Upload
anoop-uchagawkar
View
212
Download
0
Embed Size (px)
Citation preview
8/15/2019 Transport in Highly Heterogeneous Porous Media-From Direct Simulation to Macroscale Two Equation or Mixed Model
1/6
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/228645799
Transport in Highly Heterogeneous PorousMedia: From Direct Simulation to Macro-Scale
Two-Equation Models or Mixed Models
ARTICLE in CHEMICAL PRODUCT AND PROCESS MODELING · JUNE 2008
DOI: 10.2202/1934-2659.1130
CITATIONS
7
READS
48
2 AUTHORS:
Debenest Gerald
Institut de Mécanique des Fluides de Toulo…
114 PUBLICATIONS 355 CITATIONS
SEE PROFILE
Michel Quintard
Institut de Mécanique des Fluides de Toulo…
564 PUBLICATIONS 4,393 CITATIONS
SEE PROFILE
Available from: Michel Quintard
Retrieved on: 17 March 2016
https://www.researchgate.net/?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_1https://www.researchgate.net/profile/Michel_Quintard?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_7https://www.researchgate.net/institution/Institut_de_Mecanique_des_Fluides_de_Toulouse?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_6https://www.researchgate.net/profile/Michel_Quintard?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_5https://www.researchgate.net/profile/Michel_Quintard?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_4https://www.researchgate.net/profile/Debenest_Gerald?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_7https://www.researchgate.net/institution/Institut_de_Mecanique_des_Fluides_de_Toulouse?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_6https://www.researchgate.net/profile/Debenest_Gerald?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_5https://www.researchgate.net/profile/Debenest_Gerald?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_4https://www.researchgate.net/?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_1https://www.researchgate.net/publication/228645799_Transport_in_Highly_Heterogeneous_Porous_Media_From_Direct_Simulation_to_Macro-Scale_Two-Equation_Models_or_Mixed_Models?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_3https://www.researchgate.net/publication/228645799_Transport_in_Highly_Heterogeneous_Porous_Media_From_Direct_Simulation_to_Macro-Scale_Two-Equation_Models_or_Mixed_Models?enrichId=rgreq-dfb00943-29c7-481c-b2fd-3c157b27a3d7&enrichSource=Y292ZXJQYWdlOzIyODY0NTc5OTtBUzoxMDIzNjMxNDI1NTc3MDhAMTQwMTQxNjY4MzcyMA%3D%3D&el=1_x_2
8/15/2019 Transport in Highly Heterogeneous Porous Media-From Direct Simulation to Macroscale Two Equation or Mixed Model
2/6
Transport in Highly Heterogeneous Porous Media: From Direct
Simulation to Macro-Scale Two-Equation Models or Mixed Models
G. Debenest1 and M. Quintard
*,1
1 Institut de Mécanique des Fluides de Toulouse*Corresponding author: Institut de Mécanique des Fluides de Toulouse, 4 allée C. Soula, 31400 Toulouse
Abstract: Flows in highly heterogeneous mediaare found in many practical fields, such as
hydrology, petroleum engineering, chemical
engineering. The case of two-region
heterogeneous media (fractured media, catalytic
beds, etc.) plays a fundamental role. The
different questions associated to this specificcase are illustrated in this paper for two different
kinds of transport: (i) flow of a slightly
compressible fluid, (ii) dispersion of a tracer.
Many different theoretical models are
implemented under COMSOL, using many of
the original features of the software. These
models correspond to direct simulation and
macro-scale or large-scale models such as fully
averaged models or mixed models.
Keywords: porous media, heterogeneous
system, double-porosity, mixed models
1. Introduction
Flow in highly heterogeneous media have
received a considerable interest in the scientific
literature since all natural media, such as the one
appearing in petroleum reservoir engineering or
hydrogeology, features more or less importantheterogeneity. Heterogeneity is known to lead to
abnormal transport mechanisms. Here, the word
abnormal means that classical convection-
dispersion equations cannot be used reliably [1].
