Modeling of Matrix-Fracture Interaction For Conventional Fractured Gas

UNIVERSITY OF CALGARY
Modeling of Matrix-Fracture Interaction for Conventional Fractured Gas

Reservoirs

by

Ehsan Ranjbar

A THESIS
SUBMITTED TO THE FACULTY OF GRADUATE STUDIES
IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF CHEMICAL AND PETROLEUM ENGINEERING
CALGARY, ALBERTA
JANUARY, 2014

Ehsan Ranjbar 2014
ii
Abstract
Modeling of matrix-fracture transfer function is important in the simulation of fluid flow in
fractured porous media using a dual-porosity concept. This transfer function is directly
related to the shape factor. One of the main focuses of this study is to find the shape factor
for the single-phase flow of compressible fluids in fractured media using the solution of
nonlinear gas diffusivity equation. The developed shape factor can be used as an input for
modeling flow of compressible fluids in dual-porosity systems. For a compressible fluid,
the consequence of a pressure boundary condition on the shape factor has not been
investigated in the previous studies. Another major purpose of this study is, therefore, to
investigate the effect of the fracture pressure on the shape factor for single-phase flow of a
compressible fluid.
Most of the developed models for fractured reservoirs assume ideal matrix block size
distribution. This assumption may not be valid in reality for naturally fractured reservoirs
and possibly lead to errors in prediction of production from the naturally fractured
reservoirs especially during early time production from the matrix blocks. The effect of
different matrix block size distributions on the single-phase matrix-fracture fluid transfer is
studied using a semi-analytical approach. The proposed model is able to simulate fluid
exchange between matrix and fracture for continuous or discrete block size distributions. In
the last part of this study we present semi-analytical solutions for release of a single-phase
liquid or gas from cylindrical (two dimensional flow) and spherical (three dimensional
flow) matrix blocks with various block size distributions and different fracture boundary
conditions. This solution can be simplified to model flow of slightly compressible fluids
like water or oil in dual-porosity media.
The approximate semi-analytical model for the matrix fracture transfer function presented
in this study for different cases is verified using single-porosity, fine-grid, numerical
simulations. This model can recover the shape factor of slightly compressible fluids
reported in the literature. In the proposed semi-analytical model for all cases the pressure
variability of viscosity and isothermal compressibility is considered by solving the
nonlinear partial differential equation of compressible fluid flow in the fractured media.

iii
Acknowledgements
I would like to express my deepest gratitude to all those who provided me the opportunity
to complete this thesis. I have worked with a great number of people. It is my pleasure to
thank them for their supports and help during my PhD program at the University of
Calgary.
First of all I would like to express deepest appreciation to my supervisors Dr. John Chen
and Dr. Hassan Hassanzadeh for their unconditional support during the course of this
study. This work would not have been possible without their valuable comments,
encouragements, guidance and aids. Thanks Dr. Hassanzadeh for providing me this
opportunity to work under his supervision and enthuse and enrich my growth as a student
and researcher. I really appreciate Dr. Hassanzadeh for hours of inspiring discussions,
productive suggestions and patience when the results were not approaching. Thank you Dr.
Chen for your outstanding support, and trusting my abilities to discover different thoughts.
I would like to be grateful Dr. Mingzhe Dong and Dr. Christopher Clarkson for serving on
my supervisory committee and their support during my research. I would also like to thank
Dr. Fanhua Zeng and Dr. Xin Wang as members of my examining committee. I wish to
thank Dr. Mehran Pooladi-Darvish for teaching me the advanced reservoir engineering
course which was really helpful during the period of this study.
Financial support of NSERC/AERI (AIEES)/Foundation CMG and iCORE (AITF) Chairs
Funds and the Department of Chemical and Petroleum Engineering at the University of
Calgary is acknowledged.
Collective and individual acknowledgments are also owed to my past and present
colleagues at the University of Calgary and Reservoir Simulation Group. I would like to
thank one of my best friends Mr. Hamid Emami Meybodi for his helps during my PhD at
the University of Calgary.
Lastly, and most significantly, I wish to thank my parents, Shamsi and Ali who educated,
rose, and advised me. They are my exceptional source of encouragement and motivation. I
would like to thank my sister, Arezoo, and my brother, Omid, for their help and support
which are always a part of my life.

iv
Dedication

I dedi c at e t hi s w or k t o my Mot her f or
her unc ondi t i onal l ove and suppor t
v
Table of Contents
Abstract ............................................................................................................................... ii
Acknowledgements ............................................................................................................ iii
Dedication........................................................................................................................... iv
Table of Contents ................................................................................................................ v
List of Tables .................................................................................................................... viii
List of Figures and Illustrations .......................................................................................... ix
CHAPTER ONE: INTRODUCTION ................................................................................. 1
1.1 Background ............................................................................................................... 1
1.2 Dual-porosity system ................................................................................................. 1
1.3 Shape factor concept ................................................................................................. 3
1.4 Motivations and objectives ........................................................................................ 4
1.5 Components and outline of this study ....................................................................... 5
References ....................................................................................................................... 8
CHAPTER TWO: EFFECT OF FRACTURE PRESSURE DEPLETION REGIMES
ON THE DUAL-POROSITY SHAPE FACTOR FOR FLOW OF
COMPRESSIBLE FLUIDS IN FRACTURED POROUS MEDIA ........................ 10
Abstract ......................................................................................................................... 10
2.1 Introduction ............................................................................................................. 12
2.2 Methodology ........................................................................................................... 16
2.2.1 Constant fracture pressure ............................................................................... 19
2.2.2 Linearly declining fracture pressure ............................................................... 19
2.2.3 Exponentially declining fracture pressure ....................................................... 20
2.3 The approximate analytical solutions ...................................................................... 21
2.3.1 Constant fracture pressure ............................................................................... 22
2.4 Model verification ................................................................................................... 29
2.5 Results ..................................................................................................................... 33
2.5.1 Comparison of model with Warren and Root model ...................................... 33
2.6 Conclusions ............................................................................................................. 38
Nomenclature ................................................................................................................ 40
References ..................................................................................................................... 42
Appendix 2.A: Solution of the gas diffusivity equation for different fracture
depletion conditions .............................................................................................. 46
2.A1: Linearly declining fracture pressure .............................................................. 46
2.A2: Exponentially declining fracture pressure ...................................................... 49
CHAPTER THREE: ONE DIMENSIONAL MATRIX-FRACTURE TRANSFER IN
DUAL-POROSITY SYSTEMS WITH VARIABLE BLOCK SIZE
DISTRIBUTION ...................................................................................................... 54
Abstract ......................................................................................................................... 54
vi
3.1 Introduction ............................................................................................................. 55
3.2 Matrix block size distributions ................................................................................ 58
3.2.1 Uniform or rectangular distribution ................................................................ 59
3.2.2 Exponential distribution .................................................................................. 59
3.2.3 Normal or Gaussian distribution ..................................................................... 60
3.2.4 Linear distribution ........................................................................................... 60
3.2.5 Log-normal distribution .................................................................................. 60
3.3 Equivalent lengths for different matrix block size distributions ............................. 62
3.3.1 Equivalent length concept ............................................................................... 62
3.3.2 Discrete matrix block size distribution ........................................................... 62
3.3.3 Continuous matrix block size distribution ...................................................... 63
3.4 Mathematical model for flow of compressible fluid in fractured media with
variable block size distribution .............................................................................. 64
3.5 Slightly compressible fluids .................................................................................... 68
3.6 Validation ................................................................................................................ 69
3.7 Results ..................................................................................................................... 73
3.7.1 Rectangular, discrete, normal and log-normal distributions ........................... 74
3.7.2 Linear distribution ........................................................................................... 75
3.7.3 Exponential distribution .................................................................................. 77
3.8 Discussion ............................................................................................................... 78
3.9 Conclusions ............................................................................................................. 82
Nomenclature ................................................................................................................ 83
References ..................................................................................................................... 85
Appendix 3.A: Solution of diffusion equation for variable block size distribution ...... 88
Appendix 3.B: Derivation of dimensionless rate and dimensionless cumulative
production for different matrix block size distribution ......................................... 90
CHAPTER FOUR: SEMI-ANALYTICAL SOLUTIONS FOR RELEASE OF FLUIDS
FROM ROCK MATRIX BLOCKS WITH DIFFERENT SHAPES, SIZES AND
DEPLETION REGIMES.......................................................................................... 94
Abstract ......................................................................................................................... 94
4.1 Introduction and previous studies ............................................................................ 95
4.2 Approximate analytical solution ............................................................................. 98
4.2.1 Constant fracture pressure ............................................................................. 103
4.2.2 Variable fracture pressure ............................................................................. 105
4.2.3 Variable block size distribution (multiple blocks) ........................................ 107
4.3 Model verification ................................................................................................. 112
4.4 Results ................................................................................................................... 116
4.4.1 Effect of fracture pressure depletion regime ................................................. 116
4.4.2 Block size distribution effect ........................................................................ 120
4.4.3 Comparison of different block geometries .................................................... 122
4.5 Conclusions ........................................................................................................... 125
Nomenclature .............................................................................................................. 126
References ................................................................................................................... 128
Appendix 4.A: Analytical solution for cylindrical blocks .......................................... 133
4.A1: Constant fracture pressure ............................................................................ 133
4.A2: Variable fracture pressure ............................................................................ 136
vii
4.A3: Variable block size distributions .................................................................. 141
Appendix 4.B: Analytical solution for spherical blocks ............................................. 144
4.B1: Constant fracture pressure ............................................................................ 144
4.B2: Variable fracture pressure ............................................................................ 147
4.B3: Variable Block size distributions ................................................................. 149
CHAPTER FIVE: CONCLUSIONS AND RECOMMENDATIONS............................ 152
5.1 Conclusions ........................................................................................................... 152
5.1.1 Fracture pressure boundary condition effect for a slab-shaped block .......... 152
5.1.3 Block size distribution effect for slab-shaped blocks ................................... 153
5.1.4 Block geometry effect ................................................................................... 153
5.2 Recommendations ................................................................................................. 154
APPENDIX A: COPYRIGHT PERMISSIONS ............................................................. 155

viii
List of Tables
Table 2.1: Stabilized values of the shape factor and time at which the pressure
disturbance reaches the inner boundary for different depletion regimes in the
fracture. ........................................................................................................................ 38
Table 3.1: Observed frequency of matrix block size in the soil column (Gwo et al.,
1998) ............................................................................................................................ 63
Table 3.2: Block size distribution for the discrete distributions. ......................................... 72
Table 3.3: Values of dimensionless equivalent length for different matrix block size
distributions. ................................................................................................................ 81
Table 4.1: Different probability distribution function and their equivalent radius. .......... 111
Table 4.2: Data used for semi-analytical and numerical models. ...................................... 115
Table 4.3: Stabilized values of the shape factor based on this study and literature
models. ....................................................................................................................... 116
Table 4.4: Values of dimensionless equivalent radius for different matrix block size
distributions. .............................................................................................................. 120

ix
List of Figures and Illustrations
Figure 1.1: An idealized dual-porosity system (Lemonnier and Bourbiaux, 2010). ............. 2
Figure 1.2: Road map of this study. ...................................................................................... 7
Figure 2.1: Schematic of the matrix-fracture model. ......................................................... 17
Figure 2.2: Comparison of the matrix-fracture cumulative fluid production obtained
from the approximate analytical solution and the numerical model of Eclipse for
constant fracture pressure. ........................................................................................... 30
Figure 2.3: Comparison of the developed shape factor model with literature models for
slightly compressible fluid in the case of linearly declining fracture pressure. .......... 30
slightly compressible fluid in the case of exponentially declining fracture pressure
for small values of exponent (k=0.0001). ................................................................... 31
(k=0.632). .................................................................................................................... 32
(k=1). ........................................................................................................................... 32
Figure 2.7: Comparison of the developed model with numerical and Warren and Root
model. .......................................................................................................................... 33
Figure 2.8: Comparison of the shape factor for linearly declining and constant fracture
pressure. ....................................................................................................................... 35
Figure 2.9: Shape factor comparison for different exponents for exponentially declining
fracture pressure. ......................................................................................................... 36
Figure 2.10: Comparison of the dimensionless shape factor for different pressure
depletion regime in the fracture. .................................................................................. 37
Figure 3.1: Different probability density functions. ............................................................ 61
Figure 3.2: Illustration of a matrix-fracture system and its boundary conditions (left)
and representation of fractured reservoirs in the case of non-ideal matrix block
size distribution (right). ............................................................................................... 65
Figure 3.3: Comparison of the presented model with literature models (Chang 1995,
Hassanzadeh and Pooladi-Darvish 2006) for slightly compressible fluid and ideal
block size distributions. ............................................................................................... 71
x
Figure 3.4: Comparison of the presented semi-analytical model with the numerical
results for the first and second case. Continuous lines are predictions by analytical
model. Dots and dashes are from numerical simulations. ........................................... 73
Figure 3.5: Dimensionless rate versus dimensionless time for ideal, rectangular,
discrete, normal and log-normal distributions when F
h
=0.1. ...................................... 75
Figure 3.6: Comparison of dimensionless rate for linearly increasing (m=100/81,
b=35/81), linearly decreasing (m=-100/81, b=145/81) distribution with F
h
=0.1
and ideal matrix block size distributions. .................................................................... 76
Figure 3.7: Comparison of dimensionless rate for exponential (exponent values of -20,
-5, 5 and 20) with F
h
=0.1 and ideal block size distributions. ..................................... 77
Figure 3.8: Dimensionless cumulative production versus dimensionless time for
different matrix block size distributions with F
h
=0.1. Blocks showing the
dimensionless equivalent size of the matrix block L
c
/L
cmax
. ........................................ 79
Figure 3.9: Dimensionless shape factor for different matrix block size distributions.
Blocks showing the dimensionless equivalent size of the matrix block L
c
/L
cmax
. ....... 81
Figure 4.1: Schematic representation of the problem for a cylindrical block. .................. 102
Figure 4.2: Comparison of the presented model with the numerical simulation for 2D
flow (cylindrical block approximation). .................................................................... 114
Figure 4.3: Comparison of the presented model with the numerical simulation for 3D
flow (spherical block approximation). ...................................................................... 114
Figure 4.4: Dimensionless rate versus dimensionless time for different fracture
depletion regimes for a cylindrical block. ................................................................. 118
Figure 4.5: Dimensionless cumulative release versus dimensionless time for different
fracture depletion regimes for a cylindrical block. .................................................... 119
fracture depletion regimes for a spherical block. ...................................................... 119
block size distribution and cylindrical blocks. .......................................................... 121
block size distribution and spherical blocks .............................................................. 121
block geometries ........................................................................................................ 123
Figure 4.10: Normalized cumulative fluid release versus square root of scaled time ....... 124

Chapter One: Introduction
1.1 Background
Large portion of the production of natural gas occurs from fractured formations including
naturally fractured reservoirs (NFR), coal bed methane (CBM), shale and tight fractured
gas reservoirs. A naturally fractured reservoir (NFR) is defined as a reservoir that contains
fractures created by natural processes, such as diastrophism or volume shrinkage. These
natural fractures can have a positive or negative effect on fluid flow (Aguilera, 1995;
Ordonez et al., 2001). NFRs are usually thought to consist of an interconnected fracture
network, which provides the main flow paths (fractures have high permeability and low
storage volume), and the reservoir rock or matrix, which acts as the main source of the
fluid storage (matrix blocks have low permeability and high storage volume) (Beckner,
1990). On the other hand in these reservoirs, the major storage for the reservoir fluids is in
the matrix whereas flow primarily occurs in the highly conductive fractures (Chen, 1989).
In such a system, in addition to the intrinsic properties of the matrix and fracture, the
interaction between the matrix and fractures should be modeled accurately.
In general, there are computational challenges in upscaling of fluid flow in fractured
formations. Upscaling of transport and flow parameters for porous media has been
investigated from decades and a range of upscaling techniques have been introduced (Deng
et al., 2010). Flow and transport in fractured porous media are often described using a
dual-porosity model.
1.2 Dual-porosity system
The dual-porosity approach has been used in numerical simulation of groundwater, oil, and
gas flow in fractured porous media. This model assumes that the porous medium includes
two different regions, one related with the macropore or fracture network with high
permeability and the other with a less permeable and more porous system of rock matrix
blocks. Dual-porosity models assume that fluid flow can be described by two equations for
matrix and fractures, which are coupled using a term describing the exchange of fluid
between the two pore regions (Gerke and van Genuchten, 1993). Figure 1.1 shows a
schematic representation of an ideal dual-porosity system.

Chapter 1. Introduction
2

Figure 1.1: An idealized dual-porosity system (Lemonnier and Bourbiaux, 2010).

In the literature, two approaches have been widely used to model dual-porosity systems
mathematically including non-coupled (Chang, 1995; Hassanzadeh and Pooladi-Darvish,
2006, Ranjbar and Hassanzadeh, 2011), and coupled (Hassanzadeh et al., 2009) methods.
The equations that are expressed for coupled dual-porosity models are as follows:
q p
k
t
p
c
f
f f
f f
+ V =
c
c
2
| ,
(1.1)
m
m m
m m
p
k
t
p
c
2
V =
c
c
| ,
(1.2)
m m
m
p
V
A k
q V = ) (
.
(1.3)
Equations (1.1) and (1.2) are used to explain the pressure diffusion of single-phase slightly
compressible fluid in fractured and matrix medium respectively. In Equation (1.1), q is the
source term that stands for the net addition of fluid to the fracture system from the matrix
blocks, per unit of total volume. Equation (1.3) is used to couple Equations (1.1) and (1.2)
and shows the matrix-fracture interflow or matrix-fracture transfer function. In these
equations subscripts f and m stand for fracture and matrix, respectively. The p
f
and p
m

represent the fluid pressure in the fracture and matrix respectively, is the fluid viscosity,
| is the porosity of the medium and k is the permeability. Laplacian operator
2
V represents
3
the divergence of gradient and can be used for different coordinate system and different
matrix block geometries (Cartesian, Cylindrical or Spherical).
A coupled dual-porosity model can be formulated by assuming that at each point, there is a
matrix block with specified shape. Inside each block the fluid pressure p
m
is time and space
dependent (Equation (1.2)). Matrix-fracture interflow term q does not appear explicitly in
Equation (1.2). This interflow is assumed to be distributed through the fracture media as a
source/sink term; the interflow enters the matrix blocks only at their boundaries
(Zimmerman et al., 1992).
Non-coupled approach is an alternative method which is used to determine the matrix-
fracture transfer rate for fractured reservoirs. In this approach the pressure diffusion is
solved in the matrix block and fracture is used as a boundary for the matrix system. The
solution of matrix diffusion equation is used to determine the matrix-fracture transfer
function. Equation (1.4) represents the pressure diffusion in the matrix block
m
m m
m m
p
k
t
p
c
2
V =
c
c
| .
(1.4)
The fracture pressure is assumed as a boundary condition for this partial differential
equation (PDE) to determine the matrix-fracture transfer function. As we will discuss later
this fracture pressure can be a constant or vary with time based on a pre-specified time
function.
It should be mentioned that Equation (1.4) is used for a slightly compressible fluid and is a
linear partial differential equation and this equation can be solved by common methods like
Laplace transform or separation of variables.
1.3 Shape factor concept
Currently, the dual-porosity approach is one of the computationally efficient and widely
used methods to model fluid flow in fractured reservoirs. In this approach, the matrix and
fracture are separated into two different media, each with its own properties. A transfer
function has been used to represent the matrix-fracture interaction and govern the mass
transfer between the matrix blocks and the fractures (Barrenblat et al., 1960). The rate of
mass transferred from the matrix to the fracture is directly proportional to the shape factor.
For modeling of naturally fractured reservoirs, an accurate value of the shape factor is
4
required to account for both the transient and pseudo-steady state behaviour of the matrix-
fracture interaction and also the geometry of the matrix-fracture system.
In the dual-porosity models and pseudo-steady state conditions the blocks are treated as a
lumped system. In this approach the matrix-fracture transfer function is expressed as
follows:
) (
f m
m
p p
k
q = o
.
(1.5)
In this equation o is called shape factor. This parameter is a function of fracture spacing
and geometry of the blocks and has the dimension of reciprocal of area.
1.4 Motivations and objectives
Study of gas flow in fractured porous media is important in a variety of engineering fields.
In hydrology there exist a large number of mathematical models (numerical, analytical or
semi-analytical) to simulate the flow of compressible fluids in underground environments
and structured soils (You et al., 2011). Flow of compressible and slightly compressible
fluids (water or oil) in fractured reservoirs has been studied extensively with applications
in prediction of production rates and well testing. Therefore, the flow of compressible
fluids like gases and air in fractured porous media is important in hydrological,
environmental and petroleum engineering.
Flow and transport in fractured porous media are often described using a dual-porosity
model. It is emphasized that it is not practical to model a large scale fractured reservoir
based on a fine grid approach due to the requirement of large computational time. The
presented semi-analytical models in this study which are based on a non-coupled approach
can be incorporated into numerical models for accurate modeling of the amount of
transferred fluids between matrix and fractures using available dual-porosity formulation.
In other words, this study is important to reduce the computational time for large scale
simulations of gas flow in fractured porous media and can be nested in a numerical model
to resolve subgridblock scale flows.
The developed model in this study can be used to simulate single-phase compressible fluid
flow in the fractured porous media. The presented models can handle the variable fracture
pressure and variable block size distribution for different block geometries. It also can be
5
used for slightly compressible fluid without requiring numerical inversion or infinite series
calculation, in an efficient manner.
1.5 Components and outline of this study
This is a paper-based thesis with each chapter published in a peer reviewed journal.
Chapters are written so that they can be followed independently where each chapter
consists of separate abstract, introduction and previous studies, materials and
methodologies, verification, results, conclusions, nomenclatures, references and
appendices.
In Chapter two we investigate the effect of pressure decline in the fracture on the matrix-
fracture transfer shape factor for a compressible fluid. We study the influence of the
fracture pressure (as a boundary condition for the matrix block) on the shape factor for
flow of a compressible fluid in a dual-porosity model, which has not been investigated in
the former works. To determine the value of the shape factor, diffusivity equation for gas
flow, which is a nonlinear PDE, is solved using the combination of the heat integral
method, the method of moments and Duhamels theorem and this solution is used to
evaluate the shape factor.
Chapter three represents a new semi-analytical approach to consider the effect of variable
matrix block size distribution on flow of a compressible fluid in dual-porosity media. In
addition, due to the analytical nature of the proposed model, it can be used to model flow
of slightly compressible fluids in fractured media. The proposed model is a new semi-
analytical approach that can handle the variable matrix block size distribution for the flow
of both compressible and slightly compressible fluids in the fractured media. In addition to
the continuous block size distribution, the proposed model is also capable to model matrix-
fracture transfer when there is discrete block size distribution using structural information
of a porous medium.
Finally in Chapter four a new semi-analytical model for different matrix block geometries
(cylindrical and spherical) for flow of compressible and slightly compressible fluids in
fractured porous media is developed. This approximation is used to derive the matrix-
fracture fluid transfer for 2D flow or slab-shaped blocks surrounded by two sets of
fractures and 3D flow or slab-shaped blocks surrounded by three sets of fractures. The
6
model can handle various block size distributions and different pressure regimes in the
fracture in the case of different geometries. At the end in Chapter five, conclusions of the
thesis are presented leading to the recommendations for future works. Figure 1.2 shows the
road map of the study.

7

Figure 1.2: Road map of this study.

Slab-Shaped Matrix Block
Cylindrical & Spherical Matrix Block
Matrix-Fracture Transfer Function for
Fractured Gas Reservoirs
Fine Grid Numerical Simulation and Comparison with
Literature Model for Validation (Chapters two to four)
Variable Fracture Pressure
(Chapter two)
Variable Block Size
Distribution (Chapter three)
Constant Fracture Pressure
(Chapter four)
Variable Fracture Pressure
(Chapter four)
Variable Block Size
Distribution (Chapter four)
8
References
Aguilera, R. (1995). Naturally fractured reservoirs. Tulsa, Oklahoma: PennWell Press.
Barenblatt, G.E. Zheltov, I.P. Kochina, I.N. (1960). Basic concepts in the theory of seepage
of homogeneous liquids in fissured rocks (strata). J. Appl. Math. Mech., 20, 852-864.
Beckner, B.L. (1990). Improved modeling of imbibition matrix/fracture fluid transfer in
double porosity simulators. PhD dissertation, Stanford University.
Chang, M.M. (1995). Analytical solution to single and two-phase flow problems of
naturally fractured reservoirs: theoretical shape factor and transfer functions. PhD
dissertation, University of Tulsa.
Chen, Z.X. (1989). Transient flow of slightly compressible fluids through double-porosity,
double-permeability systems-A state-of-the-art review. Transp. Porous Med., 4,147-84.
Deng, H. Dai, Z. Wolfsberg, A. Lu, Z. Ye, M. Reimus, P. (2010). Upscaling of reactive
mass transport in fractured rocks with multimodal reactive mineral facies. Water Resour.
Res., 46.
Gerke, H.H. van Genuchten, M.Th. (1993). A dual-porosity model for simulating the
preferential movement of water and solutes in structured porous media, Water Resour.
Res., 29(2), 305-319
Hassanzadeh, H. Pooladi-Darvish, M. (2006). Effects of fracture boundary conditions on
matrix-fracture transfer shape factor. Transp. Porous Med., 64, 51-71.
Hassanzadeh, H. Pooladi-Darvish, M. Atabay, S. (2009). Shape factor in the drawdown
solution for well testing of dual-porosity systems. Adv. Water Res., 32, 1652-1663.
Lemonnier, P. Bourbiaux, B. (2010). Simulation of Naturally Fractured Reservoirs. State
of the Art. Part 2: Matrix-Fracture Transfers and Typical Features of Numerical Studies,
Oil & Gas Science and Technology Rev. IFP, 65 (2), 263-286.
Ordonez, A. Penuela, G. Idrobo, E.A. Medina C.E. (2001). Recent advances in naturally
fractured reservoir modeling. CT&F Ciencia, Tecnologia y Futuro., 2, 51-64.
Ranjbar, E. Hassanzadeh, H. (2011). Matrix-fracture transfer shape factor for modeling
flow of a compressible fluid in dual-porosity media. Adv. Water Res., 34(5), 62739.
You, K. Zhan, H. Li, J. (2011). Analysis of models for induced gas flow in the unsaturated
zone, Water Resour. Res., 47, W04515, doi: 10.1029/2010WR009985.
9
Zimmerman, R.W. Chen, G. Bodvarsson, S. (1992). A dual-porosity reservoir model with
an improved coupling term. Seventeenth Workshop on Geothermal Reservoir
Engineering, Stanford.

Chapter Two: Effect of Fracture Pressure Depletion Regimes on the Dual-Porosity
Shape Factor for Flow of Compressible Fluids in Fractured Porous Media
1

Abstract
A precise value of the matrix-fracture transfer shape factor is essential for modeling fluid
flow in fractured porous media by a dual-porosity approach. The slightly compressible
fluid shape factor has been widely investigated in the literature. In a recent study, we have
developed a transfer function for flow of a compressible fluid using a constant fracture
pressure boundary condition (Ranjbar and Hassanzadeh, 2011
2
). However, for a
compressible fluid, the consequence of a pressure depletion boundary condition on the
shape factor has not been investigated in the previous studies. The main purpose of this
chapter is, therefore, to investigate the effect of the fracture pressure
3
depletion regime
4
on
the shape factor for single-phase flow of a compressible fluid. In the current study, a model
for evaluation of the shape factor is derived using solutions of a nonlinear diffusivity
equation subject to different pressure depletion regimes. A combination of the heat integral
method, the method of moments and Duhamels theorem is used to solve this nonlinear
equation. The developed solution is validated by fine-grid numerical simulations. The
presented model can recover the shape factor of slightly compressible fluids reported in the

1
This chapter is an exact copy of: Ranjbar, E. Hassanzadeh, H. Chen, Z. (2011). Effect of Fracture Pressure
Depletion Regimes on the Dual-Porosity Shape Factor for Flow of Compressible Fluids in Fractured Porous
Media, Advances in Water Resources, Vol. 34 (12), Page: 1681-1693.

2
The focus of our previous study (Ranjbar, E. Hassanzadeh, H. (2011), Matrix-fracture transfer shape factor
for modeling flow of a compressible fluid in dual-porosity media, Advances in Water Resources, 34(5), page
627-639) was to find the shape factor for the single-phase flow of compressible fluids (gases) in fractured
porous media for the case of constant fracture pressure. In this study (Ranjbar and Hassanzadeh, 2011), a
theoretical analysis of the constant fracture pressure shape factor for the flow of a compressible fluid in
fractured porous media was presented. The presented semi-analytical solution for constant fracture pressure
was validated with fine-grid numerical simulations. In this chapter we further develop our previous study to
consider the effect of pressure variation in the fracture on the matrix-fracture shape factor.

3
It is worth noting that the fracture pressure in this thesis is different than the hydraulic fracture pressure and
it implies the fluid pressure inside the fracture or the boundary condition imposed on the matrix block.

4
In this thesis the fracture depletion regime implies the pressure variations in the fracture which acts as a
boundary condition for the matrix block.

Chapter 2. Effect of fracture pressure depletion regimes on the dual-porosity
11
literature. This study demonstrates that in the case of a single-phase flow of compressible
fluid, the shape factor is a function of the imposed boundary condition in the fracture and
its variability with time. It is shown that such dependence can be described by an
exponentially declining fracture pressure with different decline exponents. These findings
improve our understanding of fluid flow in fractured porous media.

