paper

© All Rights Reserved

0 views

paper

© All Rights Reserved

- Analysis of Male and Female Handwriting
- Week4 Homeawork
- S Essentials General Mathematics Advanced 11
- Digital Signal Processing Matlab
- 8 TCCS4 Task F Georgescu-Modelling and Simulation of CO2 Leakage Mechanisms Version2[1]
- ECON 262-Mathematical Applications in Economics-Kiran Arooj
- JEE Advanced Syllabus by Entranceindia
- Matlab Tutorial
- hw4
- SPE-23575-MS - Slug Test Method in Pressure Sensitive Reservoirs
- Group Theory
- RDT Reservoir Description Tool
- Maths_SQP.pdf
- Production Technology Wellflo Project (1)
- Schelkunoff Theorems for Linear Arrays
- PETSOC-86!04!02 Skin Factor and Flow Efficiency in Wells Pro
- Octave 3 2 4 Documentation PDF
- V3I8_IJERTV3IS080539
- A Simple Technique for Calculating Natural Frequencies of Geometrically Nonlinear Prestressed Cable Structures
- BSIT 42Lab[1] Revised

You are on page 1of 14

Media

M. Tene, Delft University of Technology; M. S. Al Kobaisi, The Petroleum Institute; H. Hajibeygi, Delft University

of Technology

This paper was prepared for presentation at the SPE Reservoir Simulation Symposium held in Houston, Texas, USA, 2325 February 2015.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents

of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect

any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written

consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may

not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract

A general algebraic multiscale framework is presented for fractured porous media, which enables the

treatment of fractures with multiple length scales and wide ranges of conductivity contrasts. To this end,

fully integrated local basis functions for both matrix and fractures are constructed. These basis functions

are employed to construct the multiscale coarse system for both matrix and fractures, and then interpolate

the coarse solutions back to the fine-scale reference system. Combined with a second stage fine-scale

solver, here, ILU(0), our development leads to an iterative multiscale strategy for heterogeneous fractured

media, allowing for error reduction to any arbitrary level, while honoring mass conservation after any

multiscale finite volume stage. In order to maintain generality, it is shown that when each fracture network

is modeled using a single coarse grid cell, our formulation automatically reduces to that proposed by

Hajibeygi et al. (2011). However, in order to facilitate the treatment of general ranges of conductivity

contrasts and deviation term norms, we introduce the ability to refine the coarse grid to more than one cell

per network. Our experiments show that this flexibility results in a significant improvement on convergence properties. This added degree of flexibility is made possible through an algebraic formulation,

which leads to a multi-stage multiscale conservative linear solver for multiphase flow in fractured media.

Introduction

Mathematical formulations describing flow in hydrocarbon reservoirs entail highly heterogeneous anisotropic static and dynamic coefficients, which span several orders of magnitude over their entire length

scales (See, e.g., Aziz and Settari (2002)). Accurate treatment of these parameters using state-of-the-art

reservoir simulators is not possible due to their grid resolution (i.e., computational capability) limits. In

addition, many of the worlds hydrocarbon reserves are fractured, either naturally or as a result of artificial

fracturing processes. And, while the majority of the hydrocarbons reside in the matrix blocks, the fractures

act as lower dimensional highly conductive channels, with a significant impact on flow and transport of

fluids. As such, the presence of fractures in heterogeneous media is considered as an additional challenge

to the simulators, especially due to their scale and physical characteristics, which are significantly

contrasting with those of the matrix. As a result, efficient and accurate simulation of multiphase flow in

heterogeneous fractured porous media is a challenging task, yet crucial for several important applications,

SPE-173200-MS

including those of Enhanced and Improved Oil Recovery (EOR and IOR, respectively) and Carbon

Capture and Storage (CCS). Also note that fractures are typically of a wide range of length scales, and

their structure exhibits complex patterns: they can become tangential or cross each other at different

points, forming networks. And, their shapes can change dramatically between different reservoir layers.

Therefore, the explicit treatment of all fractures present in the entire large-scale reservoir (order of km)

is computationally intractable, as for the computational and grid generator limits.

In order to resolve these limitations, Lee et al. (2000) proposed a hierarchical modeling approach. In

their approach fractures were classified based on their length scale. The small and medium fractures

(typically, the size of one matrix grid cell) are homogenized within the matrix, altering its effective

permeability. The large-scale fractures, however, are explicitly treated in the discretized system. This

approach allows for the separation of matrix and fracture fluxes, which in turn, enables the independent

specification of computational grids for the two domains.

