ACCELERATED PERTURBATION BOUNDARY ELEMENT MODEL FOR FLOW PROBLEMS IN HETEROGENEOUS RESERVOIRS
a dissertation submitted to the department of petroleum engineering and the committee on graduate studies of stanford university in partial fulfillment of the requirements for the degree of doctor of philosophy
By Kozo Sato June 1992 c Copyright 1992 by Kozo Sato All Rights Reserved
ii I certify that I have read this thesis and that in my opin- ion it is fully adequate, in scope and in quality, as a dissertation for the degree of Doctor of Philosophy.
Roland N. Horne (Principal Adviser)
I certify that I have read this thesis and that in my opin- ion it is fully adequate, in scope and in quality, as a dissertation for the degree of Doctor of Philosophy.
Khalid Aziz
I certify that I have read this thesis and that in my opin- ion it is fully adequate, in scope and in quality, as a dissertation for the degree of Doctor of Philosophy.
Henry J. Ramey, Jr.
Approved for the University Committee on Graduate Studies:
Dean of Graduate Studies
iii Abstract
The boundary element method (BEM), successfully applied to fluid flow problems in porous media, owes its elegance, the computational efficiency and accuracy, to the existence of the fundamental solution (free-space Green’s function) for the governing equation. If there is no such solution in a closed form, the BEM cannot be applied, and, unfortunately, this is the case for flow problems in heterogeneous media. To overcome this difficulty, the governing equation is decomposed into various order perturbation equations, for which the free-space Green’s function can be found. At each level of perturbation, the solution is computed by the BEM and the sum- mation of various order perturbation solutions gives the complete solution for the original governing equation. The convergence of the perturbation series depends on the distance to the nearest singularity inherent in the equation. A greater magnitude of heterogeneity makes this distance shorter, and, hence, the rate of convergence becomes slower and eventually the series diverges. However, it is possible to elicit a significant amount of information from the perturbation series and to recover an accurate approximation to the exact solution. To this end, Pad´e approximants are employed to accelerate the rate of convergence of a slowly convergent series and to convert a divergent series into a convergent series. Two kinds of perturbation boundary element models are developed: one for steady-state flow problems, associated with the Laplace operator and the other for transient flow problems, associated with the modified Helmholtz operator in Laplace space. These models are verified against analytical solutions for simplified problems and are utilized to solve various application problems.
iv The perturbation BEM shows its utility in streamline tracking and well testing problems in heterogeneous media. The analytical nature of the solution is well pre- served through the free-space Green’s function methodologically and through the singularity programming technically. The durability of the model to rapid spatial variability in rock properties is established by using Pad´e approximants. Through the verification and application problems, it is observed that if the av- erage property value within a drainage area is not much different from the near-well property value, heterogeneity has little effect on pressure responses. The perturbation forms for steady-state and transient flow equations derived in this study should be of value in formulating any semi-analytical scheme.
v Acknowledgement
This dissertation leaves me greatly indebted to Professor Roland N. Horne, my prin- cipal advisor, for his advice, guidance, and encouragement during the course of this research. I wish to express my appreciation to Professor Khalid Aziz, my M.S. re- search supervisor, who encouraged me to come back to Stanford. Sincere thanks are due to Professor Henry J. Ramey, Jr., who served on the reading committee, and Professors Thomas A. Hewett and William E. Brigham, who participated in the examination committee. Appreciation is extended to Professor Joseph B. Keller of the Department of Mathematics who helped me to broaden my understanding of perturbation methods. I appreciate Dr. Michael F. Riley for sharing his insights and knowledge of per- turbation methods. Chick Wattenbarger, Santosh Verma, and Cesar Palagi helped me in preparing this dissertation. Many other colleagues were more helpful than they realized. Being a teaching assistant, I gained a flexible thinking about boundary element methods through their questions. Financial support for this work was provided by Teikoku Oil Company (TOC) and Japan National Oil Corporation. My special gratitude goes to my colleagues of TOC for helping me to get academic leave.