12
2.1 Introduction
In general, there are computational challenges in upscaling of fluid flow in fractured
formations. Upscaling of transport and flow parameters for porous media has been
investigated from decades and a range of upscaling techniques have been introduced (Deng
et al., 2010). Large portion of the produced natural gas occurs from fractured formations
including naturally fractured reservoirs (NFR), coal bed methane (CBM) and tight
fractured gas reservoirs. In these reservoirs, the major storage for the reservoir fluids is in
the matrix whereas flow primarily occurs in the highly conductive fractures (Chen, 1989).
Warren and Root (1963) established the dual-porosity model for modeling of a slightly
compressible fluid flow in the naturally fractured reservoirs. In the dual-porosity approach,
a fractured reservoir is divided into two media with completely different properties:
fracture and matrix. The fracture network supplies the main flow paths and the reservoir
rock or matrix acts as the major source of the fluid storage (Beckner, 1990). On the other
hand most of the fluid storage is in the matrix and fluid flows through the fractures as the
main channel. Therefore, an improved dual-porosity model should be able to accurately
account for the fracture and matrix interaction.
A dual-porosity model, which is an effective and broadly used approach for modeling and
upscaling of fluid flow in the fractured porous media, assumes that two distinct types of
porosity coexist in a representative rock volume (Chen, 1989; Cihan and Tyner, 2010; Di
Donato and Blunt, 2004; Liu and Chen, 1990; Zimmerman et al., 1996). In general,
fracture has a low storage capacity and high transmissivity and the adjacent rock matrix
has a high storage capacity and a relatively low transmissivity (Kazemi and Gilman, 1993).
Defining the transfer shape factor that accounts for the interaction among the matrix and
fracture is a great challenge in dual-porosity upscaling. In dual-porosity models the matrix-
fracture interaction is modeled through a shape factor. An equivalent fracture permeability,
matrix-permeability, matrix-fracture transfer coefficient (shape factor) and saturation
functions (for multiphase flow) are essential parameters for the dual-porosity approach.
Studies have been conducted in the past to determine the transfer shape factor for slightly
compressible fluids in the fractured reservoirs (Bourbiaux et al., 1999; Coats, 1989;
13
Kazemi et al., 1976; Quintard and Whitaker, 1996; Quintard and Whitaker, 1996; Quintard
and Whitaker, 1998; Thomas et al., 1983; Ueda et al., 1989).
A precise value of the shape factor is essential to consider transient and pseudo-steady state
performance of the matrix-fracture interaction and also geometry of the matrix-fracture
system. It should be noted that the functionality of the fracture pressure as a boundary to
the matrix blocks may also have a significant effect on the stabilized value of the shape
factor for a slightly compressible fluid (Chang, 1995; Hassanzadeh and Pooladi-Darvish,
2006).
In traditional dual-porosity formulation the flow between the matrix and the fracture is
considered by a transfer function, which acts as a source term in the governing equation for
fluid flow in the fractures. Darcys law is used in this source function over the mean path
between the matrix and the adjacent fracture. In the non-coupled dual-porosity formulation,
flow in the fractures acts as a boundary condition for flow in the matrix (Kazemi and
Gilman, 1993). This transfer function and the amount of fluid that is transferred from the
matrix to the fracture are directly proportional to the shape factor. Numerical simulation of
naturally fractured reservoirs using a dual-porosity approach requires a precise value of the
shape factor for the entire period of the production time.
In general, there are two models to consider the matrix and fracture interaction including
pseudo-steady state and transient transfer. The former model ignores the pressure transient
in the matrix while the latter model accounts for the pressure transient in the matrix. The
matrix-fracture shape factor for a slightly compressible fluid can be obtained using the
following equation (Lim and Aziz, 1995):
) (
f m
m
m
m m
p p
t
p
k
c
c
c
=
|
o ,
(2.1)
where is the fluid viscosity, c
m
, |
m
, and k
m
are the total isothermal compressibility,
porosity and permeability, respectively,
m
p shows the average pressure of the matrix
block,
f
p is the fracture pressure and is the matrix-fracture transfer shape factor with
dimension of L
2
. In a pseudo-steady state model the matrix blocks are considered as a
14
lumped system with an average pressure,
m
p , while in the transient model one needs to
find the solution of the pressure diffusivity equation given by:
t
p
c p
k
m
m m m
m
c
c
=
|
|
.
|
\
|
V V |
,
(2.2)
For a slightly compressible fluid, negligible variation of the fluid viscosity and isothermal
compressibility with the pressure leads to a linear flow equation for the pressure variation
in the matrix. This equation can be solved by common analytical or semi-analytical
methods such as the Laplace transform or separation of variables method (Zimmerman et
al., 1996; Lim and Aziz, 1995; Hassanzadeh et al., 2009; Shan and Pruess, 2005;
Zimmerman et al., 1993).
Determination of the matrix-fracture transfer shape factor for a slightly compressible fluid
based on the pseudo-steady state or transient transfer model has been studied in the past.
Investigators have considered the effect of fracture boundary conditions on the dual-
porosity formulation and shape factor for the slightly compressible fluid (Chang, 1995;
Hassanzadeh and Pooladi-Darvish, 2006; Rangel-German et al., 2010). It has been shown
that the fracture pressure and its variation with time affect the transient and pseudo-steady
state value of the shape factor.
There have been new efforts to determine the shape factors for multi-phase flow and
thermal methods in fractured porous media (Rangel-German et al., 2010; Civan and
Rasmussen, 2002; van Heel et al., 2008). There have also been a few reports in the
literature to model dual-porosity systems for compressible fluids with different approaches
than this study (Lu and Connel, 2007; Penuela et al., 2002). A more detailed review of
shape factor developments was discussed elsewhere (Ranjbar and Hassanzadeh, 2011).
Although the dual-porosity approach with the shape factor concept has some limitations, it
has been widely used and well accepted approach in hydrological sciences and petroleum
reservoir modeling. This may be because of its simplicity, computational efficiency and
flexibility in application to various fluid flow and transport problems. In addition, lack of
more advanced and efficient models that can accurately take into account the matrix-
fracture interaction have contributed to extensive use of the dual-porosity models.
Currently, the majority of commercial flow simulators use the dual-porosity approach.
15
However, it should be pointed out that a new line of attack on tackling fluid flow and
transport in fractured rocks has been recently introduced based on discrete fracture network
models (Hoteit and Firoozabadi, 2005). Hoteit and Firoozabadi (2005) presented a discrete
fracture model for single phase flow of compressible fluids in heterogeneous and fractured
media. They developed a numerical model by combining the mixed finite element and the
discontinuous Galerkin methods for multi-component gas flow. Discrete fracture model
also have been used for multiphase flow and water injection in fractured media (Karimi-
Frad and Firoozabadi, 2003; Lemmonnier and Bourbiaux, 2010).
It has been reported in the literature that in the case a slightly compressible fluid the
pseudo-steady state value of the matrix-fracture shape factor is a function of the pressure
decline regime in the fracture. Contrary to the slightly compressible fluid case, the
variation of the isothermal compressibility and viscosity with pressure cannot be ignored
when dealing with a compressible fluid. This leads to a nonlinear PDE. Therefore, the
reported shape factors for slightly compressible fluids cannot be applied for compressible
fluids or their application has not been validated in the previous studies. In a recent study
we derived the matrix-fracture shape factor for a compressible fluid in dual-porosity
systems (Ranjbar and Hassanzadeh, 2011). The effect of fracture pressure decline on the
compressible fluid shape factor has not been reported in the previous studies. In this study,
we further develop our previous study to investigate the effect of pressure decline in the
fracture on the matrix-fracture transfer shape factor for a compressible fluid.
We study the influence of the fracture pressure (as a boundary condition for the matrix
block) on the shape factor for flow of a compressible fluid in a dual-porosity model which
has not been investigated in the former works. To obtain the matrix-fracture shape factor a
nonlinear diffusivity equation is solved using the heat integral method and the method of
moments. To consider the effect of the time variation of the boundary conditions a
modified trial solution (early time) and Duhamels theorem (late time) are used to derive
the early and late time shape factors for the declining fracture pressure cases. The
developed approximate analytical solution is validated by a numerical model (Ranjbar and
Hassanzadeh, 2011). The developed shape factor model can recover predictions from the
shape factor models available in literature for a slightly compressible fluid. This shape
16
factor may find applications in dual-porosity modeling of the conventional and
unconventional naturally fractured gas reservoirs such as coalbed methane and fractured
tight gas reservoirs.
This paper is organized in a manner that follows a methodology for derivation of the shape
factor for compressible fluids. Next solution of the nonlinear diffusivity equation subject to
a declining fracture pressure is obtained using the heat integral and moment methods and
Duhamels theorem. Afterwards model verification and results are discussed followed by
conclusions.
2.2 Methodology
In this section the shape factor for flow of a compressible fluid from a matrix block under
different fracture boundary conditions is derived by taking into account the pressure
dependency of the viscosity and isothermal compressibility. Darcys law for flow of gas in
the porous media is expressed as follows:
dx
dp
B
A k
q
g
m
gsc
= ,
(2.3)
where A is the cross-section area and B
g
is the gas formation volume factor. Using the
definitions of the gas formation volume factor and real gas pseudo-pressure (Ikoku, 1992)
and writing the Darcys law over some characteristics length l leads to the following
equation:
2
f m
sc
sc m
sc
l
A
Tp
T k
q

= .
(2.4)
As shown in Equation (2.4), l is a length where the matrix pressure is equal to its average
pressure and this length changes with time during transient matrix production. For a
matrix-fracture combination shown in Figure 2.1, Equation (2.4) is multiplied and divided
by the bulk volume of the matrix-block to define the transfer function for compressible
fluids. Using the definition of the shape factor (Equation (2.5)); the final equation for the
matrix-fracture transfer function for compressible fluids (e.g. gases), is expressed as
Equation (2.6) (Ranjbar and Hassanzadeh, 2011):
17
) 2 / (
b
V l
A
= o ,
(2.5)
) (
4
f m
m
sc
b sc
sc
T
k
p
V T
q
o
= ,
(2.6)
In this equation T is the absolute temperature, is the shape factor,
m
shows the average
matrix block pseudo-pressure and
f
is the fracture pseudo-pressure.
According to the Warren and Root (1963) dual-porosity model for a slightly compressible
fluid, the interporosity flow rate per unit volume of the rock can be expressed in terms of
the accumulation rate in the matrix as follows:
t
p
c q
m
m m
c
c
= | .
(2.7)
The interporosity flow rate for compressible fluids can be expressed as follows (Ranjbar
and Hassanzadeh, 2011):
t T
c
p
V T
q
m m m
sc
b sc
sc
c
c
=
|
4
.
(2.8)
Combination of Equations (2.6) and (2.8) leads to the following equation for single-phase
shape factor of compressible fluids as follows (Ranjbar and Hassanzadeh, 2011):
( ) t k
c
m
f m m
m m
c
c
=

|
o .
(2.9)

Figure 2.1: Schematic of the matrix-fracture model.

x
h
m

L
c

F
r
a
c
t
u
r
e

M
a
t
r
i
x

18
There is another alternative to derive the shape factor for compressible fluids by integrating
of the diffusivity equation over the matrix-block volume. This method was used by
Zimmerman et al., (1993) to derive the shape factor for slightly compressible fluids and
leads to Equation (2.1). By integrating of the gas diffusivity equation over half of the
matrix block volume we reach to the following equation,
Adx
x x
k
V t
c
m m
b
m
m m
) (
2 /
1
c
c
c
c
=
c
c
}

| .
(2.10)
Simplification of Equation (2.10) leads to the following equation:
x
k
V
A
t
c
m m
b
m
m m
c
c
=
c
c
|
2 /
.
(2.11)
Using the Warren and Root (1963) approximation we have,
l x
m f
m

=
c
c
.
(2.12)
Where l is the characteristics length, which is the distance from the matrix-fracture
boundary where the matrix pressure is equal to its average pressure. By substituting this
equation in Equation (2.11) we reach to the following equation:
) (
) 2 / (
m f
m
b
m
m m
k
V l
A
t
c
| =
c
c
.
(2.13)
Using the definition of the shape factor given by Equation (2.5) we reach to the following
equation for the shape factor of compressible fluids, which is similar to the equation that
was obtained by Zimmerman et al., (1993) for slightly compressible fluids. It should be
noted that in this equation pseudo-pressure is appeared in the final equation as we are
dealing with compressible fluids.
) (
m f
m
m
m m t
k
c

|
o
c
c
= .
(2.14)
This equation is the same as Equation (2.9). It should be pointed out that for a
compressible fluid the viscosity-isothermal compressibility product is a strong function of
pressure, in Equations (2.9) or (2.14) the solution of the nonlinear gas diffusivity equation
is utilized to determine the shape factor for different pressure regimes in the fracture.
19
2.2.1 Constant fracture pressure
In this case it is assumed that at the matrix-fracture interface, the fracture pressure and
hence the pseudo-pressure is a constant. For this case the dimensionless variables are
defined as follows:
i f
i m
D

= ,
(2.15)
c
D
L
x
x = ,
(2.16)
q
q
q
m
D
= ,
(2.17)
2
c
D
L
t
t
q
= .
(2.18)
In Equation (2.17) the average hydraulic diffusivity, q is given by (Ranjbar and
Hassanzadeh, 2011):
} }

=
=
f
i
f
i
p
p m i f m
m
p
p m m
m
i f
c
dp
p p
k
dp
c
k
p p | |
q
1 1
. (2.19)
Using the definition of dimensionless variables (Equations (2.15), (2.17) and (2.18)) in the
shape factor equation (Equation (2.9)), the subsequent equation for the dimensionless
shape factor in the case of the constant fracture pressure is obtained:
D
D
D D
m
t
h
c
c
|
|
.
|
\
|
=

q
o
1
1 4
2
. (2.20)
2.2.2 Linearly declining fracture pressure
For a linearly declining fracture pressure we have the following equation for the fracture
pseudo-pressure as given by:
t
t
i f
1
), 1 ( s = o o (2.21)
where is a decline constant. For this case the dimensionless pseudo-pressure and the
dimensionless fracture pseudo-pressure are defined as follows:
20
i
m i
D

= ,
(2.22)
D
D D fD
t
t t
1
, ) ( s = k k
(2.23)
where is the dimensionless decline constant and is defined as follows:
D
t
t o
k = ,
(2.24)
Applying the explanation of the dimensionless variables, Equations (2.17), (2.18), (2.22)
and (2.23), in the shape factor equation (Equation (2.9)) leads to the following equation for
the shape factor in the case of the linearly declining fracture pressure:
|
|
|
|
.
|
\
|
c
c
=
D D
D
D
D
m
t
t
h
k
q
o
4
2
.
(2.25)
2.2.3 Exponentially declining fracture pressure
For this case the fracture pressure declines exponentially with time according to the
following equation:
) ( exp ) ( t
i f
o + =

,
(2.26)
where ) ( =
t
f
. For an exponential decline, the dimensionless pseudo-pressure
and the fracture dimensionless pseudo-pressure are defined as follows:
i
i m
D D
t

) ( ,
(2.27)
) ( exp 1 ) (
D D fD
t t k = ,
(2.28)
where is the dimensionless decline constant and is defined in Equation (2.24). Using the
definition of dimensionless variables (Equations (2.17), (2.18), (2.27) and (2.28)) in the
shape factor equation (Equation (2.9)) leads to the following equation for the shape factor
in the case of the exponentially declining fracture pressure:
)) ( exp 1 (
4
2
D D
D
D
D
m
t
t
h
k
q
o

c
c
= .
(2.29)
21
2.3 The approximate analytical solutions
The compressible fluid diffusivity equation for linear flow can be stated as:
t k
c
x
m
m
m m m
c
c
=
c
c |
2
2
.
(2.30)
Strong pressure dependence of the viscosity and isothermal compressibility leads to a
nonlinear partial differential equation (PDE) for compressible fluid flow in fractured
porous media. Solution of this PDE cannot be obtained by common methods like Laplace
transform or separation of variables. Equation (2.30) in term of the matrix hydraulic
diffusivity,
m
= k
m
/c
m
|
m
is expressed as follows:
2
2
) (
x
p
t
m
m
m
c
c
=
c
c
q
.
(2.31)
In Equation (2.31), hydraulic diffusivity is a space and time dependent parameter. To solve
Equation (2.31), we neglect the isothermal compressibility-viscosity product variation with
space and effect of the space is considered by a correction factor, (Ranjbar and
Hassanzadeh, 2011; Agarwal, 1979). Fine-grid numerical simulations are used to
determine this correction factor. Since the gas compressibility is orders of magnitude larger
than the rock compressibility we ignore the rock compressibility in the solution (Agarwal,
1979; Ikoku, 1992). Therefore, we reach the following PDE with the initial and boundary
conditions.
) ) ( (
x
t
x t
m
m
m
c
c
c
c
=
c
c
|q
, (2.32)
i m
t = = 0 , (2.33a)
0 0 =
c
c
=
x
x
m
, (2.33b)
f m c
L x = = .
(2.33c)
In Equations (2.32) and (2.33), is used to correct the effect of space on the hydraulic
diffusivity; L
c
is characteristic length of the matrix-block which is half of the matrix-block
thickness (h
m
). Figure 2.1 illustrates a graphical representation of the matrix-fracture
system.
22
A solution for the constant fracture pressure is given in our recent work (Ranjbar and
Hassanzadeh, 2011). Since this solution will be used as a basis for the time-dependent
boundary condition, the final form of the solution is given in the following. Using an
integral method (Finlayson, 1972; Goodman, 1964; Pooladi-Darvish, 1994; Zimmerman
and Bodvarsson, 1989) the early time solution for constant fracture pressure can be found
as follow (Ranjbar and Hassanzadeh, 2011):
1
3
1 1 1
3
1
24
1
, )
24
1
1 (
24 24
) 1 24 (
D
D
D D
D
D D D D
D D D
D
t
t
x
t t
t x
|q |q |q |q
|q
<
=
+
= .
(2.34)
1
1
24
1
,
4
24
D
D
D D
D
t
t
|q
|q
< = . (2.35)
where
D
is the average pseudo-pressure and
D1
is hydraulic diffusivity of the fracture in
dimensionless form and is expressed as follows:
q
|
q
q
q
m f f m
D m
D
c k
x
/
) 1 (
1
=
=
= .
(2.36)
The late time solution can be obtained by the method of moments (Chang, 1995; Ames,
1965; Crank, 1975) as given by (Ranjbar and Hassanzadeh, 2011):
, )) exp( 686 . 5 ) exp( 541 . 0 (
24
1
)) exp( 175 . 6 ) exp( 793 . 1 (
)) exp( 489 . 0 ) exp( 252 . 1 1 ( ) , (
3
2 1
1
2
2 1
2 1
D D D
D
D D D D
D D D D D
x t t
t x t t
t t t x

|q

+
> +
+ + =

(2.37)
1
2 1
24
1
), exp( 148 . 0 ) exp( 790 . 0 1
D
D D D D
t t t
|q
> = .
(2.38)
where
1 1
486 . 2
D
|q = , (2.39)
1 2
181 . 32
D
|q = , (2.40)
Substituting early and late time average dimensionless pseudo-pressures (Equations (2.35)
and (2.38)) and their derivatives in Equation (2.20) leads to the following equation for the
23
dimensionless shape factor for flow of a compressible fluid from a matrix block subject to
a constant fracture pressure boundary condition:
1 1
1 2
24
1
,
1
)
24 4
1
(
6 4
D
D
D D D D
D
m
t
t t
h
|q |q q
|q
o <
=
(2.41)
1 2 1
2 1 1 2
24
1
,
) exp( 148 . 0 ) exp( 790 . 0
) exp( 763 . 4 ) exp( 964 . 1 4
D
D
D D
D D
D
D
m
t
t t
t t
h
|q

q
|q
o >
+
+
= .
(2.42)
where parameters and
D
were obtained by matching the early and late time cumulative
production from the matrix to the fracture by a numerical flow simulator (Geo-Quest,
2009).
For the linearly declining fracture pressure the diffusivity equation and its initial and
boundary conditions are expressed as follows:
) ) ( (
D
D
D
D D
D
x
t
x t c
c
c
c
=
c
c
|q
,
(2.43)
0 0 = =
D D
t , (2.44a)
, 0 0 =
c
c
=
D
D
D
x
x

(2.44b)
D D fD D D
t t x k = = = ) ( 1 . (2.44c)
For the early time solution of these equations we assume that it has the following form:
3
)
) ( 1
1
1 ( ) , (
D
D
D D D D
t
x
t t x
o
k
= .
(2.45)
When the boundary condition changes with time the penetration depth in the heat balance
integral method (HBIM) is found by solving the following ordinary differential equation
(Mitchel and Myers, 2010):
) (
) (
) 1 (
) (
1
) ( ) (
2
D
D fD
D
D D fD
D
t
t n
n
t
n
t t
dt
d
o
u
u o
o +
=
(
+
.
(2.46)
In this equation n is the exponent in the trial solution, n=3 for our case and o o =1 . In
Equation (2.46), can be found from the following equation (Mitchel and Myers, 2010):
24
1 2
) 1 (
2

c
c
=
n
n n
t
fD
D
fD
o
u .
(2.47)
Solving Equation (2.46) for a linearly declining fracture pressure leads to the following
penetration depth for this case:
D D
t
1
8 1 |q o ~ . (2.48)
It should be noted that Equation (2.48) is obtained by assuming =0 in Equation (2.46)
(Mitchel and Myers, 2010). If we do not use the assumption of =0 Equation (2.46) cannot
be solved analytically. The derivation of this equation is shown in Appendix 2.A1 in more
details. Our numerical results show that we can obtain a more accurate solution if we use
the following equation for the penetration depth:
D D
t
1
9 1 |q o = . (2.49)
The early time solution is valid till the penetration depth reaches the inner boundary, so we
can find the time at which the pressure disturbance reaches the boundary (t
*
) as follows:
1
* *
1
9
1
9 1 0
D
D
t t
|q
|q = = .
(2.50)
Therefore, the early time solution of the partial differential equation (2.43) with the
boundary conditions (2.44) can be expressed as follows:
1
3
1
9
1
, )
9
1
1 ( ) , (
D
D
D D
D
D D D D
t
t
x
t t x
|q |q
k <
= .
(2.51)
Integrating of Equation (2.51) over the matrix block volume, leads to Equation (2.52) for
the early time average dimensionless pseudo-pressure:
1
2 / 3 1
9
1
,
4
9
D
D D
D
D
t t k
|q
|q
< = . (2.52)
The time dependence of the boundary condition for the late time solution can be
considered using Duhamels theorem. When the fracture pseudo-pressure varies with time
(Equation (2.44c)), Duhamels theorem provides the basis to solve the problem with
variable boundary conditions based on the solution provided for the constant fracture
pseudo-pressure. Using Duhamels theorem (Chang, 1995; Ozisik, 1993; Polyanin, 2001)
25
the solution of partial differential Equation (2.43) with the boundary conditions, Equations
(2.44b) and (2.44c) can be expressed as:
t t t d t x
t
D
t
D D D fD
D
D
}

c
c
=
0
) , ( ) ( . (2.53)
In Equation (2.53),
D
within the integral is the solution when 1 =
fD
and
D
on the left-
hand side is the solution of PDE (2.43) when the matrix-fracture boundary condition
changes with time.
Using Duhamels theorem leads to the following late time solution for the case of the
linearly declining fracture pressure:
1
2
3 2
2
1
3 2
1
9
1
), 1 ) )(exp( 086 . 1 086 . 0 (
47256 . 5
) 1 ) )(exp( 314 . 3 314 . 2 (
55790 . 0
) , (
D
D D D D
D D D D D D D
t t x x
t x x t t x
|q
k
k
> +
+ + =

(2.54)
Derivation of Equation (2.54) is shown in Appendix 2.A1 in more details. The late time
average matrix block pseudo-pressure for the linearly declining fracture pressure is
obtained as follows:
1
2
2
1
1
1
0
9
1
), 1 ) (exp(
14229 . 0
) 1 ) (exp(
81416 . 0
) , (
D
D D D D
D D D D D
t t t t
dx t x
|q
k
k

>
= =
}

(2.55)
Using the average pseudo-pressure and its derivative in the shape factor equation (Equation
(2.25)) results in the following equations for the early and late time shape factors in the
case of the linearly declining fracture pressure:
1 1
1 2
9
1
,
1
) 1
4
9
(
9
2
3
D
D
D
D
D
D
D
m
t
t
t
h
|q |q
|q
q
o <
=
(2.56)
1 2 1
2 1 1 2
9
1
,
) 1 ) (exp( 00442 . 0 ) 1 ) (exp( 32750 . 0
) exp( 14229 . 0 ) exp( 81416 . 0 1 4
D
D
D D
D D
D
D
m
t
t t
t t
h
|q

q
|q
o >
+

=
(2.57)
It should be noted that for the linearly declining fracture pressure, the shape factor for
compressible fluid is not a function of the dimensionless decline constant, ; a similar
26
observation was reported by Hassanzadeh and Pooladi-Darvish (2006) for flow of a
slightly compressible fluid in fractured porous media.
For the exponentially declining fracture pressure the solution of the following PDE should
be used in Equation (2.29) to derive the shape factor for this case:
) ) ( (
D
D
D
D D
D
x
t
x t c
c
c
c
=
c
c
|q
,
(2.58)
0 0 = =
D D
t , (2.59a)
, 0 0 =
c
c
=
D
D
D
x
x

(2.59b)
) exp( 1 ) ( 1
D D fD D D
t t x k = = = (2.59c)
The following function is assumed for the early time solution which satisfies the outer
boundary condition:
3
)
) ( 1
1
1 ))( exp( 1 ( ) , (
D
D
D D D D
t
x
t t x
o
k
= .
(2.60)
Solving the ODE of Equation (2.46) leads to the following equation for the penetration
depth in the case of an exponentially declining fracture pressure:
)
) (
) exp( 1 ) exp( 1
2
( 12 1
1
D
D
D D
D D
t
t erf
t t
t
k
k
k
t
k
|q o

~ . (2.61)
The derivation of this equation is illustrated in Appendix 2.A2. It should be noted that in
Equation (2.61), erf(x) is the error function defined as follows:
}

=
D
t
y
D
dy e t erf
k
t
k
0
2
2
) ( . (2.62)
Since Equation (2.61) is obtained based on an approximation of u=0 in Equation (2.46),
our numerical results show that one can obtain a more accurate solution if we use the
following equation for penetration depth:
)
) (
) exp( 1 ) exp( 1
2
( 96 . 12 1
1
D
D
D D
D D
t
t erf
t t
t
k
k
k
t
k
|q o

= . (2.63)
27
It should be noted that the effect of pressure disturbance will reach the inner boundary
when 0 = o and for the exponential decline we cannot obtain an explicit equation for t
*

and t
*
is determined for any values of k by making Equation (2.63) equal to zero.
Therefore, the early time solution of Equations (2.58) and (2.59) can be expressed as
follows:
* 3
1
, )
) (
) exp( 1 ) exp( 1
2
( 96 . 12
1
1 (
)) exp( 1 ( ) , (
t t
t
t erf
t t
t
x
t t x
D
D
D
D D
D D
D
D D D D
<

=
k
k
k
t
k
|q
k

(2.64)
Integrating of Equation (2.64) over the matrix block volume, leads to Equation (2.65) for
the early time average dimensionless pseudo-pressure:
*
1
, )
) (
2 ( 96 . 12
4
) exp( 1
t t
t
t erf
t
t
D
D
D
D D
D
D
<

=
k
k t
|q
k
(2.65)
Duhamels theorem and the verified solution of the constant fracture pressure boundary
condition lead to the following equation for the late time dimensionless pseudo-pressure in
the case of the exponentially declining fracture pressure:
), exp(
) 086 . 1 086 . 0 )(
) exp( ) exp(
) exp( 1
( 48743 . 1
) 314 . 3 314 . 2 )(
) exp( ) exp(
) exp( 1
( 48743 . 0
) 086 . 1 086 . 0 )( exp( )
) exp( ) exp(
) exp( 1
( 48743 . 1
) 314 . 3 314 . 2 )( exp( )
) exp( ) exp(
) exp( 1
( 48743 . 0
) exp( 1 ) , (
3 2
* *
2
*
3 2
* *
1
*
* 3 2
2 * *
2
*
3 2
1 * *
1
*
D
D D
D D
D D D D
D D D
D D D D
t
x x
t t
t
x x
t t
t
t t x x t
t t
t
x x t
t t
t
t t x
k
k
k
k
k
k
k
k
k
k
(
(
(
(
+

+

+
> +

+
+

=

(2.66)
More details about the derivation of Equation (2.66) are discussed in Appendix 2.A2.
Integrating over the matrix-block bulk volume results in the following equations for the
average dimensionless pseudo-pressure in the case of the exponentially declining fracture
pressure:
28
), exp( )
) exp( ) exp(
) exp( 1
( 03867 . 0 )
) exp( ) exp(
) exp( 1
( 71132 . 0
) exp( )
) exp( ) exp(
) exp( 1
( 03867 . 0
) exp( )
) exp( ) exp(
) exp( 1
( 71132 . 0 ) exp( 1 ) , (
* *
2
*
* *
1
*
*
2 * *
2
*
1 * *
1
*
D
D D
D D D D D
t
t t
t
t t
t
t t t
t t
t
t
t t
t
t t x
k
k
k
k
k
k
k
k
k
k

+

> +

=
(2.67)
Using the average pseudo-pressure and its derivative in Equation (2.29) leads to the
following equations for the early and late time shape factors in the case of the
exponentially declining fracture pressure:
*
1
2 / 3
1 1
2
,
)) exp( 1 ( )
) (
2 ( 96 . 12
4
) exp( 1
)
) (
2 ( 2
)
) (
2 ( )
) ( 2
) ( ) exp(
(
)) exp( 1 ( 96 . 12 )
) (
2 ( 96 . 12
) exp( 1 2
) exp(
1
t t
t
t
t erf
t
t
t
t erf
t
t
t erf
t
t erf
t
t
t
t
t
t erf
t
t
t
h
D
D
D
D
D D
D
D
D
D
D
D
D
D
D
D
D
D D
D
D
D D
D
D
D
m
<

(
(
(
(
(
|
|
.
|
\
|
+ +

+

=
k
k
k
t |q
k
k
k
t
k
k
t
k
k t k k
k |q
k
k
t |q
k
k k
q
o

(2.68)
*
2 * *
2
*
1 * *
1
*
* *
2
*
* *
1
*
2 * *
2
*
2
1 * *
1
*
1
* *
2
*
* *
1
*
2
,
) exp( )
) exp( ) exp(
) exp( 1
( 03867 . 0 ) exp( )
) exp( ) exp(
) exp( 1
( 71132 . 0
) exp( )
) exp( ) exp(
) exp( 1
( 03867 . 0 )
) exp( ) exp(
) exp( 1
( 71132 . 0
)) exp( )
) exp( ) exp(
)) exp( 1 (
( 03867 . 0 ) exp( )
) exp( ) exp(
)) exp( 1 (
( 71132 . 0
)
) exp( ) exp(
) exp( 1
( 03867 . 0 )
) exp( ) exp(
) exp( 1
( 71132 . 0 1 ) exp(
4
t t
t
t t
t
t
t t
t
t
t t
t
t t
t
t
t t
t
t
t t
t
t t
t
t t
t
t
h
D
D D
D
D D
D
D
m
>
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k k
q
o

(2.69)
29
Based on Equations (2.68) and (2.69), for an exponentially declining fracture pressure the
shape factor for a compressible fluid is a function of the decline exponent, k. Similar
observations have been made in the previous studies for a slightly compressible fluid
(Chang, 1995; Hassanzadeh and Pooladi-Darvish, 2006).
2.4 Model verification
The developed shape factor was validated by a fine-grid single porosity model (Eclipse
100). The total cumulative production from the matrix to the fracture based on the
simulator was used to find the correction factor () and
D
and to validate the presented
model. Figure 2.2 shows the matrix-fracture cumulative fluid production versus time for a
case with =0.7, T=93.3C and pressure drawdown of 45 to 22.5 MPa. The values obtained
for the correction factor (), the matching parameter (
D
), the average hydraulic diffusivity
(q ) and the dimensionless fracture hydraulic diffusivity (
D1
) are 0.730, 0.3127, 0.03457
and 0.3691, respectively. In the model verification studies, a slab-shaped matrix-block with
thickness (h
m
) of 4m, permeability of 1mD, and porosity of 0.1 are considered. We use the
same reservoir data and parameters throughout this paper. As illustrated in Figure 2.2 the
approximate analytical model based on this study is in a good agreement with the fine grid
numerical simulation. More details about the numerical simulations and more validation
cases are discussed elsewhere (Ranjbar and Hassanzadeh, 2011).
The developed model with 1
1
= = =
D D
q q | must reproduce the shape factor for a slightly
compressible fluid. For additional validation of the developed model, the shape factor
derived in this study is evaluated with the shape factor of the slightly compressible fluid.
Outcomes show that the models developed for different boundary conditions can reproduce
the slightly compressible fluid shape factor for the complete period of time.
Figure 2.3 compares the developed shape factor model for a slightly compressible fluid (
1
1
= = =
D D
q q | ) and models available in the literature (Chang, 1995; Hassanzadeh and
Pooladi-Darvish, 2006) when the fracture pressure declines linearly with time. According
to this figure the presented model shows an acceptable match with other models.

30
Time (sec)
0 500 1000 1500 2000
C
u
m
u
l
a
t
i
v
e

F
l
u
i
d

P
r
o
d
u
c
t
i
o
n
(
S
m
3
)
0
100
200
300
400
500
600
700
Numerical
Compressible Fluid Model

Figure 2.2: Comparison of the matrix-fracture cumulative fluid production obtained from the
approximate analytical solution and the numerical model of Eclipse for constant fracture pressure.
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

S
h
a
p
e

f
a
c
t
o
r

(
o

h
m
2
)
1
10
100
1000
Hassanzadeh & Pooladi-Darvish Model
Chang Model

Figure 2.3: Comparison of the developed shape factor model with literature models for slightly
compressible fluid in the case of linearly declining fracture pressure.

31
Figures 2.4, 2.5 and 2.6 demonstrate the comparisons between the presented shape factor
models in this study with the literature models for a slightly compressible fluid when the
fracture pressure declines exponentially with time for different values of the decline
exponent. These figures demonstrate that the presented model can reproduce the slightly
compressible fluid shape factor with an acceptable accuracy.
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

S
h
a
p
e

F
a
c
t
o
r
(
o

h
m
2
)
1
10
100
1000

compressible fluid in the case of exponentially declining fracture pressure for small values of exponent
(k=0.0001).

