MODELING OF MULTI-STAGE FRACTURED HORIZONTAL WELLS...

191
MODELING OF MULTI-STAGE FRACTURED HORIZONTAL WELLS A Thesis Submitted to the Faculty of Graduate Studies and Research in Partial Fulfillment of the Requirements for the Degree of Master of Applied Science in Petroleum Systems Engineering University of Regina By Shanshan Yao Regina, Saskatchewan December 2013 Copyright 2013: Shanshan Yao

Transcript of MODELING OF MULTI-STAGE FRACTURED HORIZONTAL WELLS...

MODELING OF MULTI-STAGE FRACTURED

HORIZONTAL WELLS

A Thesis

Submitted to the Faculty of Graduate Studies and Research

in Partial Fulfillment of the Requirements

for the Degree of

Master of Applied Science

in Petroleum Systems Engineering

University of Regina

By

Shanshan Yao

Regina, Saskatchewan

December 2013

Copyright 2013: Shanshan Yao

UNIVERSITY OF REGINA

FACULTY OF GRADUATE STUDIES AND RESEARCH

SUPERVISORY AND EXAMINING COMMITTEE

Shanshan Yao, candidate for the degree of Master of Applied Science in Petroleum Systems Engineering, has presented a thesis titled, Modeling of Multi-Stage Fractured Horizontal Wells, in an oral examination held on November 22, 2013. The following committee members have found the thesis acceptable in form and content, and that the candidate demonstrated satisfactory knowledge of the subject material. External Examiner: Dr. Tsun Wai Kelvin Ng, Environmental Systems Engineering

Co-Supervisor: Dr. Fanhua Zeng, Petroleum Systems Engineering

Co-Supervisor: Dr. Gang Zhao, Petroleum Systems Engineering

Committee Member: Dr. Farshid Torabi, Petroleum Systems Engineering

Committee Member: Dr. Exeddin Shirif, Petroleum Systems Engineering

Chair of Defense: Dr. Chun-Hua Guo, Department of Mathematics & Statistics

I

ABSTRACT

Horizontal wells stimulated by multiple fractures unlock tight formations and

shale gas reservoirs that used to be considered as uneconomic plays. However,

the popularity of such techniques presents new challenges to the reservoir and

fractures evaluation. Fractured horizontal wells’ pressure and production rate

behaviour exhibit complex trends that are quite different from previous horizontal

and fractured vertical wells. To facilitate the pressure and rate analysis, this

work developed semi-analytical models under different assumptions and

comprehensively described the fluid flow of a multi-stage fractured horizontal

well in a bounded reservoir.

The governing Partial Differential Equations (PDEs) in this work are highly

nonlinear, and, therefore, analytical methods are not applicable to obtaining

results of drawdown and build-up tests. The semi-analytical modeling method

here shows advantages in applicability over the analytical modeling.

For fractured horizontal wells with constant fracture conductivities, four

kinds of fluid flow, including flow from the reservoir to fractures and to the

horizontal wellbore, flow inside the fractures as well as inside the horizontal

wellbore were all taken into consideration. Standard type curves for transient

pressure analysis were documented. The unique pressure behaviour reveals

that multiple fractures play a more important role than the horizontal wellbore in

the whole system. Also, the applicability of these type curves in the transient

II

pressure analysis was proved when compared with the method based on typical

characteristic lines.

The stress-sensitive hydraulic fracture conductivity was the other factor

incorporated into the semi-analytical models. A series of type curves were also

generated to evaluate the fractures’ stress-dependent characteristics. Stress-

dependent conductivities can cause pressure and pressure derivative curves to

increase rapidly, which seem to be “apparent boundary-dominated flow”. The

influence of these changing conductivities strongly depends on the fractures’

properties (i.e., fracture stress-sensitive characteristic value df and initial fracture

conductivity CfDi.).

When the bottomhole flowing pressure remains constant, the semi-

analytical model can be further used to study the fractured horizontal wells’

production rate behaviour. After the comparison with analytical solutions, it can

be concluded that horizontal wells with multi-stage fractures produce as if the

fractures worked individually at early times. Moreover, if considering the stress-

sensitive fracture conductivities, the straight lines of reciprocal rates vs. time

exhibit special slopes deviate from 1/4 and 1/2 for bilinear and linear flow,

respectively.

In addition to the advantages over analytical methods, the methodology and

models presented are flexible and widely applicable as numerical models.

Remarkable progress can be achieved if the methodology is extended to solve

flow problems in complex fracture systems and dynamic matrix permeability.

III

ACKNOWLEDGEMENTS

I would like to take this opportunity to express my sincere appreciation and

gratitude to my co-supervisors, Dr. Fanhua Zeng and Dr. Gang Zhao, for their

guidance and support throughout my studies. Their encouragement, expertise,

advice, and enthusiasm helped me accomplish this study.

I also would like to thank my family: Jianhua Yao and Xianxia Zeng (my

parents) and Yulong Yao (my brother) for their endless love and understanding

during my graduate studies.

Acknowledgment is due to the Faculty of Graduate Studies and Research at

the University of Regina for financial support in the form of scholarships.

Furthermore, I am thankful to the members of my examination committee and

their valuable suggestions in this study. I also thank Heidi Smithson for her

proofreading.

I would also like to thank my colleagues, Ms. Lijuan Zhu, Ms. Suxin Xu, Mr.

Tao Jiang, Mr. Xinfeng Jia, and Mr. Zuojing Zhu for their care and helpful

discussion regarding this work.

IV

DEDICATION

To

My best friend and companion, Mr. Ning Ju,

and my loving parents Mr. Jianhua Yao and Ms. Xianxia Zeng.

V

TABLE OF CONTENTS

ABSTRACT ······················································································ I

ACKNOWLEDGEMENTS ······································································ III

DEDICATION ···················································································· IV

TABLE OF CONTENTS ·········································································V

LIST OF TABLES ················································································ IX

LIST OF FIGURES ················································································X

LIST OF APPENDICES ······································································ XIV

NOMENCLATURE ··············································································XV

CHAPTER 1 INTRODUCTION ······························································ 1

1.1 Multi-stage hydraulic fracturing ························································ 1

1.2 Scope and objectives of this study ···················································· 4

1.3 Organization of this dissertation ······················································· 5

CHAPTER 2 LITERATURE REVIEW ····················································· 7

2.1 Modeling multi-stage fractured horizontal wells ··································· 8

2.1.1 Analytical modeling ·········································································· 8

2.1.2 Numerical modeling ······································································· 10

2.1.3 Summary ····················································································· 11

2.2 Modeling horizontal wells with stress-sensitive hydraulic fractures ········ 11

2.2.1 Laboratory observations of stress-sensitive hydraulic fractures ················ 11

2.2.2 Modeling stress-sensitive hydraulic fractures ······································· 12

2.2.3 Fluid flow modeling with stress-sensitive hydraulic fractures ···················· 13

VI

2.2.4 Summary ····················································································· 15

2.3 Multi-stage fractured horizontal well production rate analysis ··············· 16

2.3.1 Decline curve analysis ···································································· 16

2.3.2 Type curve analysis ······································································· 18

2.3.3 Summary ····················································································· 20

2.4 Chapter Summary ······································································· 21

CHAPTER 3 METHODOLOGY ··························································· 23

3.1 Green’s functions and source/sink solutions ····································· 23

3.1.1 Green’s and source/sink function ······················································ 23

3.1.2 Newman product ··········································································· 25

3.2 Laplace transformation ································································· 26

3.3 Continuity conditions ···································································· 26

3.4 Constructing and solving linear equation systems ······························ 28

3.5 Chapter summary ········································································ 29

CHAPTER 4 MULTI-STAGE HYDRAULICALLY FRACTURED

HORIZONTAL WELLS ··················································· 30

4.1 Model and algorithm ···································································· 30

4.1.1 Dimensionless variables ·································································· 32

4.1.2 Mathematical model ······································································· 34

4.1.3 Algorithm ····················································································· 37

4.2 Model validation ·········································································· 43

4.3 Results and discussion ································································· 46

4.3.1 Effect of fluid flow from reservoir to horizontal wellbore··························· 46

4.3.2 Effect of horizontal wellbore pressure drop ·········································· 49

VII

4.3.3 Effect of fracture stages ·································································· 51

4.3.4 Effect of gas desorption ·································································· 61

4.3.5 Effect of skin factors ······································································· 64

4.4 Field examples ··········································································· 68

4.4.1 No.1 Build-up test analysis ······························································ 68

4.4.1 No.2 Build-Up test analysis ······························································ 69

4.5 Chapter summary ········································································ 75

CHAPTER 5 HYDRAULICALLY FRACTURED WELLS WITH STRESS-

SENSITIVE CONDUCTIVITIES ········································· 77

5.1 Model and algorithm ···································································· 78

5.2 Model validation ·········································································· 82

5.3 Results and discussion ································································· 85

5.3.1 Pressure behaviour characteristics ··················································· 85

5.3.2 Effect of degree of stress-sensitivity ··················································· 92

5.3.3 Effect of degree of conductivity loss ··················································· 96

5.3.4 Effect of initial conductivity ····························································· 100

5.3.5 Stress-sensitive conductivity ·························································· 102

5.4 Field example ··········································································· 105

5.4.1 Analysis without corrections in matrix permeability ······························ 105

5.4.2 Analysis with corrections in matrix permeability only ···························· 110

5.5 Chapter summary ······································································ 116

VIII

CHAPTER 6 PRODUCTION RATE ANALYSIS ··································· 118

6.1 Model and algorithm ·································································· 119

6.2 Model validation ········································································ 120

6.3 Results and discussion ······························································· 122

6.3.1 Comparison with analytical solutions of single-fractured wells ················ 122

6.3.2 Effect of stress-sensitive fracture conductivity ···································· 135

6.4 Field examples ········································································· 142

6.4.1 Marcellus shale gas well A ····························································· 142

6.4.2 Marcellus shale gas well B ····························································· 147

6.5 Chapter summary ······································································ 152

CHAPTER 7 Conclusions and Recommendations ····························· 153

7.1 Conclusions ············································································· 153

7.2 Recommendations ···································································· 155

List of References ··········································································· 156

APPENDIX A SOURCE FUNCTIONS ················································· 167

APPENDIX B SOLUTIONS OF FLUID FLOW INSIDE HYDRAULIC

FRACTURES ······························································ 170

IX

LIST OF TABLES

Table 4.1 Basic input parameters. ··················································· 71

Table 5.1 Reservoir, well, fracture and fluid data. ······························· 88

Table 5.2 Results of field case analysis. ········································· 114

X

LIST OF FIGURES

Figure 1.1 Schematic of a multi-stage fractured horizontal well. ················ 3

Figure 4.1 A multi-stage fractured horizontal well in a box-shaped reservoir.

···················································································· 31

Figure 4.2 Diagram showing fracture and horizontal wellbore discretization.

···················································································· 39

Figure 4.3 Flow chart for modeling and solving process. ······················ 44

Figure 4.4 Model validation with Kappa Ecrin. ····································· 45

Figure 4.5 Effect of fluid flow from reservoir to the wellbore. ··················· 47

Figure 4.6 Flow distribution along the fracture and wellbore. ·················· 48

Figure 4.7 Effect of horizontal wellbore pressure drop. ·························· 50

Figure 4.8 Effect of fracture stages when xf is constant. ························ 55

Figure 4.9 Effect of fracture stages on the productivity index with costant

xf. ················································································· 56

Figure 4.10 Effect of fracture stages when fracture volume is constant. ····· 57

Figure 4.11 Effect of fracture stages on the productivity indxex with cosntant

Vf. ················································································ 58

Figure 4.12 Non-uniform fractures along the horizotnal wellbore. ·············· 59

Figure 4.13 Effect of non-uniform fractures on the pressure behaviour . ····· 60

Figure 4.14 Effect of gas desorption on pressure behaviour . ··················· 63

Figure 4.15 Effect of skin factors on the pressure behaviour . ·················· 66

Figure 4.16 Effect of skin factors on the flow distribution along the fracture. 67

XI

Figure 4.17 Field production data. ······················································ 70

Figure 4.18 Plot of adjusted pressure vs. fourth root time. ······················· 72

Figure 4.19 Type curve match for the first test. ····································· 73

Figure 4.20 Type curve match for the second test.································· 74

Figure 5.1 Discretizing the fracture system. ········································ 80

Figure 5.2 Discretizing the fracture conductivity. ·································· 80

Figure 5.3 Model validation. ···························································· 84

Figure 5.4 Transient pressure behaviour with stress-sensitive hydraulic

fractures. ······································································· 89

Figure 5.5 Modified smooth conductivity curve. ···································· 90

Figure 5.6 Pressure behavior based on modified conductivity change. ····· 91

Figure 5.7 Normalized fractures conductivities change with stress (Abass et

al., 2009 and Zhang, et al., 2013) ······································· 94

Figure 5.8 Type curves showing the effect of stress-sensitive conductivity,

CfDi=50, CfDmin=0.25. ························································ 95

Figure 5.9 Normalized fracture conductivity (Abass et al., 2009) ············· 98

Figure 5.10 Type curves showing the effect of minimum conductivity, CfDi=50

and df=1×10-7Pa-1. ··························································· 99

Figure 5.11 Type curves showing the effect of initial conductivity,

CfDi/CfDmin=200. ······························································ 101

Figure 5.12 Variation of fracture conductivity ratio with different df and CfDi.104

XII

Figure 5.13 Gas and water production rates, and calculated flowing

bottomhole pressures (Clarkson et al., 2012). ··················· 107

Figure 5.14 Pressure and pressure derivative without corrections in

matrix permeability (Clarkson et al., 2012). ························ 108

Figure 5.15 Type curve matching results without corrections in

matrix permeability. ······················································ 109

Figure 5.16 Pressure and pressure derivative with corrections in

matrix permeability (Clarkson et al., 2012). ························ 112

Figure 5.17 Type curve matching results with corrections in matrix

permeability. ······························································· 113

Figure 5.18 Production rate prediction. ············································· 115

Figure 6.1 Model validation. ·························································· 121

Figure 6.2 Results comparison for bilinear flow. ································ 124

Figure 6.3 Results comparison for linear flow. ·································· 127

Figure 6.4 Production rates with different fracture half-lengths. ············ 130

Figure 6.5 Results comparison for compound linear flow. ··················· 131

Figure 6.6 Results comparison for boundary-dominated flow. ·············· 134

Figure 6.7 Production curves with different df. ·································· 137

Figure 6.8 Conductivity curves with different df.·································································· 138

Figure 6.9 Production curves with different initial conductivity CfDi. ······· 140

Figure 6.10 Conductivity curves with different CfDi. ······························ 141

Figure 6.11 Plot of reciprocal gas rate vs. fourth root of time. ················ 143

Figure 6.12 Plot of reciprocal gas rate vs. square root of time. ·············· 144

XIII

Figure 6.13 Type curve matching for Marcellus shale gas well A. ·········· 146

Figure 6.14 Diagnostic plot of production rate vs. time. ························ 149

Figure 6.15 Plot of reciprocal gas rate vs. square root of time. ·············· 149

Figure 6.16 Type curves based on parameters from analytical

solutions. ···································································· 150

Figure 6.17 Type curve matching result by semi-analytical model. ········· 151

XIV

LIST OF APPENDICES

APPENDIX A Source functions ························································· 167

APPENDIX B Solutions of fluid flow inside hydraulic fractures ················· 170

XV

NOMENCLATURE

a = reservoir length, L, m

b = reservoir width, L, m

ct =total compressibility, L2/m, pa-1

tc =total compressibility at pressure in the region of influence, L2/m, pa-1

cg = gas compressibility, L2/m, pa-1

C =gas concentration at the surface of pore walls, n/L3, mole/ m3

Cw =wellbore storage, L5/m, m3/Pa

Cf = fracture conductivity, L3, m3

Cη = fracture diffusivity, dimensionless

df = fracture stress-dependent characteristic, L2/m, Pa-1

h = net-pay thickness, L, m

J=pseudo-steady state productivity index, L5/(t∙m), m3/(Pa∙s)

k = reference permeability, L2, m2

k =permeability at pressure in the region of influence, L2, m2

ks = fracture-face skin zone permeability, L2, m2

NRe= Reynolds number, dimensionless

NRe,w = Inflow Reynolds number, dimensionless

p = pressure, m/L2, Pa

pa=adjusted pressure, m/L2, Pa

q = flow rate, L3/t, m3/s

qf = flow rate normal to fracture, L2/t, m2/s

XVI

qh = flow rate normal to horizontal wellbore, L2/t, m2/s

Q=reference flow rate, L3/t, m3/s

RD=dimensionless desorption storability ratio

sck = choked-fracture-skin factor, dimensionless

sff = fracture-face skin factor, dimensionless

S= strength of source, dimensionless

t = time, t, s [hr]

T =temperature, T, K

u=Laplace variable

v = velocity, L3/s, m3/s

wf = fracture width, L, m

xck = choke length in one fracture wing, L, m

xf =fracture half-length, L, m

z = deviation factor, dimensionless

μ = viscosity, m/(L2∙t), Pa∙s

=viscosity evaluated at pressure in the region of influence, m/L, Pa∙s

ρ = density, m/L3, g/cm3

ϕ= porosity, fraction

Ω= segment number

Γ=segment interface

τ=time variable in integration, t, s

Subscripts

D = dimensionless

XVII

f = fracture

h= horizontal well

i=initial condition.

min=minimum value.

r = reference length

wf= wellbore bottomhole condition

sc= standard conditions

1

CHAPTER 1

INTRODUCTION

1.1 Multi-stage hydraulic fracturing

Hydraulic fracturing is a technique during which typically a mixture of water,

propping agents (usually sands), and chemicals is pumped at sufficiently high

rates and pressure into the pay zone to create fractures (Ralph,1983). Created

fractures can provide conduits for gas, oil, and water to easily migrate towards

the well. The fracture usually has two wings extending in opposite directions

from the well.

The first hydraulic fracturing experiment was conducted in 1947 at the

Hugoton gas field, Grant Country, Kansas, USA (Clark, 1949). In 1949, two

commercial hydraulic fracturing treatments were introduced into the industry in

Oklahoma and Texas. Since then, the hydraulic fracturing technique has

evolved into a standard operating practice and approximately 2.5 million of such

treatments have been performed globally.

As target formations became deeper, hotter, and lower in permeability,

massive hydraulic fracturing (MHF) emerged in 1968 to address the associated

challenges. The definition of MHF varies but generally refers to the treatments

injecting up to 3.8×103 m3 (1×106 gal) fracturing fluid and more than 1.36×106 Kg

(3×106 lbm) propping agent (Ben and Spencer, 1993). MHF treatments

significantly improve wells’ productivity by creating large and high-conductivity

fractures, especially in tight and shale formations.

2

Horizontal well completion is another technique that has proved to be much

more effective than vertical wells for tight chalk and shale formations. The

horizontal wellbore extends horizontally within the target formation to a

predetermined bottomhole location. This lateral wellbore makes it easier to

conduct multiple fracturing treatments for one well.

In the late 1980s, operators in Texas began to combine horizontal drilling

with MHF treatments. The hydraulic fracturing technique evolved from single-

stage operation to multi-stage (40 plus) fracturing. Multi-stage fracturing begins

at the toe of the long horizontal wellbore and extends down to the heel. For each

stage, the corresponding wellbore section is isolated, and then, water is pumped

to crack the formation. Sand carried along with the water can prop the facture.

Figure 1.1 shows the schematic of a multi-stage fractured horizontal well.

Multi-stage fractured horizontal wells are crucial to unconventional reservoir

development. Prior to the popularity of multi-stage fractured horizontal wells,

unconventional resources were always overlooked by operators. Take the

Barnett Shale as an example. Before 1980, Barnett Shale was known to have

essentially zero permeability and, thus, was considered uneconomic. However,

the gas production from multi-stage fractured horizontal wells in Barnett Shale

increased to about 5 Bcf per day in 2010 with application of multi-stage fractured

horizontal wells .( United States Department of Energy, 2011).

3

(Original in color)

Figure 1.1― Schematic of a multi-stage fractured horizontal well (Mitch, 2010).

4

Several mechanisms have been proposed to explain the advantages of

multi-stage hydraulic fracturing over other techniques. At first, multiple hydraulic

fractures provide more “superhighways” than single-stage fractured vertical

wells. Moreover, the increasing stages can substantially enlarge the contacted

reservoir area. Furthermore, there always exist open natural fractures in tight

and shale formations. The interaction between hydraulic fractures and the

natural-fracture network expands the stimulated reservoir volume (SRV) beyond

which the remaining reservoir is usually negligible (Medeiros et al. 2008;

Mayerhofer et al. 2010).