A particular class of heterogeneous systems
shows easily this abnormal behavior: two-region
systems, often called also double-porosity
systems. Examples of such systems are given in
Fig. 1. The left picture corresponds to a fractured
system in which the ω-region permeability isseveral orders of magnitude lower than the
permeability of the η-region. Advection is themajor transport mechanisms in the fractures,
while diffusion is dominant in the matrix blocks.
These systems are also called immobile-mobile
systems. If the ratio of permeability is closer
than 1, some advection may occur also through
the ω-region, and the resulting systems are calledmobile-mobile system.Of course, the Darcy-scale equations relevant to
macro-scale description may be solved directly
to capture the transport phenomena within such
systems. This poses specific numerical problems
especially in the case of fracture systems forwhich the volume fraction of the η-region,
denoted φ η , is small. The major problem,however, for real-scale applications, is that direct
simulation is limited to a small number of matrix blocks and fractures.
ω
η
Figure 1. Example of double-porosity systems
Alternative models must be sought, which can be
divided into two different classes. The first class
corresponds to fully averaged models, or two-
equation models. Transport equations in both
regions are homogenized over a representative
volume of the heterogeneous system [2]. Hence,
the numerical model requires only two large-
scale grids, for each homogenized region, whichis efficient in terms of computation. An
exchange term must be implemented to connect
the two sets of large-scale equations. One
understands that this single exchange terms may
have difficulties in representing all the detailed
behavior associated to the exchange between the
fractures and the many blocks within a numerical
cell. If one wants to keep some of these details,
an alternative model, called mixed model, may
be used. It is based on a fully averaged equation
associated to the η-region, coupled to a smaller-scale model with the original equations (hence
Debenest, G. and Quintard, M., 2006. Transport in Highly Heterogeneous Porous Media:
From Direct Simulation to Macro-Scale Two-Equation Models or Mixed Models, COMSOL
2006. COMSOL, Paris, pp. 1-5.
8/15/2019 Transport in Highly Heterogeneous Porous Media-From Direct Simulation to Macroscale Two Equation or Mixed Model
3/6
the name mixed model). Those different
strategies are schematically represented in Fig. 2.
{{
{
Direct model
Two-equation model
Mixed model
Figure 2. Schematic representation of the threedifferent models
The resulting PDEs have many different and
interesting features that may be solved nicely
using the different approaches offered in
COMSOL. This is illustrated in this paper.
2. Direct simulation
The problems posed by direct simulations are
two folds: (i) high contrast of properties, (ii)
small volume fraction of one of the region in the
case of fractured media. If one considers, for
instance, the flow of a slightly compressible fluid
over the system represented Fig. 3, the equationsto be solved are:
1 P c P
t
βη
η βη βη
β
∂
∂ µ
= ∇ ⋅ ⋅ ∇
K (1)
and a similar equation in the other region, with
continuity of flux and pressure over all interfaces
between matrix blocks and fractures. This
transport equation for the fractures can easily be
transformed into a PDE over a fracture seen as a
line (in 2D) or plane (in 3D), we have
exchange with surroundingmatrix blocks
1
1 .
f f
P
h c h P t
P
βη
η βη βη β
ηω βω βω
β
∂
∂ µ
µ
= ∇ ⋅ ⋅ ∇
+ ⋅∇n
K
K(2)
where derivatives are taken along the fractures.
This can be solved nicely in COMSOL using the
application mode “Weak Form, Boundary” (see
for instance the application to adsorption
problems in COMSOL manuals). An example of
the obtained pressure field at a given time is
shown Fig. 4.
Figure 3. Simulation over a fractured system.
Figure 4. Pressure field (a well is located at the
center)
In this paper, we consider the flow of one phase
together with the dispersion of a tracer. For now
on we focus on the tracer dispersion problem, i.e.
