Upload
weldsv1
View
220
Download
0
Embed Size (px)
Citation preview
8/9/2019 1__ResSimCh7
1/57
CONTENTS
1. INTRODUCTION
1.1 Averaging
1.2 Effective, equivalent and pseudo permeability
1.3 When to Use Upscaling
2. SINGLE-PHASE UPSCALING
2.1 Introduction
2.2 Analytical Scale-up Methods
2.3 Tensor Permeability
2.4 Numerical Methods
2.5 Examples
2.6 Simulation with Full Permeability Tensors
2.7 Awkward Cases
2.8 Summary of Single-Phase Scale-up
3. TWO-PHASE UPSCALING
3.1 Introduction
3.2 Balance Of Forces
3.3 Steady-State Methods
3.4 Dynamic Methods
3.5 Additional Upscaling Methods
3.6 Summary of Two-Phase Scale-up
4. PRACTICAL APPLICATIONS
4.1 Introduction
4.2 Upscaling Stages
4.3 Recent Approaches
4.4 SPE Upscaling Study
4.5 Additional Topics
4.6 Upscaling Packages
4.7 Summary
6. REFERENCES
7 Permeability Upscaling
8/9/2019 1__ResSimCh7
2/57
2
Learning Objectives
After reading through this Chapter, the Student should be able to do
the following:
• Dene effective permeability, and understand when the denition applies.
• Recognise the difference between effective and equivalent permeability, and
pseudo relative permeability.
• Know how to calculate the effective permeability for simple models
analytically.
• Realise that permeability is a tensor property, and be able to distinguish
between the terms of the tensor (kxx, kxy, etc.).
• Understand numerical upscaling for single-phase flow, and know when to
apply different boundary conditions.
• Know how to carry out steady-state upscaling for two-phase ow.
• Become familiar with conventional two-phase dynamic upscaling
techniques, particularly the Kyte and Berry and Stone methods.
• Be aware of a range of other two-phase upscaling techniques.
• Judge when upscaling from the core-scale to the geological model cell is
required.
• Identify which upscaling methods are suitable for different stages of
upscaling
8/9/2019 1__ResSimCh7
3/57
Institute of Petroleum Engineering, Heriot-Watt University 3
7 Permeability Upscaling
1. INTRODUCTION
Reservoir modelling frequently involves generating multi-million cell models, which
are too large for carrying out ow simulations using conventional techniques. The
numbers of cells must therefore be reduced by “upscaling” (Figure 1). This is what most
engineers refer to when they talk about upscaling. However, as shown in this Chapter,
upscaling may be used to calculate effective properties at a variety of scales.
Geological model
Full-field model
Some quantities, such as porosity and water saturation, are easy to upscale, because
they may be averaged arithmetically. However, other quantities - notably permeability
- are much more difcult to upscale.
1.1 AVERAGINGThis simple example shows how to average porosity and permeability. Suppose we
have 10 grid blocks of size 1 cm3, 4 of which have a porosity of 0.15, and 6 of which
have a porosity of 0.20, as shown in Figure 2.
φ = 0.15 φ = 0.20
Sw = 0.40S
w = 0.50
The average porosity, φ , is given by:
φ = = × + ×
=total pore volume
total volume
4 0 15 6 0 20
100 18
. .. . (1)
When averaging the water saturation, we need to take the porosity into account. In theprevious example, suppose the water saturation was 0.5 in the blocks with porosity of
0.15, and 0.4 in the blocks with porosity 0.2, then the average water saturation is:
Figure 2.
Example for averaging porosity and water
saturation
Figure 1.
Upscaling to reduce
the size of a geological
model for full-eld ow
simulation. The model on
the left contains 20 million
grid cells and the one on
the right contains about
300,000 cells.
8/9/2019 1__ResSimCh7
4/57
4
S
total amount of water
total pore volumew = =
× × + × ×
× + ×
=+
= =
4 0 15 0 5 6 0 20 0 4
4 0 15 6 0 20
0 3 0 48
1 8
0 78
1 80 433
. . . .
. .
. .
.
.
.. . (2)
1.2 Effective, Equivalent And Pseudo PermeabilityWe cannot, in general, average the permeability in such a simple way, because the
effect on ow of a block of rock depends on the direction in which uid is owing
through the rock, and on the surrounding rock. Because of these problems, we
give two different denitions for single-phase permeability upscaling, and another
denition for two-phase ow.
Note that, when we upscale permeability, we are assuming the Darcy’s Law holds at
the coarse-scale as well as the ne-scale.
1.2.1 Effective PermeabilityIn scale-up, we wish to replace a ne-scale heterogeneous permeability distribution,
with a single effective permeability value. This process is called homogenisation, and
has been studied by researchers in a variety of elds. The true effective permeability
of a medium is an intrinsic property of that medium, and does not depend on the
boundary conditions. (Boundary conditions are the pressures, pressure gradients
and ows surrounding the model.) However, a true effective permeability may only
be calculated in limited cases. Figure 3 shows two examples, one stochastic and the
other deterministic, where the calculation of effective permeability is valid. In each
case, it is assumed that a permeability pattern, of length-scale a, extends over a large
region of size c. The condition for being able to calculate the effective permeability
of a region of size, b, is that a
8/9/2019 1__ResSimCh7
5/57
Institute of Petroleum Engineering, Heriot-Watt University 5
7 Permeability Upscaling
1.2.2 Equivalent Permeability
If we do not have a wide separation of distance scales, then the “averaged” permeabilityis not strictly speaking an effective permeability. It will depend on the boundary
conditions, and should be called the equivalent permeability. In reservoir simulation,
it is often the case that we are only calculating the equivalent permeability. However,
people usually refer to it as the effective permeability.
1.2.3 Pseudo permeabilityThere is a third type of permeability “average”, and this is the pseudo permeability.
When we are dealing with two-phase ow, we need to scale-up the relative permeabilities
as well as the absolute permeability. (See section 3.) Pseudo relative permeabilities
(pseudos) are tables of numbers, which allow us to reproduce a ne-scale simulation
on a coarse grid. They depend on many factors, such as the injection rate, and
the direction. In addition, pseudos may be calculated so that they compensate for
numerical dispersion.
Oil
Water
Oil
Water
Fine
Coarse
Time
R e c o v e r y
I P
Scale-up using pseudos is not a rigorous process, and is subject to error, unless carried
out carefully. There are several methods, each of which may perform well in some
situations and badly in others.
Figure 4
Example of a permeability
distribution where only an
equivalent permeability may
be calculated
Figure 5
A ne- and a coarse-scale
simulation of a waterood.
The pseudo relative
permeabilities are
calculated so that the
recovery at the coarse-scale
matches that for the ne-
scale
8/9/2019 1__ResSimCh7
6/57
6
Before describing how to calculate effective, equivalent and pseudo permeabilities,
we outline different situations where upscaling may be necessary.
1.3 When To Use UpscalingMost reservoirs are modelled using, what is commonly termed a “ne-scale geological
model”. This is a stochastic model with grid cells of size approximately 50 m in the
horizontal directions, and about 0.5 m in the vertical. There are typically about 107
such cells in a full eld model. These cells must be reduced in number to about 104 for
full-eld simulation. However, each of the grid cells in the geological model is likely
to be heterogeneous, containing, for example, sedimentary structures. Petrophysical
data (permeabilities, relative permeabilities, and capillary pressures) are acquired from
core plugs, which are only a few cm long. When small-scale structure is present,
petrophysical data should be upscaled before being applied to the grid blocks of thegeological model.
For convenience, we split upscaling into two separate stages (see gure 6). Stage
1 is upscaling from the smallest scale at which we may treat the rock as a porous
medium (rather than a network of pores), up to the scale of the stochastic geological
model, i.e. from the mm - cm scale to the m - Dm scale. Stage 2 is upscaling from
the stochastic geological model to the full-eld simulation model. (Note that each
stage could involve several levels of upscaling.
1.3.1 Stage 1 UpscalingThis level of upscaling is frequently ignored by engineers, and “rock” relative
permeability curves are applied directly to the geological model. However, work
carried out at Heriot-Watt University has demonstrated that small-scale structures,
such as sedimentary lamination may have a signicant effect on oil recovery (Corbett
et al., 1992; Ringrose et al., 1993; Huang et al., 1995). For example, in a wateroodof a water-wet rock, water is imbibed into the low permeability laminae, and oil
becomes trapped in the high permeability laminae.
Figure 6
Two separate stages of
upscaling. (Geological
model taken from “TenthSPE Comparative Solution
Project: A Comparison of
Upscaling Techniques”, by
Christie and Blunt, 2001)
8/9/2019 1__ResSimCh7
7/57
Institute of Petroleum Engineering, Heriot-Watt University 7
7 Permeability Upscaling
The Geopseudo Method is an approach, where upscaling is carried out in stages, using
geologically signicant length-scales. Models of typical sedimentary structures arecreated and permeability values are assigned to the laminae (from probe permeameter
measurements, or by analysing core plug data). Relative permeabilities and capillary
pressure curves are also assigned to each lamina-type (by history matching SCAL
experiments on core plugs). Flow simulations are carried out to calculate the effective
single-phase permeability and the two-phase pseudo parameters. Additional stages
of modelling and upscaling may be required - e.g. upscaling from beds to bed-sets
or facies associations.
