Investigation of EOR Performance in Shale Oil Reservoirs by Cyclic Gas Injection

Tao Wan, M.S.

A Dissertation

Submitted to the Graduate Faculty of

Texas Tech University

in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

PETROLEUM ENGINEERING

Approved

Dr. James J. Sheng

Chair of Committee

Dr. Habib K. Menouar

Dr. Lloyd Heinze

Dr. Marshall Watson

Dr. Mohamed Y. Soliman

Dr. Mark Sheridan Dean of the Graduate School

May, 2015

Texas Tech University, Tao Wan, May 2015

Acknowledgements

First and foremost I would like to express sincere gratitude to the committee members, Dr. James J. Sheng, Dr. Habib K. Menouar, Dr. Lloyd Heinze, Dr. Marshall Watson and Dr. Mohamed Soliman. I should give my strong appreciation to my advisor Dr. Sheng for working with me patiently and directing me throughout this expedited process. I was grateful to him for assisting me in improving my writing skills for journal publication. I was impressed by his intensity and enthusiasm in doing every piece of work. His advice on both research as well as on career has been priceless to me. My time at Texas Tech U was made enjoyable in large part due to talk with Dr. Sheng, faculty and our research groups.

I gratefully acknowledge the Petroleum Engineering Department for the financial support and funding sources from Dr. Sheng that made my Ph.D. work possible.

Texas Tech University, Tao Wan, May 2015

Table of Contents

ACKNOWLEDGEMENTS ...... II

ABSTRACT ...... VI

LIST OF TABLES ...... VIII

LIST OF FIGURES ...... X

NOMENCLATURE ...... XV

CHAPTER 1 ...... 1

INTRODUCTION ...... 1

Objectives ...... 4 Organization of this dissertation ...... 5

CHAPTER 2 MODELING OF THE EOR PROCESS IN STIMULATED SHALE OIL RESERVOIRS BY CYCLIC GAS INJECTION ...... 6

2.1. Abstract...... 6 2.2. Introduction ...... 6 2.3. Description of Models ...... 8 2.3.1. MMP determination ...... 10 2.4. Simulation Results and Discussion ...... 17 2.5. Summary ...... 32

CHAPTER 3 EVALUATION OF THE EOR POTENTIAL IN FRACTURED SHALE OIL RESERVOIRS BY CYCLIC GAS INJECTION ...... 40

3.1. Abstract...... 40 3.2. Introduction ...... 41 3.3. Model Setup and Validation ...... 43 3.4. Simulation Results and Discussion ...... 45 3.5. Conclusions ...... 55

iii

Texas Tech University, Tao Wan, May 2015

3.6. References ...... 55

CHAPTER 4 COMPOSITIONAL MODELING OF THE DIFFUSION EFFECT ON EOR PROCESS IN FRACTURED SHALE OIL RESERVOIRS BY GAS FLOODING ...... 59

4.1. Abstract...... 59 4.2. Introduction ...... 60 4.3. Description of Mathematical Model ...... 62 4.4. Model Validation ...... 64 4.5. Simulation Results and Discussion ...... 70 4.6. Conclusions ...... 81 4.7. References ...... 82

CHAPTER 5 EVALUATE THE EOR POTENTIAL OF CO 2 DISPLACEMENT IN SHALE RESERVOIRS USING STAGGERED ZIPPER FRACTURED HORIZONTAL WELLS ...... 87

5.1. Abstract...... 87 5.2. Introduction ...... 88 5.3. Model Description ...... 90 5.4. Results and Discussion ...... 94 5.5. Conclusions ...... 102 5.6. References: ...... 103

CHAPTER 6 NUMERICAL SIMULATION OF THE EXPERIMENTAL DATA IN LIQUID-RICH SHALES BY CYCLIC GAS INJECTION ...... 106

6.1. Introduction ...... 106 6.2. Material and Methods ...... 108 6.3. Simulation model description ...... 110 6.4. Results and Discussion ...... 113 6.5. Conclusions ...... 124

Texas Tech University, Tao Wan, May 2015

6.6. Acknowledgments ...... 125 6.7. References ...... 125

CHAPTER 7 CONCLUSIONS ...... 129

The contribution of this study ...... 131 Recommendations for future work ...... 131 References ...... 132

Texas Tech University, Tao Wan, May 2015

Abstract

The primary oil recovery factor from shale oil reservoirs is only a small percentage of the in-place reserves. The low primary recovery efficiency and the abundance of shale reservoirs provide huge potential for enhanced oil recovery. Well production performance in shale oil or gas reservoirs strongly depends on the size of fracture-network. The induced fractures that connect the natural fracture complexity provide high flow capacity for injected fluid to access the hydrocarbons trapped in shale matrix. However, in such a system, flooding may not be effective to enhanced oil recovery because the injected fluids may break-through to production wells via the fracture network. A cyclic injection scheme is one way to solve this problem.

A simulation approach is used to evaluate the EOR potential from cyclic gas injection. The reason why we initiate such a project is because there is limited research on studying the approaches on recovering shale oil. Conventional gas flooding may not be a good candidate for improving oil recovery in shale oil reservoirs because of the reasons stated above. The deficiency of gas fingering is avoided by using the gas huff-and-puff method (also called cyclic gas injection in this dissertation). Cyclic gas injection is not subject to early breakthrough. It can take the advantage of natural fractures to increase the contact area of the injected solvent with reservoir rocks.

This dissertation also addresses the role of diffusion effect on EOR process in fractured shale oil reservoirs by gas flooding in field-scale displacements and in the presence of viscous flow. A dual permeability model coupled with diffusion that characterized the dispersive-convective flux through nanopores in shale oil reservoirs during gas injection process is presented. The results produced by this model are in good agreement with experimentally measured data in shale rocks. The significance of inclusion of matrix- fracture diffusion rate in the oil phase and diffusion within the matrices in tight shale oil reservoirs is highlighted. The diffusion effect on gas flooding efficiency is summarized, which is implemented in two hydraulically zipper fractured horizontal wells in liquid-rich vi

Texas Tech University, Tao Wan, May 2015 shales.

vii

Texas Tech University, Tao Wan, May 2015

List of Tables

Table 2.1. Peng-Robinson EOS Fluid Description ...... 12

Table 2.2. Binary interaction coefficients ...... 12

Table 2.3. Reservoir Fluid Composition ...... 12

Table 2.4. Pressure dependent mixing parameter ...... 13

Table 2.5. Black-oil and solvent PVT table ...... 13

Table 2.6. Reservoir properties for the model input ...... 13

Table 3.1. Eagle Ford Fluid properties ...... 45

Table 4.1. Core sample properties ...... 66

Table 4.2. Compositional description of reservoir fluid in experimental simulation ...... 66

Table 4.3. Binary coefficient used for experimental simulation ...... 66

Table 4.4. Peng-Robinson EOS Fluid Description ...... 71

Table 4.5. Binary coefficient for Eagle Ford fluid and reservoir flooding case ...... 71

Table 4.6. The effective diffusion coefficients of different components at 2000 psi ...... 71

Table 4.7. Reservoir properties for the model input ...... 71

Table 4.8. Required inputs for the NPV calculations ...... 79

Table 5.1. Peng-Robinson EOS Fluid Description of Eagle Ford Condensate lumping ... 91

Table 5.2. Reservoir properties for the model input ...... 92

viii

Texas Tech University, Tao Wan, May 2015

Table 6.1. Properties of Soltrol-130 (Chevron-Phillips Chemical Company LP) ...... 108

Table 6.2. Properties of C 15 at 95 ˚F ...... 110

Table 6.3. Reservoir and fluid properties used in this study ...... 111

Table 6.4. Measured shale oil recovery factor and oil saturation with a deletion time of 0.05-hr (Yu, 2015) ...... 113

Table 6.5. Measured recovery factor and oil saturation with a deletion time of 40-hr (Yu, 2015) ...... 114

Table 6.6. Operational schedules in a cycle ...... 119

Texas Tech University, Tao Wan, May 2015

List of Figures

Figure 2.1-P-T phase envelopes with varying composition for reservoir fluid ...... 9

Figure 2.2-Pseudo-ternary diagram cell-to-cell MMP simulation ...... 11

Figure 2.3-Slim-tube simulation of recovery for lean gas injection...... 11

Figure 2.4-Yuan’s MMP correlation graph ...... 12

Figure 2.5-Effect of solution lean gas on swelling of reservoir fluid ...... 14

Figure 2.6-Effect of lean gas mole fraction on relative volume ...... 14

Figure 2.7-Reservoir model ...... 17

Figure 2.8-Simulation model of natural fracture-network (implicit) ...... 19

Figure 2.9-Effect of numerical dispersion on oil R.F. vs. time ...... 19

Figure 2.10-Effect of numerical dispersion on C1 prod. vs. time ...... 20

Figure 2.11-Comparison of oil recovery for four component model vs. fully compositional model...... 21

Figure 2.12-Comparison of oil R.F. for different fracture spacing ...... 22

Figure 2.13-Impact of hydraulic fracture on oil R.F...... 22

Figure 2.14-Extrapolation of recovery to zero grid-block size (slim-tube model) ...... 24

Figure 2.15-Gas production rate (RC) and hydrocarbon pore volume injection ...... 25

Figure 2.16-Oil production rate and average reservoir pressure (unit fracture) ...... 26

Figure 2.17-C20 recovery vs. time at 0.2 PVI...... 26 x

Texas Tech University, Tao Wan, May 2015

Figure 2.18-Impact of natural fracture spacing and hydraulic fracture on CO 2 huff-puff oil recovery, BHP of producer = 2500 psi ...... 28

Figure 2.19-BHP effect on ultimate oil recovery (Fracture spacing = 50-ft) ...... 29

Figure 2.20-Reservoir pressure, CO 2 mole fraction in oil phase and oil viscosity variations during CO 2 huff-and-puff ...... 31

Figure 2.21-Period oil production (unit fracture production, entire horizontal well production should be 20 times) ...... 32

Figure 3.1-Stimulated fracture-network in SRV (Dx=200) ...... 46

Figure 3.2-Stimulated fracture-network in SRV (Dx=100) ...... 46

Figure 3.3-Impact of stimulated fracture-network spacing on EOR ...... 47

Figure 3.4-Impact of the conductivity of fracture-network in the SRV on production profiles ...... 48

Figure 3.5-Irreversible process of fracture conductivity reduction ...... 49

Figure 3.6-Stress-dependent F CD of fracture-network effect on EOR ...... 50

Figure 3.7-Permeability distribution for the non-uniform fracture network ...... 51

Figure 3.8-Effect of stress-dependent FCD on EOR performance ...... 51

Figure 3.9-Non-SRV embedded in SRV (Hydraulic fracture spacing = 400-ft) ...... 53

Figure 3.10-Hydraulic fracture spacing effect on EOR ...... 54

Figure 3.11-Fracture scenarios effect on EOR (DX=200-ft) ...... 54

Figure 4.1-Comparative cumulative oil recovery ...... 67

Texas Tech University, Tao Wan, May 2015

Figure 4.2-2D single porosity model with 10 x 10-m matrix blocks ...... 68

Figure 4.3-Comparison of C 3 recovery for the dual permeability model with single porosity model ...... 69

Figure 4.4-Comparison of C 3 recovery by CO 2 injection in shale gas reservoirs ...... 70

Figure 4.5-The horizontal well pair perforated and stimulated in a staggered pattern ..... 72

Figure 4.6-Simulation Unit ...... 73

Figure 4.7-Comparison of Coats’s model results with our model for matrix permeability km = 1 mD ...... 74

Figure 4.8-Effect of diffusion in the oil phase and within matrices on shale oil recovery 74

Figure 4.9-Peclet number in the oil phase in the matrix (k m=1 md) at 7000 days ...... 75

Figure 4.10-Peclet number in the oil phase in the shale matrix (k m=1E-04 md) at 7000 days ...... 75

Figure 4.11-Oil recovery vs. PVI ...... 77

Figure 4.12-Oil recovery vs. time ...... 77

Figure 4.13-Comparative NPV by two different injection rates ...... 79

Figure 4.14-Comparison of gas production rates and cumulative gas injection by two different injection rates ...... 80

Figure 4.15-Effect of natural fracture spacing on gas injection performance ...... 80

Figure 5.1A-The horizontal well pair perforated and stimulated in a staggered pattern ... 89

Figure 5.2-The horizontal well pair stimulated in a staggered pattern ...... 90

xii

Texas Tech University, Tao Wan, May 2015

Figure 5.3-Phase diagram of gas condensate behavoir ...... 91

Figure 5.4-Reservoir pressure changes during gas injection ...... 92

Figure 5.5-Effect of solution gas on swelling of ...... 93

Figure 5.6-Relative volume curve reservoir fluid by CO 2 ...... 94

Figure 5.7-The liquid dropout curve for constant-composition expansion experiment at 335F on the gas condensate mixture...... 94

Figure 5.8-Effect of numerical dispersion on C 1 recovery vs. time ...... 96

Figure 5.9-Effect of numerical dispersion on gas R.F. vs. time ...... 96

Figure 5.10-Unit fracture controlled stimulated reservoir volume vs. entire horizontal well SRV results ...... 97

Figure 5.11-BHP impact on gas recovery performance ...... 97

Figure 5.12-Gas recovery factor-Darcy flow vs. gas recovery-non-Darcy flow ...... 98

Figure 5.13-Impact of hydraulic fracture spacing between injection well and production well on gas recovery and C9 recovery ...... 99

Figure 5.14-Global mole fraction of CO 2 and C 1 changes during CO 2 flooding ...... 99

Figure 5.15-Condensate saturation distribution in or at vicinity of the fracture ...... 100

Figure 5.16-Gas relative permeability on or vicinity of the fracture ...... 101

Figure 6.1-Experimental setup and apparatus ...... 110

Figure 6.2-Base simulation model ...... 111

Figure 6.3-Shale oil production response in cyclic gas injection processes from 1th-5th xiii

Texas Tech University, Tao Wan, May 2015 cycle ...... 114

Figure 6.4-Effect of depletion time on CGI recovery performance ...... 115

Figure 6.5-Effect of grid block size on calculated oil R.F...... 116

Figure 6.6-Effect of fracture permeability on cyclic gas injection performance ...... 117

Figure 6.7-Comparison of simulation results with experimental data (0.05 hours) ...... 117

Figure 6.8-Comparison of simulation results and experimental data (40 hours) ...... 118

Figure 6.9-Comparison of simulated oil saturation and measured data ...... 119

Figure 6.10-Pressure variations in one cycle of huff-n-puff process (40-hour depletion) ...... 119

Figure 6.11-Comparison of simulation results and experimental data (Pi=1000 psi, depletion time = 40-hr) ...... 121

Figure 6.12-Effect of diffusion on ultimate oil recovery ...... 121

Figure 6.13 -Effect of soak duration on production response ...... 122

th th Figure 6.14-N2 mole fraction in the shale matrix from 1 -8 cycle (1-hour soak time) 123

xiv

Texas Tech University, Tao Wan, May 2015

Nomenclature

B = the phase formation volume factor

͖*=1/ ̼*, ͖"=1/ ̼"

BHP= bottom-hole flowing pressure

̾$% =Binary diffusion coefficient between component i and j in the mixture

Fo= Forchheimer Number

FCD = fracture conductivity

͂2% = the well bore pressure head between the connection and the well’s bottom-hole datum depth.

HF = Hydraulic fractures

̈́=Mass flux

k=Permeability

͟-=Relative permeability

Kfeff =effective fracture permeability

Lx , L y , L z =Fracture spacing in x, y and z direction

MMP= minimum miscibility pressure

NF = natural fracture-network

PV= pore volume

Texas Tech University, Tao Wan, May 2015

PVI= injected pore volume

͊2=the bottom hole pressure of the well

Po,P g = Oil and gas phase pressure, respectively

Pcog =Capillary pressure between oil and gas

ͥ$,%= the component i molar rate in the j phase

ͥ+,%= the volumetric flow rate of phase p in connection j at reservoir condition

ͦ*= the pressure equivalent radius

ͦ2=the well bore radius

Rs=solution gas, scf/STB

Rv=oil vapor in gas, STB/scf

So, S g= Saturation of oil and gas, respectively

ͩ=Velocity

Vblock =Volume of the grid-block

&=Porosity

= the angle of segment of connecting with the well

!= mixing parameter

./* =stock tank oil density

./" =stock tank gas density

xvi

Texas Tech University, Tao Wan, May 2015

=non-Darcy beta factor

σ =Shape factor, 1/ft 2

τ ij =mass transfer of component i in phase j caused by both convection and diffusion

*, "=Viscosity of oil and gas, cp

ω ij =The mass fraction of component i in phase j divided by the total mass of all components in that same phase

= Molar density

Subscripts c=components f=fracture i=Component index j=Phase index, o=oil phase, g=gas phase m=matrix

xvii

Texas Tech University, Tao Wan, May 2015

Chapter 1

Introduction

Cyclic gas injection is an effective and quick responding enhanced-oil-recovery method in intensely naturally fractured or hydraulically fractured reservoirs. The short payout is a good characteristic to attract industry’s interest in investing in these projects. One of the limitations of gas or water injection in tight shale oil reservoirs is that the fluid injectivity is low due to the nature of very low permeability of shale. EIA (2013) published a report about the world shale gas and shale oil resource assessment, which provided the oil recovery efficiency factor of the 28 U.S. tight oil plays. The primary oil recovery factors in shale oil plays range from 1.2% to 8.4%. Another challenge of gas flooding is that the injected gas is subject to early breakthrough in densely fractured shale gas or oil reservoirs, resulting in poor flooding performance. Cyclic gas injection (CGI) in a single horizontal well is not affected by early gas breakthrough. Compared to gas flooding, cyclic gas injection is a more effective recovery process in tight shale oil reservoirs. This dissertation presents our simulation work on using cyclic injection method to improve shale oil recovery. Core flooding and simulation outputs showed that it is favorable to implement cyclic gas injection enhanced oil recovery process in shale oil reservoirs.

Enriched-gas displacements are widely used as secondary recovery process because it is possible to obtain high local displacement efficiencies for enriched gas with reservoir oil

(Johns et al. 2000). Laboratory studies (Zhang et al. 2004) of the effect of CO 2 impurities

on oil recovery efficiency showed that the addition of nitrogen or methane in the CO 2 injection stream tends to increase the minimum miscibility pressure. Shyeh-Yung (1995) found that lower solvent mobility may still recover oil efficiently in field displacements owing to better sweep efficiency compensating for the loss in local displacement efficiency. Chapter 3 examines the effect of injection gas composition effect in recovering oil from liquid-rich shales.

Texas Tech University, Tao Wan, May 2015

Integrated technologies including argon ion-milled scanning electron microscopy (SEM), lateral resistivity logging and core analyses were used by Kurtoglu (2013) to characterize the mineralogy, pore structure and fracture properties in the Bakken formation. She discovered that micro-fracture network in shale reservoirs provides the main pathway for injected fluid to transport to the matrix and contact with oil in the matrix. It is concluded that the well productivity in the Bakken formation is largely attributed to the presence of micro-fracture network. Landry et al. (2014) used the SEM to study the Eagle Ford shale matrix-fracture connectivity and they observed two types of fractures. The first type of calcite filled fracture is thick and short parallel to bedding, which is less likely to significantly influence fluid flow. The second type of fracture is thin and long oriented preferentially perpendicular to bedding that contributes to fluid flow and well productivity in shale reservoirs. Chapter 3 mainly addresses the effect of natural fracture properties on well production performance during enhanced oil recovery process in shale oil reservoirs by cyclic gas injection.

Most of the available literature on performance of cyclic gas injection focused on reservoir conditions that have high permeability. Recent studies (Chen et al. 2014; Gamadi, et al. 2013; Wan et al. 2013) showed that cyclic gas injection could be a viable method to improve the oil recovery in shale oil reservoirs. Simulation results showed cyclic gas injection combined with modern technologies such as horizontal well drilling and hydraulic fracturing has achieved promising recovery results in low permeability formation. Kovscek et al. (2008) presented a series of experimental results of using CO 2 injection to improve oil recovery in low permeability shale rocks (0.02-1.3 mD). Countercurrent flow and concurrent injection schemes were employed to evaluate the oil recovery potential. Their experimental results showed that the incremental oil recovery from near miscible CO 2 injection is around 35%, in which 25% oil recovery obtained for the countercurrent flow mode and 10% for the concurrent flow. Wang et al. (2013)

reported experimental results of CO 2 huff-and-puff process operated in a 973 mm long composite core with an average permeability of 2.3 mD. Their experimental results showed that the first operation cycle contributes above half of the total oil production and additional oil produced from subsequent cycles is significantly decreased compared to

Texas Tech University, Tao Wan, May 2015 previous cycles. Kurtoglu (2013) evaluated the feasibility of enhanced oil recovery by conventional gas injection and gas huff-n-puff in Bakken fields using simulation techniques. He used a dual-porosity reservoir model to simulate the CO 2 huff-n-puff flooding performance in the Bakken field. Unfortunately, the diffusion effect was not included in their model because of numerical convergence issues. However, studies (Javadpour et al. 2007; Sakhaee-Pour and Bryant 2012; Ozkan et al. 2010) suggest that molecular diffusion is an important recovery mechanism in the mobilization and recovery of oil in very low permeability shale oil or gas reservoirs. Chapter 4 is dedicated to investigating the effect of diffusion on shale oil recovery performance by secondary gas injection process.

Cyclic gas flooding in conventional field applications (Miller et al. 1998; Lino 1994; Bardon et al. 1986; Gondiken 1987) has been performed successfully. In shale gas or oil reservoirs, the presence of fissures or induced hydraulic fractures provides highly conductive paths for injected gas to diffuse or penetrate into the nano-permeable matrix, which makes it favorable to perform cyclic gas injection. Chen et al. (2014) investigated the effect of reservoir heterogeneity on CO 2 huff-n-puff recovery process using (UT- COMP) simulation approaches. Gamadi et al. (2013) presented a series of experimental data of cyclic gas injection in Barnett, Mancos and Eagle Ford shale cores. They investigated the effect of injection pressure, soaking time and the number of injection cycles on oil recovery performance by N 2 huff-n-puff process. Improved oil recovery efficiency factor was observed both in their simulation work and experimental work.

However, the available literature provides limited information on the laboratory examination of the applicability of gas huff-n-puff in very low permeability shale rocks. Very limited field or laboratory data are available on the performance of cyclic gas injection in shale oil reservoirs. In this dissertation, chapter 6 focuses on examining the relevant parameters that affect the performance of cyclic gas injection process in detail. The principle recovery mechanisms in shale oil reservoirs were discussed. Chapter 6 interrelated numerical simulation approach with the laboratory data to analyze the significance of possible parameters that have on the performance of cyclic injection recovery process. 3

Texas Tech University, Tao Wan, May 2015

Sanaei et al. found that desorption has negligible effect on gas and condensate recoveries from Eagle Ford shales. It is believed that desorption effect has to be considered in shales with high total organic content (TOC). Otherwise, an underestimation of ultimate gas recovery is expected. The significance of desorption effect on oil or gas recovery is not considered in this dissertation.

Objectives

The objectives of this work include:

1. Evaluate the performance of cyclic gas injection to improve oil recovery in liquid- rich shales using numerical simulation approaches.

2. Evaluate the effects of fracture spacing, the size of fracture-network, fracture connectivity (uniform and non-uniform) and stress-dependent fracture-network conductivity on production performance of shale oil reservoirs by secondary cyclic gas injection.

3. Simulate the importance of molecular diffusion in fractured shale oil reservoirs. Proper representation of diffusion-convention mass transfer between matrix and fracture is critical to examining the recovery mechanism in shale reservoirs.

4. Investigate the significance of possible factors on gas huff-n-puff recovery process in shale oil reservoirs via simulation approaches.

5. Combined simulation approaches with experimental data to show the potential of cyclic gas injection method in shale oil reservoirs and discuss the diffusion effect in laboratory-scale flooding in liquid-rich shales.

Our simulation results have demonstrated the potential of gas huff-n-puff injection to improve oil recovery in shale oil reservoirs. We also examined the effect of diffusion on improved oil recovery performance by cyclic injection process. Our simulation results benchmarked with experimental observations showed that molecular diffusion played a significant role in the mobilization of oil in lab scale. The Computer Modeling Group 4

Texas Tech University, Tao Wan, May 2015

(CMG) software is used in this dissertation to perform the numerical simulation studies.

Organization of this dissertation

Chapter 1 gives an introduction, background and objectives of this study.

Chapter 2 describes the simulation results of the effect of injected fluid compositions on enhanced oil recovery process.

Chapter 3 primarily discusses the effect of fracture properties on enhanced oil recovery process by cyclic gas injection. It first only considers the presence of natural fractures. Then, both natural fracture-network and hydraulic fractures are included to show the impact of stress-dependent fracture conductivity on gas injection performance.

Chapter 4 presents the effect of diffusion on well productivity performance in liquid-rich shales by gas flooding. This chapter focuses on discussing the recovery mechanism in recovering oil in tight shale reservoirs.

Chapter 5 proposes an approach to increase stimulated reservoir volume in shale oil or gas reservoirs. The design of staggered pattern of zipper frac is to expose more reservoir rocks to injected solvent. The existence of micro-fracture network or highly conductive induced fractures is a key factor that impacts oil recovery by secondary cyclic gas injection.

Chapter 6 presents a numerical simulation study of experimental data in shale oil reservoirs by cyclic nitrogen injection at an immiscible condition. A model is developed to achieve a good history matching of the experimental data. Then, this model is used to evaluate the significance of possible factors that affect the secondary recovery performance in shale oil reservoirs.

Chapter 7 provides the conclusions of this dissertation, recommendations for future work and contributions of this study.

Texas Tech University, Tao Wan, May 2015

Chapter 2 Modeling of the EOR Process in Stimulated Shale Oil Reservoirs by Cyclic Gas Injection

2.1. Abstract

Cyclic gas injection is considered as an effective and quick responding recovery process that has been widely used in the worldwide oil industry. Cyclic gas injection technique was introduced in our earlier publication (Wan et al., 2013a) for improving oil recovery in hydraulically fractured shale oil reservoirs. In this chapter, we focus on the effect of injected gas composition on oil recovery. Different injection gas scenarios such as lean gas, rich gas and CO 2 were included in the simulation models to represent the EOR mechanisms of vaporizing, condensing or a combined condensing/vaporizing process.