vi This dissertation is dedicated to my parents, Hisashi and Shizu Sato
vii Contents
Abstract iv
Acknowledgement vi
1 Introduction 1
2 Review of Literature 5 2.1 Boundary Element Method ...... 5 2.1.1 Indirect and Direct Methods ...... 6 2.1.2 Applications to Flow in Homogeneous Media ...... 7 2.1.3 Applications to Flow in Heterogeneous Media ...... 8 2.2 Perturbation Method ...... 10
3 Mathematical Preliminaries 12 3.1 Fluid Flow Equation ...... 12 3.1.1 Dimensionless Equations ...... 14 3.1.2 Laplace Space Formulation ...... 15 3.2 Green’s Function Method ...... 16 3.2.1 Laplace Operator ...... 17 3.2.2 Modified Helmholtz Operator ...... 19 3.3 Boundary Element Method ...... 21 3.3.1 Boundary Discretization ...... 22 3.3.2 Boundary Solutions ...... 25 3.3.3 Interior Solutions ...... 27
viii 3.4 Perturbation Method ...... 27 3.4.1 Steady-State Flow Problems ...... 29 3.4.2 Transient Flow Problems ...... 30 3.4.3 Convergence of Perturbation Series ...... 33 3.5 Improvement of Perturbation Series ...... 34 3.5.1 Analysis of Series ...... 34 3.5.2 Pad´e Approximants ...... 36
4 Development of a Steady-State Flow Model 40 4.1 Boundary Integral Equations ...... 40 4.2 Treatment of Wells ...... 43 4.2.1 Well Singularity ...... 43 4.2.2 Pressure Specified Wells ...... 44 4.3 Evaluation of Integrals ...... 45 4.3.1 Boundary Integrals ...... 45 4.3.2 Domain Integrals ...... 46 4.4 Evaluation of Pad´e Approximants ...... 50 4.5 Streamline Tracking ...... 51 4.5.1 Euler and Modified Euler Methods ...... 52 4.5.2 Displacement Performance ...... 54
5 Development of a Transient Flow Model 56 5.1 Boundary Integral Equations ...... 56 5.2 Treatment of Wells ...... 59 5.2.1 Well Singularity ...... 59 5.2.2 Pressure Specified Wells ...... 60 5.3 Evaluation of Integrals ...... 60 5.3.1 Boundary Integrals ...... 61 5.3.2 Domain Integrals ...... 63 5.4 Evaluation of Pad´e Approximants ...... 65 5.5 Transient Pressure Tests in Wells ...... 66 5.5.1 Well with Skin and Wellbore Storage ...... 66
ix 5.5.2 Pressure Derivatives ...... 68
6 Computational Procedures 69 6.1 Steady-State Flow Model ...... 70 6.2 Transient Flow Model ...... 75
7 Results and Discussion 81 7.1 Verification of the Steady-State Flow Model ...... 81 7.1.1 Effectiveness of Pad´e Approximants ...... 82 7.1.2 Reservoir of Harmonically Varying Permeability ...... 95 7.1.3 A Well in an Exponentially Heterogeneous Reservoir ..... 97 7.1.4 Effectiveness of Modified Euler Method ...... 104 7.2 Verification of the Transient Flow Model ...... 111 7.2.1 Optimal Number of Stehfest Sampling Times ...... 112 7.2.2 Reservoir of Harmonically Varying Permeability ...... 114 7.2.3 Reservoir of Varying Porosity ...... 118 7.2.4 Wells in an Exponentially Heterogeneous Reservoir ...... 118 7.2.5 Importance of Well-Singularity Treatment ...... 126 7.2.6 Effects of Skin and Wellbore Storage ...... 126 7.3 Application Problems ...... 131 7.3.1 Full-Field Problem ...... 131 7.3.2 Repeated Five-Spot Pattern ...... 138 7.3.3 Type Curves for Idealized Heterogeneities ...... 159
8 Conclusions and Recommendations 177
Nomenclature 181
Bibliography 185
A Dirac Delta Function 196
B Selected Theorems on Pad´e Approximants 200 B.1 Uniqueness Theorem ...... 200
x B.2 Invariance Theorems ...... 201
C Extra Evaluation of Influence Coefficients 204 C.1 Directional Derivatives ...... 204 C.2 Singular Integrals ...... 206
D Exact Solutions for Verification Problems 208 D.1 Wells in a Heterogeneous Reservoir ...... 208 D.1.1 Transient Pressure Solution ...... 208 D.1.2 Steady-State Pressure Solution ...... 212 D.2 Equidimensional Laplace-Space Equations ...... 212 D.2.1 Reservoir of Varying Permeability ...... 213 D.2.2 Reservoir of Varying Porosity ...... 214
E Multiple-Boundary Problems 216 E.1 Mathematical Considerations ...... 216 E.2 Results and Discussion ...... 218
F Input/Output Samples 230
G FORTRAN Code 244
xi List of Tables
7.1 Coefficients of Taylor series and perturbation series expansions. . . . 89 7.2 Comparison of perturbation BEM results and exact solutions for the flow problem in harmonically heterogeneous media...... 99 7.3 Boundary solutions for the right-angled wedge problem...... 107 7.4 Sensitivity of the perturbation BEM to the number of Stehfest sam- pling times...... 113 7.5 Comparison of pressure responses at (0.4,0.5) computed with and with- out the singularity programming...... 127 7.6 Comparison of pressure responses at (0.6,0.5) computed with and with- out the singularity programming...... 128 7.7 Parameters for the full-field problem...... 134 7.8 Comparison of well responses in a synthetic reservoir with and without variability in permeability...... 135 7.9 Parameters for the repeated five-spot pattern flooding...... 139