Even though the homogenization of small-scale fracture leads to a significant reduction in the overall

Degrees Of Freedom (DOF), the flow problem remains still challenging, even for the most efficient

numerical solvers. This makes multiscale methods (See, e.g., Cortinovis and Jenny (2014); Efendiev and

Hou (2009); Hajibeygi and Tchelepi (2014); Hajibeygi et al. (2012); Hou and Wu (1997); Hou et al.

(1999); Jenny et al. (2003); Kippe et al. (2008); Lee et al. (2008); Zhou and Tchelepi (2012); Zhou et al.

(2011) and referenced therein) quite attractive options for next-generation reservoir simulation. The

application of multiscale methods to fractured porous media was first proposed by Hajibeygi et al. (2011).

They developed a sequential strategy on the basis of expressing the pressure inside fracture networks in

terms of an average value and fine-scale deviation terms. As such, the matrix grid block pressure could

be coupled to the average pressure for each fracture network and solved for implicitly using the Multiscale

Finite Volume (MSFV) method. Then, the deviation terms were treated in a separate second stage.

Therefore, as shown in their numerical experiments, their method was efficient for media with highly

conductive fracture networks which span short spatial length scales (relative to that of the domain).

However, for test cases where the average network pressure is not a good zero-order guess for the whole

fracture pressure distribution, the convergence rate of their method would degrade.

In this work, we propose a general and novel multiscale method for hierarchical fracture modeling in

large-scales heterogeneous hydrocarbon reservoirs. Given a fine-scale discrete system, which describes

flow in the matrix (prescribed with effective parameters, after the homogenization of small scale fractures)

and fractures (with lengths larger than a fine grid cell size), our method constructs an appropriate

coarse-scale solution, on the basis of coarsening strategies (i.e., multiscale superposing expressions) for

both matrix and fracture domains.

In order to recover the fine-scale solution, a new prolongation operator is assembled, which contains

modified multiscale basis functions to capture heterogeneities, as well as fully-integrated fracture and well

basis functions, which account for external mass transfers, in a conservative fashion. The basis functions

are solved locally over dual-coarse grid blocks, similar to the original MSFE of, e.g., Hou and Wu (1997)

and MSFV of Jenny et al. (2003), having the fracture and well coarse nodes set to 0. Well functions are

computed by setting the well pressure to 1, while all other coarse nodes to 0, over perforated dual-coarse

blocks (similar as Jenny and Lunati (2009), but here algebraically formulated and implemented for 3D

problems). The novel development of this work, i.e., general fracture functions (interpolators inside

fracture domains), on the other hand, are first localized within their respective fracture networks,

neglecting matrix-fracture couplings, and setting the pressure in the corresponding fracture coarse node

to 1. These solutions are used as Dirichlet condition to calculate fracture basis functions inside perforated

matrix dual-coarse grid blocks.

Our novel multiscale method for fractured media reduces in the limit to that of Hajibeygi et al. (2011)

if only one fracture coarse node is considered. In this case, the basis function for fracture domain reduces

to constant 1 along the fracture network, thus it employs only one additional DOF per fracture network

SPE-173200-MS

implicitly coupled within the coarse-scale matrix equation. More importantly, when more fracture coarse

nodes are prescribed, the fracture pressure is shown to be captured more accurately, also leading to

significant improvements for the solvers convergence rate. Our numerical results show that only a few

DOF per network are necessary to obtain a reasonable tradeoff between convergence rate and computational complexity.

Following the traditional algebraic multiscale formulation, our method is general, in the sense that it

can be easily extended to account for complex physics, such as gravity, as in Wang et al. (2014), or

compressibility (Tene et al. 2014). Furthermore, it allows for the direct implementation as a back-box

procedure in existing reservoir simulators.

The paper is structured as follows. First we study the fine-scale discrete system describing flow in

fractured media using the hierarchical fracture modeling approach. Then, in Section 3, our general

multiscale method for fractured media is presented. Section 4 consists of numerical results on both 2D and

3D test cases, and, finally, conclusions and plans for future research are drawn in Section 5.

The single-phase flow equations for fractured media can be written as

(1)

(2)

where superscripts m and f denote quantities and operators for matrix and fracture media, respectively.

