k rw o   k rg  o  
Stone, 1973
Fayers and Matthew (SPE J, 1984) showed that
Stone I generally performs better than Stone II:
Saturation-weighted method:
(Baker, SPE 17369, 1988)
Relative permeability is a weighted average of two-phase rel perms:

 S w  S wi  kro  w    S g  S gr  kro  g 
k ro 
 S w  S wi    S g  S gr 
 So  Soi  krw o    S g  S gr  k rw g 
k rw 
 So  Soi    S g  S gr 
 S w  S wi  k rg  w    So  Soi  k rg  o 
k rg 
 S w  S wi    So  Soi 
 Stone I Stone II

Saturation-weighted Baker claimed a

better fit than
Stone models

Baker, 1988
Delshad and Pope method:
(Delshad and Pope, TiPM, 1989)
Weighted average with adjustable parameters

k ro  k o
row aS 
o 1  S   1  a  S 1  S 
w
 
o g


1  S w  S g  S or
So 
1  S wr  S gr  S or
S w  S wr
Sw 
1  S wr  S gr  S or
S g  S gr
Sg 
1  S wr  S gr  S or

Sg
S or  bS orw  1  b  S org b  1
1  S wr  S org
Which is better?
Stone I and II often yield curved isoperms
Saturation-weighted and Delshad/Pope often perform better
 Stone I Stone II

Saturation- Delshad/Pope
weighted

Delshad and
Pope, 1989
3-phase rel perm wrap-up

Lots of models out there (we only discussed 4)

All have assumptions
Many are strictly empirical
Many processes occur during 3-phase flow (trapping, layer
drainage, miscibility effects, wettability effects, etc. etc.)

This is the forefront of research… experimental data are very

difficult to collect (this department has some of the best capabilities
in the world, however) and the basic physics of the process are not
well understood.