32
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

S
h
a
p
e

F
a
c
t
o
r
(
o

h
m
2
)
1
10
100
1000
Chang Model

compressible fluid in the case of exponentially declining fracture pressure (k=0.632).
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

S
h
a
p
e

F
a
c
t
o
r
(
o

h
m
2
)
1
10
100
1000

compressible fluid in the case of exponentially declining fracture pressure (k=1).
33
2.5 Results
In this section a comparison of the developed model with the Warren and Root model
(1963) is presented and then the behaviour of the shape factor for different fracture
pressure depletion regimes for flow of a compressible fluid through dual-porosity media is
described.
2.5.1 Comparison of model with Warren and Root model
Warren and Root (1963) used a pseudo-steady state approach to derive the shape factor for
a slightly compressible fluid in the fractured media. They derived the following shape
factor for slightly compressible fluid for different sets of fractures,
2
) 1 ( 4
m
h
n n +
= o .
(2.70)
where n is the number of sets of fractures. For one set of fracture the derived value of the
shape factor is (12/h
m
2
). Figure 2.7 shows the comparison of the matrix-fracture cumulative
fluid production based on the Warren and Root shape factor, presented semi-analytical
model (time dependent shape factor) and numerical results.
Time (sec)
0 200 400 600 800 1000
C
u
m
u
l
a
t
i
v
e

F
l
u
i
d

P
r
o
d
u
c
t
i
o
n
(
S
m
3
)
0
100
200
300
400
500
600
700
800
Numerical
Warren & Root

Figure 2.7: Comparison of the developed model with numerical and Warren and Root model.
34
Based on Figure 2.7 it can be concluded that using a constant slightly compressible fluid
shape factor (Based on Warren and Root model) for flow of compressible fluids leads to
large error in prediction of the cumulative production from the matrix.
Figure 2.8 shows the shape factor for the linearly declining fracture pressure derived from
Equations (2.56) and (2.57). The shape factor for a constant fracture pressure is also shown
in this figure for comparison. As illustrated in this figure, for the linearly declining fracture
pressure, the transient period for the linear decline is longer than that of the constant
fracture pressure and the shape factor is stabilized at value of 10.38 when t
D
is about 3.81.
For the case of the constant fracture pressure the stabilized value of the shape factor is 8.57
when dimensionless time is about 0.7. The same behaviour has been reported for a slightly
compressible fluid by Chang (1995) and Hassanzadeh and Pooladi-Darvish (2006) in the
case of a slightly compressible fluid. From this figure it can be concluded that the transient
and pseudo-steady state values of the shape factor for a linear decline is larger than those
of the constant fracture pressure. It should be noted that like slightly compressible fluids
the value of the depletion rate has no impact on the transient and stabilized value of the
shape factor.
35
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

S
h
a
p
e

F
a
c
t
o
r

(
o

h
m
2
)
1
10
100
1000
Linear decline Fracture Pressure

Figure 2.8: Comparison of the shape factor for linearly declining and constant fracture pressure.

Figure 2.9 shows the effect of the exponent of the exponential decline on the shape factors
based on Equations (2.68) and (2.69). Different values of the decline exponent ranging
from 0.0001 to 1,000 are used. A large decline factor implies fast pressure depletion in the
fracture while a small value represents a slow depletion. As illustrated in Figure 2.9 for
small values of (<0.1), the shape factor begins at large values and subsequently
converges to a stabilized value of 10.38 as compared to 8.57 for a constant fracture
pressure case. As the value of the exponent increases (fast depletion) the transient and
pseudo-steady state values of the shape factor tend to those of the constant fracture
pressure boundary condition. When >10 the constant fracture pressure and the
exponentially declining fracture pressure have the same value of the stabilized shape
factor. Similar behaviour was reported for a slightly compressible fluid and exponentially
declining fracture pressure by Chang (1995) and Hassanzadeh and Pooladi-Darvish (2006).
36
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

S
h
a
p
e

f
a
c
t
o
r

(
o

h
m
2
)
1
10
100
1000
k=0.0001
k=0.4
k=100
k=1000

Figure 2.9: Shape factor comparison for different exponents for exponentially declining fracture
pressure.

Figure 2.10 shows the dimensionless shape factor for flow of a compressible fluid in a
dual-porosity medium for various pressure depletion regimes in the fracture. Results show
that the stabilized values of the shape factor vary from 8.57 for the constant fracture
pressure to 10.38 for the linearly declining fracture pressure. For the exponentially
declining fracture pressure the stabilized values vary between these two limits. For a very
small value of the exponent (slow depletion) the stabilized value is the same as that for the
linearly declining fracture pressure. However, as the value of the exponent increases, the
stabilized values of the shape factor for the exponential decline shift to the constant
fracture pressure value. At a large value of the exponential decline (fast depletion), the
constant and exponentially declining fracture pressures have the same stabilized value of
8.57 for the shape factor.
h
m
2
=8.57
h
m
2
=9.31
h
m
2
=10.38
37
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

S
h
a
p
e

f
a
c
t
o
r

(
o

h
m
2
)
1
10
100
1000
Exponential(k=0.0001)
Exponential(k=0.4)
Exponential(k=1000)
Linear decline

Figure 2.10: Comparison of the dimensionless shape factor for different pressure depletion regime in
the fracture.

The above results show that both the transient and pseudo-steady state values of the single-
phase shape factor depend on how the fracture pressure changes with time. It should be
noted for a linearly declining fracture pressure the stabilized value of the shape factor is
independent of the decline rate. On the other hand, for an exponential decline the stabilized
value of the shape factor depends on the decline exponent. Furthermore, the time
dependence of the fracture boundary condition on the stabilized value of the shape factor
can be described by using an exponentially declining regime with different decline
exponents. In such cases, the small decline exponents replicate the linear pressure decline
in the fracture whereas a large decline exponent reproduces the constant fracture pressure
boundary condition. Table 2.1 shows the stabilized values of the single-phase shape factor
and the time at which the effect of pressure disturbance reaches the inner boundary (t
*
) for
different pressure depletion regimes in the fracture. It should be pointed out that the
developed model is applicable for single-phase flow of a compressible fluid in the
fractured media.
h
m
2
=8.57
h
m
2
=9.31
h
m
2
=10.38
38
Table 2.1: Stabilized values of the shape factor and time at which the pressure disturbance reaches the
inner boundary for different depletion regimes in the fracture.
Depletion regime in the fracture t* Stabilized value of the dimensionless shape factor
Linear Decline 0.412 10.38
Exponential Decline (k=0.0001) 0.429 10.38
Exponential Decline (k=0. 4) 0.416 9.31
Exponential Decline (k=1000) 0.154 8.57
Constant Fracture Pressure 0.155 8.57

2.6 Conclusions
The following major conclusions are made as a result of this study:
- The matrix-fracture shape factor for single-phase flow of compressible fluids
illustrates a transient period and then stabilizes to a stable value throughout pseudo-
steady state transfer.
- The approximate analytical solution presented revealed that the matrix-fracture
transfer shape factor for single phase flow of a compressible fluid in the dual-
porosity media is a function of the pressure depletion regime in the fracture.
- Based on the pressure depletion regime in the fracture the stabilized value of the
shape factor varies between two limits. The upper limit is obtained for a linearly
declining fracture pressure which corresponds to a slow pressure depletion regime.
The lower limit is derived for the constant fracture pressure boundary conditions
where depletion takes place faster.
- When the fracture pressure depletes exponentially with time, the stabilized value of
the shape factor falls between those values of the constant fracture pressure and
linearly declining fracture pressure. This stabilized value is a function of the
exponent . For small exponent values the stabilized shape factor has the same
value as that for the linearly declining fracture pressure. For large exponent values,
39
the stabilized value of the shape factor is equal to that for a constant fracture
pressure.
- The pseudo-steady state time (stabilization time) of the shape factor increases as the
fracture boundary condition changes from a fast depletion regime toward a slow
depletion regime.

40
Nomenclature
A Cross-sectional area [L
2
]
B
g
Gas formation volume factor
c
m
Matrix total compressibility [LT
2
/M]
h
m
=2L
c
Matrix block length [L]
k
m
Matrix permeability [L
2
]
l Time dependent length where pressure is average pressure [L]
L
c
Matrix block characteristic length [L]
n Exponent in the polynomial trial solution using HBIM
m
p
Matrix-block pressure [M/LT
2
]
m
p
Average matrix-block pressure [M/LT
2
]
f
p
Fracture pressure [M/LT
2
]
sc
q
Matrix-fracture fluid transfer [L
3
/T]
q
Interporosity flow rate per unit volume of rock [1/T]
t Time [T]
t* Dimensionless time at which the pressure disturbance reach to the boundary
t
D
Dimensionless time
T Reservoir temperature [K]
V
b
Matrix-block volume [L
3
]
x
D
Dimensionless distance

Greek Symbols

o Decline constant [1/T]
Space correction factor

Gas specific gravity
o
Penetration depth
c Proportionality constant for penetration depth
m
q
Matrix hydraulic diffusivity [L
2
/T]
q Average hydraulic diffusivity [L
2
/T]
D
q
Dimensionless hydraulic diffusivity
1 D
q
Dimensionless fracture hydraulic diffusivity
k Dimensionless decline constant

Dimensionless exponent of solution of gas diffusivity equation using
moment method

Fluid viscosity [M/LT]
o Shape factor [1/L
2
]
t Duhamel s variable
41
| Porosity
m

Average matrix-block pseudo-pressure [M/LT
3
]
f

Fracture pseudo-pressure [M/LT
3
]
i

Initial pseudo-pressure [M/LT
3
]
D

Dimensionless pseudo-pressure
fD

Dimensionless fracture pseudo-pressure
D

Average dimensionless pseudo-pressure

Fracture pseudo-pressure when t
D
tends to infinity [M/LT
3
]

Subscripts

D Dimensionless
f Fracture
g Gas
i Initial condition
m Matrix
SC Standard conditions

42
References
Agarwal, R.G. (1979). Real gas pseudo-time-A new function for pressure build-up analysis
of MHF gas wells. SPE paper 8279.
Ames, W.F. (1965). Nonlinear partial differential equations in engineering. New York:
Academic Press.
Beckner, B.L. (1990). Improved modeling of imbibition matrix/fracture fluid transfer in
double porosity simulators. PhD dissertation, Stanford University.
Bourbiaux, B. Granet, S. Landereau, P. Noetinger, B. Sarda, S. Sabathier, J.C. (1999).
Scaling up matrix-fracture transfer in dual-porosity models: theory and application. SPE
Paper 56557.
double-permeability systems-A state-of-the-art review. Transp. Porous Med., 4, 147-
184.
Cihan, A. Tyner, J.S. (2011). 2-D radial analytical solutions for solute transport in a dual-
porosity medium. Water Resour. Res., 47(4), doi: 10.1029/2009WR008969.
Civan, F. Rasmussen, M.L. (2002). Analytical hindered-matrix-fracture transfer models for
naturally fractured petroleum reservoirs. SPE paper 74364.
Coats, K.H. (1989). Implicit compositional simulation of single-porosity and dual-porosity
reservoirs. SPE paper 18427.
Crank, J. (1975). The mathematics of diffusion. Oxford: Clarendon Press.
Deng, H. Dai, Z. Wolfsberg, A. Lu, Z. Ye, M. Reimus, P. (2010). Upscaling of reactive
mass transport in fractured rocks with multimodal reactive mineral facies. Water Resour.
Res., 46, W06501, doi: 10.1029/2009WR008363.
Di Donato, G. Blunt, M.J. (2004). Streamline-based dual-porosity simulation of reactive
transport and flow in fractured reservoirs. Water Resour. Res., 40, W04203, doi:
10.1029/2003WR002772.
43
Finlayson, B.A. (1972). The method of weighted residuals and variational principles. New
York: Academic Press.
Geo-Quest, (2009). Eclipse 100 technical descriptions 2009.1. Geo-Quest, Schlumberger.
Goodman, T.R. (1964). Application of integral methods to transient nonlinear heat transfer.
Advances in Heat Transfer, 1, 51-122, San Diego, CA: Academic.
matrix-fracture transfer shape factor. Transp. Porous Med., 64, 51-71.
solution for well testing of dual-porosity systems. Adv. Water Res., 32, 1652-1663.
Hoteit, H. Firoozabadi, A. (2005). Multicomponent fluid flow by discontinuous Galerkin
and mixed methods in unfractured and fractured media. Water Resour. Res., 41,
W11412, doi: 10.1029/2005WR004339.
Ikoku, C.U. (1992). Natural gas reservoir engineering. Florida: Krieger Press.
Karimi-Fard, M. Firoozabadi, A. (2003). Numerical Simulation of Water Injection in
Fractured Media Using the Discrete-Fracture Model and the Galerkin Method. Soc. Pet.
Eng. J., 6(2), 117-126.
Kazemi, H. Gilman, J.R. (1993). Multiphase flow in fractured petroleum reservoirs. In: J.
Bear, C. F. Tsang, and G. de Marsily (eds.), Flow and Contaminant Transport in
Fractured Rock. Academic Press, San Diego, 267-323.
Kazemi, H. Merrill, L.S. Porterfield, K.L. Zeman, P.R. (1976). Numerical simulation of
water-oil flow in naturally fractured reservoirs. Soc. Pet. Eng. J., 16(6), 317-326.
Lemonnier, P. Bourbiaux, B. (2010). Simulation of naturally fractured reservoirs. State of
the art. Part 1: Physical mechanisms and simulator formulations. Oil and Gas Science
and Technology, 65(2), 239-262.
Lim, K.T. Aziz, K. (1995). Matrix-fracture transfer shape factors for dual-porosity
simulators. J. Pet. Sci. Eng., 13, 169-178.
Liu, M.X. Chen, Z.X. (1990). Exact solution for flow of slightly compressible fluids
through multiple-porosity, multiple-permeability media. Water Resour. Res., 26(7),
1393-1400.
44
Lu, M. Connel, L.D. (2007). A dual-porosity model for gas reservoir flow incorporating
adsorption behavior-part I. Theoretical development and asymptotic analysis. Transp.
Porous Med., 68,153-73.
Mitchel, S.L. Myers, T.G. (2010). Improving the accuracy of heat balance integral methods
applied to thermal problems with time dependent boundary conditions. Int. J. Heat Mass
Transfer, 53, 3540-3551.
zisik, M.N. (1993). Heat Conduction. United States: John Wiley & Sons Inc.
Penuela, G. Civan, F. Hughes, R.G. Wiggins, M.L. (2002). Time-dependent shape factors
for interporosity flow in naturally fractured gas-condensate reservoirs. SPE paper 75524.
Polyanin, A.D. (2001) Handbook of linear partial differential equations for engineers and
scientists. United States: CRC Press.
Pooladi-Darvish, M. Tortike, W.S. Farouq Ali, S.M. (1994). Non-isothermal gravity
drainage under conduction heating. Petroleum Society of CIM and AOSTRA. Paper NO
94-65.
Quintard, M. Whitaker, S. (1996) Transport in chemically and mechanically heterogeneous
porous media. I: theoretical development of region-averaged equations for slightly
compressible single-phase flow. Adv. Water Res., 19(1), 29-47.
Quintard, M. Whitaker, S. (1996). Transport in chemically and mechanically
heterogeneous porous media II: comparison with numerical experiments for slightly
compressible single-phase flow. Adv. Water Res., 19, 49-60.
Quintard, M. Whitaker, S. (1998). Transport in chemically and mechanically
heterogeneous porous media III: Large-scale mechanical equilibrium and the regional
form of Darcy's law. Adv. Water Res., 21(7), 617-29.
Rangel-German, E. Kavscek, A.R. Akin, S. (2010). Time-dependent shape factors for
uniform and non-uniform pressure boundary conditions. Transp. Porous Med., 83, 591-
601.
flow of a compressible fluid in dual-porosity media. Adv. Water Res., 34(5), 62739.
Shan, C. Pruess, K. (2005). An analytical solution for slug tracer tests in fractured
reservoirs. Water Resour. Res., 41:W08502, doi:10.1029/2005WR004081.
45
Thomas, L.K. Dixon, T.N. Pierson, R.G. (1983). Fractured reservoir simulation. Soc. Pet.
Eng. J., 23(1), 42-54.
Ueda, Y. Murata, S. Watanabe, Y. Fanatsu, K. (1989). Investigation of the shape factor
used in the dual-porosity reservoir simulator. SPE Paper 19469.
van Heel, A.P.G. Boerrigter, P.M. van Dorp J.J. Thermal and hydraulic matrix-fracture
interaction in dual permeability simulation. Soc. Pet. Eng. J. (reservoir evaluation and
engineering), 11(4), 735-749.
Warren, J.E. Root, P.J. (1963). The behavior of naturally fractured reservoirs. Soc. Pet.
Eng. J., 3(3), 245-255.
Zimmerman, R.W. Bodvarsson, G.S. (1989). An approximate solution for one-dimensional
absorption in unsaturated porous media. Water Resour. Res., 25(6), 14221428.
Zimmerman, R.W. Chen, G. Hadgu, T. Bodvarsson, G.S. (1993). A numerical dual-
porosity model with semi-analytical treatment of fracture/matrix flow. Water Resour.
Res., 29(7), 2127-2137.
Zimmerman, R.W. Hadgu, T. Bodvarsson, G.S. (1996). A new lumped-parameter model
for flow in unsaturated dual-porosity media. Adv. Water Res., 19(5), 317-327.

46
Appendix 2.A: Solution of the gas diffusivity equation for different fracture depletion
conditions
2.A1: Linearly declining fracture pressure
In this case the following PDE with initial and boundary conditions (Equations (2.A1.2a)-
(2.A1.2c)) should be solved:
) ) ( (
D
D
D
D D
D
x
t
x t c
c
c
c
=
c
c
|q
,
(2.A1.1)
0 0 = =
D D
t , (2.A1.2a)
, 0 0 =
c
c
=
D
D
D
x
x

(2.A1.2b)
D D fD D D
t t x k = = = ) ( 1 . (2.A1.2c)
For the early time solution we use the following trial solution:
3
)
) ( 1
1
1 ( ) , (
D
D
D D D D
t
x
t t x
o
k
= .
(2.A1.3)
When the boundary condition changes with time the penetration depth in the heat balance
integral method (HBIM) is found by solving the following ordinary differential equation
(Mitchel and Myers, 2010):
) (
) (
) 1 (
) (
1
) ( ) (
2
D
D fD
D
D D fD
D
t
t n
n
t
n
t t
dt
d
o
u
u o
o +
=
(
(
+
.
(2.A1.4)
In this equation n is the exponent in the trial solution (n=3 for our case) and o o =1 . In
Equation (2.A1.4), can be obtained by using the following equation:
1 2
) 1 (
2

c
c
=
n
n n
t
fD
D
fD
o
u .
(2.A1.5)
Assume that
D
t c o = and substituting this equation in the ODE equation of (2.A1.4)
leads to following ODE for the linearly declining fracture pressure:
D
D
D D D
dt
t
t n
n
t t
d
c
k c k
=
(
(
+
0
1
.
(2.A1.6)
47
It should be noted that this ODE is obtained by assuming =0 in Equation (2.A1.4)
(Mitchel and Myers, 2010). Integrating of Equation (2.A1.6) leads to the following
equation for :
8
3
) 1 ( 2
3
=
+
=
=
c c
n
n n
. (2.A1.7)
So we reach the following equation for penetration depth in the case of a linear PDE:
D D D
t t t 8 1 8 ~ = = o c o . (2.A1.8)
In the case of a nonlinear PDE we have the following equation for the penetration depth:
D D
t
1
8 1 |q o ~ . (2.A1.9)
As was illustrated in the model verification section we can obtain a more accurate solution
if we use the following equation for the penetration depth:
D D
t
1
9 1 |q o = . (2.A1.10)
Therefore, the early time solution of the partial differential Equation (2.A1.1) with the
boundary conditions (2.A1.2) can be expressed as follows:
1
3
1
9
1
, )
9
1
1 ( ) , (
D
D
D D
D
D D D D
t
t
x
t t x
|q |q
k <
= .
(2.A1.11)
Integrating of Equation (2.A1.11) over the matrix block volume, leads to the following
equation for the early time average dimensionless pseudo-pressure:
1
2 / 3 1
9
1
,
4
9
D
D D
D
D
t t k
|q
|q
< = . (2.A1.12)
For the late time solution in the case of the linearly declining fracture pressure the
following PDE should be solved. It should be noted that the initial condition for the late
time solution comes from the early time solution (Equation (2.A1.11)).
) ) ( (
D
D
D
D D
D
x
t
x t c
c
c
c
=
c
c
|q
,
(2.A1.13)
3
1 1
)
9
(
9
1
D
D
D
D
D
x t
|q
k
|q
= = , (2.A1.14a)
, 0 0 =
c
c
=
D
D
D
x
x

(2.A1.14b)
48
D D fD D D
t t x k = = = ) ( 1 . (2.A1.14c)
The time dependence of the boundary condition for the late time solution can be
considered by Duhamels theorem. When the fracture pseudo-pressure varies with time
(Equation (2.A1.14c)), Duhamels theorem provides the basis to solve the problem with
variable boundary conditions based on the solution provided for the constant fracture
pseudo-pressure. Using Duhamels theorem (Chang, 1995; zisik, 1993; Polyanin, 2001)
the solution of PDE (2.A1.13) with conditions (2.A1.14) can be expressed as:
1 0
9
1
, ) , ( ) (
D
D
t
D D D fD
D
D
t d t x
t
D
|q
t t t >
c
c
=
}
. (2.A1.15)
In this equation,
D
fD
(Equation (2.A1.16))
and
D
on the left-hand side is the solution of the PDE when the matrix-fracture boundary
condition changes with time.
. )) exp( ) exp( (
)) exp( 086 . 1 ) exp( 314 . 3 (
)) exp( 086 . 0 ) exp( 314 . 2 1 ( ) , (
3
2 2 1 1
2
2 2 1 1
2 2 1 1
D D D
D D D
D D D D D
x t m t m
x t m t m
t m t m t x

+
+
+ + + =

(2.A1.16)
Derivation of Equation (2.A1.16) has been shown in our previous study (Ranjbar and
Hassanzadeh, 2011). Using Duhamels theorem and substituting Equations (2.A1.16) and
(2.A1.14c) in Equation 2.A1.15 lead to the following late time solution for the case of the
linearly declining fracture pressure:
1
3
2
2
2
1
1
1
2
2
2
2
1
1
1
2
2
2
1
1
1
9
1
, ) 1 ) (exp( ) 1 ) (exp(
) 1 ) (exp(
086 . 1
) 1 ) (exp(
314 . 3
) 1 ) (exp(
086 . 0
) 1 ) (exp(
314 . 2
) , (
D
D D
D
D D
D
C
D D
D D D D D D
t x t
m
t
m
x t
m
t
m
t
m
t
m
t t x
|q
k
k
>
+
+
)
`
+ + =

(2.A1.17)
49
The initial condition is used to find m
1
and m
2
as follows:
1 1
*
9
, 0
9
1
D D
D
D C t t
|q
k
|q
= = = =
(2.A1.18)
Solving the system of Equations (2.A1.18) leads to the following values for m
1
and m
2

47256 . 5 , 55790 . 0
2 1
= = m m (2.A1.19)
Using these values in Equation (2.A1.17) and some simplifications leads to the following
late time pseudo-pressure for the linearly declining fracture pressure:
1
2
3 2
2
1
3 2
1
9
1
), 1 ) )(exp( 086 . 1 086 . 0 (
47256 . 5
) 1 ) )(exp( 314 . 3 314 . 2 (
55790 . 0
) , (
D
D D D D
D D D D D D D
t t x x
t x x t t x
|q
k
k
> +
+ + =

(2.A1.20)
The late time average matrix block pseudo-pressure for the linearly declining fracture
pressure is obtained as follows:
1
2
2
1
1
1
0
9
1
), 1 ) (exp(
14229 . 0
) 1 ) (exp(
81416 . 0
) , (
D
D D
D D D D D D D
t t
t t dx t x
|q
k
k
>
= =
}

(2.A1.21)
2.A2: Exponentially declining fracture pressure
When the fracture pressure changes exponentially with time we have the following PDE
with these initial and boundary conditions:
) ) ( (
D
D
D
D D
D
x
t
x t c
c
c
c
=
c
c
|q
,
(2.A2.1)
0 0 = =
D D
t ,
(2.A2.2a)
, 0 0 =
c
c
=
D
D
D
x
x

(2.A2.2b)
) exp( 1 ) ( 1
D D fD D D
t t x k = = = . (2.A2.2c)
For the early time the following trial solution is suggested which satisfies the outer
boundary condition:

* 3
, )
1
1
1 ))( exp( 1 ( ) , ( t t
x
t t x
D
D
D D D D
<
=
o
k (2.A2.3)
50
When the boundary condition change with time, the penetration depth is found by solving
the following ODE (Mitchel and Myers, 2010):
) (
) (
) 1 (
) (
1
) ( ) (
2
D
D fD
D
D D fD
D
t
t n
n
t
n
t t
dt
d
o
u
u o
o +
=
(
(
+
.
(2.A2.4)
where:
1 2
) 1 (
2

c
c
=
n
n n
t
fD
D
fD
o
u .
(2.A2.5)
By assuming =0 and
D
t c o = and substituting Equation (2.A2.2c) in Equation (2.A2.4)
we reach the following ODE:
D
D D D
t
t n
n
t t
d
)) exp( 1 (
1
)) exp( 1 ( k
c
k c
=
(
(
+

.
(2.A2.6)
Solving this ODE for , substituting it in the penetration depth equation and taking n=3
lead to the following equation for the penetration depth for the linear partial differential
equation:
)
) (
) exp( 1 ) exp( 1
2
( 12 1
D
D
D D
D
t
t erf
t t
t
k
k
k
t
k
o

~ . (2.A2.7)
And for the nonlinear partial differential equation (Equations (2.A2.1) and (2.A2.2)) we
reach the following equation for the penetration depth:
)
) (
) exp( 1 ) exp( 1
2
( 12 1
1
D
D
D D
D D
t
t erf
t t
t
k
k
k
t
k
|q o

~ . (2.A2.8)
Our numerical results show that we can increase the accuracy of the solution if we use the
following equation for the penetration depth:
)
) (
) exp( 1 ) exp( 1
2
( 96 . 12 1
1
D
D
D D
D D
t
t erf
t t
t
k
k
k
t
k
|q o

= . (2.A2.9)
Therefore, the early time solution in Equations (2.A2.1) and (2.A2.2) can be expressed as
follows:
51
* 3
1
, )
) (
) exp( 1 ) exp( 1
2
( 96 . 12
1
1 (
)) exp( 1 ( ) , (
t t
t
t erf
t t
t
x
t t x
D
D
D
D D
D D
D
D D D D
<

=
k
k
k
t
k
|q
k

(2.A2.10)
Integrating of Equation (2.A2.10) over the matrix block volume, leads to the following
equation for the early time average dimensionless pseudo-pressure:
*
1
,
) (
2 ( 96 . 12
4
) exp( 1
t t
t
t erf
t
t
D
D
D
D D
D
D
<

=
k
k t
|q
k
(2.A2.11)
For the late time solution in the case of the exponentially declining fracture pressure the
following PDE should be solved. It should be noted that the initial condition for the late
time solution comes from the early time solution (Equation (2.A2.10)).
) ) ( (
D
D
D
D D
D
x
t
x t c
c
c
c
=
c
c
|q
,
(2.A2.12)
3 * *
)) exp( 1 (
D D D
x t t t k = = ,
(2.A2.13a)
, 0 0 =
c
c
=
D
D
D
x
x

(2.A2.13b)
) exp( 1 ) ( 1
D D fD D D
t t x k = = = . (2.A2.13c)
Using Duhamels theorem (Equation (2.A1.15)) and the solution of the constant fracture
pressure (Equation (2.A1.16)) leads to the following late time solution for the case of the
exponentially declining fracture pressure:
52
,
) exp( ) (
) exp( ) 1 ( ) exp( ) 1 (
) exp( )
2
086 . 1 314 . 3
(
) exp( ) 1 ( 086 . 1 ) exp( ) 1 ( 314 . 3
) exp( )
086 . 0 314 . 2
(
) exp( ) 1 ( 086 . 0 ) exp( ) 1 ( 314 . 2
) exp( 1 ) , (
3
2
2
1
1
2
2
2
2 1
1
1
1
* 2
2
1
1
2
2
2
2 1
1
1
1
2
2
1
1
2
2
2
2 1
1
1
1
D
D
D
D
D D
D D
C
D
C
D D
D
D D
D D D D
x
t
m m
t m t m
t t x
t
m m
t m t m
t
m m
t m t m
t t x
+
+
+
+
+
+
> +
+
+
+
+
+
+
+
+
+
+
+
+
+ =

k
k
k
k
k
k

|
k

k
k
k
k
k
k

k
k
k
k
k
k

k

(2.A2.14)
1
and m
2
as follows:
) exp( 1 , 0
* *
t D C t t
D
k = = = (2.A2.15)
1
and m
2

\
|

+
=

+
=
) exp( ) exp(
) exp( 1
48743 . 1
) exp( ) exp(
) exp( 1
48743 . 0
* *
2
*
2
2
* *
1
*
1
1
t t
t
m
t t
t
m
k
k
k
k
k
k
k
k

(2.A2.16)
Using these values in Equation (2.A2.14) and simplifying lead to the following late time
pseudo-pressure for the exponentially declining fracture pressure:
53
), exp(
) 1 086 . 1 086 . 0 )(
1
( 48743 . 1
) 314 . 3 314 . 2 )(
1
( 48743 . 0
) 086 . 1 086 . 0 )( exp( )
1
( 48743 . 1
) 314 . 3 314 . 2 )( exp( )
1
( 48743 . 0
) exp( 1 ) , (
3 2
3 2
* 3 2
2
3 2
1
* *
2
*
* *
1
*
* *
2
*
* *
1
*
D
D D
t t
t
D D
t t
t
D D D D
t t
t
D D D
t t
t
D D D D
t
x x
e e
e
x x
e e
e
t t x x t
e e
e
x x t
e e
e
t t x
k
k
k
k
k
k
k
k
k
k
(
(
(
(
+
> +
+
+

(2.A2.17)
Integrating over the matrix block volume results in the following equations for the average
dimensionless pseudo-pressure in the case of the exponentially declining fracture pressure:
), exp( ))
1
( 03867 . 0 )
1
( 71132 . 0 (
) exp( )
1
( 03867 . 0
) exp( )
1
( 71132 . 0 ) exp( 1 ) , (
* *
2
*
* *
2
*
* *
2
*
* *
1
*
*
2
1
D
t t
t
t t
t
D D
t t
t
D
t t
t
D D D D
t
e e
e
e e
e
t t t
e e
e
t
e e
e
t t x
k
k
k
k
k
k
k
k
k
k
> +

(2.A2.18)

Chapter Three: One Dimensional Matrix-Fracture Transfer in Dual-Porosity Systems
with Variable Block Size Distribution
1

Abstract
Most of the developed models for fractured reservoirs assume ideal matrix block size
distribution. This assumption may not be valid in reality for naturally fractured reservoirs
and possibly lead to errors in prediction of production from the naturally fractured
reservoirs especially during a transient period or early time production from the matrix
blocks. In this study, we investigate the effect of variable block size distribution (VBSD)
on one dimensional flow of compressible fluids in fractured reservoirs. The effect of
different matrix block size distributions on the single phase matrix-fracture transfer is
studied using a recently developed semi-analytical approach. The proposed model is able
to simulate fluid exchange between matrix and fracture for continuous or discrete block
size distributions using probability density functions or structural information of a
fractured formation. The presented semi-analytical model demonstrates a good accuracy
compared to the numerical results. There have been recent attempts to consider the effect
of variable block size distribution in naturally fractured reservoir modeling for slightly
compressible fluids with a constant viscosity and compressibility. The main objective of
this study is to consider the effect of variable block size distribution on a one dimensional
matrix-fracture transfer function for single phase flow of a compressible fluid in fractured
porous media. In the proposed semi-analytical model the pressure variability of viscosity
and isothermal compressibility is considered by solving the nonlinear partial differential
equation of compressible fluid flow in the fractured media. The closed form solution
provided can be applied to flow of compressible fluids with variable matrix block size
distribution in naturally fractured gas reservoirs.