Therefore, km stands for effective matrix permeability (matrix rock and homogenized small-scale fractures), and kf for fractures permeability. Also, and are the fluid density (single-phase flow) and rock

porosity, respectively. These equations are to be solved for matrix and fracture pressures pm and pf, on

matrix m and fracture f domains. Note that fractures are in a lower dimensional space, i.e. f c Rn-1,

than matrix (reservoir rock) n c Rn. Also, qmw and qfw are the matrix and fracture external source terms

due to injection/production wells, respectively, described by Peaceman (1977) as, e.g., for matrix

(3)

where productivity index is PI. Similarly, the fracture-matrix coupling term, i.e., qfm is formulated as

(4)

which is based on a linear pressure distribution assumption between the two overlapping fracture and

matrix grid cells. The Qm and Qf terms describe other external source terms for matrix and fractures (e.g.,

gravity terms). The CI is the connectivity index, defined on a discrete system (grid cells) and bears

similarity to PI, which is the productivity index in the Peaceman well model. As described by Hajibeygi

et al. (2011), the CIij is defined as

(5)

where Aij is the area fraction of fracture element i overlapping with the matrix element j, and is the

average distance of the two elements. The k* is the effective permeability (weighted harmonic average)

between the overlapping fracture and matrix nodes. For highly conductive fractures, it becomes nearly km,

while for impermeable fractures it becomes close to kf.

The main advantage of this type of formulation is that the matrix and fracture grids are totally

independent. Their formulations also can be described easily by considering appropriate physics for each

medium, with high applicability to highly fractured reservoirs or dynamically generated (and closed)

SPE-173200-MS

models, e.g., geothermal and tight-gas/oil formations modeling and simulations, (i.e., when geomechanics effects are considered).

Equations (1) and (2) are discretized over fine-grid cells based on the resolution of the heterogeneous

permeability and small-scale fractures present in the domain. When nonlinearities are present (e.g.,

compressible flows, where the density and porosity terms are pressure-dependent), the discrete nonlinear

equations need to be linearized, leading to,

(6)

and then iteratively solved until the converged solution is achieved. Note that system in (6) shows an

implicit treatment of the coupling between fracture and matrix through the Afm entries, and that the matrix

system A can be non-symmetric, due to the compressibility effects, as described by Tene et al. (2014) and

Hajibeygi and Jenny (2009).

Developing an efficient solution strategy for the linear system (6) is quite challenging for several

reasons. On the one hand, the size of this system can exceed 107 for realistic test cases. On the other hand,

the value of the condition numbers for the system matrix is worsened by high contrasts between reservoir

properties (permeability is highly heterogeneous over large scales).

Clearly a classical upscaling method cannot be employed here due to the highly resolved fractures,

which could have important effect on the mass transport. Multiscale methods have the unique advantage

of being conservative (MSFV), while honoring fine-scale data. They operate by solving coarse-scale

systems, the solution of which will be interpolated (prolonged) back to the original resolution. This would

allow one to calculate the error (residual) in fulfilment of the fine-scale system (6) and employ a few

iterations adaptively, when necessary (see, e.g., Hajibeygi et al. (2012) and Hajibeygi and Jenny (2011)).

with Complex Wells

In this paragraph, we describe how the system (6) is solved using a multiscale finite volume (MSFV),

multiscale finite element (MSFE), or a hybrid multiscale finite element and volume (MSMIX) methods.

Given a reservoir with Nf fracture networks and Nw wells, the Algebraic Multiscale Solver for flows

with Fractures and Wells (Frac-Well-AMS) approximates the matrix and fracture pressure fields as

(7)

and

(8)

respectively, where Ncm and Ncfi are the number of coarse cells for matrix and fracture network i,

,

respectively (Figures 1(b) and 1(c)). Here, basis functions associated with matrix coarse cells

and matrix-well effects

are employed in order to capture the effects of

matrix-fracture effects

m

all the important factors in the matrix pressure field p . Important, and the novel aspect of this work, is

that the pressure field inside fracture field, i.e., pf, is also described with a multiscale formulation (8). This

means that the fracture grid cells are also decomposed into coarse blocks, similar to the matrix, i.e.,

fractures are also mapped to the coarse scale and back to the original resolution. More specifically, each

fracture network f is decomposed into Ncfi coarse grid blocks (Figure 1(c)), for which a basis function is

calculated. Also, one can also use the same formulation for wells, which are typically assigned a single

DOF, due to their homogeneous properties.

