Вы находитесь на странице: 1из 11

Simulation of Two-Phase Flow in Reservoir

Rocks Using a Lattice Boltzmann Method


Thomas Ramstad, SPE, Pl-Eric ren, SPE, and Stig Bakke, Numerical Rocks

Summary
We present results from simulations of two-phase flow directly on
digitized rock-microstructure images of porous media using a lattice
Boltzmann (LB) method. The implemented method is performed on
a D3Q19 lattice with fluid/fluid and fluid/solid interaction rules to
handle interfacial tension and wetting properties. We demonstrate
that the model accurately reproduces capillary and wetting effects
in pores with a noncircular shape. The model is applied to study
viscous coupling effects for two-phase concurrent annular flow in
circular tubes. Simulated relative permeabilities for this case agree
with analytical predictions and show that the nonwetting-phase
relative permeability might greatly exceed unity when the wetting
phase is less viscous than the nonwetting phase.
Two-phase LB simulations are performed on microstructure
images derived from X-ray microtomography and process-based
reconstructions of Bentheimer sandstone. By imposing a flow
regulator to control the capillary number of the flow, the LB model
can closely mimic typical experimental setups, such as centrifuge
capillary pressure and unsteady- and steady-state relative permeability measurements. Computed drainage capillary pressure curves
are found to be in excellent agreement with experimental data.
Simulated steady-state relative permeabilities at typical capillary
numbers in the vicinity of 105 are in fair agreement with measured
data. The simulations accurately reproduce the wetting-phase relative permeability but tend to underpredict the nonwetting-phase
relative permeability at high wetting-phase saturations. We explain
this by pointing to percolation threshold effects of the samples. For
higher capillary numbers, we correctly observe increased relative
permeability for the nonwetting phase caused by mobilization
and flow of trapped fluid. It is concluded that the LB model is a
powerful and promising tool for deriving physically meaningful
constitutive relations directly from rock-microstructure images.
Introduction
Modeling of multiphase flow in hydrocarbon reservoirs requires
accurate estimates of constitutive relations, such as capillary pressure and relative permeability curves. These relations are typically
measured experimentally. Experimental data, however, are often
scarce because the measurements can be difficult and time consuming to perform, especially under reservoir conditions. Typically, a
single set of constitutive relations is assumed and applied to the
whole field or to a few major rock types. This is almost invariably
incorrect. Constitutive relations are a direct manifestation of the
complex rock microstructure, the physical characteristics of the
solid, and the fluid properties. Multiphase properties such as relative permeability can, therefore, vary considerably throughout the
reservoir, depending on the local pore structure, wettability trends,
and fluid interactions.
An alternative or complementary technique to derive transport
properties of rocks is to start from a description of the microstructure and then follow a process-modeling approach. In the past
decades, several methods have been proposed for the acquisition of
3D representations of rock microstructures. Such methods include
direct imaging using X-ray microtomography (Dunsmoir et al.

Copyright 2010 Society of Petroleum Engineers


This paper (SPE 124617) was accepted for presentation at the SPE Annual Technical
Conference and Exhibition, New Orleans, 47 October 2009, and revised for publication.
Original manuscript received for review 19 August 2009. Revised manuscript received for
review 20 January 2010. Paper peer approved 25 January 2010.

2010 SPE Journal

1991; Spanne et al. 1994; Arns et al. 2004), process- or geologically based reconstructions (Bryant et al. 1993; Bakke and ren
1997; ren and Bakke 2002), and statistical reconstructions (Adler
et al. 1990; Yeoung and Torquato 1998; Wu et al. 2006). Macroscopic material and transport properties, in turn, may be derived
directly from the microstructure images by solving the governing
transport equations numerically within the context of the images.
Different static petrophysical properties, such as permeability,
formation factor, nuclear-magnetic-resonance (NMR) relaxation,
and elastic properties, have been computed successfully following
this approach (Arns et al. 2002, 2004; Knackstedt et al. 2004; Jin
et al. 2007; ren and Bakke 2002; ren et al. 2002, 2007).
Predictions of constitutive relations from the underlying rock
microstructure are made commonly using network-modeling techniques. The premise of the network model is that the complex pore
space can be represented by an equivalent network of interconnected pores. Individual pore elements are represented by simplified
geometrical shapes that are amenable to analytical expressions for
capillary entry pressure, hydraulic conductances, and phase saturations. Fluid/fluid interfaces, thus, can be determined accurately;
hence, the resolution is infinitely good. Since the pioneering
work of Fatt (1956), network models have been used extensively
to study a huge range of transport phenomena in porous media
(Blunt 2001). The predictive capabilities of network models have
improved greatly with recent developments in constructing geologically realistic pore networks from microstructure images (Lindquist
et al. 1996; ren and Bakke 2003; Sheppard et al. 2005; Silin and
Patzek 2006; Jiang et al. 2007). Predictive modeling of a variety
of processes, including two- and three-phase relative permeability,
has been demonstrated for a number of different rocks (ren et al.
1998; Lerdahl et al. 2000; ren and Bakke 2003; Valvatne and
Blunt 2004; Piri and Blunt 2005; ren et al. 2006; Svirsky et al.
2007; Nardi et al. 2009).
Despite the success of using microstructure images combined
with network modeling to derive constitutive relations, a number of
limitations exist. Most network models are restricted to quasistatic
conditions that neglect the effects of viscous forces on fluid distribution and on flow or mobilization of discontinuous phases. Accurately
capturing the effects of viscous forces in multiphase network models
has proved elusive and continues to be an area of active research.
Furthermore, extraction of topological and geometrical equivalent
pore networks from rock-microstructure images remains a challenging and uncertain task (Dong et al. 2008) that involves geometrical
simplifications that inevitably have an effect on flow predictions.
A way to surmount some of these limitations is to simulate the fluid
dynamics of multiphase flow directly on microstructure images.
However, this is a difficult task within conventional computationalfluid-dynamics methods, such as the finite-difference and finite-element methods. The difficulties are mainly because of the complex
pore space and inherent free-boundary issues, such as breaking and
merging of interfaces that often depend on externally imposed conditions (e.g., flow rate or applied pressure drop).
A computational-fluid-dynamics method that has gained popularity in recent years is the LB method. LB models are kinetic in
origin and fall into the category of mesoscopic modelsbetween
microscopic and macroscopic. Hence, they are capable of capturing
microscopic effects as well as reproducing macroscopic behavior.
In contrast to conventional methods, LB models do not track interfaces. Sharp interfaces can be maintained automatically, and macroscopic behaviors such as interface dynamics arise naturally from
the microscopic effects. Furthermore, because of the simplified
1

Flow direction

Fig. 1Sketch of the boundary conditions. The left figure illustrates the imposed void layers, while the right figure shows the
bounce-back scheme in wall/particle collision and the concept
of periodic boundary conditions.