Although multi-stage fractured horizontal wells are efficient, analyzing and

predicting such wells’ performance are challenging not only because of

potentially complex reservoir behaviour (dual porosity/dual permeability, multi-

layer, stress-dependent porosity and permeability, multi-phase flow, etc.), but

also because of the complex sequence of flow regimes that evolve over time

during production (Clarkson and Pedersen, 2010). Hence, advanced production

analysis methods are highly appreciated.

1.2 Scope and objectives of this study

The objectives of this study are to develop a set of semi-analytical

methodologies to rigorously model the fluid flow behaviour in reservoirs with

multi-stage fractured horizontal wells, to present standard type curves for

identifying matrix and fracture properties with transient pressure and production

rate data, and to help petroleum engineers better understand the fractured

horizontal wells’ influence on reservoir development.

5

This work focuses on single-phase slightly compressible fluid flow in porous

media. The multi-stage fractured horizontal well exists in a whole homogeneous

reservoir. Hydraulic fracture conductivities could be constant or change along

with pressure. All above phenomena are rigorously modeled.

The influence of different parameters, including fracture half-lengths,

fracture conductivities, fracture stages, and the horizontal wellbore contribution,

on pressure and production rate behaviour is studied based on sensitivity

analysis. These results can be further applied in optimizing hydraulic fracturing

treatments. This study also derives analytical solutions that can be used to

evaluate and predict the production rates.

This study provides a general approach to accurately model complex fluid

flow in the reservoir with multi-stage fractured horizontal wells. This approach

can be further coupled with geomechanical methods in modeling the fracture-

formation interaction and rigorous PVT behaviour variation in transient pressure

and/or production rate analysis.

1.3 Organization of this dissertation

This dissertation is presented in seven chapters. Chapter 1 introduces the

background of multi-stage fractured horizontal wells. The objectives and

possible applications of this work are outlined. A comprehensive literature

review of scientific research on modeling multi-stage fractured horizontal wells is

conducted and limitations of current research are discussed in Chapter 2.

Chapter 3 presents the general methodology used in developing semi-analytical

models in this work. Chapter 4 shows a semi-analytical model and analyzes the

6

transient pressure behaviour of a multi-staged fractured horizontal well with

constant fracture conductivities. Type curves are shown with different

combinations of parameters. One field case is also analyzed. The effect of

stress-dependent hydraulic fracture conductivities on the transient pressure

behaviour is studied in Chapter 5. Another field example is introduced in this

chapter. In Chapter 6, the transient production rate behaviour of the fractured

horizontal well is analyzed based on my semi-analytical models and two field

examples. Several analytical solutions are also derived and tested in field

applications. Finally, Chapter 7 draws conclusions and provides

recommendations.

7

CHAPTER 2

LITERATURE REVIEW

Although multi-stage hydraulic fracturing is important for many

unconventional reservoirs, it is difficult to evaluate the fractures’ properties and

predict the wells’ performance since the transient pressure and production rate

behaviours are influenced by many factors, such as the reservoir permeabilities

and fracture conductivities. Therefore, accurately modeling and measuring the

effect of each parameter is necessary for better understanding the fractured

well/reservoir system.

Great effort and attempts have been made to model fractured wells during

recent decades. Related research in the literature and the related methodologies

used to model fractured wells are reviewed in this chapter. Based on the

objectives of this study, all these efforts and attempts in the literature can be

classified into the following three main categories:

Modeling multi-stage fractured horizontal wells with constant

properties,

Modeling hydraulically fractured wells with stress-sensitive

conductivities.

Modeling the production behaviour of fractured horizontal wells,

The literature review is also presented under the same categories as above.

8

2.1 Modeling multi-stage fractured horizontal wells

2.1.1 Analytical modeling

Although the Green’s function method had long been known, it was not

widely used in modeling flow behaviour in reservoirs until 1973. In 1973,

Gringarten and Ramey used the Green’s and source functions with the Newman

product method to generate reservoir transient-flow problem solutions. Pressure

response integration to an instantaneous source was applied to describe the

pressure behaviour of a continuous plane/slab source.

Gringarten et al. (1974) applied the source functions to model the pressure

behaviour of a well with a single infinite-conductivity vertical fracture. Compared

with Russell and Truitt’s work (1964), this new model is specifically useful in the

analysis of short-time field data. The analysis based on this model can provide

information concerning permeabilities, fracture lengths, and distance to a

symmetrical drainage limit.

The assumption of infinite conductivity is inapplicable for long and/or low-

conductive fractures. Then, Cinco-Ley and Samaniego (1981) developed a

mathematical model for the finite-conductivity fracture and used Laplace

transformation to solve the corresponding PDEs. Bilinear flow, which had never

been considered before, appears in the generated type curves.

Horizontal well drilling is another important technique in developing

reservoirs. In 1989, Babu and Odeh integrated point source functions and used

the simplified equations to describe the pressure behaviour of a horizontal well

9

in bounded reservoirs. However, the horizontal well must be parallel to one of

the boundaries.

Originally, horizontal wells were thought to have infinite conductivity, which

may lead to erroneous evaluation. In 1999, Penmatcha and Aziz proposed a

transient reservoir/wellbore coupling model for finite-conductivity horizontal wells

based on Babu and Odeh’s solutions. It was concluded that ignoring the

wellbore pressure drop could overpredict horizontal wells’ productivity.

All the above analytical solutions for fractured vertical wells and horizontal

wells laid a strong foundation for modeling multi-stage fractured horizontal wells

analytically.

In 1994, Guo and Evans developed a systematic methodology for modeling

a horizontal well intersecting multiple random discrete fractures, but the

interference between fractures was ignored, the study of which was undertaken

by Horne and Temeng (1995) who used the superposition principle on the basis

of Babu and Odeh’s solutions. In 1997, Chen and Raghavan rewrote Horne and

Temeng’s solutions by Laplace transformation according to Ozkan and

Raghavan who documented an extensive library of transient pressure solutions

for a wide variety of wellbore configurations in 1991.

One disadvantage of the source/sink function method is its inherent

singularity where the point source/sink is placed. In order to avoid this limitation,

Valkó and Amini (2007) developed the distributed volumetric sources (DVS)

method, which can also model the transient pressure behaviour.

10

In 2009, Ozkan, Raghavan, and Kazemi further presented an analytical tri-

linear flow model for fractured horizontal wells intercepted with natural fractures.

The dual-porosity inner reservoir between hydraulic fractures in this model is

naturally fractured. Mayerhofer et al. (2010) classified the stimulated reservoir

volume (SRV) in the fractured horizontal well system and suggested that

modeling SRV is important for evaluating the stimulation performance since the

contribution beyond the SRV can be ignored.

2.1.2 Numerical modeling

In numerical modeling of fractured horizontal wells, much effort has been

dedicated to representing fractures accurately and effectively.

In 1996, Herge studied the numerical simulation of fractured horizontal wells

in detail. He indicated that numerical models of fractured horizontal wells should

be dependent on the study objectives. If the early-time transient pressure

behaviour is required, explicit modeling of fractures and small grids near

fractures is appropriate. If the fractures are assumed to be infinite-conductive, it

is recommended to simply connect the wellbore to fractures and specify

connection factors.

In shale gas reservoirs, complex fractures networks are always generated

during multi-stage hydraulic fracture treatments. In 2009, Cipolla et al. proposed

the “LS-LR-DK” model to simulate fractured horizontal wells. In this model, the

single-plane propped fracture is modeled using a locally refined grid. Moreover,

the dual-permeability method is applied to represent fractures in both the

stimulated and unstimulated volumes. At last, grids in the stimulated volume are

11

further locally refined logarithmically. Also, the gas desorption effect and stress-

dependent fracture emerges in the model.

2.1.3 Summary

Both analytical solutions and numerical modeling are widely used for

fractured wells. Compared with analytical models, numerical solutions are

numerically unstable and time-consuming with excessively fine grid blocks.

Although accurate, analytical solutions are still limited to a series of assumption

and simplification. For example, in most cases, only one fracture is selected

from the whole system for detailed study under the assumption that all fractures

are the same. Therefore, a comprehensive semi-analytical model is required for

the multi-stage fractured horizontal wells, which incorporates analytical source

solutions and numerical discretization. Some phenomena that are visible in

unconventional reservoirs should also be added into this new mathematical

model.

2.2 Modeling horizontal wells with stress-sensitive hydraulic

fractures

2.2.1 Laboratory observations of stress-sensitive hydraulic fractures

Friedel et al. (2007) suggested the dependency of propped hydraulic

fracture permeabilities on reservoir pressure according to the experimental data

from Core Lab. Abass et al. (2009) and Zhang et al. (2013) also performed

experiments to investigate the propped fracture permeability vs. stress. Their

12

results indicate that the fracture conductivity can be reduced to a few to

hundreds of times.

2.2.2 Modeling stress-sensitive hydraulic fractures

Best and Katsube (1995) proposed that there is no sufficient support to

firmly describe the relationship between hydraulic fracture conductivities and the

stress change. The relationship between hydraulic fracture conductivities and

the effective stress is seldom discussed in the literature, while several

correlations between the matrix and natural fracture permeability and effective

stress have been presented.

Jones (1975) first derived the relationship between the permeability and

stress for a carbonate reservoir core sample containing natural fractures. Based

on Jones’ experimental data, a linear correlation between cubic root of

permeability and logarithm of confining pressure is established. However, the

permeability is referred to as mean permeability for the whole system rather than

fracture permeability itself, and the applicability of the relationship is subject to

proof for other kinds of rocks except carbonate rock.

Pedrosa (1986) and Yilmaz et al. (1991) summarized the effects of pressure

on matrix permeability and applied the rock permeability modulus γ as a

measure of dependency on pore pressure. The permeability can be expressed

as (Yilmaz et al., 1991):

13

pp

i

iekk

·············································································· (2.1)

where ki is the permeability at initial condition and γ is the formation permeability

modulus. Equation (2.1) is suitable for different rock types and γ is determined

by the rock characteristics.

Raghavan and Chin (2002), Rutqvist et al. (2002), and Minkoff et al. (2003)

proposed a series of more comprehensive correlations for stress-dependent

matrix and natural-fracture permeability, respectively. All those equations are

similar in form, and in them, permeability reduces exponentially with

stress/pressure change. I chose the simple but practical equation from

Raghavan and Chin (2002):

ffpd

fifekk

. ·········································································· (2.2)

In Equation (2.2), df is a characteristic parameter of the rock type ǀ, which is

determined experimentally.

According to Berumen and Tiab’s (1996) work, the above correlations for

stress-sensitive matrix/natural fracture permeabilities can be revised and further

applied to hydraulic fractures.

2.2.3 Fluid flow modeling with stress-sensitive hydraulic fractures

At first the stress-sensitive matrix permeability was incorporated into

reservoir flow models for transient pressure analysis. A number of investigators,

such as Vairogs et al. (1971), Raghavan et al. (1972), and Samaniego et al.

(1977), defined different kinds of pseudopressure functions, which include

pressure-dependent fluid and rock properties, to solve nonlinear flow equations.

Solutions are only suitable for vertical wells in a radial homogeneous reservoir.

14

Pedrosa (1986) incorporated the matrix permeability modulus γ into

mathematical modeling of a stress-sensitive formation. The simplification of

permeability change with the Taylor expansion of γ is used for an approximate

analytical solution under constant boundary conditions. There is no need to input

pressure vs. permeability data into the mathematical model.

Based on Pedrosa’s work, Celis et al. (1994) extended the matrix

permeability modulus to analytically model transient and pseudo-steady state

transfer between matrix and stress-sensitive natural fractures. The natural

fracture permeability modulus is created, which is similar with the matrix

permeability modulus.

Berumen and Tiab (1996) and Pedroso et al. (1997) modeled vertical wells’

pressure behaviour with pressure-dependent-conductivity hydraulic fractures by

defining the hydraulic fracture permeability modulus. Hydraulic fracture

permeability is also assumed to change exponentially. This shows that not

considering the pressure-dependency of hydraulic fracture conductivities may

lead to incorrect estimates of the fracture-formation properties.

In 2000, Poe proposed the production analysis model to analyze the rate

behaviour of fractured wells subject to stress-dependent variation of both the

intrinsic formation and hydraulic fracture properties. In 2009, Cipolla et al.

modeled the well performance with stress-sensitive partially propped fractures in

shale gas reservoirs by using numerical simulation. It showed that significant

reductions in fracture conductivities are likely with increasing ultimate gas

recovery.

15

To avoid complex calculation, Clarkson et al. (2012) approximated the

hydraulic fracture conductivity changes by a time-dependent skin factor in rate

transient analysis of multi-stage fractured horizontal wells. In the linear flow

regime, the one-half slope trend in the square-root time plot can be represented

as (Clarkson et al., 2012):

btmq

pmpm

g

wfi

)()(, ···························································· (2.3)

where m and b are the slope and intercept of the square-root time plot,

respectively. Changes in fracture conductivity can cause long-term changes of b,

the dynamic skin factor.

2.2.4 Summary

The above literature review shows some available results in modeling wells

with stress-dependent fractures by applying the pseudo-variables, dynamic skin,

and numerical simulation. Most of them focus on single fracture and/or

production rate analysis. Moreover, the accuracy and applicability of the above

methods cannot meet the requirements of transient pressure analysis. Therefore,

a more detailed study on the transient pressure behaviour of vertical/horizontal

wells with one or more fractures is required.

16

2.3 Multi-stage fractured horizontal well production rate

analysis

2.3.1 Decline curve analysis

Nearly all production decline curve analyses are based on or related with

Arps’ empirical decline equation (Arps, 1944):

b

i

i

tbD

qtq

/1]1[

)(

, ······································································ (2.4)

where 10 b . 0b indicates an exponential decline while 1b refers to

harmonic decline and 10 b defines the hyperbolic decline. The larger the b

value, the smaller the decline rate becomes.

For a low-permeability tight gas reservoir with fractured wells, a single

decline equation with b<1 cannot evaluate the production. The optimized

exponent always exceeds the unit. Maley (1985) argued that no theoretical basis

is set for limiting b to a value less than 1. Moreover, he proposed that the Arps’

decline equation with b=2 can better approximate the decline in the linear flow

regime.

In practice, field operating conditions keep changing during production,

which makes it difficult to apply decline equations, especially in tight formations.

Even for long periods of operational stability, Arps’ equations may be insufficient

to represent actual production behaviour. In 1988, Robertson further modified

the Arps’ equation as

N

N

o

at

atqtq

exp1

exp1)(

, ··························································· (2.5)

17

with 10 .This equation can be applied to match both the early-stage

production with an hyperbolic curve and the late-time term results with an

exponential curve.

The Robertson’s equations had no physical basis, and such decline

behaviour is very unlikely in nature. Also, Cox et al. (2002) investigated the

applicability of Arps’ decline equation in fractured tight gas reservoirs. It was

concluded that Arps’ decline curve analysis is suitable for wells whose drainage

remains constant (i.e., boundary-dominated flow (BDF) appears). However, for

tight-gas and shale-gas wells, transient flow may last for many years. Even if a

best match is obtained with b larger than the unit by Arps’ equation, the future

performance and remaining reserves can be greatly overestimated.

Cheng et al. (2008) proposed an improved technique of analyzing transient-

flow-dominated production data by Arps’ decline equation. They determined the

b value for BDF a priori and then conducted history matched for multiple periods

of late-stage production data in a backward way. The final parameters for the

latest history match can project future production. Despite some limitations, it

still produces far better results than conventional methods.

In 2008, Kupchenko et al. studied the production decline for hydraulically

fractured vertical wells in tight gas reservoirs and summarized the decline

exponent b in bilinear, linear, and pseudo-radial flow regimes. It was proposed to

use b=2 during the linear flow regime and classic Arps’ hyperbolic decline for the

BDF.

18

Duong (2011) provided a new empirical approach for predicting

performance of fractured wells in unconventional reservoirs. The dimensionless

time and rates in the model are described as:

maxt

tt

m , ················································································· (2.6)

and

11

max

1

m

mt

m

m

m

met

q

q. ···································································· (2.7)

To reduce uncertainties, the best first 3-month average rate is used as qmax in

the model. The larger m becomes, the bigger the decline rate will be.

2.3.2 Type curve analysis

The traditional decline curve analysis is deficient during transient flow, and

the numerical models are too complicated to be always available to practicing

engineers. Therefore, some researchers recommend using modern decline

analysis methods. Fetkovich (1980) integrated analytical solutions of a radial

flow system into Arps’ empirical decline equation and presented generated type

curves in log-log plots in dimensionless forms. The dimensionless liquid

production rate and time are shown as (Fetkovich, 1980)

)(

5.0)ln()(3.141

wfi

w

e

Dd

ppkh

r

rtBq

q

, ·················································· (2.8)

and

5.0)ln(15.0

100634.0

22

w

e

w

ewt

Dd

r

r

r

rrC

ktt

. ································· (2.9)

19

Such Fetkovich type curves are widely applied in production rate analysis, but in

fact, the curves are only strictly applicable for slightly-stimulated wells exhibiting

radial flow, which usually is not the case for fractured wells. Then, Carter (1985)

and Palacio and Blasingame (1993) filled this gap in Fetkovich decline curves

and provided a new set of type curves for analyzing transient linear flow of

fractured gas wells.

In 1998, Agarwal et al. developed the Agarwal-Gardner type curves for

vertically fractured wells by using pressure transient analysis concepts for the

first time. Rate-time ( wDp/1 vs.

DAt ), rate-cumulative ( wD

p/1 vs. wDDApt / ), and

cumulative-time production ( wDDApt / vs.

DAt ) decline type curves and derivative

curves are presented. These type curves make a clearer distinction between

transient flow and BDF periods than previous curves. I can not only estimate the

gas/oil-in-place but also the reservoir permeability, skin factor, fracture length,

and fracture conductivities using these type curves.

In tight gas production, the assumption of infinite-conductivity fractures is

typically inadequate. Then, in 2003, Pratikno, Rushing, and Blasingame

developed new type curves for a well with a finite-conductivity vertical fracture

centered in a bounded, circular reservoir based on the analytical transient flow

solutions given by Cinco-Ley and Meng (1988).

For multi-stage fractured horizontal wells, it becomes more complex in the

production rate analysis because several fractures work together along one

horizontal wellbore. Lin and Zhu (2010) developed a corresponding semi-

analytical model by using the slab source method to predict the performance of

20

multi-stage fractured horizontal wells under a constant pressure condition. Each

fracture is regarded as an individual source, and the interference between

fractures is included by the superposition principle. It proves that for low-

permeability reservoirs, multi-stage hydraulic fracturing can increase production

dramatically. Regrettably, no further detailed analysis was undertaken based on

Lin and Zhu’s model.

Then, Bello and Wattenbarger (2010) used a linear dual-porosity model to

model multi-stage fractured horizontal wells in shale gas reservoirs and

generated type curves. In the type curves, five flow regimes have been identified

as: 1) transient drainage in fractures; 2) bilinear flow; 3) infinite-acting flow; 4)

transient drainage from the matrix; and 5) boundary-dominated flow, which

provides reference for future transient rate analysis.

2.3.3 Summary

In previous decades, significant advances have been achieved in the

development of analytical models and corresponding type curves for analyzing

and forecasting production rates for fractured wells in unconventional reservoirs.

Despite such advancement, empirical decline curves remain popular for

forecasting fractured wells’ performance. However, all aforementioned methods

are limited in production analysis of fractured horizontal wells. No type curves

generated by accurate mathematical model are reported for detailed transient

rate analysis. In field application, more simple but reasonable methods are also

in need for evaluating and predicting production at different flow regimes,

especially the long linear flow regime. Through comparing analytical solutions of

21

fractured vertical wells and semi-analytical modeling of multi-stage fractured

wells, the relationship among fractures in a multi-stage fractured horizontal well

can be further investigated.

2.4 Chapter Summary

The above literature review outlines the progress in modeling the fluid flow

of multi-stage fractured horizontal wells in three categories. The first section

covers modeling of the transient pressure behaviour of fractured wells with

constant fracture properties. The second section reviews the attempts that were

made to solve problems of stress-sensitive hydraulic fracture conductivities in

fractured wells. Then, in the final section, the production rate instead of transient

pressure is analyzed to evaluate the performance of multi-stage fractured

horizontal wells. As can be seen, the studies on all three different aspects of

multi-stage fracture horizontal wells have limitations, which are summarized as:

On the one hand, numerical solutions are uncertain and time-consuming

with excessively fine grid blocks to represent fractures. On the other hand,

analytical solutions are limited to a series of assumptions and

simplifications despite their accurateness.

Much attention has been focused on describing the stress-sensitive

matrix and natural fracture permeability in transient pressure analysis. No

summarization specifically about hydraulic fracture conductivities vs.

stress change is provided even though the importance of stress-sensitive

hydraulic fracture conductivities has been proven by experiments.

22

No type curves computed with accurate mathematical models are

reported for detailed transient rate analysis of multi-stage fractured

horizontal wells with/without stress-dependent fracture conductivities.