In the nest-scale model, the grid cells may be a mm cube, or less. If we upscale to
blocks of 50 m x 50 m x 0.5 m, we are upscaling by a factor of at least 5 x 104 in the
horizontal directions and 500 in the vertical.
Low Perm High Perm
Individual Rel. Perm Curves
Pseudo Rel. Perm CurvesEffective Perm
1.3.2 Stage 2 UpscalingWhen engineers talk about upscaling, they usually refer to reducing the number of
cells in the geological model from about 107 to about 104, in order to carry out full-
eld simulation. Note that this is equivalent to upscaling by a factor of about 10 in
each direction, so the scale-up factor for Stage 2 upscaling is much smaller than for
Stage 1 upscaling. Once again two-phase (or even three-phase!) upscaling should be
carried where necessary, although frequently, only single-phase upscaling is used.
2. SINGLE-PHASE UPSCALING
2.1 IntroductionSingle-phase upscaling is relatively easy, compared with two-phase scale-up. We shall
assume in this section that we have steady-state, linear ow, with no sources or sinks
in the region of interest. In most cases, we shall also assume that we have widely
separated distance scales, so that we are calculating true effective permeabilities. We
start with simple cases, where the permeabilities may be averaged.
Figure 7
Illustration of the
Geopseudo Method
8/9/2019 1__ResSimCh7
8/57
8
2.2Analytical Scale-up Methods
2.2.1 Flow Parallel to Uniform Layers
∆
x
P1 P2
k i, tiQi
Consider a set of (innite) parallel layers of thickness, ti and permeability k
i, where i
= 1, 2, .. n (the number of layers). The effective permeability of these layers is given
by the arithmetic average , ka.
k k
t k
t eff a
i i
i
n
i
i
n= = =
=
∑
∑
1
1
. (3)
(Equation (3) may be proved by applying a xed pressure gradient along each
layer.)
Example 1
x
z
t1 = 3 mm, k 1 = 10 mD
t2 = 5 mm, k 2 = 100 mD
1
2
Suppose we have two layers as shown in Figure 8. The effective permeability for
ow in the x-direction is given by Equation (3), and is:
k a
=× + ×
+
= =3 10 5 100
3 5
530
866 25. mD
Figure 8
Along-layer ow
Figure 9
A simple, two-layer example
8/9/2019 1__ResSimCh7
9/57
Institute of Petroleum Engineering, Heriot-Watt University 9
7 Permeability Upscaling
2.2.2 Flow Across Uniform Layers
∆x
Q
k i, ti∆Pi
For ow perpendicular to the layers, the effective permeability is given by the
harmonic average, kh:
k k
t
t
k
eff h
i
i
n
i
ii
n= = =
=
∑
∑
1
1
. (4)
(Equation (4) may be proved by assuming a constant ow rate through each layer.)
Example 2
Equation (4) may be used to calculate the effective permeability for ow across the
two layers in the model shown in Figure 9, i.e. ow in the z-direction.
k h
=+
+
= =3 5
3 10 5 100
8
0 3522 86
.. mD.
From Examples 1 and 2, we see that the permeability is different in different directions.
In reservoirs with approximately horizontal layers, the arithmetic average may be
used for calculating the effective permeability in the horizontal direction, and the
harmonic average may be used for calculating the effective permeability in the
vertical direction.
2.2.3 Flow through Correlated Random FieldsFigure 11 shows an example of a correlated random permeability distribution.
(“Correlated” means that areas of high or low permeability tend to be clustered, so
that the spatial distribution is smoother that a totally random one.) The “correlation
length” is approximately the size of patches of high or low permeability. The longer
the correlation length, the longer will be the range of the semi-variogram for the
permeability distribution.
Assuming that we are averaging over many correlation lengths (i.e. we have a large
separation of distance scales), permeability should be isotropic (same in the x-, y- andz-directions). The effective permeability for a random permeability distribution is
proportional to the geometric average, which is given by:
Figure 10
Across-layer ow
8/9/2019 1__ResSimCh7
10/57
10
k k
ng
ii
n
= ( )
=∑exp ln1 (5)
where i = 1, 2, .. n is the number of cells in the distribution.
Correlation Length
The results given below have been derived theoretically for log-normal distributions,
with a standard deviation of σY, where Y = ln(k). The results depend on the number
of dimensions:
k eff = k g 1 −σY 2
2( ) in1 D
k eff = k g in 2 D
k eff = k g 1 +σY 2
6( ) in 3 D
(6)
These formulae are approximate, and assume σY is small (< 0.5). (Gutjahr et al,
1978.) The 1D result is an approximation of the harmonic average. Note that the
results do not depend on the correlation length of the eld, provided it is much
smaller that the system size.
Also note that ka > k
g > k
h, and the effective permeability always lies between the
two extremes: ka and k
h.
Example 3
Figure 11
A correlated, random
permeability distribution
Figure 12
A random arrangement of
the permeabilities in the
simple example
8/9/2019 1__ResSimCh7
11/57
Institute of Petroleum Engineering, Heriot-Watt University 11
7 Permeability Upscaling
Suppose that the permeability values in the simple example are jumbled up, so that
there are 75 small cells of 10 mD, and 125 small cells of 100 mD. See Figure 12.The effective permeability of this model will be:
pseudos - solid linesave. rock curves - dashed lines
Water Saturation
R e l a t i v e P e r m e a b i l i t y
0 0.2 0.4 0.6 0.8
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
keff = k g = 1075× log(10 )+125× log(100)
200
= 1075+ 250200
= 101.625
= 42.17mD.
2.2.4 Additional Averaging MethodsTo increase the accuracy of permeability averaging, some engineers use power
averaging. The power average is dened as:
k
k
n p
i
i
n
=
=∑ α
α
1
1/
(7)
where α is the power.
Also, sometimes, engineers use a combination of the arithmetic and harmonic averages,
e.g. they take the arithmetic average of the permeabilities in each column and them
calculate the harmonic average of the columns.
2.3 Tensor PermeabilityNow suppose that we have layers which are tilted at an angle to the horizontal, as
in Figure 13.
8/9/2019 1__ResSimCh7
12/57
12
∆P
net flow in x-dir
net flow in z-dir
x
z
A pressure gradient has been applied in the x-direction. This will obviously give rise to
a ow in the x-direction. The uid takes a path through the medium, so that it expends
a minimum amount of energy. There will be a component of ow up the high perm,
and only a small amount of ow across the low perm laminae, as shown. This gives
rise to a net ow in the z-direction, or crossow. Here, the term crossow is used
to describe ow perpendicular to the applied pressure gradient. When calculating
the effective permeability of this model, we need to take this crossow into account.
This may be done using a tensor effective permeability, k , where:
k
k k k
k k k k k k
xx xy xz
yx yy yz
zx zy zz
=
(8)
The rst index applies to the ow direction, and the second to the direction of the
pressure gradient. For example kxy
gives the ow in the x-direction caused by a
pressure gradient in the y-direction. The terms kxx
, kyy
, kzz
are known as the diagonal
terms. These are the terms which are usually considered - the horizontal and vertical
permeabilities, kh and k
v, respectively. The other terms, which describe the crossow,
are the off-diagonal terms.
With tensor permeabilities, Darcy’s Law becomes:
u = − ⋅ ∇k
µ P, (9)
where u is the Darcy velocity (vector) and P is pressure (scalar).
u k P
xk
P
yk
P
zxx xy xz x
= − ∂
∂ +
∂∂
+ ∂
∂
1
µ (10)
u k P
x k P
y k P
zy yx yy yz= −
∂∂ +
∂∂ +
∂∂
1
µ (11)
Figure 13
Crossow due to tilted
layers. White represents
high perm and dark
represents low perm
8/9/2019 1__ResSimCh7
13/57
Institute of Petroleum Engineering, Heriot-Watt University 13
7 Permeability Upscaling
u k
P
x k
P
y k
P
zz zx zy zz= −
∂
∂ +
∂
∂ +
∂
∂
1
µ (12)
In the model with horizontal layers (Figure 9), there is no crossow, so the tensor
is diagonal:
k =
66 25 0
0 22 86
.
.
or in 3D:
k =
66 25 0 0
0 66 25 0
0 0 22 86
.
.
.
In the random model (Figure 11), there will be little net crossow, so the effective
permeability tensor will be approximately:
k =
42 17 0
0 42 17
.
.