Darcy-scale equations of the type
( )*t
C C C η
η η η η η η ε ε ∂ + ⋅ ∇ = ∇ ⋅ ⋅∇
∂V D (3)
The velocity field, Vη, will be computedindependently from Darcy’s law.
As an illustration, we will consider the system
represented Fig. 5 (after [3]). Low permeability
nodules are scattered through a highly permeable
medium. The direct dispersion problem is solved
and the concentration field at a given time is
represented Fig. 5. The necessary PDEs are
directly available in COMSOL.
8/15/2019 Transport in Highly Heterogeneous Porous Media-From Direct Simulation to Macroscale Two Equation or Mixed Model
4/6
We now discuss the two different large-scale
models that may be associated to this system,i.e., the two-equation model and the mixed
model.
Figure 5. Concentration field at t =10000s field(injection from the left, initial zero concentration)
3. Two-equation model
The two-equation dispersion model is
characterized by the following equations [2]
( )
( )
*
* * ** *
* **
*
* * ** *
* **
t
...
t
...
C C C
C C
C C C
C C
η
η η η η ηη η
η ω
ω
ω ω ω ω ωω ω
η ω
ε ϕ
α
ε ϕ
α
∂ + ⋅ ∇ = ∇ ⋅ ⋅ ∇ ∂
− +−
∂ + ⋅ ∇ = ∇ ⋅ ⋅ ∇ ∂
+ +−
V D
V D
(4)
in which the concentrations are those of a
representative volume of the heterogeneous
system. In the case of the problem represented
Fig. 5, the large-scale equations may be solved in
1D, i.e., the Darcy-scale 2D problem is replaced
at the large-scale by two 1D equations. These
equations can be solved easily in COMSOL.
A more interesting aspect for COMSOL
application is the calculation of the effective properties in Eq. (4) from a representative Unit
Cell of the porous medium. The PDEs to be
solved are given in the Appendix. In particular,
the mass exchange coefficient, α *, in Eqs. (4)requires the solving of an integro-differential
system for a r -field, which in facts links the
concentration deviation to the difference in the
large-scale concentrations, i.e., ( )* *C C η ω − .This is beyond the scope of this paper to give
theoretical details, and we refer the reader to [2]
for a complete explanation. Two computational
difficulties can be found in these problems: (i)
the use of periodicity boundary conditions, (ii)
the existence of an integral involving the
solution as a source term (the term α *). Thosetwo features are directly available from
COMSOL! Figs. 6 and 7 show an example of
solution (the unit cell is simply a single nodule
within the η-region) in the case of a diffusiveregime or a more convective regime.
Figure 6. r-field (diffusive regime)
Figure 7. r-field (convective regime)
4. Mixed model
The equations for the mixed model are the
following. The transport equation in the η-regionis homogenized, we have
( ) ( )*
* * ** *
*
exchange with ω
1 .
A
C C C
t
C dAV
ωη
η
η η η η η η η η
ηω ω ω ω
ϕ ε ϕ ε
ε ∞
∂+ ∇ ⋅ = ∇ ⋅ ⋅ ∇
∂
+ ⋅ ∇∫
V D
n D
(5)
8/15/2019 Transport in Highly Heterogeneous Porous Media-From Direct Simulation to Macroscale Two Equation or Mixed Model
5/6
where we need to compute the flux exchanged
with the embedded nodules. For our particular
application, this equation will be solved in 1D.The exchanged flux will be calculated by solving
several independent Darcy-scale problems over
several nodules. These problems are given by
( )*t
C C ω ω ω ω ω ε ε
∂= ∇ ⋅ ⋅ ∇
∂D (6)
within each nodule, and a mixed boundary
condition
*
Darcy-scalelarge-scale
atC C Aη ω ωη = (7)
which connects the (here) 2D Darcy-scale
problem to the 1D large-scale problem.