Simple and convenient correlations are also in need study in field

applications.

Thus, studies on multi-stage fractured horizontal wells need further

improvement. New methodologies and models should be tried to provide more

accurate, effective, and comprehensive solutions for fluid flow problems of multi-

stage fractured horizontal wells. In this work, I present three semi-analytical

models that can obtain as accurate of analytical solutions and that can work as

powerfully and flexibly as numerical models. In Chapters 4, 5, and 6, the

proposed models are presented with respect to their applications in transient

pressure and production rate analysis for multi-stage fractured horizontal wells

with/without stress-sensitive conductivities.

23

CHAPTER 3

METHODOLOGY

In this chapter, a semi-analytical methodology is presented to solve non-

linear mathematical models in the following chapters. This methodology includes

deriving source/sink functions, Laplace transformation, coupling solutions by

continuity conditions, and constructing and solving the linear equation systems.

3.1 Green’s functions and source/sink solutions

The Green’s and source/sink function method is a powerful way to solve a

wide variety of reservoir flow problems, especially for reservoirs with complex

well geometries.

3.1.1 Green’s and source/sink function

Assuming constant permeabilities, porosity, fluid viscosity, small pressure

gradient everywhere, and no gravity effect, the diffusivity equation for the

reservoir can be written as (Gringarten and Ramey, 1973)

0),(

),(2

t

tMptMp . ······························································· (3.1)

The solution ),( tMp is uniquely determined by the initial and boundary

conditions. The instantaneous Green’s function, tMMG ,, , with respect to

Equation 3.1, can represent the pressure response of the point M’(x’,y’,z’) at

time t , which is stimulated by an instantaneous fictitious source with unit

strength at the point M(x,y,z) at the time with zero initial and boundary

conditions (Gringarten and Ramey, 1973).

24

If the reservoir produces at a certain flux rate and the Green’s function can

be found, then the pressure ),( tMp in the reservoir with initial condition ),( tMpi

is given by (Gringarten and Ramey, 1973)

t

SMS

e

w

t

Dww

t

dMdSMn

tMMGMp

Mn

MptMMG

ddMtMMGMqC

tMp

ee

w

0

0

)(])(

),,(),(

)(

),(),,([

),,(),(1

,

,

···································································································· (3.2a)

where

),(,,)(),( tMpMdtMMGMptMpD

i . ···································· (3.2b)

In Equations 3.2a and 3.2b, ),( w

Mq is the withdrawal (source) or injection (sink)

rate per unit volume at each point of the source/sink and n

is differentiation

normal to the boundary element )( MdSe

in the outward direction.

The above pressure drop consists of two parts. One part is responsible for

the source/sink. The other part refers to the outer boundary conditions. If the

domain D is infinite or infinite with zero boundary conditions, the second part

would disappear. Therefore, I can simplify Equations 3.2a and 3.2b for infinite

reservoirs. If the fluid rate is assumed to be uniform over the entire source

volume, a new simple equation is obtained (Gringarten and Ramey, 1973):

t

t

dtMSqC

tMp0

),()(1

),(

, ·················································· (3.3a)

where

25

wD

wwdMtMMGtMS ),,(),( . ····················································· (3.3b)

),( tMS is the instantaneous uniform flux source function. The integration

over time can generate the continuous source function. Zhao (1999) provided

the methods to calculate source functions accurately. Liu (2006) further

generated a series of source functions in modeling the reservoir flow with

wormholes. Basic source functions based on Zhao (1999) and Liu’s (2006) work

are listed in Appendix A.

3.1.2 Newman product

Although Green’s and source/sink functions are powerful in reservoir

unsteady flow problems, obtaining such functions is a great challenge.

According to Newman (1936), the solution of a three-dimensional heat

conduction problem can be represented as the product of three one-dimensional

problem solutions. When it refers to the pressure rather than heat, the solutions

can be visualized as the product of instantaneous functions from the one-

dimensional (or one- and two-dimensional) supposed sources/sinks if the real

one is the intersection of several hypothetical sources/sinks. For instance, an

infinite line source can be regarded as the intersection of two infinite plane

sources that are perpendicular to each other.

As for finite reservoirs, the images method is useful. The source/sink

functions for a source/sink that is located in a reservoir with straight boundaries

is the algebraic sum of the source itself and its images. All the necessary

source/sink functions are also listed in Appendix A.

26

3.2 Laplace transformation

At early times, Equation 3.3a is not sufficiently accurate. Thus, the Laplace

transformation to the variable, t, is necessary to obtain better numerical

calculation results. For the function f(t), its Laplace transformation is

0

)()( dtetfuLut

. ···································································· (3.4)

Since the expression for pressure drop equals the integration of the product

of two distinct terms, strength of source and source function, the convolution

theory becomes useful in Laplace transformation. For Equation 3.3, its Laplace

transformation can be rewritten as:

deMSuq

CuMp

u

t

0

,1

, . ··············································· (3.5)

The solutions in Laplace domain should be transformed into real-time

domain for analysis. Stehfest algorithm (Stehfest, 1970) inversely transforms the

Laplace-domain solutions into real-time domain.

3.3 Continuity conditions

In order to solve non-linear mathematical models, it is necessary to

discretize the reservoir into segments. Zeng and Zhao (2009) gave details about

semi-analytical methods of discretizing reservoir system and coupling solutions.

Such semi-analytical methods effectively eliminate the truncation error in

numerical simulation while achieving the same accuracy as the analytical

method.

27

Solutions at adjacent segments are coupled together with continuity

conditions, which include pressure- and flux-continuity. Supposing that two

segments, i and 1

i , are adjacent, the pressure and flux continuity conditions

in Laplace domain are

1,11,,,

)()(

iiiiii

upup , ······························································ (3.6a)

and

1,11,,,

)()(

iiiiii

uquq . ······························································· (3.6b)

Equations 3.6a and 3.6b cannot be applied when the two adjacent

segments are in different Laplace domains. Based on Zeng’s (2008) work, the

continuity conditions for segments within different Laplace domains are derived

according to the Stehfest inverse Laplace transformation algorithm. The

pressure in the real-time domain should be consistent with

iiiikk

tptp

,,

)()( , ································································· (3.7)

where 1

kkkttt . The Stehfest algorithm for inverse Laplace transformation

shows (Zeng, 2008)

L

Lii

Lii

N

i jiLkkupV

ttp

,,)(

2ln)( , ····················································· (3.8)

and

L

Lii

Lii

N

i jiLkkupV

ttp

,,)(

2ln)( . ················································ (3.9)

Substitution of Equations 3.8 and 3.9 in Equation 3.7 gives

L

Lii

L

L

Lii

L

N

i jiLk

N

i jiLkupV

tupV

t ,,

)(2ln

)(2ln

, ·································· (3.10)

28

where L

N is an integer controlling the number of terms in inverse Laplace

transformation. Therefore, the pressure-continuity condition for two adjacent

segments in different Laplace domains is

iiL

iiL

jkjkup

tup

t

,,

)(1

)(1

. ······················································· (3.11)

With the same procedure, the flux-continuity condition for two adjacent

segments in different Laplace domains can be expressed as (Zeng, 2008):

iiL

iiL

jkjkuq

tuq

t

,,

)(1

)(1

. ························································ (3.12)

The above continuity conditions are derived based on the Stehfest algorithm,

which requires that the functions in real-time domain have no discontinuities,

salient points, sharp peaks, or rapid oscillation. Therefore, the continuity

conditions derived above are applicable only when the pressure and flux

functions are continuous in the time domain of interest.

3.4 Constructing and solving linear equation systems

Applying the pressure- and flux-continuity conditions at each interface

between every two adjacent segments generates a linear equation system at

each time step. After solving the linear equation systems, the flux distribution in

Laplace domain can be mapped. Corresponding pressure distribution can be

calculated based on the flux distribution. Finally, the Stehfest algorithm for

inverse Laplace transformation is employed to calculate the flux and pressure

distribution in the real-time domain.

29

3.5 Chapter summary

This chapter shows general semi-analytical methodologies used to solve

non-linear mathematical problems in this thesis. The following chapters will use

such methodologies to deal with different mathematical models.

30

CHAPTER 4

MULTI-STAGE HYDRAULICALLY FRACTURED

HORIZONTAL WELLS

In a post-peak-oil world, oil demand will surpass crude oil production.

Fortunately, the difference between supply and demand can be made up from

an increase in unconventional oil and gas production such as shale gas, tight

gas, and oil, coalbed methane and gas hydrates. The horizontal well multi-stage

fracturing technique makes unconventional reservoir production economically

viable and more efficient.

4.1 Model and algorithm

Figure 4.1 is a diagrammatic representation of a multi-stage fractured

horizontal well in a box-shaped reservoir. In order to derive a semi-analytical

model, the following assumptions are made:

The fractured horizontal well is located in a homogeneous box-shaped

reservoir. All the boundaries are closed boundaries.

The model is derived for single-phase flow.

The fluid flow from the reservoir directly to the horizontal wellbore is

considered.

The reservoir could be isotropic or anisotropic.

The horizontal wellbore is parallel to reservoir boundaries.

31

Figure 4.1―A multi-stage fractured horizontal well in a box-shaped reservoir.

A

B

H

32

The fractures are vertical, symmetrical and perpendicular to the horizontal

well. Each hydraulic fracture is equally spaced along the horizontal well.

If the reservoir is anisotropic, the geometric mean of permeabilities from

three dimensions is chosen as the reference permeability. Additionally, the

hydraulic fractures are not necessarily assumed to be the same in properties.

However, creating equally spaced hydraulic fractures with similar properties is a

common practice unless there is significant difference among fractures in field

application. Furthermore, it is very difficult to discern individual fracture

properties from transient pressure data alone (Raghavan et al, 1997).

4.1.1 Dimensionless variables

At first, I will define the dimensionless pressure and time as

Bq

ppkhp

sc

i

D

, ············································································ (4.1)

and

2

rt

D

LC

ktt

, ················································································· (4.2)

where

3zyx

kkkk . ·················································································· (4.3)

In Equations (4.1), (4.2), and (4.3), Lr means a reference length and k becomes

reference permeability.

To consider variable gas compressibility cg (p) and viscosity μ (p) in gas wells’

transient pressure analysis, the adjusted pressure can be used (Olarewaju and

Lee, 1989):

33

dpp

pp

p

pr

r

ar

)(

)(

, ········································································· (4.4)

where pr is a reference pressure and μr and ρr are the viscosity and density

under the reference pressure. Usually, the standard pressure is taken as the

reference pressure.

As Agarwal (1979) indicated, when analyzing build-up test results, it is

useful to make time transformations. However, for drawdown tests, there is no

such need.

The dimensionless distances in x-, y-, and z-direction are defined as

xr

D

k

k

L

xx , ·················································································· (4.5)

yr

D

k

k

L

yy , ············································································ (4.6)

and

zr

D

k

k

L

zz , ············································································· (4.7)

For the hydraulic fracture system, the dimensionless fracture conductivity

and fracture diffusivity are defined as

r

ff

fD

kL

wkC , ············································································· (4.8)

and

k

C

C

kC

t

tff

f

, ·········································································· (4.9)

34

respectively, where kf is the fracture permeability and Ctf is the fracture total

compressibility.

The dimensionless flow rates are expressed as :

Q

qq

f

fD , ················································································· (4.10)

Q

Lqq

rrf

rfD , ·············································································· (4.11)

Q

Lqq

rrh

rhD . ············································································· (4.12)

4.1.2 Mathematical model

The mathematical model depicting the fluid flow in a reservoir with a

fractured horizontal well consists of three parts: (1) the fluid flow in the formation,

(2) the flow in the fracture, and (3) the pipe flow in the wellbore (Pipe flow will be

discussed in the next section). Fractures are regarded as plane sources with

non-uniform flux distribution. Compared with the whole reservoir, the horizontal

wellbore is considered to be a line source.

The equation describes the pressure drop in the formation as

t

pC

z

p

y

p

x

pk

t

2

2

2

2

2

2

. ······················································ (4.13)

For fractures, the inner boundary conditions is

),( tyqx

pkh

rf

xxF

,

ff

yyy ··················································· (4.14)

and the initial condition

iptzyxp )0,,,( . ····································································· (4.15)

35

For the horizontal wellbore, the inner boundary condition is expressed as

)(2

2,

tqr

prk

rhH

zrrw

, ······························································ (4.16)

and the outer boundary conditions can be written as

0),,,(

x

tzyxp, at 0x or Ax , ·················································· (4.17)

0),,,(

y

tzyxp, at 0y or By , ················································· (4.18)

0),,,(

z

tzyxp, at 0z or Hz . ·················································· (4.19)

The mathematical model for the pressure drop inside fractures can be derived

as

t

pC

hw

tyq

y

pkf

tff

f

rfff

,

2

2

, ······················································ (4.20)

And the boundary conditions as,

hBy

fpp

2

··············································································· (4.21)

0

f

yy

f

y

p ·············································································· (4.22)

and initial condition as

ifptyp )0,( . ········································································ (4.23)

The dimensionless mathematical model for fluid flow in the formation is

represented as

D

D

D

D

D

D

D

D

t

p

z

p

y

p

x

p

2

2

2

2

2

2

, ·························································· (4.24)

36

With the inner boundary conditions,

),(DDDrf

xxD

Dtyq

x

p

DFD

,

fDDfD

yyy ······································ (4.25)

DrhD

D

HzrrD

D

Dtq

H

r

pr

DDwDD

22

,

, ··················································· (4.26)

and the outer boundary conditions,

0

D

D

x

p, at 0

Dx or

DDAx , ······················································· (4.27)

0

D

D

y

p, at 0

Dy or

DDBy , ······················································ (4.28)

0

D

D

z

p, at 0

Dz or

DDHz . ······················································ (4.29)

and the initial condition,

0)0,,,( DDDDD

tzyxp , ······························································ (4.30)

The corresponding dimensionless equations for flow inside hydraulic

fractures can be reformulated as

D

fD

fD

rfD

D

fD

t

p

CC

q

y

p

1

2

2

, ······························································ (4.31)

and boundary conditions,

hDBy

fDpp

D

D

2

·········································································· (4.32)

0

DfD

yyD

fD

y

p··········································································· (4.33)

and initial condition,

37

0)0,( DDfD

typ . ······································································ (4.34)

4.1.3 Algorithm

A multi-stage fractured horizontal well in a box-shaped reservoir is

separated into four sub-systems: formation/fracture sub-system,

formation/horizontal well sub-system, fracture sub-system, and horizontal

wellbore sub-system. Each sub-system is discretized and solved individually in

Laplace domain. As such, the solutions for the above sub-systems in Laplace

domain are coupled based on the interface pressure- and flux-continuity

conditions. Finally, Stehfest’s (1970) Laplace inversion algorithm is applied to

determine the corresponding pressure distribution in the real-time domain.

4.1.3.1 Fracture sub-system solution

In this fracture sub-system, each hydraulic fracture is supposed to have a

finite-conductivity. Fluid flow in the fractures is simplified as 1D linear flow, which

is similar to fractured vertical wells. Furthermore, each hydraulic fracture is

discretized into equal segments. In each segment, fluid flows from the reservoir

qrf beside the inside flow qf (Figure 4.2a). At interfaces between adjacent

segments, solutions are coupled with equal flow rate and pressure.

Based on the results of Van Kruysdijk (1988), the dimensionless pressure at

xD (xDi-1 <xD < xDi) in Laplace domain for segment i is (Van Kruysdijk, 1988):

rfDiifDiifDiiDfDiqCqBqAuxp

1),( , ni 1 ··································· (4.35)

where

38

Cuxx

Cuxx

DiD

fD

i

DiD

DiD

e

e

Cuxx

CuC

A

)(

)(2

11

1

1

cosh21

, ················· (4.36)

Cuxx

Cuxx

DDi

fD

i

DDi

DDi

e

e

Cuxx

CuC

B

)(

)(2

1

cosh21

, ···················· (4.37)

uC

CC

fD

i

. ·············································································· (4.38)

In this case, u is the Laplace variable.

Usually, choked fracture skin effect refers to the presence of a damaged

near-wellbore zone with a reduced conductivity in a hydraulic fracture (Romero

et al., 2003). The extra pressure drop caused by the choked skin factor sck can

be expressed as:

u

sp

ck

D , ··············································································· (4.39)

And (Romero et al., 2003)

1,ckf

f

f

ck

ckk

k

x

xs

, ···································································· (4.40)

where kf, ck is the choked fracture permeability and xck is length of the choked

zone.

4.1.3.2 Horizontal well sub-system solution

In previous papers (Babu and Odeh, 1988; Ozkan et al., 2009), the

horizontal wellbore is always simplified as an infinite-conductivity “pipe”. In this

work, the wellbore pressure drop, which has never been included in fractured

39

(a)

(b)

(c)

Figure 4.2―Diagram showing fracture and horizontal wellbore discretization.

qout qin

qr

40

horizontal well model before, is discussed in detail. The pressure drop is the

result of frictional, radial influx and accelerational effects. The wellbore is divided

into m segments as shown in Figure 4.2b. Center points of each segment are

taken as the reference nodes among which the pressure drops are calculated.

The frictional pressure loss ∆pfric can be calculated as (Brown, 2003):

2

2v

d

Lfp

fric

, ······································································ (4.41)

where f is friction factor and v is fluid velocity. The friction factor is the explicit

approximation of the implicit Colebrook-White friction factor (Swamee and Jain,

1976).

Fluid flow directly from the reservoir and hydraulic fractures would cause a

pressure drop in the horizontal wellbore. Ouyang et al. (1996) proposed that the

radial influx effect could be incorporated with the frictional effect by introducing a

modified friction factor f*. For laminar flow, I have

6142.0

Re,

Re

*4303.01

64

wN

Nf , ·························································· (4.42)

and for turbulent flow (Ouyang et al., 1996),

3978.0

Re,0

*0153.01

wNff , ···························································· (4.43)

where f0 is the friction factor without consideration of the radial influx and

NRe,w=qrρ/πμ is the inflow Reynolds number. It is clear that the influx rate can

increase the pressure drop in the laminar flow regime. Conversely, pressure

drop is reduced in turbulent flow.

From the horizontal well toe to heel, fluid velocity keeps increasing, and this

accelerational effect causes an extra pressure drop. For each segment, the

41

pressure drop caused by the accelerational effect can be expressed as a

function of the change in momentum ∆M:

)(22

inoutaccvv

A

Mp

. ·························································· (4.44)

Here vin and vout are fluid flow velocities into and out of the wellbore segment,

respectively.

The total pressure drop is the sum of the aforementioned three kinds of

pressure loss. Then, the total pressure loss is converted into Laplace domain by

dividing the Laplace variable u.

4.1.3.3 Formation/Fracture and Formation/Horizontal Well Sub-system

Solutions

The source function method is effective in dealing with a wide variety of

reservoir flow problems. Instantaneous source functions proposed by Gringarten

and Ramey (1973) are applied in this paper. Compared with the whole reservoir,

each hydraulic fracture segment with unit source strength can be treated as a

finite vertical plane source that can be written as the intersection of an infinite

slab source and an infinite plane source in infinite slab reservoirs:

0

),(D

ut

yxDDdteSSuyp D . ··························································· (4.45)

where

1

2

22)(

exp

)(cos

)(cos

11

n

D

DD

D

DD

D

DD

x

A

tn

A

xxn

A

xxn

AS

, ······················· (4.46)

42

n

DD

DDD

DD

DDD

DD

DDD

DD

DDD

y

t

nByyerf

t

nByyerf

t

nByyerf

t

nByyerf

S

2

2

2

2

2

2

2

2

2

1 . ················· (4.47)

Every horizontal wellbore segment is regarded as a line source in a box-shaped

reservoir. The corresponding line source function is a combination of an infinite

slab source function and two infinite plane source functions.

Filter cake and polymer accumulation in fractures reduce the permeability

normal to the fracture wall, which is known as the fracture-face skin effect. As

such, a corresponding extra pressure drop should be added to each segment

solution in the formation/fracture sub-system. To this end, Cinco-Ley and

Samaniego (1981) described the fracture-face skin factor in terms of damage

penetration and damaged permeability (Cinco-Ley and Samaniego, 1978):

)1(2

s

s

f

ff

k

kw

xs

, ···································································· (4.48)

where ws is fracture-face skin zone width and ks is reduced permeability. The

dimensionless pressure drop caused by the facture-face skin is the product of

the skin factor and dimensionless flow rates from the reservoir to the fractures.

For a horizontal well, any pressure field deviation from perfect radial flow in

the well vicinity can be accounted for by the horizontal well skin factor. The

corresponding dimensionless pressure drop is calculated as the same way as

fracture skin factors.

43

4.1.3.4 Coupling the solutions

The solutions from different sub-systems are coupled together between any

two adjacent segments. For the pressure calculation in the horizontal wellbore,

unknown wellbore flow distribution becomes necessary when coupling solutions.