SPE-173200-MS

(9)

where

(10)

is the matrix prolongation operator, having three sub-columns of matrix-matrix, matrix-fracture, and

matrix-wells. Also,

(11)

Similarly, for fracture medium, according to Eq. (8), one can define:

(12)

with

SPE-173200-MS

(13)

and

(14)

Starting from the given fine-grid cells (Figure 1(a)), multiscale methods impose a coarse (Figures 1(b))

and dual-coarse grid (Figures 1(d)) (See, e.g., Jenny et al. (2003, 2006)). The former is anon-overlapping

decomposition (with the total number of Ncm), each having a coarse node inside (typically, but not strictly,

the center cell). By connecting the coarse nodes, the dual-coarse grid cells are obtained (with the total

number of Ndm), an overlapping partition of the domain with the overlap width of one sets of fine cells

(boundary cells), which serves as support for the basis functions.

The same strategy can be now proposed for fractures, where for their given fine grid cells, a

non-overlapping coarse partitioning is considered with Ncfi coarse grids for fracture network fi.

Basis functions for matrix domain are computed over matrix dual-coarse grid cells as

(15)

(16)

(17)

For all above equations, zero Drichlet condition is set to at all coarse nodes except the one for which

its basis function is to be calculated. Moreover, all of these basis functions are subject to reduceddimensional boundary conditions. Note that the basis functions are calculated locally, hence will be

updated adaptively if the fine-scale properties change beyond a threshold value as described, e.g., by

Hajibeygi and Jenny (2011).

For fracture medium, similarly, basis functions are obtained as

(18)

Dirichlet condition at fracture coarse nodes xk. It is important to note that, for

subject to

the basis functions of fractured medium, the matrix-fracture and well-fracture coupling is not considered,

in order to ensure localization. Though the coupling terms can be easily considered, and will positively

impact the solution quality, their consideration would lead to additional computational complexity. This

formulation for mapping the fracture pressure from fine to coarse scale results in an efficient and accurate,

yet flexible, multiscale strategy. The detailed coupling between the two media will be obtained as a

high-frequency error in a separate second stage smoothing procedure.

Figure 2 depicts an example of a basis and a fracture function in a dual-coarse block. Notice that the

sum of the basis and fracture basis functions form a partition of unity.

SPE-173200-MS

Figure 2A basis function (a), a fracture function (b), and summation of all four basis functions (c) with fracture functions (d). Note that sum of

all basis and fracture functions (and well functions, if they perforate the shown dual cell) is 1.

Figure 32D test case setup: permeability field (a) and geometries of fractures and wells (b).

Once the prolongation operator is computed, the restriction operator (i.e., map from coarse to fine

scale) needs to be defined. As for having three distinct features in the domain, i.e., matrix, fracture, and

wells, the restriction operator can be defined in much more general form than those previously published

works. We allow for having a full Finite-Volume restriction, as in the classic MSFV, a full Finite Element

method, R PT, or a Mixed Finite Element-Volume, where the matrix is restricted according to Finite

Elements, while the fractures and wells use Finite-Volumes.

Numerical Results

In order to assess the performance of our newly developed multiscale framework for fractured porous

media, we investigate two test cases. The first is a 2D reservoir with one central fracture network,

composed of 60 interconnected segments, and two horizontal wells. In the second scenario, as a proof of

concept, we extrude the 2D fracture network along the Z axis, in order to present the behavior of a 3D

5-spot reservoir in a synthetic manner, i.e., assuming only 1D flow inside fracture plates.

2D test case

For the 2D case, we consider the heterogeneous permeability field depicted in Figure 3(a), traversed by

a rate-constrained injector on the left edge and by a pressure-constrained producer on the right edge,

respectively (Figure 3(b)). There is a fracture network located in the center of the reservoir, whose

permeability was considered to be 2 orders of magnitude higher than the matrix average.

In the fine-scale, there are 4096 matrix cells (64 x 64) and the fracture network was discretized into

60 segments. Using multiscale, we reduce the number of degrees of freedom (DOF) of the single-phase

pressure solution to 64 coarse-scale matrix blocks (each containing 8 x 8 fine cells), accompanied by

SPE-173200-MS

Figure 4 Convergence behavior of the algebraic multiscale fracture solver with different fracture coarsening strategies.

different coarsening strategies for the fracture network. The fracture control volumes are constructed by

first specifying the distribution of the coarse nodes within the fracture network (depicted as green lines

on the left-hand-side plots of Figure 5) and then assigning the remaining segments (red lines) to their

closest nodes, in terms of number of hops (dots).