xii List of Figures
3.1 Local ξη coordinate system...... 24
4.1 Quadrature points and weights of four-point Gaussian quadrature. . . 48
6.1 Flow chart for the steady-state flow model...... 71 6.2 Flow chart for the transient flow model...... 76
7.1 Schematic configuration of the problems in Sections 7.1.1 and 7.2.1-3. 83 7.2 Comparison of perturbation BEM results and exact solutions for the
steady-state flow problems with k = exp(xD) and k = exp(3xD). . . . 84 7.3 Error profiles of 40- and 80-node discretizations for the steady-state
flow problem with k = exp(3xD)...... 85 7.4 Various order perturbation solutions for the steady-state flow problems
with k = exp(5xD) and k = exp(7xD)...... 87 7.5 Locations of the nearest singularities for the steady-state flow problems
with k = exp(5xD) and k = exp(7xD)...... 88 7.6 Comparison of perturbation BEM results and exact solutions for the
steady-state flow problems with k = exp(5xD) and k = exp(7xD). . . 91 7.7 Error profiles of 40- and 80-node discretizations for the steady-state
flow problem with k = exp(7xD)...... 92 7.8 Various order Pad´e approximant solutions for the steady-state flow
problems with k = exp(7xD), k = exp(11xD), k = exp(15xD), and
k = exp(19xD)...... 96 7.9 Permeability variation given by Eq. 7.18 with a =3,b = 2, and c =1. 98
xiii 7.10 Schematic configuration of the problems in Sections 7.1.3 and 7.2.4. . 100 7.11 Comparison of perturbation BEM results and exact solutions for the
steady-state flow problem with k = exp{3(xD − xwD)}...... 101 7.12 Comparison of perturbation BEM results and exact solutions for the
steady-state flow problem with k = exp{5(xD − xwD)}...... 102 7.13 Comparison of perturbation BEM results and exact solutions for the
steady-state flow problem with k = exp{7(xD − xwD)}...... 103 7.14 Schematic configuration of the right-angled wedge problem...... 106 7.15 Streamlines tracked with (a) regular elements and (b) singular elements for the right-angled wedge problem, the Euler method...... 108 7.16 Streamlines tracked with (a) regular elements and (b) singular elements for the right-angled wedge problem, the modified Euler method. . . . 109 7.17 Effects of Δθ on (a) computational effort and (b) computational accu- racy for the right-angled wedge problem...... 110 7.18 Comparison of BEM results and exact solutions for the one-dimension-
al transient flow problem with nst =8...... 115 7.19 Comparison of perturbation BEM results and exact solutions for the 2 transient flow problem with k =(1+xD) ...... 116 7.20 Comparison of perturbation BEM results and exact solutions for the 2 transient flow problem with k =(1+5xD) ...... 117 7.21 Comparison of perturbation BEM results and exact solutions for the −2 transient flow problem with φ =(1+xD) ...... 119 7.22 Comparison of BEM results and exact solutions for transient pressure responses at wells in a homogeneous reservoir...... 121 7.23 Comparison of perturbation BEM results and exact solutions for tran- sient pressure responses at wells in an exponentially heterogeneous reservoir...... 122 7.24 Comparison of BEM results and exact solutions for pressure derivatives at wells in a homogeneous reservoir...... 124 7.25 Comparison of perturbation BEM results and exact solutions for pres- sure derivatives at wells in an exponentially heterogeneous reservoir. . 125
xiv 7.26 Comparison of BEM results and exact solutions for transient pressure
responses at a well with S = 0 and various CD’s...... 129 7.27 Comparison of BEM results and exact solutions for transient pressure