Therefore, a hypothetical uniform initial flow distribution along the wellbore and

an iterative process are applied in the model. The procedure of coupling all

solutions to generate a linear equation system is shown in Figure 4.3.

4.2 Model validation

KAPPA Ecrin, a commercial well-testing software, verified my model. The

wellbore pressure drop cannot be considered in KAPPA analytical models.

Therefore, results without horizontal wellbore pressure drops from my work were

compared with those obtained from KAPPA Ecrin. Figure 4.4 shows the

comparison of the two methods for a horizontal well with four identical hydraulic

fractures. The input parameters also appear in this figure.

The comparison suggests the results from this work are consistent with the

results from the commercial software. As expected, there are four flow regimes

for the multiple-fracture system in Figure 4.4: bilinear/linear flow, early-radial

flow, compound-linear flow (CLF), and pseudo-radial flow, which is similar with

the conclusion of Chen and Raghavan (1997). Therefore, the mathematical

model and its algorithm are reliable. Moreover, it is difficult to track the early-

time pressure behaviour in KAPPA Ecrin. However, the model provided here

can provide very early-stage pressure behaviour and guarantee accuracy.

44

Figure 4.3―Flow chart for modeling and solving process.

newDDqq

,

ebsqqDnewD

,

No

Laplace

Inversion

BqAD

Initial influx

distribution along

the horizontal well

Vector B: consisting of

hDp between adjacent

horizontal segments

Coefficient Matrix

A

Solutions of each

fracture segment

in fracture sub-

system

Solutions of each

fracture segment in

formation/fracture sub-

system

Solutions of each

segment in

formation/horizontal

well sub-system

Yes

ttt newDinitialDqq

,,

45

Figure 4.4―Model validation with KAPPA Ecrin.

0.01

0.1

1

10

100

1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2

Pw

D, d

Pw

D/d

lnt D

tD

This Work

Kappa Results

rw 0.05m h 50 m porosity 0.2 viscosity 0.001Pa.s Q 100 m3/day k 0.004mD Xf 50m kfwf 6×10-12m3

46

4.3 Results and discussion

4.3.1 Effect of fluid flow from reservoir to horizontal wellbore

Figure 4.5 compares the pressure behaviour with and without the

contribution of the horizontal wellbore for 3-, 10-, and 20-stage fractured

horizontal wells, respectively. Correspondingly, 3, 10, and 20 are the stage

numbers for these fractured horizontal wells. Here, wellbore pressure drops

were not taken into consideration. Fluid flow directly from the reservoir could blur

the differentiation between bilinear/linear and early-radial flow when stage

number was small. Moreover, fluid flow directly from the reservoir reduces the

pressure drop slightly until the pseudo-steady-state (PSS) regime is reached.

Horizontal wellbore contributions are weakened greatly when more fractures

are created. A reasonable explanation can be found by comparing fracture

contributions to the total production. Figure 4.6 shows the production proportion

of fractures and the horizontal wellbore to the total production. The upper lines

correspond to fractures while the lower ones are for wellbores. Contributions

from a horizontal wellbore increase over time. In fact, the ratio increases from

0.09 to 0.33 for a 3-stage fracture horizontal well. Although a horizontal wellbore

produces more oil, fractures produce much more fluids than wellbores. This

phenomenon becomes more evident when the fracture stage number increases

to 20; at that level, the fracture production ratio remains around 0.99. Actually,

the stages of some open-hole multi-stage fractured horizontal wells can reach

more than 20. Therefore, the flow directly into the wellbore can be ignored when

the stage number for a fractured horizontal well is large enough.

47

Figure 4.5―Effect of fluid flow from reservoir to the wellbore.

1E-3

1E-2

1E-1

1E+0

1E+1

1E+2

1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2

Pw

D

tD

3 Fractures

3 fracutes, considering wellbore

10 fractures

10 fractures, considering wellbore

20 Fractures

20 Fractures, considering wellbore

48

Figure 4.6―Flow distribution along the fracture and wellbore.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

Pro

po

rtio

n o

f to

tal p

rod

cu

tio

n

tD

3 Fractures

10 Fractures

20 Fractures

49

4.3.2 Effect of horizontal wellbore pressure drop

Figure 4.7 compares the pressure behaviour from a 6-stage fractured

horizontal well under different production rates and reservoir permeabilities

(k=400 md, 40 md and 0.4 md). The pressure loss along the horizontal wellbore

can increase the early-stage pressure drops slightly but only when reservoir

permeability becomes large enough (k>100 md). Certainly the conclusion would

be different under low-permeability condition. There is no significant difference

between pressure drops—with and without considering wellbore pressure loss—

when permeability is less than 1 md. Furthermore, a higher production rate can

increase the pressure deviation if considering the wellbore pressure drop.

Similarly, when production rate is high but permeability is very low, the

drawdown ratio ( 0.36 kPa with wellbore contribution vs. 0.36 kPa without

wellbore contribution at k=0.4 mD) at t=0.001 day (Figure 4.7) is close to unit,

meaning it is the permeability rather than production rate that is the key

parameter that amplifies the horizontal wellbore pressure loss. For tight

formation and shale gas reservoirs, permeability is usually in the range of

microdarcies or below, and, as such, wellbore pressure loss has little influence

on the transient pressure behaviour. Therefore, the effect of pressure drops

inside the horizontal wellbore of multi-stage fractured horizontal wells can be

ignored in tight and shale formations.

50

(a)

(b)

Figure 4.7―Effect of horizontal wellbore pressure drop.

1E-2

1E-1

1E+0

1E+1

1E+2

1E+3

1E-4 1E-2 1E+0 1E+2

Pw

D, d

Pw

D/d

lnt D

tD

k=400md

k=40md

k=0.4md

Without horizontal well pressure drop

Q=100 m3/day

1E-2

1E-1

1E+0

1E+1

1E+2

1E+3

1E+4

1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2 1E+3 1E+4

Pw

D, d

Pw

D/d

lnt D

tD

k=400md

k=40md

k=0.4md

Without horizontal well pressure drop

Q=1000 m3/day

51

4.3.3 Effect of fracture stages

For multi-stage fractured horizontal wells, fracture stages have significant

influence on productivity and economic benefits. Here, fracture stage effects

were analyzed in three steps without changing the fracture width.

4.3.3.1 Constant fracture half-length

Figure 4.8 gives the pressure responses and their logarithmic derivatives

with different fracture stages and the same fracture half-lengths for a 600 m

horizontal well. Figure 4.9 presents the dimensionless relative productivity

indexes over time under different fracture numbers for the fractured wells. PSS

productivity index, J, has been accepted as a criterion for evaluating fracture

performance, whose dimensionless form is defined as (Zeng, 2008):

wfD

D

pJ

1 , ···············································································(4.49)

where pwfD is the dimensionless flowing bottomhole pressure.

According to Figures 4.8 and 4.9, a larger stage number makes a larger

total fracture volume, which leads to less pressure drop and higher productivity.

Such well productivity increment decreases with increasing fracture stages while

also increasing multi-stage fracturing cost. Hence, optimum fracture stage

numbers (fracture spacing) must exist under constant fracture half-lengths .

In Figure 4.9, the wells’ productivities are linearly related to the fracture

numbers in the early-stage production. For example, the 40-stage fractured

horizontal well works as well as 40 separate fractured vertical wells in the

bilinear/linear flow regime. Therefore, to analyze early-time pressure data of

52

fractured horizontal wells, a conventional analysis method for bilinear/linear flow

can be easily applied.

After the bilinear/linear flow, interference between fractures appears. The

duration of bilinear/linear flow becomes shorter and interference appears earlier

with more fractures along the wellbore. In fact, when the number of fractures

increases to 40, pressure and pressure derivative curve slopes are close to units

in the interference period after tD=0.002. This apparent BDF is similar to the PSS

flow. In fact, this phenomenon means the appearance of SRV. The key

parameter controlling SRV is the ratio of the fracture spacing and fracture half-

length. When the fracture spacing is significantly smaller than the half-length,

SRV appears clearly.

4.3.3.2 Constant total fracture volume

Figure 4.10 compares the pressure and corresponding derivative curves

under the constant total fracture volume. Figure 4.11 provides the relative

productivity index as it changes over fracture stages and time. As shown in

Figures 4.10 and 4.11, the pressure drop is reduced with more fractures in early-

stage production while productivity growth tends to slow down. Correspondingly,

more fractures will cause a high production rate if the bottomhole pressure (BHP)

is constant.

The appearance of SRV reverses such a trend. For example, the

dimensionless pressure derivative of the well with10 fractures tended to surpass

one with less than 10 fractures after tD=0.002. This suggests that the pressure

drop of a 10-stage fractured horizontal well increases more quickly than a well

53

with less than 10 fractures. Furthermore, the 10-stage fractured well’s

productivity would become lower than that of the 4-stage fractured horizontal

well (Figure 4.11). Therefore, in the long term, creating more fractures is not

always better. As discussed in 4.3.3.1, there also exists an optimum

combination of fracture stage and fracture half-lengths when the total fracture

volume remains constant.

4.3.3.3 Non-uniform fractures

For many, natural fractures in shale formations are critical in controlling

shale gas well productivity (Gaskari and Mohaghegh, 2006). Unpropped

fractures induced by hydraulic fracturing treatments can also exist alongside

pre-existing open natural fractures.

In Figure 4.12, I simply added small fractures beside each hydraulic fracture

to simulate natural and induced fractures for a 7-stage fractured horizontal well.

These seven hydraulic fractures have different lengths. At first, 7 small fractures

were added on the right of hydraulic fractures as natural/induced fractures. Then,

another 7 larger fractures were created. Each natural/induced fracture can have

different half-lengths and conductivities but, for the sake of simplicity, 7 small

and 7 larger natural/induced fractures possessed similar properties.

Figure 4.13 shows the pressure responses of natural/induced fractures. It is

clear that natural/induced fractures work as if improving hydraulic fracture

conductivities. So, the hydraulic fracture conductivity can be overestimated if

existing natural/induced fractures are ignored. However, it is difficult to estimate

54

natural/induced fractures with only transient pressure behaviour data shown in

Figure 4.13.

55

Figure 4.8―Effect of fracture stages when xf is constant.

1E-4

1E-3

1E-2

1E-1

1E+0

1E+1

1E+2

1E-6 1E-4 1E-2 1E+0 1E+2

Pw

D d

Pw

D/d

lnt D

tD

xf=250m,n=2

Xf=250m, n=4

Xf=250m, n=10

Xf=250m, n=40

56

Figure 4.9―Effect of fracture stage on the productivity index with costant xf.

1

3

5

7

9

11

13

15

17

19

21

0 5 10 15 20 25 30 35 40 45

JD

r

Fracture Stage

tD=1E-6

tD=1E-5

tD=1E-4

tD=1E-3

tD=1E-2

tD=1E-1

tD=1E0

tD=1E1

57

Figure 4.10―Effect of fracture stages when fracture volume is constant.

1E-3

1E-2

1E-1

1E+0

1E+1

1E+2

1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2

Pw

D

dln

Pw

D/d

lnt D

tD

Xf=250,n=2

Xf=167,n=3

Xf=125,n=4

Xf=100, n=5

Xf=83,n=6

Xf=71,n=7

Xf=62.5,n=8

Xf=55.5,n=9

Xf=50,n=10

58

Figure 4.11―Effect of fracture stages on the productivity indxex with cosntant Vf.

0.8

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

0 2 4 6 8 10 12

JD

r

Fracture Stages

tD=1E-6

tD=1E-5

tD=1E-4

tD=1E-3

tD=1E-2

tD=1E-1

59

Figure 4.12―Non-uniform fractures along the horizotnal wellbore.

60

Figure 4.13―Effect of non-uniform fractures on the pressure behaviour.

0.001

0.01

0.1

1

10

100

1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2

Pw

D d

Pw

D/d

lnt D

tD

Series1

Series3

Series5

7 Fractures

14 Fractures

21 Fractures

61

4.3.4 Effect of gas desorption

Gas desorption is an important phenomenon in shale gas and coalbed

reservoirs. During reservoir depletion, absorbed gas desorbs from the surface

because of the thermodynamic equilibrium when pressure changes. Gas

desorption is governed by van der Waals’ force, which makes hydrocarbon

molecules detach from the solid surface of adsorbents. In shale gas reservoirs,

the pore size is almost in nanoscale. The exposed area in nanopores is much

larger than that of conventional reservoirs. Therefore, a large amount of gas is

adsorbed at the large surface area of Kerogen materials in the reservoir far

before drilling.

According to Kucuk and Sawyer (1980), the dimensionless desorption

storability ratio is defined as

tt

gD

D

C

cR

, ············································································· (4.50)

where gD

c is the desorbed gas storativity and ϕD=RT( dC/d(p/z)) and C is the gas

condensation at the pore wall surface. D

R can be easily added to the existing

mathematical model as an extra source term in the diffusivity. Therefore, a

similar PDE can be derived for gas reservoir including desorption

t

p

k

RC

y

p

x

paDtaa

)1(

2

2

2

2

, ···················································· (4.51)

Figure 4.14 illustrates the effect of gas desorption on the pressure response.

The dimensionless desorption storability ratio represents how much gas is

desorbed when compared with free gas. As shown in Figure 4.14, the pressure

62

drop with gas desorption is lower than without desorption. As a result, gas

desorption acts as an extra source in the reservoir and puts off the appearance

of BDF.

63

Figure 4.14―Effect of gas desorption on pressure behaviour.

1E-3

1E-2

1E-1

1E+0

1E+1

1E+2

1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2

Pa d

Pa/d

lnt D

tD

No desorption

RD=1.0

RD=2.0

RD=3.0

64

4.3.5 Effect of skin factors

Figure 4.15 shows pressure and derivative curves with different skin factors

for a 3-stage fractured horizontal well. Here, I assumed that the skin factors

along fractures and the wellbore remain constant. Actually, I can generate the

pressure data for varying skin factors. However, I cannot discern different skin

factors along fractures and along the wellbore in field application. In Figure 4.15,

fracture choked skin factor sck increases the pressure drop significantly and is

followed by the fracture face-skin factor sff. With the same skin factor value, the

horizontal well skin factor sh only increases the dimensionless pressure a little.

The surface area of fractures is much larger than in the horizontal wellbore and

most of the fluid flows to the wellbore through fractures. Therefore, relatively

small fracture skin factors can make a more significant difference than the same

horizontal wellbore skin factors. Moreover, fracture-face skin factors can mask

the bilinear/linear flow regime and make it difficult to catch bilinear/linear flow in

log-log plots of pressure and pressure derivatives.

Figure 4.16 compares the flow distribution with different skin factors at

tD=10-6. The effect of fracture-face skin factor seems more complex than the

choked skin factor. The fracture-face skin factor reduces the amount of fluid

from the reservoir to fractures and increases pressure drops. Moreover, the

fracture-face skin factor tries to reduce the influx difference to reach an even

flow distribution along fractures.

The choked skin factor increases the pressure drop inside fractures since

the choked skin factor reduces the conductivities of fractures in the vicinity of the

65

well. Also, the choked skin factor reduces the amount of fluid influx from the

reservoir to fractures. However, it has no influence on the influx distribution

pattern along fractures. Obviously, the horizontal well skin factor also has little

influence on the influx along fractures.

As far as the total skin factor is concerned, the extra pressure drop is not

equal to the sum of the extra pressure drop caused by each skin factors.

Ultimately, different skin factors can interact with each other and finally strike a

balance.

66

Figure 4.15―Effect of skin factors on the pressure behaviour.

1E-3

1E-2

1E-1

1E+0

1E+1

1E+2

1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2

Pw

D d

Pw

D/d

lnt D

tD

No skin factor

Sff=1.0

Sck=1.0

Sh=1.0

Total skin factor

67

Figure 4.16―Effect of skin factors on the flow distribution along the fracture.

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.0 0.2 0.4 0.6 0.8 1.0

qrf

D

XfD

No skin factor

Sff=1.0

Sck=1.0

Sh=1.0

Total skin factor

tD=1E-6

68

4.4 Field examples

Using my model, I performed a pressure transient analysis based on field

data. The well is a multi-stage fractured horizontal well in a typical shale gas

reservoir in Sichuan, China. The pressure data came from two build-up tests

after almost one-year production. Figure 4.17 shows the gas production rates

from August 2, 2011, to May 30, 2012. Table 4.1 summarizes the basic input

parameters for this example.

4.4.1 No.1 Build-up test analysis

At first, I completed a qualitative type-curve analysis. Since the duration of

the 1st build-up test is short and the reservoir permeability is extremely low, I

found no characteristic pressure responses such as half-slope trend in CLF and

the unit-slope line at late times. In addition, I cannot detect the fracture half-

length directly from the plot of ∆pa―t1/2 because no linear-flow trend appears in

the build-up test. However, what I can distinguish from the log-log plot is the

bilinear flow. Accordingly, I plotted ∆pa vs. t1/4

in Figure 4.18. The slope of the

straight line through the data is m=90 MPa/hr1/4. Using this slope, I estimated the

fracture conductivity wfkf =2.88×10-15 m3.

Next, I compared the real field data with different type curves to find the

most appropriate fracture half-length and fracture conductivity. Figure 4.19

shows the matching results with my model. In general, the match is satisfactory.

I chose a constant value Cw=9.20×10-6 m3/Pa (0.4 bbl/psi) to account for the

wellbore storage effect. In fact, in Figure 4.19, I can see that the wellbore

storage factor is not strictly constant and varies from 7×10-6 to 1×10-5 m3/Pa (0.3

69

to 0.5 bbl/psi) during the build-up test. Furthermore, the fracture half-length was

estimated as 137 m and the fracture conductivity was 4×10-16 m3. This is the

best match for the whole data set. As such, I can also conclude that the area of

the SRV should be 67.7 acres (L×2 x f =274000 m2). In fact, the fracture

conductivity is not as good as the one shown in my type curve. Higher fracture

conductivity can make the pressure derivative bend upward with a greater angle

as seen at the end of this test. However, in Figure 4.19, the derivative is almost

flat at the end. Therefore the fracture conductivity should be a little lower than

my expectation.

4.4.1 No.2 Build-up test analysis

After analyzing the No.1 build-up test, I recognized that using specific lines

to evaluate fractures is limited for shale gas wells. Thus, I directly employed the

type curve matching method to conduct quantitative analysis for the No.2 build-

up test. Figure 4.20 shows the matching results with my model. In this case, the

wellbore storage factor increases to 2.3×10-5 m3/Pa (1.0 bbl/psi), while the

average fracture’s half-length was estimated as 100 m. Moreover, its

conductivity was recalculated as 6.2×10-16 m3. Compared to the No.1 analysis

results, the half-length is shorter and conductivity is higher. This is reasonable.

In fact, part of the proppant flows back to the wellbore along with gas; therefore

the corresponding fracture section is almost closed without proppant support,

but the remaining part of the fracture swells after extra proppant squeezes in.

70

Figure 4.17―Field production data.

0

2000

4000

6000

8000

10000

12000

14000

16000

0 1000 2000 3000 4000 5000 6000 7000

Pro

du

ctio

n r

ate

, m3 /

Day

Time,hr

71

TABLE 4.1―BASIC INPUT PARAMETERS

Initial Pressure, pi (MPa) 18.62

Formation thickness, h(m) 38

Formation temperature, T(°C) 65

Porosity, 0.03871

Average formation permeability,

k(m2) 3.41×10-20

Wellbore radius, rw (m) 0.05

Specific gravity, γ 0.554

Horizontal well length,L (m) 1000

Number of hydraulic fractures, n 12

Initial Gas Saturation, Sg (fraction) 40.14

72

(Original in color)

Figure 4.18―Plot of adjusted pressure vs. fourth root time.

0

100

200

300

400

500

600

0 1 2 3 4 5 6

Ad

juste

d p

ressu

re,

Mp

a

Fourth root of time

Slope m=90

73

(Original in color)

Figure 4.19―Type curve match for the first test.

1

10

100

1000

0.01 0.1 1 10 100 1000

Ad

juste

d p

ressu

re c

han

ge,

MP

a

Adjusted time, hr

Observed

Semi-analytical model

74

(Original in color)

Figure 4.20―Type curve match for the second test.

0.1

1

10

100

0.01 0.1 1 10 100

Ad

juste

d p

ressu

re c

han

ge,

MP

a

Adjusted time, hr

Field data

Semi-analytical result

75

4.5 Chapter summary

In this chapter, a semi-analytical model has been built to study the transient

pressure behaviour of multi-stage fractured horizontal wells in a rectangular

reservoir. The effect of different parameters on well production has been studied

in detailed. Several conclusions can be derived:

The model and its algorithm were validated to be applicable in transient

pressure analysis for the multi-stage fractured horizontal well. It can be

further extended to study dual-porosity reservoirs and reservoirs with

complex fractures.