2.3.1 Flow Through Tilted Layers
θ
x
z
x’
z’
If the layered model in the previous example is tilted, the effective permeabilities in
the x’ and z’ directions will be the same as before (i.e. the arithmetic and harmonic
averages, respectively). However, in the x-z co-ordinate system, the effective
permeability should be represented by a full tensor. The terms of the tensor may becalculated from the arithmetic and harmonic averages, as follows:
Figure 14
Layers tilted at an angle of θ
to the horizontal
8/9/2019 1__ResSimCh7
14/57
14
k
k k k k
k k k k
a h a h
a h a h=
+ −( )−( ) +
cos sin sin cos
sin cos sin cos
2 2
2 2
θ θ θ θ
θ θ θ θ (13)
This formula is obtained by rotating the co-ordinate axes through an angle θ. (Youare not required to know the proof). This example is in 2D, so only the k
xx, k
xz, k
zx
and kzz
are shown. Further rotations may be carried out around the x’ or z’ axes to
obtain a full 3D tensor.
Note that:
• The tensor is symmetric (kxz
= kzx
).
• Depending on the sign of θ, the off-diagonal terms may be positive ornegative.
Example 4
Suppose the example in Figure 9 is rotated by 30º, and calculate the effective
permeability tensor.
cos230 = 0.75, sin230 = 0.25, sin30.cos30 = 0.433.
From before, ka = 66.25 mD, and k
h = 22.86 mD.
kxx
= 66.25 x 0.75 + 22.86 x 0.25 = 55.40 mD,
kxz
= (66.25 - 22.86) x 0.433 = 18.79 mD,
kzz
= 66.25 x 0.25 + 22.86 x 0.75 = 33.71 mD.
k =
55 40 18 79
18 79 33 71
. .
. .
2.4 Numerical MethodsIn general, the permeability distribution will not be simple enough for us to be able to
calculate the effective permeability analytically, and we will have to perform a numerical
simulation. We can use a nite difference method to calculate the pressures.
2.4.1 Recap on Flow SimulationThe continuity equation tells us that there is no net accumulation or loss of uid
within a grid block:
qxin
+ qzin
= qxout
+ qzout
(14)
(We are assuming incompressible rock and uids, here.)
8/9/2019 1__ResSimCh7
15/57
Institute of Petroleum Engineering, Heriot-Watt University 15
7 Permeability Upscaling
q xin q xout
q zin
q zout
i,j i+1,ji-1,j
i,j+1
i,j-1
x
z
Darcy’s law is used to express the ows in terms of the pressures and permeabilities.
For example, if the grid block in Figure 14 is of length Δx and height Δz and width
Δy, and if it is centred on location (i,j), then:
qk z y P P
x xin
x i j i j i j = −
−( )− −, / , , ,1 2 1∆ ∆∆µ
(15)
where kx,i-1/2,j
is the harmonic average of the permeabilities in the x-direction in blocks
(i-1,j) and (i,j). (The harmonic average is used, because ow from one cell to the
next, is equivalent to ow across two layers. See Section 2.2.2) The other ows are
calculated in a similar manner.
It is useful to use the transmissibilities, Tx = kxΔzΔy/Δx and Tz = kzΔxΔy/Δz. We cantherefore derive the pressure equation:
T T T T P
T P T P
T P T P
x i j x i j z i j z i j i j
x i j i j x i j i j
z i j i j z i j i j
, / , , / , , , / , , / ,
, / , , , / , ,
, , / , , , / ,
− + − +
− − + +
− − + +
+ + +( )− −
− −
1 2 1 2 1 2 1 2
1 2 1 1 2 1
1 2 1 1 2 1
== 0
(16)
An equation is set up for each Pij, 1 = 1, 2, .. nx and j = 1, 2, .. nz. The transmissibilities
are known, and using the appropriate boundary conditions, we can solve this set of
linear equations to obtain the pressure in each grid block. The effective permeability isthan calculated from the total ow and the total pressure drop, as described below.
2.4.2 Boundary ConditionsBoundary conditions are required to specify what happens at the edges of the
model.
Figure 15
Recap on ow simulation
8/9/2019 1__ResSimCh7
16/57
16
(a) No-Flow Boundaries
no flow through the sides
no flow through the sides
P1 P2
The pressure is xed on two sides of the model, (four sides in 3D) and no ow is
allowed through the others sides of the model (see gure 16). This type of boundary
condition is suitable for models where there is little crossow: for example, models
with approximately horizontal layers, or a random distribution.
A diagonal effective permeability tensor may be calculated using the method outlined
below. See Figure 17.
on left face
Pressure= P2Pressure= P1
on right face
L
Area, AFlow Rate, Q
x
y
z
1 Solve the steady-state equation to give the pressures, Pij
, in each grid block.
2 Calculate the inter-block flows in the x-direction using Darcy’s Law.
(See Equation 15.)
3 Calculate the total flow, Q, by adding the individual flows between two y-z
planes. (Any two planes will do, because the total ow is constant.)
4 Calculate the effective permeability for flow in the x-direction, using
the equation:
Q
k A P P
L
eff x
=
−, ( )1 2
µ (17)
Repeat the calculation for ow in the y- and z-directions, to obtain keff,y
and keff,z
.
Figure 17
The calculation of effective
permeability with no-ow
boundary conditions
Figure 16
No-ow boundary
conditions
8/9/2019 1__ResSimCh7
17/57
Institute of Petroleum Engineering, Heriot-Watt University 17
7 Permeability Upscaling
(b) Periodic Boundary ConditionsPeriodic boundary conditions are useful for calculating the effective permeabilities
of crossbeds or other sedimentary structures, when using the Geopseudo approach.
(See Section 1.3.1.) We assume that there are innitely repeated geological structures
in each direction. The use of periodic boundary conditions ensures that we have an
innitely repeated pattern of ows and pressure gradients. In the example shown in
Figure 18, there is a net pressure gradient in the x-direction. The blocks are numbered
i = 1, 2, ..nx in the x-direction, and j = 1, 2, .. nz in the z-direction.
x
z
P(i,0) = P(i,nz)
P(i,nz+1) = P(i,1)
P(0,j) = P(nx,j)+∆P P(nx+1,j) = P(1,j)-∆
A full permeability tensor may be calculated using periodic boundary conditions.
When a pressure gradient is applied in the x-direction there will be ow in the x-
direction, and there may also be ow in the z-direction due to internal heterogeneity.
These ows can be used to calculate the kxx
and kzx
tensor terms. Then a pressure
gradient is applied in the z-direction to obtain the kzz
and kxz
terms. (In 3D, a pressure
gradient should also be applied in the y-direction.)
(c) Linear Pressure Boundary ConditionsThese are similar to the no-ow boundary conditions, but the pressure is varied
linearly along the sides of the model (instead of specifying no ow).
P1 P2
P1 P2
P1 P2
Figure 18 Periodic boundary
conditions
Figure 19
Linear pressure boundary
conditions
8/9/2019 1__ResSimCh7
18/57
18
Some engineers use no-ow boundary conditions for calculating the effective
permeability in the x-direction, and linear pressure boundary conditions for calculatingthe effective permeability in the z-direction (vertical direction). This is sometimes
useful when upscaling sand/shale sequences.
The effective permeability calculated using linear pressure boundary conditions is
always greater to or equal to the effective permeability calculated using periodic
boundary conditions, which is always greater to or equal to the effective permeability
calculated using no-ow boundary conditions.
(d) Jackets or Skins
To reduce the effect of boundary conditions when calculating the effective permeability,
some engineers perform the simulations on a larger grid than necessary. The extragrid blocks round the edges are referred to as a “jacket” or “skin”. See Figure 20.
k eff
calculated for
this block
boundary conditions applied
to outer edges of model
2.5 ExamplesThese examples were calculated using the Geopseudo Toolkit software package
developed at Heriot-Watt. In addition to demonstrating the scale-up methods, they
show the limitations of nite difference simulation. The nite difference method is
only valid when ΔP/Δx → dP/dx, i.e. when the model is very nely discretised.
2.5.1 Scale-up of Stochastic DistributionsIn 2D, the effective permeability for a correlated random distribution should be the
geometric average.
(a) Log-normal distribution, poorly correlated, with σ = 1Since this is a stochastic distribution, every realisation will be slightly different.
We only show one realisation here.
Figure 20
Example of a ow jacket
round a model. In this case
the jacket is 4 cells thick
Figure 21
Contour plot of a
poorly-correlated stochastic
distribution.
(A 40 x 40 grid was used.)
8/9/2019 1__ResSimCh7
19/57
Institute of Petroleum Engineering, Heriot-Watt University 19
7 Permeability Upscaling
The geometric average of the permeabilities (calculated by computer using Equation
5) was 151 mD, so the analytical results were:
k =
151 0
0 151
The effective permeability was also calculated numerically using both no-ow
boundary conditions (NOF) and periodic boundary conditions (PBC). (The
Geopseudo Toolkit does not use linear pressure boundary conditions.) The
numerical result was:
k kNOF PBC
=
=
132 0
0 137
131 1
1 137
The off-diagonal terms are negligible compared with the diagonal ones, so no-ow
boundaries are adequate in this case. The numerical result has underestimated the
analytical one, because the grid is too coarse.