Our simulation results indicate that the stimulated natural fractures are critical to improving oil recovery and well productivity performance in shale oil reservoirs. Since the interaction of the induced hydraulic fractures with pre-existing natural fractures and fissures makes the hydraulically fractured reservoir modeling very challenging in shale oil/gas reservoirs, a dual-continuum model was used by varying the fracture permeability and intensity to attain a better characterization of the natural fractures. We conclude that cyclic gas injection in shale oil reservoirs employing hydraulically stimulated fractures is feasible to improve substantial amounts of oil production than primary production.

2.2. Introduction

Cyclic gas injection is an efficient enhanced-oil-recovery method in intensely naturally fractured or hydraulically fractured reservoirs. Artun et al. (2011) used a neural-network based proxy models to perform the parametric study of the cyclic gas injection process in order to develop optimized design schemes to maximize the production efficiency of the process in the naturally fractured reservoirs. Artificial neural networks model has the 6

Texas Tech University, Tao Wan, May 2015 advantage of extracting complex and non-linear relationships, but this black-box does not represent the actual physics phenomena of reservoir fluid flow and production performance. Ivory et al. (2010) investigated the cyclic solvent injection process in Cold Lake and Lloydminster heavy oil reservoirs. Their experiment result indicated that the potential secondary recovery was 50% by cyclic solvent (28% C 3H8 -72% CO 2) injection process after primary production. Haines et al. (1990) studied the cyclic natural gas injection for enhanced recovery of light oil from waterflooded fields. Their core flooding and numerical simulation results indicated that response to natural gas huff-n-puff enhanced oil recovery was a function of cyclic injection process variables, such as cycles used and gas slug size. They concluded that repressurization and gas relative permeability hysteresis are the major recovery mechanisms. Alves et al. (1990) used a Peng-Robinson equation-of-state (EOS) to predict the potential of miscible gas displacement as a secondary production in tight carbonate formation in which rich gas or lean gas were used as injection gas. The PR-EOS fluid model has to be tuned and compared with the PVT experimental data (Moortgat et al. 2009). Vega et al. (2010) performed an experimental study to investigate the oil recovery from CO 2 injection in 1.3 mD permeability siliceous reservoirs, as they showed the improved oil recovery was significant. Unfortunately, their compositional simulation results could not reproduce their experimental results.

Cyclic gas recovery in field applications (Miller et al. 1998; Lino 1994; Bardon et al. 1986; Gondiken 1987) has been performed successfully. The above cases demonstrate that cyclic gas injection played an important role in enhanced-oil-recovery in conventional reservoirs. Recent studies (Chen et al. 2013; Gamadi, et al. 2013; Wan et al. 2013) showed that cyclic gas injection could be a viable method to improve the oil recovery in shale oil reservoirs, but those studies were performed without detailed analysis of the phase behavior of reservoir oil/injected gas system and the effect of injected gas compositions on enhanced oil recovery process.

Texas Tech University, Tao Wan, May 2015

2.3. Description of Models

The reservoir fluid composition data (There was no field shale oil compositional data available when this work was done in 2013) is from fifth SPE comparative solution project published data (Killough et al. 1987) which aimed to illustrate the comparison results between a four-component black-oil miscible flood simulator and a fully compositional reservoir simulation model. Later on, Pu (2013) presented the Bakken reservoir fluid composition data which was collected from a well T23N R58E SEC19 of the Elm Coulee Field. Kurtoglu (2013) presented the Bakken PVT data in Reunion Bay, Bailey and Murphy Creek field. She also used the software to tune the fluid composition data to match the lab measured data. It is very difficult to obtain in-situ fluid composition data. Fig.2.1 shows the P-T phase envelope and quality fraction lines of reservoir oil calculated by Heidemann method (1980). Table 2.1 presents the pseudo-component description and input for Peng-Robinson equation of state calculations. Binary interaction coefficients were given in Table 2.2. The initial reservoir oil compositions are shown in

Table 2.3 which represent extremely light oil. The injected solvent contained 77% C 1, 20%

C3 and 3% C 6. Todd and Longstaff (1972) proposed a method of modeling miscible gas displacement performance without reproducing the fine structure of the flow. One prominent feature of their method is that it requires modifying the physical properties and flowing characteristics of the miscible fluids in a three-phase black-oil model. A mixing parameter ! is required to determine the degree of miscibility between the miscible fluids within a grid-block. Fayers et al. (1992) recommended a formula (Eq. 2.1) to quantify the mixing parameter by comparing the effective mobilities in the Todd-Longstaff model with the Koval model (1963). Table 2.4 shows the pressure dependent mixing parameter.

u x = 1 − 4 log 0.78 + 0.22 * / log * (2.1) ! ʮ ʦ "ʧ ʯ ʦ "ʧ

Texas Tech University, Tao Wan, May 2015

Bubblepoint=2302

Critical point

Figure 2.1-P-T phase envelopes with varying composition for reservoir fluid

In order to compare the four-component Todd-Longstaff miscible flood model with fully compositional simulation results, a differential liberation expansion was simulated by Winprop to generate the black-oil PVT properties that correspond with the EOS characterization (Killough et al. 1987). The Coats (1982) method uses the conservation equations to obtain the oil phase properties such as formation volume factor and gas-oil ratio.

+ = + (2.2) ͦ " " * . * ͦ ͥ " " * . * ͥ ͐ Ƴ͖ ͍ ͖ ͌ ͍ Ʒ ͐ Ƴ͖ ͍ ͖ ͌ ͍ Ʒ

+ = + (2.3) ͦ " 1 " * * ͦ ͥ " 1 " * * ͥ ͐ Ƴ͖ ͌ ͍ ͖ ͍ Ʒ ͐ Ƴ͖ ͌ ͍ ͖ ͍ Ʒ

Where ͖* = 1/̼*, ͖" = 1/̼". The stock tank densities are obtained from the output of

the separators at the saturation pressure. ͌. and ͖* can be obtained by using Whitson and Torp (1983) method by flashing the reservoir fluids at surface separation conditions. Most existing commercial software uses this method to simulate the differential libration expansion. Table 2.5 presents the black-oil and solvent PVT data generated from Table 9

Texas Tech University, Tao Wan, May 2015

2.1 EOS characterization.

2.3.1. MMP determination

Local displacement efficiency by gas or solvent injection process is closely dependent upon the minimum miscibility pressure (MMP). Most gas or solvent injection cases operate in a regime in which true miscibility is not achieved, but high recoveries could be possibly attained even so. The analytical method (Wang et al. 1998; Yuan et al.2005) for MMP determination focused on finding the key crossover tie lines for a dispersion-free displacement when one of the key tie lines becomes a critical tie-line (a tie line of zero length). Johns and Orr (1993) developed a method to find the key crossover tie lines from the geometric construction in the analytical solution to control the development of miscibility in condensing/vaporizing systems. Jessen (1998) developed an algorithm based on the key tie line approach to reduce the calculation of MMP time consumption and improve the method robustness. In this study, we use cell-to-cell simulation method provided by Winprop to determine the MMP of the given solvent composition and reservoir fluids. The pseudo-ternary diagram (Fig.2.2) is generated from the calculations to study the vaporization or extraction process which interprets MMP of solvent as 3440 psi. We also developed a slim-tube simulation model that has good agreement with the calculated MMP by tie-line method. The grid sensitivity was run by using 10, 20, 50, 100 1D gridblocks to illustrate the effect of numerical dispersion. The slim-tube simulation results from these different grid-block sensitivity studies show that there is only slight difference in cumulative oil recovery vs. time if the number of grid-blocks is larger than 100. Then, slim tube displacement simulation was performed at eight different pressures

(Fig.2.3) by employing 100 1D grid-blocks. CO 2 is more favorable to achieve miscibility with reservoir oil than lean gas. The miscibility development is achieved by a combined

vaporizing/condensing mechanism. The miscibility achieved at 2300 psia by pure CO 2 displacement calculated by the key tie line method. Yuan et al. (2005) proposed a correlation to calculate the MMP for CO 2 floods that gives us 2617 psi. As compared to what we get from key tie-line method, the result from correlation overestimates a little bit 10

Texas Tech University, Tao Wan, May 2015 the minimum miscibility pressure.

Figure 2.2-Pseudo-ternary diagram cell-to-cell MMP simulation

100 Recovery at 1.2 PV

R.F. % at 1.2 PV at % R.F. 60

50 2000 2500 3000 3500 4000 4500 Pressure, psi

Figure 2.3-Slim-tube simulation of recovery for lean gas injection

Texas Tech University, Tao Wan, May 2015

4000

3500

3000

2500

2000 MMP for CO2injection

1500 0.4 0.3 350 300 0.2 250 0.1 200 150 0 100 mol% of C2-6 MWc7+

Figure 2.4-Yuan’s MMP correlation graph Table 2.1. Peng-Robinson EOS Fluid Description

Components Initial Pc (atm) Tc (k) Acentric Fac. MW Vc Parachor Comp. C1 0.5 45.44 190.6 0.013 16.04 0.0998 39.84318 C3 0.03 41.94 369.8 0.1524 44.1 0.2005 126.7644 C6 0.07 29.73 507.4 0.3007 86.18 0.3698 250.622 C10 -15 0.2 20.69 617.7 0.4885 142.29 0.6297 403.6545 C15 -20 0.15 13.61 705.6 0.65 206 1.0423 560.6208 C20+ 0.05 11.02 766.7 0.85 282 1.3412 724.5072

Table 2.2. Binary interaction coefficients

C1 C3 C6 C10 -15 C15 -20 C20+ C1 zero 0.0 0.0 0.0 0.05 0.05 C3 0.0 zero 0.0 0.0 0.005 0.005 C6 0.0 0.0 zero 0.0 0.0 0.0 C10 -15 0.0 0.0 0.0 zero 0.0 0.0 C15 -20 0.05 0.005 0.0 0.0 zero 0.0 C20+ 0.05 0.005 0.0 0.0 0.0 zero

Table 2.3. Reservoir Fluid Composition

Comp. Initial Comp. Injection solvent Comp. C1 0.5 0.77 C3 0.03 0.20 C6 0.07 0.03 C10 -15 0.2 0.00 C15 -20 0.15 0.00 C20+ 0.05 0.00

Texas Tech University, Tao Wan, May 2015

Table 2.4. Pressure dependent mixing parameter

Pressure (psi) Oil viscosity Gas viscosity Solvent viscosity ! 3500 0.175 0.0214 0.031 0.730605 4000 0.167 0.0232 0.034 0.733841 4500 0.159 0.025 0.037 0.736953 4800 0.155 0.0261 0.038 0.738665

Table 2.5. Black-oil and solvent PVT table

Pressure GOR Bo Bg (bbl/ft3) Oil Visc Gas Visc Solvent Visc (PSIA) (SCF/STB) (cp) 14.7 0.0 1.03480 0.211416 0.310 0.0107 0.011 500.0 117.600 1.10170 0.00592417 0.295 0.0127 0.012 1000.0 222.600 1.14780 0.0028506 0.274 0.0134 0.013 1200.0 267.700 1.16770 0.0023441 0.264 0.0138 0.014 1500.0 341.400 1.19970 0.0018457 0.249 0.0145 0.016 1800.0 421.500 1.23500 0.0015202 0.234 0.0153 0.018 2000.0 479.000 1.26000 0.00136023 0.224 0.0159 0.019 2302.3 572.800 1.30100 0.0011751 0.208 0.0170 0.022 2500.0 634.100 1.32780 0.00110168 0.200 0.0177 0.023 3000.0 789.300 1.39560 0.0009852 0.187 0.0195 0.027 3500.0 944.400 1.46340 0.0009116 0.175 0.0214 0.031 4000.0 1099.500 1.53210 0.0008621 0.167 0.0232 0.034 4500.0 1254.700 1.59910 0.0008224 0.159 0.0250 0.037 4800.0 1347.800 1.63980 0.0008032 0.155 0.0261 0.038

Table 2.6. Reservoir properties for the model input

Initial Reservoir Pressure 6425 psia Reservoir Temperature 320 Fo Saturation Pressure 2302 psia Rock Compressibility 5.0E-06 Porosity 6 % Permeability of shale 100 nano-Darcy Water Density 62.4 lb/cuft

Texas Tech University, Tao Wan, May 2015

4000 5.5 Psat 3800 5

3600 Swelling Factor 4.5

3400 4

3200 3.5

3000 3

2800 2.5

Saturation Pressure, psia Saturation 2600 2

2400 1.5

2200 1 0 0.2 0.4 0.6 0.8 1 Injected lean gas mole composition

Figure 2.5-Effect of solution lean gas on swelling of reservoir fluid

Lean gas injection

2.7 0, mole fraction 0.2, mole fraction

2.2 0.4, mole fraction

0.6, mole fraction 1.7 Relative Volume Relative 1.2

0.7 0 1000 2000 3000 4000 5000 6000 7000 Pressure, psia

Figure 2.6-Effect of lean gas mole fraction on relative volume

Texas Tech University, Tao Wan, May 2015

The swelling test was simulated by varying proportions of injection gas mixed with original reservoir oil. The initial reservoir pressure is 6425 psi which is far above the bubble point pressure at 2302 psi. When CO 2 or gas dissolves in oil, the liquid volume will expand and increase. Fig.2.5 shows the effect of lean gas injection on the saturation pressure. Constant composition expansion experiment (CCE) is started at a pressure higher than the saturation point. As the pressure is lowered, oil volume expands and is recorded. When the pressure dropped below the bubble point pressure, the measured volume will increase rapidly because gas evolves from the oil. CCE simulation can give us information about the relative volumetric amounts of oil and gas in the reservoir at various stages of the lifetime of the reservoir (Pedersen et al. 2006). Fig.2. 6 illustrates the effects of injected gas mole fraction on the relative volumes. As injected gas mole fraction increases, the relative volume goes up because more gas can evolve from the oil when pressure drops below the bubble-point pressure.

The reservoir rock properties we used in this model are based on the published data in Eagle Ford shale as was previously used as shown in Table 2.6 (Hsu and Nelson, 2002; Chaudhary et al., 2008; Bazan et al., 2010; Wan et al., 2013). The initial reservoir pressure for this field is 6,425 psi. The producer is subject to minimum bottom-hole pressure constraint (BHP) of 2500 psi and is produced for 1800 days (5 years) as the natural depletion. Local displacement efficiency by gas or solvent injection process is related to the controlled BHP that dictates whether local miscible flooding can occur or not. In this study, we propose cyclic CO 2/solvent injection in a single hydraulically stimulated horizontal well to improve oil recovery in 0.0001-mD shale oil reservoirs. The horizontal well drilling combined with hydraulic fracturing technique was undertaken in combination with cyclic gas injection, which indicates as an effective technique for improving oil recovery in shale oil reservoirs. The dimension of the shale reservoir is 2000-ft long ×1000-ft wide ×200-ft thick, as shown in Fig.2.7. The horizontal well is stimulated with 10 transverse fractures each placed 200-ft apart. We will simulate only one single hydraulic fractured stimulated reservoir volume on the basis of flow symmetry.

Texas Tech University, Tao Wan, May 2015

The field cumulative oil production and production rate can be obtained simply multiplied by the number of effective fractures.

Due to the complex propagation of induced hydraulic fractures in gas shale reservoirs associated with the interaction of pre-existing fissures or natural fractures, it has been suggested that modeling the complex fracture system should be classified into three categories (Cipolla et al., 2008; Cipolla et al., 2009; Dershowitz et al., 2011):

1: Planar hydraulic fractures provide the dominated conductive flow path to the wellbore if the injected proppant is mainly concentrated in the single hydraulic fracture.

2: Planar hydraulic fractures act as the primary flow path to the wellbore supplemented by the natural fracture if there is slight amount of frac fluids leakage.

3: Natural fracture-network dominates the flow path in the gas reservoirs if the proppant is evenly distributed in the entire fracture-network.

According to the classifications of stimulated reservoirs, the reservoir simulation in this study is separated into two stages:

1. Investigate the cyclic gas injection EOR in naturally fractured shale oil reservoirs, without induced hydraulic fractures.

2. Investigate the gas huff-n-puff EOR in hydraulic fractures combined with natural fractures.

Texas Tech University, Tao Wan, May 2015

Figure 2.7-Reservoir model 2.4. Simulation Results and Discussion

In this study, the dual permeability model was used to simulate the natural fracture- network. The dual permeability model allows the communication between matrices of the adjacent grid blocks in addition to the expected inter-block fracture to fracture flows and the matrix to fracture fluid flow. In the base model, the fracture spacing was assumed to be 200-ft in both X and Y direction. The natural fracture network is assumed to be contained within an orthogonal system of continuous, uniformly spaced and constant width. In each of these blocks, two perpendicularly crossed 0.001-ft wide fracture (approximately actual fracture width) is assumed to exist which runs through the 200-ft x

200-ft block. Therefore, the average fracture porosity is 0.00001=2xV frac /V block (2x (0.001 x 200 x 200 ) / (200 x200x200 )). The fracture spacing is used to calculate the shape factor developed by Gilman and Kazemi (1983). Specifying small values of fracture spacing will result in a large value of shape factor, hence the matrix-fracture transfer rate will also increase (CMG Manual, 2009).

The simulation model assumes that the SRV consists of a uniform network-fracture with conductivity of 4 mD-ft (4 mD-ft is a reasonable value used for the conductivity in the 17

Texas Tech University, Tao Wan, May 2015 stimulated volume as discussed by Rubin (2010)). The 200-ft x 1000-ft single fracture controlled SRV is simulated by using 80 blocks (50 ft x 50 ft). In the case with 4 mD-ft fracture, the effective fracture permeability equals to 0.08 mD (4/50) which should conserve the fracture conductivity of 4 mD-ft. The non-Darcy flow effect should be considered for gas injection in fractured reservoirs. The Forchheimer equations are written as (Dake, 1978):

2.73 × 10 ͥͤ = + ͦ, = (2.4) ͤ͘ ͥ.ͥͤͨͩ ͦ͘ ͟ ͟

= 1 + , = (2.5) ͤ͘ ͟ ʚ ̀ͣ ʛ ̀ͣ ͦ͘ ͟ Fo is referred to as the Forchheimer Number (CMG Manual, 2009). In actual 0.001-ft width size fracture model, the fracture permeability is 4000 mD (F CD =4 md-ft), whereas the effective permeability is 0.08 mD in 50-ft grid-block model to preserve the fracture conductivity. Similarly, the hydraulic fracture permeability is 83300 mD (Hydraulic fracture conductivity = 83.3 md-ft) and the effective fracture permeability should be 41.65 mD in a 2-ft width pseudo LS-LGR-DK (Rubin 2010; Wan et al. 2013) model. This model uses fine grid LGR near the fracture regions so better representation of pressure and saturation changes could be captured near the fractures. The non-Darcy flow Forchheimer correction factor was used in this model to ensure the coarse fracture model will produce similar results to the actual 0.001-ft width fracture model.

Texas Tech University, Tao Wan, May 2015

Figure 2.8-Simulation model of natural fracture-network (implicit)

10 9 10x50x1 8 4x20x1 7 6 1x5x1 5

Oil R.F. R.F. Oil % 4 3 2 1 0 0 1000 2000 3000 4000 5000 6000 Time, days

Figure 2.9-Effect of numerical dispersion on oil R.F. vs. time

Fig.2.8 shows the base reservoir model using 80 grid blocks of 50 ft x 50 ft to simulate a unit fracture-network. In this study, we examined the impact of grid refinement on oil recovery by performing a series of numerical sensitivity calculations. Figs.2.9 and 2.10 illustrate that using course grid-blocks is producing similar results with more refined

Texas Tech University, Tao Wan, May 2015 grid-blocks. The induced numerical dispersion effect is not that significant. The reservoir initially operates in naturally depletion for 1800 days. Then, we start 20 cycles of secondary recovery. Each cycle consists of 100 days gas/solvent injection and 100 days production. The injection rate decreases rapidly in the huff phase due to low permeability in shale rocks. The minimum bottom-hole pressure of the production well is 2500 psi.

900000 800000 10x50x1 700000 4x20x1 600000 1x5x1 500000 400000 300000 C1 produced , lb , produced C1 200000 100000 0 0 1000 2000 3000 4000 5000 6000 Time, days

Figure 2.10-Effect of numerical dispersion on C1 prod. vs. time

Texas Tech University, Tao Wan, May 2015

10 9 Black oil model 8 7 Composition model 6 5 4

Oil Recovery, % Oil 3 2 1 0 0 1000 2000 3000 4000 5000 6000 Time, days

Figure 2.11-Comparison of oil recovery for four component model vs. fully compositional model

Fig.2.11 compares the oil recovery efficiency factor of the four-component model with a fully compositional model for the scenario that the average reservoir pressure was maintained well above the original saturation pressure. Although the black-oil model gives close results with the composition model, the black-oil model lack the ability to consider the mass transfer between injected solvent and the reservoir oil. Vaporizing/condensing gas drive process controls the development of miscibility in gas displacement in which gas extracts intermediate molecular weight hydrocarbon components from oil and oil may take up components from the gas phase. For immiscible condition, the black-oil model gives lower oil recovery compared to the compositional model because the black-oil model can’t carry condensable liquids in the gaseous phase (Killough et al. 1987). The oil recovery given by the base model is low because of the sparse fracture spacing (200-ft) and no presence of hydraulic fracture.

Texas Tech University, Tao Wan, May 2015

60 100-ft fracture spacing

50 200-ft fracture spacing

40 50-ft fracture spacing

Oil Recovery, % Oil 20

0 0 1000 2000 3000 4000 5000 6000 Time, days

Figure 2.12-Comparison of oil R.F. for different fracture spacing

Hydraulic fractures + natural fracture-network 20 Fracture-network

10 Oil Recovery, % Oil

0 0 1000 2000 3000 4000 5000 6000 Time, days

Figure 2.13-Impact of hydraulic fracture on oil R.F.

Texas Tech University, Tao Wan, May 2015

Fig.2.12 shows the impact of natural fracture spacing on the ultimate oil recovery factor by lean gas (77% C 1, 20% C 3 and 3% C 6) huff-and-puff process. As shown in the graph, reducing fracture spacing from 200-ft to 50-ft would result in an almost six fold increase in cumulative oil recovery factor. The results show that natural fracture spacing has a significant impact on enhanced oil recovery. The fracture-network complexity is critical to well productivity in shale reservoirs because they maximize fracture-surface contact area with the shale through both size and fracture density (Mayerhofer et al.2010). The well production performance in shale oil reservoirs is strongly dependent upon the fracture complexity. Fig.2.13 compares the shale oil recovery from a case that hydraulic fractures connect to the natural fracture-network and a case that has natural fracture- network exclusively. The role of hydraulic fractures in well productivity in shale reservoirs is observed. The introduction of hydraulic fractures on the base model of natural fracture-network gives much higher oil recovery than the base model (natural fracture-network). The hydraulic fracture conductivity is set as 83.3 mD-ft. Cipolla (2009) showed that the presence of highly connected fracture-network can reduce the requirement for fracture conductivity in shale gas reservoirs. If the fracture density exceeds beyond a point, the addition of hydraulic fractures would not significantly improve oil recovery.

Lean gas, rich gas and CO 2 injection performance

Texas Tech University, Tao Wan, May 2015

99 94 89 84 Solvent 79 Solvent + 50% C1 C1 recovery at 1.2 recovery PV at 74 6 C Solvent + 20% C3 69 64 0 0.05 0.1 0.15 0.2 Square root of grid-block size (Dimensionless grid block size)

Figure 2.14-Extrapolation of recovery to zero grid-block size (slim-tube model)

Fig.2.14 shows effects of numerical dispersion on the component of C 6 recovery after 1.2 pore volumes injection (PVI) slim-tube test. It also shows the recovery as a function of

the dilution or enrichment of incipient solvent with C 1 or C 3. The slim-tube displacements were performed at 320 ˚F and 2500 psig with high matrix permeabilities. The solvent or

diluted solvent with C 1 will become immiscible flooding at this designated pressure, which is below the required minimum miscibility pressure (3400 psi) for incipient solvent to be miscible with reservoir fluid. An increase of C 1 content in the solvent makes it more difficult to develop miscibility with oil. Solvent was injected at a rate of 10% pore volumes per day to ensure a low pressure drawdown and low dispersion. The use of square root of dimensionless grid block size method will give approximately linear extrapolation to the correct answer. The dimensionless grid block size is the actual grid block size divided by the slim-tube length (Stalkup, 1987). For example, use of 100 grid- blocks with an equal block size of 0.1-ft will result in a square root of dimensionless

block size at 0.1. Fig.2.14 shows that as the incipient solvent is diluted by C 1, the extrapolated recovery of injection gases falls off. However, the addition of 20% C 3 in the incipient solvent gives an effective miscible displacement. It suggests that enrichment of injected solvent by C 3 or heavier components contributes to achieve miscibility with

Texas Tech University, Tao Wan, May 2015

reservoir oil. The MMP required for the enriched solvent with 20% C 3 to form miscibility with reservoir oil is reduced to 2210 psi. Zick (1986) proposed a vaporizing/condensing drive mechanism to explain this. The light intermediates of the injection gas condense into oil when gas first contacts with oil, while the light gas will pick up small amount of middle intermediates from oil during the gas transportation. The intermediates that were originally present in the gas combined with those stripped from oil will make the gas become richer that is able to condense with fresh oil downstream. When injection gas enrichment exceeds a critical value, displacement behavior of the reservoir fluid becomes a miscible flooding. It important to notice immiscible displacements are not as efficient as miscible displacements, but may still recover oil and the gas utilization factor is higher for the immiscible flooding.

Figure 2.15-Gas production rate (RC) and hydrocarbon pore volume injection

Texas Tech University, Tao Wan, May 2015

Figure 2.16-Oil production rate and average reservoir pressure (unit fracture)

50-ft fracture-network, BHP=2500

C1 20 CO2 Solvent + 20% C3 Solvent 15

10 C20 R.F R.F % , C20

0 0 1000 2000 3000 4000 5000 6000 7000 Time, days

Figure 2.17-C20 recovery vs. time at 0.2 PVI

Texas Tech University, Tao Wan, May 2015

Fig.2.15 displays the gas injection rate and the total injected hydrocarbon pore volume. The injector maintained a injection rate at 240 rft 3/day at reservoir condition. In order to compare the performance of miscible and immiscible flooding, it is critical to keep injected pore volume the same for each case. For the same injection pressure, different gas components have different injection rates. The inflow performance relationship in terms of the volumetric production or injection rate of each phase is expressed as Eq. 2.6 and 2.7.