The fluid flow directly from reservoir to horizontal wellbore can change the

early-stage pressure and its derivative if fewer fractures are created along

the wellbore (n is in the range of 2 to 10). The pressure drop along the

wellbore can be ignored for typical multi-stage fracture horizontal wells in

tight formations and shale gas reservoirs.

When fracture half length is constant, there exists an optimum fracture

stage number considering the increase of the well’s productivity. When

total fracture volume is constant, there is also an optimum fracturing

design to obtain the highest profit. Therefore, an optimum fracturing

treatment must exist if all the necessary parameters are considered.

Different kinds of skin factors influence the pressure behaviour in different

aspects. Among all skin factors, the choked skin factor is the most

damaging to fractured wells’ productivity.

76

In field applications, pressure response might not display any

characteristic flow behaviour at early times for shale-gas wells.

Techniques based on the slopes of characteristic lines might not be

guaranteed. Type curve matching can be a candidate for analyzing the

overall pressure behaviour with a model to obtain a better solution.

77

CHAPTER 5

HYDRAULICALLY FRACTURED WELLS WITH STRESS-

SENSITIVE CONDUCTIVITIES

This chapter aims to derive solutions for the non-linear mathematical model

and accordingly describe the pressure behaviour of fractured wells with stress-

dependent conductivities. A box-shaped homogeneous reservoir with all closed

boundaries is considered, and the fractured well is located at the center of the

reservoir. In order to simulate the well/reservoir system by a mathematical

model, a few assumptions must be made first:

1. The porous media is isotropic, homogeneous, and bounded with upper

and lower impermeable strata.

2. Reservoir parameters such as porosity ϕ and permeability k are kept

constant.

3. Both the reservoir and hydraulic fractures are filled with a single fluid,

which could be gas or slightly compressible fluid with constant viscosity.

4. All the hydraulic fractures are vertical, symmetric and fully penetrate the

formation. For the multi-stage fractured horizontal well, properties of each fracture

at the initial time are assumed to be identical.

5. The fluid flow in the system is subject to Darcy’s law and the fluid flow in

hydraulic fractures is described with the Forchheimer equation.

The fracture conductivity is determined by fracture width and fracture

permeability. Although fracture width wf changes during production, considering

78

fracture permeability stress-dependency seems more practical based on the

literature review and experimental work. Also the fracture diffusivity Cη changes

in a way that is consistent with the conductivity.

5.1 Model and algorithm

The mathematical model for hydraulically fractured wells with stress-

sensitive conductivities is similar to those with constant conductivities. The

comprehensive reservoir system consists of two sub-systems: the fracture sub-

system and the matrix sub-system. The contribution from the horizontal wellbore

is ignored because of its minor influence. As shown in Figure 5.1, each fracture

(if multiple fractures are created along the horizontal wellbore) is divided into

several segments, and each segment owns a uniform influx from the

corresponding matrix segment, qrfi,j in m3/(s∙m), and two node fluxes from

adjacent fracture segments, qfi,j-1 and qfi,j in m3/s. At the end of each time step,

the fracture segment solutions in Laplace domain are expressed as a linear

combination ofqfi,j-1,qfi,j andqrfi,j, which corresponds to a linear format solution

withqrfi,j for the adjacent matrix segments. Then, the pressure equivalent and

flux continuity conditions are applied to couple solutions at interfaces of any two

adjacent segments from the two sub-systems at the same time point in Laplace

domain. Also fracture segments closest to the wellbore in different fractures are

coupled by equivalent pressure values, and the sum of all fluid flow into

fractures should be the constant production rate Q. As a result, fluxes at each

interface and the pressure distribution in the reservoir can be achieved.

79

However, the differences of such a stress-sensitive model from the model

with constant fracture conductivities are still obvious. For the fracture sub-

system, although each fracture’s conductivity remains constant during each time

step (Figure 5.2), they become different at different time steps. Meanwhile,

matrix properties remain unchanged at all times. This would cause a remarkable

difference in Laplace transformation for the above two sub-systems, which

brings challenges to coupling solutions on interfaces. Moreover, fracture

conductivities act as a function of stress/pressure, and stress/pressure is

influenced by fracture and reservoir characteristics in turn, which requires an

iterative process in modeling.

In order to couple fracture and reservoir sub-systems, the fracture sub-

system solutions with changing conductivities in Laplace domain should be

reformulated at first. For segment j of fracture i at time step k, I have

D

fD

fD

rfD

D

fD

Dt

p

CC

q

y

p

y

1 for

jDiDjDiyyy

,1,

,

k

DD

k

Dttt

1

············· (5.1a)

bounded with

jfDi

fDyyD

fDq

Cy

p

jDiD

,

1

,

1,

1

,1

jfDi

fDyyD

f

qCy

p

jDiD

D , ··························································· (5.1b)

80

(a) Discretizing each fracture into segments

(b) Flux at segment j for fracture i

Figure 5.1―Discretizing the fracture system

Figure 5.2―Discretizing the fracture conductivity

n n-1 … 3 2 1 0

n n-1 … 3 2 1

81

and zero initial condition. The above mathematical model’s solution is obtained

as in Appendix B:

rfDiDji

jfDiDji

jfDiDjiD

jfDi qydqyCqybuyp*

,1,

*

,,

*

,,

*

)()()(),( , ··············· (5.2a)

where

Cuyy

Cuyy

jDiD

fD

Dji

jDiD

jDiD

e

e

Cuyy

CuC

yb

)(

)(2

,

,

,

,

1

])cosh[(21

)( ············ (5.2b)

Cuyy

Cuyy

DjDi

fD

Dji

DjDi

DjDi

e

e

Cuyy

CuC

yc

)(

)(2

1,

,

1,

1,

1

])cosh[(21

)( ········· (5.2c)

uC

Cd

fD

ji

,. ············································································ (5.2d)

In Equations (5.2a), (5.2b), (5.2c), and (5.2d), u’ is the Laplace variable forD

t .

The matrix sub-system solutions are the same as ones in the previous

model with constant properties. Then, solutions from two sub-systems in

different Laplace domains are coupled using methods mentioned in Chapter 3.

The next problem is to find the correlation between fracture conductivities

(which is defined by stress-dependent permeability) and stress/pressure. I chose

the simple but practical equation from Raghavan and Chin (2002) as shown

below:

ffffif

pd

fi

ppd

fifekekk

. ························································· (5.3)

In Equation (5.3), fi

k is the permeability at initial condition and f

d is the fracture

stress-sensitive characteristic. In fact, both the fracture conductivity f

C and

82

fracture diffusivity

C change when fracture permeability is stress-dependent.

Originally, Equation (5.3) was derived for rocks but was extended to modeling

hydraulic fractures by Berumen and Tiab (1996) and Pedroso et al. (1997). In

fact, my semi-analytical model can incorporate different stress-dependent

conductivity equations if other appropriate expressions are created. In the model,

I use the same df for each stage when the horizontal well is multi-stage fractured.

Furthermore, it is not realistic for the hydraulic fracture permeability to approach

zero, and, therefore, a minimum limit kfmin is set in the model.

The flux distribution and pressure distribution at each time step can be

mapped by coupling solutions and numerical inverse Laplace transformation,

which determines the fracture properties for next time step.

5.2 Model validation

The model and its algorithm were validated through comparison with two

limiting cases with constant fracture properties. The two cases in Figure 5.3

show the pressure behaviour of a 6-stage fractured horizontal well with initial

fracture conductivity CfD=10 and minimum conductivity CfDmin=0.5. The fracture

stress-dependent characteristic f

d was chosen to be 10-7 Pa-1. At the beginning,

pressure behaviour of the stress-sensitive case is consistent with the case

CfD=10. When the fracture conductivity reduces along with pressure, the

pressure behaviour deviates from the typical linear flow with a slope larger than

1/2. When CfD reaches the lowest value, the pressure derivative falls back to the

curve of CfD=0.5. This comparison suggests that the pressure behaviour

83

represented by this semi-analytical model changes gradually among the two

extreme cases, which provides indirect proof for the applicability of this model.

84

Figure 5.3―Model validation.

tD

1e-7 1e-6 1e-5 1e-4 1e-3 1e-2 1e-1 1e+0

Pw

D, dP

wD/d

lnt D

0.0001

0.001

0.01

0.1

1

10

CfD=0.5

CfDi=10,df=1E-7 Pa-1, CfDmin=0.5

CfD=10

85

5.3 Results and discussion

The pressure behaviour of stress-sensitive hydraulic fractures is

represented by following type curves with different combinations of variables.

The variables of interest included fracture conductivity Cf, the minimum fracture

conductivity CfDmin and stress-dependent characteristic df. Effects of these

variables on pressure behaviour were analyzed in detail. Table 5.1 lists the

properties of the fractured well, fluid, and formation used in section 5.3.

5.3.1 Pressure behaviour characteristics

Stress-sensitive hydraulic fracture conductivity’s existence has a significant

impact on the pressure behaviour of fractured wells. Figure 5.4 shows the

dimensionless wellbore pressure response and its derivative on a log-log plot

when fractures are stress-sensitive. At first, the stress-sensitive effect gradually

increases the wellbore pressure drop and pressure derivatives, which deviates

from typical flow regimes, especially bilinear/linear flow. When fracture

conductivity decreases as pressure drops, the slope of the linear flow regime

becomes increasingly higher than 1/2, which is the “signature” of linear flow

when fracture conductivity is constant. Likewise, the slope of bilinear flow

becomes larger than 1/4. Moreover, it should be noted that the amount of the

slope’s increase varies according to different combinations of parameters. If the

fracture conductivity changes rapidly, the wellbore pressure derivative curves

might display a slope that is close to or even larger than unit and/or overstrides

corresponding pressure curves, which can be considered as apparent BDF. This

conclusion is consistent with Berumen and Tiab’s work (1996). This special

86

feature can be considered as a sign of severe fracture conductivity reduction

during production.

When the fracture conductivity reduces to CfDmin and remains constant, the

pressure curve falls back to that with constant conductivity CfD=CfDmin, and,

therefore, pressure derivative has a rapid decline and a hump forms. As shown

in Figures 5.3 and 5.4, the sudden drop appears at the downward section of the

hump in pressure derivative curves, which is caused by the sudden conductivity

change at CfD=CfDmin. In order to eliminate the sudden drop in the pressure

derivative curve, I divided the CfD vs. P into three parts and each part is

described by a specific equation to smooth the conductivity curve. Figure 5.5

gives the smooth conductivity curve and original one. It shows that conductivity

gradually reduced to the minimum conductivity without sudden change in the

smooth conductivity curve.

Figure 5.6 shows the dimensionless pressure and pressure derivative

curves based on the smooth conductivity change in Figure 5.5. The sudden drop

is replaced by a slowly varying downward section in Figure 5.6, which means

that the derivative curve is smooth as long as the conductivity changes gradually.

In Figures 5.5 and 5.6, I applied a new conductivity model that is more

complex than Equation (5.3). But these new equations for CfD vs. P which aim at

smoothing the conductivity change are hypothetical. Large amount of accurate

experimental data are necessary to get reliable new equations for the smooth

conductivity change. Therefore, Equation (5.3) would still be used in the

87

following discussions to describe the stress-sensitive conductivity behavior until

new equations are obtained in the future work.

88

TABLE 5.1―RESERVOIR, WELL, FRACTURE AND FLUID DATA

Reservoir Size, a×b×h 1000m×300m ×50m

Formation Permeability, k 0.004 md

Formation Porosity, ϕ 0.09(fraction)

Total Compressibility, Ct 5×10-8 Pa-1

Fluid Viscosity, μ 0.001Pa∙s

Fracture Width, wf 0.005m

Fracture Stage, n 6

Horizontal Well Length, Lh 600m

Production Rate, Q 1×102m3/day

Reference Length, Lr 600m

89

Figure 5.4―Transient pressure behaviour with stress-sensitive hydraulic

fractures.

tD

1e-6 1e-5 1e-4 1e-3 1e-2 1e-1 1e+0

Pw

D,

dP

wD

/dln

t D

0.001

0.01

0.1

1

10

CfDi

=50,df=2.5E-7 Pa

-1

CfDi

=50,df=0

Effective Pressure, MPa0 5 10 15 20 25

CfD

0

10

20

30

40

50

df=2.5E-7 Pa-1

90

Figure 5.5―Modified smooth conductivity curve.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0E+0 1E+7 2E+7 3E+7 4E+7

CfD

/CfD

i

Effective Pressure, Pa

Conductivity curve with Eq. 5.3

Smooth conductivity curve CfD=CfDi*exp(-df*∆pf), df=10-7 Pa-1 for ∆pf<1.6×107Pa;

CfD=CfDi*(-0.0046∆pf3+0.0734∆pf

2-0.3844∆pf+0.7768)/6.8×106, for

∆pf<3.8×107Pa;

CfD=CfDmin, for ∆pf>3.8×107Pa

91

Figure 5.6―Pressure behavior based on modified conductivity change.

0.0001

0.001

0.01

0.1

1

10

1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1

Pw

D, dP

wD/d

lnt D

tD

df=1E-7 Pa-1

CfDi=5 CfDmin=0.3

92

5.3.2 Effect of degree of stress-sensitivity

Equation (5.3) can be applied in modeling the effect of stress-sensitive

conductivity. The range of df value can be determined through matching different

experimental data. In this work, df values were estimated based on the

experimental results from Friedel et al. (2007), Abass et al. (2009), and Zhang et

al. (2013). Figure 5.7 presents part of the above experimental data and the

regression results with Equation (5.3). The discrete points represent the relative

conductivity measured at corresponding closure stress. The regression results

suggest that df value varies between 5×10-7 Pa-1 and 5×10-8 Pa-1. Therefore, I

choose df within such a range in order to discuss the effect of stress-sensitive

fractures.

In Figure 5.8, I chose four df values increasing from 5×10-7 to 5×10-8 Pa-1

under the same CfDi and CfDmin. The value of df can influence the rate of

dimensionless pressure increase (i.e., the slope of the upward portion of the

hump in pressure derivative curves). The dimensionless times corresponding to

the zenith of four humps are 3×10-5, 1×10-4, 6×10-4, and 2×10-3, respectively,

from largest to smallest df, which demonstrates that it takes less time for fracture

conductivity to reduce to CfDmin when df is larger. It can also be found that a

relatively small change in df value can cause a big difference in the pressure

behaviour. This is reasonable since df indicates the exponential relationship

between fracture conductivity and stress/pressure change. At last, when df has

a small value such as 5×10-8 Pa-1 in Figure 5.8, the slope of long linear flow

regime is too close to 1/2 to discern the effect of stress-sensitive conductivities.

93

On the whole, the pressure deviation will appear sooner and pressure derivative

slopes become larger with bigger df values and vice versa.

94

(Original in color)

Figure 5.7―Normalized fractures conductivities change with stress (Abass et al.,

2009 and Zhang, et al., 2013)

0.001

0.01

0.1

1

0 500 1000 1500 2000 2500 3000 3500 4000

CfD

/CfD

i

Closure stress, psi

df=5.5E-8, 9.5E-8, 1.1E-7,1.2E-7, 1.3E-7 and 2.6E-7 Pa-1

95

Figure 5.8―Type curves showing the effect of stress-sensitive conductivity,

CfDi=50, CfDmin=0.25.

tD

1e-6 1e-5 1e-4 1e-3 1e-2 1e-1 1e+0

Pw

D, dP

wD/d

lnt D

0.001

0.01

0.1

1

10

CfDi=50,df=5E-7 Pa-1

CfDi=50,df=2.5E-7Pa-1

CfDi=50,df=1E-7Pa-1

CfDi=50,df=5E-8Pa-1

CfDi=50,df=0

Effective pressure, MPa

0 20 40 60 80 100 120

CfD

0

10

20

30

40

50

df=5E-7 Pa-1

df=2.5E-7 Pa-1

df=1E-7 Pa-1

df=5E-8 Pa-1

96

5.3.3 Effect of degree of conductivity loss

It should be noted that there exists a minimum conductivity CfDmin in my

model. I assumed that the hydraulic fracture conductivity stops changing when

it reaches CfDmin. As shown in Figure 5.7, the exponential model cannot simulate

actual stress-dependent conductivity behaviour satisfactorily when effective

stress becomes large because the conductivity reduction rate decreases with

increasing effective stress. It could be inferred that the conductivity will remain

nearly constant when effective stress exceeds a certain “threshold” value. Figure

5.9 gives the conductivity behaviour for a fracture with 100 mesh proppants

according to Abass et al. (2009). The fracture sustains 66% of its original

conductivity after 6000 psi, which supports my inference. The exact minimum

values CfDmin are different for different hydraulic fractures and can be determined

based on experiments. CfDmin/CfDi could be as large as 0.66 in Figure 5.9 or as

small as 0.004 in Figure 5.7. In my study, CfDmin/CfDi values were set to be 0.005,

0.02, and 0.1.

Although CfDmin has no relation with the hump’s slope, CfDmin is the key factor

that influences the zenith of the hump. Figure 5.10 gives the pressure behaviour

with different conductivity minimum values when CfD and df remain

unchangeable. Despite different CfDmin, the trend of pressure and pressure

derivative for the three cases in Figure 5.10 remain consistent under the same

CfDi and df. Certainly, the existence of CfDmin prevents the curves from growing

upward. When CfDmin becomes smaller, the pressure can continue growing

quickly under the influence of df. When CfDmin/CfDi is large, such as 60%, it is

97

possible that the hump disappears. When the fracture just loses a small part of

its conductivity, it is hard to discern the difference in the pressure behaviour with

and without stress-sensitive conductivity. In fact, such property is determined by

fracture characteristics. However, I can estimate the minimum conductivity value

through experiments in the laboratory. Generally, proppants with higher loading

strength have a larger CfDmin value, which can prevent serious fracture

conductivity reduction.

98

Figure 5.9―Normalized fracture conductivity (Abass et al., 2009).

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

CfD

/CfD

i

Effective stress, psi

99

Figure 5.10―Type curves showing the effect of minimum conductivity, CfDi=50

and df=1×10-7Pa-1.

tD

1e-6 1e-5 1e-4 1e-3 1e-2 1e-1 1e+0

pw

D,d

Pw

D/d

lnt D

0.001

0.01

0.1

1

10

CfDi=50,df=1E-7Pa-1

, CfDmin=0.25

CfDi=50,df=1E-7Pa-1

,CfDmin=1

CfDi=50,df=1E-7Pa-1

,CfDmin=5

CfDi=50, df=0

P,KPa0 10000 20000 30000 40000 50000 60000

CfD

0

10

20

30

40

50

df=1E-7Pa-1

.CfDmin=5

df=1E-7Pa-1

.CfDmin=1

df=1E-7Pa-1

.CfDmin=0.25

100

5.3.4 Effect of initial conductivity

Conductivity reduction should have different impacts on pressure behaviour

when initial conductivity conditions vary. A spectrum of CfD was tested in Figure

5.11. In Figure 5.11, the ratios of CfDmin to CfDi are the same and CfD vs. ∆P is in

semi-log plot at the lower right corner. In Figure 5.11, it is observed that the

lower the initial fracture conductivity is, the higher the effect of stress-sensitive

fracture conductivity becomes. At first, pressure and pressure derivative

deviation from typical constant-conductivity type curves appear much earlier for

fractured wells with lower CfDi under the same df value. For example, the

influence of stress-sensitive conductivity appears at about tD=1×10-6 when

CfDi=50 while it starts no later than tD=1×10-7 when CfDi=0.5. Secondly, when the

fracture conductivity reaches CfDmin, the remaining conductivity may still lead to

linear flow after the hump for high initial conductivity value. However, for

fractures with low initial conductivities, the hump may be followed by bilinear

flow or even boundary dominated flow, which significantly decrease wells’

productivity. It can be concluded that the stress sensitivity does more harm to

the hydraulic fractures with low initial conductivities.

101

Figure 5.11―Type curves showing the effect of initial conductivity,

CfDi/CfDmin=200.

tD1e-7 1e-6 1e-5 1e-4 1e-3 1e-2 1e-1 1e+0

Pw

D,d

Pw

D/d

lnt D

0.0001

0.001

0.01

0.1

1

10

CfDi=50,df=1E-7 Pa-1

CfDi=5,df=1E-7 Pa-1

CfDi=0.5,df=1E-7 Pa-1

Effective Pressure, MPa0 20 40 60 80 100 120

CfD

0.001

0.01

0.1

1

10

100

CfDi=50

CfDi=5

CfDi=0.5

102

5.3.5 Stress-sensitive conductivity

In addition to the pressure behaviour, the relationship of the stress-sensitive

conductivity vs. time in production can also be represented. Figure 5.12 shows

the ratios of CfD to CfDi over time in semi-log plots for CfDi=50, CfDi=5 and CfDi=0.5

with different df. It is observed that higher df value reduces the fracture

conductivity more severely. For example, the fracture loses approximately 60%

of its conductivity in about 15 days when df equals 5×10-7 Pa-1. As discussed

before, the conductivity’s rapid reduction changes the slopes of bilinear/linear

flow regimes. Correspondingly, higher df makes the conductivity reduction

appear sooner and faster in Figure 5.12. I also compared the conductivities’

response when different initial conductivities were set in the model. It illustrates

that small stress-dependent characteristic values (df =1×10-7, and 5×10-8Pa-1)

have nearly the same effect on fractures with either high or low initial

conductivities. However, with high stress sensitivities (df =5×10-7, and 2.5×10-

7Pa-1), fracture conductivities with low initial values reduce more quickly than

those with high initial conductivity values.