(b) Log-normal distribution, well-correlated, with σ = 1.
In this case, the geometric average was calculated to be 119 mD, so theanalytical results was:
k =
119 0
0 119
The numerical results were:
k kNOF PBC
=
=
−
−
114 0
0 120
122 2
2 118,
Here the numerical results are closer to the analytical ones, because we have a
well-correlated grid, so the nite difference method is more accurate. Again, the
off-diagonal terms are negligible.
Figure 22
Contour plot of a well-
correlated permeability
distribution.
(An 80 x 80 grid was used.)
8/9/2019 1__ResSimCh7
20/57
20
2.5.2 Scale-up of a Structured Model
Figure 23 shows a model of alternating layers of 20 mD and 100 mD, of equalthickness, and inclined at an angle of 20.56° (= arctan(0.5)) to the horizontal.
The analytical solution may be calculated using Equation 13.
kk k k k
k k k k
a h a h
a h a h
= + −( )−( ) +
=
cos sin sin cos
sin cos sin cos
. .
. .
2 2
2 2
54 7 10 7
10 7 38 7
θ θ θ θ
θ θ θ θ
The model was gridded coarsely (8 x 4 blocks), and then more nely (24 x 12 blocks)
(Figure 24), and the effective permeability was calculated numerically, using both
no-ow and periodic boundary conditions.
The 8 x 4 grid gives:
k kNOF PBC=
=
44 9 0 0
0 0 36 6
47 8 6 1
6 1 36 9
. .
. . ,
. .
. .
The 24 x 12 grid gives:
k kNOF PBC
=
=
48 4 0 0
0 0 38 6
52 4 9 2
9 2 38 1
. .
. .,
. .
. .
A simulation with a 96 x 48 grid gave:
k kNOF PBC=
=
50 0 0 0
0 0 39 2
54 1 10 3
10 3 38 5
. .
. . ,
. .
. .
Figure 23
Model of tilted layers
Figure 24
Left: 8 x 4 grid. Right: 24
x 12 grid
8/9/2019 1__ResSimCh7
21/57
Institute of Petroleum Engineering, Heriot-Watt University 21
7 Permeability Upscaling
The 8 x 4 grid is obviously much too coarse, and underestimates the effective
permeability. The 24 x 12 grid, although it looks much better, still under-estimatesthe analytical results. A 96 x 48 grid gives a result which is very close to the analytical
answer. However, people rarely perform simulations with such a rened grid, because
they are too time-consuming.
We can see from these results that using no-ow boundaries gives a reasonable
approximation to the diagonal terms of the tensor, but produces zero off-diagonal
terms.
(Note that there are other methods of discretising the pressure equations, which
will give a better answer with fewer gridblocks. However, these methods are not
described here, because they are more complicated.)
2.6 Simulation With Full Permeability TensorsHaving calculated full effective permeability tensors using periodic boundary
conditions, we need special software to handle them at the larger scale. Conventional
nite difference simulators use a 5-point scheme in 2D and a 7-point scheme in 3D,
and only take diagonal tensors - e.g. ECLIPSE 100 allows you to specify PERMX,
PERMY and PERMZ. Simulation with full tensors is more complicated and more
time-consuming, but some packages allow the user to input full tensors. For example,
ECLIPSE now allows users to input PERMXY, etc.
In 2D, a 9-point scheme is required to take account of crossow. This means that
there are 9 terms in each of the pressure equations, as illustrated in Equation 18.
a P a P a P a P a P
a P a P a P a P
i j i j i j i j i j
i j i j i j i j
1 2 1 3 1 4 1 5 1
6 1 1 7 1 1 8 1 1 9 1 1 0
, , , , ,
, , , , .
− − − −
− − − − =
− + − +
− − − + + − + + (18)
The coefcients, ai, in Equation 18 depend on the transmissibilities between the
blocks. There are several different methods of discretisation which give slightly
different results. To extend this to 3D, we need either a 19-point scheme or a 27-point
scheme. See Figure 25. (The 19-point scheme leaves out the 8 corners of the cube.)
Obviously, it takes longer to solve equations with a larger number of terms.
i-1,j-1 i,j-1 i+1,j-1
i-1,j i,j i+1,j
i-1,j+1 i,j+1 i+1,j+1
x
z
a) b)
Figure 25
a) A 9-point scheme for 2D.
b) A 27-point scheme for 3D
8/9/2019 1__ResSimCh7
22/57
22
The examples in Section 2.5 show that sometimes the off-diagonal elements of the
permeability tensor (kxy, etc) are negligible, so the limitations of using a 5-point(2D) or a 7-point (3D) scheme are not serious. Figure 26 shows examples of models
where the off-diagonal terms of the effective permeability tensor will be zero, or very
small, while Figure 27 shows examples of models where the off-diagonal terms may
be signicant.
a)
c)
b)
a)
c)
d)
b)
Figure 26
(a) innite parallel layers,
with one co-ordinate axis
aligned with the layers;
(b) models with an axis
of symmetry along a
co-ordinate axis (2D cross-
beds viewed along paleo-
current direction);
(c) a correlated random
distribution, where the
correlation length is very
much larger, or very much
smaller than the coarse
block size
Figure 27
(a) layers tilted with respect
to the co-ordinate axes;
(b) crossbeds or ripples;
(c) correlated random
distributions where the
correlation length is similar
to the coarse block size;
(d) models with faults or
fractures at an angle to the
co-ordinate axes.
8/9/2019 1__ResSimCh7
23/57
Institute of Petroleum Engineering, Heriot-Watt University 23
7 Permeability Upscaling
In layered systems, the size of the off-diagonal term may be gauged from Equation
13 in Section 2.3.3:
k xz
= (k a-k
h) sinθcosθ (19)
This is a maximum for θ = 45o, and increases as (ka-k
h) increases. Therefore, full
permeability tensors become more important as the angle of the lamination or bedding
increases, and as the permeability contrast increases.
2.7 Awkward CasesIn most cases, upscaling will introduce some error. However, when the permeability
contrasts are large, the errors may also be very large. Examples of “awkward
cases” for upscaling are:
• Low perm shales in a high perm sand
• High perm channels in a low net/gross region
• High perm fractures in a low perm matrix
• Low perm faults in a high perm region
For example, consider the sand/shale model in Figure 28. The total model is split
into 3 coarse blocks. However, there is a shale lying across each coarse block. If
the shale has zero permeability, each coarse block will have zero permeability in
the z-direction (vertical). However, uid can ow through the model vertically, as
shown. When upscaling models such as this one, care has to be taken to make the
coarse blocks large enough to include a representative elementary volume (REV), sothat the upscaled permeabilities are reasonable. (See King et al, 1998 for a practical
approach to upscaling in turbidite reservoirs.)
In awkward cases, non-uniform scale-up may be used to increase the accuracy of the
simulation. This means that the coarse blocks are not all the same size. Durlofsky
et al. (1997) assign smaller coarse grid cells to regions where there are high ow
rates.
For cases with high permeability streaks, the effective medium boundary conditions
(EMBCs) improve the accuracy (Wallstrom et al., 2000). See Section 3.5.
Figure 28
Sand/shale model
8/9/2019 1__ResSimCh7
24/57
24
2.8 Summary Of Single-Phase Scale-Up
• In some simple permeability distributions, the effective permeability may be
calculated analytically
- Continuous parallel layers
- Correlated random elds
• Permeability is actually a tensor quantity
- 4 terms in 2D
- 9 terms in 3D
• In most cases, it is more accurate to perform a numerical simulation
• Several types of boundary conditions are commonly used
- No-ow
- Periodic - Linear pressure gradient
• Some commercial simulators now take full permeability tensors
• Figure 26 shows examples of where diagonal tensors are adequate. Figure 27
shows examples of where full tensors will probably be more accurate.
• The most awkward cases to upscale are those with high permeability contrasts.
3. TWO-PHASE UPSCALING
3.1 IntroductionTwo-phase scale-up is more complicated than single-phase scale-up. In addition to
scaling up the absolute permeability, we need to consider:
• Relative permeabilities and capillary pressure
• Well locations
• Injection rates
(Well locations should also be considered in connection with single-phase ow.
However, in Section 2, we ignored sources and sinks, for simplicity.)
In general, we have one uid displacing another, so the ood is not in steady-state.
We have to simulate ne-scale oods in representative parts of the reservoir to obtain
pseudo relative permeabilities. Sometimes, a ood may approach a steady state overa small region, in which case, the phase permeabilities may be considered effective
permeabilities (rather than pseudos) provided the condition of widely separated
distance scales holds, as discussed in Section 1.
When upscaling, we should use the phase permeabilities:
k f = k
absk
rf(20)
Where “f” stands for uid - oil, gas or water. Generally, we assume that both the absolute
and the relative permeabilities are homogeneous and isotropic at the smallest scale
(diagonal tensor with kxx
= kzz
). As we upscale, we may require full tensors for the
phase permeabilities, k o and k
w. (However, full two-phase tensors are rarely used.)