Consequently, we vary the number of coarse cells for fractures and investigate the performance of the

iterative method. As second stage (smoother), we employ 5 iterations of ILU(0), as an optimum choice

suggested by the systematic studies of Tene et al. (2014) and Wang et al. (2014). We consider subsets of

1, 3, 4, 6, 10 and 15 DOF from the original 60. It is important to note that we do not perform the correction

stage of Hajibeygi et al. (2011), where the fine-scale fracture pressure deviations are determined in a

second stage. In our framework, the implicitly computed fracture functions serve this purpose, instead.

However, in experiments with only 1 DOF, the single fracture function will, by definition, show constant

pressure within its fracture network. As such, these cases must rely on the smoother to account for the

fine-scale deviations.

The right plots in Figure 5 shows the pressure solution after the first multiscale iteration for different

number of fracture coarse grids. For higher numbers of coarse DOF for fractures, the pressure after the

first multiscale stage is already a very good approximation to the fully converged solution and, as such,

only a small number of multiscale iterations is needed to converge. However for 1 to 4 DOF, some

noticeable features are not captured and, hence, extra effort is required to achieve convergence, which is

translated into an increased number of both multiscale and smoother iterations.

Illustrated also in Figure 4, the case of 1 coarse DOF for fracture entails slow convergence rate,

because of its lack of effective interpolation within the fracture network. However, the rate is dramatically

improved when only a few extra DOF are added, since the fracture functions are becoming effective as

interpolators within the network, thus offloading the extra effort from the smoother. It is also interesting

to note that there is a small difference between the experiments with 6 and 10 DOF. Therefore, the number

of fracture coarse nodes should not be the only criterion in developing an effective strategy. Rather, the

positioning of these nodes also plays an important role. Finally, we see that with 15 DOF, we obtain a

convergence behavior quite similar to the case when all 60 DOF are considered. Therefore, considering

the efficiency-accuracy tradeoff, the case with 15 seems quite efficient. Of course the CPU time should

be measured for a more reliable assessment.

SPE-173200-MS

Figure 5Fracture coarsening strategies. The left plots depict the fracture coarse nodes (in green), while the figures on the right show the pressure

solutions after the first multiscale iteration (without smoothing).

10

SPE-173200-MS

Figure 6 Components of the multiscale prolongation operator for a fracture coarsening strategy with 3 DOF: (a-c) three assembled fracture

functions, (d) sum of the basis functions, (e) injector well function, (f) producer well function. Note that each fracture function (a, b, and c) and well

function (e and f) plot is obtained by assembling the locally computed fracture and well functions over matrix dual cells. Also, (a, b, and c) plots

illustrate the fracture functions for matrix medium.

SPE-173200-MS

11

Figure 73D 5-spot test case setup: geometry of the medium with wells (a) and fracture plates (b) are shown. Also shown are the permeability field

(c) and the pressure solution (d). Note that vertically lumped grid cells are considered for fractures, thus, 60 grid cells are used to discretize the

fracture plates along their longitudinal direction.

Figure 6 provides further insights into the way our newly developed multiscale method performs

interpolation for the matrix domain, for the test case at hand. This time, the fracture network is divided

into 3 coarse blocks. Figures 6 (a)-(c) depict the footprints of the corresponding fracture functions on the

matrix, which are assembled after being computed locally over dual-coarse cells. Plot (d) shows the sum

of the basis functions, where we can clearly distinguish the outline of the fracture network, and, finally,

the two well functions are given in subplots (e), for the rate-constrained injector, and (f) for the pressure

constrained producer. As mentioned before, these functions reside in the columns of the multiscale

prolongation operator and it can be clearly seen that, together, they form a partition of unity, which

guarantees the monotonicity of the pressure solution.

3D 5-spot test case

The performance of our devised multiscale framework for a 3D reservoir with 262144 (6x 64 64)

fine-scale grid cells is tested. The fracture network was obtained by extruding those used in the previous

12

SPE-173200-MS

Figure 8 Convergence histories with different fracture coarsening strategies for 3D quarter-five-spot test case. Note that our algebraic multiscale

fracture solver is compared with SAMG in terms of number of iterations.

paragraph along the Z axis (Figure 7), between levels 20 and 44. For the sake of simplicity, we do not

account for the vertical inter fluxes within the fracture plates, i.e., the number of degrees of freedom within

the fracture network remains 60. In other words, vertically lumped grid cells are considered for fractures.