responses at a well with S = 20 and various CD’s...... 130 7.28 Configuration of the full-field problem...... 133 7.29 Displacement profiles after 0.2 PV of injection with the permeability variation ignored (left) and honored (right)...... 136 7.30 Displacement profiles after 0.5 PV of injection with the permeability variation ignored (left) and honored (right)...... 137 7.31 Pressure responses at injection and production wells in a homogeneous repeated five-spot pattern...... 140 7.32 Streamlines and tracer fronts at breakthrough in a homogeneous re- peated five-spot pattern...... 143 7.33 Comparison of simulated tracer breakthrough curve and exact solution for a homogeneous repeated five-spot pattern...... 144 7.34 Tracer breakthrough curves simulated by ECLIPSE for a homogeneous repeated five-spot pattern...... 145 7.35 Permeability variation given by Eq. 7.47 with a = 22 and b =3.... 146 7.36 Streamlines and tracer fronts at breakthrough in a repeated five-spot pattern with the permeability variation given by Eq 7.47 with a =22 and b =3...... 147 7.37 Tracer breakthrough curve for a repeated five-spot pattern with the permeability variation given by Eq 7.47 with a = 22 and b =3..... 148 7.38 Pressure responses at injection and production wells in a repeated five- spot pattern with the permeability variation given by Eq 7.47 with a = 22 and b =3...... 150
7.39 Permeability variation given by k = exp(3xD)...... 152 7.40 Pressure responses at injection and production wells in a repeated five-
spot pattern with the permeability variation given by k = exp(3xD). . 153 7.41 Streamlines and tracer fronts at breakthrough in a repeated five-spot
pattern with the permeability variation given by k = exp(3xD). . . . 154
xv 7.42 Tracer breakthrough curve for a repeated five-spot pattern with the
permeability variation given by k = exp(3xD)...... 155
7.43 Permeability realization (No.1) for k50 = 10, σln(k) =0.7, and the range of300ft...... 157
7.44 Permeability realization (No.2) for k50 = 10, σln(k) =0.7, and the range of300ft...... 158 7.45 Pressure responses at injection and production wells in a repeated five- spot pattern with the permeability realization No.1...... 160 7.46 Pressure responses at injection and production wells in a repeated five- spot pattern with the permeability realization No.2...... 161 7.47 Streamlines and tracer fronts at breakthrough in a repeated five-spot pattern with the permeability realization No.1...... 162 7.48 Tracer breakthrough curve for a repeated five-spot pattern with the permeability realization No.1...... 163 7.49 Streamlines and tracer fronts at breakthrough in a repeated five-spot pattern with the permeability realization No.2...... 164 7.50 Tracer breakthrough curve for a repeated five-spot pattern with the permeability realization No.2...... 165 7.51 The linear variation in porosity (above) and the permeability variation given by the Kozeny-Carman equation (below)...... 167 7.52 Type curve for a single well in a square flow domain with the lin- ear variation in porosity and the permeability variation given by the Kozeny-Carman equation...... 168 7.53 The concave variation in porosity (above) and the permeability varia- tion given by the Kozeny-Carman equation (below)...... 169 7.54 Type curve for a single well in a square flow domain with the con- cave variation in porosity and the permeability variation given by the Kozeny-Carman equation...... 170 7.55 The convex variation in porosity (above) and the permeability variation given by the Kozeny-Carman equation (below)...... 172
xvi 7.56 Type curve for a single well in a square flow domain with the con- vex variation in porosity and the permeability variation given by the Kozeny-Carman equation...... 173 7.57 Type curve for a single well in a square flow domain with the concave variation in porosity...... 175 7.58 Type curve for a single well in a square flow domain with the permeabil- ity variation given by the Kozeny-Carman equation (for the concave variation in porosity)...... 176
A.1 δ sequence defined by Eq. A.6...... 197
D.1 Image-well location for a single well in the center of a square...... 211
E.1 Multiple-boundary system...... 217 E.2 Streamlines in a repeated five-spot pattern with an impermeable cen- tered circle...... 219 E.3 Streamlines in a repeated five-spot pattern with a constant-pressure centered circle...... 220 E.4 Streamlines in a repeated five-spot pattern with impermeable and constant-pressure circles...... 222 E.5 Streamlines in a repeated five-spot pattern with a N45W/S45E frac- ture, coarse discretization...... 223 E.6 Streamlines in a repeated five-spot pattern with a N/S fracture, coarse discretization...... 224 E.7 Dimensionless inward-flux distribution along the N/S fracture, coarse discretization...... 226 E.8 Dimensionless inward-flux distribution along the N/S fracture, fine dis- cretization...... 227 E.9 Streamlines in a repeated five-spot pattern with a N/S fracture, fine discretization...... 228 E.10 Streamlines in a repeated five-spot pattern with a N/S off-centered fracture...... 229