To obtain effective (or pseudo) relative permeabilities, the absolute permeability must
be scaled-up separately. Then terms of the phase permeability tensor are divided by
8/9/2019 1__ResSimCh7
25/57
Institute of Petroleum Engineering, Heriot-Watt University 25
7 Permeability Upscaling
the appropriate term of the absolute permeability tensor. For example:
krf,xz
= kf,xz
/kabs,xz
3.2 Balance Of ForcesThe paths which uids take through the reservoir depend on the forces acting on
them. There are three types of forces:
• Viscous,
• Gravity,
• Capillary.
(The viscous force is caused by the injection of uid.)
A ternary diagram (Figure 29) may be used to show the balance of forces. We may
estimate the balance by comparing the pressure gradients due to each force acting
on the uids in a model. (Actually the force balance depends on the location in the
model, and varies with time, but for choosing a scale-up method, it is sufcient to
make a rough estimate of the average force balance.)
viscous dominant
gravity negligible
forces equal
capillary
gravityviscous
If the injection rate is very high the viscous pressure drop will be large compared with
capillary pressure gradients. If, in addition, the densities of the uids are similar,
the ood may be considered to be viscous-dominated.
If the viscous and gravity forces are small, the ood will be capillary-dominated.
This is most likely to occur at the small scale, particularly where we have ne-scale
lamination with large permeability contrasts over the scales of mm-cm. If
the injection rate is very low, there may be time for the uids to come to
capillary equilibrium over small distances (< 1 m). This means that the ood
is approximately in a steady-state. We can take advantage of this when we are
upscaling. See Section 3.3.1.
Alternatively, if the capillary forces are negligible, and the injection rate is very
low, the ood may be gravity-dominated, and we may have vertical equilibrium. In
this case, the uids become segregated immediately during the ood. The verticalequilibrium assumption (VE) may be used to scale-up in the z-direction, reducing
the number of dimensions in the simulation from 3 to 2, or from 2 to 1.
Figure 29
Ternary diagram showing
the balance of forces
8/9/2019 1__ResSimCh7
26/57
26
3.3 Steady-State Methods
These are actually quasi steady-state methods. We assume that within a short intervalof time the zone of interest is in a steady-state, but we allow the uid saturation to
change gradually, so that a full range of saturation is obtained. At steady-state, the
water saturation does not change with time, i.e. ∂Sw / ∂t = 0, so the continuity equation
becomes:
∇ • =uf 0, (22)
where u is Darcy velocity, and f is uid. From Darcy’s law:
∇ • • ∇( ) =k Pf f
0. (23)
3.3.1 Capillary-EquilibriumAssume that the injection rate is very low, gravity forces are negligible, and
that the uids have come into capillary equilibrium over a small distance (< 1
m). This means that the saturation distribution is determined by the capillary
pressure curves.
The method is as follows:
1. Choose a Pc level.
2. Determine the water saturations, and then the rel perms.
3. Calculate the pore volume-weighted average water saturation
4. Calculate the phase permeabilities: ko = k
absk
ro, k
w = k
absk
rw.
5. Calculate the effective water phase permeability.
6. Calculate the effective oil phase permeability.
7. Calculate the effective relative permeabilities, krw
= kw/k
abs, etc.
8. Repeat the process with another value of Pc.
Steps 5 and 6 may be carried out analytically or numerically, depending on the
distribution. The upscaled phase permeability may be a full or a diagonal tensor.
Example 5
Consider a model with two layers of equal thickness, as shown in Figure 30. The
absolute permeabilities are 100 mD and 20 mD. Assume that the porosity in each layer
is equal to 0.2. The rel perm and Pc curves for each layer are shown in Figure 31.
8/9/2019 1__ResSimCh7
27/57
Institute of Petroleum Engineering, Heriot-Watt University 27
7 Permeability Upscaling
Using the arithmetic and harmonic averages (Section 2), the effective permeability
is:
k =
60 0 0 0
0 0 33 3
. .
. .
0
2
4
6
8
10
12
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Water Saturation
C a p P r e s s u r e
R e l P e r m
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Water Saturation
lo
hi
hi
hi
lo
lo
Suppose we choose a capillary pressure of Pc = 0.45.
In the high perm layer: Sw = 0.34, k
rw = 0.0013, k
w = 0.13, k
ro = 0.5, k
o = 50.
In the low perm layer: Sw = 0.44, k
rw = 0.0016, k
w = 0.032, k
ro = 0.48, k
o = 9.6.
Since the layers are of equal width, the average saturation is Sw = 0.39.
Using the arithmetic and harmonic averages from Section 2:
k k
k k
w rw
o ro
=
=
=
=
0 081 0 000
0 000 0 051
0 00135 0 00000
0 00000 0 00154
29 8 0 0
0 0 16 1
0 50 0 00
0 00 0 48
. .
. .
. .
. .
. .
. .
. .
. .
Note that the kv/kh ratio ( = kzz
/kxx
) is different for oil and water. It is 0.63 for water,
but only 0.54 for oil.
Effective relative permeability curves may be derived by repeating this calculationfor a range of capillary pressure values.
Figure 30
Model with horizontal
layers
Figure 31
Rel perm and Pc curves for
each lamina
8/9/2019 1__ResSimCh7
28/57
28
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Water Saturation
k rox
k roz
k rwz
k rwx
R e l P e r m
The capillary-equilibrium method is useful as a quick method for scaling up from
the lamina to the bed scale. However, it is only valid in cases where the ow rate
is very low.
3.3.2 Viscous-Dominated Steady-StateWhen the capillary pressure is negligible, the viscous-dominated steady-state method
may be used. It may seem strange to use a steady-state method when the ood is
certainly not in a steady state, because there will be a Buckley-Leverett shock front
moving through the model. However, the method works surprisingly well. In thiscase, it is assumed that the fractional ow of water, f
w = q
w/q
t, is constant throughout
the model. The method proceeds in a similar manner to the capillary-equilibrium
method, except different levels for f w are chosen.
Recently a new steady-state method for intermediate viscous/capillary ratios has been
developed (Stephen and Pickup, 2000).
3.4 Dynamic MethodsFor dynamic (or non steady-state methods), we need to perform a two-phase ow
simulation on a ne grid. There are basically two types of non steady-state scale-
up:
(a) Weighted Pressure Methods
There are various ways of averaging the pressure, but we shall concentrate on the
Kyte and Berry (1975) method.
(b) Total Mobility Methods
In these methods, we scale-up the total mobility, and use the average fractional ow
to calculate the pseudo relative permeabilities.
The total mobility is:
λ λ λ µ µ
t o w
ro
o
rw
w
k k= + = + . (24)
Figure 32
Effective rel perm curves
in the horizontal and
vertical directions (x- and z-directions)
8/9/2019 1__ResSimCh7
29/57
Institute of Petroleum Engineering, Heriot-Watt University 29
7 Permeability Upscaling
The fractional ow is the ow of water divided by the total ow:
f q
q q
q
qw
w
o w
w
t
=
+
= (25)
Again there are a number of variations of this method, but we concentrate on the
Stone (1991) method.
The “traditional” upscaling methods assume that permeability is a diagonal tensor.
More recently there have been attempts to calculate full two-phase tensors. The
Pickup and Sorbie (1996) method is described in the last subsection.
3.4.1 The Kyte and Berry MethodA simple version of the Kyte and Berry (1975) method is presented here, using the
grid shown in Figure 33.
i=1 2 3 4 5 6 7 8 9 10
j=1
2
3
4
5 ∆z
∆x
∆X ∆Z
The diagram shows two coarse grid blocks, each of which is made up of 5 x 5 neblocks. The equations below show how to calculate the pseudo rel perms and capillary
pressure for the left coarse block.
The rst step is to perform a ne-scale, two-phase simulation (in ECLIPSE), saving
the pressures and inter-block ows at specied intervals of time. The method proceeds
as follows:
1. Calculate the effective absolute permeability in the area shown in Figure 34.
Figure 33
Model used for describing
the Kyte and Berry Method.
The thickness of the model
in Δ y.
8/9/2019 1__ResSimCh7
30/57
30
Kyte and Berry approximate the effective permeability using the arithmetic average
in each column, and then taking the harmonic average of the columns. The area
between the two coarse blocks is used, for reasons explained below.
∆
∆
∆
∆ ∆
z k
x
z k
x x
i i
i
ij ij
ij i j j
=+( )+=
∑11
5
2, (26)
where Δxij and Δz
ij are the sizes of the ne blocks. (In this case, all the blocks are
of equal size.)
k
X
Z x
z k
I
i
i ii
=
=
∑
∆
∆ ∆
∆3
7 (27)
where Δx and Δz are the dimensions of the coarse blocks, andk I
is the requiredeffective absolute permeability.
The pseudos are then calculated, at certain times during the simulation. (These are
the times at which the restart les are written in the Eclipse simulation.)