handling of complicated boundary conditions, LB models are a


powerful method for simulating flow in systems with arbitrarily
complex geometry, such as porous media. A variety of applications
of the general LB model with extensions and tailored modifications have been applied to porous media, such as single-phase
flow [e.g., Jin et al. (2004) and Succi et al. (1989)], two-phase
flow [e.g., Martys and Chen (1996), Chen and Doolen (1998),
Ferrol and Rothmann (1995), Pan et al. (2004), and Schaap et al.
(2007)], multicomponent reactive transport (Kang et al. 2006), and
recently for testing of hypotheses linked to enhanced oil recovery
(Pride et al. 2008).
In this paper, we report LB simulations of flow in porous
media with immiscible fluid pairs, such as oil and water, which
we refer to as two-phase flow. We start with a brief description of
the general LB method, including preliminaries and our specific
implementations of the method for two-phase flow. We show that
the implemented LB model accurately reproduces capillary pressure effects in pores with a noncircular shape. Next, we investigate
the effects of viscous coupling on relative permeability in idealized
geometries. Finally, the model is applied to simulate two-phase
flow under water-wet conditions directly on rock-microstructure
images of Bentheimer sandstone. The images were acquired from
X-ray microtomography and from process-based reconstructions.
The simulated capillary pressure and relative permeability curves
are compared with experimental data (ren et al. 1998).
The LB Method
The LB method is an attractive and versatile tool for simulating
flow in porous media. This is mainly because of its computational
efficiency, programming simplicity, and inherent parallelism. The
LB method is a mesoscopic method for obtaining numerical
solutions for the macroscopic behavior of incompressible hydrodynamics. This is achieved by solving a discretized Boltzmann
equation of fluid-particle distributions that move and interact on
a regular lattice with very limited degrees of freedom. These are
represented in the lattice by directional links in which the distributions are allowed to move. The underlying assumption is that the
overall fluid behavior depends very little on individual particles of
the fluid; rather, it is the collective effects of many particles that
govern the dynamics. Although the LB method consists of highly
coarse-scaled particles and the interactions on the microscopic
scale are strongly simplified, they collectively reproduce macroscopic behaviors and take small-scale effects into account. For our
LB simulations, we employ the so-called D3Q19 lattice, which is
a 3D lattice with 18 links representing 19 velocities (including the
zero velocity). A detailed description and review of the LB method
and the lattice that we use in this study can be found in Rothmann
and Zaleski (1997), Succi (2001), Chen and Doolen (1998), and
Chen et al. (2009).
In the following, we describe key issues involved in the implementation of the model and in extending the model to simulation of
immiscible two-phase flow that exhibits surface tension at fluid/fluid
interfaces and specific wetting properties at fluid/solid interfaces.
2

Boundary Conditions. Two kinds of boundary conditions are used


in the LB simulations: no-slip boundary conditions at solid/fluid
interfaces and periodic flow conditions at the inlet and outlet. The
no-slip boundary condition is implemented using the bounce-back
scheme where a particle distribution that propagates onto a solid node
scatters back with no relaxation and has its momentum reversed. The
scheme is very easy to implement numerically, but, in its simplest
form, it treats the walls position half-way between the void nodes
that represent the pore space and the solid matrix. This leads to a
simplification of the medium that is claimed to be only first-order
accurate. For very complex geometries, this may lead to artifacts in
the simulations (Manwart et al. 2002; Maier et al. 1996).
At the inlet and outlet of the model, we apply periodic boundary conditions (i.e., the fluid that propagates out of the model will
enter it at the opposite end). Periodic boundary conditions are
implemented by adding four extra void layers at the inlet and outlet
faces of the model. A schematic of this is presented in Fig. 1, along
with the implementation of the bounce-back scheme.
For large model sizes, it is necessary to run a single simulation
on several computer processors (CPUs) in a parallel environment.
Hence, the model needs to be separated into domains that are
assigned to different CPUs that communicate through a messagepassing interface. How the model is split into domains depends
on the workload and structure of the model grid and the number
of CPUs. However, the void layers mentioned in the preceding
paragraph are distributed among the domains so that they follow
the global structure in Fig. 1.
Driving of the fluids is accomplished by applying a uniform
body force to the particle distributions. This is equivalent to an
external force acting on the system and adds extra momentum to
the particle distribution (Buick and Greated 2000). If the force acts
in the x direction, the macroscopic momentum balance is altered
and the local fluid pressure will change as P = P0Fbxx in the
flow direction, which is the same as applying a pressure gradient
P/L. According to the weight functions of the D3Q19 lattice, the
relation between the body force and the pressure gradient is
1
P
0 Fbx =
.
3
Lx

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)

Fluid/Fluid Interaction. Several LB models exist to incorporate


surface tension and the subsequent phase separation. One of the
first was the color-gradient model proposed by Gunstensen and
Rothmann (1992). Later, Shan and Chen (1993, 1994) introduced a pseudopotential model based on microscopic interactions
between the phases. A thermodynamically consistent two-phase
model based on a free-energy approach was introduced by Swift
et al. (1995). In addition, LB models derived from kinetics equations exist (Guo and Zhao 2005). Collectively, these models have
been tested and implemented on a broad range of multiphase-flow
scenarios. All of the models have their strengths and weaknesses.
However, there is no method or class thereof that has been
reported to be superior to the others (Succi 2001). The method
of choice depends on personal preference and on the specific
system that is studied. In our work, we use the color-gradient
method with later modifications (Latva-Kokko and Rothmann
2005). This is mainly because (1) the value of the surface tension
in this method is controlled by a single free tuning parameter and
(2) it is easy to control the contact angle and, thus, the wetting
properties of the solid.
When a fluid system consists of two or more immiscible
phases, the fluids must be treated explicitly with their specific
properties. However, the fluid phases coexist on the same lattice
and constitute a collective momentum and mass distribution. In
the case of two-phase flow, the local fluid-particle distribution
Ni is split in two parts, Ni = Ri+Bi, which we call red and blue,
respectively (space and time coordinates are skipped for brevity).
The two immiscible phases have a physical tendency to demix
and to minimize diffusion. This is the notable effect of surface
tension. It acts locally as an anisotropic pressure field and gives
rise to capillary forces.
2010 SPE Journal

In order to implement the surface tension, something must be


said about the separation of interfaces. This is achieved through
a color gradient:
I
I
I I
I I
f = ci R j ( x + ci ) B j ( x + ci ). . . . . . . . . . . . . . . . . . . . . (2)
i

The effect of the surface tension is applied as a local perturbation


to the particle distribution:
I I 2
I ci f
1
N i N i + 36 wi  f I I , . . . . . . . . . . . . . . . . . (3)
2
ff

and we exclude rest particles. The strength of the perturbation is


set by the parameter , and the surface tension  is given by

=

192 
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)


where  is the inverse relaxation time and the constant 192 is


determined by the lattice (Rothmann and Zaleski 1997). The
single tuning parameter  is used to scale the magnitude of the
actual surface tension. The two phases are separated according to
the method presented by Latva-Kokko and Rothmann (2005), and
each phase population is updated by
Ri =

R
RB
Ni + 
N eq cos . . . . . . . . . . . . . . . . . . . (5)
R+ B
( R + B )2 i

and
Bi =

B
RB
Ni 
N eq cos , . . . . . . . . . . . . . . . . . . (6)
R+ B
( R + B )2 i

where R and B are the local densities of the two phases and 
1.0. The angle between the gradient and the velocity direction is
given by cos .
Fluid/Solid Interaction. Forces acting between the solid s and
the fluids determine the wetting properties of the system. These
are characterized by a static contact angle
according to the force
balance at the three-phase contact line: RBcos
= sRsB. In the
color-gradient model, the contact angle is controlled by a single
parameter that gives the concentration of a fluid component at
the solid nodes. This parameter is constant and affects the color
gradient and the calculation of the surface tension. For a perfectly
sharp interface, the total color gradient at a line perpendicular to
the interface will, from Eq. 2, be
I
f = R + B = 2 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7)
which gives full surface tension. At the solid, this will be altered
so that the tension balance reads
2 cos
=  (1 + )  (1 ), . . . . . . . . . . . . . . . . . . . . . . . . (8)
where 1 1 determines which fluid is wetting and the value
of the contact angle cos
= .