xvii Chapter 1
Introduction
Numerical techniques for solving partial differential equations describing various phys- ical processes can be categorized into two distinct classes: the domain methods and the boundary methods. Finite difference and finite element methods fall in the first class, and boundary element methods constitute the second. The domain meth- ods have enjoyed much popularity and have been extensively used in the realm of petroleum engineering, as they have in many other disciplines. Although the bound- ary element method is being accepted as a promising rival of domain methods in some areas such as electrostatics and elastostatics, it has not received the attention it deserves in the petroleum industry. While there may be peripheral reasons for this, of interest here are purely method- ological reasons. The principal differences between these two types of numerical schemes can be illustrated by comparing their advantages and disadvantages mani- fested when applied to fluid flow problems. The domain methods require the domain subdivision to evaluate the solution at the nodes or grid points. The solutions between these network nodes are expressed in an interpolated form in terms of the values at the nodes. A system of linear algebraic equations is constructed by relating the solutions at nodal locations to the partial differential equation governing fluid flow. This process is not restricted by the nature of the governing equation, and flow problems with nonlinearity and/or heterogeneous domain properties are well handled by the domain methods. The resulting system
1 CHAPTER 1. INTRODUCTION 2
of equations is large but sparse (usually banded), and this special structure of the solution matrix makes the solution amenable to fast matrix inversion algorithms. When applied to fluid flow problems, the domain methods suffer from numerical dispersion and grid orientation effects (Aziz and Settari, 1979) due to the domain dis- cretizations. They cause serious problems in forecasting displacement performances, and several attempts to alleviate these unfavorable phenomena have been reported. Among these are the nine-point finite difference scheme by Yanosik and McCracken (1979), the two-point upstream weighting by Todd et al. (1972), the harmonic to- tal mobility method by Vinsome and Au (1981), the flux limiter method by Sweby (1984), the flux corrected transport by Zalesak (1979), the total variation diminish- ing mid-point scheme by Rubin and Blunt (1991), the uniform hexagonal gridding by Pruess and Bodvarsson (1983), the orthogonal curvilinear gridding by Hirasaki and O’Dell (1970), and the perpendicular bisectors by Heinemann et al. (1991). However, none of these modifications can overcome the problems entirely. Another shortcoming of the domain methods lies in the well treatment. The flowing bottom-hole pressure of a well is usually related to the pressure calculated for the block containing the well through well models. The most widely accepted well model in finite difference reservoir simulators is the type proposed by Peaceman (1978). In this model, the producing-block pressure is interpreted as a flowing pressure at an equivalent radius from the block center. Under the conditions of isotropic permeabilities, square gridblocks, steady-state flow, and a well in the center of an interior block, the equivalent radius is obtained as r0 =0.208Δx, where Δx is the length of the block edge. Several modifications to this well model have been conducted by Kuniansky and Hillestad (1980), Peaceman (1983), Abou-Kassem and Aziz (1985), Peaceman (1987), Babu and Odeh (1989), and Palagi and Aziz (1992) to relax the conditions made in the original model. These authors have successfully derived well models applicable (at least technically) to more than one centered or off-centered well in a rectangular gridblock with anisotropic permeability. However, the condition of steady-state (or pseudo-steady-state) flow has not been removed. When a high level of accuracy is required in modeling individual well behavior, for instance, in well testing problems, the domain methods would not give satisfactory results due to the CHAPTER 1. INTRODUCTION 3