Generally, the stress-sensitive effect on pressure behaviour of fractured

wells is mainly based on df, CfDmin and CfDi. Among these three factors, df has the

greatest impact on pressure behaviour. Higher df can make the pressure

increase more rapidly and the pressure derivate curve’s slope turns steeper.

The hump will appear sooner when df is large. The minimum conductivity CfDmin

determines the zenith of the hump. The fracture with low initial conductivity is

more susceptible to stress change during production.

103

Usually, the hump’s appearance could be determined by df, CfDmin and CfDi.

However, it is not usual to catch the complete hump in the field. As is discussed,

the hump is notable only when df is large and CfDmin is small. However, the well

would be shut-in for improvement if the fracture conductivity becomes too small

to produce economically. Therefore, it is difficult to see the downward part of the

hump. The upward part before the zenith in the hump is more common in many

wells with stress-sensitive hydraulic fractures whose slope is related with df and

CfDi.

104

Fig.10― Variation of fracture conductivity ratio with different df and CfDi.

Figure 5.12― Variation of fracture conductivity ratio with different df and CfDi.

tD

1e-8 1e-7 1e-6 1e-5 1e-4 1e-3 1e-2

CfD

/CfD

i

0.0

0.2

0.4

0.6

0.8

1.0

CfDi

=50

CfDi

=5

CfDi

=0.5

df=5E-8, 1E-7, 2.5E-7and 5×10-7 Pa

-1

105

5.4 Field example

The field data were first used by Clarkson et al. (2012). The horizontal well

located in the Haynesville Shale is multi-stage fractured in 18 stages with eight

perforation clusters per stage spaced approximately 27 ft apart. Gas and water

production data for about 700 days were collected in a log-log scale plot in

Figure 5.13.

Clarkson et al. (2012) analyzed the production data in detail. They started

with assuming both constant matrix permeability and fracture conductivity. Then,

they considered dynamic matrix permeability only, which was followed by adding

stress-sensitive fracture conductivity into the analysis. Although the system

permeability and fracture half-length could be estimated in Clarkson et al.’s work,

no information about fracture conductivity reduction could be derived. Based on

the type curve matching by my model, however, it becomes easier to track

fracture properties. No other additional completion/reservoir details and well

locations were provided because of operator confidentiality. Therefore, in order

to evaluate the changing fracture conductivity, my transient pressure analysis is

directly based on results from Clarkson et al.

5.4.1 Analysis without corrections in matrix permeability

Figures 5.14a and 5.14b show the rate-normalized pseudopressure versus

time and rate-normalized pseudopressure derivatives versus time, respectively.

In Figures 5.14a and 5.14b, no corrections in the pseudopressure and time

function for dynamic matrix permeability are made by Clarkson et al. (2012). For

many fractured horizontal wells in shale formations, the linear flow may last for

106

several years. However, the linear flow in Figures 5.14b is very short, which is

followed by apparent BDF with pressure and its derivatives rising rapidly. This

phenomenon could be an obvious index for the decreasing fracture conductivity

and matrix permeability, which provides the basis for my analysis.

In my transient pressure analysis, assuming either one fracture per stage or

one fracture per cluster is acceptable. Here, I chose the horizontal well model

with 18 fractures instead of 144 fractures because of computation efficiency. In

accordance with Clarkson et al.’s results, I generated a series of type curves to

measure the fractures’ stress-dependent characteristic df. Figures 5.15a and

5.15b present the matching results. The matrix permeability used in my model is

based on Clarkson et al.’s results. The fracture conductivity reduces quickly from

1.34×10-13 m3 to 1.54×10-17 m3 with df =1.97×10-7 Pa-1 when matrix permeability

remains constant during 700 days of production.

107

(Original in color)

Figure 5.13― Gas and water production rates and calculated flowing bottomhole

pressures (Clarkson et al. 2012).

108

(a)

(b)

Figure 5.14―Pressure and pressure derivative without corrections in matrix

permeability (Clarkson et al., 2012).

109

(a)

(b)

Figure 5.15―Type curve matching results without corrections in matrix

permeability (Original in color).

110

5.4.2 Analysis with corrections in matrix permeability only

Clarkson et al. (2012) also considered the changing matrix permeability and

revised the pseudopressure and pseudotime functions. The modified variables

are provided below (Al-Hussainy et al, 1966):

pdpz

pk

kpmpm

i

wf

p

pgi

wfi

)(2**

, ················································ (5.4)

and (Thompson et al., 2010)

t

tgi

itg

a

c

dtpk

k

c

t0

* )(

. ································································· (5.5)

In Equation (5.4) and Equation (5.5), the matrix permeability k changes from

initial value ki as described in Equation (5.3).

Figures 5.16a and 5.16b show the pressure behaviour while considering

stress-sensitive matrix permeability. The linear flow period becomes longer than

the no-correction case in Figures 5.14a and 5.14b. The data are also pushed to

the right. Figures 5.17a and 5.17b give my new match results. They show that

the hydraulic fracture conductivity changes to 7.11×10-18 m3 from 4.837×10-14 m3

with df =1.59×10-7 Pa-1.

Table 5.2 compares my results and the analysis from Clarkson et al. (2012)

considering only stress-sensitive matrix permeability. Stress-sensitive

characteristic df values and initial fracture conductivities in my two analysis

results are similar while the matrix permeabilities are different. This is because

the matrix permeability is assumed to reduce over time in the second analysis,

and constant matrix permeability is one of important assumptions before

developing the model. Consequently, a lower changeless average permeability

111

value is attained in my matching results to replace the stress-sensitive matrix

permeability.

Based on the analysis results with only stress-sensitive hydraulic fractures,

the forecast of production rate is provided. It is assumed that the bottomhole

pressure is constant at 2000 psi in the next 300 days. The prediction with both

stress-sensitive hydraulic fractures and matrix permeability are impracticable

because of lack of accurate pressure data. At the same time, the prediction

without stress-sensitivity from the aforementioned semi-analytical model is also

shown for comparison in Figure 5.18. The apparent BDF is regarded as real

BDF in the model with constant fracture properties. The result indicates that the

well’s productivity could be underestimated when the stress-sensitivity is

mistaken for the influence of boundaries.

112

(a)

(b)

Figure 5.16―Pressure and pressure derivative with corrections in matrix

permeability (Clarkson et al. 2012).

113

(a)

(b)

Figure 5.17―Type curve matching results with corrections in matrix

permeability (Original in color).

114

TABLE 5.2 RESULTS OF FIELD CASE ANALYSIS

Without

corrections in matrix

permeability

With corrections in matrix

permeability

Clarkson et al.’s results with

dynamic matrix permeability only

Matrix permeability, k

2.82×10-18 m2 (2.82×10-3 mD)

8.164×10-19 m2

(8.16×10-4 mD) 1.569×10-18 m2 (1.57×10-3 mD)

Matrix permeability modulus, γ

― ― 5.8×10-8 Pa-1 (4×10-4 psi-1)

Fracture conductivity, kfwf

1.34×10-13 m3 (439 mD∙ft)

4.837×10-14 m3 (158.7 mD∙ft)

Fracture half-length, xf

90 m (295 ft)

90 m (295 ft)

103.6 m (340 ft)

Fracture stress-sensitive

characteristic, df

1.97×10-7 Pa-1 (1.36×10-3 psi-1)

1.59×10-7 Pa-1 (1.1×10-3 psi-1)

115

(Original in color)

Figure 5.18―Production rate prediction.

100

1000

10000

100000

0 200 400 600 800 1000

Gas R

ate

, M

scf/

Day

Time, days

Gas rate

Semi-analytical result without dynamic matrix permeability and fracture conductivity

Semi-analytical result without correction for matrix permeability

116

5.5 Chapter summary

The study of stress-sensitive hydraulic fractures is crucial for optimizing and

monitoring fracturing treatments, especially for unconventional reservoirs. In this

chapter, I modeled and discussed the pressure behaviour of wells with stress-

dependent hydraulic fractures. Based on my work, several conclusions can be

drawn:

A semi-analytical model has been established and type curves are

generated based on the model for pressure transient analysis of fractured

wells.

When the stress-sensitive fracture conductivity reduces, the slopes of

bilinear/linear flow become larger than 1/4 and 1/2, respectively. A hump

can be formed in the pressure derivative curve, and its slope is close to or

even larger than unit.

Large df could make the pressure derivative curve slopes increase

quickly and accelerate the hump’s appearance. The minimum

conductivity CfDmin can stop the conductivity from decreasing to a very

small value in this model.

The pressure deviation from that with constant fracture conductivities is

influenced by initial fracture conductivities. The lower the initial fracture

conductivity, the greater the effect of stress-sensitive conductivity.

When df value is small, it has a similar influence on the conductivity ratio

no matter whether the fracture has a high or low initial conductivity. When

117

df becomes larger, the fractures with low initial conductivities lose their

conductivities more quickly.

The applicability of this semi-analytical model is proved in the field case

study based on an example from the literature. It shows that incorrect

conclusions could be drawn if the stress-sensitive effect of hydraulic

fractures is ignored.

118

CHAPTER 6

PRODUCTION RATE ANALYSIS

Production rate analysis has been widely used by petroleum engineers to

evaluate reservoir performance and predict wells’ production. For multi-stage

fractured horizontal wells, the production rate analysis in this chapter also

considers the stress-sensitive hydraulic fracture conductivity. Traditional

analytical methods might cause errors in predicting wells’ performance under

complex well/reservoir conditions. Therefore, in addition to analytical solutions,

this work provides accurate semi-analytical modeling for multi-stage fractured

horizontal wells with constant bottomhole pressure.

At first, a few assumptions, listed below, should be made, as in Chapter 5:

1. The target formation is isotropic, homogeneous, and bounded with upper

and lower impermeable strata. All the boundaries are closed boundaries.

2. Reservoir parameters, such as porosity ϕ and permeability k are kept

constant during production.

3. Both the reservoir and hydraulic fractures are filled with single-phase

fluid, which could be gas or slightly compressible fluid with constant viscosity.

4. All the hydraulic fractures are vertical, symmetric, and fully penetrate

the formation.

119

6.1 Model and algorithm

The mathematical model for multi-stage fractured horizontal wells under

constant-pressure condition is similar to that with constant production rate. The

fluid flow from reservoir to wellbore is negligible. Dimensionless definitions of

variables are the same as in section 4.1. Since the production rate changes over

time in constant-pressure production, a reference oil production rate Qr is

necessary in dimensionless definition:

r

wfi

D

Q

ppkhp

)( , ········································································· (6.1)

r

rrf

rfD

Q

Lqq , ············································································· (6.2)

r

f

fD

Q

qq . ················································································· (6.3)

For gas wells, I also have (Al-Hussainy et al., 1966)

scr

wfisc

DTpQ

pmpmkhTp

))()(( , ······························································· (6.4)

where Tsc and psc are the standard temperature and pressure, respectively.

The reference rate Qr can be any value but remains consistent throughout

the whole model. In the following discussion, the reference rate Qr is 105 m3/day.

The whole system is still seperated into two parts: one is the fluid flow system

inside the hydraulic fractures and the other is the fluid flow system inside the

reservoir. Each hydraulic fracture is divided into several segments, and each

segment owns a uniform influx from the reservoir systemqrfi,j and two node

fluxes from adjacent fracture segments inside fractures qrfi,j-1 andqrfi,j. At the

120

end of every time step, the segment solutions for the reservoir system that are in

a linear format with jrfi

q,

are coupled with fracture segment solutions by using

the pressure equivalent and flux continuity conditions at the interfaces of any

two adjacent segments at the same time point in Laplace domain. In addition,

the pressure of fracture segments closest to the wellbore should be equal to the

constant flowing bottomhole pressure. Finally, fluxes at each interface and

pressure distribution in the reservoir can be achieved.

For fractured wells with stress-sensitive hydraulic fractures, the algorithm is

presented in Chapter 3 and Chapter 5. Fracture conductivities change as

described by Equation (5.3). The difference of this constant-pressure model with

dynamic conductivities when compared with the constant-rate stress-sensitive

model is that here the initial pressure is chosen to be the pressure distribution

when fractured wells start to produce rather than Pi.

6.2 Model validation

This study presents a rigorous semi-analytical model for a 6-stage fractured

horizontal well located in the center of a rectangular closed reservoir. The model

was validated with the numerical case from KAPPA Ecrin. Figure 6.1 shows the

comparison between these two cases and a satisfactory match is reached. I

preferred the numerical model to the analytical model in KAPPA Ecrin because

the data from the analytical KAPPA model become unstable in the BDF regime.

121

Figure 6.1―Model validation.

1E-6

1E-5

1E-4

1E-3

1E-2

1E-1

1E+4 1E+5 1E+6 1E+7 1E+8 1E+9 1E+10 1E+11

q,m

3/s

t,s

Semi-analytical Result

Kappa numerical result

122

6.3 Results and discussion

6.3.1 Comparison with analytical solutions of single-fractured wells

As discussed in Chapter 4, there exists a bilinear flow regime, linear flow

regime, CLF regime, and BDF regime for a multi-stage fractured horizontal well

with constant fracture conductivities in a box-shaped closed reservoir. The

simple analytical solutions under the constant-rate condition for most of the

above flow regimes have been discussed at length in the literature. Similarly, in

order to simplify forecasting multi-stage fractured horizontal wells’ production

under the constant-pressure condition, I derived analytical solutions of single-

fractured wells that are then compared with my semi-analytical results.

6.3.1.1 Bilinear flow

For fractured wells, bilinear flow is the first identifiable flow regime if

fractures have finite conductivities. The constant-pressure solution for bilinear

flow can be derived based on the constant-rate solution. For the constant-rate

case, the following equation is used (Azari et al., 1991):

41

4

522

1t

C

p

fD

D

. ································································ (6.5)

Then, Equation (6.5) is transformed into Laplace domain. In Laplace domain, the

relationship between constant-pressure and constant-rate solutions is

2

1

upq

DD . ············································································ (6.6)

Therefore, after inverse Laplace transformation, the analytical solution for the

constant-pressure case can be written as

123

4/1

4/3

2/12/14/1)()(7347.0

)(

tB

pphwkCk

tqwfifft

for oil, ··················· (6.7a)

and

4/1

2/12/14/1

)()(002007.0

)(

tT

mmhwkCk

tqpwfpiffit

for gas. ········· (6.7b)

I established a semi-analytical model for a 6-stage fractured horizontal well

with fracture conductivity CfD=1. Then, the semi-analytical result is compared

with the above analytical solutions in Figure 6.2. It can be found that a product of

q(t) from the analytical solution and stage number n in the semi-analytical model

can match my semi-analytical results perfectly for early-stage production data. It

shows that these six fractures work independently in the bilinear flow regime,

and interference between different fractures has not appeared yet.

The bilinear flow period in the constant-pressure case diminishes at about

the same time as does the constant-rate case (Azari et al., 1991). The end time

of bilinear flow can be expressed as

4

2

09.0

r

f

fD

fD

L

x

C

t for r

f

fD

L

xC

2

, ·················································· (6.8a)

2

2

01.0

r

f

fD

L

xt for

r

f

fD

r

f

L

xC

L

x

2

1.0 , ············································· (6.8b)

2

fDfDCt for

r

f

fD

L

xC

1.0 . ·························································· (6.8c)

According to Equations (6.8a), (6.8b), and (6.8c), the end time tD of bilinear flow

for the case in Figure 6.2 should be 7.72×10-6, which approximately agrees with

my semi-analytical results, 4×10-6.

124

Figure 6.2―Results comparison for bilinear flow.

1E-6

1E-5

1E-4

1E-3

1E-2

1E-1

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

qD

tD

Semi-analytical results

Analytical solution for bilinear flow

125

6.3.1.2 Linear flow

Linear flow occurs when the fracture has a higher conductivity. The

analytical solution for constant-rate production in linear flow is

DwDtp . ············································································· (6.9)

Based on Equation (6.6), the analytical solution for the constant-pressure case

can be expressed as

21

24

t

B

pphxCk

tqwfift

for oil, ············································· (6.10a)

and

21

2010926.0

t

T

pphxCk

tqwfift

for gas. ·································· (6.10b)

To test the applicability of Equations (6.10a) and (6.10b), it is necessary to

compare such analytical solutions with my semi-analytical model. Figure 6.3

shows the comparison for a 6-stage fractured horizontal well with fracture

conductivity CfD=10. In the linear flow period, which appears after the bilinear

flow, the product of production rate q(t) in Equation (6.10a) and stage number n

nearly equals my semi-analytical result. This means that the six fractures in a 6-

stage fractured horizontal well also work independently during linear flow in this

period.

For high-conductivity fractures, linear flow might be the first recognizable

flow regime in pressure and/or production rate analysis. Thus, it is important to

study the duration of linear flow. Linear flow always begins after the bilinear flow.

126

The end time of linear flow is related to the investigation distance L, which can

be obtained from the following equation (Wattenbarger et al., 1998):

tC

ktL

9856.1 . ······································································ (6.11)

For a multi-stage fractured horizontal well, linear flow ends when L reaches

12

n

LL

h , ·············································································· (6.12)

where hL is the length of a horizontal well and n is the stage number.

Substituting Equation (6.12) into Equation (6.11), I can get the end time of linear

flow as

k

C

n

Lt

th

2

106341.0

. ····························································· (6.13)

The end time tD for the case in Figure 6.3 was estimated to be 1.29×10-3, which

approximately matches the result in Figure 6.3.

Since my semi-analytical model is complex and there are diverse

combinations of parameters in the model, it would take much more time to

match the field data. Also, for fractured wells in tight formations and shale gas

reservoirs, bilinear/linear flow lasts for a long period. Thus, it is more convenient

to use simple Equations (6.7), (6.8), (6.10), and (6.13) to analyze and predict

well production in bilinear and linear flow regimes.

127

Figure 6.3―Results comparison for linear flow.

1E-6

1E-5

1E-4

1E-3

1E-2

1E-1

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

qD

tD

Semi-analytical result

analytical result for linear flow

128

6.3.1.3 Compound linear flow

Compound linear flow is similar to the aforementioned linear flow with one

half slope in production rate curves in the log-log plot. However, the difference

for compound linear flow is that the flow during this period is perpendicular to the

direction of linear flow.

Figure 6.4 shows the production rates for 6-stage fractured horizontal wells

with different fracture half-lengths among which only the well with fracture half-

length xf=40m exhibits the compound formation linear flow clearly. It can be

concluded that the appearance of a compound linear flow regime depends upon

the fractures’ half-length xf . This conclusion agrees with Chen and Raghavan’s

(1997) results. When xf=40 m and ye (inter-well spacing) =400 m, the ratio of xf

vs. ye is 0.1, which indicates that ye must be at least 10 times greater than xf if

the compound linear flow is obvious. Accordingly, it is very unlikely for the

compound linear flow to appear in field applications and, therefore, any

departure from the linear flow in production rate analysis should not be

interpreted as compound linear flow without verification.

Similar to Equation (6.10), the analytical solution for compound linear flow

under the constant-pressure condition is

21

22

t

B

pphxCktq

wfict

for oil, ········································· (6.14a)

and

21

2010926.0

t

T

pphxCktq

wfict

for gas. ····························· (6.14b)

129

where xc is the hypothetical half-length and works as the fracture half-length.

Results from my semi-analytical model and analytical solutions are plotted in

Figure 6.5. In Figure 6.5, the curve with xc= 340 m gives the best match. Hence

the length of influence area in compound linear flow can be further calculated as

2xe=680 m, which is greater than the horizontal well length of 600 m. So the area

of influence of a multi-stage fractured horizontal well in compound linear flow

should be larger than the product of the fracture length and horizontal well

length, 2xf∙Lr.

According to Liang et al. (2012), the end time of compound linear flow can

be estimated as

2

2

2

r

f

D

L

xt .··············································································· (6.15)

If the compound linear flow exists, it appears later with a larger fracture half-

length. Based on Equation (6.15), the compound linear flow in Figure 6.5 should

begin at approximately tD=9×10-3, which agrees with my semi-analytical results.

130

Figure 6.4―Production rates with different fracture half-lengths.

1E-6

1E-5

1E-4

1E-3

1E-2

1E-1

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

qD

tD

Xf=80m

Xf=70m

Xf=60m

Xf=50m

Xf=40m

131

Figure 6.5―Results comparison for compound linear flow.