Physical Systems Vs. Lattice Systems. Physical units must be


assigned to the system if we want to perform dimensional analysis
toward real fluid systems. The LB model itself operates strictly
with units associated with the lattice. These must be rescaled in
order to obtain physical ones by means of typical quantities that
represent the system under investigation. In the following, subscript P denotes physical units. The main physical quantities are
Length unit a; the physical length between the lattice nodes:
xP = ax.
Mass unit m0; rescales the mass density: P =  m0/a3.
Time unit t0; the characteristic time scale that, for example,
scales the fluid velocity uP = u a/t0.
The first two scaling units are straightforward to set because
they are directly related to the fluid properties and the grid model.
The time is more delicate to set. One approach that we have used
here is to define the value from the physical viscosity of the fluids
P = a2/t0. The lattice value of the kinematic viscosity is set before
the simulation, and, with a known physical value, the typical time
unit can be obtained that again affects the physical value of other
parameters, such as pressure, PP = P m0/(at02) and surface tension
P =  m0/t02.
Results and Discussion
In this section, we present and discuss the results of two-phase LB
simulations conducted on digitized models of porous media. We
first show that the model accurately reproduces capillary pressure
effects in pores with a noncircular shape. Then, we investigate the
effects of viscous coupling on relative permeability for concurrent
flow in a single pore. Finally, we report capillary pressure and relative permeability curves computed on rock-microstructure images
of Bentheimer sandstone. The simulated results are compared with
experimental data (ren et al. 1998).
Capillary Pressure. In real porous rocks, two (or even three)
phases can be present in a single pore at capillary equilibrium.
The wetting phase is present in corners of the pore with nonwetting fluid occupying the center of the pore. Representing pores as
angular tubes, thus, is more realistic than the commonly assumed
cylindrical tube (Blunt 2001). Equilibrium capillary pressure relations for angular tubes can be derived from a free-energy balance
(Mason and Morrow 1991; ren et al. 1998) and is given by
Pc =

 1 + 2 G cos

F (
, G ) , . . . . . . . . . . . . . . . . . . . . . (9)

where r is the inscribed radius and G = A/O is the shape factor,


where A is the cross-sectional area and O is the perimeter length.
The function F(
, G) depends on the actual corner shape (i.e.,
angle) and the value of the contact angle. For small contact angles,
F(
, G) is close to unity (ren et al. 1998).
To verify this relation, we use the LB model to compute capillary
pressure in a square tube at strongly water-wet conditions. Fig. 2
shows the setup for the numerical experiment. Clearly, the blue
fluid is strongly wetting and exists as arc menisci in the corners
at equilibrium. To obtain an accurate measurement of the capillary
pressure, we measure the pressure inside each phase along a line
that runs through the middle of the tube normal to the cross section, as illustrated in Fig. 3 (left). The capillary pressure is given

Fig. 2Capillary tube with square cross section both at initial state (left) and at equilibrium (right). The nonwetting fluid is dark
colored, while the wetting fluid is wired. The contact angle is small, cos = 0.9. The system size is LxLyLz = 402020.
2010 SPE Journal

0.04
Nonwetting
Wetting

LB simulation
Analytic

0.35

0.03

0.34

Pc

Dimensionless Pressure

0.36

0.02

Pc

0.33

0.01

0.32
0

10

20
x

30

40

0.05

0.1

0.15

0.2

1/r

Fig. 3The left figure shows a profile of the pressure in each phase and how the capillary pressure is measured along a line
running through the middle of the tube normal to the cross section. At right is a figure showing the scaling of the capillary pressure according to Eq. 9. The mach is excellent for larger inscribed radii. The lattice surface tension = 0.1.

by the fluid pressure difference between the nonwetting and the


wetting phase. Fig. 3 (right) shows that the computed results are
in excellent agreement with the predictions from Eq. 9.

and

Relative Permeability. Immiscible two-phase flow in porous media


is commonly described in terms of a generalized Darcys law using
the concept of relative permeability, kr. This is an empirical extension
of the single-phase Darcys law based on the idea that each fluid
flows in separate pore channels. Although it is recognized that kr
depends on pore structure, saturation, and the balance between viscous and capillary forces, quantified by the capillary number Nca,

Eqs. 11 and 12 show that the wetting-phase relative permeability


depends only on its own saturation while the nonwetting-phase
relative permeability depends both on saturation and on the viscosity contrast M. To verify the relations, we compute relative
permeability curves for concurrent annular flow in circular tubes.
To measure the relative permeabilities, we first calculate the total
momentum of each phase at a given wetting-phase saturation Sw:

 (Q / A )
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10)


I
 ( Sw ) =  u . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13)

N ca =

it is assumed that kr does not depend on fluid properties, such as


viscosities and densities. However, a typical situation for immiscible two-phase flow is that both fluids flow simultaneously in
a single pore. Viscous coupling or lubrication of the nonwetting
phase by a wetting film might then be important.
To investigate this, we consider concurrent annular flow in a
cylindrical pore. The wetting fluid flows along the pore walls, with
nonwetting fluid in the center. The relative permeability for this
case depends on the viscosity ratio M = nw/w and is given by
(Goldsmith and Mason 1963)
krw = Sw2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11)

Relative Permeability

1
0.8
0.6
0.4

Analytic
krnw
krw

0.2
0
0

0.2

0.4

0.6

0.8

Sw
Fig. 4Relative permeability for concurrent flow in a cylindrical tube with radius r = 20 voxels and viscosity contrast M = 1.
The simulated data are compared to the analytic predictions of
Eqs. 11 and 12.
4

2
krnw = 2Snw M + Snw
(1 2 M ). . . . . . . . . . . . . . . . . . . . . . . . . (12)

This value is normalized with the single-phase momentum to find


the fractional flow. From the single-phase global pressure drop
P(S = 1), the relative permeability in each phase is given as
kr  =

 ( Sw ) P ( S = 1)
. . . . . . . . . . . . . . . . . . . . . . . . . (14)
 ( S = 1) P ( Sw )

Fig. 4 compares the simulated relative permeabilities for the


equal-viscosity case with the predictions from Eqs. 11 and 12. The
computed relative permeabilities are in excellent agreement with
the analytical solutions. The relative permeability of both phases
is less than unity, and we note that krw+krnw = 1.
Simulated results for M = 0.1 and M = 10 are shown in Fig. 5. For
M = 0.1, both relative permeabilities are again less than unity. However,
the nonwetting-phase relative permeability is reduced compared with
the equal-viscosity case. Although both phases are subject to the same
pressure gradient, the nonwetting phase is slowed down by the wetting
phase, which is more viscous. The velocity profile of the wetting phase
close to the wall of the tube is practically unaffected by the presence of
the nonwetting phase, which is expected according to Eq. 11. Detailed
velocity profiles are provided in Fig. 5. For M = 10, the LB simulations are again consistent with the analytical solutions. We note that
krnw is considerably larger than unity for almost all saturations. This is
caused by viscous coupling and the lubricating effects of the wetting
phase, which are captured satisfactorily in our model. Compared with
the analytical solution, we compute a slightly lower nonwetting-phase
relative permeability. This is mainly because of finite-size-resolution
effects that have an effect on the numerical distinction (i.e., sharpness) of the fluid/fluid interface. With a sharper interface, momentum
transfer between the phases is more accurate, and we observe better
agreement with Eq. 12 as the system size is increased. If the sharpness
is not sufficient, the effective viscosity contrast will be altered and the
sum krw+krnw will become different. For M = 1.0, we observe that the
resolution is of less importance because krw+krnw = 1. For a comparison,
2010 SPE Journal