mathematical weakness in well models. In the boundary element method, only the boundary is discretized into elements. If the fundamental solution (or the so-called free-space Green’s function) is found to the relevant partial differential equation in an infinite domain (or free space), a system of linear algebraic equations can be constructed by imposing that the prescribed conditions at boundary nodes are satisfied by a superposed form in terms of the free- space Green’s function. This process is contingent upon the existence of the free-space Green’s function, and cannot be applied directly to the problems for which such a function does not exist. In addition, since the concept of superposition is required, this formulation process is limited to linear partial differential equations. Although its applicability is somewhat limited, the boundary element method manifests several superiorities over the domain methods. The most notable advantage is the high degree of accuracy that results from its sound mathematical foundations. The flexibility in defining boundary geometries and conditions is another feature to be emphasized. From the computational point of view, a reduction in dimensionality of the problem leads to a populated but much smaller system of algebraic equations than the counterpart in the domain methods. Once these equations are solved, the solution at any interior point can be readily computed. When applied to fluid flow problems, the boundary element method overcomes the above-mentioned shortcomings inherent in the domain methods. As no domain discretization is required, flow is not restricted by the grid system but determined purely by the governing equation. Discretization errors are introduced only on the boundary, and, hence, numerical dispersion and grid orientation effects related to the domain discretization can be removed completely. As for the well treatment, since the analytical nature of the solution is preserved through the free-space Green’s function, there remains no difficulty in evaluating the flowing bottom-hole pressure without any prominent assumptions. Despite these substantial features, the boundary element method has not been appreciated among petroleum engineers. The main reason for this seems to lie in the fact that the two important aspects, heterogeneity and multiphase flow, are excluded from its range of applicability due to the requirements of free-space Green’s functions CHAPTER 1. INTRODUCTION 4
and the restrictions of linearity. This study was initiated to relax the first condition and develop a boundary ele- ment method applicable to fluid flow problems in heterogeneous media. To achieve this end, the governing equation is manipulated and decomposed into a series of perturbation equations, the solutions for which are amenable to boundary element methods. Following the review of literature (Chapter 2), Chapter 3 discusses the pre- liminary concepts utilized in the mathematical manipulations and the model devel- opment. In Chapters 4 and 5, steady-state and transient flow models are developed systematically, and their computational procedures are summarized in Chapter 6. Chapter 7 is devoted to the verification of the models and also to the illustration of their practical use. The contributions made and the findings obtained in the course of this study are summarized in Chapter 8. Related subjects are given in Appendices A through D, and multiple-boundary problems are briefly discussed in Appendix E. Appendices F and G provide Input/Output samples and the program listing. Chapter 2
Review of Literature
The first section discusses the literature on boundary element methods. While bound- ary element methods have been applied to several engineering fields, we limit our discussion to the references considered historically important or related to this study. In particular, emphasis is placed upon those references which simulate flow in hetero- geneous media. The second section discusses the literature on perturbation methods, which play a key role in this study.