-+ ,% = − − (2.6) +,% 2% ͟ % 2 2% ͥ ͎ +,% Ƴ͊ ͊ ͂ Ʒ

( ( $,% = $,% *,% *% + $,% ",% ",% (2.7) ͥ ͬ ͖ ͥ ͭ ͖ ͥ

Fig.2.16 shows daily oil production rate and average reservoir pressure changes during cyclic gas injection process. In the first few cycles, the average reservoir pressure slowly builds up and maintains about 4,000 psi which is sufficient for the solvent to achive miscible displacement with reservoir oil. The oil production rate responses agree with the average reservoir pressure. After producing several number of cycles, the oil recovery efficiencies start to decrease. It should be noticed that all the production rate or injection rate in this study are referred as unit fracture value. Fig.2.17 shows the C 6 component recovery after 20 cycles of gas injection in a 50-ft spacing of fracture-network at 0.01 PV

injection for each cycle. CO 2 and solvent enriched with C 3 give miscible displacement.

They perform better than incipient solvent or C 1. The MMP required for methane to form miscibility with reservoir oil is too high that can’t happen at current injection case. Even so, the improved oil recovery due to immiscible methane injection is remarkable.

CO 2 Huff-n-Puff

Carbon dioxide is believed to be a favorable injection fluid for implementing EOR projects because CO 2 has the advantage of extracting or vaporizing some hydrocarbon

Texas Tech University, Tao Wan, May 2015

components when it comes in contact with reservoir oils. CO 2 recovery mechanisms include CO 2 dissolution in the oil leads to oil swelling, oil viscosity reduction, vaporization of intermediate to heavy hydrocarbons and development of multi-contact miscibility (Whitson et al).

Figure 2.18-Impact of natural fracture spacing and hydraulic fracture on CO 2 huff-puff oil recovery, BHP of producer = 2500 psi

Texas Tech University, Tao Wan, May 2015

90 7000 BHP=500 80 BHP=2500 6000 70 Average pressure for 5000 60 BHP=500 50 4000

40 3000

Oil recovery, Oil % 30 2000 20

1000 pressure,psia Averagereservoir 10

0 0 0 1000 2000 3000 4000 5000 6000 7000 Time, days

Figure 2.19-BHP effect on ultimate oil recovery (Fracture spacing = 50-ft)

The results shown in Fig.2.18 indicate that the natural fracture density (spacing) is a dominant factor that dictates well production performance in shale oil reservoirs by secondary cyclic CO 2 injection. It shows the effect of an increase of fracture permeabilities by induced hydraulic fractures is not as significant as the increase of fracture-network densities on shale oil recovery efficiency. In other words, the conductivity requirement is not as critical as fracture spacing to cyclic gas injection enhanced oil recovery process. Fig.2.18 shows that when the natural fracture-network spacing reduced from 200-ft to 50-ft, it results in an increase of 50% oil production. However, it is important to notice that the model assumes that the conductivity of natural fracture is constant (4 md-ft). In field cases, without hydraulic fracturing stimulation, the conductivity of natural fracture is not likely to reach 4 md-ft. More importantly, natural fracture-network properties (conductivity, density, size) are highly dependent on hydraulic fracture treatment. For instance, high injection pressure in hydraulic fracture treatment may cause slip on natural fractures resulting in an increase of the conductivity of fracture-network. The presence of highly conductive hydraulic fracture also provides 29

Texas Tech University, Tao Wan, May 2015 more contact area with injected solvent which is considered as a major influence on recovery efficiencies. The relationship between natural fracture network properties and hydraulic fracturing has not been investigated in this work by including a geomechanical model. The comparisons of results for the various flowing bottom-hole pressure of producer are presented in Fig.2.19 . The reservoir pressure in each gridblock is controlled above the minimum miscibility pressure with 2500 psi of bottom-hole-flowing pressure. When the BHP of the producer is 500 psi, the reservoir pressure in most blocks will drop below the MMP during the production phases. Lowering the bottom-hole pressure has the advantage of improving inflow performance by gaining higher pressure drawdown, thus recovering more oil.

Texas Tech University, Tao Wan, May 2015

Figure 2.20-Reservoir pressure, CO 2 mole fraction in oil phase and oil viscosity variations during CO 2 huff-and-puff

Fig.2.20 shows the reservoir pressure, CO 2 mole fraction in oil phase and oil viscosity variations during CO 2 huff-n-puff process at time 0, after 1800 days of primary recovery, after the first 100 days of injection, after the first 100 days of production and after 20 cycles of secondary recovery. As cyclic CO 2 injection continues in 20 cycles, there is increasing amount of CO 2 retained in the oil phase as shown in Fig.2.19. This loss of CO2 in the oil production cycle is actually a form of geological storage as CO 2 will be contained within the reservoir (Whittaker et al. 2013). With an increase the amount of

CO 2 dissolved in oil, it results in an oil viscosity reduction from initial 0.295 cp to 0.08

cp that allows the oil to flow more easily toward the production well. CO 2 cycling process can be repeated several times, but the oil recovery efficiency decreases (Fig.2.21). It is important to determine an optimized injection length in each cycle to make the displacements most efficient. Long injection cycle will result in a loss of production time during the injection period. Oil will be pushed far away from the fractures by long injection time that makes it more difficult to produce back, which is not a good strategy.

Texas Tech University, Tao Wan, May 2015

Figure 2.21-Period oil production (unit fracture production, entire horizontal well production should be 20 times) 2.5. Summary

The objective of this chapter is to evaluate the response of cyclic gas injection as an enhanced-oil-recovery method in intensely naturally fractured and hydraulically fractured reservoirs using a compositional CMG model. The compositional model was compared with generated black-oil model. They give similar results in the case of miscible displacements. Detailed reservoir black-oil and gas condensate PVT characterization analysis was performed to illustrate the lean gas injection and CO 2 injection mechanisms. A series of slim-tube simulations were performed to study the interaction of injected solvent with reservoir oil. Miscible and immiscible gas flooding performances were compared at the same injection pore volumes for each case. When injection solvent enrichment exceeds a certain value, miscible displacement can happen at a lower pressure.

Enriched solvent or CO 2 performed better than pure solvent or C1 because it is difficult for C1 to develop miscible displacement with reservoir oil at a low pressure.

Texas Tech University, Tao Wan, May 2015

The impact of fracture densities of fracture network on oil recovery were investigated in this chapter, the simulation results indicate that smaller fracture spacing is crucial to improving oil recovery in shale oil reservoirs. The primary decision in designing fracture treatment in shale oil reservoirs is to exploit fracture complexity. The role of hydraulic fractures should be honored because hydraulic fracture treatment induces the fracture complexity to shale reservoirs. CO 2 cycling process can be repeated several times, but the oil recovery efficiency decreases with the increase of cycles. Next chapter will focus on investigating the effect of stress-dependent natural-fracture network permeability on cyclic gas injection performance in shale oil reservoirs.

Texas Tech University, Tao Wan, May 2015

References

Alves, F.B., Mostafa, I. 1990. Compositional Simulation Study of Miscible Gas Injection Pilot Project. Paper 21344 presented at the Middle East Oil Show, 16-19 November 1991, Bahrain. Doi: 10.2118/21344-MS.

Artun, Emre, et al. 2011. Performance and Economic Evaluation of Cyclic-Pressure Pulsing in Naturally Fractured Reservoirs. JCPT . 50 (9): 24-36. Doi: 10.2118/129599-PA.

Bardon, C.P., Karaoguz, D., and Tholance, M. 1986. Well Stimulation by CO2 in the Heavy Oil Field of Camurlu in Turkey. Paper SPE 14943 presented at the SPE Enhanced Oil Recovery Symposium, Tulsa, 20–23 April. Doi:10.2118/14943-MS.

Bazan, L.W., Larkin, S.D., Lattibeaudiere, M.G., Palisch, T.T. 2010. Improving Production in the Eagle Ford Shale with Fracture Modeling, Increased Fracture Conductivity, and Optimized Stage and Cluster Spacing Along the Horizontal Wellbore. SPE 138425 presented at Tight Gas Completions Conference, San Antonio, Texas, USA, 2-3 November 2010. Doi: 10.2118/138425-MS.

Chen, C., Balhoff, B., Mohanty, K. K., 2013. Effect of Reservoir Heterogeneity on Improved Shale Oil Recovery by CO2 Huff-n-Puff, SPE 164553 presented at 2013 SPE Unconventional Resources Conference, Woodlands, Texas, USA,10-12 Apr. Doi: 10.2118/164553-MS.

Chaudhary, A.S., Ehlig-Economides, C., Wattenbarger, R. 2011. Shale Oil Production Performance from a Stimulated Reservoir Volume. Paper SPE 147596 presented at Annual Technical Conference and Exhibition, Denver, Colorado, USA, 30 October-2 November . Doi: 10.2118/147596-MS.

Cipolla, C.L., Warpinski, N.R., Mayerhofer, M.J., Lolon, E.P., and Vincent, M.C. 2008a. The Relationship Between Fracture Complexity, Reservoir Properties, and Fracture

Texas Tech University, Tao Wan, May 2015

Treatment Design. Paper SPE 115769 presented at the SPE Annual Technical Conference and Exhibition, Denver, 21–24 September. Doi: 10.2118/115769-MS.

Cipolla, C.L. et al. 2009. The Effect of Proppant Distribution and Un-Propped Fracture Conductivity on Well Performance in Unconventional Gas Reservoirs. Paper 119368 presented at SPE Hydraulic Fracturing Technology Conference, 19-21 January 2009, The Woodlands, Texas. Doi; 10.2118/119368-MS.

CMG Manual. 2009. Modelling Non Darcy Flow in Hydraulic Fractures Accurately Using a Grid Based Approach. IMEX, Advanced Oil/Gas Reservoir Simulator Version, PP. 167-190 .

Coats, H. 1982. Simulation of Gas Condensate Reservoir Performance. SPE paper 10512 presented at the Sixth SPE Symposium on Reservoir Simulation, New Orleans, January 31st - February 3rd.

Dake, L.P. 1978. Fundamentals of Reservoir Engineering. Elesvier Scientific Publishing Company.

Dershowitz, W., W. Doe, T. 2011. Modeling Complexities Of Natural Fracturing Key In Gas Shales. The American Oil & Gas Report. Aug.2011.

Fayers, F.J., Blunt, M.J., Christie, M.A. 1992. Comparisons of Empirical Viscous- Fingering Models and Their Calibration for Heterogeneous Problems. SPEJ . 7 (2):195-203. Doi: 10.2118/22184-PA.

Gamadi, T.D., Sheng, J.J., Soliman, M.Y. 2013. An Experimental Study of Cyclic Gas Injection to Improve Shale Oil Recovery. Paper 166334 presented at the SPE Annual Technical Conference and Exhibition, 30 September- 2 October 2013, New Orleans, Louisiana, USA. Doi: 10.2118/166334-MS.

Gilman, J. R. and Kazemi, H. 1983. Improvements in Simulation of Naturally Fractured Reservoirs. SPEJ. 23 (4):695-707. 35

Texas Tech University, Tao Wan, May 2015

Gondiken, S. 1987. Camurlu Field Immiscible CO2 Huff and Puff Pilot Project. Paper SPE 15749 presented at the Middle East Oil Show, Bahrain, 7–10 March. Doi: 10.2118/15749-MS.

Haines, H. K., Monger, T. G. 1990. A Laboratory Study Of Natural Gas Huff 'N' Puff. Paper presented at the Annual Technical Meeting, Jun 10 - 13, 1990, Calgary, Alberta. Doi: 10.2118/90-78.

Heidemann, R.A., and Khalil, A.M. 1980.The Calculation of Critical Points. AIChE J . 26 (5) :769-779.

Hong, K.C. 1982. Lumped-Component Characterization of Crude Oils for Compositional Simulation. Paper 10691 presented at SPE Enhanced Oil Recovery Symposium, 4-7 April 1982, Tulsa, Oklahoma. Doi: 10.2118/10691-MS.

Hoteit, H., Firoozabadi, A. 2006. Numerical Modeling of Diffusion in Fractured Media for Gas Injection and Recycling Schemes. Paper presented at SPE Annual Technical Conference and Exhibition, 24-27 September 2006, San Antonio, Texas, USA. Doi: 10.2118/103292-MS.

Hsu, S.-C., and Nelson, P.P. 2002. Characterization of Eagle Ford Shale. Engineering Geology,Volume 67, PP. 169-183.

Ivory, John, Chang, J., Coates, R., Forshner, K. 2010. Investigation of Cyclic Solvent Injection Process for Heavy Oil Recovery. Journal of Canadian Petroleum Technology . 49 (9):22-33. Doi: 10.2118/140662-PA.

Jessen, K., Michelsen, M., Stenby, E. H. 1998. Global approach for calculation of minimum miscibility pressure. Fluid Phase Equilibria. 153 (2):251-263.

Johns, R.T., Dindoruk, B., Orr, F.M. 1993. Analytical Theory of Combined Condensing/Vaporizing Gas Drives. SPEJ . 1 (2): 7-16. Doi: 10.2118/24112-PA.

Texas Tech University, Tao Wan, May 2015

Killough, J.E., Kossack, C.A. 1987. Fifth Comparative Solution Project: Evaluation of Miscible Flood Simulators. Paper SPE 16000 presented at SPE Symposium on Reservoir Simulation, 1-4 February 1987, San Antonio, Texas. Doi: 10.2118/16000- MS.

Koval, E.J. 1963. A Method for Predicting the Performance of Unstable Miscible Displacement in Heterogeneous Media. SPEJ . 3 (2):145 – 154. Doi: 10.2118/450- PA.

Lino, U. de R.A. 1994. An Evaluation of Natural Gas Huff ‘n’ Puff Field Tests in Brazil. Paper 26974 presented at the SPE Latin America/Caribbean Petroleum Engineering Conference, 27-29 April 1994, Buenos Aires, Argentina. Doi: 10.2118/26974-MS.

Mayerhofer, M.J., Lolon, E.P., Warpinski, N.R., Cipolla, C.L., Walser, D., Rightmire, C.M., et al. 2010. What Is Stimulated Reservoir Volume? SPEJ . 25 (1):89-98. Doi: 10.2118/119890-PA.

Miller, Bernard J., Bretagne G.P.1988. Field Case: Cyclic Gas Recovery for Light Oil- Using Carbon Dioxide/Nitrogen/Natural Gas. Paper 49169 presented at the SPE Annual Technical Conference and Exhibition, 27-30 September 1998, New Orleans, Louisiana. Doi: 10.2118/49169-MS.

Orangi, A et al. 2011. Unconventional Shale Oil and Gas-Condensate Reservoir Production, Impact of Rock, Fluid, and Hydraulic Fractures. Paper 140536 presented at SPE Hydraulic Fracturing Technology Conference, 24-26 January 2011, The Woodlands, Texas, USA. Doi:10.2118/140536-MS.

Pedersen, K. S., Christensen, P. L., & Azeem, S. J. 2006. Phase behavior of petroleum reservoir fluids. CRC Press.

Rubin, B. 2010. Accurate Simulation of Non Darcy Flow in Stimulated Fractured Shale Reservoirs. Paper SPE 132093 presented at Western Regional Meeting, Anaheim,

Texas Tech University, Tao Wan, May 2015

California, USA, 27-29 May. Doi: 10.2118/132093-MS.

Stalkup, F.I. 1987. Displacement Behavior of the Condensing/Vaporizing Gas Drive Process. SPE 16715 presented at SPE Annual Technical Conference and Exhibition, 27-30 September, Dallas, Texas. -MS. Doi: 10.2118/16715-MS.

Todd, M.R., Longstaff, W.J. 1972, The Development, Testing, and Application Of a Numerical Simulator for Predicting Miscible Flood Performance. Journal of Petroleum Technology . 24 (7):874-882.

Vega, B., O’Brien, W.J., Kovscek, A.R. 2010. Experimental Investigation of Oil Recovery from Siliceous Shale by Miscible CO2 Injection. Paper SPE 135627 presented at the SPE Annual Technical Conference and Exhibition held in Florence, Italy, 19–22 September. Doi:10.2118/135627-MS.

Wan, T., Sheng, J., Soliman, M.Y. 2013a. Evaluation of the EOR Potential in Shale Oil Reservoirs by Cyclic Gas Injection. Paper SPWLA-D-12-00119 presented at the SPWLA 54th Annual Logging Symposium held in New Orleans, Louisiana, June 22- 26, 2013.

Wan, T., Sheng, J.J., Soliman, M.Y. 2013b. Evaluate EOR Potential in Fractured Shale Oil Reservoirs by Cyclic Gas Injection. Paper 168880 presented at the Unconventional Resources Technology Conference held in Denver, Colorado, USA, 12-14 August.

Wang, Y. and Orr, F.M., Jr. Calculation of Minimum Miscibility Pressure. Paper SPE 39683 presented at SPE/DOE Improved Oil Recovery Symposium, 19-22 April 1998, Tulsa, Oklahoma. Doi: 10.2118/39683-MS.

Whitson, C.H., and Torp, S.B. 1983. Evaluating Constant-Volume Depletion Data. SPEJ . 35 (03):610-620. Doi: SPE-10067-PA.

Whittaker, S., Perkins, E. 2013. Technical Aspects of CO2 Enhanced Oil Recovery and 38

Texas Tech University, Tao Wan, May 2015

Associated Carbon Storage. Global CCS institute, Nov, 2013.

Yuan, H., Johns, R.T. 2005. Simplified Method for Calculation of Minimum Miscibility Pressure or Enrichment. SPEJ . 10 (4):416-425. Doi: 10.2118/77381-PA.

Yuan, H., Johns, R.T., Egwuenu, A.M. 2005. Improved MMP Correlations for CO2 Floods Using Analytical Gasflooding Theory. SPEJ . 8 (5):418-425. Doi: 10.2118/89359-PA.

Zick, A.A. 1986. A Combined Condensing/Vaporizing Mechanism in the Displacement of Oil by Enriched Gases. Paper 15493 presented at SPE Annual Technical Conference and Exhibition, 5-8 October, New Orleans, Louisiana. Doi: 10.2118/15493-MS.

Texas Tech University, Tao Wan, May 2015

Chapter 3 Evaluation of the EOR Potential in Fractured Shale Oil Reservoirs by Cyclic Gas Injection

3.1. Abstract

The current technique to produce shale oil is to use horizontal wells with multi-stage stimulation. However, the primary oil recovery factor is only a few percent. The low recovery and the abundance of shale reservoirs provide huge potential for enhanced oil recovery.

Well productivity in shale oil and gas reservoirs primarily depends upon the size of fracture network and the stimulated reservoir volume (SRV) which provides highly conductive conduits to communicate the matrix with the wellbore. The natural fracture complexity is critical to the well production performance and it also provides an avenue for injected fluids to displace the oil. However, the disadvantage of gas flooding in fractured reservoirs is that injected fluids may break through to production wells via the fracture network. Therefore, a preferred method is to use cyclic gas injection to overcome this problem.

In this chapter, we used a numerical simulation approach to evaluate the EOR potential in fractured shale oil reservoirs by cyclic gas injection. Simulation results indicate that the stimulated fracture network contributes significantly to the well productivity via its large contact volume with the matrix, which prominently enhances the macroscopic sweep efficiency in secondary cyclic gas injection process. In our previous simulation work, the EOR potential was evaluated in hydraulic planar traverse fractures without considering the propagation of natural fracture-network. In this chapter, we examined the effect of fracture networks on shale oil secondary production performance. The impacts of fracture spacing and stress-dependent fracture conductivity on the ultimate oil recovery are investigated. The results presented in this chapter demonstrate an EOR potential by cyclic 40

Texas Tech University, Tao Wan, May 2015 gas injection in shale oil reservoirs. The objective of this chapter focus on evaluating the effects of fracture spacing, the size of fracture-network, fracture connectivity (uniform and non-uniform) and stress-dependent fracture-network conductivity on production performance of shale oil reservoirs by secondary cyclic gas injection.

3.2. Introduction

This chapter builds upon and extends our earlier work that investigated the impact of the planar fractures on the enhanced oil recovery by cyclic gas injection (Wan et al. 2013). We investigated the impact of different well operating schedules (injection time and production time in each cycle), degree of in-situ miscibility between solvent with oil and propped planar hydraulic fractures spacing on the ultimate oil recovery. The previous work mainly focused on examining the effects of planar hydraulic fractures on the well primary production and secondary gas injection performance without taking into account of the contribution from fracture-network.

In naturally fractured shale formations, the fracture network permeability is sensitive to changes in stress, especially by secondary gas injection process (Palmer and Mansoori, 1998). There are some difficulties in simulation of fluid flow in unconventional reservoirs because a large permeability contrast exists between the hydraulic fractures and their neighboring tight shale matrix (Moinfar et al., 2013). Rubin (2010) used a logarithmically spaced, finely-gridded local grid refinement (LGR) method based on dual permeability model to model gas flow from unconventional shale gas reservoirs inside the SRV region. Unrefined dual permeability grids were used to simulate the fluid flow outside the SRV because of low pressure drop. The LS-LR-DK (logarithmically spaced, locally refined and dual permeability) model is able to accurately simulate the gas flow in fractured shale gas reservoirs. Rubin also compared the results with the actual 0.001-ft width fracture reference model and they matched very well. The fracture-network characterization in shale reservoirs is a challenging task because the location of proppant transported by fracturing fluids and fracture conductivity are difficult to be determined. The initial

Texas Tech University, Tao Wan, May 2015 microseismic-mapping work in the Barnett shale (Fisher et al. 2002; Fisher et al. 2004) has shown that the fracture propagation can be highly unpredictable and complex, ranging from simple planar fractures to very complex fracture systems. In horizontal well multiple-cluster completions, the propagation of subsequent fractures is affected by the reorientation of stress field from the previous propagated fractures. The occurring of stress shadow effect in multiple clusters of fracturing completion makes the simulation of hydraulic fracturing in naturally fractured shale plays more complex. The drawback of using a single-porosity approach to simulate intensely fractured shale oil reservoirs is that it would require tremendous number of fine grid-block when fracture-network spacing is narrow. The advantage of LS-LR-DK model over a single-porosity model appears in the modeling of low permeability shale reservoirs because it can produce similar results as the finely gridded single-porosity model with much less LGR grids as showed by Rubin (2010). The LS-LR-DK model can be implemented with significantly less number of grids in comparison with an explicit representation of both matrix and fracture regions.

Chen et al. (2013) investigated the effect of reservoir heterogeneity on improved shale oil recovery by CO 2 huff and puff. In their simulation model, they used the log-normally distributed permeability field and Dykstra-Parsons function to represent the permeability heterogeneity in shale reservoirs. However, this permeability heterogeneity representation cannot physically simulate the fracture distributions, fracture network complexity and fluids flow in naturally fractured reservoirs. Cipolla et al. (2008) examined the effect of fracture characteristics on gas well performance. In their fracture model, two types of fractures were assumed based on proppant distribution: planar fracture and fracture- network in which proppant was evenly distributed in.

Their reservoir simulation results indicated that with an increase of fracture complexity it requires lower conductivity to achieve the same production compared with production from simple fracture system. Cipolla et al. (2011) and Moinfar et al. (2013) showed that the well primary productivity was significantly affected by stress-dependent network- fracture conductivity in shale gas reservoirs. Tremendous work presented approaches for 42

Texas Tech University, Tao Wan, May 2015 reservoir modeling the primary production in naturally fractured shale gas reservoirs (Mayerhofer et al. 2006; Cipolla et al. 2010). But limited work was dedicated to addressing the impact of fracture network characteristics on improved oil recovery process in shale oil reservoirs. The objective of this chapter is to investigate the applicability of cyclic gas injection in very low permeability shale oil reservoirs with the LS-LR-DK approach and examine the characteristics of fracture network on enhanced oil recovery process.

In this chapter, the simulation results showed that in the uniform conductivity of unpropped fracture-network, production from shale oil reservoirs is affected by the reduction of natural fracture conductivity due to stress effect, resulting in a lower recovery. However, the presence of highly conductive propped hydraulic fractures that connect to the natural fractures can compensate the recovery loss due to reduction of the network-fracture conductivity by stress effect.

3.3. Model Setup and Validation

The LS-LR-DK method is used to simulate the fluid flow in fractured shale oil reservoirs. In this chapter, we use finely gridded dual permeability LGR cells to model the fracture- network and hydraulic fractures inside the SRV region. We simply use coarse dual permeability grids to model the non-SRV region because the pressure drop in non-SRV is low. The use of coarse dual permeability grids is enough to capture the low pressure drop resolution in the shale non-SRV. The structure of this chapter is first to examine the effect of uniform conductivity of fracture network inside the SRV on improving shale oil recovery by cyclic gas injection scheme. We then consider the effect of non-uniform conductivity of fracture-network by including hydraulic fractures that connect the natural fracture complexity.