6
Analytic
krnw r = 20
krw r = 20
krw r = 40
krnw r = 40

0.8
0.6

Relative Permeability

Relative Permeability

0.4
0.2
0

4
Analytic
krnw r = 20
krw r = 20
krw r = 40
krnw r = 40

3
2
1
0

0.2

0.4

0.6

0.8

Sw

8.0105

1.2104

LB simulations, nw
LB simulations, w
Analytic

6.0105
4.0105
2.0105
0.0
0

10

20
Position z

0.2

0.4

0.6

0.8

Sw

Dimensionless Fluid Velocity

1.0104
Dimensionless Fluid Velocity

30

40

1.0104

LB simulations, nw
LB simulations, w
Analytic

8.0105
6.0105
4.0105
2.0105
0.0
0

10

20
Position z

30

40

Fig. 5Relative permeability for concurrent flow in a cylindrical tube with radius r = 20 and r = 40 voxels and viscosity contrast
M = 0.1 (left) and M = 10 (right). The simulated data are compared to the analytic predictions of Eqs. 11 and 12. We observe that,
with finer resolution, the simulated results for the nonwetting fluid approach the analytic predictions. In the lower figures are
the corresponding velocity profiles for a nonwetting saturation Snw = 0.5. The velocity of the nonwetting fluid in the center of
the tube is decreased for M = 0.1 (lower left) and enhanced for M = 10 (lower right). As is visible on both the lower figures, the
interface between the wetting and nonwetting fluid is not perfectly sharp because of low resolution. The analytic comparison is
performed by using the methodology of Kang et al. (2004) and Yiotis et al. (2007), and, because the match is fair for M = 0.1, it
suggests that the velocity profile of the nonwetting phase from the LB simulation is underpredicted. This is also evident in the
relative permeability data.

we refer to studies by Kang et al. (2004) and Yiotis et al. (2007) that
investigated two-phase concurrent flow in a 2D channel.
Bentheimer Sandstone. Two-phase oil/water relative permeability
and capillary pressure relations were measured on three homogeneous core samples of Bentheimer sandstone. This water-wet
sandstone is well sorted and composed mainly of quartz (7080%),
feldspar (2025%), and authigenic clays (23%). The measured
porosity and permeability of the samples ranged from 0.23 to 0.24
and from 2820 to 2930 md, respectively. Relative permeabilities
were measured by the steady-state method. Further details of the
experiments are given by ren et al. (1998). We acquired rockmicrostructure images of Bentheimer sandstone by two methods:
X-ray microtomography (denoted MCT) and process-based reconstructions (denoted PBM). The resolution of the MCT image is 6.67
m, and the size is 5003 voxels. The corresponding values for the
PBM are 5 m and 1,0003 voxels. Fig. 6 depicts the grain matrix and
associated pore space of a reconstructed Bentheimer sandstone.
The multiphase LB simulations are computationally expensive.
Thus, it is desirable to perform the simulations on samples that are
small but still large enough to capture a representative volume.
However, rocks are heterogeneous at all scales, including the
pore scale, and it is difficult to determine a priori a representative
sample size. Fig. 7a shows the directionally averaged two-point
correlation function C2 for the microtomographic data. The decay
length L, determined at C2 = 0, is found with L = 40 voxels (i.e.,
L = 267 m). The rapid decline of the function shows that the sample is fairly homogeneous and that little spatial correlation exists
2010 SPE Journal

beyond L = 40 voxels. Fig. 7b shows the local porosity distribution


( , L) (Hilfer 1991) for the MCT sample. ( , L) measures the
empirical probability of finding the local porosity in a cubic cell
of linear dimension L. The porosity distributions displayed in Fig.
7b were computed with L = 128 voxels. Although the side length
of the measuring box is more than 3 times larger than the decay
length, significant variation in porosity is observed. The width of
the distribution ranges from 0.15 to 0.28, with the peak (most-probable porosity) occurring at = 0.21. This demonstrates the spatial
nonstationarity of porosity.

Fig. 6Reconstructed digital grid of Bentheimer sandstone. The


left figure shows the grain matrix, and the right figure shows the
pore space. The images correspond to the PBM2 model.
5

0.3
0.25
Frequency

0.8

C2

0.6
0.4
0.2

0.2
0.15
0.1
0.05

0
0

20

40
60
L, voxels

80

0
0.14

100

(a)

0.16

0.18

0.2
0.22
Porosity

0.24

0.26 0.28

(b)

4000

28

3500

26

3000

24

Formation Factor

Permeability, md

Fig. 7The left plot (a) shows the average two-point correlation function for the original microtomographic data, while the right
plot (b) shows the local porosity distribution.

2500
2000
Size 1283
Size 2563
Size 5003

1500
1000
500
0.16

0.18

0.2

0.22
0.24
Porosity

0.26

Size 1283
Size 2563
Size 5003

22
20
18
16
14

0.28

12
0.16

0.18

0.2

0.22
0.24
Porosity

0.26

0.28

Fig. 8Average permeability (left) and formation factor (right) values as a function of the porosity for different sample sizes.
Their behavior falls onto a linear dependency on the porosity.

To investigate the resulting spatial nonstationarity of the effective transport properties, we computed absolute permeability and
formation factor on nonoverlapping subsamples of size 1283 and
2563 voxels. The results of the calculations are displayed in Fig. 8.
Although the porosity, permeability, and formation-factor for the
subsamples of size 1283 vary widely, the porosity-vs.-permeability
and porosity-vs.-formation factor trends are well constrained. The
results for the larger subsamples fall right on the trends as do
those computed on the full MCT sample. These trends thus exhibit
spatial stationarity.
The results suggest that, for the Bentheimer sandstone, representative effective properties might be obtained from calculations
on smaller subsamples. The actual LB simulations were performed
on subsamples of size 1283 and 2563 voxels in order to reduce
memory and computational expenses. Table 1 summarizes the
computed porosity and permeability of the various samples used
in this work. For all the simulations, we applied periodic boundary conditions in the flow direction (see Fig. 1) and impermeable
boundaries on the other four faces parallel to the flow direction.
The applied body forces were equal for both phases in every
simulation run. The fluid properties that we used are summarized
in Table 2. The only parameter that is not linked a priori to the
actual fluid properties is the time scale t (i.e., the time it takes
to perform one timestep). However, it affects all other physical
quantities in which time is included; hence, t must be selected
carefully. The contact angle is set to
= 35, which is consistent
with the observed water-wet behavior of Bentheimer sandstone.
Capillary Pressure. Simulated primary-drainage capillary pressure curves were obtained for Sample MCT1. The procedure is
6

to mimic the behavior of an experimental centrifuge setup, which


gives the capillary pressure as the pressure head, Pc = P (Dullien
1992). This is accomplished by first injecting nonwetting phase
at a low flow rate and then allowing the system to reach steady
state, thus obtaining one point on the capillary pressure curve. The
system is assumed to have reached steady state when the change
in saturation meets the criterion
S =

S (t ) S (t T )
, . . . . . . . . . . . . . . . . . . . . . . . . . . (15)
S (t )

where T is the sampling rate (T = 100). If S drops below the


desired value, the flow rate is increased by a predefined constant
TABLE 1GRID AND PETROPHYSICAL PROPERTIES
OF THE DIFFERENT SAMPLES OF BENTHEIMER
SANDSTONE*
Sample

Size
3

Resolution

k (md)

Porosity

PBM

1,000

5.0 m

2685

0.218

PBM1

128

5.0 m

2470

0.234

PBM2

256

5.0 m

2487

0.220

MCT

500

6.67 m

1918

0.206

MCT1

128

6.67 m

3463

0.252

MCT2

256

6.67 m

2074

0.222

* The average absolute permeability values k are found from LB simulations,


and the intergranular porosity values are taken from image analysis of the grid
models.