The reservoir has a heterogeneous permeability field and is driven by 5 wells vertical wells: a rateconstrained injector, placed in the middle, and 4 pressure-constrained producers, lated in the corners.

Finally, the steady-state single-phase pressure solution is also depicted in Figure 7.

Figure 8 depicts the convergence history of the multiscale solver using the 3D-equivalent fracture

coarsening strategies to those shown previously, in Figure 5. The same trend as in the previous experiment

is noticed, which means that the proposed multiscale method can efficiently capture the fine-scale solution

using only 15 DOF to represent the flow in the fracture network. For reference, we also show the results

obtained using the SAMG commercial solver, developed by Fraunhofer SCAI (Stuben 2010), which, just

like the multiscale solver, was run in a Richardson loop. The different behavior can be, most likely,

attributed to the matrix heterogeneity which is better captured by SAMGs adaptive coarse grid generation

strategies, whereas for our solver we prescribed a regular coarse grid of 8 8 8 blocks.

Conclusions

In this paper, we proposed a novel general multiscale framework for efficient and accurate simulation of

flow through heterogeneous porous media with complex wells and fractures of various length scales. We

introduced, for the first time, the possibility to prescribe an arbitrary number of degrees of freedom per

fracture network at coarse scale. This was done in order to obtain the desired tradeoff between

convergence properties and computational expense. Such a novel formulation was made possible by the

computation of fully-integrated fracture and well functions, in full agreement with the classical multiscale

algebraic formulations. As such, the algorithm can be directly implemented as a black-box procedure

into the existing and next-generation reservoir simulators, and poses no algorithmic restrictions to

accommodate the inclusion of complex physics such as compressibility, capillarity, gravity and geomechanics. These extensions are the subject of our ongoing research.

SPE-173200-MS

13

Our numerical experiments on 2D and 3D fractured reservoirs revealed that the new method is quite

efficient and flexible, i.e., it can effectively reduce the number of required degrees of freedom per fracture

network, at the same time maintaining appropriately fast convergence rates.

As an extension of this work, an effective and automatic coarsening strategy for both matrix and

fractures are quite important. This is also the topic of our future research.

Acknowledgments

Financial supports of PI/ADNOC are greatly acknowledged.

References

K. Aziz and A. Settari. Petroleum Reservoir Simulation. Blitzprint Ltd., Cagary, Alberta, 2002.

D. Cortinovis and P. Jenny. Iterative galerkin-enriched multiscale finite-volume method. J. Comp.

Phys., 227:248 267, 2014.

Y. Efendiev and T. Y. Hou. Multiscale Finite Element Methods: Theory and Applications. Springer,

2009.

H. Hajibeygi and P. Jenny. Multiscale finite-volume method for parabolic problems arising from

compressible multiphase flow in porous media. J. Comput. Phys., 228:5129 5147, 2009.

H. Hajibeygi and P. Jenny. Adaptive iterative multiscale finite volume method. J. Comput. Phys.,

230(3):628 643, 2011.

H. Hajibeygi and H. A. Tchelepi. Compositional multiscale finite-volume formulation. SPE Journal,

19(2):316 326, 2014.

H. Hajibeygi, D. Karvounis, and P. Jenny. A hierarchical fracture model for the iterative multiscale

finite volume method. J. Comput. Phys., 230(24):8729 8743, 2011.

H. Hajibeygi, S. H. Lee, and I. Lunati. Accurate and efficient simulation of multiphase flow in a

heterogeneous reservoir by using error estimate and control in the multiscale finite-volume

framework. SPE Journal, 17(4):10711083, 2012.

T. Y. Hou and X.-H. Wu. A multiscale finite element method for elliptic problems in composite

materials and porous media. J. Comput. Phys., 134:169 189, 1997.

T. Y. Hou, X.-H. Wu, and Z. Cai. Convergence of a multiscale finite element method for elliptic

problems with rapidly oscillating coefficients. Math. Comp., 68(227):913943, 1999.

P. Jenny and I. Lunati. Modeling complex wells with the multi-scale finite volume method. J. Comput.

Phys., 228:687702, 2009.