The reservoir rock and fluid properties used in this model are based on the published data in Eagle Ford shale as previously used (Bazan, 2010). The initial reservoir pressure for this field is 6,425 psi. The producer is subject to minimum bottom-hole pressure 43

Texas Tech University, Tao Wan, May 2015 constraint (BHP) of 2500 psi and opened production for 1800 days (5 years) as primary production. The bubble point pressure of the reservoir is 2398 psi. The reservoir rock properties are referred to Table 3.1 (Wan et al. 2013). The reservoir fluid properties are presented in Table 3.1. The injection gas is comprised of the produced gas from the reservoir and the gas produced from previous gas lift injection. The dimension of fracture-network is 2,000 ft×1,000 ft. We assume the fracture network is orthogonal to each other and centered in the grid-blocks. The simulation model assumes that the SRV consists of a uniform network-fracture with a conductivity of 4 mD-ft (this value is reasonable according to Cipolla et al. 2008). The fracture network is assumed to be contained within an orthogonal system of continuous, uniformly spaced and constant width. The distance between network-fracture in both the X and Y directions is 200 ft in the base model. Later on we will examine the fracture spacing effect on the enhanced oil recovery performance. The local grid refinement is applied to discretize each large 200 ft×200 ft block in the SRV region into 9×9 logarithmically spaced smaller cells, as shown in Fig.3.1. In each of these cells, two perpendicularly crossed fractures are assumed to exist. We performed grid sensitivity study using finer grids with better resolution. It is found that further grid refinement does not affect the numerical simulation results. We use a computationally efficient 2-foot width pseudoization fracture model for representing the fluids flow in the fracture (Better discussion referred to Rubin, 2010). The shale matrix permeability is 0.0001 mD. The 2-foot width fracture pseudoization model was already tested to be able to accurately simulate the fluid flow in fractured shale gas reservoirs by including the non-Darcy flow correction (CMG Manual, 2009). The 2-foot width fracture model is validated against the reference model that represents fractures by narrow 0.001-ft width cells and they both produce similar results. The 2-foot width fracture pseudoization model has to preserve the same fracture conductivity as the standard fracture model. The fracture permeability of the 0.001-ft width fracture model is

4000 mD (F CD =0.001×4000=4 mD-ft). Thus, the permeability of the 2-ft pseudoization fracture model is 2 mD (F CD =2×2=4 mD-ft).

Texas Tech University, Tao Wan, May 2015

Table 3.1. Eagle Ford Fluid properties

3 3 Pressure (psi) Rs (ft /bbl) Bo (bbl/STB) Eg (ft /bbl) Oil viscosity (cp) Gas viscosity (cp) 14.69 4.68 1.09917 4.10159 0.9026 0.013601 332.47 65.28 1.12711 95.3676 0.7194 0.013905 650.24 140.36 1.16295 191.364 0.5973 0.014385 968.02 223.32 1.20393 291.506 0.5154 0.015001 1285.79 311.98 1.24913 394.75 0.4567 0.015745 1603.57 405.21 1.29803 499.604 0.4126 0.016612 1921.34 502.26 1.3503 604.264 0.3779 0.01759 2239.11 602.62 1.40566 706.874 0.3498 0.018664 3218.4 929.14 1.59372 995.379 0.2889 0.022371 4859.2 1521.47 1.95964 1360.49 0.2299 0.028854 6500.0 2193.14 2.37939 1609.67 0.1946 0.034795

3.4. Simulation Results and Discussion

We only simulated one branch of fractures instead of modeling the entire stimulated reservoir volume considering the flow symmetry, as illustrated in Fig.3.1. The implicit assumption is that the fracture-network has the identical properties such as the same the same geometric shape, spacing, aperture and the same conductivity. Although these parameters may vary substantially in real reservoirs, the more effective simulation of complex fracture networks or fracture connectivity by using discrete fracture-network models (DFN) is out of the scope of chapter. We specified the maximum surface solvent rate of 80 Mscf/day for injector in a unit (200-ft×1000-ft×200-ft) SRV region and the maximum allowable surface injection pressure is 6000 psi. For primary production stage, the producer opens for producing 1800 days. Then, the well operation mode changes into huff-n-puff for secondary recovery. In the huff-n-puff process, each cycle consists of 100 days of solvent injection and 100 days of producing. We implement 60 cycles of cyclic gas injection after 1800 days of primary production. Fig.3.1 and Fig.3.2 showed the pressure distribution after the 60 cycles of gas huff and puff for the entire stimulated fracture-network with 200-ft spacing and 100-ft spacing, respectively.

Texas Tech University, Tao Wan, May 2015

Figure 3.1-Stimulated fracture-network in SRV (Dx=200)

Figure 3.2-Stimulated fracture-network in SRV (Dx=100)

Texas Tech University, Tao Wan, May 2015

80 Entire SRV, DX=100-ft 70 Entire SRV, DX=200-ft Simulation unit, DX=100-ft 60 Simulation unit, DX=200-ft 50

20 OOIPrecovery factor, %

0 0 2000 4000 6000 8000 10000 12000 14000 Time (days)

Figure 3.3-Impact of stimulated fracture-network spacing on EOR

The uniform fracture conductivity (4 mD-ft) of stimulated fracture network with 200-ft spacing is displayed in Fig.3.1. Fig.3.3 compares the production profiles for a network- fracture with spacing of 200-ft to 100-ft. After 60 cycles of cyclic gas injection, the oil recovery is increased from 5.5% of primary recovery to 45% of ultimate oil recovery for the 200-ft spacing of fracture-network, while it obtains 70% of ultimate oil recovery for 100-ft spacing of fracture-network. The oil response after 60 cycles of gas injection is encouraging. The injected gas increases the reservoir energy and dissolves in the crude oil to reduce its viscosity. In shale oil reservoirs, increasing the fracture-network density is one avenue to increase the contact volume between injected solvent and shale matrix. The above simulation results indicate that the stimulated network-fracture spacing has a significant impact on the production performance of shale oil reservoirs. The stimulated fracture-network density and complexity are the key components that dictate the well productivity in shale oil reservoirs.

Impact of Fracture-network Conductivity on Cyclic Gas Injection EOR Performance

Texas Tech University, Tao Wan, May 2015

80 4 mD-ft, DX=100-ft 70 2 mD-ft, DX=100-ft

60 4 mD-ft, DX=200-ft 2 mD-ft, DX=200-ft 50

20 OOIPrecovery factor, %

0 0 2000 4000 6000 8000 10000 12000 14000 Time (days)

Figure 3.4-Impact of the conductivity of fracture-network in the SRV on production profiles

Fig.3.4 presents the results of the conductivity of fracture-network in the SRV on ultimate oil production profiles by cyclic gas injection. It is noted that the impact of fracture- network conductivity is not as significant as stimulated fracture spacing. For 100-ft spacing fracture networks, there is roughly 6% more oil production due to an increase of fracture conductivity from 2 md-ft to 4 md-ft, while more than 27% oil production could be achieved with fracture network spacing decrease from 200-ft to 100-ft.

Impact of Stress-Dependent Conductivity of the Uniform Conductivity Fracture-Network on Production Performance

Production performance of shale reservoirs is primarily dictated by the size or conductivity of network of fractures. It is critical to consider the productivity losses in reservoirs due to stress-dependent fracture-network permeability reduction. If the stress- dependent fracture-network conductivity reduction process is irreversible, the conductivity will never be able to recover its initial value with the decreasing of closure pressure ( Fig.3.5). On the other hand, if we assume the stress-dependent fracture-network 48

Texas Tech University, Tao Wan, May 2015 conductivity reduction process is reversible, the fracture conductivity will return to its initial value with the increase of reservoir pressure by gas injection.

Figure 3.5-Irreversible process of fracture conductivity reduction

Cipolla et al. (2010) studied the impact of stress dependent network-fracture conductivity on well performance in shale gas reservoirs. Their simulation results showed that the well productivity in many shale gas reservoirs is impaired by the insufficient fracture conductivity. Raghavan and Chin (2004) presented correlations to evaluate the stress- induced productivity losses in stress-sensitivity reservoirs. However, their focus was on investigating the effect of pressure-dependent rock-matrix permeability, not the fracture- network permeability, on production performance. Cho et al. (2012) used the experimental data from Bakken cores to screen and calibrate the correlations proposed by Raghavan and Chin (2004). They modified those correlations and applied them to analyze the effect of fracture permeability on the productivity of shale gas reservoirs. The modified model prediction obtained a good history match to the performance of field examples. The correlation developed by Cho et al. (2012) is used in this chapter to provide us a clear understanding about stress-dependent fracture-network permeability effect on EOR performance in shale oil reservoirs.

Texas Tech University, Tao Wan, May 2015

80 Without stress effect, DX=100-ft Reversible stress effect, DX=100-ft 70 Irreversible stress effect, DX=100-ft Without stress effect, DX=200-ft 60 Reversible stress effect, DX=200-ft Irreversible stress effect, DX=200-ft 50

20 OOIP recovery OOIP factor, %

0 0 2000 4000 6000 8000 10000 12000 14000 Time (days)

Figure 3.6-Stress-dependent F CD of fracture-network effect on EOR

Fig.3.6 illustrates that the stimulated fracture-network spacing has more significant effect on improved oil recovery than stress-dependent conductivity. For 100-ft spacing uniform fracture-network, the oil recovery loss due to the irreversible reduction of fracture conductivity is roughly 10% compared to the case without stress-dependent effect. However, the secondary shale oil recovery from 100-ft spacing fracture-network is substantially higher than 200-ft scenarios.

Impact of Stress-Dependent Conductivity of the Non-Uniform Conductivity Fracture-Network on Production Performance

Texas Tech University, Tao Wan, May 2015

FCD=83.3

FCD= 83.

FCD=4

Figure 3.7-Permeability distribution for the non-uniform fracture network

Fig.3.7 presents the fracture conductivity distribution for the non-uniform fracture network. The unpropped network fracture (blue lines) conductivity in the SRV remains to be 4 mD-ft. The hydraulic fracture (red lines) conductivity is 83.3 mD-ft. The non- uniform conductivity of fracture-network is represented by the infinite conductivity primary hydraulic fractures connecting with finite conductivity natural fractures.

50 Without stress effect Stress-dependent FCD 40 Irreversible Stress-dependent FCD

Primary 30 Recovery Secondary Recovery 20 Dx=200 Ultimate% R.F., Oil 10 Hydraulic fractures communicate with natural fracture network

0 0 2000 4000 6000 8000 10000 12000 14000 Time, days

Figure 3.8-Effect of stress-dependent FCD on EOR performance

Fig.3.8 considers the effect of stress-dependent conductivity of the non-uniform

Texas Tech University, Tao Wan, May 2015 conductivity fracture-network on well performance in tight shale reservoirs. The simulation results from Fig.3.8 indicate that the impact of stress-dependent fracture conductivity on ultimate oil recovery is inconsequential in the presence of hydraulic fractures communication with the fracture-network. As compared to Fig.3.6, without the presence of highly conductive propped fractures, the results illustrate that the existence of highly conductive hydraulic fractures connecting to the fracture- network complexity can reduce the requirement for fracture conductivity. When the long, narrow and relatively high conductivity closely spaced directional fracture swarms are communicated, hydraulic fracture length requirements are very minimum and highly conductive fractures are not demanded (Gale et al. 2007). In many shale gas reservoirs, the primary gas production is strongly dependent on fracture conductivity. However, it is interesting to note that the well completion design for improving oil recovery in shale oil reservoirs by secondary cyclic gas injection is different from the primary production from shale gas reservoirs. The fracture complexity is more critical than fracture conductivity in the cyclic gas injection enhanced-oil-recovery process. When a certain level of fracture conductivity has been reached, increasing the fracture conductivity has a diminishing effect on improved oil recovery. The simulation results from Fig. 3.8 indicate that the conductivity of fracture network is not the dominating factor anymore if the fracture networks were communicated with the propped hydraulic fractures.

The Effect of Stimulated Fracture-Network Size on Cyclic Gas Injection EOR Performance

Texas Tech University, Tao Wan, May 2015

HF=400

Non-SRV

Figure 3.9-Non-SRV embedded in SRV (Hydraulic fracture spacing = 400-ft)

Fig.3.9 shows a realistic case in the field in which the stimulated rock volume is not always continuous and is separated by the non-SRV region. The hydraulic fracture spacing is dependent upon the number of fracture treatment stages along the lateral. Implementing more fracturing stages results in smaller fracture spacing. Exploiting the fracture complexity in unconventional reservoirs is a main consideration for improving shale oil recovery by secondary cyclic gas injection as discussed. Slick water has been the primary fracturing fluid in treating shale gas reservoirs because low-viscosity water leaks off easily to fracture networks to widen the zone of stimulation away from a single fracture plane. Fig.3.9 presented a case of 400-ft spacing hydraulically fractured shale model in which 200-ft interval is not stimulated between two adjacent fracturing stimulated volumes. The fracture permeability in non-SRV is 0.001-mD that is 10 times higher than the shale matrix. The relationship between hydraulic fracturing spacing and the well performance is illustrated in Fig.3.10 . The results indicate that the hydraulic fracturing spacing is a dominant factor that dictates secondary well performance because the size of induced natural fracture-network closely depends on the spacing of hydraulic fracturing treatment.

Texas Tech University, Tao Wan, May 2015

HF spacing=200-ft 50 HF spacing =400-ft HF spacing=600-ft 40

10 Ultimate shale recoveryoil factor, %

0 0 2000 4000 6000 8000 10000 12000 14000 Time (days)

Figure 3.10-Hydraulic fracture spacing effect on EOR

45 HF + NF Fracture-network only 40 Hydraulical fractures only 35

Ultimate recovery shaleoil factor, % 5

0 0 2000 4000 6000 8000 10000 12000 14000 Time (days)

Figure 3.11-Fracture scenarios effect on EOR (DX=200-ft)

The results in Fig.3.11 show that the ultimate shale oil recovery from planar hydraulic fracture is approximately 30% by gas huff-n-puff process that is consistent with the results of previous work (Wan et al.2013). Fig.3.11 shows that the ultimate oil recovery from a system of hydraulic fractures connected to natural fractures is the highest, which obtains about 20% more oil recovery than production exclusively from propped hydraulic fractures. There is slight difference between production from fracture-network connecting to hydraulic fractures system and exclusive fracture-network system. The priority in optimizing fracture treatments for improving oil recovery in shale oil reservoirs should be maximizing the fracture-network complexity.

Texas Tech University, Tao Wan, May 2015

3.5. Conclusions

Natural fractures are critical to the well productivity in shale oil reservoirs. A distinct feature of exploiting shale oil reservoirs by secondary cyclic gas injection is that the fracture-network spacing is more critical than fracture conductivity. The fracture-network spacing is predominating and plays a more important role than fracture network conductivity in enhancing oil recovery.

If the an infinite conductivity primary hydraulic fractures were connected to a finite conductivity network, the hydraulic fracture or fracture-network conductivity reduction due to stress effect will not result in significant ultimate oil recovery loss by using cyclic gas injection technique.

The priority in designing fracture treatments for enhancing oil recovery in shale oil reservoirs should be maximizing the fracture network complexity rather than spending money by pumping large volumes of proppant to maximize the fracture conductivity. It is recommended that creating dense fracture spacing in shale oil reservoirs will lead to more effective completion designs and bring more productivity to enhanced oil recovery by cyclic gas injection process.

The main well productivity is from the stimulated reservoir volume (SRV). The number of hydraulic fracturing treatment stages is crucial to improving the stimulated rock volume, which provides the main drainage area and the flow conduits for hydrocarbons.

The role of diffusion in a field-scale displacement on enhanced-oil-recovery process by cyclic gas injection will be discussed in next chapter.

3.6. References

Bazan, L.W., Larkin, S.D., Lattibeaudiere, M.G., and Palisch, T.T. 2010. Improving Production in the Eagle Ford Shale with Fracture Modeling, Increased Fracture Conductivity, and Optimized Stage and Cluster Spacing Along the Horizontal 55

Texas Tech University, Tao Wan, May 2015

Wellbore. SPE 138425 presented at Tight Gas Completions Conference, San Antonio, Texas, USA, 2-3 November 2010. Doi: 10.2118/138425-MS.

Chen, C., Balhoff, M., and Mohanty, K. K. 2013. Effect of Reservoir Heterogeneity on Improved Shale Oil Recovery by CO Huff-n-Puff. Paper SPE 164553 presented at the SPE Unconventional Resources Conference - USA, Apr 10 - 12, The Woodlands, TX, USA. Doi: 10.2118/164553-MS.

Cho, Y., Apaydin, O.G., and Ozkan, E. 2012. Pressure-Dependent Natural-Fracture Permeability in Shale and its Effect on Shale-Gas Well Production. Paper SPE 159801 presented at SPE Annual Technical Conference and Exhibition, 8-10 October 2012, San Antonio, Texas, USA. Doi:10.2118/159801-MS.

Cipolla, C.L., Warpinski, N.R., Mayerhofer, M.J., Lolon, E.P., and Vincent, M.C. 2008. The Relationship Between Fracture Complexity, Reservoir Properties, and Fracture Treatment Design. Paper SPE 115769 presented at the SPE Annual Technical Conference and Exhibition, Denver, 21–24 September. Doi: 10.2118/115769-MS.

Cipolla, C.L., Lolon, E.P., Erdle, J.C., and Rubin,B. 2010. Reservoir Modeling in Shale- Gas Reservoirs. SPEREE 13 (04): 638-653. Doi: 10.2118/125530-PA.

Cipolla, C., Weng, X., Mack, M., Ganguly, U., Gu, H., Kreese, O., and Cohen, C. 2011. Integrating micro-seismic mapping and complex fracture modeling to characterize fracture complexity, SPE 140185 presented at the SPE Hydraulic Fracturing Technology Conference, The Woodlands, Texas, 24-26 January. Doi: 10.2118/140185- MS.

CMG Manual. 2009. Modelling Non Darcy Flow in Hydraulic Fractures Accurately Using a Grid Based Approach. IMEX , Advanced Oil/Gas Reservoir Simulator Version, PP. 167-190 .

Fisher, M.K., Wright, C.A., Davidson, B.M., Goodwin, A.K., Fielder, E.O., Buckler, W.S.,

Texas Tech University, Tao Wan, May 2015

and Steinberger, N.P. 2002. Integrating Fracture Mapping Technologies to Optimize Stimulations in the Barnett Shale. Paper SPE 77441 presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, 29 September–2 October. Doi: 10.2118/77441-MS.

Fisher, M.K., Heinze, J.R., Harris, C.D., Davidson, B.M., Wright, C.A., and Dunn, K.P. 2004. Optimizing Horizontal Completion Techniques in the Barnett Shale Using Microseismic Fracture Mapping. Paper SPE 90051 presented at the SPE Annual Technical Conference and Exhibition, Houston, 26–29 September. Doi: 10.2118/90051-MS.

Gale, J.F., Reed, RM, and Holder, J. 2007. Natural fractures in the Barnett Shale and their importance for hydraulic fracture treatments. AAPG Bulletin 91 (4) : 603-622.

Mayerhofer, M.J., Lolon, E.P., Youngblood, J.E., and Heinze, J.R. 2006. Integration of Microseismic-Fracture-Mapping Results With Numerical Fracture Network Production Modeling in the Barnett Shale. Paper SPE 102103 presented at the SPE Annual Technical Conference and Exhibition, 24-27 Septembe, San Antonio, Texas, USA. Doi: 10.2118/102103-MS.

Moinfar, A., Varavei, A., Sepehrnoori, K., and Johns, R.T. 2013. Development of a Coupled Dual Continuum and Discrete Fracture Model for the Simulation of Unconventional Reservoirs. Paper SPE 163647 presented at the SPE Reservoir Simulation Symposium, The Woodlands, TX, USA, Feb 18 – 20. Doi: 10.2118/163647-MS.

Palmer, I., Mansoori, J. 1998. How Permeability Depends on Stress and Pore Pressure in Coalbeds: A New Model. SPE Reservoir Evaluation & Engineering 1(06):539-544. Doi: http://dx.doi.org/10.2118/52607-PA.

Raghavan, R., and Chin, L.Y. 2004. Productivity Changes in Reservoirs With Stress- Dependent Permeability. Paper SPE 88870 presented at the SPE Annual Technical 57

Texas Tech University, Tao Wan, May 2015

Conference and Exhibition, 29 September-2 October, San Antonio, TX. Doi: 10.2118/88870-PA.

Rubin, B. 2010. Accurate Simulation of Non Darcy Flow in Stimulated Fractured Shale Reservoirs. Paper SPE 132093 presented at Western Regional Meeting, Anaheim, California, USA, 27-29, May. Doi: 10.2118/132093-MS.

Wan, T., Sheng, J., Soliman, M.Y. 2013. Evaluation of the EOR Potential in Shale Oil Reservoirs by Cyclic Gas Injection. Paper SPWLA-D-12-00119 presented at the SPWLA 54th Annual Logging Symposium held in New Orleans, Louisiana, June 22- 26, 2013.

Texas Tech University, Tao Wan, May 2015

Chapter 4 Compositional Modeling of the Diffusion Effect on EOR Process in Fractured Shale Oil Reservoirs by Gas Flooding

4.1. Abstract

Gas injection is considered as an effective recovery process that has been widely used in the worldwide. There are limited pilot field projects conducted on EOR process by gas injection in shale oil reservoirs. Although many studies have been conducted on gas injection in tight gas or oil reservoirs, the main recovery mechanism in shale oil reservoirs is not well understood. Diffusion plays an important role in the oil recovery process in fractured shale reservoirs. Most of the current studies on diffusion are performed in such a way that the producing pressure is equal to the initial reservoir pressure or core pressure, thus, the convective displacement is eliminated or minimized. One of the challenges is to evaluate the role of diffusion in field scale displacements in the presence of viscous flow. This chapter discusses the role of diffusion in improving oil recovery in fractured shale oil reservoirs.

Hoteit and Firoozabadi (2009) investigated the diffusion effect on recovery performance in a fractured gas/condensate reservoir. Their simulation results showed that molecular diffusion has a significant effect on gas recovery if the reservoir pressure is below the minimum miscible pressure. Modeling of the diffusion effect on ultimate oil recovery in extensively fractured shale reservoir is crucial to the development of these marginal shale oil or gas projects. Evaluation of the recovery contribution from diffusion will provide important insights into the recovery mechanisms in intensely fractured shale gas/oil reservoirs. Currently, a majority of the diffusion models were developed on the basis of the single-porosity model that demands tremendous grid refinement in intensely fractured shale oil reservoirs. The grid refinement is necessary surrounding the fracture intersections that makes the system become computationally expensive. In this chapter,

Texas Tech University, Tao Wan, May 2015 the matrix-matrix and matrix-fracture diffusion is coupled in a dual permeability model to overcome the drawback of single-porosity model. The simulation results demonstrate that the enhanced oil recovery by gas injection process in the Eagle Ford shale oil reservoir will benefit from matrix-matrix and matrix-fracture molecular diffusion.

4.2. Introduction

It is recognized that diffusion is an important recovery mechanism in recovering oil by gas injection in fractured reservoirs (Coats, 1989; Da Silva et al. 1989; Karimaie et al. 2007; Morel et al.1990). Ertekin et al. (1986) derived a slippage factor under the assumption that the driving mechanisms exerted by the concentration and pressure field are acting in parallel. The combination of Darcian flow velocity and molecular diffusion velocity yields a slippage factor that depends on composition, pressure and saturation. Followed Ertekin’s work, a lot of successive studies on gas diffusion (Allan et al. 2012; Javadpour 2009; Ozkan et al. 2010; Sakhaee-Pour and Bryant 2012; Shi et al.2013) used the dusty gas model (DGM) to model gas flow through nanoscale pores and throats based on the assumption that overall flow rate is a linear combination of gas transport mechanisms. Ozkan et al. (2010) incorporated Knudsen flow into the dual porosity formulation to simulate gas migration from matrix to fracture system. Javadpour (2009) developed the concept and formulation of apparent gas permeability in shale by adding the Knudsen diffusion and viscous forces as a total mass flux, similar to Ertekin’s model. Roy and Raju (2003) stated that different flow regimes are dependent on the Knudsen number. They modeled gas flow characteristics through microchannels and nanopores beyond the slip flow regime. They found that in the case of nanopore systems the continuum assumption is not valid. Although some studies (Grogan & Pinczewski, 1987; Darvish et al., 2006) suggests that molecular diffusion is an important recovery mechanism in the mobilization of oil in laboratory-scale floods, little work has addressed the role of diffusion and determined the time scales necessary for diffusion to be an effective recovery mechanism in the reservoir-scale flooding. This work focuses on examining the contribution of diffusion at reservoir flooding conditions. 60

Texas Tech University, Tao Wan, May 2015

The study of diffusion effect on fluid flow dynamics in shale resource plays that have complex pore networks starts to draw operators’ attention (Javadpour et al. 2007; Sakhaee-Pour & Bryant 2012; Schettler et al. 1989; Yuan et al. 2013). Most of the studies (Ghorayeb and Firoozabadi 2000; Hoteit et al. 2006, 2009, 2011; Jamili 2010) on numerical modeling of the diffusion role in fractured media used a single porosity, dual- continuum model to simulate naturally fractured reservoirs. The fractures are set as high permeable blocks and fine grid blocks are demanded surrounding the fractures. The construction of fracture-network requires using very fine grid-block near the fractures to accurately capture the rapid pressure, saturation changes and multi-phase flow effects surrounding the fractures. The disadvantage of this approach is self-explanatory because it demands a large number of refined grids and a huge amount of computing time. Hoteit and Firoozabadi (2009) presented a numerical simulation model of molecular diffusion for gas injection in 10-m spacing of fracture network. It demands more than one day of running time on a 2.5-GHz, Pentium 4 PC. It is almost impractical to use such a grid- refinement and multi-component EOS model in a full-field simulation. The deficiency of the single-porosity diffusion model is avoided by using a dual permeability model to represent fracture network that includes the diffusion transfer flux in it. The aim of incorporating diffusion flow into a dual permeability model is to significantly reduce the runtime but produce the same accurate results as a single-porosity diffusion model. Coats (1989) proposed a fully implicit numerical model for compositional simulation of fluid flow that includes the effect of diffusion in the dual-porosity model. He solved the diffusion equation in 1D and extended this approach to compositional simulation. However, this model only considered the gas phase diffusion flux between matrix and fracture. Matrix-matrix diffusion and diffusion rates in the oil phase were neglected. Jamili (2010) used a dual-continuum approach to examine the mass transfer between the fractures and the matrix blocks in naturally fractured reservoirs. The matrix is discretized into fine grids and the fractures act as the boundaries of the matrix. It simply comprises of the matrix with four fractures surrounding it. A lot of current studies on modeling the diffusion effect on gas flooding efficiency used a single porosity model to explicitly

Texas Tech University, Tao Wan, May 2015 represent the fracture network (Ghorayeb and Firoozabadi (2000); Hoteit et al. 2006, 2009, 2011). In this chapter, we coupled the diffusion equation in a dual permeability model that can properly simulate the fluids flow in shale oil reservoirs.