2.1 Boundary Element Method
Although the boundary element method (BEM) has drawn attention only recently, its mathematical backbone, the method of integral equations, can be traced to pre- computer times. Kellogg (1929), for example, applied Fredholm integral equations to solve potential problems. The Fredholm integral equations follow from the represen- tation of harmonic potentials by single-layer or double-layer potentials and lead to the so-called indirect BEM. An alternative derivation of the integral equation can be performed through the application of the second form of Green’s theorem, in which a harmonic function is defined as the superposition of the single-layer and the double- layer potential. The so-called direct BEM takes its origin from this alternative. Analytical solutions for integral equations are obtained using the Green’s function for the geometry which satisfies the prescribed boundary conditions of the problem.
5 CHAPTER 2. REVIEW OF LITERATURE 6
The utility of the Green’s function method was discussed by Carslaw and Jaeger (1959) in heat conduction and by Gringarten and Ramey (1973) in petroleum engi- neering literature. However, the Green’s function is limited to very simple geometries, and for “real world” problems a numerical means, the BEM, must be introduced. Historically, the BEM has developed in two parallel ways. One of these is a physical approach and the other is a mathematical approach, which are discussed in the next section.
2.1.1 Indirect and Direct Methods
Basic principles of the present-day BEM were formed in the 1960s as a consequence of the emergence of powerful computers. Jaswon (1963) and Symm (1963) presented a numerical technique to solve boundary-value problems of potential theory formulated by the Fredholm integral equations. This technique can be classified as the indirect BEM, since the unknowns are not the physical variables of the problem but some weighting factors (source densities). After solving for source densities, unspecified boundary parameters are obtained in terms of these densities. Notable predecessors of this method in the same decade were Hess and Smith (1967), who developed powerful programs for the solution of boundary-value problems governed by elliptic partial differential equations, and Harrington et al. (1969), who applied the indirect BEM to solve electrical engineering problems with the Robin type boundary condition. The direct BEM, in which the intermediate step is eliminated by use of the second form of Green’s theorem (or other adequate fundamental integral theorems), was also originated in the work of Jaswon (1963) and Symm (1963), and the results using this method were reported by Jaswon and Ponter (1963). In elastostatics, Rizzo (1967) used the Somigliana identity to formulate the direct BEM, and Cruse and Rizzo (1968) extended the formulation to elastodynamics. The direct BEM seems more appealing to engineers than does the indirect method, since the unknowns bear some physical relation to the problem. Most of the boundary element methods found in the literature on flow mechanics take the direct formulation. CHAPTER 2. REVIEW OF LITERATURE 7
2.1.2 Applications to Flow in Homogeneous Media
Despite the close relationship between the potential theory and the fluid flow me- chanics, it appears fair from the engineering point of view to consider that Liggett and Liu were the first to use the BEM in solving flow problems in porous media. In 1977, Liggett (1977) and Liggett and Liu (1977) employed the BEM to study problems concerning seepage with a free surface, and Liu and Liggett (1977) solved two-dimensional unconfined flow problems in which the response of a water table to artificial recharge was simulated by the BEM. Liu and Liggett (1978) and Liu and Liggett (1979) extended the work done in 1977 and showed that the BEM is a numerical technique suitable to certain problems in flow mechanics. Following this breakthrough, special problems were addressed by several authors. Lennon