P. Jenny, S. H. Lee, and H. A. Tchelepi. Multi-scale finite-volume method for elliptic problems in

subsurface flow simulation. J. Comput. Phys., 187:4767, 2003.

P. Jenny, S. H. Lee, and H. A. Tchelepi. Adaptive fully implicit multi-scale finite-volume method for

multi-phase flow and transport in heterogeneous porous media. J. Comput. Phys., 217:627641,

2006.

V. Kippe, J. E. Aarnes, and K.-A. Lie. A comparison of multiscale methods for elliptic problems in

porous media flow. Comput. Geosci., 12(3):377398, 2008.

S. H. Lee, C. L. Jensen, and M. F. Lough. Efficient finite-difference model for flow in a reservoir with

multiple length-scale fractures. SPE J., 3(5):268 275, 2000.

S. H. Lee, C. Wolfsteiner, and H. A. Tchelepi. Multiscale finite-volume formulation for multiphase

flow in porous media: black oil formulation of compressible, three-phase flow with gravity.

Comput. Geosci., 12(3):351366, 2008.

D. W. Peaceman. Interpretation of well-block pressures in numerical reservoir simulation. SPE 6893,

SPE-AIME 52nd Annual Fall Technical Conference and Exhibition, Denver, USA, Oct 9-12,

1977.

K. Stuben. SAMG Users Manual. Fraunhofer Institute SCAI, 2010.

14

SPE-173200-MS

M. Tene, H. Hajibeygi, Y. Wang, and H. Tchelepi. Compressible algebraic multiscale solver (c-ams).

Proceedings of the 14th European Conference on the Mathematics of Oil Recovery (ECMOR),

Catania, Sicily, Italy, 2014.

Y. Wang, H. Hajibeygi, and H. A. Tchelepi. Algebraic multiscale linear solver for heterogeneous

elliptic problems. Journal of Computational Physics, 259:284 303, 2014.

H. Zhou and H. A. Tchelepi. Two-stage algebraic multiscale linear solver for highly heterogeneous

reservoir models. SPE J., SPE 141473-PA, 17(2):523539, 2012.

H. Zhou, S. Lee, and H. Tchelepi. Multiscale finite-volume formulation for saturation equations. SPE

J., 17(1):198 211, 2011.

- Analysis of Male and Female HandwritingUploaded byInternational Journal of Computer and Communication System Engineering
- Week4 HomeaworkUploaded byAbdallaIsmail
- S Essentials General Mathematics Advanced 11Uploaded bytl102938
- Digital Signal Processing MatlabUploaded byCatalin del Bosque
- 8 TCCS4 Task F Georgescu-Modelling and Simulation of CO2 Leakage Mechanisms Version2[1]Uploaded bysaladinayubi1234
- ECON 262-Mathematical Applications in Economics-Kiran AroojUploaded byHaris Ali
- JEE Advanced Syllabus by EntranceindiaUploaded byEntranceindia
- Matlab TutorialUploaded byrtloquias
- hw4Uploaded byRavi Sharma
- SPE-23575-MS - Slug Test Method in Pressure Sensitive ReservoirsUploaded byLevinska Primavera
- Group TheoryUploaded byMinhaz Pathan
- RDT Reservoir Description ToolUploaded bykumarddn
- Maths_SQP.pdfUploaded byNirvana Adithya Visiobibliophobiatic Guy
- Production Technology Wellflo Project (1)Uploaded byLoh Chun Liang
- Schelkunoff Theorems for Linear ArraysUploaded bylautnertaylor
- PETSOC-86!04!02 Skin Factor and Flow Efficiency in Wells ProUploaded byMiguel Angel Gonzales
- Octave 3 2 4 Documentation PDFUploaded byRobert
- V3I8_IJERTV3IS080539Uploaded bynilesh
- A Simple Technique for Calculating Natural Frequencies of Geometrically Nonlinear Prestressed Cable StructuresUploaded bySebastián Francisco
- BSIT 42Lab[1] RevisedUploaded byAnkita Sharma
- Modeling Methods Study GuideUploaded bysean_terry
- Assignment 1Uploaded byAnonymous k2scQvN2
- 0. PET524 IntroUploaded by56962645
- p250-fredmanUploaded byasdf
- lec34Uploaded byMishal Limbu
- DeterminantsUploaded byAlhilali Ziyad
- 2D StandardizationUploaded byparvaneh.e
- paper1.pdfUploaded byAfnan Ali
- JR - A Generlzd & Effct Algo for Est Trnsit Route ODs From Passngr CountsUploaded byochaerry
- On JAM of Triangular Fuzzy Number MatricesUploaded byijsret