4.3. Description of Mathematical Model

The species balance for component i in an nc-component system is given by (Hoteit and Firoozabadi 2009; Jamili 2010) in the following convection-diffusion equation:

  ∂     φρωS  +∇⋅ ρω u + J  = 0, i=1,...,nc j=o, g (4.1) ∂t ∑jjij ∑ jijj ∑ ij, j  j j  accumulation convective flux diffusion flux 

The velocity for each phase is described by Darcy’s law and the diffusion is followed by Fick’s law

kk =−rj ∇ + uj() p jρ j g j=o, g (4.2) µ j

J= − S D ∇ ij, φρ jjij ω ij, i=1,...,n c ; j=o, g (4.3)

The composition !$ is constrained by

nc ω = ∑ ij 1, j=o, g (4.4) i=1

Therefore, the governing matrix flow equation of the diffusion model can be expressed as:

∂     S  +∇⋅ u + J + = ∑φρωjjij   ∑() ρω jijj ij,  ∑ τ mfij , 0, i=1,...,nc j=o, g (4.5) ∂t j j j    m  m

The fluids flow conservation equation in the fracture is:

∂     φρωS   +∇⋅ ρω u + J  − τ = 0, i=1,...,nc j=o, g (4.6) t ∑jjij  ∑() jijj ij, ∑ mfij , ∂ j j j    f  f

Texas Tech University, Tao Wan, May 2015

The matrix-fracture transfer term can be obtained by modifying the equation proposed by Kazemi et al. (1976). In our work, the transfer of a component between matrix-fracture by diffusion is simply described by Fick’s law that preserves the same form as viscous displacement. Modeling of matrix-to-cleat diffusion in the coal-bed methane adopts the same approach to simulate the diffusive mass flow rate of a species (CMG, 2011).

The mass transfer formulation between the matrix and fracture is expressed as:

k ρ  =rj j + − τσmfij,()()p jm ,,- p jf ρωω jjij SD ijm ,, ijf  i=1,2...nc , j=o,g (4.7) µ j 

= + pg p o p cog (4.8)

S+ S = 1 o g (4.9)

k k k  σ =4 x +y + z  (4.10) L2 L 2 L 2 x y z 

Where %denotes the molar density of phase j (j=o, g). ͍%,ͤ%,͟-% , ͩ% and % represent the saturation, pressure, relative permeability, velocity and viscosity of phase j, respectively.

The subscript i with i=1,…,nc corresponds to the components. ̾$% is the effective diffusion coefficient of component i in the phase j. ̈́$,%is the mass flux of component i in

the phase j by diffusion. !$% represents the molar fraction of component i in the phase j.

(! ,$% is the matrix-fracture transfer of component i in the phase j.

Equation 4.1 can be derived from the fundamental phase conservation equation (Lake, 1989). Jamili (2010) developed the same mathematical model as Hoteit’s model to simulate diffusion and convection mechanisms for gas injection in naturally fractured reservoirs. The governing matrix transport equation for each species in the oil and gas phases due to convective and molar diffusive flux is shown in Eq.4.5. The governing flow equation in the fracture is represented by Eq.4.6. Eqs.4.5 and 4.6 are developed on the basis of Eq.4.1. The governing equations of matrix and fracture are similar to the dual permeability formulation, except that the diffusion of components in the oil and gas 63

Texas Tech University, Tao Wan, May 2015 phases is considered. We only included the molecular diffusion in this chapter without containing mechanical dispersion. We used the Gilman and Kazemi shape factor to characterize the matrix-fracture transfer coefficient.

The system of diffusion equations in the dual permeability model consists of 2×(2n c+4) equations. The 2×(2n c+4) unknowns are , , , , , …, and Ƴ͊* ͊" ͍* ͍" ͬͥ ͬͦ ͬ) Ʒ(

, , , , , …, . There are 2n c equations from Eqs. 4.5 and 4.6 and 2nc Ƴ͊* ͊" ͍* ͍" ͬͥ ͬͦ ͬ) Ʒ! equations that can be derived from thermodynamic equilibrium (fugacity) equations. The rest of the required equations are from Eqs.4.4, 4.8 and 4.9, that is, mass fraction conservation equation, capillary pressure equation and saturation equations in the matrix and fracture. We used an adaptive-implicit approach that is developed by (Collins et al. 1992) to solve the system of equations. The essence of adaptive-implicit approach is that the task of solving the flow equations is independent from solving the equilibrium equations. It avoids the drawback of solving the flow and phase equilibrium equations simultaneously that has a high level of complexity to find the solution. The solution method of this model is well-established in the literature (Collins et al. 1992; Jamili 2010).

The difference between this model and the previously developed model by Hoteit and Firoozabadi (2009) is that we coupled the diffusion mechanism into a dual permeability model and included the matrix-fracture mass transfer due to diffusive flux (as shown in Eq.4.7). The difference between this model and the Coats (1989) model is that the latter does not include diffusion rate of a component in the oil phase and the diffusive flux within matrices and fractures.

4.4. Model Validation

Our numerical simulation results from the model are benchmarked using experimental data and published numerical simulation results. Then, simulation results of gas diffusion in nanopores are presented and compared with conventional numerical model by Coats

Texas Tech University, Tao Wan, May 2015

(1989). Kovscek et al. (2008) reported a series of experimental results of using CO 2 injection to enhance oil recovery in low permeability shale rocks (0.02-1.3 mD). Initially, the shale core sample is saturated with live oil at 1300 psi. Then, it is depleted to a pressure at 350-psi. Two CO 2 injection modes followed the primary depletion. Countercurrent flow and concurrent injection schemes were employed to evaluate the oil recovery potential after primary production. The core sample is placed in the horizontal direction. Gravity segregation effect is not considered in their study. In the countercurrent mode, CO 2 is flushed through the inlet at constant pressures while the outlet is sealed. A sketch of experimental apparatus for countercurrent CO 2 flow and concurrent flow is shown by Kovscek et al. (2008) (their Fig.4.5 and Fig.4.6). The injected carbon dioxide diffuses from fractures into the porous matrix to displace oil. The experimental setup of countercurrent mode is designed to evaluate the oil recovery in the absence of viscous displacement. The experimental results for 0.023 mD shale rock sample showed no incremental oil production during the countercurrent flooding stage. Concurrent flow is performed at an injection pressure at which viscous displacement becomes the dominant oil recovery mechanism. Later, Vega et al. (2010) tried to simulate the miscible gas injection process (1.32 mD shale sample) including countercurrent and concurrent flow modes. But the individual experimental recovery process was unable to be matched. There is a huge gap between their simulation results and experimental data. They used a Carman-Kozeny type of porosity-permeability correlation to generate the permeability distribution of the shale rock sample. There is no evidence to support that the approach they used can represent the fracture and matrix permeability distribution in shale. The fact that there is significant deviation between their simulation results and experimental results is an indication that the complexity of fracture network in shale rocks is not fully characterized by using this correlation. In this chapter, we will use a dual permeability model to simulate the experimental results of gas injection in ultra-tight siliceous shale of 0.023 mD presented in Kovscek’s work ( Table 4.1). The matrix permeability of the sample is 0.023 mD. The fracture permeability used in dual permeability model is 1 mD. The lumped reservoir fluids pseudo-component description is given by Vega et al. (2010).

Texas Tech University, Tao Wan, May 2015

Table 4.2 and 4.3 summarizes the compositional description of reservoir fluid and binary interaction coefficients used in experimental simulation. Sigmund (1976) method is used to model the molecular diffusion and calculate binary diffusion coefficient between component i and j in the mixture.

Table 4.1. Core sample properties

Property Sample

Length, cm 7.3

Diameter, cm 3.2

Matrix Permeability, mD 0.023

Porosity, fraction 0.3

Initial pressure, psi 1400

CO 2 injection Yes

Table 4.2. Compositional description of reservoir fluid in experimental simulation

Component Mole Fraction Pc (atm) Tc (K) Acentric factor Mol. Weight (g/mol) CO 2 0.0036 72.8 304.2 0.225 44.01 C1 0.1565 45.4 190.6 0.008 16.04 C2-C3 0.0815 45.05 338.93 0.1246 36.99 C4-C6 0.1002 34.28 462.55 0.2258 70.14 C7-C15 0.4494 24.13 627.20 0.3984 135.20 C16 -C34 0.1576 12.44 828.07 0.8581 305.01 C35+ 0.0512 6.50 786.09 0.8421 644.85

Table 4.3. Binary coefficient used for experimental simulation

Component CO 2 C1 C2-C3 C4-C6 C7-C15 C16 -C34 C35+

CO 2 zero 0.103 0.0 0.0 0.0 0.0 0.0 C1 0.103 zero 0.0 0.0 0.0 0.0 0.0 C2-C3 0.0 0.0 zero 0.0 0.0 0.0 0.0 C4-C6 0.0 0.0 0.0 zero 0.0 0.0 0.0 C7-C15 0.0 0.0 0.0 0.0 zero 0.0 0.0 C16 -C34 0.0 0.0 0.0 0.0 0.0 zero 0.0 C35+ 0.0 0.0 0.0 0.0 0.0 0.0 zero

Texas Tech University, Tao Wan, May 2015

40 35 Primary depletion Concurrent stage 30 Countercurrent 25 20 Experimental Results 15 Simulation Results Oil Recovery, % Oil 10 5 0 1 10 100 Days

Figure 4.1-Comparative cumulative oil recovery

Fig.4.1 shows the cumulative oil recovery of the simulation model contrasting with the experimental data at depletion, countercurrent flow and concurrent flow stages. Results produced by the model are in good agreement with the experimental results. There is little noticeable oil recovery from the countercurrent flow stage because the production is performed in a short span of time and there is no viscous displacement and gravity drainage assisted recovery. Another reason may attribute to that the fluid is injected at high flow velocities so that diffusion is unable to effectively equalize concentration in such a short residence time. Most of the oil production is recovered during pressure- driven stage.

The model is also compared with Hoteit and Firboozabadi (2009) numerical model. In

Hoteit’s example, the reservoir fluid is simplified that only contains C1/C 3 mixture. Methane is injected as a solvent at a rate of 1.3x10 -4 pore volume per day to displace

C1/C 3 mixture. The reservoir domain is assumed to be a 2D cross section with 500-m length and 100-m height, as shown in Fig.4.2. In Hoteit’s model, different sizes of matrix blocks (100x10, 10x10 and 10x5) were used to construct the fracture-networks that have different fracture spacing. The natural fracture spacing is adjusted by varying the sizes of matrix blocks. The fracture aperture is equal to 0.5 mm. Grid refinement is needed for the

Texas Tech University, Tao Wan, May 2015 area surrounding the fractures. The drawback of this approach is that it would require tremendous grid blocks when the fracture-network spacing is smaller. The most important advantage of a dual permeability model over a single-porosity model is that it does not require detailed grid-refinement because the natural fractures are incorporated into the model. For example, if the spacing of fracture-network is 10-m, we can use 10mx10m (32.8 ft x32.8 ft) blocks in which two perpendicularly crossed 0.001-ft wide fractures (approximately actual fracture width) are assumed to exist which runs through each 32.8ft x32.8 ft block. Therefore, the input fracture porosity in the dual permeability model is 6.1E-05 = (2xV frac /V block ). The simulation results in Hoteit’s simulation example 1 was reproduced. Methane is injected at the top of the right corner matrix blocks to

displace C 3 in the domain. The fracture network spacing is 10-m apart away in both X and Z directions. We used 10 x 10 m grid-blocks in dual permeability model that implies the fracture spacing is 10-m. The fracture porosity is 6.1E-05. The same diffusion coefficients as Hoteit was used. They are in the range of 8.7 x 10 -8 and 2.2 x 10 -9 m2/sec in gas and oil phase, respectively. The bottom-hole flowing pressure at the production well is maintained equal to the initial reservoir pressure to eliminate the viscous displacement caused by pressure gradient. The reservoir model setup probes recovery by gas injection in fractured reservoirs under a zero pressure gradient. In this case, the gravity drainage and diffusion flow becomes the dominant recovery mechanisms in fractured reservoirs.

Figure 4.2-2D single porosity model with 10 x 10-m matrix blocks

Texas Tech University, Tao Wan, May 2015

100 Dual Perm-With Diffusion 90 Dual Perm -No Diffusion 80 70 Hoteit (2009) -Without Diffusion 60 Hoteit (2009) -With Diffusion 50 40 30 C3 Recovery, C3 fraction 20 10 0 0 20 40 60 80 100 PVI, % Figure 4.3-Comparison of C 3 recovery for the dual permeability model with single porosity model

Our simulation results are well-matched with the single porosity model that uses grid- block refinement, as shown in Fig.4.3. It takes only few minutes to simulate the test by using the dual permeability model coupled with diffusion, while their model takes hours to run (Hoteit and Firboozabadi, 2009). Fig.4.3 clearly shows that diffusion contributes to a large percentage of oil recovery in absence of pressure gradient driven convective transport.

Then, the importance of diffusion effect on fractured shale gas reservoirs was investigated. Rock and fluid properties (relative permeabilities, phase behavior and grid- block distribution) were the same as used in the Hoteit’s model. The matrix permeability was simply changed into 0.0001-mD. The C 3 recovery, with and without considering

diffusion effect, by CO 2 injection is presented in Fig.4.4. It obtained 86% of C 3 recovery

in CO 2 injection process due to contribution of diffusion in absence of viscous displacement. Fig.4.4 indicates that diffusion plays a crucial role for improving gas recovery in fractured shale reservoirs. Similar observations of the importance of diffusion effects on shale gas production have been reported in the literature (Javadpour et al. 2007; Sakhaee-Pour & Bryant 2012; Schettler et al. 1989; Yuan et al. 2013). 69

Texas Tech University, Tao Wan, May 2015

100 90 Without Diffusion in Shale 80 With Diffusion in Shale 70 60 50 40 30 20 C3 Recovery, C3 fraction 10 0 0 20 40 60 80 100 PVI, %

Figure 4.4-Comparison of C 3 recovery by CO 2 injection in shale gas reservoirs 4.5. Simulation Results and Discussion

The main objective of this chapter is to model the dispersive-convective flux through nanopores in shale oil reservoirs during gas injection process using the dual permeability model coupled with diffusion. The reservoir fluid and rock properties used in this model are based on the published data in Eagle Ford shale as previously used (Wan et al. 2014). The initial reservoir oil compositions are shown in Table 4.4 which represents light oil. Table 4.5 presents the binary coefficients used for Eagle Ford fluid. The effective diffusion coefficients for different components in the oil and gas phases are presented in Table 4.6 which were calculated from Wilke-Chang equation (Wilke and Chang, 1955). Hydrodynamic dispersion includes both molecular diffusion and mechanical dispersion. The molecular diffusion coefficients between component i and j in the gas phase are calculated by Sigmund (1976) method. The Wilke-Chang equation is used to estimate effective oil phase molecular diffusion coefficients. The basic reservoir properties are presented in Table 4.7.

Texas Tech University, Tao Wan, May 2015

Table 4.4. Peng-Robinson EOS Fluid Description

Components Initial Pc (atm) Tc (k) Acentric Fac. MW Vc Comp. C1 0.5 45.44 190.6 0.013 16.0 0.0998 CO 2 0.0 72.8 304.2 0.225 44.01 0.094 C2-3 0.03 41.94 369.8 0.152 44.1 0.2005 C4-6 0.07 29.73 507.4 0.300 86.2 0.3698 C7-10 0.2 20.69 617.7 0.488 142.2 0.6297 C11-15 0.15 13.61 705.6 0.65 206 1.0423 C16+ 0.05 11.02 766.7 0.85 282 1.3412

Table 4.5. Binary coefficient for Eagle Ford fluid and reservoir flooding case

Component C1 CO 2 C2-C3 C4-C6 C7-C10 C11 -C15 C16+ C1 zero 0.103 0.0 0.0 0.0 0.05 0.05 CO 2 0.103 zero 0.135 0.0 0.0 0.0 0.0 C2-C3 0.0 0.135 zero 0.0 0.0 0.005 0.005 C4-C6 0.0 0.0 0.0 zero 0.0 0.0 0.0 C7-C10 0.0 0.0 0.0 0.0 zero 0.0 0.0 C11 -C15 0.05 0.0 0.005 0.0 0.0 zero 0.0 C16+ 0.05 0.0 0.005 0.0 0.0 0.0 zero

Table 4.6. The effective diffusion coefficients of different components at 2000 psi

Components C1 CO 2 C2-C3 C4-C6 C7-C10 C11 -C15 C16+

2 Gas Phase, cm /s 1.30E-04 2.13E-04 1.60E-04 1.07E-04 7.80E-05 5.77E-05 4.88E-05

2 Oil Phase, cm /s 4.38E-05 3.99E-05 3.11E-05 1.93E-05 1.53E-05 1.21E-05 8.82E-06

Table 4.7. Reservoir properties for the model input

Initial Reservoir Pressure 6425 psia Reservoir Temperature 160 F o Saturation Pressure 2302 psia Rock Compressibility 5.0E-06 Porosity 6 % Permeability of shale matrix 100 nano-Darcy

The diffusion effect on gas flooding efficiency in liquid-rich shale is examined that is implemented in two hydraulically zipper fractured horizontal wells. Considering that CO 2

Texas Tech University, Tao Wan, May 2015

EOR process is limited by the cost and availability of CO 2 in the field, it is favorable to use produced natural gas as the injection gas. The composition of injector fluid is

specified as 70% C 1, 20% C 3 and 10% C 6. Reinjection of produced gas into existing reservoirs is an economically available avenue to perform EOR projects due to low natural gas prices. We applied staggered zipper frac technique to stimulate two adjacent horizontal wells to maximize the exposure of reservoir rock. Two horizontal wells drilled through the reservoir, which are closely spaced, one adjacent to the other, forming a gas injector and producer pair, as shown in Fig.4.5. The dimension of the shale reservoir is 2000-ft long ×1000-ft wide ×200-ft thick. An effective fracture permeability of 0.04 mD is used to simulate the natural fractures in the stimulated reservoir volume (SRV). The hydraulic fracture conductivity is 83.3 mD-ft. The hydraulic fracture half-length is 250-ft. The horizontal well is stimulated with 10 transverse hydraulic fractures at the spacing of 200-ft apart from each other. The LS-LR-DK (logarithmically spaced, locally refined and dual permeability) model is used to accurately simulate the fluid flow in fractured shale oil reservoirs. The LGR coupled to standard DK grids is used to represent the hydraulic fractures owing to the high permeability contrast between fracture and matrix. Better discussions about the advantages of this LS-LR-DK model can be found in the literature (Rubin 2010).

Figure 4.5-The horizontal well pair perforated and stimulated in a staggered pattern

Texas Tech University, Tao Wan, May 2015

Figure 4.6-Simulation Unit

The producer is subject to minimum bottom-hole pressure constraint (BHP) of 2000 psi. Our operational bottom-hole flowing pressure is close to the minimum miscibility pressure requirement for injected solvent to achieve miscible flooding with reservoir oil. We specified the solvent injection rate at 1 PVI/10000 days for injector and the maximum allowed bottom-hole injection pressure is 5000 psi. Another similar case with an injection rate at 1 PVI/5000 days is also simulated to investigate the effects of injection rate on oil recovery in multistage hydraulically fractured horizontal wells. In our simulation model, we only simulate a half unit of the hydraulic fracture controlled volume based on the symmetry of fluid flow, as shown in Fig.4.6. The red lines represent hydraulic fractures. In Fig.4.6, we used a fine logarithmic gridding surrounding the hydraulic fractures and near the wellbore. The simulation domain was a two-dimensional Cartesian section (x-y). Twenty two gridblocks were used in the x direction with a variable gridblock size and forty two gridblocks in the y direction. We performed grid sensitivity study and found that further grid refinement did not affect the numerical results.

Texas Tech University, Tao Wan, May 2015

The matrix permeability is 1-md 70

Coats model (1989) 60 Our simulation model

20 C15 component Recovery,%

0 0 1000 2000 3000 4000 5000 6000 7000 Time (days)

Figure 4.7-Comparison of Coats’s model results with our model for matrix permeability km = 1 mD

The matrix permeability is 0.0001-md 18 Coats model (1989) 16 Our simulation model

6 C15 component Recovery,% 4

0 0 1000 2000 3000 4000 5000 6000 7000 Time (days)

Figure 4.8-Effect of diffusion in the oil phase and within matrices on shale oil recovery 74

Texas Tech University, Tao Wan, May 2015

Matrix permeability = 1 md 10

4 Peclet number

0 5 10 15 20 Grid away from the producer

Figure 4.9-Peclet number in the oil phase in the matrix (k m=1 md) at 7000 days

-3 x 10 Matrix permeability = 1E-04 md 4

3.5

2.5

1.5 Peclet number 1

0.5

0 5 10 15 20 Grid away from the producer

Figure 4.10-Peclet number in the oil phase in the shale matrix (k m=1E-04 md) at 7000 days

Figs.4.7 and 4.8 compare our simulation results with Coats (1989) model output results for the high matrix permeability case (1 mD) and the shale matrix permeability case (0.0001-mD). In Coats’s model, it only includes the gas phase diffusion term in the molecular flux between a fracture block and a matrix block. He did not consider the 75

Texas Tech University, Tao Wan, May 2015 diffusion rate of components in the oil phase and the diffusive flux from matrix to matrix, which may be an important recovery mechanism in tight shale oil or gas reservoirs. Fig.4.7 shows the results produced by Coats’s model are in good agreement with our simulation results for conventional reservoirs (k m=1 md). The importance of matrix- matrix diffusion and diffusion rate in the oil phase is observed in Fig.4.8 in the low permeability shale case. Fig.4.8 shows that the impact of matrix-fracture diffusion rate in the oil phase and matrix-matrix diffusion on ultimate oil recovery is significant in tight shale reservoirs (k m= 1e-04 mD). In very low permeability shale, it is essential to model the diffusion rate in the oil phase and within matrices. Figs.4.9 and 4.10 show the Peclet number in the oil phase in the matrix blocks along the horizontal producer (1~22, 32, 1). The Peclet number is defined as the ratio of convective transport rate to dispersive transport rate (Lake, 1989). The Peclet number is a measure of the relative importance of advection to diffusion. Perkins and Johnson (1963) have studied the influence of diffusion and dispersion on miscible displacement processes. They found that when Peclet number is less than 0.02, the transport is controlled by diffusion. The Peclet number in the shale matrix is considerably lower than that of in the conventional reservoirs (as shown in Figs.4.9 and 4.10). This observation tends to confirm the idea that diffusion plays an important role in recovering oil from nano-scale shale oil reservoirs. Without including the matrix-fracture diffusion in the oil phase and matrix-matrix diffusion, the oil recovery from the shale matrix is significantly lower. It reflects that matrix-fracture diffusion in the oil phase and diffusion within matrices play a crucial role in recovering shale matrix oil.

Texas Tech University, Tao Wan, May 2015

10 Diffusion, 1 PVI/10000D

Shalerecoveryoil factor, % Diffusion, 1 PVI/5000D 5 Without Diffusion, 1 PVI/10000D Without Diffusion, 1 PVI/5000D

0 0 0.2 0.4 0.6 0.8 1 PVI

Figure 4.11-Oil recovery vs. PVI

10 Diffusion, 1 PVI/5000D

Shalerecoveryoil factor, % Without Diffusion, 1 PVI/5000D 5 Diffusion, 1 PVI/10000D Without Diffusion, 1 PVI/10000D 0 0 2000 4000 6000 8000 10000 12000 Time (days)

Figure 4.12-Oil recovery vs. time

Texas Tech University, Tao Wan, May 2015

Fig. 4.11 and Fig.4.12 show the effect of diffusion and injection rate on shale oil recovery. One can observe that the diffusion plays a role in improving oil recovery at two hydraulically fractured horizontal wells flooding case. With the diffusion, the velocity of a component is owing to the summation of both convection and diffusion velocities. Vapor-vapor diffusion is about tenfold faster than vapor-liquid diffusion (da Silva and Belery 1989). Thus, the injected gas will diffuse from the fracture to the matrix with a velocity considerably higher than the velocity of crude oil being flushed out by convective flow. In terms of one pore volume injection, the scenario with an injection rate at 1 PVI per 10000 days achieved higher oil recovery than the rate of 1 PVI per 5000 days (Fig.4.11). If the injected fluid flows at a high interstitial velocity through the porous medium, the oil recovery due to diffusion will be impaired because there is less residence time for fluids in high concentration side to diffuse in low concentration pore spaces (Perkins and Johnston, 1963). However, Fig.4.12 shows that higher gas injection rate achieves higher oil recovery at a given injection time because it requires more pore volumes of solvent injection.

The incremental oil recovery due to diffusion by using lower injection rates at a given pore volume injection has raised the consideration on the trade-offs between the potential economic gains and maximization of ultimate hydrocarbons recovery. We performed an economic analysis to evaluate the potential economic benefits from these two injection rates. The study assesses shale oil production in two different injection schemes (1PVI/10000D and 1PVI/5000D) in order to determine how to extract shale oil economically in the field. Well economics vary greatly across the basin as a function of productivity, geology, drilling and stimulation cost. Presumably, in unconventional reservoirs, it will cost approximately $4 million to drill a horizontal well and $3 million to stimulate and complete it (Alexander et al., 2009). Table 4.8 presents the input economic parameters for calculation of the NPV. All capital expenditures occur in the year of first production. The capital investment in shale oil production includes drilling, completion and stimulation costs. The operating expenditure is assumed to be $4/boe. The oil prices are assumed to be flat over the life cycle of production ($100 per barrel). 78

Texas Tech University, Tao Wan, May 2015

Royalty is levied on gross production at a rate of 20%.