2010 SPE Journal

TABLE 2FLUID PROPERTIES USED IN THE LB SIMULATIONS


Parameter

Lattice value

Physical value

Model

PBM1

PBM2

MCT1

MCT2

Time scale t

1.0

1.0 s

1.0 s

1.5 s

1.5 s

Mass density of wetting fluid, w

1.0

1000 kg/m

1000 kg/m

Mass density of nonwetting fluid, nw

1.0

Surface tension,

0.1

12.5 mN/m

Viscosity of wetting fluid, w

0.05

1.25 mm /s

Viscosity of nonwetting fluid, nw

J (S ) =

Considering the small size of the sample, the agreement between


computed and measured capillary pressures is encouraging. Fig.
9 shows that both the simulations and the experiments display
L-shaped curves with an extended plateau region. This reflects
the homogeneity and narrow pore-size distribution of Bentheimer
sandstone. The predicted irreducible wetting-phase saturation Swi is
larger than the measured one (0.09 vs. 0.05). This is mainly because
of resolution effects. At low Sw, the hydraulic continuity of the wetting phase is mainly maintained through wetting films present in
corners and crevices of the pore space. These are difficult to capture
accurately, unless the sample resolution is very good. Wetting films
thus tend to collapse at a lower capillary pressure in the simulations
than in the experiments, and, as a result, Swi increases.
Relative Permeability. There are two main types of experimental
setups for measuring relative permeability: unsteady state and
steady-state flow. It is our goal to be able to closely reproduce both
of these experimental setups by direct simulations. To this end, we
have performed tests for both unsteady- and steady-state flow.
Unsteady-State Flow. We examined unsteady-state flow in
Sample PBM1 by injecting nonwetting fluid into the sample that is

1000 kg/m

1000 kg/m

1000 kg/m

1000 kg/m

13.2 mN/m

1.48 mm /s

1.25 mm /s

1.48 mm /s

1.48 mm /s

initially fully saturated with wetting fluid. This mimics a primarydrainage displacement. As before, we applied periodic boundary
conditions, but wetting fluid that exited the model changed to nonwetting fluid and was reinjected at the inlet face. This procedure
can be reversed easily to mimic an imbibition process (i.e., injection of wetting phase). A sequence of snapshots from a simulated
primary drainage and subsequent imbibition process is shown in
Fig. 10. The capillary number for the simulations is relatively high,
Nca = 103, and it is obvious that viscous effects are prevalent. The
reason we chose the high Nca was to avoid numerical instabilities
and to obtain a complete invasion within a relatively short time.
Simulated primary-drainage relative permeabilities are shown
in Fig. 9. In the unsteady-state simulations, we measure the fractional flow in Eq. 14 only at the outlet of the model. Consequently,
the fractional flow of the injected nonwetting fluid is zero before
breakthrough. However, the global pressure drop increases before
breakthrough, caused by capillary forces at the interfaces. This, of
course, affects the relative permeability of the wetting phase, which
decreases instantly. We also observe strong initial transient effects
that lead to krw > 1 at high Sw. For the applied body force and flow
rate, the final wetting-phase saturation Swi = 0.22. Also shown in
Fig. 11 are relative permeabilities obtained from network-model
simulations at quasistatic conditions. As expected, the relative
permeabilities from the high-Nca LB simulations are larger than
those for quasistatic conditions (i.e., Nca 0). It is well established
that relative permeabilities increase and tend toward straight lines
when Nca increases (Bardon and Longeron 1980).
Steady-State Flow. To mimic steady-state-flow measurements,
we randomly distribute the fluid phases inside the void space
according to a target saturation value. Flow at a given Nca is then
commenced and proceeds until steady state is reached, thus obtaining one point on the kr-vs.-Sw curve. Steady state was assumed to
be reached when the relative linear drift in the overall volumetric
flux yielded a regression coefficient b < 105. This was obtained
statistically from the time evolutions of the distributions. As for
the unsteady-state case, we apply periodic boundary conditions,
but phases cross the boundaries freely. Phase saturations thus are
10

LB simulations
Experimental

LB simulations
Experimental
6

Leverett J-function

Leverett J-function

13.2 mN/m

1.48 mm /s

1.25 mm /s

Pc
k
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (16)
 cos

1000 kg/m

1.25 mm /s

until the system reaches a new equilibrium state. It is important to


choose the value of in Eq. 15 carefully. Obviously, if is large,
the computational time is reduced, but sufficient time will not be
allowed for the system to reach equilibrium. The predicted capillary pressure for a given saturation thus will be too high. In the
following, we use = 5107. This value is based on numerical
experiments.
Fig. 9 compares computed and measured capillary pressure
curves. To account for variations in porosity and permeability
between the samples, the capillary pressures are given in terms of
the dimensionless Leverett J-function:

12.5 mN/m

0.05

1000 kg/m

0.1

0
0

0.2

0.4

Sw

0.6

0.8

0.2

0.4

Sw

0.6

0.8

Fig. 9Measured vs. computed capillary pressure curves for Bentheimer sandstone. Computed results are for Sample MCT1. Four
different runs were started at different initial saturations to obtain the total simulation curve. The experimental data are taken from
three individual laboratory experiments. The dashed line is a guide to the eye following the mean of the experimental data.
2010 SPE Journal

S w = 0.92

S w= 0.61

S w = 0.38

S w = 0.22

S w= 0.79

S w = 0.91

Fig. 10Snapshots of simulated primary drainage (top) and subsequent imbibition (bottom) in sample PBM1. The dark fluid is
nonwetting, while the lighter gray colors indicate the surface of the porous matrix. The wetting fluid is not displayed so as to
maintain visual clarity in the pictures.

constant during the simulation. Similar to experimental setups, we


measure the invariant fractional flux. We do this over the entire
sample to obtain better averaged values. The global pressure drop
over the model was the same in both phases because the same
body force was applied to each phase. This results in a zero capillary pressure gradient, which is strictly valid only if the saturation
profile is uniform, something we assume for these simulations.
We simulated steady-state relative permeabilities on samples
PBM2 and MCT2 for Nca = 105, which is a typical value for
experimental steady-state measurements. The value of Nca is controlled using a regulator that adjusts the body force Fb to produce
a constant volumetric flux. Computed data are compared with
experimentally measured primary-drainage relative permeabilities
in Fig. 12. We first note that the simulated results for samples
PBM2 and MCT2 are similar. This suggests that the pore structure
of the two samples is also similar. The simulated krw is in good
agreement with the experimental one across the entire saturation
range. The predicted krnw is in fair agreement with the experimental
data at low and intermediate wetting-phase saturations. At high
Sw, the simulated results underpredict krnw. Percolation threshold
effects, caused by low Nca and small sample size, are the main

cause for the low krnw. This is also demonstrated clearly in Fig. 11,
which shows that, even at high Nca, nonwetting-phase breakthrough
occurs at relatively high nonwetting-phase saturations (Snw 0.2).
For larger systems, breakthrough happens at lower Snw and percolation threshold effects are reduced.
Fig. 13 shows the effects of increasing Nca on the relative
permeabilities of sample MCT2. Increasing the capillary number
from Nca = 105 to Nca = 104 has only minor effects on the wettingphase relative permeability. However, it has a strong effect on the
nonwetting-phase relative permeability, especially at high Sw. Our
simulations show that this is chiefly because of mobilization and
flow of disconnected clusters of nonwetting fluid. This agrees with
previously reported observations (Avraam and Payatakes 1999; Li
et al. 2005). Finally, we note that the scatter of the experimentally
measured krnw values is bounded by the simulated krnw curves for
Nca = 105 and Nca = 104.
Conclusion
We have simulated immiscible two-phase flow in porous media
using a color-gradient-based two-phase LB model. We show that
the model accurately captures capillary pressure and wetting