et al. (1979) showed the formulation of axisymmetric flow problems and applied the BEM to evaluate the free surface in pumping wells. Lafe et al. (1980) discussed three types of singularities: corner flow, interzonal flow between zones of different permeabilities, and well singularities. The applicability of the BEM was extended to three-dimensional flow problems by Lennon et al. (1980), to nonlinear interfacial boundary conditions by Liu et al. (1981), and to complex (leaky, layered, confined, unconfined, and nonisotropic) aquifers by Lafe et al. (1981). The first paper discussing transient flow in porous media was presented by Liggett and Liu (1979). In this study, real-space and Laplace-space formulations were com- pared. Time-stepping schemes in a real-space formulation were later investigated by Taigbenu and Liggett (1986) and Aral and Tang (1988). The text by Liggett and Liu (1983) is a compilation of the above-mentioned papers. In the realm of petroleum engineering, Masukawa and Horne (1988) and Numbere and Tiab (1988) used the BEM for tracking streamlines, and Kikani and Horne (1988) solved well testing problems by the BEM. A version of the BEM was utilized in evaluating the pressure response of a vertical fracture by some authors (Gringarten et al., 1974: Cinco-Ley et al., 1978: Cinco-Ley and Meng, 1988: van Kruysdijk, 1988). The boundary conditions considered in the literature above were all assumed de- terministic. Cheng and Lafe (1991) claimed that hydrological events were better de- scribed as random phenomena and developed the BEM for the solutions of stochastic CHAPTER 2. REVIEW OF LITERATURE 8
groundwater flow problems. It was found that the direct BEM leads to a much length- ier formulation for the integral equations for covariances, and the indirect BEM was employed. Another interesting study to be covered is the formulation in terms of complex variables. Hunt and Isaacs (1981) utilized the Cauchy integral theorem to formulate the integral equation applicable to two-dimensional steady-state flow. The main advantages of this method are its compactness and the fact that the solution yields both the potential and the stream function. Hromadka and Lai (1987) discussed this formulation in detail.
2.1.3 Applications to Flow in Heterogeneous Media
The treatment of heterogeneity in the BEM was addressed by Lafe et al. (1981), who invoked the piecewise homogeneous assumption and used the zoning technique. In this technique, the flow domain is divided into piecewise zones each with a constant property. Interzonal compatibility relations are taken into account in forming a so- lution system. Kikani and Horne (1989) showed the zoning technique can be used to simulate the transient pressure behavior in sectionally homogeneous reservoirs, and Masukawa and Horne (1988) used the same concept on the moving fluid interface during an immiscible displacement process. The piecewise homogeneous assumption, however, does not always provide a good representation of an actual reservoir. In addition, when a large number of zones are required, the computational efficiency of the BEM is limited. The approaches presented by Clements (1980) and Cheng (1984) required no such assumption. When the formation property variation is represented by a certain function, the Green’s function can be derived such that the governing equation with a given variation is satisfied. As a popular permeability variation in the literature, the following function
was considered: k x α = 1+β (2.1) k0 d where k0 is the reference permeability at x =0,d the reference distance in the direction of the permeability variation, and α and β the parameters of curve fitting. CHAPTER 2. REVIEW OF LITERATURE 9
Clements (1980) gave the Green’s function in a double series truncated at a certain number of terms for α =2n (n =0, ±1, ±2, ···). Cheng (1984) considered the variation with α = 2 and two more harmonically varying permeabilities suitable to the transformation suggested by Georghitza (1969):