1
This chapter is an exact copy of: Ranjbar, E. Hassanzadeh, H. Chen, Z. (2012). One Dimensional Matrix-
Fracture Transfer in Dual-Porosity Systems with Variable Block Size Distribution, Transport in Porous
Media, Vol. 95 (1), Page: 185-212.
Chapter 3. One dimensional matrix-fracture transfer in DP systems with VBSD
55
3.1 Introduction
One of the most common approaches for the modeling of naturally fractured reservoirs is
through dual-porosity, which treats the fracture and matrix systems as separate domains. In
this approach matrix (main domain for fluid storage) and fracture (main domain for fluid
flow) are connected with a cross-flow transfer function (Chen, 1995). The transfer term,
which is directly proportional to a shape factor, is one of the key parameters in a pressure
transient analysis of dual-porosity reservoirs. This parameter considers the effect of the
fracture spacing distribution into the pressure transient response equations. Studies have
been done to evaluate the shape factor for flow of slightly compressible fluids in the
fractured media (e.g., Warren and Root, 1963; Kazemi et al., 1976; Lim and Aziz, 1995;
Chang, 1995; Zimmerman et al., 1993, 1996; Civan and Rasmussen, 2002; Penuela et al.,
2002; Hassanzadeh and Pooladi-Darvish, 2006; Lu and Connel, 2007; van Heel et al.,
2008; Hassanzadeh et al., 2009; Rangel-German et al., 2010). In our previous studies we
have developed a model for a compressible fluid shape factor and considered the effect of
fracture pressure depletion regimes on this shape factor (Ranjbar and Hassanzadeh, 2011;
Ranjbar et al., 2011).
One of the main difficulties with dual-porosity models proposed by Warren and Root
(1963) is that all fractures are assumed to be evenly spaced throughout the reservoir.
Measurements of fracture spacing in geological formations have shown that formations
have a broad distribution of matrix block sizes due to lithologic variations, variable rock
stresses, and diagenetic processes (Reis, 1998). Therefore, for accurate prediction of mass
transfer between the matrix and fracture systems the effect of variable block size
distribution should be considered. There are recent studies on the effect of variable block
size distribution in the fractured reservoir modeling for slightly compressible fluids. This
variability in the matrix block size distribution influences the hydrocarbon recovery
especially during early time productions from matrix blocks (Rodriguez et al., 2001).
Therefore, for accurate prediction of production, it is essential to characterize and model
fractured porous media by considering variable matrix block size features (Belani, 1988).
Consequently, dual-porosity models such as Barenbelatt et al., (1960) and Warren and
Roots (1963) are an over-simplification of real fractured formations for fluid flow in
56
fractured porous media and cannot be considered as a comprehensive model for the
fractured reservoirs.
The main objective of this study is to investigate the effect of variable matrix block size
distribution on the matrix-fracture fluid transfer for compressible fluids in fractured media.
The semi-analytical model presented does not require numerical inversion or infinite series
calculations and may find applications for modeling the flow of slightly compressible
fluids in fractured porous media with variable matrix block size distribution.
During the last two decades, the modeling of variable matrix block size distribution in
naturally fractured reservoirs has been the focus of numerous investigations (Cinco-Ley et
al., 1985; Abdassah and Ershaghi, 1986; Belani and Jalali-Yazdi, 1988; Johns and Jalali-
Yazdi, 1991; Jelmert, 1995; Reis, 1998; Gwo et al., 1998). A brief review of some of the
developed models that consider the effect of the variable block size distribution on the
recovery from matrix blocks and the shape factor in the case of a slightly compressible
fluid is discussed in this section.
Zimmerman and Bodvarsson (1995) modeled collections of matrix blocks of different sizes
with a single equivalent block with an equivalent average (volumetrically-weighted)
radius. They showed that the collection of matrix blocks performs like a single matrix
block with an equivalent radius. Their results were validated during early times. However,
in the longtime limit, no effective radius was rigorously defined that can be used as an
equivalent radius.
Rodriguez et al., (2001) presented a variable matrix block size model for flow of slightly
compressible fluids by considering uniform, exponential and normal distribution of matrix
block size distributions. They obtained their solution in the Laplace space involving an
integral, which was numerically evaluated.
There are also recent attempts to consider block size distribution for gas reservoirs with
approaches different from the semi-analytical approach presented in this study (Fathi and
Akkutlu, 2009; Lu and Connel, 2011). In addition discrete fracture representation has been
used for random fracture network with various fracture densities and conductivities
(Bogdanov et al., 2003; Mourzenko et al., 2011).
57
According to the previous studies it can be concluded that the transient pressure response
for the variable matrix block size distribution is different from that of the conventional
sugar cube model of Warren and Root for slightly compressible fluids. It should also be
pointed out that most of the presented models for variable matrix block size distribution in
the literature do not provide a simple prediction for the pressure transient response. Most of
these models provide either a solution in the Laplace space that requires a numerical
inversion into the time domain, or they require the use of infinite series calculations. The
effect of variable block size distribution on flow of compressible fluid, which leads to a
nonlinear partial differential equation, has not been widely investigated in the past. The
main goal of this study is to present a new semi-analytical approach to consider the effect
of variable matrix block size distribution on flow of a compressible fluid in dual-porosity
media. In addition, due to the analytical nature of the proposed model, it can be used to
model flow of slightly compressible fluids in fractured media. The proposed model in this
study is a new semi-analytical approach that can handle the variable matrix block size
distribution for the flow of both compressible (nonlinear partial differential equation) and
slightly compressible fluids (linear partial differential equation) in the fractured media. In
addition to the continuous block size distribution, the proposed model is also capable to
model matrix-fracture transfer when there is discrete block size distribution using structural
information of a porous medium.
In the presented model a combination of a heat integral method, the method of moments
and a concept of equivalent length are used to consider the effect of variable block size
distribution on the matrix-fracture fluid transfer. This new model can be simply
programmed into a spreadsheet for quick use. Results of this study show that the variability
of matrix block size distribution has a significant effect on the transient values of the
matrix-fracture fluid transfer and shape factor for compressible and slightly compressible
fluids.
This paper is organized as follows: First, a short study on the different matrix block size
distributions is presented. In the next step, the definition of equivalent length is used to
derive the equivalent length for different matrix block size distributions. After that, a
mathematical model for variable matrix block size distribution of compressible fluid flow
58
in fractured media is developed. Subsequently, the developed model is applied to slightly
compressible fluids. Afterwards, validation of the proposed model is discussed. Finally, the
results and discussions are presented leading to conclusions.
3.2 Matrix block size distributions
Geological and well logging studies have shown that fractured formations have a wide
distribution of matrix block sizes due to lithologic variations, variable rock stresses, and
diagenetic processes. To improve the modeling of fluid flow in fractured formations one
needs to consider variation in the matrix block size distribution. Although these parameters
are practically impossible to measure directly because of the shortage of three-dimensional
access to representative rocks, they can be estimated from fracture spacing measured on
orthogonal fracture sets from outcrops. Fracture spacing is classically measured using a
scanline along an outcrop or a core, where the successive spacing between fractures is
recorded as they intersect the scanline (Reis, 1998).
In general, variable matrix block size distribution is reported based on two approaches
including discrete block size distribution and continuous block size distribution. In the case
of discrete block size distribution structural information of a fractured rock is reported
based on the observed fracture spacing and the frequency of the matrix block size
distribution for a finite number of block sizes (Gwo et al., 1998). In the continuous block
size distribution the actual geological formation is reported based on a probability density
function. Exponential and linear distributions are two most probable continuous block size
distributions existing in outcrops (Segal, 1981).
For a discrete distribution we have the following equation:
=
=
N
i
ci i
L f
1
1 ) ( (3.1)
where L
c
is the matrix block characteristic length (half of the matrix block thickness for a
slab-shaped matrix block, h
m
), f
i
(L
ci
) is the fraction of the block volume of size L
ci
and N is
the number of block sizes, which is finite. Continuous block size distribution is reported
using a probability density function (PDF), which is a positive function. The area under the
curve of a PDF as illustrated in Equation (3.2) is equal to one.
59
}
=
max
min
1 ) (
c
c
L
L
c c
dL L f (3.2)
In Equation (3.2), f(L
c
) is the probability density function with the random variable L
c
and
describes the distribution of the block size. Since the distribution functions are usually
expressed in dimensionless form we use the following dimensionless variables to express
the PDFs in dimensionless form:
max c
c
D
L
L
L =
(3.3)
) ( ) (
max c c D D
L f L L f = (3.4)
max
min
c
c
h
L
L
F =
(3.5)
In the following, a brief description of different probability density functions is presented.
3.2.1 Uniform or rectangular distribution
In this case the following probability density function is used to describe the matrix block
size distribution:
min max
1
) (
c c
c
L L
L f
=
(3.6)
When enough data about matrix block size distribution does not exist and the matrix block
size distribution is unknown this distribution may be used. Equation (3.6) in the
dimensionless form can be expressed as follows:
h
D D
F
L f
=
1
1
) (
(3.7)
3.2.2 Exponential distribution
The corresponding equation for an exponential probability density function is expressed as
follows:
max min
) (
c c
c
L L
L
c
e e
e
L f
o o
o
o

= (3.8)
where is an exponential distribution constant. Using Equations (3.3) to (3.5) in Equation
(3.8) leads to the following dimensionless exponential probability distribution function:
60
a F a
L a
D D
e e
ae
L f
h
D

= ) ( (3.9)
In Equation (3.9), F
h
defines the range of block sizes for exponential functions and a is a
dimensionless exponential distribution constant.
The literature values for the dimensionless exponential constants are -20, -5, 5, and 20 (Rodriguez
et al., 2001). The positive values of a (e.g., 20 or 5) represent distributions with a high fraction of
small blocks or high density of fractures. For sparsely fractured distribution or a high fraction of
large blocks the values of the exponential constant are small (a= -20 or -5).
3.2.3 Normal or Gaussian distribution
The following equation shows the normal or Gaussian function:
) ) (
2
1
exp(
2
1
) (
2
2
2
M L L f
D D D

=
o
to

(3.10)
In Equation (3.10), M is the mean and
2
is the variance of the distribution. Using the
values for mean and variance in the Gauss function leads to the following normal matrix
block size distribution (Rodriguez et al., 2001):
)
`
=
2
2
)]
2
1
( [
) 1 (
18
exp
) 1 ( 2
6
) (
h
D
h h
D D
F
L
F F
L f
t
(3.11)
3.2.4 Linear distribution
For the linear block size distribution we have the following dimensionless equation to
describe matrix block size distribution:
b mL L f
D D D
+ = ) ( (3.12)
In Equation (3.12), m>0 means a low fracture density or large matrix block sizes. For a
high proportion of small matrix blocks m is less than zero. In general, two cases for the
linear PDF is considered in the literature as linear decreasing distribution (m=-100/81;
b=145/81) and linear increasing distribution (m=100/81; and b=35/81) (Rodriguez et al.,
2001). It should be noted that linear distribution changes to the uniform distribution if we
use m=0 and b=10/9, when F
h
=0.1 (Rodriguez et al., 2001).
3.2.5 Log-normal distribution
A log-normal distribution is a probability distribution which is defined based on the
following equation,
61
)
`

=
2
ln
2
ln
2
ln
2
] ) [ln(
exp
2
1
) (
o
to
M L
L
L f
D
D
D D

(3.13)
Different values for variance and average matrix block size length has been used in the
literature for log-normal distribution (Leung and Zimmerman, 2012; Reis, 1998). In this
study we used the value of 0.5 for standard deviation of spacing logarithm and -1.0 for the
mean of spacing logarithm. Figure 3.1 compares the different distribution functions for
specific values of standard deviation, mean, m, b, and a when F
h
=0.1.

Figure 3.1: Different probability density functions.

L
D
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
f
D
(
L
D
)
0
1
2
3
4
5
6
Rectangular
Linear Increasing (m=100/81)
Linear Decreasing (m=-100/81)
Exponental (a=-20)
Exponential (a=-5)
Exponential (a=5)
Exponential (a=20)
Normal (o=0.15, M=0.55)
Log-Normal (o=0.5, M=1)
62
3.3 Equivalent lengths for different matrix block size distributions
3.3.1 Equivalent length concept
It has been shown that the collection of matrix blocks with different distributions performs
like a single matrix block with an equivalent length (Zimmerman and Bodvarsson, 1994).
For the case of discrete matrix block size distribution the equivalent half matrix block size
is defined as follows (Gwo et al., 1998):
ci
N
i
ci i
N
i
ci
t
i
t
N
i
ci i
N
i
i
N
i
ci i
ce
L L f L
N
N
N
L N
N
L N
L

= =
=
=
=
= = = =
1 1
1
1
1
) ( (3.14)
where N
t
is the total number of matrix blocks and f
i
(L
ci
) is the probability of the matrix
block size of L
ci
. Equation (3.14) (using Equation (3.3)) can be expressed in the
dimensionless form as follows:
Di
N
i
ci i De
L L f L

=
=
1
) ( (3.15)
For the case of continuous block size distribution the following equation is used to find the
half matrix block equivalent length:
}
=
max
min
) (
c
c
L
L
c c c ce
dL L f L L
(3.16)
Using Equations (3.3) to (3.5) in Equation (3.16) leads to the following equation for the
dimensionless equivalent length in the case of continuous matrix block size distribution:
}
=
1
) (
h
F
D D D D De
dL L f L L
(3.17)
3.3.2 Discrete matrix block size distribution
Equation (3.15) is used to calculate the equivalent length for the case of discrete matrix
block size distribution when the observed frequency of matrix block sizes is available for a
finite number of matrix blocks. Table 3.1 shows the observed fracture spacing and the
frequency of matrix block size distribution reproduced by Gwo et al., (1998) for fractured
Saprolite.
63
To calculate the equivalent length using the structure of porous media, the dimensionless
length (L
D
) and probability of matrix blocks (f(L
ci
)) are calculated as illustrated in Table
3.1. Using Equation (3.15) the calculated value for the equivalent length of the data
provided in Tables 3.1 is 0.489. This method can be used to find the equivalent length for
any observed frequency of matrix block sizes.

Table 3.1: Observed frequency of matrix block size in the soil column (Gwo et al., 1998)
Block Size (h
m
),
meter
Block size characteristic
length (L
c
), meter
Average frequency (top
and bottom of soil (N
i
)
L
D
=L
c
/L
cmax
f(L
ci
)=N
i
/N
t

0.003 0.00150 1.5 0.167 0.088
0.005 0.00250 4.5 0.278 0.265
0.0075 0.00375 3.5 0.417 0.206
0.01 0.00500 3.5 0.556 0.206
0.0125 0.00625 2.0 0.694 0.118
0.018 0.00900 2.0 1.000 0.118

3.3.3 Continuous matrix block size distribution
Substituting the probability density functions for rectangular (Equation (3.7)), exponential
(Equation (3.9)), Gaussian or normal (Equations (3.10) or (3.11)), linear (Equation (3.12))
and log-normal (Equation (3.13)) distributions in the equivalent length equation (Equation
(3.17)) leads to the following equations for the equivalent length in the case of different
probability density functions.
Rectangular distribution :
2
1
h
De
F
L
+
= (3.18)
Exponential distribution :
) (
) 1 ( ) 1 (
a aF
a aF
h
De
e e a
e a e aF
L
h
h

+ +
= (3.19)
Normal distribution :
(
(
(
(
+
(
2
2
2
2
2
) 1 (
2
) (
2
2 2
2
)
2
1
( )
2
(
2
1
o o
t
o
o o
M F M
h
De
e e
M
erf
F M
erf M L
h
(3.20)
Linear distribution : ) 1 (
2
) 1 (
3
2 3
h h De
F
b
F
m
L + = (3.21)
64
Log-normal distribution:
(
(
+
=
+
]
2
[ ]
2
) ln(
[
2
2
ln
2
ln ln
2
ln
2
ln
2
2
ln
ln
o
o
o
o
o
M
erf
F M
erf
e
L
h Ln
M
De

(3.22)
These derived equivalent lengths for discrete (Equation (3.15)) and continuous (Equations
(3.18)-(3.22)) block size distributions should be replaced in diffusivity equation and shape
factor equation to calculate the average matrix block pressure or matrix-fracture transfer
shape factor for different matrix block size distributions in the case of compressible fluid
flow in fractured media. This will be discussed in the following sections.
3.4 Mathematical model for flow of compressible fluid in fractured media with
variable block size distribution
In this section the mathematical model for the matrix-fracture transfer function for a
compressible fluid is derived by considering the variable matrix block size distribution. In
this derivation the dependence of the compressible fluid hydraulic diffusivity on pressure is
considered by solving a nonlinear flow equation in porous media. The diffusivity equation
for a slab shape matrix can be expressed as:
2
2
) (
x
p
t
m
m
m
c
c
=
c
c
q
(3.23)
In Equation (3.23),
m
is matrix hydraulic diffusivity and is the ratio of matrix permeability
to product of matrix porosity, fluid viscosity and matrix (fluid and rock) compressibility.
It should be pointed out that
m
is strongly pressure-dependent; since pressure is space- and
time-dependent; the matrix hydraulic diffusivity also varies with time and space. For
solving this PDE, we suppose that the hydraulic diffusivity is time-dependent and the
effect of space dependence is accounted for by introducing a correction factor, . This
correction factor is then obtained using fine-grid numerical simulations (Ranjbar and
Hassanzadeh, 2011). Therefore, in the case of variable matrix block size distribution the
following PDE with the initial and boundary conditions should be solved.
) ) ( (
x
t
x t
m
m
m
c
c
c
c
=
c
c
|q
(3.24)
i m
t = = 0 (3.25a)
65
0 0 =
c
c
=
x
x
m
(3.25b)
f m ce
L x = =
(3.25c)
It should be noted that in Equation (3.25c), L
ce
is equivalent length which is not a constant
and is a function of block size distribution (Equations (3.14) and (3.16)). Figure 3.2 (left)
shows a matrix-fracture system and its boundary conditions. A schematic of the model for
variable block size distribution is illustrated in Figure 3.2 (right). This figure demonstrates
a discrete fracture model with non-uniform block size distribution and boundary conditions
for the blocks.

Figure 3.2: Illustration of a matrix-fracture system and its boundary conditions (left) and
representation of fractured reservoirs in the case of non-ideal matrix block size distribution (right).

For solving Equation (3.24) with the boundary conditions in Equation (3.25), we define the
average hydraulic diffusivity over the matrix-fracture pressure drawdown as follows:
} }

=
=
f
i
f
i
p
p m i f m
m
p
p m m
m
i f
c
dp
p p
k
dp
c
k
p p | |
q
1 1
(3.26)
It should be pointed out that in this equation k
m
, c
m
and
m
| are matrix permeability,
compressibility and porosity respectively, is fluid viscosity and p
f
and p
i
are fracture and
initial pressure, respectively. Equations (3.24) and (3.25) are made dimensionless by
defining the following dimensionless variables:
0 =
c
c
x
m

f m
=
66
i f
i m
D

=
(3.27)
max c
D
L
x
x =
(3.28)
q
q
q
) (
) (
t
t
m
D
=
(3.29)
2
max c
D
L
t
t
q
=
(3.30)
max c
ce
De
L
L
L =
(3.31)
Using Equations (3.27) to (3.31) in Equations (3.24) and (3.25), we obtain the following
dimensionless PDE with the initial and boundary conditions (Equations (3.33a), (3.33b),
(3.33c)):
) ) ( (
D
D
D
D D
D
x
t
x t c
c
c
c
=
c
c
|q

(3.32)
0 0 = =
D D
t (3.33a)
0 0 =
c
c
=
D
D
D
x
x

(3.33b)
1 = =
D De D
L x (3.33c)
In Equation (3.33c), L
De
is used to consider the effect of variable block size distribution;
this equivalent length is calculated based on Equations (3.15) or (3.17). A heat integral
method (Goodman, 1964; Finlayson, 1972; Pooladi-Darvish et al., 1994) and the method of
moments (Ames, 1965; Crank, 1975) are used to find the early and late time solutions of
this nonlinear PDE, respectively. The following equations give the solution of the PDE:
1
2
3
1
24
, )
24
1 (
D
De
D
D D
D De
D
L
t
t
x L
|q |q
<
=
(3.34)
67
, )) exp(
686 . 5
) exp(
541 . 0
(
24
)) exp(
175 . 6
) exp(
793 . 1
(
)) exp( 489 . 0 ) exp( 252 . 1 1 ( ) , (
3
2 3 1 3
1
2
2
2 2 1 2
2 1
D D
D
D
D
D
De
D D D
D
D
D
D D D D D
x t
L
t
L
L
t x t
L
t
L
t t t x

|q

+
> +
+ + =

(3.35)
More details on the derivation of these two equations are shown in Appendix 3.A.
Integrating over the bulk volume of a matrix block leads to the following equation for the
early and late time average dimensionless pseudo pressures:
1
2
1
24
3
1
24 4
24
)
24
1 (
1
1
D
De
D
De
D D
L
t L
D
D D
D De
De
D
L
t
L
t
dx
t
x L
L
De
D D De
|q
|q
|q
|q
< =
=
}

(3.36)
1
2
2 1
0
24
) exp( 148 . 0 ) exp( 790 . 0 1
1
D
De
D D D
L
D D
De
D
L
t t t dx
L
De
|q
> = =
}
(3.37)
In these equations the fracture dimensionless hydraulic diffusivity,
D1
,
1
and
2
are
defined as follows, respectively:
q
|
q
q
q
m f f m
D m
D
c k
x
/
) 1 (
1
=
=
= (3.38)
2
1
2 2
1
1
181 . 32 486 . 2
De
D
De
D
L L
|q
|q
=
(3.39)
In our previous studies the following equation was derived for the dual-porosity shape
factor of compressible fluid in fractured media (Ranjbar and Hassanzadeh, 2011; Ranjbar
et al., 2011):
( ) t k
c
m
f m m
m m
c
c
=

|
o
(3.40)
Using the definition of dimensionless variables (Equations (3.27) to (3.31)) in this equation
leads to the following dimensionless shape factor for variable block size distribution:
D
D
D D
c
t
L
c
c
|
|
.
|
\
|
=

q
o
1
1 1
2
max
(3.41)
In this equation,
D
is not a constant and is a function of pressure, specific gravity and
temperature since the viscosity-compressibility product is dependent on these parameters.
68
Numerical simulation is used to find the dimensionless hydraulic diffusivity (Ranjbar and
Hassanzadeh, 2011):
Substituting Equation (3.31) into Equation (3.41) leads to the following dimensionless
shape factor for a case with variable matrix block size distribution:
D
D
D D
De
ce
t
L
L
c
c
|
|
.
|
\
|
=

q
o
1
1
2
2
or
D
D
D D
De
me
t
L
h
c
c
|
|
.
|
\
|
=

q
o
1
1 4
2
2
(3.42)
In Equation (3.42), h
me
is the equivalent matrix block length (thickness). Substituting
Equations (3.36) and (3.37) and their derivatives in Equation (3.42) leads to the following
equation for the early and late time dimensionless shape factors in the case of variable
matrix block size distribution, respectively:
1
2
1
1 2
24
,
1
) 1
4
24
(
6
D
De
D
D
De
D D
D
D
De
me
L
t
t
L
t
L
h
|q |q
|q
q
o <
= (3.43)
1
2
2
1
2
1
2
1
2
1
1 2
24
,
)
181 . 32
exp( 148 . 0 )
486 . 2
exp( 790 . 0
)
181 . 32
exp( 763 . 4 )
486 . 2
exp( 964 . 1
4
D
De
D
De
D D
De
D D
De
D D
De
D D
D
D
me
L
t
L
t
L
t
L
t
L
t
h
|q
|q |q
|q |q
q
|q
o >
=

(3.44)
It should be noted that for different matrix block size distribution L
De
is calculated using
Equations (3.16) and Equations (3.18) to (3.22) and substituted in Equations (3.36), (3.37),
(3.43) and (3.44) to calculate the average pseudo-pressure and shape factor for each
distribution.
3.5 Slightly compressible fluids
The above presented model can be also used to consider the matrix block size distribution
for flow of a slightly compressible fluid in fractured porous media.
It should be pointed out that in the case of a slightly compressible fluid flow in fractured
media due to the constant viscosity and compressibility of such a fluid, the hydraulic
diffusivity is constant and the matrix flow equation is a linear partial differential equation
as follows:
69
2
2
x
p
t
p
m
mi
m
c
c
=
c
c
q (3.45)
i m
p p t = = 0 , 0 0 =
c
c
=
x
p
x
m
,
f m ce
p p L x = = (3.46)
In this case, 1
1
= = =
D D
q q | (Ranjbar and Hassanzadeh, 2011). Using these values in the
compressible fluid model, the solution for slightly compressible fluid flow in fractured
media with variable block size distributions can be obtained. Using the same procedure as
above, the following equations for the average dimensionless matrix block pressure and
matrix-fracture transfer shape factor are obtained in a closed form without requiring a
numerical inversion into the time domain or the use of an infinite series evaluation:
24 4
24
2
De
D
De
D
D
L
t
L
t
p < = (3.47)
24
)
181 . 32
exp( 148 . 0 )
486 . 2
exp( 790 . 0 1
2
2 2
De
D
De
D
De
D
D
L
t
L
t
L
t
p >
= (3.48)
24
) 1
4
24
(
6
2
2 De
D
De
D D
De
me
L
t
L
t t
L
h <
= o (3.49)
24
)
181 . 32
exp( 148 . 0 )
486 . 2
exp( 790 . 0
)
181 . 32
exp( 763 . 4 )
486 . 2
exp( 964 . 1
4
2
2 2
2 2
2 De
D
De
D
De
D
De
D
De
D
me
L
t
L
t
L
t
L
t
L
t
h >
= o (3.50)
In these equations the equivalent length derived in Section 3.3 for different discrete and
continuous matrix block size distribution is used to derive the average dimensionless
pressure and dimensionless shape factor for different matrix block size distributions for
slightly compressible fluids.
3.6 Validation
In order to validate the presented model by the equivalent length concept, the solution for
matrix blocks with ideal block size is derived, and it is shown that the presented model can
reproduce the literature model for the ideal block size distribution. At the end of this
70
section the developed semi-analytical model for two different discrete block size
distributions in the case of compressible fluid is compared with the fine grid numerical
results, which show an acceptable accuracy.
In the case of the ideal cubic sugar of the Warren and Root model, the distribution function
can be expressed using a Dirac delta function which is described as follows:
=
=
= A =
1
1 0
) 1 ( ) (
D
D
D D D
L
L
L L f (3.51)
The Dirac delta function has the following property:
1 ) 1 ( ) ( ) ( ) (
1
= A = = A
} }

h
F
D D D De D D D D D
dL L L L a f dL a L L f
(3.52)
Substituting the equivalent length being equal to unity in the shape factor equations
(Equations (3.43) and (3.44)) can recover the proposed model for the shape factor of
compressible fluid flow in fractured porous media (Ranjbar and Hassanzadeh, 2011).
Figure 3.3 compares the slightly compressible fluid shape factor based on the shape factor
derived in this study with equivalent length of one (Equations (3.49) and (3.50)) and the
previous models in the literature (Chang, 1995; Hassanzadeh and Pooladi-Darvish, 2006).
As illustrated in this figure there is an accurate match between the presented model and the
previous models. It should be pointed out that the stabilized value of the slightly
compressible fluid shape factor based on the presented model is 9.9443, which is very
close to t
2
as was reported by Lim and Aziz (1995).
71

Figure 3.3: Comparison of the presented model with literature models (Chang 1995, Hassanzadeh and
Pooladi-Darvish 2006) for slightly compressible fluid and ideal block size distributions.

In this part we present the comparison of our model with a fine-grid, single porosity
numerical model (Eclipse 100) for two different discrete block size distributions. For
reaching this goal we determine the cumulative fluid exchange as a function of real time
based on the developed model (Equation (3.B.15) in Appendix 3.B) and compare this with
the numerical results. We have considered the gas specific gravity of 0.7, temperature of
93.3C and pressure draw-down of 45 to 22.5 MPa for these two matrix block size
distributions. The values used for and
D
are 0.73 and 0.3327, respectively. It should be
pointed out that in our previous studies (Ranjbar and Hassanzadeh, 2011 and Ranjbar et al.,
2011) for the compressible fluid flow we did not consider the early time cumulative
production, since it was a short time period (about 17 seconds). But in this study we
consider the early time cumulative production, which results in a change in
D
from 0.3127
to 0.3327. All the other properties are the same as the base case in our previous study
(Ranjbar and Hassanzadeh, 2011).
Now we will consider two discrete models and compare our results with the numerical
results. Following table shows the discrete block size distribution for these two cases. The
initial pressure is 45 MPa and fracture pressure is 22.5 MPa. All other properties have the
same value as was reported by Ranjbar and Hassanzadeh (2011).
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

S
h
a
p
e

F
a
c
t
o
r
1
10
100
1000
Presented Model
Previous Models
72

Table 3.2: Block size distribution for the discrete distributions.
Block Size (h
m
), meter
(First case)
N
i
f(h
mi
)
Block Size (h
m
), meter
(Second case)
N
i
f(h
mi
)
4 1 1/3 4 1 0.2
2 1 1/3 2 2 0.4
3 1 1/3 1.5 1 0.2
3 1 0.2

The arithmetic equivalent length of the matrix blocks and reservoir length for case one and
two are calculated based on the following equations respectively:
m L m L m
h
L m h h f h
c R
me
ce mi
N
i
mi i me
2 , 9 3 2 4 , 5 . 1
2
, 3 ) (
max
1
= = + + = = = = =
=
(3.53)
m L m L m
h
L m h h f h
c R
me
ce mi
N
i
mi i me
2 , 5 . 12 3 5 . 1 ) 2 2 ( 4 , 25 . 1
2
, 5 . 2 ) (
max
1
= = + + + = = = = =
=
(3.54)
Figure 3.4 demonstrates the comparison of the presented semi-analytical model (Based on
Equation (3.B.15)) with the numerical results for these two discrete distributions. As
illustrated in this figure the presented semi-analytical model can predict the behaviour of
the compressible fluid flow for different block size distributions with an acceptable
accuracy.
73

Figure 3.4: Comparison of the presented semi-analytical model with the numerical results for the first
and second case. Continuous lines are predictions by analytical model. Dots and dashes are from
numerical simulations.

3.7 Results
In the following two sections we focus on the dimensionless rate, dimensionless
cumulative production and dimensionless shape factor for different matrix block size
distributions in the case of single-phase compressible fluid flow in fractured porous media.
The dimensionless shape factor is calculated based on Equations (3.43) and (3.44). In these
sections a base case with the gas specific gravity of 0.7, temperature of 93.3C and
pressure draw-down of 45 to 22.5 MPa for different matrix block size distribution is
considered. The values used for different parameters are reported elsewhere (Ranjbar and
Hassanzadeh, 2011) except for
D
that has changed from 0.3127 to 0.3327 due to
consideration of early time cumulative production. The dimensionless rate and
dimensionless cumulative production for different matrix block distributions versus
Time (Sec.)
0 200 400 600 800 1000 1200 1400 1600
C
u
m
u
l
a
t
i
v
e

F
l
u
i
d

E
x
c
h
a
n
g
e

(
S
m
3
)
0
500
1000
1500
2000
2500
3000
3500
4000
4500
Case2
Case1
74
dimensionless time for compressible fluid flow can be calculated using the following
equations:
| |

> +
<
=
1
2
2 1 2
1
1
2
1
24
, ) exp( 763 . 4 ) exp( 964 . 1
24
,
8
24
D
De
D D D
De D
D
D
De
D
D De D
D
D
L
t t t
L
L
t
t L
q
|q

q
k|q
|q q
|q k
(3.55)
| |

>
<
=
1
2
2 1
1
2
1
24
, ) exp( 148 . 0 ) exp( 790 . 0 1
24
,
4
24
D
De
D D D
D
D
De
D D
De D
D
D
L
t t t
L
t t
L
Q
|q

q
k
|q q
|q k

(3.56)
In these two equations is the ratio of L
R
to L
cmax
for the distributions. In the results and
discussion section we have used the value of 16.6 for , which is the case for the discrete
fracture model (Table 3.1). Derivation of Equations (3.55) and (3.56) is discussed in
Appendix 3.B in more details. In the case of a slightly compressible fluid the production
rate and cumulative production can be calculated by putting 1
1
= = =
D D
q q | in these two
equations. It is worth noting that L
cmax
is used as the matrix block characteristic length to
scale the time, the production rate and the cumulative production, which is independent of
the matrix block size distribution. Therefore, it is possible to compare the dimensionless
rate and the cumulative production for different matrix block size distributions as presented
in the following subsections.
3.7.1 Rectangular, discrete, normal and log-normal distributions
Figure 3.5 compares the dimensionless rate versus dimensionless time on the log-log plot
for the discrete fracture, rectangular, normal and log-normal distributions and an ideal
block size distribution with the equivalent length of 0.489, 0.550, 0.548 and 0.414 and
1.000, respectively. According to this figure at the early time of the production the smallest
equivalent length distribution (log-normal) produces with the largest dimensionless
production rate and the reservoir is depleted more quickly than the other distributions. It
should be pointed out that at the early time of the production the dimensionless rate is
inversely proportional to the square root of the dimensionless time, which is characteristic
of early time pressure diffusion. According to this figure based on the equivalent lengths of
75
the distributions the early time matrix-fracture fluid transfer is lowest for the highest
equivalent length. Results show that the early time fluid transfer increases as the equivalent
length decreases and the smallest early time production rate belongs to the ideal
distribution (large equivalent length L
De
=1). The matrix blocks are depleted more quickly
for distribution with smaller equivalent lengths.