1
krnw
krw
krw Network
krnw Network

0.8

Relative Permeability

Relative Permeability

0.6
0.4
0.2
0

0.1

0.01

0.001
0

0.2

0.4

Sw

0.6

0.8

krnw
krw
krw Network
krnw Network

0.2

0.4

Sw

0.6

0.8

Fig. 11Relative permeability for the primary-drainage part of the simulation shown in Fig. 10. Comparison with network simulations for very low capillary numbers is given for illustration. The left figure presents the relative permeability on a linear scale,
while the right figure shows the same data on a logarithmic scale for comparison.
8

2010 SPE Journal

1
krnw (LB)
krw (LB)
krnw (Experimental)
krw (Experimental)

0.8

Relative Permeability

Relative Permeability

0.6
0.4
0.2
0

krnw (LB)
krw (LB)
krw (Experimental)
krnw (Experimental)

0.8
0.6
0.4
0.2
0

0.2

0.4

Sw

0.6

0.8

0.2

0.4

Sw

0.6

0.8

Fig. 12Steady-state relative permeability for LB simulations on samples MCT2 (left) and PBM2 (right) compared with data for
primary-drainage steady-state experiments. The wetting-phase relative permeabilities are in good agreement with the experimental
ones for both samples. For the nonwetting phase, the agreement is fair.

effects in pores having noncircular shapes. The model is used to


study the effects of viscous coupling on relative permeability in
capillary tubes. We find that the nonwetting-phase permeability
may greatly exceed the single-phase permeability when the wetting
phase is less viscous than the nonwetting phase. The simulated
relative permeabilities agree well with analytical solutions, and we
conclude that the model is capable of handling viscous coupling
effects caused by contrasts in fluid viscosities.
We have applied the model to simulate two-phase flow directly
on rock-microstructure images of Bentheimer sandstone. The
images were obtained from X-ray microtomography and processbased reconstructions. Simulated capillary pressure and relative
permeability relations are compared with experimental data from
three water-wet Bentheimer core plugs. The computed drainage
capillary pressure curve is in good agreement with experimental
data across most of the saturation range, and we conclude that the
proposed approach for numerically measuring capillary pressure
curves is physically sound.
We demonstrate that the model can be used to mimic both
unsteady- and steady-state relative permeability measurements.
Simulated steady-state relative permeabilities are in fair agreement
with experimental steady-state data at similar Nca values (Nca
105). The model accurately predicts the wetting-phase relative
permeability but tends to underpredict the nonwetting-phase relative permeability at high wetting-phase saturations. This is mainly
because of finite-size percolation threshold effects. For increasing
Nca (Nca 104), we observe that the nonwetting-phase relative
permeability increases, which agrees with theory and shows that
the model can handle a wide range of flow rates.

Relative Permeability

1
krnw
krw
krnw
krw
krnw
krw

0.8
0.6

Nca = 105
Nca = 105
Nca = 104
Nca = 104

(LB)
(LB)
(LB)
(LB)

(Experimental)
(Experimental)

0.4
0.2
0
0

0.2

0.4

Sw

0.6

0.8

Fig. 13Steady-state relative permeability for sample MCT2 at


different capillary numbers.
2010 SPE Journal

The work presented here is only the first part of a larger


research effort aimed at developing a more fundamental processbased approach for acquiring constitutive relations for reservoir
rocks. Although these preliminary results are encouraging and suggest that two-phase LB simulations can be a powerful and versatile
tool for obtaining physically meaningful constitutive relations,
a significant amount of work remains. One needs to investigate
systematically and verify that the LB model accurately captures
and reproduces the effects of different rock microstructures, complex wettability, and different displacement processes. This work
is ongoing.
Nomenclature
a = lattice constant
A = cross-sectional area
B = particle distribution of the blue phase
I
c = lattice velocity in the i direction
cs = lattice sound speed
CI2 = two-point correlation function
f = LB color gradient
F = function that affects the capillary pressure in noncircular
geometries
Fb = body force
G = shape factor
k = absolute permeability
kr = relative permeability
L = length
m = mass
M = dynamic-viscosity ratio
N = particle distribution
Nca = capillary number
Neq = pseudoequilibrium distribution function
O = circumference
P = pressure
Pc = capillary pressure
Q = volumetric flux
r = radius
R = particle distribution of the red phase
S = saturation
t = time
T = sampling time
u = fluid velocity
w = weight function
I
x = position
 = phase-separation controlling parameter
P = pressure drop
9

t





0







=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

timestep
steady-state criterion
contact-angle controlling parameter
static contact angle
dynamic viscosity
kinematic viscosity
particle momentum
fluid-particle density
fluid density
surface tension
relaxation time
porosity
local porosity distribution
relaxation parameter
surface-tension perturbation constant

Subscripts
0 = used to indicate typical unit for conversion from lattice
units to physical units
i = lattice direction
nw = nonwetting
P = physical
w = wetting
x, y, z = spatial direction
 = phase
Acknowledgments
The authors thank Numerical Rocks for permission to publish this
paper. This study was partly financed by the Norwegian Research
Council through PETROMAKS grant Towards a Digital Core Laboratory. A substantial amount of the computing time was granted
by NOTUR. Thomas Ramstad thanks Olav Aursj and Henning
Knudsen at the University of Oslo for very helpful discussions.
References
Adler, P.M., Jacquin, C.G., and Quiblier, J.A. 1990. Flow in simulated porous
media. Int. J. Multiphase Flow 16 (4): 691712. doi: 10.1016/03019322(90)90025-E.
Arns, C.H., Knackstedt, M.A., Pinczewski, V., and Garboczi, E.J. 2002.
Computation of linear elastic properties from microtomographic
images: Methodology and agreement between theory and experiments.
Geophysics 67 (5): 13961405. doi: 10.1190/1.1512785.
Arns, C.H., Knackstedt, M.A., Pinczewski, V., and Martys, N.S. 2004.
Virtual permeametry on microtomographic images. J. Pet. Sci. Eng. 45
(12): 4146. doi: 10.1016/j.petrol.2004.05.001.
Avraam, D.G. and Payatakes, A.C. 1999. Flow Mechanisms, Relative
Permeability, and Coupling Effects in Steady-State Two-Phase Flow
through Porous Media. The Case of Strong Wettability. Ind. Eng. Chem.
Res. 38 (3): 778786. doi: 10.1021/ie980404o.
Bakke, S. and ren, P.-E. 1997. 3-D Pore-Scale Modeling of Sandstones
and Flow Simulations in the Pore Networks. SPE J. 2 (2): 136149.
SPE-35479-PA. doi: 10.2118/35479-PA.
Bardon, C. and Longeron, D.G. 1980. Influence of Very Low Interfacial
Tensions on Relative Permeability. SPE J. 20 (5): 391401. SPE-7609PA. doi: 10.2118/7609-PA.
Blunt, M.J. 2001. Flow in porous mediapore network models and multiphase flow. Current Opinion in Colloid & Interface Science 6 (3):
197207. doi: 10.1016/S1359-0294(01)00084-X.
Bryant, S.L., Cade, C.A., and Melor, D.W. 1993. Permeability prediction
from geological models. AAPG Bulletin 77 (8): 13381350.
Buick, J.M. and Greated, C.A. 2000. Gravity in a lattice Boltzmann model.
Phy. Rev. E 61 (5): 53075320. doi: 10.1103/PhysRevE.61.5307.
Chen, C., Lau, B.L.T., Gaillard, J., and Packman, A.I. 2009. Temporal
evolution of pore geometry, fluid flow, and solite transport resulting
from colloid deposition. Water Resources Research 45: W06416. doi:
10.1029/2008WR007252.
Chen, S. and Doolen, G.D. 1998. Lattice Boltzmann Method for Fluid
Flows. Annual Review of Fluid Mechanics 30: 329364. doi: 10.1146/
annurev.fluid.30.1.329.
10