Table 4.8. Required inputs for the NPV calculations

Economic Parameters Unit Value Discount rate fraction/yr 10% Working interest fraction/yr 100% Royalty rate fraction/yr 20% Net revenue interest fraction/yr 80% Severance tax fraction 4.6% Ad valorem tax fraction 3% CAPEX, $ $MM 14 (2 wells) OPEX, $ $/boe 4 Oil price, $ $/barrel 100

15 Primary 10

5 Secondary recovery

0 0 5 10 15 20 25 30 35 -5

NPV, $Millions NPV, -10 1PVI/10000 days -15 1PVI/5000 days -20 Years

Figure 4.13-Comparative NPV by two different injection rates

5 x 10 60 2.5

50 2

40 1.5

1 20 1 PVI/10000D, Gas prod rate 1 PVI/5000D, Gas prod rate 0.5 10 1 PVI/10000D, Cumu gas injection Production gas rate, Mscf/day 1 PVI/5000D, Cumu gas injection Cumulative gas injection, Mscf 0 0 0 2000 4000 6000 8000 10000 12000 Time,Days

Texas Tech University, Tao Wan, May 2015

Figure 4.14-Comparison of gas production rates and cumulative gas injection by two different injection rates

Fig.4.13 shows the results of the net present value (NPV) calculations for two different injection rates at one pore volume injection. Fig.4.14 presents the comparison of gas production rates and cumulative gas injection by two different injection schemes. The insignificant difference of NPV in Fig.4.13 reflects the project cannot achieve more return by enduring the injection time for a given injection pore volume. Although it yielded higher oil recovery by using slower injection rate that extends injection time, the NPV was even lower. Depreciation of cash flow has a negative influence on return that comes from diffusion contributed oil recovery. Along with the continuous operating expenditures including compression cost and gas reinjection cost, it is not a good idea to take advantage of diffusion increased oil recovery by extending injection time. The NPV calculations aim to provide an insight for designing the solvent injection rates in shale oil reservoirs and to estimate the project’s return.

Injection rate = 1 PVI/5000 D 40 200-ft, Diffusion 35 200-ft, Without Diffusion 100-ft, Diffusion 30 100-ft, Without Diffusion 50-ft, Diffusion 25 50-ft, Without Diffusion

10 Shalefield oil recovery, %

0 0 1000 2000 3000 4000 5000 6000 7000 Time (days)

Figure 4.15-Effect of natural fracture spacing on gas injection performance

Fig.4.15 compares the effect of natural fracture spacing on gas injection recovery performance. It is observed that the diffusion effect is not quite significant at the primary production stage. The importance of molecular diffusion is magnified in gas injection 80

Texas Tech University, Tao Wan, May 2015 enhanced recovery process in highly fractured shale oil reservoirs with smaller fracture spacing. The increase of oil recovery by inclusion of diffusion effect for the case of 50-ft fracture spacing is much higher than 200-ft spacing case. In densely fractured reservoirs, diffusion recovery mechanism predominates with a decrease of fracture spacing.

It is noteworthy that selection of different shape factors may lead to different reservoir behaviors. Gilman and Kazemi shape factor assumes that all fractures are instantly immersed in water and under quasi-steady state conditions (Rangel-German and Kovscek, 2003). One deficiency in use of Kazemi’s shape factor is that assumptions of pseudo- steady state and instantly immersed fractures may break down in shale reservoirs. The shape factors are treated as adjustable parameters which are used to reproduce observed field or laboratory results. However, this approach does not necessarily capture the physics of matrix-fracture transfer flow behavior. Bahrami et al. (2008) presented an approach of integrating image log data associated with well-test analysis to determine the shape factor. Combining of core analysis, well-test data and well-logging data might be an effective approach to estimate the shape factor. More effort is needed to develop a shape factor that can characterize the transient behavior and match the production performance in shale reservoirs.

4.6. Conclusions

In this chapter, we coupled the diffusion equation with a dual permeability model so that the enhanced oil recovery processes by gas flooding can be properly simulated. There are difficulties to use a single porosity dual-continuum model to match the experimental results of gas flooding in the low permeability shale rocks, because it can not properly capture the matrix-fracture mass transfer rates of densely distributed microfractures. The results produced by dual permeability model coupled with diffusion are in good agreement with the experimentally measured data in shale rocks. The simulation work addressed the role of diffusion in the field scale flooding. The impact of matrix-fracture diffusion rate in the oil phase and matrix-matrix diffusion on oil recovery is significant in

Texas Tech University, Tao Wan, May 2015 tight shale oil reservoirs. It is observed that the interstitial velocity of oil phase in the shale matrix is considerably lower than that of the conventional reservoirs. In very low permeability shale oil or gas reservoirs, the dominant recovery mechanism is by diffusion according to the analysis of Peclet number. One noticeable finding of gas injection in shale reservoirs is that without including the matrix-fracture diffusion in the oil phase, it results in lower oil recoveries. It is essential to model the matrix-fracture diffusion rate in the oil phase and diffusion within the matrices for tight shale oil reservoirs. In conventional reservoirs, Coats’s (1989) model is able to produce good results because the transport is controlled by viscous flow.

The importance of diffusion effect on fractured shale gas reservoirs and shale oil reservoirs is observed. It is important to notice that the oil recovery by including diffusion effect is much higher than the case without considering diffusion in shale oil reservoirs. In the cases where the convective flux for a component is negligible because of low pressure drawdown, diffusion due to compositional differences between matrix and fracture tends to become the main recovery mechanism. From a reservoir management perspective, relying on diffusion enhanced oil recovery by decreasing injection rate is not the best way to exploit shale resource plays.

4.7. References

Alexander, T., Baihly, J., Boyer, C., Clark, B., Jochen, V., Calvez, J.L., Lewis, R., Thaeler, J. and Toelle, B.E. 2011. Shale Gas Revolution. Oilfield Review Autumn 23 (3):40-55.

Allan, A.M. and Mavko, G. 2012. The Effect of Adsorption and Diffusion on the Gas Permeability of Kerogen. SEG-2012-0076 presented at 2012 SEG Annual Meeting, 4- 9 November, Las Vegas, Nevada.

Bahrami, H., Siavoshi, J., esmaili, S., Karimi, M.H., and Bahraie, R. 2008. Estimating Fracture Permeability and Shape Factor by Use of Image Log Data in Welltest Analysis. SPE 114594 presented at SPE Asia Pacific Oil and Gas Conference and 82

Texas Tech University, Tao Wan, May 2015

Exhibition, 20-22 October, Perth, Australia.

CMG Manual. 2011. Coal Matrix-to-Cleat Modelling Parameters. GEM Advanced Compositional and GHG Reservoir Simulator, Version 2011, pp. 423-424 and 1155- 1225.

Coats, K.H. 1989. Implicit Compositional Simulation of Single-Porosity and Dual- Porosity Reservoirs. Paper SPE 18427 presented at the SPE Symposium on Reservoir Simulation, Houston, 6–8 February.

Collins, D.A., Nghiem, L.X., and Li, Y-K., Grabonstotter,J.E. 1992. An Efficient Approach to Adaptive-Implicit Compositional Simulation With an Equation of State. SPERE 2 (7):259-264.

Da Silva, F.V. and Belery, P. 1989. Molecular diffusion in naturally fractured reservoirs: a decisive recovery mechanism. SPE Paper 19672 presented at the SPE Annual Technical Conference and Exhibition, 8-11 October, San Antonio, Texas.

Darvish, G.R., Lindeberg, E., Holt, T., Utne, S.A., and Kleppe, J. 2006. Reservoir- Conditions Laboratory Experiments of CO2 Injection Into Fractured Cores. Paper SPE 99650 presented at the SPE Europec/EAGE Annual Conference and Exhibition, Vienna, Austria, 12–15 June.

Ertekin, T., King, G.R., and Schwerer, F.C.1986. Dynamic Gas Slippage: A Unique Dual- Mechanism Approach to The Flow of Gas in Tight Formations. SPEFE 1 (1):43-52. SPE 12045-PA.

Ghorayeb, K. and Firoozabadi, A. 2000. Numerical Study of Natural Convection and Diffusion in Fractured Porous Media. SPE J 5 (1): 12-20. SPE-51347-PA.

Grogan, A.T. and Pinczewski, W.V. 1987. The Role of Molecular Diffusion Processes in Tertiary CO2 Flooding. SPE J 5 (39): 591-602.

Texas Tech University, Tao Wan, May 2015

Hoteit, H. and Firoozabadi, A. 2009. Numerical Modeling of Diffusion in Fractured Media for Gas-injection and-Recycling Schemes. SPE J 14 (2):323-337. SPE- 103292-PA.

Hoteit, H. 2011. Proper Modeling of Diffusion in Fractured Reservoirs. Paper SPE 141937 presented at SPE Reservoir Simulation Symposium, 21-23 February, The Woodlands, Texas.

Hoteit, H. and Firoozabadi, A. (2006). Numerical Modeling of Diffusion in Fractured Media for Gas-injection and-Recycling Schemes. Paper SPE 103292 presented at Annual Technical Conference and Exhibition, 24-27 September, San Antonio, Texas, USA.

Jamili, A. 2010. Modeling effects of diffusion and gravity drainage on oil recovery in naturally fractured reservoirs under gas injection. Ph.D dissertation at the University of Kansas.

Javadpour, F., Fisher, D., and Unsworth, M. 2007.Nanoscale Gas Flow in Shale Gas Sediments. Journal of Canadian Petroleum Technology 46 (10):55-61. Doi: 10.2118/ 10.2118/07-10-06.

Javadpour, F. 2009. Nanopores and Apparent Permeability of Gas Flow in Mudrocks (Shales and Siltstone). JCPT 48 (8): 16-21. Doi: http://dx.doi.org/10.2118/09-08-16- DA.

Karimaie, H. 2007. Aspects of water and gas injection in fractured reservoirs. PhD thesis, Norwegian University of Science and Technology.

Kazemi, H., Merrill, L. S., Porterfield, K.L. and Zeman, P.R. 1976. Numerical Simulation of Water-Oil Flow in Naturally Fractured Reservoirs. SPE J 16 (6):317-326. SPE- 5719-PA.

Kovscek, A.R., Tang, G.Q., and Vega, B. 2008. Experimental Investigation of Oil 84

Texas Tech University, Tao Wan, May 2015

Recovery from Siliceous Shale by CO2 Injection. Paper SPE 115679 presented at the SPE Annual Technical Conference and Exhibition held in Denver, Colorado, 19–22 September.

Lake, L.W.1989. Enhanced Oil Recovery, Prentice-Hall, Englewood Cliffs, New Jersey.

Morel, D., D., Bourbiaux, B., Latil, M., and Thiebot, B.1990. Diffusion Effects in Gas- Flooded Light Oil Fractured Reservoirs. SPE Paper SPE 20516 presented at the Annual Technical Conference and Exhibition, Sept. 23-26.

Ozkan, E., Raghavan, R.S., and Apaydin, O.G. 2010. Modeling of Fluid Transfer From Shale Matrix to Fracture Network. Paper 134830 presented at SPE Annual Technical Conference and Exhibition, 19-22 September, Florence, Italy.

Perkins,T.K. and Johnston,O.C. 1963. A Review of Diffusion and Dispersion in Porous Media. SPE J 3 (1):70-84. SPE-480-PA.

Rangel-German, E. R. and Kovscek, A. R. 2003. Time-Dependent Matrix-Fracture Shape Factors for Partially and Completely Immersed Fractures. SPE 84411 presented at SPE Annual Technical Conference and Exhibition, 5-8 October, Denver, Colorado.

Roy, S., Raju, R., Chuang, H.F., Cruden, B.A., and Meyyappan, M. 2003. Modeling Gas Flow through Microchannels and Nanopores. Journal of Applied Physics 93 (8): 4870- 4879.

Rubin, B. 2010. Accurate Simulation of Non Darcy Flow in Stimulated Fractured Shale Reservoirs. Paper SPE 132093 presented at Western Regional Meeting, Anaheim, California, USA, 27-29 May.

Sakhaee-Pour, A. and Bryant, S. 2012. Gas Permeability of Shale. SPE J 15 (4): 401-409. SPE-146944-PA.

Schettler, PD., Parmely, CR., and Lee, WJ. 1989. Gas storage and transport in Devonian 85

Texas Tech University, Tao Wan, May 2015

shales. SPEFE 4 (3):371–376. SPE-17070-PA.

Shi, J., Zhang, L., Li, Y., Yu, W., He, X., Liu, N., Li, X., and Wang, T. 2013. Diffusion and Flow Mechanisms of Shale Gas through Matrix Pores and Gas Production Forecasting. Paper 167226 presented at SPE Unconventional Resources Conference Canada, 5-7 November, Calgary, Alberta, Canada.

Sigmund, P.M. 1976. Prediction of Molecular Diffusion At Reservoir Conditions. Part 1- Measurement And Prediction of Binary Dense Gas Diffusion Coefficients. JCPT 15 (2):48-57.

Vega, B., O’Brien, W.J., and Kovscek, A.R. 2010. Experimental Investigation of Oil Recovery from Siliceous Shale by Miscible CO2 Injection. Paper SPE 135627 presented at the SPE Annual Technical Conference and Exhibition held in Florence, Italy, 19–22 September.

Wan, T., Meng, X., Sheng, J., and Watson, M. 2014. Compositional Modeling of EOR Process in Stimulated Shale Oil Reservoirs by Cyclic Gas Injection. Paper SPE 169069 presented at SPE Improved Oil Recovery Symposium, 12-16 April, Tulsa, Oklahoma, USA.

Wilke, C.R. and Chang, P. 1955. Correlation of Diffusion Coefficients in Dilute Solution. AIChE J 1 (2): 264-270.

Yuan, W., Pan, Z., Li, X., Yang, Y., Zhao, C., Connell, L.D., Li, S., and He, J. 2013. Experimental study and modelling of methane adsorption and diffusion in shale. Fuel 117 (Part A):509-519.

Texas Tech University, Tao Wan, May 2015

Chapter 5

Evaluate the EOR Potential Of CO 2 Displacement In Shale Reservoirs Using Staggered Zipper Fractured Horizontal Wells

5.1. Abstract

Modified zipper frac technique is developed in a manner different from zipper frac in which the fractures are stimulated in a staggered pattern. The benefit of the modified zipper frac is that it will improve the contact area with the reservoir and increase the effective stimulated volume. Studies showed that enhancing fracture complexities in shale gas resources is critical to improving stimulation treatment and well production performance. CO 2 injection EOR process under the miscible flooding condition can significantly reduce oil viscosity. Oil viscosity reduction combined with the increased contact area by hydraulic fractures could be the dominant recovery mechanisms. Problems associated with gas injection in conventional well patterns such as early breakthrough and channeling through high permeability zones will not likely happen in nano-permeable shale oil or gas reservoirs, except that two wells are communicated by fractures.

In this chapter, we propose gas injection to enhance gas condensate recovery in a horizontal well pair, a gas injection well and a production well, which is stimulated in a staggered zipper frac pattern. The approach integrates the advantages of hydraulic zipper fracturing, horizontal wells and miscible gas flooding. Miscible gas flooding has shown the IOR potential in shale gas and oil reservoirs in this simulation work. We develop a compositional model to simulate complex interactions between the injected gas and reservoir fluids. Our simulation results of the Eagle Ford Shale indicate that the secondary recovery is increased after 5 years of primary recovery in a 100-ft fracture spacing staggered zipper fracture pattern. The investigation of CO 2 injection in a 87

Texas Tech University, Tao Wan, May 2015 staggered zipper fractured horizontal well pair provides an insight into the EOR performance in nano-shale reservoirs.

5.2. Introduction

In this chapter, staggered zipper frac technique was used to stimulate two adjacent horizontal wells to maximize the exposure of new reservoir rock. Although the laterals expose more of the nano-Darcy permeability shale rocks, the desired end result in hydraulic fracture stimulation is to maximize coverage around each lateral. The benefit of modified zipper frac technique from a geomechanical consideration is to increase stimulated reservoir volume and complexity development in successive fracturing stages because the net pressure created by the stimulation stage on the adjacent first well help divert the fracture direction (Belhadi, J. et al 2011; Rafiee et al. 2012).

Numerous numerical simulations (Mayerhofer et al. 2006; Warpinski et al. 2009) show that well productivity in shale oil and gas reservoirs primarily depends upon the size of fracture network and the stimulated reservoir volume (SRV), which provides highly conductive conduits to communicate the matrix with the wellbore. The natural fracture complexity is critical to the well production performance and it also provides an avenue for injected fluids to displace the oils (Cipolla et al. 2008; Wan et al 2013; Cipolla et al. 2011 and Moinfar et al. 2013).

Two horizontal wells are drilled through the reservoir and casing is set and cemented in place. The two horizontal wells are closely spaced, one above the other, forming a gas injector and producer pair like steam-assisted gravity drainage (SAGD) system. Hydraulic fracturing is introduced to improve the flow capacity of the reservoir and well productivity performance in shale reservoirs. The horizontal well fracture stimulation requires pumping fluids into a wellbore at high rate and pressure that is too high for the formation to accept without breaking. The perforating design plays a critical role in the stimulation treatment. The effectiveness of the perforating process in this two horizontal well pair cased-hole completion system depends on the perforating location. Before 88

Texas Tech University, Tao Wan, May 2015 selecting components for a perforating job, the first task is to understand how to avoid the hydraulic fracturing in the injection well directly communicating with producer. This requires perforating in a staggered pattern of effective entrance hole through the pipe and cement. Once the perforation and staging design were completed for one lateral, the perforation scheme for successive laterals was based on a staggered design with respect to adjacent wells (Belhadi, J. et al 2011). There is no perforating in the injector at the location where there are perforations in the producer. The primary objective is to prevent the created two wings of the injector from communicating with the producer. In this case, there is no connection between the propagated hydraulic fractures in the injector and the production horizontal well. The injected solvent would not break through to the producer by the created hydraulic fractures of the injector. The injected CO 2 from the hydraulic fractures of the injector effectively pushes the oil towards the hydraulic fractures of the production well and components in the injection gas dissolve in the oil phase as chemical equilibrium is established.

Figure 5.1A-The horizontal well pair perforated and stimulated in a staggered pattern

Texas Tech University, Tao Wan, May 2015

Figure 5.1B-Unit fracture SRV

Figure 5.2-The horizontal well pair stimulated in a staggered pattern 5.3. Model Description

The reservoir fluid composition description and phase behavior is from fifth SPE published data (Pope et al; Wu et al; Wang et al) which is aimed to model the behavior of a gas condensate fluid using six hydrocarbon components. Table 5.1 presents the pseudo- component description and input for Peng-Robinson equation of state calculations. The initial reservoir fluid compositions are also given in Table 5.1 which represent a gas condensate, as shown in Fig.5.3. The reservoir rock properties we used in this model are based on the published data in Eagle Ford shale as was previously used (Hsu and Nelson, 2002; Chaudhary et al., 2008; Bazanet al., 2010; Wan et al., 2013). The dimension of the 90

Texas Tech University, Tao Wan, May 2015 shale reservoir is 2000-ft long×1000-ft wide×200-ft thick, as shown in Fig.5.4. The producer is located at the beneath of reservoir domain and the injection well is placed at the upper layer. The producing horizontal well is stimulated with 10 transverse fractures each placed 200-ft apart as well as the injection well. The fracture spacing between the injector and producer is 100-ft. The initial reservoir pressure for this field is 6,425 psi. The producer is subject to minimum bottom-hole pressure constraint (BHP) of 500 psi and is produced for 1800 days (5 years) as the natural depletion. Our operational flowing bottom-hole pressure 500 psi is far below the dew-point pressure which results in the condensate dropout.

Table 5.1. Peng-Robinson EOS Fluid Description of Eagle Ford Condensate lumping

Components Initial Pc Tc (k) Acentric MW Shift Comp. (atm) Fac. CO 2 0.1618 72.8 304.2 0.225 44.01 0.094 C1 0.7098 45.4 190.6 0.008 16.043 0.099 C2-3 0.079 45.7 330.5 0.119 36.0 0.16964 C4-6 0.026 34.9 453.8 0.226 72.0 0.2986 C9 0.02 26.0 606.0 0.359 128.0 0.4904 C22+ 0.0034 14.9 869.7 0.788 310.0 1.0589

Dew point line

Figure 5.3-Phase diagram of gas condensate behavoir

Texas Tech University, Tao Wan, May 2015

Figure 5.4-Reservoir pressure changes during gas injection Table 5.2. Reservoir properties for the model input

Initial Reservoir Pressure 6425 psia o Reservoir Temperature 335F Saturation Pressure 4456 psia Rock Compressibility 5.0E-06 Porosity 6 % Permeability of shale 100 nano-Darcy Water Density 62.4 lb/cuft Hydraulic fracture conductivity 83.3 md-ft

CO 2 recovery mechanisms include CO 2 dissolution in the oil leads to an increase in its volume, oil viscosity reduction, vaporization of intermediate to heavy hydrocarbons and development of multi-contact miscibility (Whitson et al). The swelling test was simulated by varying proportions of injection gas mixed with original reservoir oil. The swelling test provides useful phase and volumetric data to investigate how a reservoir fluid reacts with gas injection. Fig.5.5 shows that the reservoir fluids remain in a single gas phase preceding to gas injection because the initial reservoir pressure at 6425 psi is higher than

saturation pressure. With increasing injection percentage of CO 2, the saturation pressure

decreases. The swelling factor increases with increasing CO 2 solubility. Fig.5.6 shows the simulated relative volumes in a CCE experiment at 335 F for the gas condensate mixture. Fig.5.7 shows the simulated liquid dropout curve in a CVD experiment at 335 F for the

Texas Tech University, Tao Wan, May 2015

gas condensate mixture. Liquid dropout curve expressed as V ro =V o/V s. Relative oil volume is defined as the volume of oil, V o at a given pressure, divided by the original saturation volume. The relative volume provides a measurement of the average reservoir oil saturation that will develop during depletion of a gas condensate reservoir (Whitson et al). The reservoir oil saturation can be calculated from V ro with So=(1-Sw)V ro.

Liquid dropout starts at the dew point pressure 4456 psi and continue to increase until the pressure reducing to 1500 psi when a maximum of condensate liquid reached.

5 4400 Psat 4.5

Swelling Factor 4 3900 3.5

3 3400 2.5

2900 2 Saturation Pressure, psia Saturation 1.5

2400 1 0 20 40 60 Injected CO2, mole composition %

Figure 5.5-Effect of solution gas on swelling of

Texas Tech University, Tao Wan, May 2015

4.5

3.5

2.5

2 Relative Volume Relative 1.5

0.5 1000 2000 3000 4000 5000 Pressure

Figure 5.6-Relative volume curve reservoir fluid by CO 2

90 1.4 80 1.2 70 1 60

0.8 50 40 0.6 30 prod,% Gas Liquid volume 0.4 Liquid dropout Liquid 20 Gas prod 0.2 10 0 0 1000 1500 2000 2500 3000 3500 4000 4500 5000 Pressure

Figure 5.7-The liquid dropout curve for constant-composition expansion experiment at 335F on the gas condensate mixture. 5.4. Results and Discussion

In this study, we examined the impact of grid refinement through numerical sensitivity 94

Texas Tech University, Tao Wan, May 2015 calculations. Figs.5.8 and 5.9 show the reservoir model using 10 x 21 coarse grids produces similar results to refined 22 x 21 grids. Fig.5.8 illustrates that using course grids for hydraulic fracture representation (as shown in Fig.5.1B) is getting close to the result as 21 grid-blocks. There is no large error caused by numerical dispersion. From these results we found that a 10 x 21 grid discretization around the fractures is able to properly model the rapidly varying pressure in and near the fractures. In Fig.5.8 and 5.9, the gas recovery ‘tail’ flatten out is due to the software results output issue which does not precisely reflect the simulation cases. On the basis of symmetry, Fig.5.1B simulates only a unit fracture SRV consisted of two ‘mirror’ half fractures stimulated reservoir volume. Fig.5.10 shows the production performance from the unit fracture controlled SRV produces the similar results with the simulated entire fractured horizontal well controlled SRV. Initially the reservoir operates in naturally depletion for 1800 days. Then, we start

CO 2 injection for 4000 days. The injection well is constraint to inject at a maximum injection pressure 7000 psi. Fig.5.11 shows the results for the different flowing bottom- hole pressure of the producer on the gas recovery factor. The BHP used in the base case simulation is 500 psi. The dew point pressure is 4456. The largest pressure drops occur near producing wells and hydraulic fractures. Condensate liquid saturation will build up near a well because of drawdown below the dewpoint pressure. The results in Fig.5.11 take into account the effect of condensate blockage, namely gas productivity index reduction due to build up of condensate in the near wellbore region. The benefit of large pressure drawdown in the producing well to the gas inflow performance overwhelms the gas recovery loss due to gas relative permeability reduction and condensate dropout. There is no visible distinction for the BHP of 500 psi and 1000 psi cases. It illustrates that the gas recovery loss due to condensate banking effect is equivalent to the benefit from pressure drawdown.

Texas Tech University, Tao Wan, May 2015

90 6X21X2 80 22X21X2 70 60 10X21X2 50 40

C1 Recovery C1 30 20 10 0 0 1000 2000 3000 4000 5000 6000 7000 Time, days

Figure 5.8-Effect of numerical dispersion on C 1 recovery vs. time

90 6X21X2 80 22X21X2 70 10X21X2 60 50 40 30 Gas Recovery, Gas % 20 10 0 0 1000 2000 3000 4000 5000 6000 7000 Time, days

Figure 5.9-Effect of numerical dispersion on gas R.F. vs. time

Texas Tech University, Tao Wan, May 2015

90 7000 80 6000 70 5000 60 50 4000

40 3000 Entire HW 30 Gas Recovery, Gas % Unit Fracture 2000 20 Average reservoir pressure Averagereservoir Entire HW PAVG 1000 10 Unit Fracture PAVG 0 0 0 2000 4000 6000 Time, days

Figure 5.10-Unit fracture controlled stimulated reservoir volume vs. entire horizontal well SRV results

90 BHP=500 80

70 BHP=1000

60 BHP=2500

Gas Recovery, Gas % 30

0 0 1000 2000 3000 4000 5000 6000 7000 Time, days

Figure 5.11-BHP impact on gas recovery performance

The non-Darcy flow effect should be considered for gas injection in fractured reservoirs.

Texas Tech University, Tao Wan, May 2015

The description of non-Darcy effect in hydraulic fractures is similar to the content presented in chapter 2.

Figure 5.12-Gas recovery factor-Darcy flow vs. gas recovery-non-Darcy flow

Fig.5.12 compares the condensate and gas recovery results of Darcy flow model with the non-Darcy flow model. The non-Darcy flow corrected model in 2-ft wide fracture conduits exhibit a little bit lower recovery than Darcy flow model. Rubin (2010) showed that the non-Darcy correction could be used to accurately model a pseudo fracture 1000 times wider than the actual fracture size in fractured shale gas reservoirs for horizontal wells. Fig.5.13 shows the impact of fracture spacing on the cumulative gas recovery factor. As shown in the graph, reducing fracture spacing from 200-ft to 100-ft would result in a more than two fold increase in cumulative gas recovery efficiency factor. The horizontal well production is the total sum of all fracture network segments which is strongly dependent on the fracture spacing. Increasing stimulation stages can be used to increase fracture-network size and SRV. Mayerhofer et al. (2010) illustrated that well production in shale reservoirs is related directly to reservoir volume stimulated during the fracture treatments which is in context with the fracture spacing.