2. Calculate the average water saturation:
S
S x z
x z
w
w ij ij ij ij
i j
ij ij ij i j
= ==
==
∑∑
∑∑
, φ
φ
∆ ∆
∆ ∆
1
5
1
5
1
5
1
5
(28)
where φij is the porosity.
3. Calculate the total ow of oil and water out of the left coarse block (Figure 35).
q q f f j j
==
∑ 51
5
, ,
(29)
where qf5,j
is the ow of uid “f” from ne block number (5,j).
Figure 34
The area used for
calculating the effective
absolute permeability
8/9/2019 1__ResSimCh7
31/57
Institute of Petroleum Engineering, Heriot-Watt University 31
7 Permeability Upscaling
Figure 35
Calculation of the total ow
4. Calculate the average phase pressures in the central column of each coarse block.
In this example, we use the ne blocks in columns 3 and 8, the shaded areas in
Figure 36.
I II
In the Kyte and Berry method, the pressures are weighted by the phase permeabilities
times the height of the cells (which in this case are all the same size). This is so thatmore weight is given to regions where there is greater ow. However, there is no
scientic justication for using this weighting. In the rst coarse block (numbered,
I), the average pressure is:
P
k k z P g D D
k k z
fI
j rf j f j f j
j
j rf j
j
=
− −( )=
=
∑
∑
3 3 3 3 3
1
5
3 3 3
1
5
∆
∆
ρ ( )
(30)
where D3j is the depth of cell (3,j) andD is the average depth of coarse cell I. The
term gρf (D3j -D) is to normalise the pressure to the grid block centre. The averagepressure for coarse block II is calculated in the same manner, but using column 8
instead of column 3. The pressure difference is then calculated as:
∆P P Pf fI fII= − (31)
5. The pseudo rel perms are then calculated using Darcy’s law. Firstly, calculate the
pseudo potential difference. (Potential is dened as Φ = P-ρgz, so that the ow rateis proportional to ∇Φ.)
∆Φ ∆ ∆f f f P g D= − ρ (32)
where ΔD is the depth difference between the two coarse grid centres. Then:
Figure 36
Averaging the phase
pressures
8/9/2019 1__ResSimCh7
32/57
32
k
q X
Zkrf f f
I I=
−µ ∆
∆ ∆Φ (33)
6. Calculate the pseudo capillary pressure using:
P P Pc oI wI= − (34)
The Eclipse PSEUDO package can be used for calculating Kyte and Berry pseu-
dos.
3.4.2 Discussion on Numerical DispersionOne advantage of pseudo-isation methods, such as that of Kyte and Berry is that they
can take account of numerical dispersion. When a simulation is carried out using a
larger grid, the front between the oil and water becomes more spread out. However,
the Kyte and Berry method counteracts this effect by calculating the ows on the
down-stream side of the coarse block, instead of the middle. This is illustrated by
a simple example of a homogeneous grid. Figure 37 shows an example of input rel
perm curves (“rock” curves).
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Water Saturation
R e l
P e r m
If the water saturation is Sw = 0.5, the rock curves show that there is a small amount
of oil and water owing. However, when the average saturation, is 0.5 in the coarse
block, the distribution could be as shown in Figure 38.
coarse
block water
oil
Since the water has reached only half way across the coarse block, there should be
no water owing out of the right side. The Kyte and Berry method calculates the
pseudo rel perms using the ow on the down stream side of the coarse block, to
Figure 37
Example of “rock” curves
Figure 38
Example of the water
saturation in a coarse block
8/9/2019 1__ResSimCh7
33/57
Institute of Petroleum Engineering, Heriot-Watt University 33
7 Permeability Upscaling
prevent water breaking through too soon. The pseudo water rel perm curve is moved
to the right, relative to the rock curves, as shown in Figure 39.
rock
curves
pseudo kro
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Water Saturation
pseudo krw
R e l
P e r m
3.4.3 Disadvantages of the Kyte and Berry MethodThere are certain problems with the Kyte and Berry method.
• Negative rel perms are produced, if ∆Φ has the same sign as qf .
• Innite rel perms occur if ∆Φ f is zero.
• The method of averaging the pressures, using rel perm as a weighting function,
may cause errors when the fluids are separated due to gravity. For floods whichare gravity-dominated, the TW method works better (Section 3.5).
• Non-zero pseudo capillary pressure may be produced, even if there is no capillary
pressure in the ne-scale simulation. This is because a different weighting is used
for calculating the average pressure in each phase.
• The capillary pressure may be different in different directions, because only the
central column is used for averaging the pressures.
Because of the rst two disadvantages, i.e. negative, or innite rel perms, pseudos
obtained from packages like the PSEUDO must be vetted before using at the coarse
scale. Often “odd” values of relative permeability are set to zero.
The last two disadvantages may be overcome by using a pore volume weighted
average of the pressures over the entire coarse block. This method is also available
in ECLIPSE. The last two advantages may be overcome by using a pore volume
weighted average of the pressures over the entire coarse block. In the ECLIPSE
PSEUDO package, the capillary pressure is now calculated in this way. There is also
a Pore Volume Weighted (PVW) method, in which the rel perms are also calculated
from pressures which have been averaged using pore volume weighting, rather than
rel perm weighting.
Good reviews of various methods for calculating pseudos are presented in Barker
and Dupouy (1996) and Barker and Thibeau (1997).
Example 6 - A Simple Kyte & Berry Calculation
Figure 40 shows an 11 x 3 ne grid. The rst and last columns are for injection and
Figure 39
Example pseudo rel perm
curves
8/9/2019 1__ResSimCh7
34/57
34
production. The central 9 columns make up 3 coarse blocks, each of which is 3 x
3 ne blocks. The calculations below show how to derive the pseudo rel perms forthe central coarse block, at one specied time. We shall ignore gravity in this simple
simulation, so we have effectively a 1D model.
Assume that the blocks are homogeneous, with porosity = 0.2, and permeability kabs
= 100 mD = 0.1 D. Also assume that the viscosity of water is μw = 1 cP, and that
of oil is μo = 5 cP. The “rock” rel perms and capillary pressures were supplied by
tables. You only need to know kr and P
c for a few water saturation values, and these
are supplied below.
After simulating for 200 time steps, suppose the water saturation distribution is as
follows:
I II III 0 1 2 3 4 5 6 7 8 9 10
(inj) (prod)
Sw 0.59 0.56 0.54 0.53 0.51 0.47 0.41 0.32 0.28 0.27 0.27
1. Calculate the average water saturation in coarse block II:
Sw
= 0.46.
2. Calculate the total ow of each phase out of the central coarse block. The pressures,
which were calculated using a simulator, are as follows:
I II III
0 1 2 3 4 5 6 7 8 9 10
(inj) (prod)
Po 11.12 11.03 10.93 10.81 10.67 10.51 10.35 10.21 10.14 10.07 10.0 (atms)
Pc 0.15 0.16 0.17 0.18 0.19 0.21 0.27 0.53 2.72 4.25 4.25
(atms)
Pw 10.97 10.87 10.76 10.63 10.48 10.30 10.08 9.68 7.42 6.82 5.75
(atms)
The phase perms (in Darcies) are shown for each block, below. These were calculated
using the water saturations above, the rel perm tables, and the absolute perm, which
is 0.1 D.
Figure 40
Grid used for simple Kyte
and Berry example
8/9/2019 1__ResSimCh7
35/57
Institute of Petroleum Engineering, Heriot-Watt University 35
7 Permeability Upscaling
I II III
0 1 2 3 4 5 6 7 8 9 10 (inj) (prod) kw 0.012 0.009 0.008 0.006 0.005 0.003 0.001 0.000 0.000 0.000 0.000
(D)
ko 0.000 0.000 0.004 0.006 0.008 0.013 0.028 0.059 0.080 0.085 0.085
(D)
Use Darcy’s law to calculate the ows out coarse block II, i.e. ows out of column
6.
q k P
x
q q qf f
f
f f f
i
f = = = ×=
∑µ
∆
∆
,1
3
3
ΔPf is the pressure drop between ne blocks 6 and 7 (shown using dark shading in the
table), and Δx is the distance between the centres of these blocks, and so is 1 cm.
I II III
0 1 2 3 4 5 6 7 8 9 10
(inj) (prod) qw
0.0012
qo 0.0023
The units forq are cc/s.
3. Calculate the average pressures in the central columns of coarse blocks II and III.
Since the pressures are constant within each column, the averaging is trivial. The
pressures to use for calculating the pseudos are shown using the light shading in the
pressure table above.
The pressure drop between the centres of blocks II and III is therefore:
∆Pw = 10.30 - 7.42 = 2.88 atms.
∆Po = 10.51 - 10.14 = 0.37 atms.
4. Calculate the pseudo phase permeability using Darcy’s law again.
k
q x
A P f
f f =
µ ∆
∆
Here Δx is the distance between the centres of the coarse blocks II and III, and is 3cm, and A is the cross-sectional area of the coarse blocks, which is 3 cm2.