Chen, S., Diemer, K., Doolen, G.D., Eggert, K., Fu, C., Gutman, S., and
Travis, B.J. 1991. Lattice gas automata for flow through porous media.
Physica D: Nonlinear Phenomena 47 (12): 7284. doi: 10.1016/01672789(91)90281-D.
Constantinides, G.N. and Payatakes, A.C. 1996. Network simulation of
steady-state two-phase flow in consolidated porous media. AIChE
Journal 42 (2): 369382. doi: 10.1002/aic.690420207.
Dong, H., Fjeldstad, S., Alberts, L., Roth, S., Bakke, S., and ren, P.E.
2008. Pore network modeling on carbonate: a comparative study of
different micro-CT network extraction methods. Presented at the 22nd
International Symposium of the Society of Core Analysts, Abu Dhabi,
UAE, 29 October2 November.
Dullien, F.A.L. 1992. Porous Media: Fluid Transport and Pore Structure,
second edition. New York: Academic Press.
Dunsmuir, J.H., Ferguson, S.R., DAmico, K.L., and Stokes, J.P. 1991.
X-Ray Microtomography: A New Tool for the Characterization of Porous
Media. Paper SPE 22860 presented at the SPE Annual Technical Conference and Exhibition, Dallas, 69 October. doi: 10.2118/22860-MS.
Fatt, I. 1956. The Network Model of Porous Media. I. Capillary Pressure
Characteristics. Trans. AIME, 207: 144.
Ferrol, B. and Rothmann, D. 1995. Lattice Boltzmann simulations of flow
through fontainebleau sandstone. Transport in Porous Media 20 (12):
320. doi: 10.1007/BF00616923.
Goldsmith, H.L. and Mason, S.G. 1963. The flow of suspensions through
tubes: II. Single large bubbles. Journal of Colloid and Interface Science 18: 237261.
Gunstensen, A.K. and Rothmann, D.H. 1992. Lattice-Boltzmann Studies of
Immiscible Two-Phase Flow Through Porous Media. J. Geophys. Res.
98 (B4): 64316441. doi: 10.1029/92JB02660.
Gunstensen, A.K., Rothman, D.H., Zaleski, S., and Zanetti, G. 1991.
Lattice Boltzmann model of immiscible fluids. Phys. Rev. A 43 (8):
43204327. doi: 10.1103/PhysRevA.43.4320.
Guo, Z. and Zhao, T.S. 2005. Finite-difference-based lattice Boltzmann
model for dense binary mixtures. Phys. Rev. E 71 (2): 026701. doi:
10.1103/PhysRevE.71.026701.
Hilfer, R. 1991. Geometric and dielectric characterization of porous media.
Phys. Rev. B 44 (1): 6075. doi: 10.1103/PhysRevB.44.60.
Jiang, Z., Wu, K., Couples, G., van Dijke, M.I.J., Sorbie, K.S., and Ma, J. 2007.
Efficient extraction of networks from three-dimensional porous media.
Water Resour. Res. 43: W12S03. doi: 10.1029/2006WR005780.
Jin, G., Patzek, T.W., and Silin, D.B. 2004. Direct Prediction of the Absolute Permeability of Unconsolidated and Consolidated Reservoir Rock.
Paper SPE 90084 presented at the SPE Annual Conference and Exhibition, Houston, 2629 September. doi: 10.2118/90084-MS.
Jin, G., Torres-Verdin, C., Radaelli, F., and Rossi, E. 2007. Experimental
Validation of Pore-Level Calculations of Static and Dynamic Petrophysical Properties of Clastic Rocks. Paper SPE 109547 presented
at the SPE Annual Technical Conference and Exhibition, Anaheim,
California, USA, 1114 November. doi: 10.2118/109547-MS.
Kang Q., Lichtner P.C., and Zhang D. 2006. Lattice Boltzmann pore-scale
model for multi-component reactive transport in porous media. J. Geophys. Res. 111: B05203. doi: 10.1029/2005JB003951.
Kang Q., Zhang D., and Chen S. 2004. Immiscible displacement in a channel: simulations of fingering in two dimensions. Advances in Water
Resources 27 (1): 1322. doi: 10.1016/j.advwatres.2003.10.002.
Knackstedt, M.A., Arns, C.H., Limaye, A., Sakellariou, A, Senden, T.J.,
Sheppard, A.P., Sok, R.M., and Pinczewski, W.V. 2004. Digital Core
Laboratory: Properties of reservoir core derived from 3D images. Paper
SPE 87009 presented at the SPE Asia Pacific Conference on Integrated
Modelling for Asset Management, Kuala Lumpur, 2930 March. doi:
10.2118/87009-MS.
Koplik, J. and Lasseter, T.J. 1985. Two-Phase Flow in Random Network
Models of Porous Media. SPE J. 25 (1): 89100. SPE-11014-PA. doi:
10.2118/11014-PA.
Lallemand, P. and Luo, L.-S. 2000. Theory of the lattice Boltzmann method:
Dispersion, dissipation, isotropy, Galilean invariance, and stability. Physical Review E 61 (6): 65466562. doi: 10.1103/PhysRevE.61.6546.
Latva-Kokko, M. and Rothmann D.H. 2005. Diffusion Properties of Gradient-Based Lattice Boltzmann Models of Immiscible Fluids. Phys. Rev. E
71 (5): 056702. doi: 10.1103/PhysRevE.71.056702.
Lerdahl, T.R., ren, P.E., and Bakke, S. 2000. A Predictive Network Model
for Three-Phase Flow in Porous Media. Paper SPE 59311 presented at
2010 SPE Journal

the SPE/DOE Improved Oil Recovery Symposium, Tulsa, 35 April.