Solution is carried out in the following manner,using the multi-domain feature of COMSOL,
and the possibility to exchange information
between the two domains. Eq. (5) is
implemented in a 1D domain, with source terms
coming from the flux calculation at the nodule
boundaries and Eqs. (6) and (7) are solved other
10 nodules, as shown in Fig. 8. The arrows
illustrate the links between the two domains.
Figure 8. The two-domain representation of the mixedmodel.
Fig. 9 shows the 1D concentration field for the
η-region, and the concentration fields within thenodules, at a given time.
Figure 9. One example of mixed model concentrationfields.
Finally, it is interesting for the physicist to
compare the results obtained from the different
models. As an example, we will consider thetime evolution of the exit concentration, as
shown in Fig. 10.
0.001
0.101
0.201
0.301
0.401
0.501
0.601
0.701
0.801
0.901
1 201 401 601 801
time (minutes)
C o n c e n t r a t i o n
direct model
two-equationmodel
mixed model
Figure 10. Exit concentration versus time for the
different models.
We see that the slightly more complex mixed
models offer a better representation than the two-
equation model. In terms of computational
resources, of course, the direct computation is
very heavy, the mixed model still requires 2D or
3D calculations, while the two-equation model
does not require detailed Darcy-scale
simulations. The choice between the different
possibilities will be made depending on the
required accuracy, and the computational
requirements.
7. Conclusions
In this paper, we have provided an example
of the use of COMSOL for a comprehensive
analysis of transport through two-region
heterogeneous porous systems. Many questions
are raised from a physicist point of view when
considering these problems. In particular, which
large-scale models may be used to avoid a heavy
direct computation of the Darcy-scale equations?Answering these questions requires a thorough
theoretical analysis. The analysis leads to several
sets of PDEs, which have many peculiarities
which are often a problem in terms of
computational implementation: integro-differential equations, periodicity conditions,
coupling between domains of different
dimensions, …
It is remarkable that all these difficulties
were solved readily with COMSOL through the
8/15/2019 Transport in Highly Heterogeneous Porous Media-From Direct Simulation to Macroscale Two Equation or Mixed Model
6/6
available tools, without the need of special
developments.
8. References
1. Cushman, J.H. and Ginn, T.R., 1993. Nonlocal
Dispersion in Media with Continuously Evolving
Scales of Heterogeneity. Transport in Porous
Media, 13: 123-138.
2. Cherblanc, F., Ahmadi, A. and Quintard, M.,2003. Two-medium description of dispersion in
heterogeneous porous media: Calculation of
macroscopic properties. Water Resources Res.,
39(6): SBH 6-1:6-20.3. Zinn, B. et al., 2004. Experimental
visualization of solute transport and mass
transfer processes in two-dimensional
conductivity fields with connected regions of
high conductivity. Environ Sci Technol., 38(14):
3916-3926.
9. Acknowledgements
This work has received partial support from
INSU/CNRS, and Institut Français du Pétrole.
10. Appendix
The “closure problem” ([2]) to be solved inorder to calculate the effective properties of
the two-equation model is given for two
unknown fields r η and r ω by
( ) ( ) 1 *r r η η η η η ϕ α ∗ −∇ ⋅ = ∇ ⋅ ⋅ ∇ −V D
r r at Aη ω ηω= + 1 ,
atr r Aηω η η ηω ω ω ηω ∗ ∗⋅ ⋅ ∇ = ⋅ ⋅∇n D n D
( ) ( )1r r ω ω ω ω ω ϕ α ∗ − ∗∇ ⋅ = ∇ ⋅ ⋅ ∇ +V D
Periodicity:
( ) ( ) , ( ) ( ) , 1, 2, 3i ir r r r iη η ω ω + = + = =r r r rℓ ℓ
r r ηη
ωω
n s l q= =0 0,
where
( )* *1
A
r r dA
V ωη ηω η η η η α
∞
= − ⋅ − ⋅∇∫ n V D
Similar problems are available to calculate the
large-scale dispersion tensors.