Texas Tech University, Tao Wan, May 2015

Figure 5.13-Impact of hydraulic fracture spacing between injection well and production well on gas recovery and C9 recovery

Figure 5.14-Global mole fraction of CO 2 and C 1 changes during CO 2 flooding

Texas Tech University, Tao Wan, May 2015

Fig.5.14 shows global mole fraction of CO 2 and C 1 changes at time 0, after 1800 days of

primary depletion and after 4000 days of CO 2 flooding. The initial reservoir gas viscosity at 335 Fcalculated by Jossi-Stiel-Thodos (JST) correlation (Fong and Nghiem, 1980) is

0.0373 cp. The CO 2 viscosity at 335 F and at 6425 psi is 0.052 cp. When injected CO 2 contacts with reservoir fluids, the injected CO 2 effectively pushes C 1 and other components towards the hydraulic fractures of production well. Injected CO 2 channel,

bypassing most of in-situ fluid, is not observed. After 4000 days of continuous CO 2 injection from the top of the reservoir, most of the C 1 and other components have been uniformly swept to the hydraulic fracture of the producing well.

0.3

0.25

0.2 BHP=500, 1 10 2

0.15 BHP=500, 2 11 2

0.1

Condensate saturation Condensate 0.05

0 0 2000 4000 6000 Time, days

Figure 5.15-Condensate saturation distribution in or at vicinity of the fracture

100

Texas Tech University, Tao Wan, May 2015

1 Gas relative perm at 1 10 2 0.9 Gas relative perm at 2 11 2 0.8

0.7

0.6

Gas relative relative perm Gas 0.5

0.4

0.3 0 1000 2000 3000 4000 5000 6000 Time, days

Figure 5.16-Gas relative permeability on or vicinity of the fracture

It is well known that well deliverability would be compromised by accumulation of condensate near the wellbore. However, the presence of hydraulic fracture can mitigate the condensate banking effect due to large pressure drawdown. Pope et al (1998) and Fan et al (2005) developed a radial symmetric and single well model to demonstrate that condensate blockage could result in a rapid falloff in productivity. Narayanaswamy et al. (1999) concluded that when both non-Darcy and capillary number effects were considered, the capillary number effect can in some case overshadow the two-phase non- Darcy effects which results in smaller descrease in productivity indx (PI) than the case with capillary number effect only. In Pope’s model, they purposely set the initial reservoir pressure below the dewpoint that allows the condensate saturation goes up immediately. Fig.5.15 shows that as pressure declines rapidly in the fracture block (1, 10, 2), condensate saturation build-up very fast due to the large pressure drawdown in the fracture blocks. However, the condensate dropout near the fracture blocks (2, 11, 2) is not as high as the fracture blocks. Condensate blockage at the face of hydraulic fracture results in gas relative permeability reduction.

101

Texas Tech University, Tao Wan, May 2015

5.5. Conclusions

The objective of this chapter is to propose a new approach to implant the staggered zipper frac technique to stimulate a horizontal well pair. This requires perforation in a staggered pattern of effective entrance hole through the pipe and cement. Gas injection EOR process was performed in this staggered zipper fractured horizontal wells. Unit fracture controlled volume simulation was able to represent the entire horizontal well recovery on the basis of flow symmetry. The phase behavior of CO 2 mixed with reservoir fluids were investigated and proved to be efficient for enhanced oil recovery projects. From the compositional modeling of the gas injection in a fractured horizontal well pair, it is

demonstrated that injected CO 2 from hydraulic fractures of the injection well uniformly pushes the reservoir fluids to the producing well and its fractures, which increases the cumulative gas recovery significantly. The well production performance is strongly dependent on the fracture spacing. The condensate saturation builds up by the large pressure drawdown at the face of hydraulic fractures and gas relative permeability reduction results were presented. The main goal of this chapter is to examine the EOR potential by gas injection in a staggered zipper fractured horizontal well pair, which tends to produce promising results. Next chapter will focus on examining the numerical simulation of experimental data about cyclic gas injection in shale rocks. The role of diffusion in laboratory floods will be discussed. The impact of withdrawal rates and soak periods on liquid recovery in shale reservoirs will be presented.

102

Texas Tech University, Tao Wan, May 2015

5.6. References:

Belhadi, J., Ramakrishnan, H., Yuyan, R. 2011. Approach Optimizes Frac Treatments. American Oil and Gas Report.

Cipolla, C., Weng, X., Mack, M., Ganguly, U., Gu, H., Kreese, O., and Cohen, C. 2011. Integrating micro-seismic mapping and complex fracture modeling to characterize fracture complexity. SPE 140185 presented at the SPE Hydraulic Fracturing Technology Conference, The Woodlands, Texas. Doi: 10.2118/140185-MS.

Fan, L., Harris, B.W., Jamal, A., Kamath, J., Mott, R., Pope, G.A., Shandrygin, A., Whitson, C.H. 2005. Understanding Gas-Condensate Reservoirs. Oilfield Review, Vol.17.

Fong, D.K.S., and Nghiem, L.X. 1980. A Viscosity Model for Reservoir Fluids. Computer Modelling Group Research Report R7.02.

103

Texas Tech University, Tao Wan, May 2015

Hsu, S.-C., and Nelson, P.P. 2002. Characterization of Eagle Ford Shale. Engineering Geology, Volume 67, PP. 169-183.

Mayerhofer, M.J., Lolon, E.P., Warpinski, N.R., Cipolla, C.L., Walser,D., Rightmire, C.M. 2010. What Is Stimulated Reservoir Volume? SPEJ, Vol.25, No1. Pp 89-98. Doi: 10.2118/119890-PA.

Moinfar, A., Varavei, A., Sepehrnoori, K., Johns, R.T. 2013. Development of a Coupled Dual Continuum and Discrete Fracture Model for the Simulation of Unconventional Reservoirs. Paper SPE 163647 presented at the SPE Reservoir Simulation Symposium, The Woodlands, TX, USA. Doi: 10.2118/163647-MS.

Narayanaswamy, G., Pope, G.A., Sharma, M.M. 1999. Predicting Gas Condensate Well Productivity Using Capillary Number and Non-Darcy Effects. SPE paper 51910 presented at SPE Reservoir Simulation Symposium, Houston, Texas.

Pope, G.A., Wu, W., Narayanaswamy, G., Delshad, M., Sharma, M., Wang, P. 1998. Modeling Relative Permeability Effects in Gas-Condensate Reservoirs. Paper SPE 49266 presented at SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana. Doi: 10.2118/49266-MS.

Rafiee, M., Soliman, M.Y., Pirayesh, E. 2012. Hydraulic Fracturing Design and Optimization: A Modification to Zipper Frac. SPE 159786 presented at SPE Annual Technical Conference and Exhibition, 8-10 October, San Antonio, Texas, USA.

Rubin, B. 2010. Accurate Simulation of Non Darcy Flow in Stimulated Fractured Shale Reservoirs. Paper SPE 132093 presented at Western Regional Meeting, Anaheim,

104

Texas Tech University, Tao Wan, May 2015

California, USA. Doi: 10.2118/132093-MS.

Straight, D. 1989. Antelope Field: Preliminary Horizontal Drilling Evaluation, Bakken Formation. Simtech Consulting Services, Inc., Golden, CO.

Wan, T., Sheng, J.J., Soliman, M.Y. 2013. Evaluate EOR Potential in Fractured Shale Oil Reservoirs by Cyclic Gas Injection. Paper 168880 presented at the Unconventional Resources Technology Conference held in Denver, Colorado, USA.

Wang, P., G.A, Pope and K. Sepehrnoori. 1997. Development of Equations of State for Gas Condensates for Compositional Petroleum Reservoir Simulation. Industrial & Engineering Chemistry Research.

Warpinski, N.R., Mayerhofer, M.J., Vincent, M.C., Cipolla, C.L., and Lolon, E.P. 2009. Stimulating Unconventional Reservoirs: Maximizing Network Growth While Optimizing Fracture Conductivity. JCPT, Vol.48, No 10. Pp 39-51. Doi: 10.2118/114173-PA.

Whitson, Curtis H. & Brule, Michael R. 2000. Phase Behavior. SPE Monograph Series Vol. 20. Richardson, Texas.

Wu, Wei-Jr, P. Wang, M. Delshad, C. Wang, G.A. Pope and M. Sharma. 1998. Modeling Non-Equilibrium Mass Transfer Effects for a Gas Condensate Field. Paper SPE 39746 presented at the Asia Pacific Conference held in Kuala Lumpur, Malaysia.

105

Texas Tech University, Tao Wan, May 2015

Chapter 6 Numerical Simulation of the Experimental data in Liquid-Rich Shales by Cyclic Gas Injection

6.1. Introduction

The objective of this chapter is to integrate the numerical simulation approaches with the experimental data to examine the role of molecular diffusion, soak period and withdrawal rates in laboratory floods. All the experimental data presented in this chapter is conducted by Yu Yang (The experimental procedures are presented based on the discussion with Yu). The simulation model history matched the experimental data clearly showed the importance of diffusion in recovering oil in laboratory-scale flooding.

Monger and Coma (1988) conducted a laboratory-scale evaluation of the CO 2 huff-n-puff (also called cyclic gas injection) process with 32 experiments. The significance of soak periods on the ultimate oil recovery is identified. The inclusion of a soak period led to an incremental oil recovery than in absence of soak periods. The soak periods are characterized by injected gas in the fracture transporting to the shale matrix by diffusion process. Experimental results by Shayegi et al. (1996) revealed that the first cycle yielded highest oil recovery. A second cycle recovered additional incremental oil with some decline in process efficiency. Laboratory results by Monger et al. (1991) showed that more favorable recovery performance is achieved by CO 2 huff-n-puff injection at near- miscible conditions than miscible. Their experimental data and reported field test results

suggest that CO 2 retention and CO 2 utilization factor increases as pressure decreases. The conclusion was made under the presumption of same mass of CO 2 injection. The

recovery performance by a large CO 2 slug injection that develops miscibility with reservoir oil is more favorable than a small CO 2 slug.

Most of the available literature on performance of cyclic gas injection focused on reservoir conditions that have high permeability. Recent studies (Chen et al. 2014; 106

Texas Tech University, Tao Wan, May 2015

Gamadi, et al. 2013; Wan et al. 2013) showed that cyclic gas injection could be a viable method to improve the oil recovery in shale oil reservoirs. Wang et al. (2013) reported

experimental results of CO 2 huff-and-puff process operated in a 973 mm long composite core with an average permeability of 2.3 mD. Their experimental results showed that the first operation cycle contributes above half of the total oil production and additional oil produced from subsequent cycles is significantly decreased compared to previous cycles. Kurtoglu (2013) evaluated the feasibility of enhanced oil recovery by conventional gas injection and gas huff-n-puff in Bakken fields using simulation techniques. She used a

dual-porosity reservoir model to simulate the CO 2 huff-n-puff flooding performance in the Bakken field. Unfortunately, the diffusion effect was not included in their model because of numerical convergence issues. However, studies (Javadpour et al. 2007; Sakhaee-Pour and Bryant 2012; Ozkan et al. 2010) suggest that molecular diffusion is an important recovery mechanism in the mobilization and recovery of oil in very low permeability shale oil or gas reservoirs. Chen et al. (2014) investigated the effect of

reservoir heterogeneity on CO 2 huff-n-puff recovery process using (UT-COMP) simulation approaches. The drawback of current simulation work on enhanced oil recovery methods in unconventional reservoirs is that there is no sound laboratory or field data to support the model prediction. Gamadi et al. (2013) presented a series of experimental data of cyclic gas injection in Barnett, Mancos and Eagle Ford shale cores. They investigated the effect of injection pressure, soaking time and the number of injection cycles on oil recovery performance by N 2 huff-n-puff process.

The available literature provides limited information for the laboratory examination of the applicability of gas huff-n-puff in very low permeability shale rocks. Very limited field or laboratory data are available on the performance of cyclic gas injection in shale oil reservoirs. The purpose of this chapter is to extend earlier work performed by our research group (Gamadi et al., 2013). In this chapter, the relevant parameters that affect the performance of cyclic gas injection process are examined in detail. The principle recovery mechanisms in shale oil reservoirs were discussed. A series of experiments using immiscible cyclic nitrogen injection in Eagle Ford shale cores were conducted by 107

Texas Tech University, Tao Wan, May 2015

Yu. The gas huff-n-puff core floods were conducted with the same 2-in long, 1.5-in diameter Eagle Ford shale cores. This chapter interrelated numerical simulation approach with the laboratory data to analyze the significance of possible parameters that have on the performance of cyclic injection recovery process.

6.2. Material and Methods

Experimental Design

All the experiments were performed at the temperature of 95 ˚F in the oven. The cores used in current experiments are from the Eagle Ford and Barnett. Each core has the same dimension with 1.5-inch in diameter and 2-inch in length. A mineral oil Soltrol-130 was used to represent the reservoir fluids. Table 6.1 shows the basic properties of Soltrol-130 provided by the manufacturer.

Table 6.1. Properties of Soltrol-130 (Chevron-Phillips Chemical Company LP)

Properties Value Boiling Point 181˚C-209˚C Specific Gravity 0.762 @ 15.6˚C (47 lb/ft3) Viscosity 1.55 cSt @ 38˚C Vapor Pressure 1.5 mmHg @ 38˚C

The experimental procedures (Core saturation and cyclic gas injection processes)

1. The cores were dried prior to the saturation process. In order to know the weight of oil saturated into the shale cores, the dry weight of each core is required. Analyses of the microstructures in tight shales suggest that the organic porosity may play a dominant role for hydrocarbon accumulation and production (Handwerger, D.A. et al., 2012). The effect of nitrogen adsorption on the organic matter was not considered in simulation work. In this study, only oil and gas phases were considered in the displacement performance (no interstitial water). Since the pore size in tight shale formations is in the magnitude of microns in size, it is very difficult to saturate the oil in shale cores up to a desired oil saturation. To 108

Texas Tech University, Tao Wan, May 2015

inject oil into tight shales, it is necessary to vacuum the cores using a vacuum pump to remove the air and create a high pressure difference between the injection line and shale cores. The cores were placed in the desiccators and vacuumed for 2 days.

2. The shale cores were saturated with Soltrol-130 oil at an injection pressure of 1000 psi for 24 hours.

3. Weigh all the cores again and record their saturated weight to calculate the oil saturation of the shale rocks.

4. After the saturation processes, the saturated shale cores were placed in a

cylindrical core container. The N 2 injection pump is opened to pressurize the system up to 1000 psi (shown in Fig.6.1). The shale cores were exposed to the injected gas that allows the injected gas to penetrate into the shale matrix. The core container is placed in the oven which was heated to a constant temperature, 95 ˚F.

5. When the pressure showed on the pressure gauge reached to 1000 psi, shut down the injection pump. In order to allow injected gas to diffuse into oil phase in the shale matrix and to make the system reach equilibrium, the cores were soaked for 48 hours under a constant pressure (1000 psi). A uniform injection pressure in the system is maintained in the soak duration.

6. Once the soak period is finished, the pressure in the system is exposed to withdrawing down to the atmospheric pressure (14.7 psi). The pressure in the tight shale cores may drop at a slower rate than that in the cylinder. In the simulation model, the permeability of the fracture space between the core container and the shale cores is assumed to be much higher than the shale matrix. Pressure in the shale cores is required to be depleted to the atmospheric pressure before processing the next stage.

109

Texas Tech University, Tao Wan, May 2015

7. After the depletion process, the shale cores are weighed again to calculate the oil production during this gas huff-and-puff process.

8. Repeated above 4-7 steps at the same injection and soaking conditions for 8 cycles to observe oil recovery profile with increasing number of cycles.

Figure 6.1-Experimental setup and apparatus 6.3. Simulation model description

Table 6.2. Properties of C 15 at 95 ˚F

C15 at 95 F liquid Ideal Cp, BTU/lbmol-R 75.82 MW, g/mol 206.00 Density, lb/ft3 47.28 Phase volume, % 100.000

A numerical simulation model (CMG-GEM model) was developed whose validity was established by accurately simulating the cyclic gas injection results performed in the laboratory. In the simulation model, C 15 is used to represent the mineral Soltrol-130 oil. Chevron-Phillips chemical company stated that the Soltrol-130 solvent is a complex mixture of many different isoparaffinic molecules. Soltrol-130 isoparaffin solvent has the

carbon number distribution in the C 11 -C15 range. Based on the given product properties, 110

Texas Tech University, Tao Wan, May 2015

C15 that has similar properties (density and viscosity) as mineral oil is used to simulate the experiments. Although other critical properties have not been analyzed in detail yet, at this point C 15 is simply used to represent the oil properties. Table 6.2 shows the properties of C 15 at 95 ˚F.

Table 6.3. Reservoir and fluid properties used in this study

Parameters Value Unit Initial core pressure 15 Psi Soaking pressure 1000 Psi Reservoir temperature 95 ˚F Porosity of shale matrix 5% value -6 -1 Compressibility of shale 5*10 psi Shale matrix permeability 0.0005 mD Fracture permeability 1000 mD Oil density 47.3 lb/ft3

Annular fracture Sector 2

Shale matrix Sector 1

Figure 6.2-Base simulation model

A radial coordinate model and Computer Modeling Group (CMG-GEM) reservoir simulator are used to simulate the cyclic gas injection process in shale cores. The height of the core holder is approximately two times higher than the cylindrical shale sample. The dimension of the core holder (a diameter of 2.4-inch and 5.6-inch in height) is

111

Texas Tech University, Tao Wan, May 2015 designed with a larger diameter and height than the shale cores, because the fracture was represented by the surrounding annular volume. All the faces of the shale sample are open during gas injection, diffusion and production stages. The permeabilities in the annular fracture space between the shale core and inside core container are set as 1000 mD (fracture permeability sensitivity studies are performed as shown in Fig.6.6). The Eagle Ford shale matrix permeability is assumed to be 0.0005 mD. In order to history match the experimental data, the model input parameters including fracture permeability, shale matrix permeability, relative permeability and diffusivities of oil and gas phases are tuned within an acceptable range. A two-dimensional radial cross section (x-z) was used to form the simulation domain. Table 6.3 shows us the reservoir rock and fluid properties in the simulation model. The simulation model has 46 layers with the highest permeability layer at the top (fracture permeability = 1000 mD). The shale matrix that has the lowest permeability is located at the bottom. A variable grid-block size ranges from 0.048 to 0.072 inch was used in the x direction and with refined gridblocks located near the fracture. The simulation domain is separated into two sectors in order to control the output of reservoir data on a regional basis. The shale cores are set as sector 1 and the fracture region is set as sector 2. The production well and injection well are located at gridblock (1,1,1) at the top of sector 2. It is important to notice that the actual volumes of oil and gas produced by wells are different from the oil and gas recoveries from sector 1. Once the production well was shut down, there were no oil and gas recoveries from the well. However, oil and gas production from sector 1 is still possible because pressure drawdown in the shale cores may be slower than the fracture space which results in oil being driven out of the sector 1 by pressure gradients. Thus, the oil and gas recoveries by wells are different from the recoveries in sector 1. The injection well is constraint to inject at a maximum injection pressure at 1000 psi and a maximum surface gas rate at 1.2 MSCF/D. The injector will automatically change its mode of control whenever the existing control mode would violate one of these limits. The production well is subject to minimum bottom-hole pressure at 14.7 psi. This allows us to conveniently implement the displacement process of the cyclic gas injection in laboratory. Once we open the injection

112

Texas Tech University, Tao Wan, May 2015 valve, pressure in the annular fracture will reach to 1000 psi within 1-2 minutes. However, it requires some time for the gas in the fracture space to diffuse into tight shale matrix and achieve pressure and concentration equilibriums.

6.4. Results and Discussion

Fig.6.3 shows oil production response from 1th-5th cycle in the course of cyclic gas injection. Cyclic gas injection was quite effective, especially in the first few cycles.

Subsequent cycles of N 2 injection recovered additional incremental oil with some decline in process efficiency. Table 6.4 and 6.5 investigated the effect of depletion time on ultimate oil recovery. These two group comparison experiments probe the relationship between the rate of withdrawal and recovery performance of shale reservoirs. Table 6.4 and 6.5 presented shale oil recovery results with 0.05-hr and 40-hr of depletion time, respectively. The core pressure was systematically decreased to 15 psi over depletion time. The depletion time refers to the time required for the initial fracture pressure declined to atmospheric pressure in a huff-n-puff cycle. Fig.6.4 shows the comparison of shale oil recovery performance during 8 cycles of gas injection process using two different withdrawal rates. The experimental results presented in Fig.6.4 showed that more oil recovery will be obtained with an increase of withdrawal rates.

Table 6.4. Measured shale oil recovery factor and oil saturation with a deletion time of 0.05-hr (Yu, 2015)

Number of cycles 1 2 3 4 5 6 7 8

Cumulative R.F. 14.23% 21.37% 26.47% 31.50% 35.72% 39.51% 42.79% 45.45%

Oil Saturation 0.86 0.79 0.74 0.68 0.64 0.60 0.57 0.55

113

Texas Tech University, Tao Wan, May 2015

Table 6.5. Measured recovery factor and oil saturation with a deletion time of 40-hr (Yu, 2015)

Number of cycles 1 2 3 4 5 6 7 8

Cumulative R.F. 11.41% 16.95% 21.65% 25.82% 29.36% 32.62% 36.15% 39.66%

Oil Saturation 0.88 0.83 0.78 0.74 0.71 0.67 0.64 0.60

Figure 6.3-Shale oil production response in cyclic gas injection processes from 1th-5th cycle

114

Texas Tech University, Tao Wan, May 2015

50% Eagle Ford shales-Effect of Production Rate 45% 40% 35% 30% 25% Experimental results-0.05 hr 20% 15% 10% Experimental results -40 hrs 5% Cumulative Oil Oil Cumulative Recovery,% 0% 0 1 2 3 4 5 6 7 8 9 Number of Cycles

Figure 6.4-Effect of depletion time on CGI recovery performance

The effect of producing rate on ultimate oil recovery has been studied by many researchers. Results from conventional reservoirs demonstrated an increase of ultimate recovery was yielded at high producing rates purely from the perspectives of reservoir flow mechanics (Beveridge, 1974; Permyakov and Gadok, 1961). Previous studies (Morel et al.1993) showed that capillary end effect comes into play in lab-scale core flooding, with an accumulation of liquid near the fracture that delays liquid production. With a fast withdrawal rate, the capillary end effect is significantly reduced. However, other investigations showed that high production rates might lead to a reduction in the ultimate recovery compared to that with a slow withdrawal rate (Miller et al. 1949). Longer producing time periods will assist in recovering oil more effectively by the mechanisms of gravity drainage and cross-current. The recovery is affected by so many parameters that the variations in recovery are not constrained to any one factor such as rates.

In this simulation study, the impact of grid refinement on the recovery performance is examined by performing a series of numerical sensitivity calculations. Fig.6.5 illustrates that using 22x1x46 2D grid-blocks is producing similar results as more refined 50x1x46 grid-blocks. Grid refinement near the transition zone between highly conductive fracture

115

Texas Tech University, Tao Wan, May 2015 and shale matrix is needed to produce steady numerical solutions. Fig.6.5 shows that a 22x1x46 grid-block model is able to properly simulate the gas injection processes. There are some unknown parameters in this simulation model such as annular fracture permeability, relative permeability curves and shale matrix permeability. The simulation model that obtains a good agreement with the laboratory data is not unique since simulation output results are affected by these unknown input parameters. Fig.6.6 investigated the effect of fracture permeability on enhanced oil recovery performance. It is seen from Fig.6.6 that annular space permeability does not affect the cumulative oil recovery significantly provided that it is larger than 1000 mD.

45 50x1x46 40 22x1x46 35 10x1x46 22x1x90 30

Cumulative recovery,% oil 10

0 1 2 3 4 5 6 7 8 Number of cycles

Figure 6.5-Effect of grid block size on calculated oil R.F.

116

Texas Tech University, Tao Wan, May 2015

45 Kf = 10-md Kf = 100-md 40 Kf = 1000-md Kf = 100,000-md 35

20 Cumulative oilrecovery,% 15

5 1 2 3 4 5 6 7 8 Number of cycles

Figure 6.6-Effect of fracture permeability on cyclic gas injection performance

Injection P = 1000 psi, depletion time = 0.05-hr 50 Experimental data 45 Simulation (W ilke-Chang) Simulation (Sigmund) 40

15 Cumulativerecovery,% oil 10

0 1 2 3 4 5 6 7 8 Number of cycles

Figure 6.7-Comparison of simulation results with experimental data (0.05 hours) 117

Texas Tech University, Tao Wan, May 2015

Injection P = 1000 psi, depletion time = 40-hr 50 Simulation (Sigmund) 45 Simulation (W ilke-Chang) Experimental data 40

15 Cumulative oilrecovery,%

0 1 2 3 4 5 6 7 8 Number of cycles

Figure 6.8-Comparison of simulation results and experimental data (40 hours)

The diffusion mechanism is considered in computing the matrix and fracture molecular fluxes. The dispersive-convective flux through nanopores in shale oil reservoirs is modeled during gas injection process. Fig.6.7 shows the comparison of simulation results with the experimental data at 1000 psi injection pressure with 0.05 hrs of depletion time. The molecular diffusion coefficients could be calculated by Sigmund (1976) method. The effective diffusion coefficients for each component of the mixture could also be estimated by Wilke-Chang approach (Wilke and Chang, 1955). The oil recovery produced by Wilke-Chang correlation is closer to the experimental data than using Sigmund’s correlation. Fig.6.8 presented the history matching results of the measured cumulative oil recovery in eight cycles with 40-hour of depletion time. There is a slight deviation between the experimentally measured data and simulation results. This is primarily due to the fact that the micro-fractures are not considered in the simulation model, while natural fractures might exist in the shale cores that result in a higher oil recovery. Fig.6.9 shows a good agreement between the measured oil saturations and calculated oil saturations by simulation.