8/9/2019 1__ResSimCh7
36/57
36
I II III
0 1 2 3 4 5 6 7 8 9 10 (inj) (prod) kw
0.00042
(D)
ko 0.031
(D)
5. Calculate the rel perms by dividing by the absolute permeability (0.1 D).
krw
= 0.004 k
ro=
0.311
6. Finally calculate the average capillary pressure, which is the difference betweenthe average oil and average water pressure in block II.
Pc=P
o- P
w= 10.51 - 10.36 = 0.21 atms.
This calculation must be repeated at a number of different times in order to calculate
the pseudo rel perms and capillary pressures for a range of water saturations.
3.4.4 Example of K&B Pseudos for a Ripple ModelFigure 41 shows a 2D model of small-scale ripples. Pseudo rel perms were calculated
for a ripple using four different ow rates. There is a factor of 10 between each
rate. Rate 1 is the fastest. Figure 42 shows the resulting pseudos (from Pickup andStephen, 2000).
200 mD
10 mD
1 cm,
18 cells
3 cm, 54 cells
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Water Saturation
rate 1rate 2
rate 3
rate 4
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Water Saturation
rate 1rate 2
rate 3
rate 4
R e l a t i v e P e r m e a b i l i t y
R e l a t i v e P e r m e a b i l i t y
Figure 41
A model of ripples (based
on the Ardross Cliff, near
St. Monance, Fife)
Figure 42
Pseudo relative
permeabilities for the ripple
model, for different ow
rates (1: very fast ... 4: very
slow)
8/9/2019 1__ResSimCh7
37/57
Institute of Petroleum Engineering, Heriot-Watt University 37
7 Permeability Upscaling
Note the following:
1. At high rates, the pseudos are shifted to the right. This is to compensate for numeri-
cal dispersion.
2. At very low ow rates (rate 4), the ood is capillary-dominated, and the oil is
trapped. The pseudo oil rel perm goes to zero at Sw = 0.46.
3.4.5 The Stone MethodIn this method, the problem of how to average the pressure is avoided by upscaling
the total mobility, and using the fractional ow to calculate the pseudo rel perms.
As with the Kyte and Berry method, a ne-scale two-phase ow simulation is per-formed, and the inter-block ows are saved at specied times. We use the same
example as before to illustrate the method.
i=1 2 3 4 5 6 7 8 9 10
j=1
2
3
4
5 ∆z
∆x
At each time interval:
1. Calculate the average water saturation, as before:
S
S x z
x z
w
w ij ij ij ij
i j
ij ij iji j
= ==
==
∑∑
∑∑
, φ
φ
∆ ∆
∆ ∆
1
5
1
5
1
5
1
5 (35)
where φij is the porosity of each block.
2. Calculate the total (relative) mobilities in column 5.
λ λ λ µ µ
t o w
ro
o
rw
w
k k= + = + . (36)
3. Average the total mobilities as follows:
Figure 43
The Stone Method. The
thickness of the model is Δ y
and is set to one here, for
convenience.
8/9/2019 1__ResSimCh7
38/57
38
λ
λ
t
x j t j
j
x j
j
T
T
= =
=
∑
∑
5 5
1
5
5
1
5 , (37)
where Tx5j
is the x-direction transmissibility in block (5,j). Transmissibility is dened
as:
T k z
xx
x=
∆
∆ (38)
(It is assumed here that absolute permeability is directional, but relative permeabilityis isotropic. Also, this is a 2D model, and we are assuming that Δy = 1.)
4. Calculate the total ow out of the coarse block, for each phase:
q qf f j
j
==
∑ 51
5
(39)
Then calculate the total ow:
q q qt o w
= + (40)
The average fractional ow of water is then calculated:
f q
qw
w
t
= (41)
5. Finally, calculate the pseudos as follows:
k f k f rw w w t ro o w t= = −( )µ λ µ λ , 1 (42)
The Eclipse PSEUDO package has an option for the Stone Method.
3.4.6 Disadvantages of the Stone MethodAlthough the Stone Method avoids problems arising due to pressure averaging, there
are other problems:
• Capillary pressure and gravity are ignored.
• The calculation of the average total mobility is inadequate when there is a
signicant variation in the total mobility.
• This method performs badly when the model is tilted.
3.5 Additional Upscaling MethodsSince the aim of two-phase upscaling is to derive rel perms, which produce the sameresults for the course-scale simulation as the ne-scale simulation, some engineers
calculate the pseudo rel perms using history matching methods for representative
8/9/2019 1__ResSimCh7
39/57
Institute of Petroleum Engineering, Heriot-Watt University 39
7 Permeability Upscaling
sections of the eld. In this case, the pseudo rel perms are assumed to have a
particular shape (e.g. a power law), and the parameters are adjusted to produce thehistory match. See, for example, Johnson et al., 1982. This approach is sometimes
referred to as the regression method.
The following methods have been developed recently to overcome some of the
drawbacks of the Kyte and Berry and Stone methods. They are not available yet
in commercial simulators. This is not an exhaustive list - more methods are being
developed all the time.
(a) The Total Mobility Method
This method was developed by Christie et al (1995). It is similar to the Stone method,
but it handles gravity better.
(b) The Hewett and Archer (HA) Method
In this method, described by Hewett and Archer (1997), the pressures are not averaged,
the pressure in centre of the coarse grid block is used directly. This method can
handle both gravity and capillary pressure.
(c) The Transmissibility-Weighted (TW) Method
This method was developed by Nasir Darman at Heriot-Watt (Darman et al, 1999).
It is similar to the Kyte and Berry method, except transmissibility weighting is used
when calculating the average pressure. The method works better than the Kyte and
Berry method in cases where gravity effects are signicant (e.g. a gas ood) and it
also works better than the Stone method in tilted reservoirs.
Recent research (Darman, Pickup and Sorbie, 2001) shows that, although the HA
method and the TW method produce different pseudo rel perms, their effects are
equivalent, because they produce the same fractional ows.
(d) Upscaling with Effective Medium Boundary Conditions (EMBCs)
Usually in scale-up, a section of a whole model is removed and a ow simulation is
carried out on this section using simple BCs, such as constant pressure at the edges
and no-ow at the sides. This allows a large ux through any high perm streaks
- much higher than you get when the simulation is carried out on the whole model.
To overcome this problem, Wallstrom et al. (2000) use effective medium boundaryconditions (EMBCs). According to effective medium theory, if we have a circular
disk of high permeability, k, embedded in an innite medium of perm ko, then the
ux (Darcy velocity) owing inside the disk is:
u k
k k du
o
o=
+ −
( )1
(43)
where d is the dimension (2 for 2D, or 3 for 3D), and uo is the ow velocity for the
innite medium, in the absence of the disk.
In the method of Wallstrom et al (2000), Effective Medium Theory was used to set up
ow boundary conditions for the ne-scale simulations. For each coarse block, the
ux entering each ne-grid cell at the inlet face, or exiting from each ne-grid cell
8/9/2019 1__ResSimCh7
40/57
40
on the outlet face, is calculated using the theory. The method used for calculating
the pseudos was a variation of the Stone method.
(e) Two-Phase Tensor Upscaling
Using any of the above methods, you can get directional pseudo rel perms, but not full
tensors. It is possible to carry out full two-phase tensor upscaling (Pickup and Sorbie,
1996), and to simulate two-phase ow using full tensor permeabilities. However,
not many people use this method yet. It may be important to use two-phase tensor
methods in reservoir with large dip and in fractured or faulted reservoirs.
3.6 Summary Of Two-Phase Scale-UpScaling up two-phase ow is more complicated than single-phase scale-up. The
movement of uids through a reservoir depends on the balance of forces - viscous,
capillary and gravity. Calculations may be simplied if one of these forces is dominant.
For example, if the injection rate is very low, and if gravitational forces are negligible,
then the uids may come into capillary equilibrium over small distances. In this case
we have a quasi steady state, and we can calculate proper effective permeabilities for
each phase (assuming that we have widely separated distances scales).
Usually, the uids are not in a steady state, and we can only calculate pseudo functions,
rather than effective properties. There are many different methods for scaling up non
steady-state two-phase ow. They tend to fall into two categories:
• Pressure Averaging Methods,
• Total Mobility Methods.
These methods are not rigorous. They are numerical “recipes” which aim to reproduce
the ne-scale hydrocarbon recovery on a coarse-scale grid. Therefore, there is no
method which is guaranteed to give the “correct” answer.
Also, pseudos are case dependent. They depend on the injection rate and on the
direction of ow (i.e. they are history-dependent). Figure 42 showed how pseudos
vary for different ow rates.
4. PRACTICAL APPLICATIONS
4.1 IntroductionHaving described the theory behind various upscaling methods, we now outline
how to apply these methods in practice. We also present a few extra topics, such as
scale-up in the vicinity of wells, grouping of pseudos.