Figure 3.5: Dimensionless rate versus dimensionless time for ideal, rectangular, discrete, normal and
log-normal distributions when F
h
=0.1.

3.7.2 Linear distribution
Two types of linear distributions including increasing linear distribution and decreasing
linear distribution are defined based on the different values of m and b as was discussed in
Section 3.2.4. The obtained values for the equivalent length for linearly increasing and
linearly decreasing are 0.625 and 0.475, respectively. It should be noted that the equivalent
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

R
a
t
e
0.01
0.1
1
10
100
1000
10000
Ideal
Rectangular
Discrete
Normal (o=0.15, M=0.55)
Log-Normal (o=0.50, M=1)
D
D
t
q
1

76
length for rectangular distribution is the average of the equivalent length for linearly
increasing and linearly decreasing distributions. Figure 3.6 compares the dimensionless
rates for different values of the dimensionless time for the linear probability density
function.
Figure 3.6 shows that at the specified values of the early dimensionless time the linearly
increasing distribution (larger block sizes) produces with a smaller production rate than the
linearly decreasing (smaller block sizes) distribution. However, as time increases the
production rate for linearly increasing becomes higher than the linearly decreasing
distribution and the depletion time for the linearly increasing distribution is larger than that
for the linearly decreasing distribution. In general, the linear distribution shows a larger
early production rate than the ideal matrix block size distribution. In addition, matrix
blocks with linear distribution are depleted more quickly than the matrix blocks with ideal
distribution.

Figure 3.6: Comparison of dimensionless rate for linearly increasing (m=100/81, b=35/81), linearly
decreasing (m=-100/81, b=145/81) distribution with F
h
=0.1 and ideal matrix block size distributions.
Dimensionless time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

R
a
t
e
0.01
0.1
1
10
100
1000
10000
Ideal Distribution
Linear Increasing
Linear Decreasing
77
3.7.3 Exponential distribution
As was mentioned in the previous sections four different common values for exponent, a,
in the literature were reported: 20, 5, -5 and -20. The obtained values for equivalent length
for these values of a, are 0.150, 0.290, 0.810 and 0.950, respectively. Figure 3.7 compares
the dimensionless rate versus dimensionless time for the ideal block size distribution and
the exponential distribution with different values of a. These results show that when the
probability of small blocks is higher (a>0), the early time production rate is larger than that
of the exponential distribution with larger matrix blocks sizes (a<0). As the probability of
large block sizes increases (i.e., a=-20) the production rate decreases and tends toward that
of the ideal block size distribution. In general, we can state that as the probability of larger
matrix block sizes increases the early time production rate decreases and it approaches that
of the ideal block size distribution. As a consequence we can conclude that for any
distributions as the equivalent length decreases (i.e. the proportion of small blocks
increases), the early-time production rate increases and the blocks are depleted more
quickly.

Figure 3.7: Comparison of dimensionless rate for exponential (exponent values of -20, -5, 5 and 20)
with F
h
=0.1 and ideal block size distributions.
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

R
a
t
e
0.01
0.1
1
10
100
1000
10000
Ideal Distribution
Exponential Distribution(a=5)
Exponential Distribution(a=-5)
Exponential Distribution(a=20)
Exponential Distribution(a=-20)
78
3.8 Discussion
In general, it can be concluded that during early time of production the dimensionless rate
for different matrix block size distributions is inversely proportional to the square root of
the dimensionless time. All of the distributions demonstrate a larger production rate in
comparison with the ideal block size distribution at the early time. As the probability of
larger matrix block sizes increases there is a decrease in the dimensionless production rate
and the blocks are depleted more gradually for larger matrix block sizes. It should be noted
that the presented model is useful for diffusion in a collection of slab (one dimensional)
shaped matrix blocks and there might be differences in the presented solution when we are
dealing with two dimensional (cylindrical) or three dimensional (spherical) fluid flow in
the case of variable block size distribution.
Figure 3.8 compares the dimensionless cumulative production versus dimensionless time
for different matrix block size distributions based on Equation (3.56). The results show that
the early time cumulative production and the time of stabilization of cumulative production
is a function of matrix block size distribution.
According to Figure 3.8, as the probability of the smaller blocks increases the blocks are
depleted more quickly. These results show that the specified discrete matrix block size
distribution has a cumulative production very close to that of the linearly decreasing matrix
block size distribution.
For the data used in this study, the rectangular and Gaussian distributions have the same
cumulative production. It should be pointed out that the changes in cumulative production
are related to the area under the dimensionless rate curve (Figures 3.5-3.7).

79

Figure 3.8: Dimensionless cumulative production versus dimensionless time for different matrix block
size distributions with F
h
=0.1. Blocks showing the dimensionless equivalent size of the matrix block
L
c
/L
cmax
.

The variations in transient dimensionless cumulative production among various block size
distributions can be explained by the values of the dimensionless equivalent length. Table
3.3 demonstrates the values of the dimensionless equivalent length for different matrix
block size distributions. According to this table the dimensionless equivalent length for the
ideal block size distribution has the maximum value and this means that for this
distribution the transient time is larger than that for the other distributions. For example, in
the case of exponential distribution with a large positive exponent (small matrix blocks or
highly fractured formations) the dimensionless equivalent length has the lowest value and
the cumulative production reaches its plateau more quickly. The time for the cumulative
production to reach its plateau seems proportional to the equivalent length.
Dimensionless Time
0 1 2 3 4 5 6 7
D
i
m
e
n
s
i
o
n
l
e
s
s

C
u
m
u
l
a
t
i
v
e

P
r
o
d
u
c
t
i
o
n
0
10
20
30
40
50
Ideal
Exponential (a=-20)
Linear Increasing (m=100/81, b=35/81)
Rectangular
Normal (o =0.15, M=0.55)
Discrete
Linear Decresing (m=-100/81, b=145/81)
Exponential (a=20)
Log-Normal (o =0.5, M=-1)
80
Figure 3.9 shows the dimensionless shape factor versus dimensionless time for different
matrix block size distribution for flow of compressible fluids in fractured media. The
results show that the stabilized value of the shape factor is not a function of matrix block
size distribution and different matrix block size distributions have the same values of 8.05
for the stabilized value of the shape factor. It should be mentioned that in our previous
study (Ranjbar and Hassanzadeh, 2011) the reported value for the stabilized shape factor
was 8.59 without considering of the early time cumulative production to match with the
numerical results. In this study we have improved our previous study and considered the
early time cumulative production and this cause a change in the matching parameter from
0.3127 to 0.3327 and also the stabilized value of the shape factor from 8.59 to 8.05. These
results show that the transient values and the time of stabilization of the shape factors are a
strong function of the matrix block size distribution. An exponential matrix block size
distribution with a large positive exponent (small matrix blocks or highly fractured
formations) has the smallest values for the dimensionless transient shape factor and
dimensionless time of stabilization. The dimensionless shape factor for the ideal block size
distribution is larger than that of other distributions and stabilizes at a larger dimensionless
time. An exponential distribution with a large negative exponent (sparsely fractured
formation) has a larger transient dimensionless shape factor and dimensionless time of
stabilization than the linear, normal, log-normal, rectangular and discrete distributions. The
transient values of the dimensionless shape factor and stabilized dimensionless time for
linear, normal, log-normal, rectangular and discrete distributions are close to each other.
Based on Figure 3.9 the behaviour of the dimensionless shape factor for normal and
rectangular distributions is similar to each other.

81

Figure 3.9: Dimensionless shape factor for different matrix block size distributions. Blocks showing the
dimensionless equivalent size of the matrix block L
c
/L
cmax
.

Table 3.3: Values of dimensionless equivalent length for different matrix block size distributions.
Block Size Distribution Dimensionless Equivalent Length (L
De
)
Ideal distribution 1.000
Exponential distribution (a=-20) 0.950
Linear increasing distribution 0.625
Rectangular distribution 0.550
Normal distribution 0.548
Discrete distribution 0.489
Linear decreasing distribution 0.475
Log-normal distribution 0.414
Exponential distribution (a=20) 0.150
Dimensionless Time
0.001 0.01 0.1 1 10 100
D
i
m
e
n
s
i
o
n
l
e
s
s

S
h
a
p
e

F
a
c
t
o
r
1
10
100
1000
Ideal Distribution
Rectangular Distribution
Linear Increasing Distribution
Linear decreasing distribution
Exponential Distribution (a=20)
Exponential Distribution (a=-20)
Discrete Distribution
Normal Distribution (o=0.15, M=0.55)
Exponential Distribution (a=5)
Exponential Distribution (a=-5)
Log-Normal Distribution (o=0.5, M=-1)
82
3.9 Conclusions
The following major conclusions are made as a result of this study:
- The results of this study show that variable block size distribution has a significant
effect on the matrix production profile during transient state.
- The results show that as the probability of larger blocks increases or the fracture
density decreases the production profile becomes closer to that for the ideal block
size distribution (Warren and Roots model).
- Among different distributions the ideal block size distribution and the exponential
distribution with very small exponent (sparsely fractured system) have the smallest
transient cumulative production and their cumulative production reaches its plateau
more gradually.
- The large exponent exponential distribution (intensively fractured system) has the
highest transient matrix-fracture fluid transfer and deplete more quickly than other
distributions.
- The transient values of the dimensionless shape factor and the dimensionless time
of stabilization are a function of matrix block size distribution.

83
Nomenclature
a Dimensionless exponential distribution constant
2
]
A(t
D
) First coefficient of the trial solution
b Intercept for linear matrix block size distribution
B(t
D
) Second coefficient of the trial solution
C(t
D
) Third coefficient of the trial solution
c
m
Matrix compressibility [LT
2
/M]
D(t
D
) Fourth coefficient of the trial solution
f
i
(L
ci
) Fraction of the block volume of size L
ci

f(L
c
) Probability density function
f
D
(L
D
) Dimensionless probability density function
F
h
Ratio of minimum block size to the maximum block size characteristic length
h
m
=2L
c
Matrix block length [L]
k
m
2
]
L
c
Matrix block characteristic length [L]
L
D
Dimensionless length
L
De
Dimensionless equivalent length
L
R
Reservoir Length
M Mean of the distribution
m Slope of linear matrix block size distribution
N Number of matrix block sizes
N
t
Total number of matrix blocks
m
p
Matrix block pressure [M/LT
2
]
D
p
Average dimensionless matrix block pressure
f
p
Fracture pressure [M/LT
2
]
i
p
Initial pressure [M/LT
2
]
D
q
Dimensionless matrix-fracture fluid transfer
sc
q
Matrix-fracture fluid transfer [L
3
/T]
D
Q
Dimensionless cumulative fluid exchange
Q Cumulative fluid exchange [L
3
]
R Residual in the method of moments
t Time [T]
t
D
Dimensionless time
V
b
Matrix block volume [L
3
]
x
D
Dimensionless distance
84
Greek Letters
Exponential distribution constant [1/L]
| Correction factor
o
Unaffected thickness of matrix block during the early time
m
q
2
/T]
q Average hydraulic diffusivity [L
2
/T]
D
q
Dimensionless hydraulic diffusivity
1 D
q
Ratio of L
R
to L
cmax

Dimensionless exponent of solution of gas diffusivity equation using the moment method

o Shape factor [1/L
2
]
2
o
Variance of the distribution
| Porosity
m

Matrix pseudo-pressure [M/LT
3
]
m

Average matrix block pseudo pressure [M/LT
3
]
f

Fracture pseudo pressure [M/LT
3
]
i

Initial pseudo pressure [M/LT
3
]
D

Dimensionless pseudo pressure
D

Average dimensionless pseudo pressure

Subscripts

D Dimensionless
e Equivalent
f Fracture
g Gas
i Initial condition
ln Log-normal
m Matrix
min Minimum
max Maximum
R Reservoir
sc Standard conditions

85
References
Abdassah, D. Ershaghi, I. (1986). Triple-porosity systems for representing naturally
fractured reservoirs. Soc. Pet. Eng. J., 1(2), 113-127.
Ames, W.F. (1965). Nonlinear partial differential equations in engineering. New York:
Academic Press.
Barenblatt, G.E. Zheltov, I.P. Kochina, I.N. (1960). Basic concept on the theory of
homogeneous liquids in fissured rocks. J. Appl. Math. Mech., 20, 852-864.
Belani, A.K. (1988). Estimation of matrix block size distribution in natuarally fractured
reservoirs. MSc Thesis, Stanford University.
Belani, A.K. Jalali-Yazdi, Y. (1988). Estimation of matrix block size distribution in
naturally fractured reservoirs. Paper SPE 18171.
Bogdanov, I.I. Mourzenko, V.V. Thovert, J.F. Adler, P.M. (2003). Pressure drawdown
well tests in fractured porous media. Water Resour. Res., 39(1), 1021.
double-permeability systems-A state-of-the-art review. Transp. Porous Med., 4,147-184.
Cinco-Ley, H. Samaniego, V. Kucuk, F. (1985). The pressure transient behavior for
naturally fractured reservoirs with multiple block size. Paper SPE 14168.
Civan, F. Rasmussen M.L. (2002). Analytical hindered-matrix-fracture transfer models for
naturally fractured petroleum reservoirs. Paper SPE 74364.
Crank, J. (1975). The mathematics of diffusion. Oxford: Clarendon Press.
Fathi, E. Akkutlu, I.Y. (2009). Matrix heterogeneity effects on gas transport and adsorption
in coalbed and shale gas reservoirs. Transp. Porous Med., 80, 281304.
Finlayson, B.A. (1972). The method of weighted residuals and variational principles. New
York: Academic Press.
86
Gwo, J.P. OBrien, R. Jardin, P.M. (1998). Mass transfer in structured porous media:
embedding mesoscale structure and microscale hydrodynamics in a two-region model. J.
Hydrol., 208, 204-222.
matrix-fracture transfer shape factor. Transp. Porous Med., 64, 51-71
solution for well testing of dual-porosity systems. Adv. Water Res., 32(11), 1652-1663.
Jelmert, T.A. (1995). The effect of a distributed block length function on double porosity
transitions during linear flow. J. Pet. Sci. Eng., 12, 277-293.
Johns, R.T. Jalali-Yazdi, Y. (1991). Comparison of transient response in intensely and
sparsely fractured reservoirs. Soc. Pet. Eng. J., 6(4), 513-518.
Kreyszig, E. (1999). Advanced engineering mathematics. United States: John Wiley &
Sons Press.
Leung, C.T.O. Zimmerman, R.W. (2012). Estimating the hydraulic conductivity of two-
dimensional fracture networks using network geometric properties. Transp. Porous
Med., 93, 777-797.
Porous Med., 68,153-173.
Lu, M. Connel, L.D. (2011). A statistical representation of the matrixfracture transfer
function for porous media. Transp. Porous Med., 86,777-803.
Mourzenko, V.V. Bogdanov, I.I. Thovert, J.-F. Adler, P.M. (2011). Three-dimensional
numerical simulation of single-phase transient compressible flows and well-tests in
fractured formations. Math. .Comput. Simul., 81, 2270-2281.
for interporosity flow in naturally fractured gas-condensate reservoirs. SPE Paper 75524.
87
Pooladi-Darvish, M. Tortike, W.S. Farouq Ali, S.M. (1994). Non-isothermal gravity
drainage under conduction heating. Petroleum Society of CIM and AOSTRA, Paper NO
94-65.
flow of a compressible fluid in dual-porosity media. Adv. Water Res., 34(5), 627-639.
Ranjbar, E. Hassanzadeh, H. Chen, Z. (2011). Effect of fracture pressure depletion regimes
on the dual-porosity shape factor for flow of compressible fluids in fractured porous
media. Adv. Water Res., 34(12), 1681-1693.
Rangel-German, E. Kavscek, A.R. Akin. S. (2010). Time-dependent shape factors for
uniform and non-uniform pressure boundary conditions. Transp. Porous Med., 83,591-
601.
Reis, J.C. (1998). Effect of fracture spacing distribution on pressure transient response in
naturally fractured reservoirs. J. Pet. Sci. Eng., 20, 31-47.
Rodriguez, N.R. Cinco-Ley, H. Samaniego, V.F. (2001). A variable block size model for
the characterization of naturally fractured reservoirs. Paper SPE 71570.
Segall, P. (1981). The development of joints and faults. PhD dissertation Stanford
University, Stanford, California.
van Heel, A.P.G. van Dorp, J.J. Boerrigter, P.M. (2008). Heavy oil recovery by steam
injection in fractured reservoirs. SPE Paper 113461.
Eng. J., 3(3), 245-255.
Zimmerman, R.W. Bodvarsson, G.S. (1995). Effective block size for imbibitions or
absorption in dual-porosity media. Geophys. Res. Lett., 22(11), 1461-1464.
Res., 29(7), 2127-2137.
for flow in unsaturated dual-porosity media. Adv. Water Res., 19(5), 317-327.
88
Appendix 3.A: Solution of diffusion equation for variable block size distribution
Integral methods and the method of moments are used to derive a closed form approximate
solution to nonlinear diffusion equations. In this appendix a heat integral method (early
time solution) and the method of moments (late time solution) are explained for solving the
diffusion equation for compressible fluids in the fractured porous media in more details.
The integral method is used to derive the early time solution (Equation (3.34)) of the
nonlinear diffusion equation as follows:
) ) ( (
D
D
D
D D
D
x
t
x t c
c
c
c
=
c
c
|q

(3.A.1)
0 0 = =
D D
t (3.A.2)
1 = =
D De D
L x (3.A.3)
0 0 0 ) (
2
2
=
c
c
=
c
c
= =
D
D
D
D
D D D
x x
t x

o (3.A.4)
The following trial solution is used in the heat integral method:
3
)
) (
1 (
D De
D De
D
t L
x L
o
=
(3.A.5)
Using this trial solution in the integral form of Equation (3.A.1) leads to the following
ordinary differential equation:
De D D D De De
L t dt L d L = = = ) 0 ( 0 12 ) ( ) (
1
o |q o o (3.A.6)
Solving this ordinary differential equation leads to the following equation for :
D D De
t L
1
24|q o = (3.A.7)
The early time solution is valid until 0 = o so
1
2
1
24
0 24
D
De
D D D De
L
t t L
|q
|q = = (3.A.8)
Therefore, the final solution for the early time has the following form which is the same as
Equation (3.34) in the text:
1
2
3
1
24
)
24
1 (
D
De
D
D D
D De
D
L
t
t
x L
|q |q
<
=
(3.A.9)
89
The late time solution (Equation (3.35)) of the nonlinear partial differential equation (PDE)
using the method of moments is explained in more details in this part.
) ) ( (
D
D
D
D D
D
x
t
x t c
c
c
c
=
c
c
|q

(3.A.10)
3
3
1
2
1
24
D
De
D
D
De
D
x
L
L
t = =
|q
(3.A.11a)
0 0 =
c
c
=
D
D
D
x
x

(3.A.11b)
1 = =
D De D
L x (3.A.11c)
The method of moments is used to obtain the late time solution of this PDE by
recommending a third order trial solution and the residual (R) as follows:
1
2
3 2
24
, ) ( ) ( ) ( ) ( ) , (
D
De
D D D D D D D D D D D
L
t x t D x t C x t B t A t x
|q
> + + + = (3.A.12)
) (
D
D
D
D D
D
x x t
R
c
c
c
c
c
c
=

|q

(3.A.13)
The unknown coefficients (A, B, C and D) in Equation (3.A.12) are found by using the
above boundary conditions and making the zero and first moments of R vanish by
enforcing the following conditions:
} }
=
c
c
c
c
c
c
=
De De
L
D
D
D
D
D D
D
L
D
dx
x
t
x t
Rdx
0 0
0 )) ) ( ( ( 0

|q
(3.A.14)
} }
=
(
c
c
c
c
c
c
=
De De
L L
D
D
D
D
D D
D
D D D
dx
x x t
x Rdx x
0 0
0 ) ( 0

|q
(3.A.15)
From the first boundary conditions, we can conclude that B = 0; the second boundary
condition (Equation (3.A.11c)) leads to:
3 2
1
De De
DL CL A = (3.A.16)
Solving Equations (3.A.14) and (3.A.15), combining the results with Equation (3.A.16)
and simplifying, we obtain the following system of ordinary differential equations (ODEs):
D
L
C
L t d
C d
De
D
De
D
D
1
2
1
48 12 |q |q
+ =
(3.A.17)
90
D
L
C
L t d
dD
De
D
De
D
D
2
1
3
1
3
140
3
40 |q |q
=
(3.A.18)
The eigenvalues of this system of ODEs are obtained as follows (Kreyszig, 1999):
2
1
2 2
1
1
181 . 32
and
486 . 2
De
D
De
D
L L
|q
|q
=
(3.A.19)
The corresponding eigenvectors are then obtained, and, finally, the following equations for
the unknown coefficients of C and D are derived.
) exp( 086 . 1 ) exp( 314 . 3
2 2 1 1 D De D De
t m L t m L C = (3.A.20)
) exp( ) exp(
2 2 1 1 D D
t m t m D + = (3.A.21)
When
1
2
24 /
D De D
L t |q = we have
3 3
/
De D D
L x = . Therefore, m
1
and m
2
are obtained using
initial condition for the late time as follows:
3 2 3 1
686 . 5 541 . 0
De De
L
m
L
m =
=
(3.A.22)
Therefore, the trial solution of the nonlinear PDE for the late time behaviour is obtained as
follows:
. )) exp(
686 . 5
) exp(
541 . 0
(
24
)) exp(
175 . 6
) exp(
793 . 1
(
)) exp( 489 . 0 ) exp( 252 . 1 1 ( ) , (
3
2 3 1 3
1
2
2
2 2 1 2
2 1
D D
De
D
De
D
De
D D D
De
D
De
D D D D D
x t
L
t
L
L
t x t
L
t
L
t t t x

|q

+
> +
+ + =

(3.A.23)
Appendix 3.B: Derivation of dimensionless rate and dimensionless cumulative
production for different matrix block size distribution
In this appendix the derivation of Equations (3.55), (3.56) and cumulative fluid exchange
as a function of time is explained in more details. The equation for the matrix-fracture
transfer function was derived as follows (Ranjbar and Hassanzadeh, 2011):
) (
4
f m
m
sc
b sc
sc
T
k
p
V T
q
o
=
(3.B.1)
For the case of variable block size distribution Equation (3.B.1) can be expressed as
follows:
91
) (
4
2
2
f m
me
me m
sc
b sc
sc
Th
h k
p
V T
q
o
= (3.B.2)
In Equation (3.B.2), T
sc
and p
sc
are standard temperature and pressure respectively, V
b
is
the bulk volume of a matrix block, T is reservoir temperature, k
m
is matrix permeability,
is the shape factor, h
me
is the equivalent matrix block thickness,
m
is the average matrix
block pseudo-pressure and
f
is the fracture pseudo-pressure. For the case of variable
block size distribution V
b
is defined as follows:
me b
h A V = (3.B.3)
Substituting Equation (3.B.3) in Equation (3.B.2) and using the definition of the
dimensionless pseudo-pressure lead to the following transfer function for a representative
matrix block in the case of variable block size distribution:
) 1 (
8
) (
2
=
D me
ce sc
i f m sc
sc
h
TL p
Ak T
q o

(3.B.4)
In the case of block size distribution the reservoir bulk volume contains the L
R
/L
ce
number
of these blocks and we have the following equation for the production rate in the bulk
volume of the reservoir:
ce
R
D me
ce sc
i f m sc
scR
L
L
h
TL p
Ak T
q ) 1 (
8
) (
2
= o

(3.B.5)
Starting with Darcys law we have the following equation for maximum fluid transfer:
max
max
) (
2
c
i f
sc
sc m
sc
L T p
AT k
q

=
(3.B.6)
Dividing both sides of Equation (3.B.5) by the maximum fluid transfer (Equation (3.B.6))
leads to the following equation for the dimensionless rate for the reservoir in the case
variable block size distribution:
2
2
2
max
2
) 1 (
4
) 1 (
4
De
D
me
ce
c R
D
me
D
L
h
L
L L h
q
k
o
= = (3.B.7)
It should be mentioned that in Equation (3.B.7), is the ratio of L
R
to L
cmax
for each
distributions. By substituting the equations for the dimensionless shape factor (Equations
(3.43) and (3.44)) and dimensionless average pseudo-pressure (Equations (3.36) and
92
(3.37)) in the Equation (3.B.7), for the entire reservoir the following equation is derived to
calculate the dimensionless matrix-fracture fluid transfer for different block size
distributions.
| |

> +
<
=
1
2
2 1 2
1
1
2
1
24
, ) exp( 763 . 4 ) exp( 964 . 1
24
,
8
24
D
De
D D D
De D
D
D
De
D
D De D
D
D
L
t t t
L
L
t
t L
q
|q

q
k|q
|q q
|q k
(3.B.8)
The cumulative fluid production is defined based on the following equation:
}
=
t
sc sc
d q Q
0
) ( t t (3.B.9)
The definition of the dimensionless rate and dimensionless time is used in this equation to
find dimensionless cumulative production as follows:
}
=
D
t
D D D
c
sc sc
d q
L
q Q
0
2
max
max
) ( t t
q
(3.B.10)
Using the definition of maximum cumulative production (Equation (3.B.10)) for any
distribution in Equation (3.B.9) leads to the following equation for the dimensionless
cumulative production:
q
2
max
max max
c
sc sc
L
q Q = (3.B.11)
}
=
D
t
D D D D
d q Q
0
) ( t t (3.B.12)
For the presented solution the dimensionless cumulative fluid production for the entire
reservoir is determined by using the following equation:
93
| |
> +
+ =
< =
=
}
}
1
2
24
2 1 2
1
1
2
1
1
2
0
1
1
24
, ) exp( 763 . 4 ) exp( 964 . 1
)
24
(
24
,
8
24
1
2
D
De
D D
t
L
D D
De D
D
D
De
D D
D
De
D D
t
D De D
D
D
D
L
t d
L
L
t Q
L
t d
L
Q
Q
D
D
De
D
|q
t t t
q
k|q
|q
|q
t
t q
|q k
|q

(3.B.13)
Integrating of Equation (3.B.13) leads to the following equation for dimensionless
cumulative production as a function of dimensionless time:
| |

>
<
=
1
2
2 1
1
2
1
24
, ) exp( 148 . 0 ) exp( 790 . 0 001 . 1
24
,
4
24
D
De
D D D
D
D
De
D D
De D
D
D
L
t t t
L
t t
L
Q
|q

q
k
|q q
|q k
(3.B.14)
Substituting for dimensionless variables (Equations (3.25), (3.27), (3.B.6) and (3.B.11))
and using the definition of in Equation (3.B.14) leads to the following equation for the
cumulative production as a function of real time. This equation was used to compare the
results of this study with the numerical results for verification.
>
(
(
(
(

<

=
q |q q |q
q |q
q q

q |q q
|q
q

1
2
2
1
2
1
1
2
1
24
,
)
181 . 32
exp( 148 . 0
)
486 . 2
exp( 790 . 0 001 . 1
2
) (
24
,
24
8
) (
) (
D
ce
ce
D
ce
D
D sc
R i f m sc
D
ce D
ce D sc
R i f m sc
L
t
t
L
t
L
T p
L Ak T
L
t
t
L T p
L Ak T
t Q

(3.B.15)

Chapter Four: Semi-analytical solutions for release of fluids from rock matrix blocks
with different shapes, sizes and depletion regimes
1

Abstract
Dual-porosity models have been extensively used to simulate the flow of fluids (water or
gas) in aggregate soils and fractured porous media. The fluid exchange between the rock
matrix blocks and the fracture network is very important in dual-porosity models. In this
study we present semi-analytical solutions for release of a single-phase liquid or gas from
cylindrical and spherical matrix blocks with various block size distributions and different
pressure depletion regimes in the fracture. The nonlinear pressure diffusivity equations for
flow of gas and air are solved analytically using an approximate integral method. It is
shown that this solution can be simplified to model flow of slightly compressible fluids
like water or oil in dual-porosity media. The effect of variable block size distribution on
the release rate for different block geometries is studied.
Practically it is not feasible to model a large scale fractured reservoir based on a fine grid
approach due to the requirement of large computational time. The presented semi-
analytical model can be incorporated into numerical models for accurate modeling of the
amount of transferred fluid between matrix and fractures using a dual-porosity approach. It
is shown that the results calculated by the developed model match well with those from
fine grid numerical simulations. Furthermore, the developed model can recover the
available solutions in the literature for slightly compressible fluids such as water or oil. It
can be used to calculate two- or three-dimensional flows in matrix blocks bounded by two
or three sets of fractures, respectively.

1
This chapter is an exact copy of: Ranjbar, E. Hassanzadeh, H. Chen, Z. (2013). Semi-analytical solutions
for release of fluids from rock matrix blocks with different shapes, sizes and depletion regimes, Water
Resources Research, Volume 49 (4), Page: 2174-2196.