doi: 10.2118/59311-MS.
Li, H., Pan, C., and Miller, C.T. 2005. Pore-Scale Investigation of Viscous
Coupling Effects for Two-Phase Flow in Porous Media. Physical
Review E 72 (2): 026705. doi: 10.1103/PhysRevE.72.026705.
Lindquist, W.B., Lee, S.-M., Coker, D.A., Jones, K.W., and Spanne, P. 1996.
Medial axis analysis of void structure in three-dimensional tomographic
images of porous media. J. Geophys. Res. 101 (B4): 82978310. doi:
10.1029/95JB03039.
Maier, R.S., Bernard, R.S., and Grunau, D.W. 1996. Boundary conditions
for the lattice Boltzmann method. Physics of Fluids 8 (7): 17881801.
doi: 10.1063/1.868961.
Manwart, C., Aaltosalmi, U., Koponen, A., Hilfer, R., and Timonen J. 2002.
Lattice-Boltzmann and finite difference simulations for the permeability for three-dimensional porous media. Phys. Rev. E 66 (1): 016702.
doi: 10.1103/PhysRevE.66.016702.
Martys, N.S. and Chen, H.D. 1996. Simulation of multicomponent fluids
in complex three-dimensional geometries by the lattice Boltzmann
method. Phy. Rev. E 53 (1): 743750. doi: 10.1103/PhysRevE.53.743.
Mason, G. and Morrow N.R. 1991. Capillary Behavior of a Perfectly Wetting Liquid in Irregular Triangular Tubes. Journal of Colloid and Interface Science 141 (1): 262274. doi: 10.1016/0021-9797(91)90321-X.
Nardi, C., Lopez, O., ren, P.E., Held, R., and Petersen, E.B. Jr. 2009.
Pore-Scale Modeling of Three-Phase Flow: Comparative Study with
Experimental Reservoir Data. Paper SCI2009-30 presented at the
International Symposium of the Society of Core Analysts, Noordwijk,
The Netherlands, 2730 September.
Olson, J.F. and Rothman, D.H. 1997. Two-fluid flow in sedimentary rock:
simulation, transport and complexity. Journal of Fluid Mechanics 341:
34370. doi: 10.1017/S0022112097005533.
ren, P.-E., Bakke, S., and Arntzen, O.J. 1998. Extending Predictive Capabilities to Network Models. SPE J. 3 (2): 324336. SPE-52052-PA.
doi: 10.2118/52052-PA.
ren, P-E. and Bakke, S. 2002. Process based reconstruction of sandstones
and prediction of transport properties. Transport in Porous Media 46
(23): 311343. doi: 10.1023/A:1015031122338.
ren, P-E. and Bakke, S. 2003. Reconstruction of Berea Sandstone and
Pore Scale Modeling of Wettability Effects. J. Pet. Sci. Eng. 39 (34):
177199.
ren, P-E., Antonsen, F., Ruesltten, H., and Bakke, S. 2002. Numerical
Simulation of NMR Responses for Improved Interpretations of NMR
Measurements in Reservoir Rocks. Paper SPE 77398 presented at the
SPE Annual Technical Conference and Exhibition, San Antonio, Texas,
USA, 29 September2 October. doi: 10.2118/77398-MS.
ren, P-E., Bakke, S., and Held, R. 2007. Direct pore-scale computation of
material and transport properties for North Sea reservoir rocks. Water
Resour. Res. 43: W12S04. doi: 10.1029/2006WR005754.
ren, P-E., Bakke, S., and Ruesltten, H.G. 2006. Digital core laboratory:
Rock and flow properties derived from computer generated rocks.
Presented at the 20th International Symposium of the Society of Core
Analysts, Trondheim, Norway, 1216 September.
Pan, C., Hilpert, M,. and Miller, C.T. 2004. Lattice-Boltzmann simulation
of two-phase flow in porous media. Water Resour. Res. 40: W01501.
doi: 10.1029/2003WR002120.
Piri, M.D. and Blunt, M. 2005. Three-dimensional mixed-wet random porescale network modeling of two- and three-phase flow in porous media. II.
Results. Phys. Rev. E 71 (2): 026302. doi: 10.1103/PhysRevE.71.026302.
Pride, S.R., Flekky, E.G., and Aursj, O. 2008. Seismic stimulation for enhanced
oil recovery. Geophysics 73 (5): O23O35. doi: 10.1190/1.2968090.
Rothmann, D.H. and Keller, J.M. 1988. Immiscible cellular-automaton fluids. Journal of Statistical Physics 52 (34): 11191127. doi:
10.1007/BF01019743.
Rothmann, D.H. and Zaleski, S. 1997. Lattice-Gas Cellular Automata:
Simple Models of Complex Hydrodynamics. Cambridge, UK: Cambridge University Press.

2010 SPE Journal

Schaap, M.G., Porter, M.L., Christensen, B., and Wildenschild, D. 2007.


Comparison of pressure-saturation characteristics derived from computed tomography and lattice Boltzmann simulations. Water Resour.
Res. 43: W12S06. doi: 10.1029/2006WR005730.
Shan, X. and Chen, H. 1993. Lattice Boltzmann model for simulating flows
with multiple phases and components. Phys. Rev. E 47 (3): 18151819.
doi:10.1103/PhysRevE.47.1815.
Shan, X. and Chen, H. 1994. Simulation of nonideal gases and liquid-gas
phase transitions by the lattice Boltzmann equation. Phys. Rev. E 49
(4): 29412948. doi: 10.1103/PhysRevE.49.2941.
Sheppard, A.P., Sok, R.M., and Averdunk, H. 2005. Improved Pore Network Extraction Methods. Paper SCA2006-P089 presented at the 19th
International Symposium of the Society of Core Analysts, Toronto,
Canada, 28 August.
Silin, D.B. and Patzek, T.W. 2006. Pore space morphology analysis using
maximal inscribed spheres. Physica A: Statistical and Theoretical
Physics 371 (2): 336360. doi: 10.1016/j.physa.2006.04.048.
Spanne, P., Thovert, J.F., Jacquin, C.J., Lindquist, W.B., Jones, K.W., and
Adler, P.M. 1994. Synchrotron Computed Microtomography of Porous
Media: Topology and Transports. Phys. Rev. Lett. 73 (14): 20012004.
doi: 10.1103/PhysRevLett.73.2001.
Succi, S. 2001. The Lattice Boltzmann Equation for Fluid Dynamics and
Beyond. Oxford, UK: Numerical Mathematics and Scientific Computation, Oxford University Press.
Succi, S., Foti, E., and Higuera, F. 1989. Three-Dimensional Flows in
Complex Geometries with the Lattice Boltzmann Method. Europhysics
Letters 10: 433. doi: 10.1209/0295-5075/10/5/008.
Svirsky, D.S., van Dijke, M.I.J., and Sorbie, K.S. 2007. Prediction of ThreePhase Relative Permeabilities Using a Pore-Scale Network Model
Anchored to Two-Phase Data. SPE Res Eval & Eng 10 (5): 527538.
SPE-89992-PA. doi: 10.2118/89992-PA.
Swift, M.R., Osborn, W.R., and Yoemans, J.M. 1995. Lattice Boltzmann
Simulation of Non-Ideal Fluids. Phys. Rev. Lett. 75: 830833. doi:
10.1103/PhysRevLett.75.830.
Valvatne, P.H. and Blunt, M.J. 2004. Predictive pore-scale modelling of
two-phase flow in mixed wet media. Water Resources Research 40:
W07406. doi: 10.1029/2003WR002627.
Wu, K., Van Dijke, M.I.J., Couples, G.D., Jiang, Z., Ma, J., Sorbie, K.S.,
Crawford, J., Young, I., and Zhang, X. 2006. 3D Stochastic Modelling
of Heterogeneous Porous MediaApplications to Reservoir Rocks.
Transport in Porous Media 65 (3): 443467. doi: 10.1007/s11242006-0006-z.
Yeoung, C.L.Y. and Torquato, S. 1998. Reconstructing random media.
Phys. Rev. E 57 (1): 495506. doi: 10.1103/PhysRevE.57.495.
Yiotis, A.G., Psihogios, J., Kainourgiakis, M.E., Papaioannou, A., and Stubos, A.K. 2007. A lattice Boltzmann study of viscous coupling effects
in immiscible two-phase flow in porous media. Colloids and Surfaces
A: Physicochemical and Engineering Aspects 300 (12): 3549. doi:
10.1016/j.colsurfa.2006.12.045.
Thomas Ramstad is a research scientist at Numerical Rocks. He
holds a PhD degree in physics from the Norwegian University
of Science and Technology (NTNU). His field of work relates to
computational physics and modeling of multiphase flow in
porous media. He has coauthored approximately 10 papers
on these topics. Pl-Eric ren is the technical director and
cofounder of Numerical Rocks. He holds a BS degree in chemical engineering from the University of Michigan and a PhD
degree in petroleum engineering from the University of New
South Wales. Previously, he worked as a research specialist with
Statoil. Stig Bakke is a senior geological adviser and cofounder
of Numerical Rocks. His research interests include reservoir geology, numerical modeling of geological processes, and 2D and
3D image analysis. Before joining Numerical Rocks, he worked
at Statoil Research Centre and at IKU/SINTEF in Trondheim.
Bakke holds an MS degree in geology from NTNU.

11

Вам также может понравиться