The choice of which method to use in often depends, in practice, on how much time
(money) and data are available. Figure 44 illustrates the range of methods from the
point of view of time taken to implement them.
8/9/2019 1__ResSimCh7
41/57
Institute of Petroleum Engineering, Heriot-Watt University 41
7 Permeability Upscaling
Upscale absolute
perms only
Simple
averaging
Numerical
simulation
Two-phase
upscaling
Steady
state
Numerical
simulation
Easy
Difficult
We should also examine the balance of forces (Section 3.2) to assess which method to
use. Figure 45 shows which methods are appropriate for different force balances. On
the whole, dynamic methods are more accurate, if you can get the boundary conditions
correct. However, if one force is dominant, you may be able to take shortcuts.
1
3 2
4
5
6
Capillary-
dominated
Gravity-
dominated
Viscous-
dominated
7
Key:
1. Capillary-dominated - use capillary equilibrium method.
2. Gravity dominated - Simulation may be in vertical equilibrium - uids
segregate under gravity. Eclipse has a vertical equilibrium option, which may
be used to convert a 3D model into a 2D one (or a 2D to a 1D).
3. Viscous-dominated - use viscous-dominated steady-state, or a dynamic
method like Kyte and Berry.
4. Capillary and gravity forces competing, negligible viscous force - can use
capillary-gravity equilibrium.
5. Both viscous and gravity forces important - the TW method works better
here than Kyte and Berry.
6. Viscous and capillary forces competing - could use Kyte and Berry here.
A new intermediate steady-state method has been developed (Stephen and
Pickup, 2000).7. All forces competing - probably should not take short-cuts in this region -
use a dynamic method, such as Kyte and Berry.
Figure 44
Upscaling methods, as a
function of ease of use
Figure 45
Balance of forces triangle
8/9/2019 1__ResSimCh7
42/57
42
4.2 Upscaling Stages
Different methods may be appropriate for different stages of upscaling.
4.2.1 Stage 1 - Geopseudo UpscalingThe rst stage of upscaling, which tends not to be given careful consideration, is
to calculate appropriate relative permeabilities and capillary pressure for the grid
cells of the stochastic geological model (with cells sizes of approximately 50 m in
the horizontal and 0.5 m in the vertical). As mentioned in section 1, the Geopseudo
Method (Corbett et al., 1992) may be used to upscale from the lamina-scale upwards
(Figure 46). Models of sedimentary structures are created, and simulations are carried
out to upscale from the lamina-scale to the bed-scale and, if necessary another step
is used to upscale to a set of beds. Corbett et al. (1992) applied the Kyte and Berry
Method to upscale hummocky cross-stratication in several stages. However, thesemodels were very simple and only in 2D. In more complex models, it is probably not
feasible to use the Kyte and Berry Method. One approach which has been adopted to
save time is the use of steady-state methods (Pickup et al. 2000) - either the capillary
equilibrium method or viscous-dominated steady-state.
Low Perm High Perm
Individual Rel. Perm Curves
Pseudo Rel. Perm CurvesEffective Perm
Example of Geoseudo Upscaling
These two examples are from Pickup et al., 2000. The rst case is a synthetic model
of a uvio-aeolian eld, and the second case is a model of a tidal deltaic reservoir.
Figure 46
Example of Geopseudo
upscaling
Figure 47
Scale-up stages used in
the uvio-aeolian example
- from the lamina-scale
models to the stochastic
geological model.
8/9/2019 1__ResSimCh7
43/57
Institute of Petroleum Engineering, Heriot-Watt University 43
7 Permeability Upscaling
Geopseudo upscaling maybe time-consuming and expensive. There is no point in
upscaling from the smallest scales, unless cores are available for the eld. Cores
must be studied to identify the sedimentary structures present, and probe permeability
measurements should be taken to populate the small-scale models. Additionally, and
very importantly reliable SCAL data is also required.
To decide whether or not it is worthwhile carrying out Geopseudo upscaling, you
need to examine the large-scale structure of the reservoir rst. If uid can ow
through massive, high permeability sandstone to the wells, it will avoid lower quality
rock which may contain laminae. In this case, Geopseudo upscaling is probably
unnecessary.
The following list of guidelines from Ringrose et al. (1999) is useful for assessing
whether Geopseudo upscaling may be necessary:
(1) Are immiscible uids owing?
(2) Are signicant small-scale heterogeneities present? Specically:
• Is the permeability contrast greater than 5:1?
• Is the layer thickness less than 20 cm?
• Is the mean permeability less than 500 mD?
(3) What is the large-scale structure of the reservoir? In many cases, large-scale
connectivity may be the dominant issue, in which case, small-scale structure may
have to be ignored. Use the Weber and van Geuns (1990) classication to describe
the large-scale structures:
• Layer cake reservoirs - small-scale structure will usually have primary
importance. • Jigsaw puzzle reservoirs - small-scale structure may be important.
• Labyrinth reservoirs - small-scale structure will usually be of secondary
importance.
Figure 48
The three models sizes used
in the tidal deltaic model:
a) a heterolithic sand
model (1m x 1m x 0.3 m),
b) the stochastic geological
(Storm), and c) the coarse-
scale model - 6 km x 2.2 km
x 0.25m.
8/9/2019 1__ResSimCh7
44/57
44
The Geopseudo method may require two stages as follows:
1. Lamina-scale (mm or less) → bed-scale (5 m x 5 m x 0.33 m)Use simple models of representative beds. In a waterood, capillary forces will
probably dominate here, so use the capillary equilibrium method. This method is
relatively fast, and is suitable for grids of up to a million blocks. (Dynamic methods,
such as the Kyte and Berry Method are valid here, but they are time consuming.)
2. Bed-scale (5 m x 5 m x 0.33 m) → geological model scale (50 m x 50 m x 1 m)This stage of upscaling takes account of the variation of beds within a genetic unit
- e.g. trends. Steady-state methods may be used for convenience. Since the model is
no longer capillary dominated at this scale, viscous-dominated upscaling may be used
(or intermediate steady-state upscaling - Stephen and Pickup, 2000). Again, dynamicmethods are more accurate here, but are time-consuming. Some people assume that,
once they have obtained the effective rel perms of a single bed, this can be applied
directly to the geological model - i.e. this level of upscaling is not necessary.
4.2.2 Stage 2 - “Conventional” UpscalingStage 2 upscaling means reducing the number of grid cells in the stochastic geological
model so that full-eld simulation may be carried out. Depending on the amount of
time available, engineers may only upscale absolute permeability. What’s more, they
may only carry out simple averaging, rather than use one of the numerical simulation
methods (Figure 44).
At the scale of the geological model, capillary forces are no longer important, but
gravity effects may be signicant. Refer back to Figure 45 for methods to use where
gravity is important.
If dynamic methods, such as that of Kyte and Berry, are to be used for upscaling from
a 3D geological model to a full-eld model, representative sections of the full model
must be simulated at the ne-scale (i.e. the geological model scale, in this case). We
assume here that the geological model is too large for two-phase ow simulation,
otherwise upscaling would not be necessary. However, the boundary conditions are
difcult to specify, since they depend on the location of the wells.
4.3 Recent ApproachesRecently, some methods have been developed to try to overcome the problem of
using appropriate boundary conditions for the ne-grid simulation. These methods
mostly involve streamline simulation.
In streamline simulation (e.g. Blunt et al., 1996), the pressure equations are solved
once (using total mobilities), and streamlines i.e. the paths along which uids will
move, are calculated. The uids are advanced along these paths to obtain the saturation
distribution at a later time. From time to time, the pressure equation is recalculated
to take account of varying mobilities as the ood progresses. This method is much
faster than conventional simulation, because the pressure equations are only solved
a few times during the simulation, instead of perhaps thousands of times. Althoughthis method is less accurate than conventional simulation, it is feasible to carry out
streamline simulations on the grids containing millions of blocks.
8/9/2019 1__ResSimCh7
45/57
Institute of Petroleum Engineering, Heriot-Watt University 45
7 Permeability Upscaling
Figure 49
Nested grid simulation
Streamlines may also be used along with upscaling methods - for example to speedup a ne-scale two-phase simulation.
4.3.1 Nested Grid SimulationThis method was developed recently by Gautier et al. (1999). The pressure equations
are solved on the coarse grid, and these are used to calculate the ows between each
coarse block (Figure 49). The pressure equations are then solved on the ne grid,
within each coarse block, using ow boundary conditions. In this way the velocity
eld is computed on the ne-grid. Streamline methods are used to move the saturations
forward in time. The pressure eld is updated periodically.
A number of other researchers have proposed similar methods, e.g. Guedes andSchiozer (2001).
Coarse-grid simulation
Fine-grid simulation
with flow BCs
4.4 SPE Upscaling StudyIn the recent SPE Comparative Solution Project, which focussed on upscaling
techniques (Christie and Blunt, 2001), two models were presented: a 2D model
with gas displacing oil, and a 3D waterood. We summarise the results of the 3D
waterood here, as