Chapter 4. Semi-analytical solutions for release of fluids from rock matrix blocks
95
4.1 Introduction and previous studies
The fluid flow and transport in fractured rocks is of great significance in many
groundwater environments (Novakowski and Lapcevic, 1994). Flow and transport in
fractured porous media are often described using a dual-porosity model. This model
assumes that the porous medium includes two different regions, one related with the
macropore or fracture network with high permeability and the other with a less permeable
and more porous system of soil aggregates or rock matrix blocks. Dual-porosity models
assume that both water flow and solute transport can be described by two equations for
matrix and fractures, which are coupled using a term describing the exchange of fluid or
solutes between the two pore regions (Gerke and van Genuchten, 1993). The dual-porosity
approach has been used in numerical simulation of groundwater, oil, and gas flow in
fractured porous media. However, for large-scale simulations the use of this approach is
limited by the huge number of grid-blocks and small time steps that are often needed to
accurately simulate the flow and are expensive in terms of computational time. Therefore,
developing analytical and semi-analytical approaches that can handle fractured porous
media or structured soils with less computational time is important (Zimmerman and
Bodvarsson, 1989).
Study of gas flow in unsaturated soils and fractured porous media is significant in a variety
of engineering fields. In agricultural engineering, airflow in the root zone due to barometric
pressure variation is important to plant growth. Recent studies have shown that airflow can
be applied to create a "dry barrier" for waste disposal facilities (Shan, 1995). Complex gas-
water processes in fracture-matrix systems are the main processes in a range of
environmental engineering systems, changing from CO
2
storage in deep geological
formations (Altevogt and Celia, 2004; Chen and Zhang, 2010) over radioactive waste
disposal in caverns to evaporation processes in the unsaturated zone (Nuske et al., 2010).
As an example, soil vapor extraction is a widely used technique to eliminate volatile
organic contaminants from the unsaturated zone (Falta, 1995). In this process gas is
induced by vapor extraction for cleaning up vadose zone contamination of volatile organic
96
chemicals. The gas flow is also important in the unsteady flow of air in an anisotropic
layer of snow (Fan and Yen, 1968).
Another important aspect of gas flow in the fractured media is air injection tests to
determine the hydrologic properties and parameters of the fracture networks (Huang et al.,
1999). Illman and Neuman (2001) developed pressure and pressure derivative type curves
for single-phase air flow. Illman (2005) analyzed the single-hole pneumatic injection and
recovery tests for determining permeability, porosity, and skin. Shan (1995) developed
analytical solutions for transient, one-dimensional gas flow caused by barometric pumping,
and applied these solutions to estimate the air permeability of the vadose zone. In
geothermal reservoirs water and vapor (steam) transport occurs in fractured porous media.
There are several studies relevant to flow of fluids in geothermal reservoirs, which is
another important application of compressible fluids (vapor) flow in the fractured media
(Pruess, 1983; Fitzgerald and Woods, 1998). Schrauf and Evans (1986) and Parker et al.,
(2006) studied the flow of gas in a single natural fracture and porous media, respectively.
In hydrology there exist a large number of mathematical models (numerical, analytical or
semi-analytical) to simulate the flow of compressible fluids in underground environments
and structured soils (You et al., 2011). McWhorter (1990) presented an exact semi-analytic
solution for transient radial gas flow and applied it for estimating the gas permeability
using pumping test data. Wang and Dusseault (1991) obtained analytical solutions for
compressible fluids flowing through a saturated poroelastic medium by solving a new
density diffusion equation. Falta (1995) developed an analytical solution for transient and
steady state compressible gas flow to a pair of horizontal wells in an unsaturated zone.
Shan et al., (1999) presented analytical solutions for transient, two-dimensional gas flow in
a vertical vadose (unsaturated) zone section, and presented techniques for approximating
the air permeability of a vertical leaky fault. Shan (2006) developed an analytical solution
for transient gas flow in a multiwell system.
In the case of gas-water displacement, Thunvik and Braester (1990) analyzed the
displacement of gas-water in fracture networks with different permeability and different
inclination by focusing mainly on the gas breakthrough time. Berger and Braester (2000)
presented a mathematical model for displacement of gas-water in the fracture networks
97
with a nonlinear system of partial differential equations and solved this system numerically
using an iterative approach.
Flow of compressible and slightly compressible fluids (water or oil) in fractured reservoirs
has been studied extensively with applications in prediction of production rates and well
testing (Warren and Root, 1963; Kazemi et al., 1976; Zimmerman et al., 1993, 1996; Civan
and Rasmussen, 2002; Penuela et al., 2002; Bogdano et al., 2003; Lu and Connel, 2007;
van Heel et al., 2008; Mora and Wattenbarger, 2009; Hassanzadeh et al., 2009; Ranjbar
and Hassanzadeh, 2010; Ranjbar et al., 2011; Mourzenko et al., 2011; Ranjbar et al., 2012;
Ye and Ayala, 2012). As an example, Hoteit and Firoozabadi (2005) developed a discrete
fracture model to simulate the flow of compressible fluids in homogeneous, heterogeneous
and fractured porous media.
There are also a number of studies that have considered the effect of block geometries on
dual-porosity fluid flow formulation (Barker, 1985; van Genuchten and Dalton, 1986;
Wuthicharn and Zimmerman, 2011). As an example, Zimmerman et al., (1990) developed
a dual-porosity model based on an approximate solution for absorption of water into slab-
shaped, cylindrical and spherical blocks.
According to the aforementioned works the flow of compressible fluids like gases and air
in porous media is important in hydrological, environmental and petroleum engineering. In
our previous studies semi-analytical models for a slab-shaped (one-dimensional) matrix
block was developed and the effect of fracture pressure depletion regimes and matrix block
size distribution was investigated (Ranjbar and Hassanzadeh, 2011; Ranjbar et al., 2011;
Ranjbar et al., 2012).
The main objective of this study is to develop a new semi-analytical model for different
matrix block geometries (cylindrical and spherical) for flow of compressible and slightly
compressible fluids in fractured porous media. It is emphasized that it is not practical to
model a large scale fractured reservoir based on a fine grid approach due to the
requirement of large computational time. The presented semi-analytical model can be
incorporated into numerical models for accurate modeling of the amount of transferred
fluids between matrix and fractures using available dual-porosity formulation. In other
words, this study is important to reduce the computational time for large scale simulations
98
of gas flow in fractured porous media and can be nested in a numerical model to resolve
subgridblock scale flows.
In the case of compressible fluids the presented model may find applications for soil vapor
extraction, geological CO
2
sequestration, hydrological determination of fracture properties
by air injection, and flow of gas in dual-porosity reservoirs. For slightly compressible
fluids the presented model is capable to model flow of water or oil in fractured media for
different block geometries and different block size distributions. In the presented study the
models for release of fluids from cylindrical (representative of two sets of fractures or 2D)
and spherical (representative of three sets of fractures or 3D) blocks are derived. The
presented models can be expressed explicitly in terms of time and does not require the
numerical inversion or infinite series calculations. This solution approach can significantly
decrease the computational time of dual-porosity models with an acceptable accuracy.
4.2 Approximate analytical solution
A transfer function is utilized to characterize the matrix-fracture interaction and determine
the mass transfer between the matrix blocks and the fractures. The rate of mass transferred
from the matrix to the fractures is directly related to the shape factor. For modeling of
naturally fractured reservoirs, an exact value of the shape factor is required to account for
both the transient and pseudo-steady state behaviour of the matrix-fracture interaction and
also the geometry of the matrix-fracture system (Ranjbar and Hassanzadeh, 2011).
In this section the dual-porosity matrix-fracture transfer function for cylindrical and
spherical blocks is derived. After that an approximation is used to derive the matrix-
fracture fluid transfer for 2D flow (cylindrical matrix blocks approximation) or slab-
shaped blocks surrounded by two sets of fractures and 3D flow (spherical matrix blocks
approximation) or slab-shaped blocks surrounded by three sets of fractures. A similar
approach has been used in the literature (Zimmerman et al., 1990; Lim and Aziz, 1995;
Hassanzadeh and Pooladi-Darvish, 2006). To derive the transfer function for dual-porosity
cylindrical and spherical blocks we use Darcys law as given by:
dr
dp
B
A k
q
g
m
sc
= .
(4.1)
99
In this equation, A is the surface area, r is the distance from the center of the cylinder or
sphere and B
g
is the gas formation volume factor (the volume of fluid at underground
conditions divided by the volume of fluid at standard surface conditions). The following
equations are used to determine the area for cylindrical and spherical blocks, respectively:
rh A
Cylinder
t 2 = ,
(4.2)
2
4 r A
Sphere
t = .
(4.3)
Using the definition of a compressible fluid formation volume factor and the surface area
of the cylinder we have:
} }
=
'
f
m
m
m
p
p sc
sc
sc
m
R
r R
dp
Z
p
p
T
Tq
h k
r
dr
t 2
. (4.4)
In Equation (4.4), r is a time-dependent radius where the matrix pressure is equal to its
average pressure, h is the height of the cylinder and and Z are the gas viscosity and
compressibility factor, respectively. In gas reservoirs pseudo-pressure transformation is
used to account for the variability of pressure with viscosity and gas compressibility factor.
The pseudo-pressure transformation, which is similar to the Kirchhoff transformation
(Tartakovsky et al., 1999), is given by:
dp
z
p
p
p
b
}
=
2 .
(4.5)
where p
b
is a reference or base pressure. Integrating of Equation (4.4) and using the
definition of the real gas pseudo-pressure (Equation (4.5)) leads to the following equations
for cylindrical and spherical blocks, respectively:
2
) ln(
2
f m
m m
m
sc
sc m
l Cylinerica sc
r R
R
h
Tp
T k
q

t

'
=
,
(4.6)
2
) ( 4
f m
m m
sc
sc m
Spherical sc
r
r R R
Tp
T k
q

t

'
'
=
.
(4.7)
The shape factor for conduction of heat through a cylindrical wall or spherical block is
defined as follows, respectively (Holman, 2010):
100
1
2
ln
2
D
D
h
S
Cylinder
t
= ,
(4.8)
) (
4
1 2
2 1
r r
r r
S
Sphere
=
t
.
(4.9)
where D
1
and D
2
are the diameters of the inner and the outer cylinders with higher and
lower temperatures, respectively, and r
2
is the larger sphere radius with lower temperature.
Using the same notion for the case of pressure diffusion, the following equations are used
to define the pressure diffusion shape factor for cylindrical and spherical blocks,
respectively. This parameter is one of the most important parameters in dual-porosity
modeling of fractured reservoirs. The amount of fluid that is transferred from matrix to
fractures is directly proportional to the shape factor:
b
m
m
Cylinder
V
r R
R
h
) ln(
2
'
=
t
o ,
(4.10)
b
m m
Sphere
V r
r R R
'
'
=
) ( 4t
o .
(4.11)
In the shape factor equation, V
b
is the bulk volume of a matrix block. Using the definition
of the shape factor (Equations (4.10) and (4.11)) in Equations (4.6) and (4.7) leads to the
following equation for the transfer function of a cylindrical and spherical block in terms of
the shape factor:

2
f m
sc
b sc m
sc
Tp
V T k
q

o

= .
(4.12)
The equation for fluid transfer in terms of time derivative can be expressed as follows
(Ranjbar and Hassanzadeh, 2011):
t T
c
p
V T
q
m m m
sc
b sc
sc
c
c
=
|
2
.
(4.13)
Solving Equations (4.12) and (4.13) for the shape factor results in the following equation:
( ) t k
c
m
f m m
m m
c
c
=

|
o .
(4.14)
101
Solution of the diffusivity equation is used in Equations (4.12) and (4.14) to derive the
matrix-fracture fluid transfer and shape factor for compressible or slightly compressible
fluids.
The diffusivity equation for flow of compressible fluid for different geometries can be
expressed as follows:
t k
c
r
r
r r
m
m
m m m I
I
c
c
=
c
c
c
c |
) (
1
,
(4.15)
where for a cylindrical block I=1 and for a spherical block I=2. It should be mentioned
that Equation (4.15) is a nonlinear partial differential equation (PDE). This nonlinearity is
due to the pressure dependence of viscosity and compressibility of the compressible fluids
like air or gas. The diffusivity equation for a cylindrical and spherical block in terms of
hydraulic diffusivity, , (ratio of matrix permeability to the product of gas viscosity, matrix
compressibility (fluid and rock) and matrix porosity) can be expressed as follows:
t
r
r
r
r
m I m I
c
c
=
c
c
c
c
q ) ( .
(4.16)
At the initial condition a pseudo-pressure in the matrix block can be obtained from the
initial pressure. At the center of the cylindrical or spherical matrix block we have no flow
boundary condition and at the outer boundary, the pressure is equal to the fracture pressure,
which may be a constant or varies with time.
Figure 4.1 is a schematic representation of the problem for a cylindrical block surrounded
by fractures. For a spherical block the problem is the same but instead of a cylinder we are
dealing with a spherical block.

102

Figure 4.1: Schematic representation of the problem for a cylindrical block.

To solve this equation the average hydraulic diffusivity over the matrix-fracture pressure
drawdown is defined as follows (Ranjbar and Hassanzadeh, 2011):
} }

=
=
f
i
f
i
p
p m i f m
m
p
p m m
m
i f
c
dp
p p
k
dp
c
k
p p | |
q
1 1
. (4.17)
To express the diffusivity equation in dimensionless form the following dimensionless
variables are defined:
i f
i m
D

= ,
(4.18)
m
D
R
r
r = ,
(4.19)
q
q
q
) (
) (
t
t
m
D
= ,
(4.20)
2
m
D
R
t
t
q
= .
(4.21)
Using these dimensionless variables in Equations (4.14) and (4.16) leads to the following
equations for the dimensionless shape factor and diffusivity equation:
D
D
fD D D
m
t
R
c
c
|
|
.
|
\
|
=

q
o
1 1
2
,
(4.22)
R
m

0 =
c
c
r

f
=
Matrix
Fracture
103
D
D I
D
D
D I
D
D
D D D
t
r
r
r
r
t r
c
c
=
c
c
c
c
q ) ( ) , ( .
(4.23)
In this equation the hydraulic diffusivity is a function of the dimensionless radius and
dimensionless time. For solving this equation we assume that
D
is only a function of the
dimensionless time. To consider the effect of the space we multiply
D
by a correction
factor and numerical simulation is used to find this correction factor (Ranjbar and
Hassanzadeh, 2011). Therefore, we have the following dimensionless equation for the
diffusivity equation of the cylindrical or spherical block:
D
D I
D
D
D I
D D D
D
t
r
r
r t
r c
c
=
c
c
c
c
|q ) ) ( ( .
(4.24)
In this equation, I=1 is used for the cylindrical block and I=2 is used for the spherical
block. This equation is solved for different fracture boundary conditions and different
block size distributions to derive the shape factor and matrix-fracture fluid transfer for flow
of compressible and slightly compressible fluids in fractured porous media or aggregate
and structured soils. It should be mentioned that the presented solution in this study
calculates the pressure distribution in the matrix block by assuming fractures as a boundary
condition.
In this case the fracture pressure at the matrix-fracture interface is a constant and we are
dealing with a single block. For the constant fracture pressure we have the following initial
and boundary conditions for the non-linear diffusivity equation (Equation (4.24)):
0 0 = =
D D
t , (4.25)
0 0 =
c
c
=
D
D
D
r
r

,
(4.26)
1 1 = = =
fD D D
r .
(4.27)
An integral method (Goodman, 1964; Zimmerman and Bodvarsson, 1989; Pooladi-Darvish
et al., 1994) and the method of moments (Ames, 1965) are used to find the early and late
time solutions of this equation respectively. After solving the diffusivity equation and
104
integration over the bulk volume of the cylindrical matrix block we obtain the early and
late time average dimensionless pseudo-pressures as follows:
1
1
48
1
,
3
48
D
D
D D
D
t
t
|q
|q
< = , (4.28)
1
48
1
, 267 . 0 664 . 0 1
2 1
D
D
t t
D
t e e
D D
|q
e e
> = ,
(4.29)
where
1
and
2
are defined as follows for a cylindrical block:

1 2 1 1
23 . 58 77 . 5
D D
|q e |q e = = . (4.30)
Using the same approach leads to the following early and late time average dimensionless
pseudo-pressures for a spherical block:
1
2 3
720
17
20
15 6
D
D D
t
|q
o o o
<
+
= ,
(4.31)
,
720
17
,
720
30 16 3
1 1
2 3 4
D
D
D
D
t t
|q |q
o o o
<
+
=
(4.32)
1
720
17
, 371074 . 0 554435 . 0 1
2 1
D
D
t t
D
t e e
D D
|q
e e
> = ,
(4.33)
where
1
and
2
for a spherical block are defined as follows:

1 2 1 1
520264 . 76 879736 . 9
D D
|q e |q e = = . (4.34)
In these equations
D1
is the fracture dimensionless hydraulic diffusivity. Derivation of
these equations is shown in Appendixes 4.A1 and 4.B1 in more details. In Equation (4.31),
is the time-dependent penetration depth for a spherical block and Equation (4.32) is used
to relate the penetration depth to the dimensionless time.
Equations (4.28) and (4.29) (for a cylindrical block) and (4.31)-(4.33) (for a spherical
block) and their derivatives with respect to time are substituted into Equations (4.12) and
(4.22) to derive the matrix-fracture transfer rate (release rate) and shape factor for the
cylindrical and spherical matrix blocks. To express the rate in the dimensionless form we
write Darcys law over the whole domain of the cylinder to reach the following equation:
) (
2 i f
m sc
b sc m
g
R Tp
V T k
q
= .
(4.35)
105
Using the definition of the dimensionless pseudo-pressure in the transfer function
(Equation (4.12)) leads to the following equation:
) 1 (
) (
2

=
D
i f m
sc
b sc
sc
T
k
p
V T
q o

.
(4.36)
Dividing Equation (4.36) to Equation (4.35) leads to the following equation for the
dimensionless release rate for the cylindrical and spherical blocks:
( )
D
D
D
D
m
g
sc
D
t I I
R
q
q
q
c
c
+
=
+
= =

q
o
1
1
) 1 (
1
2
.
(4.37)
Integrating of these equations leads to the dimensionless cumulative fluid release from a
cylindrical or spherical block with a constant fracture pressure, respectively:
( )
D
D
t
D D D
I
dt q Q
D
q
1
0
+
= =
}
. (4.38)
The early and late time solutions of the diffusivity equation (Equations (4.28) and (4.29)
and (4.31)-(4.33)) will be used in Equations (4.37)-(4.38) to determine the dimensionless
rate and the cumulative fluid release for the cylindrical and spherical blocks when the
fracture pressure is constant. It should be pointed out that the presented model can be used
to model flow of slightly compressible fluids like water or oil if we set 1
1
= = =
D D
q q | .
4.2.2 Variable fracture pressure
In this case we assume that the outer boundary condition changes with time. In this section
we consider the effect of linearly and exponentially declining fracture pressure on the dual-
porosity formulation of compressible and slightly compressible fluids.
For the variable fracture pressure the outer time-dependent boundary condition for the
diffusivity equation (Equation (4.24)) can be linear or exponential as follows, respectively:
D D fD
i
m i
D
t t k
= ) ( ,
(4.39)
) exp( 1 ) (
D D fD
i
i m
D
t t k

=
.
(4.40)
It should be noted that the initial condition and the inner boundary condition are the same
as Equations (4.25) and (4.26), respectively. Using a recently developed semi-analytical
106
method to solve the equation for a variable fracture pressure (Mitchel and Myers, 2010;
Ranjbar et al., 2011) we reach the following equations for the average dimensionless
pseudo-pressure for the linearly declining fracture pressure of a cylindrical block:
1
1
16
1
, 16
3
D
D D D
D
D
t t
t k
|q
|q < = ,
(4.41)
1 2 1
16
1
), 1 (
29639 . 0
) 1 (
69987 . 0
2 1
D
D
t t
D D
t e e t
D D
|q e
k
e
k
k
e e
> = ,
(4.42)
where
1
and
2
are determined based on Equation (4.30) in the case of a cylindrical
block. For a spherical block with linearly declining fracture pressure the following
solutions are obtained for the early and late time average dimensionless pseudo-pressure:
D D
D
D
D
D
t
or t
1
1
3 4 5
1
13
1
13
1
) 15 6 (
260
|q o
o
|q
o o o
|q
k
=
< < + =
,
(4.43)
1 2 1
13
1
), 1 (
359632 . 0
) 1 (
626902 . 0
2 1
D
D
t t
D D
t e e t
D D
|q e
k
e
k
k
e e
> = .
(4.44)
In the case of an exponentially declining fracture pressure we have the following equations
for the average pseudo-pressure of a cylindrical block:
*
1
, )
) (
2 ( 24
3
1
t t
t
t erf
t
e
D
D
D
D D
t
D
D
<
=

k
k t
|q
k
, (4.45)
*
, )
1
( 07754 . 0 )
1
( 58746 . 0
)
1
( 07754 . 0 )
1
( 58746 . 0 1
* *
2
*
* *
1
*
2
* *
2
*
1
* *
1
*
t t e
e e
e
e e
e
e
e e
e
e
e e
e
e
D
t
t t
t
t t
t
t
t t
t
t
t t
t
t
D
D
D D D
>
(
(
k
k e
k
k e
k
e
k e
k
e
k e
k
k
.
(4.46)
In the case of a spherical block and exponentially declining fracture pressures the
following equations are obtained for the average dimensionless pseudo-pressure:
107
)
) (
1 1
2
( 18 ) (
) 15 6 (
20
) exp( 1
1
* 2 3
D
D
t t D D D
D
D
D
t
t erf
e e
t t
t t
t
D D
k
k t
|q o
o o o
k
k k
=
< +

=
,
(4.47)
*
, )
1
( 060923 . 0 )
1
( 439073 . 0
)
1
( 060923 . 0 )
1
( 439073 . 0 1
*
2
*
*
*
1
*
*
2
*
2
*
*
1
*
1
*
*
t t e
e e
e
e e
e
e
e e
e
e
e e
e
e
D
t
t t
t
t t
t
t
t t
t
t
t t
t
t
D
D
D D D
>
(
(
+ =
k
e k
k
e k
k
e
e k
k
e
e k
k
k
,
(4.48)
where erf is the error function and t
*
is the time at which the effect of pressure disturbance
will reach the matrix boundary. Derivation of these equations is shown in Appendices 4.A2
and 4.B2 in more details. To determine the dimensionless rate and the cumulative fluid
release for variable fracture pressure the same approach as described in Section 4.2.1 is
used. The following equation represents the dimensionless release rate for the variable
fracture pressure of a cylindrical or spherical block:
D
D
D
D fD D
m
D
t I
t
I
R
q
c
c
+
=
+
=

q

o
) 1 (
1
)) ( (
) 1 (
2
.
(4.49)
These equations are integrated to determine the cumulative release from a cylindrical or
spherical block with time-dependent fracture pressure.
D
D
D
I
Q
q
) 1 ( +
= .
(4.50)
Equations (4.41) to (4.48) are substituted into these equations (Equations (4.49) and (4.50))
to determine the early and late time dimensionless release rates and the cumulative release
of fluids for different fracture pressure depletion regimes.
4.2.3 Variable block size distribution (multiple blocks)
In the case of multiple blocks we substitute the collection of blocks with a single block
with an equivalent radius (Zimmerman and Bodvarsson, 1995; Ranjbar et al., 2012). To
model flow of fluids in fractured media or aggregate soils with multiple cylindrical or
spherical blocks of variable block size distributions the following initial and boundary
conditions can be written:
108
t r
r
r r
m
m
m I
I
c
c
=
c
c
c
c
q
1
) (
1
,
(4.51)
i m
t = = 0 ,
(4.52)
0 0 =
c
c
=
r
r
m
, (4.53)
f m me
R r = = .
(4.54)
It should be noted that in Equation (4.54), R
me
is the equivalent cylinder or sphere radius
and is a function of block size distribution. The following equations are used to find the
equivalent radius for discrete and continuous block size distribution, respectively (Gwo et
al., 1998; Ranjbar et al., 2012):
mi
N
i
mi i
N
i
mi
t
i
N
i
mi i
t
N
i
i
N
i
mi i
me
R R f R
N
N
R N
N
N
R N
R

= = =
=
=
= = = =
1 1 1
1
1
) (
1
,
(4.55)
}
=
max
min
) (
m
m
R
R
m m m me
dR R f R R . (4.56)
In Equation (4.56), f(R
m
) is the probability density function (PDF), which is used to
represent the probability of the blocks as a function of block sizes. In the case of multiple
blocks of variable block size distribution in addition to Equations (4.18) and (4.20) the
following dimensionless variables are defined:
max m
D
R
r
r = ,
(4.57)
2
max m
D
R
t
t
q
= ,
(4.58)
max m
me
De
R
R
R = ,
(4.59)
) ( ) (
max m m D D
R f R R f = , (4.60)
max
min
m
m
h
R
R
F = .
(4.61)
109
For the multiple blocks R
mmax
, which is independent of the block size distribution, is used
to scale the time and radius. Using Equations (4.18), (4.20) and (4.57)-(4.61) in the
diffusivity equation and the shape factor equation (Equation (4.14)) for the variable block
size distribution, we reach the following equations for the diffusivity equation, the
equivalent radius and the dimensionless shape factor:
D
D I
D
D
D I
D D D
D
t
r
r
r t
r c
c
=
c
c
c
c
|q ) ) ( ( ,
(4.62)
0 0 = =
D D
t , (4.63)
0 0 =
c
c
=
D
D
D
r
r

,
(4.64)
1 = =
D De D
R r ,
(4.65)
Di
N
i
mi i De
R R f R

=
=
1
) ( ,
(4.66)
}
=
1
) (
h
F
D D D D De
dR R f R R ,
(4.67)
D
D
D D
De
me
t
R
R
c
c
|
|
.
|
\
|
=

q
o
1
1
2
2
.
(4.68)
To solve the diffusivity equation of multiple cylindrical blocks, the integral method and
method of moments are used to find the solution (Ranjbar et al., 2012). The following
equations give the early and late time average dimensionless pseudo-pressure for multiple
cylindrical blocks with block size distribution:
1
2
1
48
,
3
48
D
De
D
De
D D
D
R
t
R
t
|q
|q
< = , (4.69)
1
2
48
, 267 . 0 664 . 0 1
2 1
D
De
D
t t
D
R
t e e
D D
|q
e e
> = ,
(4.70)
where
1
and
2
are defined as follows:

2
1
2 2
1
1
23 . 58 77 . 5
De
D
De
D
R R
|q
e
|q
e

=
= .
(4.71)
Derivation of these equations is shown in Appendix 4.A3 in more details.
110
The following equations give the early and late time average dimensionless pseudo-
pressure for multiple spherical blocks with block size distribution using the integral
method and method of moments:
2
1
2 2 3 4
1
2
3
2 2 3
720
30 16 3
720
17
,
20
15 6
De D
De De
D
D
De
D
De
De De
D
R
R R
t
R
t
R
R R
|q
o o o
|q
o o o
+
=
<
+
=
,
(4.72)
1
2
720
17
, 371074 . 0 554435 . 0 1
2 1
D
De
D
t t
D
R
t e e
D D
|q
e e
> = ,
(4.73)
where
1
and
2

2
1
2 2
1
1
520264 . 76 879736 . 9
De
D
De
D
R R
|q
e
|q
e

=
= .
(4.74)
Derivation of these equations is shown in Appendix 4.B3 in more details. Table 4.1 shows
the equivalent radius and the PDF for different continuous block size distributions. For the
case of discrete block size distribution Equation (4.66) is used to find the equivalent length.
More details about these distributions and the corresponding equivalent radius are
discussed in Ranjbar et al., (2012).

111
Table 4.1: Different probability distribution function and their equivalent radius.
Distribution Distribution Function Equivalent Radius
Exponential
a F a
R a
D D
e e
ae
R f
h
D

= ) (
) (
) 1 ( ) 1 (
a aF
a aF
h
De
e e a
e a e aF
R
h
h

+ +
=
Normal
2
2
) (
2
1
2
2
1
) (
M R
D D
D
e R f

=
o
to

(
(
(
(
(
(
(
(
+
(
2
2
2
2
2
) 1 (
2
) (
2
2 2
2
)
2
1
( )
2
(
2
1
o o
t
o
o o
M F M
h
De
e e
M
erf
F M
erf M
R
h

Linear b mR R f
D D D
+ = ) ( ) 1 (
2
) 1 (
3
2 3
h h De
F
b
F
m
R + =
Log-normal
2
ln
2
ln
2
] ) [ln(
2
ln
2
1
) (
o
to
M R
D
D D
D
e
R
R f

=
(
(
(
(
(
+
=
+
]
2
[
]
2
) ln(
[
2
2
ln
2
ln ln
2
ln
2
ln ln
2
2
ln
ln
o
o
o
o
o
M
erf
F M
erf
e
R
h
M
De

For the case of variable block size distribution the following equation is used to determine
the release rate for the total volume of the reservoir:
me
mR
D
i f m
sc
b sc
scR
R
R
T
k
p
V T
q ) 1 (
) (
2

= o

.
(4.75)
In this equation, R
mR
is the total radius of the reservoir. To determine the dimensionless
rate Equation (4.75) is divided by Darcys rate, which is expressed as follows:
( )
) (
2
1
max
max i f
me m sc
sc b m
g
R TR p
T V k I
q
+
= .
(4.76)
The following equations are used to determine the dimensionless rate of release and the
dimensionless cumulative release for the variable block size distribution for a cylindrical or
spherical block:
( ) ( )
D
D
D De
D
me
D
t I R I
R
q
c
c
+
=
+
=

q
_ _
o
1
) 1 (
1
2
2
,
(4.77)
112
( )
D
D
D
I
Q
q
_
1 +
= .
(4.78)
It should be mentioned that in these two equations is the ratio of the reservoir radius to
the radius of the maximum block. In the result section we assume that this ratio is 10. The
average pseudo-pressures (Equations (4.69)-(4.74)) and their derivatives are replaced in
Equations (4.77) and (4.78) to determine the dimensionless rate of release and the
dimensionless cumulative release for cylindrical and spherical blocks with variable block
size distribution. The equations derived for the cylindrical (spherical) block can be used for
the two-dimensional (three-dimensional) flow in the slab-shaped block if the two types of
blocks have the same volume. For example, if we set t /
2 2
m m
h R = ( ) 4 /( ) 3 (
3 3
t
m m
h R = ) in
these equations the solution can be applied for two-dimensional (three-dimensional)
blocks.
4.3 Model verification
To determine the validity of the presented model we compare our results with the fine grid
numerical simulation. Cumulative release of fluids from matrix to fractures is used to
evaluate the accuracy of the presented model. In addition, comparison of the presented
model with models available in the literature for slightly compressible fluids will be used
to validate the developed model.
To show the accuracy of the results in the case of 2D and 3D flow (two and three sets of
fractures) the obtained solution for the average pseudo-pressure is used in the transfer
function equation (Equation (4.12)). Changing all the variables into the dimensional form
and integrating this equation over time the following equation is obtained for the
cumulative release of fluids for two and three sets of fractures, respectively:
>
(
(

<
=

q t |q q t
q t |q q t
|q
q
t
q t |q q t |q
1
2
23 . 58 77 . 5
2
1
2
1
2
48
, 267 . 0 664 . 0 1
48
,
3
48
2
) (
) (
2
1
2
1
D
m
t
h
t
h
m
D
m D m
D m sc
f i sc b m
h
t e e
h
h
t t
h
Th p
T V k
t Q
m
D
m
D

(4.79)
113
q |q
o o o
q |q
q |q
o o o
q q

q
e
q
e
1
2 3 4 2
1
2
384 . 0 384 . 0
1
2 2 3
720
) 30 16 3 ( 384 . 0
720
384 . 0 17
, 371074 . 0 554435 . 0 1
720
384 . 0 17
,
20
15 6
2
) (
) (
2
2
2
1
D
m
D
m
h
t
h
t
D
m
D sc
f i b sc m
h
t
h
t e e
h
t
T p
V T k
t Q
m m
+
=
>
<
+
=

(4.80)
It should be pointed out that Equations (4.79) and (4.80) are obtained from the cylindrical
and spherical blocks models with the assumption of having the same volume with blocks
formed by two and three sets of fractures, respectively. For example, in the case of a three-
dimensional block (a matrix block formed by three sets of fractures) it is assumed that the
sphere and the cube have the same volume (Lim and Aziz, 1995).
Figures 4.2 and 4.3 compare the cumulative release of gas based on the presented semi-
analytical model (Equations (4.79) and (4.80)) and the numerical results (Eclipse 100) for
two- and three-dimensional flow, respectively. Based on these figures the presented semi-
analytical model is in a good agreement with the fine grid numerical simulations.
Table 4.2 shows the data that has been used in the fine grid numerical simulations and the
presented semi-analytical model. It should be mentioned that the data for the semi-
analytical model and the matching parameters ( and
D
) are the same as those in our
previous studies (Ranjbar and Hassanzadeh, 2011; Ranjbar et al., 2012). This shows that
these parameters are not matrix block geometry dependent.

114

Figure 4.2: Comparison of the presented model with the numerical simulation for 2D flow (cylindrical
block approximation).

Figure 4.3: Comparison of the presented model with the numerical simulation for 3D flow (spherical
block approximation).