- SPE-1208-0026-JPTUploaded byDavid Montoya
- Well Production Practical Handbook ed. by Henri Cholet.pdfUploaded byDavid Montoya
- Parte 2_Capitulo 19Uploaded byDavid Montoya
- Parte 2_Capitulo 20Uploaded byDavid Montoya
- Petrel 2013 Release NotesUploaded bybigboybigfather
- Parte 2_Capitulo 18Uploaded byDavid Montoya
- Parte 2_Capitulo 17Uploaded byDavid Montoya
- Parte 2_Capitulo 15Uploaded byDavid Montoya
- Parte 1_Capitulo 14Uploaded byDavid Montoya
- Parte 1_Capitulo 10Uploaded byDavid Montoya
- 62326_07Uploaded byDavid Montoya
- SPE-13510.pdfUploaded bywsanchez_sotelo
- 2012-188Uploaded byJhon David Giraldo Rodriguez
- Saudi Aramco JournalUploaded byDavid Montoya
- Petrel 2014 Quick Start GuideUploaded byDavid Montoya
- 43504_10Uploaded byDavid Montoya
- ECLIPSE Advanced Field Management Facilities_02.pdfUploaded byDavid Montoya
- Bataweel DissertationUploaded byDavid Montoya
- FULLTEXT01 (1)Uploaded byDavid Montoya
- energies-09-00244Uploaded byDavid Montoya
- Core-Scale Simulation of Polymer Flow Through Porous MediaUploaded byDavid Montoya
- Future Characteristics of Offshore Support VesselsUploaded byDavid Montoya
- Spe 173696 MsleerUploaded byDavid Montoya
- SPE-74391-MSUploaded byDavid Montoya
- Online Field-Wide Gas-lift Optimisation (Final)Uploaded byDavid Montoya
- Viewgraphs-11V10-DualGLUploaded byDavid Montoya
- 11V9 Members Attendance (1)Uploaded byDavid Montoya
- SPE-108081-MSUploaded byDavid Montoya
- TutorialUploaded byDavid Montoya

- 9A05101 Programming in C and Data StructuresUploaded bysivabharathamurthy
- dice and card games to practice math factsUploaded byapi-296063090
- What is AQL and Why Does It MatterUploaded byhyacintum
- Pump_ZM_B.pdfUploaded byRamon Pacheco
- LTE Guidelines in ICS Designer v2Uploaded bymykeenzo5658
- Site Investigation Code of PracticeUploaded byLiam Culligan
- 9A02304 Basic Electrical & Electronics EngineeringUploaded bysivabharathamurthy
- Computer History1Uploaded bymosesals
- MHDUploaded byMelvin George
- Rotary Cutting Instruments1Uploaded byRishabh Madhu Sharan
- DBMS - Gr8AmbitionZUploaded byAmit Chhakkar
- 10-TimeUploaded bySelect dsouza
- Control postural- dislexieUploaded byEliMihaela
- Tk Test PaperUploaded byrlnandha_2006
- Colligative Properties: Cryoscopy & EbullioscopyUploaded byVimal S. Katiyar
- Architecting Applications to Use Windows Azure AppFabric CachingUploaded byVitor Meriat
- Chapter 3 CCR 4 Millikans Oil Drop ExperimentUploaded byDanka Pantic
- Properties of Stainless SteelsUploaded byngoclinh87
- 100305Uploaded byOnel Gomez
- Lycoming IO 360 M1AUploaded bypeng_dongtao2054
- 3 ITK 330 Multiple ReactionsUploaded byASUPREMEA
- simulationUploaded byvenkatesh
- SlopeUploaded byAyesha
- How Ignition Coils WorkUploaded byxLibelle
- 03_G130&G150&S150_TECH-DAT_CLUploaded byANDRES CISTERNAS
- Visual Basic 6 (VB6)Uploaded bybalajiodd
- DNA IsolationUploaded bygrjhb
- A Step-By-step Guide to SdkUploaded byIRFAN
- ResearchDownload User Guide (en)Uploaded byVladislav Nest
- 1791383751Uploaded bysachinkurhekar