1E-6

1E-5

1E-4

1E-3

1E-2

1E-1

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

qD

tD

Xf=40m,semi-analytical model result

Compound Formation Linear Flow, analytical result

132

6.3.1.4 Boundary-dominated flow

The onset of boundary dominated flow means that the area of influence of a

fractured horizontal well has reached the boundaries. The characteristic of

boundary dominated flow under the constant-rate condition is the unit slope of

both pressure and pressure derivative curves in a log-log plot. However, for

constant-pressure production, the slope of the production rate curve in the log-

log plot is no longer unit. Similar to sections 6.3.1.1 and 6.3.1.2, the analytical

constant-pressure solution for the boundary dominated flow can also be derived

from the analytical constant-rate solution. For a constant-rate production well,

the dimensionless pressure in BDF can be expressed as

batpDD . ·········································································· (6.16)

where a and b are the slope and intercept in a Cartesian coordinate system.

Following the steps in section 6.3.1.1, the dimensionless production rate under

constant flowing pressure should be

Dt

b

a

De

bq

1

. ··········································································· (6.17)

Equation (6.17) is a general exponential decline equation rather than a unit

slope equation. Figure 6.6 compares the analytical exponential solution

(a=5.4956×104 and b=4.833×105) with my semi-analytical result and BDF

exhibits an exponential trend as Eq. (6.17).

As discussed in Chapter 4, a pseudo pseudo-steady state flow regime is

very similar to boundary dominated flow under constant-rate condition, but in

constant-pressure solutions, pseudo pseudo-steady state flow also has a close-

133

to-unit slope, rather than exponential decline, so it can be further concluded that

the slope of pseudo pseudo-steady state flow under constant-rate condition is

not exactly unit or changes over time.

134

Figure 6.6―Results comparison for boundary-dominated flow.

1E-6

1E-5

1E-4

1E-3

1E-2

1E-1

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

qD

tD

Xf=50m, semi-analytical result

Analytical result

135

6.3.2 Effect of stress-sensitive fracture conductivity

The effect of stress-sensitive fracture conductivities under constant-rate

condition has already been discussed in Chapter 5. In Chapter 6, I further

analyze such influence with the constant-pressure condition. Variables such as

df and CfD are investigated in detail through my semi-analytical model.

6.3.2.1 Effect of df

The df value demonstrates the hydraulic fractures’ stress-sensitive

characteristic. As suggested in section 5.3.2, the range of df was chosen to be

between 5×10-8 Pa-1and 3×10-7 Pa-1 in this study. Figure 6.7 provides the

production rate of a 6-stage fractured horizontal well with different stress-

sensitive fracture characteristics. The fracture conductivity reduces from initial

value CfDi=1 as described in Equation (5.3). The slopes of bilinear and linear

flows with stress-sensitive conductivities are exactly between 1/4 and 1/2 (1/4

<slope<1/2), which deviate from typical 1/4 and 1/2 slopes. Such slope

characteristic can be regarded as a sign of stress-sensitive fracture

conductivities. The larger df becomes, the bigger the slope deviation of bilinear

and linear flow will be. While bilinear and linear flows are influenced by stress-

dependent characteristic, the pseudo PSS flow and BDF are not affected by

such dynamic conductivities when df is relatively small.

The fracture conductivities vs. time are plotted in semi-log scale in Figure

6.8. Originally, fracture conductivities drop quickly over time. The slopes of

conductivity curves become greater if df becomes larger. However, the

136

conductivity tends to stop changing and approaches a minimum value C*

fDmin

after a period of reduction. Such minimum value is dependent on df, *

ip and pwf :

wfif ppd

fDifDeCC

*

*

min, ····························································· (6.18)

where the initial pressure *

ip is the pressure calculated at the first time step in

production and is related to xf and CfDi.

137

Figure 6.7― Production curves with different df.

1E-6

1E-5

1E-4

1E-3

1E-2

1E-1

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

qD

tD

CfDi=3,df=0

CfDi=3,df=5E-8 Pa-1

CfDi=3,df=7E-8 Pa-1

CfDi=3,df=9E-8 Pa-1

CfDi=3,df=1E-7 Pa-1

CfDi=3,df=3E-7 Pa-1

xf=80m CfDi=1, df=0

CfDi=1, df=3E-7 Pa-1

CfDi=1, df=1E-7 Pa-1

CfDi=1, df=9E-8 Pa-1

CfDi=1, df=7E-8 Pa-1

CfDi=1, df=5E-8 Pa-1

138

Figure 6.8 ― Conductivity curves with different df .

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

CfD

/CfD

i

tD

df=5E-8 Pa-1

df=7E-8 Pa-1

df=9E-8 Pa-1

df=1E-7 Pa-1

df=3E-7 Pa-1

139

6.3.2.2 Effect of CfDi

Fractured horizontal wells with different initial fracture conductivities have

different responses to the stress-sensitive effect. Figure 6.9 provides the

production curves with initial conductivities 10, 1, and 0.1 for a 6-stage fractured

horizontal well. As the figure shows, the wells with smaller initial conductivities

are more susceptible to the stress-dependent effect. For example, when CfDi=0.1,

the difference between production rates with and without stress-sensitive

conductivities could be as much as 33%. In addition, bilinear flow is dominant at

early-stage production when CfDi=0.1, and the production rate is proportional to

stress-sensitive fracture conductivities in the bilinear flow regime. Thus, it could

be further concluded that the bilinear flow is more sensitive to the stress-

dependent conductivities than other flow regimes.

The characteristics of conductivities over time determine the trend of

production rates. Figure 6.10 gives the conductivities vs. time in a semi-log plot.

The fracture conductivity with initial value CfDi=0.1 has the maximum change

during production, which is followed by CfDi=1 and 10. This is consistent with the

conclusions from Figure 6.9. Although the conductivity changes most when

CfDi=0.1, its change rate is the lowest among these three cases. The

conductivity with CfDi=10 reaches its minimum value at tD=3×10-3 while the

conductivity with CfDi=0.1 keeps changing until tD=3×10-1.

140

Figure 6.9―Production curves with different initial conductivity CfDi.

1E-6

1E-5

1E-4

1E-3

1E-2

1E-1

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

qD

tD

CfDi=30,df=9E-8 Pa-1

CfDi=30, df=0

CfDi=3,df=9E-8 Pa-1

CfDi=3,df=0

CfDi=0.3,df=9E-8 Pa-1

CfDi=0.3, df=0

xf=80m CfDi=10,df=9E-8 Pa-1

CfDi=10,df=0

CfDi=1,df=9E-8 Pa-1

CfDi=1,df=0

CfDi=0.1,df=9E-8 Pa-1

CfDi=0.1,df=0

141

Figure 6.10―Conductivity curves with different CfDi.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0

CfD

/CfD

i

tD

CfDi=30,df=9E-8 Pa-1

CfDi=3, df=9E-8 Pa-1

CfDi=0.3, df=9E-8 Pa-1

CfDi=10,df=9E-8 Pa-1

CfDi=1,df=9E-8 Pa-1

CfDi=0.1,df=9E-8 Pa-1

142

6.4 Field examples

This section presents two field case analyses using both the analytical

solutions and my semi-analytical models, which reveals each method’s strong

and weak points, respectively.

6.4.1 Marcellus shale gas well A

The original data from the Marcellus shale gas multi-stage fractured

horizontal well A were first reported by Nobakht et al. (2012a). The well is

producing under high-drawdown conditions (90%~95% drawdown), and,

therefore, the assumption of constant flowing pressure is applicable. The 10-

stage fractured horizontal well length is 1219 m (4000 ft). The properties of this

reservoir are as follows: pi=3.65×107 Pa (5300 psi), T=339 K (150°F), h=38 m

(125 ft), ϕ=8%, Sg=76%, Sw=24%, and cf=7.25×10-10 Pa-1 (5×10-6 psi-1).

Figure 6.11 shows a plot of inverse gas rate vs. fourth root of time for this

well. The bilinear flow dominates during the production period. The straight line’s

slope is 0.4248 s0.75/m3. Using this slope and Equation (6.7b), the product of

square root of matrix permeability and fracture conductivity ff

wkk can be

estimated as 5.725×10-27 m4. However, the production data deviates from the

straight line after t=1.27×107 s (147 days) so the second plot of inverse gas rate

vs. square root of time (Figure 6.12) is provided to analyze late-time data. The

slope and intercept of the straight line are 4×10-4 s0.5/m3 and 1.2367 s/m3,

respectively.

143

Figure 6.11―Plot of reciprocal gas rate vs. fourth root of time.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

0 10 20 30 40 50 60 70 80

1/q

, s/m

3

t0.25, s0.25

144

Figure 6.12―Plot of reciprocal gas rate vs. square root of time.

y = 0.0004x + 1.2367 R² = 0.6734

2

2.2

2.4

2.6

2.8

3

3.2

3000 3500 4000 4500 5000

1/q

, s/m

3

t0.5, s0.5

145

According to Equation (6.10b), f

xk can be calculated as 4×10-7 m2. In addition,

the fracture half-length is chosen to be 150 m from microseismic mapping. Then,

the matrix permeability and fracture conductivity should be 9×10-20 m2 (90 nd)

and 1.91×10-17 m2 (0.0626 md∙ft), respectively, by coupling ff

wkk andf

xk .

My semi-analytical model is also applied and is compared with the above

method in this production rate analysis. Figures 6.13(a) and 6.13(b) give the

matching results using the semi-analytical model. Matrix permeability, fracture

conductivity and fracture half-length are revaluated as 9.5×10-19 m2 (95 nd),

1.9×10-17 m2 (0.0626 md∙ft), and 120 m (394 ft). Good agreement is achieved in

the analysis results from semi-analytical modeling and analytical solutions. At

t=1×109 s (11,574th day), boundary-dominated flow begins and the EUR

obtained using my semi-analytical is 8.84×107 m3 (3.12 Bscf). Also, the EUR in

analytical solutions for a 30-year forecast is 1.11×108 m3 (3.91 Bscf).

The method using analytical solutions is simple and fast in analyzing and

predicting production rates, especially in bilinear/linear flow regimes. Hence,

when dealing with fractured wells with recognizable flow regimes, the analytical

solutions could become a preferred candidate for production rate analysis.

146

(a)

(b)

(Original in color)

Figure 6.13―Type curve matching for Marcellus shale gas well A.

1E-2

1E-1

1E+0

1E+1

1E+5 1E+6 1E+7 1E+8 1E+9 1E+10

q, m

3/s

t, s

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1E+5 1E+7 2E+7 3E+7

q, m

3/s

t,s

147

6.4.2 Marcellus shale gas well B

This case study was introduced by Nobakht et al. (2012b) for a multi-stage

fractured horizontal well in the Marcellus shale gas reservoir. The well is 914 m

(3000 ft) long with 12 fractures. The reservoir properties are: pi=2.7×107 Pa

(3924 psi), T=378 K (220°F), h=46 m (150ft), ϕ=7%, Sgi=66%, Swi=34%,

cf=5.8×10-10 Pa-1 (4×10-6 psi-1). Also, this well is producing under high-drawdown

conditions and could be considered as constant-pressure production. Moreover,

this case takes into account the gas desorption. According to Nobakht et al.’s

work (2012b), RD is set as 2.0.

The diagnostic plot, Figure 6.14, shows no typical flow regimes such as

bilinear and linear flow. I further investigated the relationship between inverse

production rate and square root of time for this case. Unlike the diagnostic plot,

a straight line can be drawn in Figure 6.15, which means linear flow appears.

The slope and intercept of the straight line are 3×10-4 s0.5/m3 and 0.378 s/m3.

With the line slope, Equation (6.10b) and matrix permeability 5×10-19 m2 (500 nd)

assumed by Nobakht (2012b), fracture half-length should be 110 m. The

intercept is the skin factor that reflects the reduced conductivity and changes the

shape of the log-log plot of q vs. t.

In order to verify these results, type curves of q vs. t are provided using my

semi-analytical model and parameters from Figure 6.15. However, although

dimensionless fracture conductivity CfD varies from 0.1 to 1, the type curves still

cannot match the field data because pseudo pseudo-steady state flow starts and

linear flow ends earlier than field data. Adjustments are necessary to analyze

148

results based on analytical solutions. Figure 6.17 gives a new combination of

parameters and type curve that matches well with the field data based on the

semi-analytical model. The matrix permeability, fracture half-length and fracture

conductivity kfwf become 9×10-20 m2 (90 nd), 140 m (459 ft) and 5.76×10-17 m3

(0.19 md∙ft). The matrix permeability is reduced when compared with results

from analytical solutions. The matching type curve shows that the production

period of filed data mainly belongs to the transition flow regime from bilinear flow

to pseudo PSS flow. In addition, the EUR estimated by my semi-analytical

model is 8.5×107 m3 (3.0 Bscf).

When flow regimes cannot be identified, analysis based on analytical

solutions always brings errors in estimation and prediction while semi-analytical

modeling provides more appropriate matching results. One important reason is

that semi-analytical models can easily simulate transitions between different flow

regimes under different reservoir conditions. For analytical solutions, it is difficult

to determine the exact time when a specific flow regime starts and ends. In

general, semi-analytical modeling is more effective for complex reservoir

conditions.

149

Figure 6.14―Diagnostic plot of production rate vs. time.

Figure 6.15― Plot of reciprocal gas rate vs. square root of time.

0.1

1

10

1E+5 1E+6 1E+7 1E+8

q, m

3/s

t,s

1/4 slope 1/2 slope

y = 0.0003x + 0.378 R² = 0.8537

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

0 1000 2000 3000 4000 5000

1/q

, m

3/s

t0.5,s0.5

150

(a)

(b)

(Original in color)

Figure 6.16― Type curves based on parameters from analytical solutions.

0

1

2

3

4

5

6

1E+5 5E+6 1E+7 2E+7 2E+7

q, m

3/s

t,s

Series1

Series2

Series3

Field data

CfD=0.1

CfD=1

0.1

1

10

1E+5 1E+6 1E+7 1E+8

q, m

3/s

t,s

Series1

Series2

Series3

Field data

CfD=1

CfD=0.1

151

(a)

(b)

(Original in color)

Figure 6.17―Type curve matching result by semi-analytical model

0

0.5

1

1.5

2

2.5

1E+5 5E+6 1E+7 2E+7 2E+7

q, m

3/s

t,s

0.01

0.1

1

10

1E+5 1E+6 1E+7 1E+8 1E+9

q, m

3/s

t,s

152

6.5 Chapter summary

Transient production rate analysis provides another effective way to

evaluate reservoir properties and predict wells’ production in addition to transient

pressure analysis. In this Chapter, I developed a semi-analytical model for multi-

stage fractured horizontal wells with and without stress-sensitive fractures.

Moreover, the comparison between the semi-analytical model and analytical

solutions of different flow regimes reveals the mechanisms by which individual

fracture works.

The following conclusions are drawn from this study:

In bilinear and linear flow regimes, the fractured horizontal wells’

production is the sum of each fracture’s contribution.

The influence area of a fractured horizontal well is always larger than the

product of the horizontal well length and fracture half-length.

When considering stress-dependent hydraulic fracture conductivities, the

slope of bilinear flow becomes greater than 1/4 while linear flow’s slope

becomes smaller than 1/2.

The wells with smaller initial conductivity are more susceptible to stress-

dependent effects. Correspondingly, bilinear flow regime is more

sensitive to stress-sensitive fracture conductivities than other flow

regimes.

Analytical solutions are more suitable for production rate analysis with

recognizable flow regimes while semi-analytical modeling can deal with

complex reservoir problems with no clear signature of flow regimes.

153

CHAPTER 7

CONCLUSIONS AND RECOMMENDATIONS

7.1 Conclusions

The contributions of this work can be classified into the following categories:

methodology, transient pressure behaviour, and production rate analysis. The

results in each category are summarized as follows:

Methodology

1) I have introduced a general, effective, and accurate semi-analytical model

to investigate the transient pressure behaviour and transient production

rate behaviour of multi-stage fractured horizontal wells under constant

rate or pressure conditions. This methodology is as accurate as the

analytical method, as flexible and applicable as numerical models, and

also supports the stress-sensitive hydraulic fracture conductivities.

Transient pressure behaviour

2) For multi-stage fractured horizontal wells, fluid flow from the reservoir

directly to the wellbore and the pressure drop inside the wellbore could

have more obvious influence on pressure and flow distribution when

fewer fractures are created along the wellbore.

3) There exists an optimum fracture treatment design for a specific reservoir

instead of production simply being better with more and larger hydraulic

fractures. When the fracture half-length is constant, an optimal fracture

number can be obtained for the highest profit. When the total fracture

154

volume remains unchanged, I could also find the best combination of

fracture numbers and the fracture half-lengths.

4) When the stress-dependent hydraulic fracture conductivity is considered

for multi-stage fractured horizontal wells, slopes of bilinear and linear flow

deviate from 1/4 and 1/2, which helps to form a hump in the pressure

derivative curves.

5) Higher hydraulic fracture stress-sensitive characteristic df, lower minimum

fracture conductivity CfDmin, and lower initial fracture conductivity CfDi

make stress-sensitive fractures more susceptible to conductivity loss

when the stress field changes during production.

6) The type curve matching method based on my semi-analytical models is

powerful in analyzing the complex transient pressure behaviour of multi-

stage fractured horizontal wells in tight formations and shale reservoirs.

Production rate analysis

7) For multi-stage fractured horizontal wells, the total production equals the

sum of each separate fracture’s contribution during bilinear/linear flow

regimes.

8) Fractured horizontal wells with lower initial conductivities and larger

stress-sensitive characteristic df have more significant conductivity

reduction during constant-pressure production.

9) Analysis based on analytical solutions is preferable when flow regimes

are identifiable while analysis with semi-analytical modeling is more

appropriate for complex production rate behaviour.

155

7.2 Recommendations

The following recommendations for future work are made based on my work

in this study:

1) The models and/or the results presented in this work can be

extended to more complicated reservoir conditions, such as complex

fracture systems, boundary conditions for stimulated reservoir

volume, and dynamic matrix permeability.

2) This work can be further incorporated with seismic data to fully

describe fractures underground and their effect on transient pressure

behaviour.

3) A Large amount of experimental data should be studied to modify the

equations that describe the relationship between fracture

conductivities and stress.

4) A large number of cases should be studied to investigate the

relationship between pressure derivative curve slopes and stress-

sensitive characteristic for transient pressure and production rate

behaviour of fractured wells.

156

LIST OF REFERENCES

Abass H., Sierra L. and Tahini A., Optimizing Proppant Conductivity and

Number of Hydraulic Fractures in Tight Gas Sand Wells. This paper SPE

126159 was prepared for presentation at the SPE Saudi Arabia Section

Technical Symposium, Alkhobar, Saudi Arabia,9-11 May 2009.

Agarwal, R.G., Gardner, D.C., Kleinsteiber S.W. and Fussell, D.D., Analyzing

Well Production Data Using Combined Type Curve and Decline Curve

Analysis Concepts. Paper SPE 49222 was prepared for presentation at

the SPE Annual Technical Conference and Exhibition held in New

Orleans, Louisiana, 27-30 September 1998.

Al-Hussainy, R., Ramey Jr., H.J. and Crawford, P.B., The Flow of Real Gases

Through Porous Media. Journal of Petroleum Technology

(1966)18(5):624-636.

Arps, J.J., Analysis of Decline Curves. Trans. American Institute of Mining,

Metallurgical, and Petroleum Engineers (1944)160,228-247.

Azari, M., Soliman, M.Y., Wooden, W.O. and Hunt, J.L. Performance Prediction

for Finite-Conductivity Vertical Fractures. Paper SPE 22659 was prepared

for the presentation at the 66th Annual technical Conference and

Exhibition of the Society of Petroleum Engineers held in Dallas, TX,

October 6-9, 1991.

Babu, D.K. and Odeh, A.S., Productivity of A Horizontal Well. SPE Reservoir

Engineering (1989)4(4):417-421.

157

Bello, R.O. and Wattenbarger, R.A., Multi-stage Hydraulically Fractured Shale

Gas Rate Transient Analysis. Paper SPW 126854 was presented at the

SPE North Africa Techniacal Conference and Exhibition in Cairo, Egypt,

February 14-18 2010.

Ben E. L. and Spencer, C.W., "Gas in Tight Reservoirs-An Emerging Major

Source of Energy" in David G. Howell (ed.), The Future of Energy Gasses,

US Geological Survey, Professional Paper(1993)1570:233-252.

Berumen, S., Tiab, D., Effect of Pore Pressure on Conductivity and Permeability

of Fractured Rocks. This paper SPE 35694 was prepared for presentation

at the Western Regional Meeting held in Anchorage, Alaska, 22-24 May

1996.