Time(second)
0 200 400 600 800 1000 1200 1400
Q
(
S
m
3
)
0
200
400
600
800
1000
1200
1400
Semi-Analytical Model
Numerical Simulation
Time(second)
0 200 400 600 800 1000 1200 1400
Q
(
S
m
3
)
0
200
400
600
800
1000
1200
1400
Semi-Analytical Model
Numerical Simulation
115
Table 4.2: Data used for semi-analytical and numerical models.
Data for fine-grid model
Grid Dimension: 28 (22 grids for matrix and 6 grids for fracture)28 1
Grid Spacing:
x(m): 1000,20,5,0.005,0.01,0.15,0.2,0.25,0.3,0.395,0.5,0.5,0.395,0.3,0.25,0.2,0.15,0.01,0.005,5,20,1000
y: The same as x
z = 4 m (for 2D),

for 3D simulation case the same grid spacing as x is used in z-direction.
Fracture Porosity = 1
Fracture Permeability = 4,000 mD
Common data for semi-analytical and numerical models
Gas Specific Gravity = 0.7
Matrix Permeability = 1 mD = 9.86923310
-16
m
2

Matrix Porosity = 0.1
Initial Pressure = 45 MPa
Fracture Pressure = 22.5 MPa
Reservoir Temperature = 366.45K
h
m
= 4m
Data for semi-analytical model
=0.73 p
sc
=101.325 kPa
D
=0.3327 T
sc
=288.7K
D1
=0.3691
03457 . 0 = q

The presented semi-analytical model can recover the shape factors reported in the literature
for slightly compressible fluids. Table 4.3 shows the stabilized values of the shape factor
based on this study and the literature models (Lim and Aziz, 1995; Hassanzadeh and
Pooladi-Darvish, 2006) for different boundary conditions and different sets of fractures
(2D and 3D). As illustrated in this table there is an acceptable accuracy between the
stabilized values of the shape factor based on the presented semi-analytical model (
1
1
= = =
D D
q q | ) and the literature models.
116
Table 4.3: Stabilized values of the shape factor based on this study and literature models.
Pressure depletion regime Hassanzadeh and Pooladi-Darvish
(2006)
Lim and Aziz
(1995)
This Study
2D Flow (cylindrical approximation)
Constant Fracture Pressure 18.2 18.17 18.13
Linear decline 25.5
-
24.86
Exponential Decline (small exponent) 25.13
-
24.84
Exponential Decline (large exponent) 18.2
-
18.13
3D Flow (spherical approximation)

Constant Fracture Pressure 25.65 25.67 25.73
Linear decline 39
-
38.21
Exponential Decline (small exponent) 39
-
37.04
Exponential Decline (large exponent) 25.65
-
25.73

4.4 Results
In the following the effect of fracture pressure depletion regimes, block size distribution
and geometry are studied, respectively.
4.4.1 Effect of fracture pressure depletion regime
Three different depletion regimes in the fracture are considered including constant fracture
pressure, linear and exponential decline. Results of this study show that the fracture
pressure depletion regime will affect the rate of release. In fast depletion regimes like
constant fracture and large exponent exponential decline the matrix depletes more rapidly
than that in the case of the linear decline and the small exponent exponential decline.
Figure 4.4 compares the dimensionless release rate versus the dimensionless time for
different depletion regimes in the fracture for a cylindrical block. In the depletion regime
of exponential decline with a small exponent, the dimensionless gas release rate is
proportional to the square root of the time at the early time. At the middle time, the rate
stabilizes to a constant value. The rate eventually drops to zero at the late time. The gas
release rate for linearly declining fracture pressure also has the same behaviour as the
exponentially declining fracture pressure with a small exponent at the early and middle
117
times. For the linear decline, since the decline time is limited to t
D
<1/k, the pressure
decline in the fracture is not complete and the release rate does not fall to zero.
For the linear decline the rate of depletion of the matrix increases as the decline rate (k)
increases. As illustrated in Figure 4.4 for the exponential decline, as the value of the
exponent increases, the early time release rate increases and the blocks are depleted more
quickly. The exponential decline with a large exponent and constant fracture depletion
regimes behave in the same way and the block is depleted faster than other depletion
regimes. For the fast depletion regimes (constant fracture pressure and exponential decline
with a large exponent) the early time dimensionless rate varies inversely proportionally to
the square root of the time. A similar observation has been reported for the shape factor in
the case of slightly compressible and compressible fluids (Hassanzadeh and Pooladi-
Darvish, 2006; Ranjbar et al., 2011).
It should be mentioned that due to the approximate nature of the presented solution there is
a discontinuity in the flux especially for the exponential decline with a large exponent. This
discontinuity occurs at t
*
where the early and late time solutions coincide. This
discontinuity is mainly due to the change in the slope of the pressure (at t
*
the early and
late time pressures are equal but there is a little difference in their derivatives because of
the approximate nature of the solution). Similar behaviour is reported by Zimmerman et
al., (1990) using the integral approximate solutions.
118

Figure 4.4: Dimensionless rate versus dimensionless time for different fracture depletion regimes for a
cylindrical block.

Figures 4.5 and 4.6 compare the dimensionless cumulative gas release for different fracture
pressure depletion regimes for cylindrical and spherical blocks, respectively. The results
show that the time required for the cumulative gas release from a matrix block to reach its
plateau depends on the depletion regime in the fracture. Fast depletion regimes like
constant fracture pressure and exponential decline with a large exponent reach their plateau
more rapidly than those with the linear decline and the exponential decline with a small
exponent. On the other hand the exponential decline with a small exponent and linear
decline demonstrate a prolonged release period as compared to the other declines.
t
D
10
-3
10
-2
10
-1
10
0
10
1
10
2
10
3
10
4
10
5
q
D
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
1
10
2
Constant fracture pressure
Linear Decline(k=0.0001)
Exponetial Decline (k=0.0001)
Exponetial Decline (k=1)
119

Figure 4.5: Dimensionless cumulative release versus dimensionless time for different fracture depletion
regimes for a cylindrical block.

Figure 4.6: Dimensionless cumulative release versus dimensionless time for different fracture depletion
regimes for a spherical block.
t
D
10
-3
10
-2
10
-1
10
0
10
1
10
2
10
3
10
4
10
5
10
6
Q
D
10
-9
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
1
t
D
10
-3
10
-2
10
-1
10
0
10
1
10
2
10
3
10
4
10
5
10
6
Q
D
10
-9
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
1
120
4.4.2 Block size distribution effect
In this section the effect of different block size distribution on the cumulative gas release is
investigated. Table 4.4 illustrates the values of the equivalent radius for different block size
distribution. Equations (4.66) and (4.67) and Table 4.1 are used to determine the values of
the equivalent radius for different distributions. More details about the distribution and the
different values in each distribution are discussed elsewhere (Ranjbar et al., 2012). It
should be mentioned that the presented model can also be used for discrete block size
distribution as used in our previous study (Ranjbar et al., 2012).

Table 4.4: Values of dimensionless equivalent radius for different matrix block size distributions.
Block Size Distribution Dimensionless Equivalent Radius (R
De
)
Ideal distribution 1.000
Exponential distribution (a=-20) 0.950
Linear increasing distribution 0.625
Normal distribution 0.548
Linear decreasing distribution 0.475
Log-normal distribution 0.414
Exponential distribution (a=20) 0.150

Figures 4.7 and 4.8 compare the cumulative release for different distributions with
cylindrical and spherical blocks, respectively. As demonstrated in these figures the time to
reach the ultimate cumulative release is proportional to the equivalent radius of the
distribution. Based on these figures the exponential distribution with a large exponent is
depleted faster than the other distributions. The ideal distribution and exponential
distribution with a small exponent behave similarly and deplete more gradually than other
distributions.
121

Figure 4.7: Dimensionless cumulative release versus dimensionless time for different block size
distribution and cylindrical blocks.

Figure 4.8: Dimensionless cumulative release versus dimensionless time for different block size
distribution and spherical blocks
t
D
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Q
D
0
2
4
6
8
10
12
14
16
Ideal Dis.
Exponential Dis.(a=-20)
Linear Increasing Dis.
Normal Dis.
Linear Decreasing Dis.
Log-Normal Dis.
Exponential Dis.(a=20)
t
D
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Q
D
0
2
4
6
8
10
Ideal Dis.
Exponential Dis.(a=-20)
Linear Increasing Dis.
Normal Dis.
Linear Decreasing Dis.
Log-Normal Dis.
Exponential Dis.(a=20)
122
4.4.3 Comparison of different block geometries
In this section we compare the cumulative release for a slab-shaped matrix block with the
half thickness of , a cylinder and a sphere with the radius of . For the slab-shaped block
the following equation is derived to calculate the cumulative release:
D
D
D
Q
q
1
= .
(4.81)
In our previous study (Ranjbar and Hassanzadeh, 2011) the following equations were
derived for early and late time average pseudo-pressure in a slab-shaped matrix block:
1
1
24
1
,
4
24
D
D
D D
D
t
t
|q
|q
< = , (4.82)
1
181 . 32 486 . 2
24
1
, 148 . 0 790 . 0 1
1 1
D
D
t t
D
t e e
D D D D
|q
|q |q
> =

.
(4.83)
Equations (4.82) and (4.83) are substituted into Equation (4.81) to determine the
cumulative fluid release. For the cylindrical and the spherical block Equation (4.38) (I=1
for cylinder and I=2 for sphere) is derived to determine the cumulative release.
Comparison is performed for different matrix block geometries with the same
characteristic length available for release of fluid. Figure 4.9 compares the dimensionless
cumulative release for different matrix block geometries when the fracture pressure is
constant. As illustrated in this figure the slab-shaped block has the maximum value of the
cumulative release and the blocks are depleted more slowly than the cylindrical and
spherical blocks. The spherical block is depleted faster with the smallest value of the final
cumulative release. It should be mentioned that using the same characteristic length
available for release for different geometries results in blocks with different volumes to the
surface area ratio (i.e., (V/A)
slab
=2(V/A)
cylinder
=3(V/A)
sphere
). Therefore, the difference
observed in the ultimate cumulative release shown in Figure 4.9 can be explained based on
the ratio of the block volume to the block surface area for different geometries with the
same characteristic length. It should be noted that in Figure 4.9 the dimensionless time is
scaled based on characteristic length (L
c
for slab and R
m
for cylinder and sphere).
123
Therefore, different geometries should have the same characteristic length, which results in
different ratios of V/A.
Figure 4.9 shows that the stabilized values of the dimensionless cumulative release are
3.00, 1.50 and 1.00 for the slab-shaped, the cylindrical and the spherical blocks,
respectively. These values can be explained based on the ratio of the volume to the surface
area of the blocks for different geometries. The ratio of the volume to the surface area for
the slab-shaped block is , for the cylindrical block it is /2 and for the spherical block it is
/3 (Zimmerman et al., 1990). Therefore, the ultimate cumulative release is proportional to
the ratio of the volume to the surface area for different geometries (Assuming the same
characteristic length, , for different geometries). For example, the ratio of (V/A) of the
slab to the ratio of (V/A) of the cylinder is two and the ultimate cumulative release for the
slab is two times greater than that for the cylinder.

Figure 4.9: Dimensionless cumulative release versus dimensionless time for different block geometries

To normalize the cumulative release for three different geometries we scale the
dimensionless time by
2
) / /( A V t q t = and express the normalized cumulative release
(Q(t
D
)/Q
Dt
) versus the square root of the scaled time () (Zimmerman et al., 1990) where
t
D
0 1 2 3 4 5 6 7 8
Q
D
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Slab-Shaped block
Cylindrical block
Spherical block
124
Q
Dt
is the ultimate cumulative release for any geometry. As an example, the time scale for
a sphere of radius is defined as follows:
D
t
t t
9 9
) 3 / (
2 2
= = =
o
q
o
q
t .
(4.84)
Using the same approach we obtain
D
t 4 = t and
D
t = t for the cylindrical and slab-shaped
matrix blocks, respectively. Figure 4.10 demonstrates the normalized cumulative release
versus the square root of the scaled time. As a result of the scaling, the release curves
become closer to each other for different geometries as illustrated in Figure 4.10. A similar
observation has been made by Zimmerman et al., (1990) for absorption curves for different
geometries. This scaling law may find applications for irregular shaped blocks, which is
beyond the scope of this study.

Figure 4.10: Normalized cumulative fluid release versus square root of scaled time

SQRT[t|
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Q
D
/
Q
D
t
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Slab-Shaped block
Cylindrical block
Spherical block
125
4.5 Conclusions
An integral approximation method has been used to derive the solutions for nonlinear
pressure diffusion in blocks with different geometries including cylindrical and spherical
blocks. The presented solutions have considered the effect of fracture pressure depletion
regimes and the variable block size distributions or multiple blocks. The results calculated
by the approximate solutions are in good agreement with those calculated by the fine grid
numerical models.
It has been shown that the depletion time of a matrix block is a function of the fracture
pressure depletion regimes. In the case of constant fracture pressure or exponential decline
with a large exponent the block is depleted faster than that in the linear decline and the
exponentially decline with a small exponent. For the linear decline and exponential decline
with a small exponent the early time dimensionless release rate increases proportionally to
the square root of the dimensionless time, then stabilizes at a constant rate and finally falls
to zero.
Block size distribution is another important parameter that affects the matrix production
during the transient state. The blocks with a smaller equivalent radius are depleted more
quickly. For a large equivalent radius, distributions like ideal and exponential with a small
exponent (i.e., a=-20) the transient period is longer than that for the other distributions.
Finally, the normalized cumulative release from all matrix block geometries is expressed as
a function of the square root of the scaled dimensionless time, which may find application
for irregular shape matrix blocks.

126
Nomenclature
a Dimensionless exponential distribution constant
2
]
A(t
D
) First coefficient of the trial solution
b Intercept for linear matrix block size distribution
B(t
D
) Second coefficient of the trial solution
C(t
D
) Third coefficient of the trial solution
c
m
Matrix compressibility [LT
2
/M]
D(t
D
) Fourth coefficient of the trial solution
f
i
(R
mi
) Fraction of the block volume of size R
mi

f(R
m
) Probability density function
f
D
(R
D
) Dimensionless probability density function
F
h
Ratio of minimum block size to the maximum block radius
h
m
=2L
c
Matrix block thickness [L]
k
m
2
]
M Mean of the distribution
m Slope of linear matrix block size distribution
N Number of matrix block sizes
N
t
Total number of matrix blocks
p pressure [M/LT
2
]
q Matrix-fracture fluid release [L
3
/T]
Q Cumulative fluid release [L
3
]
Q
Dt
Dimensionless ultimate cumulative release
R
m
Matrix block radius [L]
R
me
Equivalent matrix block radius [L]
R Residual in the method of moments
S Heat conduction shape factor [L]
t Time [T]
t
*
Time which the effect of pressure reaches to the inner matrix boundary
V
b
Matrix block volume [L
3
]
r
D
Dimensionless radius
Greek Symbols
Radius of cylinder, sphere or half thickness of slab [L]
Correction factor
Penetration depth
127
Un-penetrated depth for spherical block
2
/T]
q
Average hydraulic diffusivity [L
2
/T]
D1
Dimensionless exponent and slope in fracture depletion regimes
Diffusion shape factor [1/L
2
]

2
Variance of the distribution
Dimensionless scale time
| Porosity
Ratio of reservoir radius to the maximum block radius
pseudo-pressure [M/LT
3
]
Dimensionless exponent of solution of gas diffusivity equation using the moment

Subscripts

D Dimensionless
e Equivalent
f Fracture
g Gas
i Initial condition
ln Log-normal
m Matrix
min Minimum
max Maximum
R Reservoir
sc Standard conditions

128
References
Altevogt, A.S. Celia, M.A. (2004). Numerical modeling of carbon dioxide in unsaturated
soils due to deep subsurface leakage.Water Resour. Res., 40, W03509.
Ames, W.F. (1965), Nonlinear partial differential equation in engineering. New York,
Academic Press.
Barker, J.A. (1985). Block-geometry functions characterizing transport in densely fissured
media. J. Hydrol., 77, 263-279.
Berger, D. Braester, C. (2000). Gas-water displacement through fracture networks, Water
Resour. Res., 36(11), 3205-3210.
Bogdanov, I.I. Mourzenko, V.V. Thovert, J.-F. Adler, P.M. (2003). Pressure drawdown
well tests in fractured porous media. Water Resour. Res., 39(1), 1021.
Chen, C. Zhang, D. (2010). Pore-scale simulation of density-driven convection in fractured
porous media during geological CO
2
sequestration, Water Resour. Res., 46, W11527,
doi: 10.1029/2010WR009453.
Civan, F. Rasmussen, M.L. (2002). Analytical hindered-matrix-fracture transfer models for
naturally fractured petroleum reservoirs, Paper SPE 74364.
Falta, R.W. (1995). Analytical solutions for gas flow due to gas injection and extraction
from horizontal wells, Ground Water, 33(2), 235-246
Fan, S.S.T. Yen, Y.C. (1968). Nonsteady compressible flow through anisotropic porous
mediums with particular reference to snow, Water Resour. Res., 4(3), 597-606.
Fitzgerald, S.D. Woods, A.W. (1998). Instabilities during liquid migration into superheated
geothermal reservoirs, Water Resour. Res., 34(9), 2089-2101.
Gerke, H.H. van Genuchten, M.Th. (1993). A dual-porosity model for simulating the
preferential movement of water and solutes in structured porous media, Water Resour.
Res., 29(2), 305-319.
Gwo, J.P. OBrien, R. Jardin, P.M. (1998). Mass transfer in structured porous media:
embedding mesoscale structure and microscale hydrodynamics in a two-region model, J.
Hydrol., 208, 204-222.
129
Hassanzadeh, H. Pooladi-Darvish, M. (2006). Effect of fracture boundary conditions on
matrix-fracture transfer shape factor, Transp. Porous Med., 64, 51-71.
solution for well testing of dual-porosity systems, Adv. Water Res., 32(11), 16521663.
Holman, J.P. (2010). Heat Transfer. Mc Graw Hill Press, Tenth Edition.
Hoteit, H. Firoozabadi, A. (2005). Multicomponent fluid flow by discontinuous Galerkin
and mixed methods in unfractured and fractured media, Water Resour. Res., 41,
W11412.
Huang, K. Tsang, Y.W. Bodvarsson, G.S. (1999). Simultaneous inversion of air-injection
tests in fractured unsaturated tuff at Yucca Mountain, Water Resour. Res., 35(8), 2375-
2386.
Illman, W.A. (2005). Type curve analyses of pneumatic single-hole tests in unsaturated
fractured tuff: Direct evidence for a porosity scale effect, Water Resour. Res., 41,
W04018.
Illman, W.A. Neuman, S.P. (2001). Type curve interpretation of a cross-hole pneumatic
injection test in unsaturated fractured tuff, Water Resour. Res., 37(3), 583-603.
Porous Med., 68,153-173.
McWhorter, D.B. (1990). Unsteady radial flow of gas in the vadose zone, J. Contam.
Hydrol., 5, 297-314.
Mitchel, S.L. Myers, T.G. (2010). Improving the accuracy of heat balance integral methods
applied to thermal problems with time dependent boundary conditions, Int. J. Heat Mass
Transfer, 53, 3540-3551.
Mora, C.A. Wattenbarger, R.A. (2009). Analysis and verification of dual-porosity and
CBM shape factors, J. Can. Pet. Technol., 48(2), 17-21.
130
Mourzenko, V.V. Bogdanov, I.I. Thovert, J.F. Adler, P.M. (2011). Three-dimensional
numerical simulation of single-phase transient compressible flows and well-tests in
fractured formations. Math. Comput. Simul., 81, 22702281.
Novakowski, K.S. Lapcevic, P.A. (1994). Field measurement of radial solute transport in
fractured rock, Water Resour. Res., 30(1), 37-44.
Nuske, P. Faigle, B. Helmig, R. Niessner, J. Neuweiler, I. (2010). Modeling gas-water
processes in fractures with fracture flow properties obtained through upscaling, Water
Resour. Res., 46, W09538, doi: 10.1029/2009WR008076.
Parker, L. Yarwood, R. Selker, J. (2006). Observations of gas flow in porous media using a
light transmission technique, Water Resour. Res., 42, W05501, doi:
10.1029/2005WR004080.
for interporosity flow in naturally fractured gas-condensate reservoirs. SPE Paper 75524.
Pooladi-Darvish, M. Torkite, W.S. Farouq Ali, S.M. (1994). Steam heating of fractured
formations containing heavy oil: basic premises and a single-block analytical model,
SPE Paper 28642.
Pruess, K. (1983). Heat transfer in fractured geothermal reservoirs with boiling, Water
Resour. Res., 19(1), 201-208.
flow of a compressible fluid in dual-porosity media. Adv. Water Res., 34(5), 627-639.
Ranjbar, E. Hassanzadeh, H. Chen, Z. (2011). Effect of fracture pressure depletion regimes
on the dual-porosity shape factor for flow of compressible fluids in fractured porous
media, Adv. Water Res., 34(12), 1681-1693.
Ranjbar, E. Hassanzadeh, H. Chen, Z. (2012). One-Dimensional matrix-fracture transfer in
dual porosity systems with variable block size distribution. Transp. Porous Med., 95(1),
185-212.
Schrauf, T.W. Evans, D.D. (1986). Laboratory studies of gas flow through a single natural
fracture. Water Resour. Res., 22(7), 1038-1050
Shan, C. (1995). Analytical solutions for determining vertical air permeability in
unsaturated soils. Water Resour. Res., 31(9), 2193-2200.
131
Shan, C. (2006). An analytical solutions for transient gas flow in a multiwell system. Water
Resour. Res., 42, W10401, doi: 10.1029/2005WR004737.
Shan, C. Javandel, I. and P. A. Whiterspoon (1999), Characterization of leaky faults: study
of air flow in faulted vadose zones, Water Resour. Res., 35(7), 2007-2013.
Tartakovsky, D.M. Neuman, S.P. Lu, Z. (1999). Conditional stochastic averaging of steady
state unsaturated flow by means of Kirchhoff Transformation. Water Resour. Res.,
35(3), 731-745.
Thunvik, R. Braester, C. (1990). Gas migration in discrete fracture networks, Water
Resour. Res., 26(10), 2425-2434.
van Genuchten, M.Th. Dalton, F.N. (1986). Models for simulating salt movement in
aggregated field soils, Geodrama, 38, 165-183.
van Heel, A.P.G., van Dorp, J.J. Boerrigter, P.M. (2008). Heavy oil recovery by steam
injection in fractured reservoirs, SPE Paper 113461.
Wang, Y. Dusseault, M.B. (1991). The effect of quadratic gradient terms on the borehole
solution in poroelastic media, Water Resour. Res., 27(12), 3215-3223.
Eng. J., 3(3), 245-255.
Wuthicharn, K. Zimmerman, R.W. (2011). Shape factors for irregularly matrix blocks.
SPE Paper 148060.
Ye, P. Ayala, L.F. (2012). A density-diffusivity approach for the unsteady state analysis of
natural gas reservoirs. J. Nat. Gas Sci. Eng, 7, 22-34.
You, K. Zhan, H. Li, J. (2011). Analysis of models for induced gas flow in the unsaturated
zone. Water Resour. Res., 47, W04515, doi: 10.1029/2010WR009985.
Zimmerman, R.W. Bodvarsson, G.S. (1989). Integral method solution for diffusion into a
spherical block. J. Hydrol., 111, 213-224.
Zimmerman, R.W. Bodvarsson, G.S. Kwicklis, E.M. (1990). Absorption of water into
porous blocks of various shapes and sizes. Water Resour. Res., 26(11), 2797 2806.
Res., 29(7), 21272137.
132
Zimmerman, R.W. Bodvarsson, G.S. (1995). Effective block size for imbibitions or
absorption in dual-porosity media. Geophys. Res. Lett., 22(11), 1461-1464.
for flow in unsaturated dual-porosity media. Adv. Water Res., 19(5), 317327.
133
Appendix 4.A: Analytical solution for cylindrical blocks
In this Appendix the solution of nonlinear gas pressure diffusion in a cylindrical matrix
block is discussed in more details.
4.A1: Constant fracture pressure
In this case as mentioned in the text the diffusivity equation and its conditions are based on
Equations (4.24) (with I=1), (4.25), (4.26) and (4.27).
The integral method is used to find the early time solution by defining the time-dependant
penetration depth, (t
D
), in which the pressure disturbance has reached that depth. For the
early time solution we have the following boundary conditions:
1 1 = =
D D
r , (4.A1.1)
. 0 , 0 ) ( 1 =
c
c
= =
D
D
D D D
r
t r

o
(4.A1.2)
Since the exact solution for the cylindrical block is in the form of first kind, the order zero
of Bessels function (J
0
(r
D
)) we suggest the following polynomial trial solution in the
integral method (Pooladi-Darvish et al., 1994):
4 2
) ( ) ( ) (
D D D D D D
r t C r t B t A + = . (4.A1.3)
Using actual and auxiliary boundary conditions in the trial solution leads to the following
equation:
2 2
2 2 2
] ) 1 ( 1 [
] ) 1 ( [
o
o

=
D
D
r
. (4.A1.4)
Using this trial solution in the integral form of the diffusivity equation (Equation (4.24),
with I=1) leads to the following solution for early time pseudo-pressure:
1 1
2
1
2
48
1
,
48
] 48 1 [
D
D
D D
D D D
D
t
t
t r
|q |q
|q
<
+
= . (4.A1.5)
It should be noted that
D1
is the dimensionless hydraulic diffusivity of the fracture at the
outer boundary, which is defined as follows:
q q
q
q q q
f f
D
D D D D
c
k x
x
1 ) 1 (
1 @
1 1
=
=
= = =
(4.A1.6)
134
Integrating over the bulk volume of the cylindrical block is used to find the average
dimensionless pseudo-pressure as follows:
} }
= =
1
0
2
1
D D D
V
D D
dr r dV
V

(4.A1.7)
Substituting Equation (4.A1.5) into Equation (4.A1.7) leads to the following equation for
the early time average dimensionless pseudo-pressure for the cylindrical block in the case
of constant fracture pressure:
1
1
1
1
48
1
,
3
48
2
D
D
D D
D D D D
t
t
dr r
|q
|q

o
< = =
}
.
(4.A1.8)
Equation (4.A1.8) is the same as Equation (4.28) in the main text.
The late time solution of the nonlinear PDE using the method of moments is explained in
more details in this part. For the late time solution the following equation is used as the
initial condition, which comes from the early time solution:
4
1
48
1
D D
D
D
r t = =
|q
.
(4.A1.9)
The inner and outer boundary conditions and the diffusivity equation are the same as
Equations (4.26), (4.27) and (4.24) (with I=1), respectively. The method of moments is
used to find the late time solution of this PDE by suggesting the following trial solution
and the residual (R) as follows:
1
4 2
48
1
, ) ( ) ( ) ( ) ( ) , (
D
D D D D D D D D D D D
t r t D r t C r t B t A t r
|q
> + + + = ,
(4.A1.10)
) (
D
D
D D
D D
D
D
r
r
r t
r R
c
c
c
c
c
c
=

|q
.
(4.A1.11)
The unknown coefficients (A, B, C and D) in Equation (4.A1.10) are found by using the
boundary conditions and making the zero and first moments of R vanish by enforcing the
following conditions:
} }
=
c
c
c
c
c
c
=
1
0
1
0
0 )) ( ( 0
D
D
D
D D
D D
D
D D
dr
r
r
r t
r Rdr

|q
,
(4.A1.12)
135
} }
=
(
c
c
c
c
c
c
=
1
0
1
0
0 ) ( 0
D
D
D
D D
D D
D
D D D D
dr
r
r
r t
r r Rdr r

|q
.
(4.A1.13)
From the inner boundary conditions, we can conclude that B = 0; the outer boundary
condition (Equation (4.27)) leads to:
D C A =1 . (4.A1.14)
Solving Equations (4.A1.12) and (4.A1.13), combining the results with Equation (4.A1.14)
and some simplification lead to a system of ordinary differential equations as follows:
D C
t d
C d
D D
D
1 1
96 20 |q |q + = ,
(4.A1.15)
D C
t d
dD
D D
D
1 1
84 21 |q |q = .
(4.A1.16)
Solving the system of the ordinary differential equations leads to the following equations
for unknown coefficients C and D:
) exp( 227 . 1 ) exp( 725 . 3
2 2 1 1 D D
t m t m C e e = , (4.A1.17)
) exp( ) exp(
2 2 1 1 D D
t m t m D e e + = , (4.A1.18)
where
1
and
2
are the eigenvalues of the system of the ordinary differential equations
and are defined based on Equation (4.30) in the main text. So we have the following
equation for the late time dimensionless pseudo-pressure:
,
48
1
, ) ( ) 227 . 1
725 . 3 ( ) 227 . 0 725 . 2 1 ( ) , (
1
4
2 1
2
2
1 2 1
2 1 2
1 2 1
D
D D
t t
D
t
t t t
D D D
t r e m e m r e m
e m e m e m t r
D D D
D D D
|q
e e e
e e e
> + +
+ + + =

(4.A1.19)
where m
1
and m
2
are obtained by using the initial condition of Equation (4.A1.9), and the
trial solution of the nonlinear PDE for the late time behaviour is obtained as in Equation
(4.A1.20). After that, Equation (4.A1.20) is integrated (Equation (4.A1.7)) over the matrix
block volume to obtain the average matrix block pseudo-pressure (Equation (4.29) in the
main text).
.
48
1
, ) 022 . 5 555 . 0 ( ) 162 . 6
067 . 2 ( ) 14 . 1 512 . 1 1 ( ) , (
1
4 2
2 1 2
1 2 1
D
D D
t t
D
t
t t t
D D D
t r e e r e
e e e t r
D D D
D D D
|q
e e e
e e e
> + +
+ + =

(4.A1.20)
136
4.A2: Variable fracture pressure
For the linearly declining fracture pressure the outer boundary condition (fracture pressure)
varies linearly with the time based on the following equation:
t
t R r
i f m m
1
), 1 ( s = = = o o . (4.A2.1)
In this case the dimensionless pseudo-pressure and fracture dimensionless pseudo-pressure
i
m i
D

= ,
(4.A2.2)
D
D fD
t
t
1
, s = k k .
(4.A2.3)
For the linearly declining fracture pressure the diffusivity equation and its initial and inner
boundary conditions are the same as Equations (4.24) (with I=1), (4.25) and (4.26),
respectively. Equation (4.A2.3) is used as the outer boundary condition. In this case the
shape factor equation has the following form for a cylindrical block:

|
|
|
|
.
|
\
|
c
c
=
D D
D
D
D
m
t
t
R
k
q
o
1
2
.
(4.A2.4)
For the early time solution of the linear decline of a cylindrical block we assume that the
trial solution has the following form:
2
2
2
)) ( 1 ( 1
1
1 ) , (
(

=
D
D
D D D D
t
r
t t r
o
k . (4.A2.5)
In Equation (4.A2.5) the terms in the bracket is the solution of the constant fracture
pressure case. In the case of variable fracture pseudo-pressure the penetration depth
(numerator in Equation (4.A2.5)) is found by solving the following ordinary differential
equation (Mitchel and Myers, 2010; Ranjbar et al., 2011):
{ }
{ }
{ }
2 2
2
2
)) ( 1 ( 1
) (
) 1 (
)) ( 1 ( 1
1
)) ( 1 ( 1 ) (
D
D fD
D
D D fD
D
t
t n
n
t
n
t t
dt
d
o
u
u o
o

+
=
(
(
+

.
(4.A2.6)
137
In this equation, n is the trial solution exponent (n=4 for the cylindrical block) and is
obtained based on the following equation:
{ }
1 2
) 1 ( )) ( 1 ( 1
2
2

c
c
=
n
n n t
t
fD D
D
fD
o
u .
(4.A2.7)
To find a semi-analytical solution for the penetration depth we assume that =0 in
Equation (4.A2.6) as assumed by Mitchel and Myers (2011). For solving Equation (4.A2.6)
we use the following substitution:
D D
t z t o = =
2
)) ( 1 ( 1 . (4.A2.8)
Solving Equation (4.A2.6) for the linear decline leads to the following equation for the
penetration depth:
3
40
1 1 ) (
1 D D
D
t
t
|q
o = . (4.A2.9)
It should be mentioned that a more accurate solution can be obtained if the following
equation is used in the trial solution in the case of the linearly declining fracture pressure
for the cylindrical matrix block:
D D D
t t
1
16 1 ) ( 1 |q o = . (4.A2.10)
In the developed solution, Equation (4.A2.10) is used for the time dependent penetration
depth. It should be noted that this solution is valid till 1-(t
D
)=0. Substituting the
penetration depth equation in the dimensionless pseudo-pressure and integrating over the
bulk volume of the matrix block leads to the following equations for the early time
dimensionless pseudo-pressure and the average dimensionless pseudo-pressure,
respectively:
1 1
2
1
2
16
1
,
16
] 16 1 [
D
D
D D
D D D
D D
t
t
t r
t
|q |q
|q
k <
+
= , (4.A2.11)
1
1
16
1
,
3
16
D
D
D D D
D
t
t t
|q
|q k
< = . (4.A2.12)
Duhamels theorem is used to find the late time solution of the diffusivity equation when
the fracture boundary condition changes with time as follows:
138
1 0
16
1
, ) , ( ) (
D
D
t
D D D fD
D
D
t d t r
t
D
|q
t t t >
c
c
=
}
. (4.A2.13)
In this equation,
D
fD
(Equation (4.A1.19))
and
D
on the left-hand side is the solution of the PDE when the matrix-fracture boundary
condition changes with time. Substituting Equations (4.A1.19) and (4.A2.3) into
Duhamels equation and using the initial condition (at
1
16
1
D
D
t
|q
= we have
4
1
16
D
D
D
r
|q
k
= ) leads to the following late time solution for the dimensionless pseudo-
pressure for the linearly declining fracture pressure and the cylindrical matrix block:
1
4 2
2
4 2
1
16
1
), 1 )( 227 . 1 227 . 0 (
57480 . 5
) 1 )( 725 . 3 725 . 2 (
58526 . 0
) , (
2
1
D
D
t
D D
t
D D D D D D
t e r r
e r r t t r
D
D
|q e
k
e
k
k
e
e
> +
+ + =
.
(4.A2.14)
Integrating this equation over the bulk volume of the cylindrical matrix block with linearly
declining fracture pressure (Equation (4.A1.7)) leads to the following equation:
1 2 1
16
1
), 1 (
29639 . 0
) 1 (
69987 . 0
2 1
D
D
t t
D D
t e e t
D D
|q e
k
e
k
k
e e
> = .
(4.A2.15)
For exponentially declining fracture pressure we have the same PDE with the same initial
and inner boundary conditions with the following outer boundary condition:
) exp( ) ( ) ( t t R r
i f m m
o + = = =