118

Texas Tech University, Tao Wan, May 2015

Injection P = 1000 psi, depletion time = 0.05-hr 1

0.9 Simulation (Wilke-Chang) Experimental data 0.8

0.7

0.6

0.5

0.4

0.3 Averageoilsaturation 0.2

0.1

0 1 2 3 4 5 6 7 8 Number of cycles

Figure 6.9-Comparison of simulated oil saturation and measured data

1 2 3

4 5 6

Figure 6.10-Pressure variations in one cycle of huff-n-puff process (40-hour depletion) Table 6.6. Operational schedules in a cycle

Time (days) 0-0.001 0.001-2 2-3.67 (40 hrs) 3.67-4

Operations Injector open Soak Producer open Further depletion Table 6.6 shows the operational schedule in a huff-n-puff cycle. Fig.6.10 displays the reservoir pressure variations during the nitrogen huff-n-puff processes at time 0, 0.001 119

Texas Tech University, Tao Wan, May 2015 days injection, 2 (soaking), 2.1 (after 0.1-day of production), 3 (after 1 day of production) and time 4 (depleted to atmospheric pressure), respectively. It is observed that the pressure buildup in the fracture space is much faster than that in the shale cores. When pressure in the fracture was increased to 1000 psi, the pressure in the inner shale cores still remained around 15 psi. The injected nitrogen front slowly propagates through the shale matrix. The region beyond the gas invaded region remains at initial reservoir pressure. As time progresses, nitrogen penetrates deep along the radial direction into the shale matrix with a progressive increase in the size of the treated shale zone. The soak period is designed to allow the injected gas in the fracture to diffuse into the shale matrix to achieve compositional and pressure equilibriums. In regarding to how long it takes to attain an equilibrium state in the system, it depends on a lot of factors, such as fracture permeability, shale cores permeability and fluids properties (diffusive velocity). During the gas production phase, the pressure drawdown in the fracture spacing decreases at a higher rate than the tight shale cores. The pressure in the fracture declines to the minimum allowable bottom-hole flowing pressure (15 psi) in a short period of time, while the pressure in the inner gridblocks of shale cores remained 1000 psi. In the production cycle, the average pressure in the fracture is lower than the shale matrix, which results in oil and nitrogen being displaced out by pressure difference. Owing to the high oil concentration and high pressure resided in the shale matrix, the transport mechanisms of oil component are a combination of diffusive and viscous displacement.

120

Texas Tech University, Tao Wan, May 2015

50 Simulation results 45 Experimental data 40

Cumulative recovery,% oil 10

0 0 5 10 15 20 25 30 Time, days

Figure 6.11-Comparison of simulation results and experimental data (Pi=1000 psi, depletion time = 40-hr)

15 Cumulativerecovery,% oil 10 With diffusion (0.05-hr depletion) With diffusion (40-hr depletion) 5 Without diffusion (0.05-hr depletion) Without diffusion (40-hr depletion) 0 1 2 3 4 5 6 7 8 Number of cycles

Figure 6.12-Effect of diffusion on ultimate oil recovery

Fig.6.11 shows comparisons between the simulation model output results and experimental results on time scale. Due to the fact that we only measured oil recovery at the end of each cycle, the oil recovery was increased instantaneously and sharply for

121

Texas Tech University, Tao Wan, May 2015 experimental data. Although there is some deviation at the first few cycles between simulation results that include molecular diffusion model with the experimental results, they are in good agreement at subsequent cycles. It is recognized that oil is continuously produced until the pressure in the shale matrix is in equilibrium with the fracture, which is clearly presented in the Fig.6.11. After having been subjected to multiple cycles of nitrogen injection, the cumulative oil recovery from the treated shale cores increased substantially. The importance of diffusion effect is observed in Fig.6.12. It is noteworthy that the impact of diffusion is more significant than production time on ultimate oil recovery. With diffusion included in the simulation model, the results exhibited roughly 10% higher oil recovery. Without considering the diffusion effect, the principle drive mechanism is viscous displacement. In absence of diffusion, the incremental oil recoveries at each cycle are increasing at roughly equal rates because applied pressure differentials in each cycle are the same. Simulation results showed that withdrawing the nitrogen in the annual fracture space at rapid rates produced a higher cumulative oil recovery than low rates, which is consistent with the experimental observations showed in Fig.6.4.

Injection P = 1000 psi, depletion time = 0.05-hr 60 Exp, Soak 1-hour Exp, Soak 3-hour 50 Exp, Soak 24-hour Exp, Soak 72-hour Simulation, Soak 1-hour 40 Simulation, Soak 24-hour

20 Cumulative oilrecovery,%

0 1 2 3 4 5 6 7 8 Number of cycles

Figure 6.13 -Effect of soak duration on production response

122

Texas Tech University, Tao Wan, May 2015

The effect of soak duration on ultimate oil response remains controversial. Monger and Coma (1988) pointed out that the cyclic process response was not sensitive to the soak duration in huff-n-puff field tests. However, Thomas and Monger (1991) showed that incremental oil recovery increased by extending the soak periods up to 4 weeks, provided that incremental oil not discounted for production lost during soak periods. Recent experimental observations reported by Yu and Sheng (2015) furthered investigation of the effect of soak duration on process response in Eagle-Ford shale reservoirs by cyclic nitrogen injection ( Fig.13). Based on their experimental data, we used a numerical simulation approach to enhance our understanding of cyclic gas recovery mechanisms. Fig.13 shows that a 24-hour soak time results in a more favorable oil recovery profile than 1-hour of soak time. However, extending the soak length from 24-hour to 72-hour has immaterial effect on the production response improvement. Production losses during the soak periods can be reduced by employing the optimal cycle length at current reservoir conditions. The comparison conducted between 24-hour and 72-hour soak showed that production response is less efficient by the long soak duration. It was speculated that experimental errors existed in the measurements of the 24-hour and 72- hour soak.

th th Figure 6.14-N2 mole fraction in the shale matrix from 1 -8 cycle (1-hour soak time) 123

Texas Tech University, Tao Wan, May 2015

The effect of soak periods during cyclic gas injection process in nano-permeable shales is presented in Fig.13. The importance of diffusion is underlined by the relation between process performance and soak duration. A soak period is dominated by diffusive transport of injected nitrogen in the fracture to the shale matrix. A soak period allows free gas to

contact with oil that is not near the core inlet. The N 2 mole fraction in the shale matrix from 1 th -8th cycle is shown in Fig.14. Nitrogen dissolved in the oil phase is increasing

progressively as the cycles proliferate. In the cases of cyclic CO 2 injection, laboratory core-flooding results by Monger and Coma (1988) suggested that hydrocarbon extraction into CO 2 rich phase also requires a soak period. Detailed discussion about the role of diffusion in a field-scale gas flooding in fractured shale oil reservoirs is referred to Wan and Sheng (2015). It is recognized that diffusion effect is an important recovery mechanism in mobilizing and recovering oil from nanopore systems. Gamadi et al. (2013)

presented the cyclic N 2 injection performance in shale cores from different fields at 1000- psi, 2000-psi and 3000-psi. Their experimental data (in their Fig.5) showed that gas injection pressure had a significant effect on incremental oil recovery. High injection pressures achieve a favorable cyclic response.

6.5. Conclusions

This study focused on evaluating the potential of oil recovery by cyclic nitrogen injection in shale oil reservoirs. Our understanding about huff-n-puff in shale plays can be furthered by comparing model predictions with experimental observations. Pressure depletion rate, soak duration and diffusion effect were discussed and how they influence process performance. Further studies are needed to better understand the mechanisms of oil recovery in shale oil reservoirs by cyclic gas injection.

1. Both experimental data and simulation results show the potential and applicability of cyclic gas injection to improve oil recovery in liquid-rich shales.

2. The first cycle yielded highest oil recovery by immiscible cyclic nitrogen injection in liquid-rich shales. Additional oil produced from subsequent cycles is 124

Texas Tech University, Tao Wan, May 2015

significantly decreased compared to previous cycles.

3. Results of the simulation and experiments showed that higher ultimate recovery was yielded at higher producing rates.

4. The simulation model that includes the molecular diffusion recovery mechanism produced well-matched results with experimental data. The diffusion effect should be considered in recovering oil from nano-permeable shale rocks by secondary cyclic injection method.

5. It requires some time for the gas in the fracture space to diffuse into the oil phase in tight shale cores and achieve pressure and concentration equilibriums with the shale cores. The significance of soak periods is observed in experimental results.

6.6. Acknowledgments

I would like to express my appreciation to ConocoPhillips. Their support is greatly appreciated. I also appreciate Yu give me the experimental data to complete this simulation work. I extend my appreciation to Computer Modeling Group (CMG) for providing the software for reservoir simulation.

6.7. References

Beveridge, S.B., Coats, K.H., Agrawal, R.K. and Modine, A.D., 1974. A Study of the Sensitivity of Oil Recovery to Production Rate. SPE 5129, Fall Meeting of the Society of Petroleum Engineers of AIME, 6-9 October, Houston, Texas.

Black, D.J., Aziz, N.I., Ren, T.X., 2011. Enhanced Gas Drainage from Undersaturated Coalbed Methane Reservoirs. No.50, The 3 rd Asia Pacific Coalbed Methane Symposium, May 3-6, Brisbane, Australia.

Chen, C., Balhoff, B., Mohanty, K. K., 2014. Effect of Reservoir Heterogeneity on

125

Texas Tech University, Tao Wan, May 2015

Primary Recovery and CO2 Huff-n-Puff Recovery in Shale-Oil Reservoirs. SPEREE 17 (03), 404-413. Doi: 10.2118/164553-MS.

Gamadi, T.D., Sheng, J.J., Soliman, M.Y., 2013. An Experimental Study of Cyclic Gas Injection to Improve Shale Oil Recovery. SPE 166334, SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 30 September–2 October.

Handwerger, D.A., Suarez-Rivera, R., Vaughn, K.I., Keller, J.F., 2012. Methods Improve Shale Core Analysis. The American Oil and Gas Reporter.

Javadpour, F., Fisher, D., and Unsworth, M., 2007. Nanoscale Gas Flow in Shale Gas Sediments. JCPT 46 (10), 55-61. Doi: 10.2118/ 10.2118/07-10-06.

Kurtoglu, B., 2013. Integrated reservoir characterization and modeling in support of enhanced oil recovery for Bakken. Ph.D. dissertation at petroleum engineering at the Colorado School of Mines.

Lofton, L.K. and Morse, R.A., 1978. The Effects Of Injection Pressure On Condensing Gas Drive Recovery. SPE 7472, SPE Annual Fall Technical Conference and Exhibition, 1-3 October, Houston, Texas.

Miller, C. C., Brownscombe, E. R., & Kieschnick Jr, W. F., 1949. A Calculation of the Effect of Production Rate upon Ultimate Recovery by Solution Gas Drive. JPT , 1(09), 235-247.

Monger, T.G., and Coma, J.M., 1988. A Laboratory and Field Evaluation of the CO2 Huff ‘n’ Puff Process for Light-Oil Recovery. SPE Reservoir Engineering , 3 (04), 1168-1176, SPE-15501-PA.

Monger, T.G., Ramos, J.C., Thomas, J., 1991. Light Oil Recovery from Cyclic CO2 Injection: Influence of Low Pressures, Impure CO2, and Reservoir Gas. SPE Reservoir Engineering , 01(6), 25-32. 126

Texas Tech University, Tao Wan, May 2015

Morel, D., Bourbiaux, B., Latil, M., Thiebot, B., 1993. Diffusion Effects in Gasflooded Light-Oil Fractured Reservoirs. SPE J . 1(02), 100-109.

Ozkan, E., Raghavan, R.S., Apaydin, O.G., 2010. Modeling of Fluid Transfer From Shale Matrix to Fracture Network. SPE 134830, SPE Annual Technical Conference and Exhibition, 19-22 September, Florence, Italy.

Permyakov, I.G. and Gadok, N.S., 1961. The Desirability of Exploitation of Oil Fields at High Rates of Oil Production. Neftyanoe Khoz , 39 (6), 33-68.

Sakhaee-Pour, A. and Bryant, S., 2012. Gas Permeability of Shale. SPE J 15 (4), 401-409. SPE-146944-PA.

Shayegi,S., Jin, Z., Schenewerk, P., Wolcott, J., 1996. Improved Cyclic Stimulation Using Gas Mixtures. SPE 36687, SPE Annual Technical Conference and Exhibition, 6-9 October, Denver, Colorado.

Sigmund, P.M., 1976. Prediction of Molecular Diffusion At Reservoir Conditions. Part 1- Measurement And Prediction of Binary Dense Gas Diffusion Coefficients. JCPT , pp 48- 57. Doi: 10.2118/76-02-05.

Wan, T., Sheng, J.J., Soliman, M.Y., 2013. Evaluate EOR Potential in Fractured Shale Oil Reservoirs by Cyclic Gas Injection. SPE 168880, Unconventional Resources Technology Conference held in Denver, Colorado, USA, 12-14 August.

Wang, Z., Ma, J., Gao, R., Zeng, F., Huang, C., Tontiwachwuthikul, P., Liang, Z., 2013. Optimizing Cyclic CO2 Injection for Low- permeability Oil Reservoirs through Experimental Study. SPE 167193, SPE Unconventional Resources Conference Canada, 5-7 November, Calgary, Alberta, Canada.

Wilke, C.R. and Chang, P., 1955. Correlation of Diffusion Coefficients in Dilute Solution. AIChE Journal 1 (2), 264-270.

127

Texas Tech University, Tao Wan, May 2015

Yu, Y., and Sheng, J.J. 2015. An Experimental Investigation of the Effect of Pressure Depletion Rate on Oil Recovery from Shale Cores by Cyclic N2 Injection. Paper URTeC 2144010 presented at the Unconventional Resources Technology Conference, San Antonio, Texas, USA, 20-22 July.

128

Texas Tech University, Tao Wan, May 2015

Chapter 7 Conclusions

The objective of this dissertation is to evaluate the response of cyclic gas injection as an enhanced-oil-recovery method in intensely naturally fractured and hydraulically fractured reservoirs. The impact of spacing of fracture network on oil recovery were investigated in chapter 3, the simulation results indicate that smaller fracture spacing is constructive to improving oil recovery in shale oil reservoirs. The primary decision in designing fracture treatments in shale oil reservoirs is to exploit fracture complexity.

In this dissertation, we also coupled the diffusion equation with a dual permeability model so that the enhanced-oil-recovery process by gas flooding can be properly simulated. There are difficulties in use a single porosity dual-continuum model to match the experimental results of gas flooding in the low permeability shale rocks, because it cannot properly capture the matrix-fracture mass transfer rates of densely distributed microfractures. The results produced by dual permeability model coupled with diffusion are in good agreement with the experimentally measured data in shale rocks. One deficiency in use of Kazemi’s shape factor in the simulation model is that assumptions of pseudo-steady state and instantly immersed fractures may break down in shale reservoirs. The shape factors are treated as adjustable parameters which are used to reproduce observed field or laboratory results. However, this approach does not necessarily capture the physics of matrix-fracture transfer flow behavior.

The simulation work in chapter 4 addressed the role of diffusion in the field scale flooding. The impact of matrix-fracture diffusion rate in the oil phase and matrix-matrix diffusion on oil recovery is significant in tight shale oil reservoirs. It is observed that the interstitial velocity of oil phase in the shale matrix is considerably lower than that of the conventional reservoirs. In very low permeability shale oil or gas reservoirs, the dominant recovery mechanism is by diffusion according to the analysis of Peclet number. One noticeable finding of gas injection in shale reservoirs is that without including the matrix- fracture diffusion in the oil phase, it results in lower oil recoveries. It is essential to model 129

Texas Tech University, Tao Wan, May 2015 the matrix-fracture diffusion rate in the oil phase and diffusion within the matrices for tight shale oil reservoirs. In conventional reservoirs, Coats’s (1989) model is able to produce good results because the transport is controlled by viscous flow.

The importance of diffusion effect on fractured shale gas reservoirs and shale oil reservoirs is observed. In the cases where the convective flux for a component is negligible because of low pressure drawdown, diffusion due to compositional differences between matrix and fracture tends to become the main recovery mechanism. From a reservoir management perspective, relying on diffusion enhanced oil recovery by decreasing injection rate is not the best way to exploit shale resource plays.

Chapter 5 proposed a new approach to implant the staggered zipper frac technique to stimulate a horizontal well pair. This requires perforating a staggered pattern of effective entrance hole through the pipe and cement. Gas injection EOR process was performed in this modified zipper fractured horizontal wells. The phase behavior of CO 2 mixed with reservoir fluids were investigated and proved to be efficient for EOR. From our compositional modeling of the gas injection in a fractured horizontal well pair, it is demonstrated that injected CO 2 from the injection well fracture uniformly pushes the reservoir fluids to the producing well and its fracture, which increases the cumulative gas recovery significantly. The well production performance is strongly dependent on the fracture spacing. The condensate saturation builds up by the large pressure drawdown at the face of hydraulic fractures and gas relative permeability reduction results were presented. The main goal of chapter 5 is to examine the EOR potential by gas injection in modified zipper fractured horizontal well pair, which tends to produce promising results.

Both experimental data and simulation results substantiated the applicability of cyclic gas injection to improve oil recovery in liquid-rich shales in chapter 6. The first cycle yielded highest oil recovery by immiscible cyclic nitrogen injection in liquid-rich shales. Additional oil produced from subsequent cycles is significantly decreased compared to previous cycles. The simulation model that includes the molecular diffusion recovery

130

Texas Tech University, Tao Wan, May 2015

mechanism produced well-matched results with experimental data. The diffusion effect should be considered in recovering oil from nano-permeable shale rocks by secondary cyclic injection method. It requires some time for the gas in the fracture space to diffuse into the oil phase in tight shale cores and achieve pressure and concentration equilibriums with the shale cores. The significance of soak periods is observed in experimental results.

The contribution of this study

Very limited work has been dedicated to studying possible avenues for enhanced oil recovery in shale oil reservoirs. Cyclic gas injection is proposed in this dissertation to improve shale oil recovery. It is discovered that the impact of matrix-fracture diffusion rate in the oil phase and matrix-matrix diffusion on oil recovery is significant in tight shale oil reservoirs. The recovery mechanisms in very low permeability shale oil and gas reservoirs were reexamined. The numerical simulation approaches with experimental data were interrelated to investigate significance of related parameters on enhanced oil recovery process. It is found that diffusion plays an important role in recovering oil from nano-permeable shale rocks by secondary cyclic injection method in laboratory-scale flooding.

Recommendations for future work

1. Affinity of organic-rich shales to carbon dioxide has been reported by many

researchers in the literature. The adsorption of CO 2 could be significant on the

finely-dispersed organics that have a large internal surface area. The higher CO 2 concentration in the fracture or adsorbed on organic-matter surface is critical to diffusive flow of injected CO2 into shale matrix. The adsorption effect of shale gas and injected gas on shale gas recovery is not included in this dissertation yet. In terms of adsorption effect, some literature states adsorbed gas comprises only a small fraction of produced gas. Sanaei et al. (2014) found that desorption has negligible effect on gas and condensate recoveries from Eagle Ford shales. It is

131

Texas Tech University, Tao Wan, May 2015

believed that desorption effect has to be considered in shales with high total organic content (TOC). Otherwise, an underestimation of ultimate gas recovery is expected. The significance of desorption effect on gas and oil recovery is not considered in this dissertation.

2. Literature points out that fracture growth in naturally fractured shale reservoirs can be very complex due to interactions between induced fractures with the pre- existing fracture network (Maxwell et al., 2002). Microseismic imaging in the Barnett shale has shown a high degree of complexity of the simulated fracture network. The stimulated fractures are interesting with pre-existing fracture network. The most difficult part is to describe the natural fractures in shale reservoirs, which are critical to well productivity by enhanced oil recovery process.

3. One common problem during CO 2 injection is that injected CO 2 could induce asphaltene precipitation that may cause formation plugging. The likelihood of asphaltene precipitation and deposition effect on oil recovery in secondary

miscible CO 2 flooding is not considered in this dissertation. In low permeability shale reservoirs, the further permeability reduction induced by wax could be an important consideration to oil recovery. Many factors can affect the asphaltene precipitation process such as the asphaltene to resin ratio, the nature of injection

gas, reservoir pressure and temperature and CO 2 concentration (Srivastava et al., 1999).

References

Johns, R.T., Pashupati, S., Subramanian, S.K. 2000. Effect of Gas Enrichment Above the MME on Oil Recovery in Enriched-Gas Floods. SPE J , 5(03):331-338. Doi: http://dx.doi.org/10.2118/65704-PA.

Kovscek, A.R., Tang, G.Q., Vega, B., 2008. Experimental Investigation of Oil Recovery 132

Texas Tech University, Tao Wan, May 2015 from Siliceous Shale by CO2 Injection. SPE 115679, SPE Annual Technical Conference and Exhibition held in Denver, Colorado, 19–22 September.

Kurtoglu, B., 2013. Integrated reservoir characterization and modeling in support of enhanced oil recovery for Bakken. Ph.D. dissertation at petroleum engineering at the Colorado School of Mines.

Landry, C., Eichhubl, P., Prodanovic, M., Tokan-Lawal, A. Matrix-Fracture Connectivity in Eagle Ford Shale. SPE paper 1922708 presented at SPE/AAPG/SEG Unconventional Resources Technology Conference, 25-27 August, Denver, Colorado, USA. DOI: http://dx.doi.org/10.15530/urtec-2014-1922708.

Maxwell, S.C., Urbancic, T.I., Steinsberger, N., Zinno, R. 2002. Microseismic Imaging of Hydraulic Fracture Complexity in the Barnett Shale. SPE 77440 presented at SPE Annual Technical Conference and Exhibition, 29 September-2 October, San Antonio, Texas. Doi: http://dx.doi.org/10.2118/77440-MS.

Pu, W. 2013. EOS Modeling and Reservoir Simulation study of Bakken Gas Injection Improved Oil Recovery in the Elm Coulee Field, Montana. A master thesis of Colorado School of Mines.

Sanaei, A., Jamili, A. 2014. Optimum Fracture Spacing in the Eagle Ford Gas Condensate Window. SPE 1922964 presented at SPE/AAPG/SEG Unconventional Resources Technology Conference, 25-27 August, Denver, Colorado, USA. Doi: http://dx.doi.org/10.15530/urtec-2014-1922964.

Shyeh-Yung, J.J. 1995. Effect of Injectant Composition and Pressure on Displacement of Oil by Enriched Hydrocarbon Gases. SPE Reservoir Engineering , 10 (02): 109-115. Doi: http://dx.doi.org/10.2118/28624-PA.

Srivastava, R.K., Huang, S.S., Dong, M. 1999. Asphaltene Deposition During CO2 Flooding. SPE Production & Facilities , 14 (04):235-245.

133

Texas Tech University, Tao Wan, May 2015

U.S. EIA/DOE. 2013. Technically Recoverable Shale Gas and Shale Oil Resources: An Assessment of 137 Shale Formations in 41 Countries Outside the United States.

Wan, T., Sheng, J.J., and Soliman, M.Y. 2013. The Evaluation of EOR Potential in Shale Oil Reservoirs by Cyclic Gas Injection. Proceedings of SPWLA 54th Annual Symposium in New Orleans, Louisiana, United States, June 22-26, 2013. SPWLA-D-12-00119.

Wan, T., Sheng, J.J., and Soliman, M.Y. 2013. Evaluation of EOR Potential in Fractured Shale Oil Reservoirs by Cyclic Gas Injection. Presented at the 2013 Unconventional Resources Technology Conference (URTeC) in Denver, Colorado, 12-14 August 2013. SPE-168880-MS.

Wan, T., Sheng, J.J., and Soliman, M.Y. 2013. December 2013 Exclusive Story. Study Evaluates EOR Potential In Naturally Fractured Shale Reservoirs. The American Oil & Gas Reporter.

Wan, T., Sheng, J.J., Meng, X., and Watson, M. 2014. Compositional Modeling of EOR Process in Stimulated Shale Oil Reservoirs by Cyclic Gas Injection. SPE 169069 presented at the Improved Oil Recovery Symposium, 12-16 April, Tulsa, Oklahoma, USA.

Wan, T., Sheng, J.J., and Watson, M. 2014. Compositional Modeling of the Diffusion Effect on EOR Process in Fractured Shale Oil Reservoirs by Gas Flooding. SPE Paper 1891403 presented at Unconventional Resources Technology Conference in Denver, Colorado, 25-27 August 2014.

Wan, T., Yu, Y., and Sheng, J.J. 2014. Comparative Study of Enhanced Oil Recovery Efficiency By CO2 Injection and CO2 Huff-n-Puff in Stimulated Shale Oil Reservoirs. 358937 Accepted for publication on AICHE annual meeting in Atlanta, GA, October, 2014.

Wan, T., Meng, X., and Sheng, J.J. 2014. Evaluation of the EOR Potential in Shale Oil

134

Texas Tech University, Tao Wan, May 2015

Reservoirs By CO2 Miscible Displacement Applied in Modified Zipper Fractured Horizontal Wells. 356067 presented at AICHE annual meeting in Atlanta, GA, October, 2014.

Wan, T., and Sheng, J.J. 2015. Compositional Modeling of the Diffusion Effect on EOR Process in Fractured Shale Oil Reservoirs by Gas Flooding . Journal of Canadian of Petroleum Technology , 54 (02):107-115. SPE-1891403-PA.

Wan, T. and Sheng, J.J. 2015. Evaluation of the EOR Potential in Hydraulically Fractured Shale Oil Reservoirs by Cyclic Gas Injection. Petroleum Science and Technology . DOI:10.1080/10916466.2015.1010041. In print.

Wan, T., Meng, X., and Sheng, J.J. 2015. Modified Zipper Frac Increases Stimulated Reservoir Volume. The American Oil &Gas Reporter , 58 (02): 83-87.

Zhang, P.Y., Huang, S., Sayegh, S., Zhou, X.L. 2004. Effect of CO2 Impurities on Gas- Injection EOR Processes. SPE 89477 presented at SPE/DOE Symposium on Improved Oil Recovery, 17-21 April, Tulsa, Oklahoma.

135