Best, M.E. and Katsube, T., Shale Permeability and Its Significance in

Hydrocarbon Exploration. Geophysics(1995) 14 (3): 165-170.

Brown, G., The History of the Darcy-Weisbach Equation for Pipe Flow

Resistance. Environmental and Water Resources History: pp 34-43.

Carter, R.D., Type Curves for Finite Radial and Linear Gas Flow Systems:

Constant Terminal Pressure Case. Journal of Society of Petroleum

Engineers (1985) 25(5): 719-728.

Celis V., Silva, R., Ramones, M., Guerra, J. and Da Prat, G., A New Model for

Pressure Transient Analysis in Stress Sensitive Naturally Fractured

Reservoirs. SPE Advanced Technology Series 1994, 2 (1): 126-135.

158

Chen, C.C. and Raghavan, R., A Multiply-Fractured Horizontal Well in A

Rectangular Drainage Region. Journal of Society of Petroleum Engineers

(1997) 2(4): 455-465.

Cinco-Ley, H. and Meng, H.Z., Pressure Transient Analysis of Wells with Finite

Conductivity Vertical Fractures in Double Porosity Reservoirs. Paper SPE

18172 presented at the SPE Annual Technical Conference and Exhibition,

Houston, TX., 5-8 October, 1988.

Cinco-Ley, H. and Samaniego-V, F., Transient Pressure Analysis for Fractured

Wells. Journal of Petroleum Technology (1981) 33(9): 1749-1766.

Cinco-Ley, H., Samaniego-V, F. and Dominguez, A. N., Transient Pressure

Behavior for a Well with Finite-Conductivity Vertical Fracture. SPE Journal

(1978)18(4):253-264.

Cheng, Y., Lee, W.J. and McVay, D.A., Improving Reserves Estimates From

Decline-Curve Analysis of Tight and Multilayer Gas Wells. SPE Reservoir

Evaluation & Engineering (2008)11(5): 912-920.

Cipolla, C.L., Lolon, E.P., Erdle, J.C. and Tathed, V., Modeling Well

Performance in Shale-Gas Reservoirs. This paper SPE 125532 was

prepared for presentation at the 2009 SPE/EAGE Reservoir

Characterization and Simulation Conference held in Abu Dhabi, UAE, 19-

21 October 2009.

Clark J.B., A Hydraulic Process for Increasing the Productivity of Oil Wells.

Journal of Petroleum Technology(1949)1(1):1-8.

159

Clarkson, C.R., Pedersen, P.K., Tight Oil Production Analysis: Adaptation of

Existing Rate-transient Analysis Techniques. Paper CSUG/SPE 137352

was presented at the Canadian Unconventional Resources& International

Petroleum Conference held in Calgary, Alberta, Canada, 19-21 October

2010.

Clarkson, C.R., Qanbari, F., Nobakht, M., Heffner, L., Incorporating

Geomechanical Changes into Rate-transient analysis: Example from the

Haynesville Shale. This paper SPE 162526 was prepared for presentation

at the SPE Canadian Unconventional Resources Conference held in

Calgary, Alberta, Canada, 30 October-1 Nov 2012.

Cox, S.A., Gilbert, J.V., Sutton, R.P. et al., Reservoir Analysis for Tight Gas.

Paper SPE 78695 was prepared for the presentation at SPE Eastern

Regional Meeting in Lexington, Kenturky, U.S.A., 23-25 October 2002.

Duong, A.N., Rate-Decline Analysis for Fracture-Dominated Shale Reservoirs.

SPE Reservoir Evaluation &Engineering (2011)14(3): 377-387.

Fetkovich, M.J., Decline Curve Using Type Curves. Journal of Petroleum

Technology (1980)32(6):1065-1077.

Friedel, T., Mtchedlishvili, G., Behr, A. et al., Comparative Analysis of Damage

Mechanisms in Fractured Gas Wells. This paper SPE 107662 was

prepared for presentation at the European Formation Damage

Conference, Scheveningen, The Netherlands, 30 May-1Jun 2007.

160

Gringarten, A.C. and Ramey Jr., H.J., The Use of Source and Green’s Functions

in Solving Unsteady-Flow Problems in Reservoirs. Journal of Society of

Petroleum Engineers (1973) 13(5):285-296.

Gringarten, A.C., Ramey, Jr., H.J. and Raghavan, R. 1974. Unsteady-State

Pressure Distributions Created by A Well with A Single Infinite-

Conductivity Vertical Fracture. Journal of Society of Petroleum Engineers

(1974) 14(4): 347-360.

Guo, G. and Evans, R.D. A Systematic Methodology for Production Modeling of

Naturally Fractured Reservoirs Intersected by Horizontal Wells. This

paper presented at International Conference on Recent Advances in

Horizontal Well Applications, Calgary, Canada, 20-23 March 1994.

Herge, T., Hydraulically Fractured Horizontal Well Simulation. Paper SPE 35506

was presented at the European 3-D Reservoir Modeling Conference,

Stavanger, Norway, April 16-17, 1996.

Horne, R.N. and Temeng, K.O., Relative Productivities and Pressure Transient

Modeling of Horizontal Wells with Multiple Fractures. Paper SPE 29891

was presented at the SPE Middle East Oil Show, Bahrain, 11-14 March

1995.

Jones Jr., F.O., A Laboratory Study of the Effects of Confining Pressure on

Fracture Flow and Storage Capacity in Carbonate Rocks. Journal of

Petroleum Technology (1975) 27(1):21-27.

Kupchenko, C.L., Gault, B.W. and Mattar, L., Tight Gas Production Performance

Using Decline Curves. This paper SPE 114991 was prepared for

161

presentation at the CIPC/SPE Gas Technology Symposium Join

Conference held in Calgary, Alberta, Canada, 16-19 June 2008.

Liang, P., Thompson, J.M. and Mattar, L., Importance of the Transition Period to

Compound Linear Flow in Unconventional Reservoirs. Paper SPE 162646

was presented at the SPE Canadian Unconventional Resources

Conference in Calgary, Alberta, Canada, October 30-November 1, 2012.

Lin, J. and Zhu, D., Modeling Well Performance for Fractured Horizontal Gas

Wells. Paper SPE 130794 was presented at the CPS/SPE International

Oil & Gas Conference in Beijing, China, June 8-10, 2010.

Liu, X., Modeling of Cold Heavy Oil Production Wells. Ph.D. dissertation.

University of Regina, 2006.

Maley, S., The Use of Conventional Decline Curve Analysis in Tight Gas Well

Applications. SPE/DOE 13898 was presented at the SPE/DOE low

Permeability Gas Reservoirs in Denver, Colorado, May 19-22, 1985.

Mayerhofer, M.J., Lolon, E.P., Warpinski, N.R., Cipolla,C.L., Walser,D., and

Rightmire,C.M., What is Stimulated Reservoir Volume? SPE Production

&Operation, (2010)25(1): 89-98.

Medeiros, F., Ozkan, E. and Kazemi, H., Productivity and Drainage Area of

Fractured Horizontal Wells in Tight Gas Reservoirs. SPE Rreservoir

Evaluation & Engineering(2008)11(5):902-911.

Minkoff, S.E., Stone. C.M., Bryant, S., Peszynska, M. and Wheeler, M.F.,

Coupled Fluid Flow and Geomechanical Deformation Modeling. Journal of

Petroleum Science and Engineering (2003) 38 (1-2): 37-56.

162

Mitch, Alberta’s Cardium Formation: Oil Play Overview, from

http://www.beatingtheindex.com/alberta-cardium-formation-oil-play-

overview.

Newman, A.B., Heating and Cooling Rectangular and Cyclindrical Solids.

Industrial and Engineering Chemistry (1936) 28(5): 545-548.

Nobakht, M., Mattar, L. and Anderson, D.M., Simplified Forecasting of

Tight/Shale-Gas Production in Linear Flow. Journal of Canadian

Petroleum Technology (2012a)51(6): 476-486.

Nobakht, M., Clarkson, C.R. and Kaviani, D., New and Improved Methods for

Performing Rate-transient Analysis of Shale Gas Reservoirs. SPE

Reservoir Evaluation& Engineering (2012b)15(3): 338-350.

Olarewaju, J.S. and Lee, W.J., New Pressure-Transient Analysis Model for

Dual-Porosity Reservoirs. SPE Formation Evaluation (1989)4(3):384-390.

Ozkan, E., Brown, M.L., Raghavan, R. and Kazemi, H., Comparison of

Fractured Horizontal-Well Performance in Conventional and

Unconventional Reservoirs. Paper SPE 121290 was presented at the

SPE Western Regional Meeting, San Jose, California, USA, 24-26 March,

2009.

Ozkan, E. and Raghavan, R., New Solutions for Well-Test-Analysis Problems:

Part1-Analytical Considerations. SPE Formation Evaluation (1991) 6(3):

359-368.

Palacio, J.C. and Blasingame, T.A., Decline-Curve Analysis Using Type Curves-

Analysis of Gas Well Production Data. Paper SPE 25909 presented at the

163

Joint Rocky Mountain Regional and Low Permeability Reservoirs

Symposium, Denver, CO, April 26-28, 1993.

Pedrosa Jr., O.A. 1986, Pressure Transient Response in Stress Sensitive

Formations. The paper SPE 15115 was prepared for presentation at the

56th California Regional Meeting of the Society of Petroleum Engineers in

Oakland, C.A. April 2-4 1986.

Pedroso, C.A., Corrêa, A.C.F., Petrobrâs S.A., A New Model of A Pressure-

Dependent-Conductivity Hydraulic Fracture in A Finite Reservoir:

Constant Rate Production Case. The paper was prepared for presentation

at the fifth latin American and Caribbean Petroleum Engineering

Conference and Exhibition held in Rio de Hanairo, Brazil, 30 Aug-3 Sep.

1997.

Penmatcha, V.R. and Aziz, K., Comprehensive Reservoir/Wellbore Model for

Horizontal Wells. Journal of Society of Petroleum Engineers (1999)

4(3):224-234.

Poe Jr., B.D., Evaluation of Reservoir and Hydraulic Fracture Properties in

Geopressure Reservoirs. This paper SPE 64732 was prepared for

presentation at the SPE Internationial Oil and Gas Conference and

Exhibition in China held in Beijing, China, 7-10 November 2000.

Pratikno, H., Rushing, J.A. and Blasingame, T.A., Decline Curve Analysis Using

Type Curves-Fractured Wells. Paper SPE 84287 was presented at the

SPE Annual Technical Conference and Exhibition in Denver, Colorado,

USA, October 5-8, 2003.

164

Raghavan, R., Chen, C.C. and Agarwal, B., An analysis of Horizontal Wells

Intercepted by Multiple Fractures. SPE J. (1997) 2(3): 235-245.

Raghavan, R. and Chin, L.Y., Productivity Changes in Reservoirs with Stress-

dependent Permeability. Paper SPE 77535 presented at the SPE Annual

Technical Conference and Exhibition, San Antonio, Texas. 29 September-

2 October, 2002.

Raghavan, R., Scorer, J.D.T., and Miller, F.G., An Investigation by Numerical

Methods of the Effect of Pressure-dependent Rock and Fluid Properties

on Well Flow Tests. Journal of Society of Petroleum Engineering

(1972)12(3):267-275.

Ralph Veatch, Overview of Current Hydraulic Fracturing Design and Treatment

Technology- Part 1, Journal of Petroleum Technology (1983)35(4): 677-

687.

Ralph Veatch, Overview of Current Hydraulic Fracturing Design and Treatment

Technology- Part 2, Journal of Petroleum Technology(1983)35(5): 853-

864.

Robertson S. Generalized Hyperbolic Equation. Paper USMS 018731 was

provided to the SPE for distribution and publication. Unsolicited in1988.

Russell, D.G., and Truitt, N.E., Transient Pressure Behaviour in Vertically

Fractured Reservoirs. Journal of Petroleum Technology

(1964)16(10):1159-1170; Trans., American Institute of Mining,

Metallurgical, and Petroleum Engineers, Vol.231.

165

Rutqvist, J., Wu, Y.S., Tsang, C.F., Bodvarsson, G., A Modeling Approach for

Analysis of Coupled Multiphase Fluid Flow, Heat Transfer, and

Deformation in Fractured Porous Rock. International Journal of Rock

Mechanics and Mining Sciences (2002)39(4):429-442.

Samaniego, V.F., Brigham, W.E., and Miller, F.G., An Investigation of Transient

Flow of Reservoir Fluids Considering Pressure-dependent Rock and Fluid

Properties. Journal of Society of Petroleum Engineers(1977) 17(2):141-

150.

Stehfest, H., Numerical Inversion of Laplace Transforms; Communications of

ACM (1970)13(1):47-49.

Thompson, J.M., Nobakht, M., and Anderson, D.M., Modeling Well Performance

Data from Overpressured Shale Gas Reservoirs. Paper CSUG/SPE

137755 was presented at the SPE Annual Technical Conference and

Exhibition, Dallas, Texas, 9-12 October 2010.

United States Department of Energy, Shale Gas: Applying Technology to Solve

America’s Energy Challenges, Pittsburgh: National Energy Technology

laboratory (NETL), 2011.

Vairogs, J., Hearn, C.L., Dareing, D.W., and Rhoades, V.W., Effect of Rock

Stress on Gas Production from Low-permeability Reservoirs. Journal of

Petroleum Technology, Sept. (1971)23(9):1161-1167.

Valkó, P.P. and Amini, S., The Method of Distributed Volumetric Sources for

Calculating the Transient and Pseudosteady-State Productivity of

Complex Well-Fracture Configurations. Paper SPE106279 was presented

166

at the SPE Hydraulic Fracturing Technology Conference, College Station,

Texas, 29-31 January, 2007.

Wattenbarger, R.A., EI-Banbi, A.H., Villegas, M.E., Bryan, M.J., Production

Analysis of Linear Flow into Fractured Tight Gas Wells. Paper SPE 39931

was prepared for presentation at the SPE Rocky Mountain Regional/Low-

Permeability Reservoirs Symposium, Denver, 5-8 April 1998.

Yilmaz, O., Nur, A. and Nolen-Hoeksema, R., Pore Pressure Profiles in

Fractured and Compliant Rocks. Unsolicited in1991.

Zeng, F., Modeling of Non-Darcy Flow in Porous Media and Its Application. Ph.D.

dissertation. University of Regina, 2008.

Zeng, F and Zhao, G., 2009. A New Model for Reservoirs with Discrete Fracture

System. Paper 2009-064 was prepared for the Canadian International

Petroleum Conference in Calgary, Alberta, Canada, 16-18 June.

Zhang, J., Kamenov, A. Zhu, D. and Hill, A.D., Laboratory Measurement of

Hydraulic Fracture Conductivities in the Barnett Shale. Paper SPE

163839 was prepared for presentation at the SPE Hydraulic Fracturing

Technology Conference, Woodlands, Texas, USA, 4-6 February 2013.

Zhao, G., Well Testing of Complex Geometry Reservoirs, Ph.D. dissertation,

University of Tulsa, 1999.

167

APPENDIX A

SOURCE FUNCTIONS

The instantaneous source functions are listed below:

(1) An infinite slab source in an infinite reservoir is (Liu, 2006)

)(2)(22

1

t

xxerf

t

xxerfISF

x

i

x

i

x

. ································ (A.1)

(2) An infinite plane source in an infinite reservoir is

)(4

exp

)(2

12

t

xx

t

IPF

xx

x

. ············································ (A.2)

(3) An infinite plane source in an infinite slab reservoir with no-flow boundaries is

(Liu, 2006)

1

2

22

exp

coscos

11

n x

x

a

tn

a

xxn

a

xxn

aNIPF

, ······················ (A.3a)

or

n

x

x

x

x

t

naxx

t

naxx

t

NIPF

)(4

2exp

)(4

2exp

)(2

1

2

2

, ························· (A.3b)

168

(4) An infinite plane source in an infinite slab reservoir with constant pressure

boundaries is

1

2

22

exp

coscos

11

n x

x

a

tn

a

xxn

a

xxn

aCIPF

, ······················ (A.4a)

or

n

x

x

x

x

t

naxx

t

naxx

t

CIPF

4

2exp

4

2exp

2

1

2

2

. ··························· (4.4b)

(5) An infinite slab source in an infinite slab reservoir with no-flow boundaries is

n

x

i

x

i

x

i

x

i

x

t

naxxerf

t

naxxerf

t

naxxerf

t

naxxerf

NISF

2

2

2

2

2

2

2

2

2

1

, ················· (A.5a)

or

1

2

22

cos2

cos2

sin

exp1

41

niiii

x

ii

ii

x

a

xna

a

xxn

a

xxn

a

tn

n

xx

a

a

xxNISF

.

··································································································· (A.5b)

169

(6) An infinite slab source in an infinite slab reservoir with a constant pressure

boundary is (Liu, 2006)

n

x

i

x

i

x

i

x

i

x

t

naxxerf

t

naxxerf

t

naxxerf

t

naxxerf

CISF

)(2

2

)(2

2

)(2

2

2

2

2

1

, ················· (A.6a)

or

a

xn

a

xxn

a

xxn

a

tn

nCISF

ii

iix

x

sin2

sin

2sinexp

1

42

22

. ························· (A.6b)

The typical pressure drop expressions in a box-shaped reservoir with

different kinds of sources are as follows:

For no-flow boundaries in x, y and z directions, the pressure drop expression

with a line source is

dNIPFNIPFNISFSzyxpzy

t

xs 0

,,. ····································· (A.7)

For no-flow boundaries in x, y, and z directions, the pressure drop expression

with a finite slab source is

t

yxsdNISFNISFSzyxp

0

),,( .··············································· (A.8)

More pressure drop expressions with different sources and outer boundary

conditions could be generated by intersection of different source functions listed

above.

170

APPENDIX B

SOLUTIONS OF FLUID FLOW INSIDE HYDRAULIC

FRACTURES

The mathematical model in segment i is

12

2

,1

DiDDi

D

fD

fD

rfDi

D

fD

yyyt

p

CC

q

y

p

, ········································· (B.1)

with initial condition,

0)0( DfD

tp , ········································································· (B.2)

and boundary conditions at Diy and 1Di

y ,

11

DiDi yfD

fD

yD

fD

C

q

y

p

,

DiDi yfD

fD

yD

fD

C

q

y

p

. ·········································· (B.3)

After Laplace transform, I have

12

2

,

DiDDifD

fD

rfDi

D

fD

yyypC

u

C

q

y

p

, ·········································· (B.4)

11

DiDiy

fD

fD

yD

fD

C

q

y

p

,

DiDiy

fD

fD

yD

fD

C

q

y

p

. ········································· (B.6)

The solution to the above system can be obtained with following source function:

00

,,

0

DDs

yyD

fD

DDfDyyP

y

pyyp

DD

. ··········································· (B.7)

The final solution can be written as

rfDi

fD

DiDs

fDiDiDs

fDiDfDiq

uC

CyyPqyyPqyp

,1,1 . ·················· (B.8)

171

For no-flow boundary segments, DD

yy 0 , with a source at 0Dy , I can get

(Van Kruysdijk, 1988)

Cuyy

Cuy

Cuyy

Cuyy

Cuy

Cuyy

Cu

yyP

DD

D

DD

DD

D

DD

DDs

0

0

0

0

0

exp

12exp

cosh2

exp

12exp

cosh2

2

1,

···································································································· (B.9)

The solution for the no-flow boundary segment DiDDiyyy

1 with a source

located at 10

DiDyy can be obtained through Equation (B.9):

0,111001

DiDiDiDDDiDDyyyyyyyy . ···························· (B.10)

Substituting Equation (B.10) into Equation (B.9), I get

Cuyy

Cuyy

Cuyy

Cu

yyP

DiD

DiDi

DiD

DiDs

1

1

1

1

exp

12exp

cosh2

2

1, . ······················ (B.11)

Likewise, the solution for the same segment with a source located at DiDyy

0

can be obtained by letting

0,00

DiDiDiDDDiDD

yyyyyyyy . ·································· (B.12)

Substituting Equation (B.12) into Equation (B.9) yields

172

Cuyy

Cuyy

Cuyy

Cu

yyP

DiDi

DiDi

DiDi

DiDs

exp

12exp

cosh2

2

1,

1. ······················· (B.13)

The solution for the fluid flow inside fractures can be expressed as (Zeng, 2008)

rfDiDifDiDifDiDiDfDiqydqycqybyp )()()()(

1

, ······························· (B.14)

where

Cuyy

Cuyy

DiD

fD

Di

DiD

DiDi

e

e

Cuyy

CuC

yb

)(

)(2

11

1

1

])cosh[(21

)( ·········· (B.15)

Cuyy

Cuyy

DDi

fD

Di

DDi

DiDi

e

e

Cuyy

CuC

yc

)(

)(2

1

])cosh[(21

)(1

·············· (B.16)

uC

Cd

fD

i

. ············································································ (B.17)