.
(4.A2.16)
In this equation,
is the fracture pseudo-pressure when time tends to the infinity. In this

case the dimensionless pseudo-pressure and the dimensionless fracture pseudo-pressure are
defined as follows:
i
i m
D

,
(4.A2.17)
) exp( 1
D fD
t k = .
(4.A2.18)
In this case the diffusivity equation and its initial and boundary conditions are the same as
Equations (4.24) (with I=1), (4.25) and (4.26), respectively, with Equation (4.A2.18) as the
139
outer boundary condition. The dimensionless shape factor equation has the following form
for a cylindrical block and exponentially declining fracture pressure:

|
|
|
|
.
|
\
|

c
c
=
)) exp( 1 (
1
2
D D
D
D
D
m
t
t
R
k
q
o .
(4.A2.19)
For the early time solution of the exponential decline we assume that the trial solution has
the following form:
2
2
2
)) ( 1 ( 1
1
1 )) exp( 1 ( ) , (
(

=
D
D
D D D D
t
r
t t r
o
k . (4.A2.20)
Using Equations (4.A2.6) and (4.A2.8) and comparison with the literature model
(Hassanzadeh and Pooladi-Darvish, 2006) leads to the following equation for penetration
depth in the case of exponentially declining fracture pressure:
)
) (
1 1
2
( 24 1 ) ( 1
1
D
D
t t D D D
t
t erf
e e
t t
D D
k
k t
|q o
k k
= . (4.A2.21)
It should be noted that the effect of pressure disturbance will reach the inner boundary
when 1-(t
D
)=0 and for the exponential decline we cannot obtain an explicit equation for t
*

and it is determined for any values of k by making Equation (4.A2.21) equal to zero.
Therefore, the early time solution for the exponentially declining fracture pressure and
cylindrical matrix block can be expressed as follows:
*
1
2
1
2
,
) (
1 1
2
( 24
) (
1 1
2
( 24 1
) 1 ( ) , (
t t
t
t erf
e e
t
t
t erf
e e
t r
e t r
D
D
D
t t D D
D
D
t t D D D
t
D D D
D D
D D
D
<
(
(
k
k t
|q
k
k t
|q
k k
k k
k
.
(4.A2.22)
The initial condition for the late time solution comes from the early time solution
(Equation (4.A2.22)) as follows:
4 * *
)) exp( 1 (
D D D
r t t t k = = .
(4.A2.23)
140
The diffusivity equation and its boundary conditions are the same as before (Equations
(4.24), (with I=1), (4.26) and (4.27)). Using Duhamels theorem (Equation (4.A2.13)) and
the solution of the constant fracture pressure (Equation (4.A1.19)) leads to the following
late time solution for the case of the exponentially declining fracture pressure:
* 4
2
2
1
1
2
2
2
1
1
1
2
2
2
1
1
2
2
2
1
1
1
2
2
1
1
2
2
2
1
1
1
, ) ( ) 1 ( ) 1 (
)
227 . 1 725 . 3
( ) 1 ( 227 . 1 ) 1 ( 725 . 3
)
227 . 0 725 . 2
( ) 1 ( 227 . 0 ) 1 ( 725 . 2
1 ) , (
2 1
2 1
2 1
t t r e
m m
e m e m
r e
m m
e m e m
e
m m
e m e m
e t r
D D
D
t t t
D
C
t t t
t t t
t
D D D
D D D
D D D
D D D
D
>
+
+
+
+
+
+
+
+
+
+
+
+
)
`
+
+
+
+
+
+
+ =

k e e
k e e
k e e
k
e k
k
e k
k
e k
e
e k
e
e k
k
e k
k
e k
e
e k
e
e k
k
e k
k
e k
e
e k
e
(4.A2.24)
1
and m
2
as follows:
) exp( 1 , 0
* *
t D C t t
D
k = = = . (4.A2.25)
1
and m
2
:

\
|
+
=
+
=
* *
2
*
* *
1
*
1
49118 . 1
1
49119 . 0
2
2
1
1
t t
t
t t
t
e e
e
m
e e
e
m
k e
k
k e
k
k
e k
k
e k
.
(4.A2.26)
Using these values in Equation (4.A2.24) and simplifying lead to the following late time
pseudo-pressure for the exponentially declining fracture pressure:
,
) 227 . 1 227 . 0 )(
1
( 49118 . 1
) 725 . 3 725 . 2 )(
1
( 49119 . 0
) 227 . 1 227 . 0 ( )
1
( 49118 . 1
) 725 . 3 725 . 2 ( )
1
( 49119 . 0 1 ) , (
4 2
4 2
* 4 2
4 2
* *
2
*
* *
1
*
2
* *
2
*
1
* *
1
*
D
D
D D
t
D D
t t
t
D D
t t
t
D D D
t
t t
t
D D
t
t t
t
t
D D D
e
r r
e e
e
r r
e e
e
t t r r e
e e
e
r r e
e e
e
e t r
k
k e
k
k e
k
e
k e
k
e
k e
k
k
(
(
(
(
+
> +
+
+
=
.
(4.A2.27)
141
Integrating of the pseudo-pressure equations over the bulk volume of the cylindrical matrix
block (Equation (4.A1.7)) leads to the following equations for early and late time average
dimensionless pseudo-pressure when the fracture pressure declines exponentially with time
for the cylindrical matrix block:
*
1
, )
) (
2 ( 24
3
1
t t
t
t erf
t
e
D
D
D
D D
t
D
D
<
=

k
k t
|q
k
. (4.A2.28)
*
, )
1
( 07754 . 0 )
1
( 58746 . 0
)
1
( 07754 . 0 )
1
( 58746 . 0 1
* *
2
*
* *
1
*
2
* *
2
*
1
* *
1
*
t t e
e e
e
e e
e
e
e e
e
e
e e
e
e
D
t
t t
t
t t
t
t
t t
t
t
t t
t
t
D
D
D D D
>
(
(
k
k e
k
k e
k
e
k e
k
e
k e
k
k

(4.A2.29)
4.A3: Variable block size distributions
In this section the solution of PDE (Equation (4.62), with I=1) with its initial and boundary
conditions (Equations (4.63) to (4.65)) for variable block size distribution is discussed in
more details.
The integral method is used to derive the early time approximate solution of the diffusivity
equation for flow of compressible and slightly compressible fluids in the cylindrical matrix
block for different block size distributions. For the early time solution, in addition to the
diffusivity equation (Equation (4.62), with I=1) and outer boundary condition (Equation
(4.65)), we have the following condition at the radius where the pressure disturbance has
reached:
. 0 , 0 ) ( =
c
c
= =
D
D
D D De D
r
t R r

o
(4.A3.1)
Similar to the single matrix block we suggest a fourth-order polynomial trial solution
(Equation (4.A1.3)) to be used in the integral method. Using the actual and auxiliary
boundary conditions (Equations (4.65) and (4.A3.1)) in the trial solution leads to the
following equations for A, B and C coefficients:
2 2 2
4
] ) ( [
] [
o
o

=
De De
De
R R
R
A ,
(4.A3.2)
142
2 2 2
2
] ) ( [
] [ 2
o
o

=
De De
De
R R
R
B ,
(4.A3.3)
2 2 2
] ) ( [
1
o
=
De De
R R
C .
(4.A3.4)
Substituting these equations for time-dependent coefficients and some simplification leads
to the following equation:
2 2 2
2 2 2
] ) ( [
] ) ( [
o
o

=
De De
De D
D
R R
R r
.
(4.A3.4)
Using this trial solution (Equation (4.A3.4)) in the integral form of the diffusivity equation
(from R
De
- to R
De
) leads to the following ordinary differential equation (ODE):
| |
2 2
2
1 2 2
) (
4
) (
6
1
o
|q
o

=
De De
De D
De De
D
R R
R
R R
dt
d
.
(4.A3.5)
Solving this ODE leads to the following equation for penetration depth and early time
pseudo-pressure:
De D D De De De
R and t R R R < = o |q o
1
2
48 , (4.A3.6)
1
2
1
2
2
1
2 2
48
,
48
] 48 [
D
De
D
D D De
D D De De D
D
R
t
t R
t R R r
|q |q
|q
<
+
= . (4.A3.7)
The early time average dimensionless pseudo-pressure is obtained by integrating over the
bulk volume of the matrix block:
1
2
1
1
2
2
1
2 2
2
48
,
3
48
48
] 48 [ 2
D
De
D
De
D D
R
R
D
D D De
D D De De D
D
De
D
R
t
R
t
dr
t R
t R R r
r
R
De
De |q
|q
|q
|q
o
< =
+
=
}

(4.A3.8)
For the late time solution we have the same PDE (Equation (4.62), with I=1) with the same
boundary conditions (Equations (4.64) and (4.65)) and the following initial condition:
4
4
1
2
1
48
D
De
D
D
De
D
r
R
R
t = =
|q
.
(4.A3.9)
The method of moments is used to find the late time solution of this PDE by suggesting a
forth-order trial solution (Equation (4.A1.10)) and the residual (R) as Equation (4.A1.11).
The unknown coefficients (A, B, C and D) in Equation (4.A1.10) are found by using the
143
} }
=
c
c
c
c
c
c
=
De De
R
D
D
D
D D
D D
D
D
R
D
dr
r
r
r t
r Rdr
0 0
0 )) ( ( 0

|q
, (4.A3.10)
} }
=
(
c
c
c
c
c
c
=
De De
R R
D
D
D
D D
D D
D
D D D D
dr
r
r
r t
r r Rdx r
0 0
0 ) ( 0

|q
.
(4.A3.11)
From the inner boundary conditions, we can conclude that B = 0; the outer boundary
4 2
1
De De
DR CR A = . (4.A3.12)
Solving Equations (4.A3.10) and (4.A3.11), combining the results with Equation (4.A3.12)
and some simplification lead to a system of ODEs as follows:
D C
R t d
C d
D
De
D
D
1 2
1
96
20
|q
|q
+ = ,
(4.A3.13)
D
R
C
R t d
dD
De
D
De
D
D
2
1
4
1
84 21 |q |q

= .
(4.A3.14)
Solving the system of ordinary differential equations leads to the following equations for
the unknown coefficients C and D:
) exp( 227 . 1 ) exp( 725 . 3
2 2
2
1 1
2
D De D De
t m R t m R C e e = , (4.A3.15)
) exp( ) exp(
2 2 1 1 D D
t m t m D e e + = , (4.A3.16)
where
1
and
2
are the eigenvalues of the system of the ordinary differential equation and
are defined based on Equation (4.71) in the main text.
The initial condition (Equation (4.A3.9)) is used to find m
1
and m
2
. Therefore, m
1
and m
2

are found by solving the following system of equations:
0 )
48
exp( 227 . 1 )
48
exp( 725 . 3
1
2
2 2
2
1
2
1 1
2
=
D
De
De
D
De
De
R
m R
R
m R
|q
e
|q
e ,
(4.A3.17)
4
1
2
2 2
1
2
1 1
1
)
48
exp( )
48
exp(
De D
De
D
De
R
R
m
R
m = +
|q
e
|q
e .
(4.A3.18)
Solving this system of equations for m
1
and m
2
and substituting in the time-dependent
coefficients of the trial solution the late time behaviour is obtained as follows:
144
.
48
, )
022 . 5 555 . 0
( )
162 . 6
067 . 2
( ) 14 . 1 512 . 1 1 ( ) , (
1
2
4
4 4
2
2
2
2 1 2
1 2 1
D
De
D D
t
De
t
De
D
t
De
t
De
t t
D D D
R
t r e
R
e
R
r e
R
e
R
e e t r
D D D
D D D
|q
e e e
e e e
> +
+
+ + =

(4.A3.19)
After that, Equations (4.A3.19) is integrated over the matrix block volume to obtain the
late time average matrix block pseudo-pressure as follows:
}
> = =
De
D D
R
D
De
D
t t
D D D
De
D
R
t e e dr r
R
0 1
2
2
48
267 . 0 664 . 0 1
2
2 1
|q

e e
. (4.A3.20)
Appendix 4.B: Analytical solution for spherical blocks
In this appendix the solution of nonlinear gas pressure diffusion in a spherical matrix block
is discussed in more details.
4.B1: Constant fracture pressure
In this section the solution of PDE (Equation (4.24), with I=2) with the initial and
boundary conditions (Equations (4.25) to (4.27)) is presented.
The integral method (Zimmerman and Bodvarsson, 1989) is used to find the early time
solution of this equation by defining the time dependant penetration depth, (t
D
), in which
the pressure disturbance has reached. For the early time solution, in addition to the outer
boundary condition (Equation (4.27)), we have the following auxiliary boundary
conditions:
0 , 0 , 0 ) ( ) ( 1
2
2
=
c
c
=
c
c
= = =
D
D
D
D
D D D D
r r
t t r

c o .
(4.B1.1)
The following third-order polynomial trial solution is suggested to be used in the integral
method:
3 2
) ( ) ( ) ( ) (
D D D D D D D D
r t D r t C r t B t A + + + = . (4.B1.2)
Using the actual and auxiliary boundary conditions in the trial solution leads to the
following equation for dimensionless pseudo-pressure:
3
3
) 1 (
) (
c
c
=
D
D
r
. (4.B1.3)
145
Using this trial solution in the integral form of the diffusivity equation (Equation (4.24),
with I=2) leads to the following ODE:
] 10 6 3 [ ) 1 ( 180
]
) 1 ( 60
10 36 45 20
[
1
3
2 3
1
3
2 3 6
1
+ + =

+ +
=
c c c c |q
c
c c c c
c
|q
d dt
dt
d
D D
D
D
.
(4.B1.4)
Solving this ODE leads to the following equation:
1
3 4
720
17 24 4 3
D
D
t
|q
c c c + +
= .
(4.B1.5)
This solution is valid till =0 or
1
720
17
D
D
t
|q
< .
Equation (4.B1.3) is integrated over the bulk volume of the matrix block to determine the
average dimensionless pseudo-pressure as follows:
1
2 3 1
3
3
2
1
2
720
17
,
20
10 6 3
) 1 (
) (
3 3
D
D D
D
D D D D D
t dr
r
r dr r
|q
c c c
c
c

c c
<
+ +
=
= =
} }

(4.B1.6)
Equations (4.B1.5) and (4.B1.6) in terms of penetration depth can be expressed as
follows:
,
720
17
,
20
15 6
1
2 3
D
D D
t
|q
o o o
<
+
=
(4.B1.7)
.
720
30 16 3
1
2 3 4
D
D
t
|q
o o o +
=
(4.B1.8)
For the late time solution we have the following initial condition (the diffusivity equation
and boundary conditions are the same as Equations (4.24), (4.26) and (4.27), respectively):
3
1
720
17
D D
D
D
r t = =
|q
.
(4.B1.9)
third order trial solution as Equation (4.B1.2) and the residual (R) as follows:
D
D
D
D
D
D D D
D
t
r
r
r t
r
R
c
c
c
c
c
c
=

|q
2 2
) ) ( ( .
(4.B1.10)
146
The unknown coefficients (A, B, C and D) in Equation (4.B1.2) are found using the
} }
=
c
c
c
c
c
c
=
1
0
2 2
1
0
0 )) ( ( 0
D
D
D
D D
D D
D
D D
dr
r
r
r t
r Rdr

|q
,
(4.B1.11)
} }
=
(
c
c
c
c
c
c
=
1
0
1
0
2 2
0 ) ( 0
D
D
D
D D
D D
D
D D D D
dr
r
r
r t
r r Rdr r

|q
.
(4.B1.12)
D C A =1 . (4.B1.13)
Solving Equations (4.B1.11) and (4.B1.12), combining the results with Equation (4.B1.13)
and some simplification leads to a system of ODEs as follows:
D C
dt
dC
D D
D
1 1
198 90 |q |q + = ,
(4.B1.14)
D C
dt
dD
D D
D
1 1
4 . 176 84 |q |q = .
(4.B1.15)
Solving the system of ordinary differential equations leads to the following equations for
unknown coefficients C and D:
) exp( 189044 . 1 ) exp( 982384 . 1
2 2 1 1 D D
t m t m C e e = , (4.B1.16)
) exp( ) exp(
2 2 1 1 D D
t m t m D e e + = , (4.B1.17)
where
1
and
2
are the eigenvalues of the system of ODEs and are defined based on
Equation (4.34) in the main text. So we have the following equation for the late time
dimensionless pseudo-pressure:
.
720
17
, ) (
) 189044 . 1 982384 . 1 ( 189044 . 0 982384 . 0 1
1
3
2 1
2
2 1 2 1
2 1
2 1 2 1
D
D D
t t
D
t t t t
D
t r e m e m
r e m e m e m e m
D D
D D D D
|q
e e
e e e e
> + +
+ + + =

(4.B1.18)
The initial condition of Equation (4.B1.9) is used to find m
1
and m
2
. Therefore, the trial
solution of the nonlinear PDE for the late time behaviour is obtained as follows:
147
.
720
17
, ) 218938 . 15 892569 . 1 (
) 095987 . 18 751798 . 3 ( 877049 . 2 859229 . 1 1
1
3
2
2 1
2 1 2 1
D
D D
t t
D
t t t t
D
t r e e
r e e e e
D D
D D D D
|q
e e
e e e e
> +
+ + + =

(4.B1.19)
After that, Equations (4.B1.19) is integrated (Equation (4.B1.6)) over the matrix block
volume to obtain the average matrix block pseudo-pressure as follows:
1
720
17
, 371074 . 0 554435 . 0 1
2 1
D
D
t t
D
t e e
D D
|q
e e
> = .
(4.B1.20)
4.B2: Variable fracture pressure
In the case of linear decline for the spherical block the outer boundary condition is given in
Equation (4.A2.3). The PDE and initial and inner boundary conditions are the same as
Equations (4.24) (with I=2), (4.25) and (4.26), respectively. For the early time solution of
the linear decline we assume that the trial solution has the following form:
3
3
3
)
1
(
) 1 (
) (
o
o
k
c
c
k
+
=
=
D
D
D
D D
r
t
r
t . (4.B2.1)
In the case of variable fracture pseudo-pressure the penetration depth is found by solving
the following ODE (Mitchel and Myers, 2010; Ranjbar et al., 2011):
) (
) (
) 1 (
) (
1
) ( ) (
2
D
D fD
D
D D fD
D
t
t n
n
t
n
t t
dt
d
o
u
u o
o +
=
(
+
.
(4.B2.2)
Using the same procedure as described in Section 4.A2 for the cylindrical block and
comparing with the literature model leads to the following equation for early time
dimensionless pseudo-pressure:
1
3
1
1
13
1
, )
13
1 13
(
D
D
D D
D D D
D D
t
t
t r
t
|q |q
|q
k <
+
= .
(4.B2.3)
Integrating over the bulk volume of the matrix block leads to the following equation for
early time average dimensionless pseudo-pressure:
1
13
1
), 15 6 (
260
1
3 4 5
1
< < + = o
|q
o o o
|q
k
or t
D
D
D
D

(4.B2.4)
Duhamels theorem (Equation (4.A2.13)) is used to find the late time solution of the
diffusivity equation in fractured media when the fracture boundary condition changes with
148
time. Substituting Equations (4.B1.18) and (4.A2.3) in Duhamels equation and using the
initial condition (at
1
13
1
D
D
t
|q
= we have
3
1
13
D
D
D
r
|q
k
= ) leads to the following late time
solution for dimensionless pseudo-pressure for the linearly declining fracture pressure and
spherical matrix block:
3
2 1
1
2
2 1
2 1
)
) 1 ( 749672 . 14 ) 1 ( 139843 . 2
(
13
1
)
) 1 ( 538009 . 17 ) 1 ( 241990 . 4
(
) 1 ( 788337 . 2 ) 1 ( 102174 . 2
2 1
2 1
2 1
D
t t
D
D D
t t
t t
D D
r
e e
t r
e e
e e
t
D D
D D
D D
e
k
e
k
|q e
k
e
k
e
k
e
k
k
e e
e e
e e
+

> +
=
.
(4.B2.5)
Integration of this equation over the bulk volume of the matrix block (Equation (4.B1.6))
from zero to one leads to the following equation for the average dimensionless pseudo-
pressure of a spherical block when the fracture pressure declines linearly with time:
1 2 1
13
1
), 1 (
359632 . 0
) 1 (
626902 . 0
2 1
D
D
t t
D D
t e e t
D D
|q e
k
e
k
k
e e
> = .
(4.B2.6)
For exponentially declining fracture pressure the same approach is used and the following
equations are obtained for early and late time dimensionless pseudo-pressure:
*
3
1
1
,
)
) (
1 1
2
( 18
)
) (
1 1
2
( 18 1
) 1 ( ) , (
t t
t
t erf
e e
t
t
t erf
e e
t r
e t r
D
D
D
t t D D
D
D
t t D D D
t
D D D
D D
D D
D
<
(
(
(
(
(
k
k t
|q
k
k t
|q
k k
k k
k
.
(4.B2.7)
149
,
) 189044 . 1 189044 . 0 )(
1
( 498783 . 2
) 982384 . 1 982384 . 0 )(
1
( 498783 . 1
) 189044 . 1 189044 . 0 ( )
1
( 498783 . 2
) 982384 . 1 982384 . 0 ( )
1
( 498783 . 1 1 ) , (
3 2
3 2
* 3 2
3 2
*
2
*
*
*
1
*
*
2
*
2
*
*
1
*
1
*
*
D
D
D D
t
D D
t t
t
D D
t t
t
D D D
t
t t
t
D D
t
t t
t
t
D D D
e
r r
e e
e
r r
e e
e
t t r r e
e e
e
r r e
e e
e
e t r
k
e k
k
e k
k
e
e k
k
e
e k
k
k
(
(
(
(
> +
+ =
.
(4.B2.8)
4.B3: Variable Block size distributions
In the case of multiple spherical blocks with variable block size distribution the diffusivity
equation (Equation (4.62), with I=2) with the initial and boundary conditions (Equations
(4.63) to (4.65)) should be solved. The integral method is used to find the early time
solution with the following auxiliary equation:
. 0 , 0 , 0 ) ( ) (
2
2
=
c
c
=
c
c
= = =
D
D
D
D
D D D De D
r r
t t R r

c o
(4.B3.1)
The third-order trial solution as Equation (4.B1.2) is suggested to be used in the integral
method. Using the actual and auxiliary boundary conditions in the trial solution leads to the
following equations for dimensionless pseudo-pressure:
3
3
3
] [
] [
] [
o
o
c
c
De D
De
D
D
R r
R
r +
=
= .
(4.B3.2)
Using this trial solution (Equation (4.B3.2)) in the integral form of Equation (4.62), with
I=2 (From to R
De
) and some simplification leads to the following ODE:
) (
3
60
10 6 3
2
1
3 2 2 3
c
|q c c c
=
(
+
De
De D De De De
D
R
R R R R
dt
d
.
(4.B3.3)
Solving this ODE leads to the following equation:
2
1
4 3 3 4
720
17 24 4 3
De D
De De De
D
R
R R R
t
|q
c c c + +
= .
(4.B3.4)
This equation in terms of penetration depth can be expressed as follows:
150
2
1
2 2 3 4
720
30 16 3
De D
De De
D
R
R R
t
|q
o o o +
= .
(4.B3.5)
The early time solution is valid till <R
De
or
1
2
720
17
D
De
D
R
t
|q
< .
The early time average dimensionless pseudo-pressure is obtained by integrating over the
bulk volume of the matrix block:
1
2
3
2 2 3
3 2
3
720
17
,
20
15 6
] [
3
D
De
D
De
De De
R
D
De D
D
De
D
R
t
R
R R
dr
R r
r
R
De
|q
o o o
o
o
c
<
+
=
+
=
}
,
(4.B3.6)
where Equation (4.B3.5) is used to relate the penetration depth to the dimensionless time.
For the late time solution we have the following initial condition and the inner and outer
boundary conditions are the same as Equations (4.64) and (4.65):
3
3
1
2
1
720
17
D
De
D
D
De
D
r
R
R
t = =
|q
.
(4.B3.7)
third-order trial solution (Equation (4.B1.2)) and the residual (R) as Equation (4.B1.10).
The unknown coefficients (A, B, C and D) in the trial solution of Equation (4.B1.2) are
found by using the boundary conditions and making the zero and first moments of R vanish
by enforcing the following conditions:
} }
=
c
c
c
c
c
c
=
De De
R
D
D
D
D D
D D
D
D
R
D
dr
r
r
r t
r Rdr
0
2 2
0
0 )) ( ( 0

|q
, (4.B3.8)
} }
=
(
c
c
c
c
c
c
=
De De
R R
D
D
D
D D
D D
D
D D D D
dr
r
r
r t
r r Rdr r
0 0
2 2
0 ) ( 0

|q
.
(4.B3.9)
4 2
1
De De
DR CR A = . (4.B3.10)
Solving Equations (4.B3.8) and (4.B3.9), combining the results with Equation (4.B3.10)
and some simplification leads to a system of ODEs as follows:
D
R
C
R t d
C d
De
D
De
D
D
1
2
1
198 90 |q |q
+ = ,
(4.B3.11)
151
D
R
C
R t d
D d
De
D
De
D
D
2
1
3
1
4 . 176 84 |q |q

= .
(4.B3.12)
Solving the system of ODEs and using the initial condition for the late time solution
(Equation (4.B3.7)) leads to the following equation for late time pseudo-pressure for
multiple spherical blocks:
1
2
3
3 3
2
2 2
720
17
, )
218938 . 15 892569 . 1
(
)
095987 . 18 751798 . 3
( 877049 . 2 859229 . 1 1
2 1
2 1 2 1
D
De
D D
t
De
t
De
D
t
De
t
De
t t
D
R
t r e
R
e
R
r e
R
e
R
e e
D D
D D D D
|q
e e
e e e e
> +
+ + + =

(4.B3.13)
where
1
and
2
are defined based on Equation (4.74) in the main text. After that,
Equation (4.B3.13) is integrated over the matrix block volume (from zero to one) to obtain
the late time average matrix block pseudo-pressure as Equation (4.73) in the main text.

Chapter Five: Conclusions and Recommendations
5.1 Conclusions
This study has presented a validated semi-analytical model for fractured gas reservoirs.
The developed mathematical model can be used for modeling of matrix-fracture transfer
function for fractured gas reservoirs using dual-porosity approach. After derivation of
matrix-fracture transfer function and shape factor for dual-porosity gas reservoirs the effect
of fracture pressure boundary condition, block size distribution and block geometry was
investigated. Due to the analytical nature of the presented solution it can be nested in
single-phase dual-porosity simulation of fractured media which is efficient in terms of
computational time.
In the presented theoretical analysis the nonlinear pressure diffusion equation for dual-
porosity gas reservoirs was solved analytically. Then, the presented semi-analytical model
was extended to consider the effect of fracture pressure boundary condition, block size
distribution and block geometry. The presented model can also be used for single-phase
flow of slightly compressible fluids in the fractured media. The following major
conclusions are made as a result of this study.
5.1.1 Fracture pressure boundary condition effect for a slab-shaped block
In the first part of this study, a theoretical analysis of the shape factor for the flow of a
compressible fluid in fractured porous media is presented. Combination of the heat integral
method, the method of moments and Duhamels theorem are used to solve the nonlinear
diffusion equation resulting from the flow of a compressible gas in a dual-porosity system.
The approximate semi-analytical solution is validated with fine-grid numerical simulations.
Similar to the flow of a slightly compressible fluid, the shape factor demonstrates a
transient behaviour and then converges to a constant value during the pseudo-steady state
period for a compressible fluid.
The approximate analytical solution presented revealed that the matrix-fracture transfer
shape factor for single-phase flow of a compressible fluid in the dual-porosity media is a
function of the pressure boundary condition in the fracture. Based on the pressure
boundary condition in the fracture the stabilized value of the shape factor varies between
two limits. The upper limit is obtained for a linearly declining fracture pressure, which
Chapter 5. Conclusions and recommendations
153
corresponds to a slow pressure depletion regime. The lower limit is derived for the constant
fracture pressure boundary conditions where depletion takes place faster.
When the fracture pressure depletes exponentially with time, the stabilized value of the
shape factor falls between those values of the constant fracture pressure and linearly
declining fracture pressure. This stabilized value is a function of the decline exponent, .
For small decline exponents the stabilized shape factor has the same value as that for the
linearly declining fracture pressure. For large exponents, the stabilized shape factor is
equal to that of a constant fracture pressure. The pseudo-steady state time (stabilization
time) of the shape factor, increases as the fracture boundary condition changes from a fast
depletion regime toward a slow depletion regime.
5.1.3 Block size distribution effect for slab-shaped blocks
Variable block size distribution has a significant effect on the matrix production profile
during transient state. The results show that as the probability of larger blocks increases or
the fracture density decreases the production profile becomes closer to that for the ideal
block size distribution (Warren and Roots model).
Among different distributions the ideal block size distribution and the exponential
distribution with very small exponent (sparsely fractured system) have the smallest
transient cumulative production and their cumulative production reaches its plateau more
gradually. The large exponent exponential distribution (intensively fractured system) has
the highest transient matrix-fracture fluid transfer and deplete more quickly than other
distributions. The transient values of the dimensionless shape factor and the dimensionless
time of stabilization are a function of matrix block size distribution.
5.1.4 Block geometry effect
An integral approximation method has been used to derive the solutions for nonlinear
pressure diffusion in blocks with different geometries including cylindrical and spherical
blocks. The presented solutions have considered the effect of fracture pressure boundary
condition and the variable block size distributions or multiple blocks. The results
calculated by the approximate solutions are in good agreement with those calculated by the
fine grid numerical models.
Chapter 5. Conclusions and recommendations
154
It has been shown that the depletion time of a cylindrical or spherical matrix block is a
function of the fracture pressure boundary conditions. In the case of constant fracture
pressure or exponential decline with a large exponent the block is depleted faster than that
in the linear decline and the exponentially decline with a small exponent. For the linear
decline and exponential decline with a small exponent the early time dimensionless release
rate increases proportionally to the square root of the dimensionless time, then stabilizes at
a constant rate and finally falls to zero.
Block size distribution is another important parameter that affects the matrix production
during the transient state for cylindrical and spherical blocks. For the cylindrical or
spherical blocks the results for different block size distribution are the same as that for
slab-shaped blocks.
5.2 Recommendations
In this study we have considered a non-coupled approach for dual-porosity modeling of
compressible fluids, extension of this model using a coupled approach is recommended for
future works. Also, we developed a single-phase model of gas flow in fractured media and
incorporating multi-phase flow (including gas and water) in fractured media is an
important area for future studies. In the case of multiphase flow gravity can have a
significant effect and should be considered in the modeling of such cases.
Adsorption plays an important role in unconventional fractured gas reservoirs such as
CBM and shale gas reservoirs. Considering the adsorption in dual-porosity modeling of
compressible fluid is recommended for future works.
Appendix A. Copyright permission
155
APPENDIX A: Copyright Permissions

156

157

Modeling of Matrix-Fracture Interaction For Conventional Fractured Gas - Ucalgary - 2014 - Ranjbar - Ehsan PDF

Enviado por

Dados do documento

Descrição original:

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Modeling of Matrix-Fracture Interaction For Conventional Fractured Gas - Ucalgary - 2014 - Ranjbar - Ehsan PDF

Enviado por

Direitos autorais:

Formatos disponíveis

UNIVERSITY OF CALGARY

is the fracture pseudo-pressure when time tends to the infinity. In this

Você também pode gostar