UNIVERSITY OF CALGARY. Application of Multi-Stage Fracturing of Horizontal Wells in SAGD Technology. Majid Saeedi A THESIS

Transcription

1 UNIVERSITY OF CALGARY Application of Multi-Stage Fracturing of Horizontal Wells in SAGD Technology by Majid Saeedi A THESIS SUBMITTED TO THE FACULTY OF GRADUATE STUDIES IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY GRADUATE PROGRAM IN CHEMICAL AND PETROLEUM ENGINEERING CALGARY, ALBERTA APRIL, 2016 Majid Saeedi 2016

2 Abstract SAGD process is used in reservoirs with continuous sand body and high vertical permeability. In western Canada, most of top tier SAGD candidates have already been developed. Many lower tier pools with interbedded shale layers (such as IHS) are available in which current SAGD practices may not deliver an efficient recovery method. The suggested method is to perform a multi-stage fracturing treatment on the injector to open up barriers with vertical hydraulic fractures before steaming. If successful, steam can pass through the barriers, effectively heat the otherwise isolated oil and recover it. This research has two major components. First, the flow mechanism and recovery response of a fractured SAGD well are discussed to show that the suggested process can improve production rate and recovery factor. The desired fracture properties and expected production rates are obtained by using analytical and numerical reservoir engineering techniques. Through numerical simulation, a matrix of sensitivity cases is made on key fracture parameters. It is shown that fractures with short half-length but high conductivity are required. Using the sensitivity results, an optimum range of the key fracturing parameters and general design criteria are determined using a new dimensionless fracture conductivity definition. In the second part, some geomechanical topics to be encountered are discussed, starting with a review of relevant hydraulic fracturing field trials in oil sands. Some interface crossing criteria are reviewed and the extended Renshaw and Pollard criterion is used to study fracture propagation through sand-shale interfaces. These results are then compared to results from a coupled reservoir-geomechanics-fracturing simulation software. It is shown that tensile hydraulic fractures can be created in the oil sands, and ii

3 fractures can cross sand-shale interfaces and be filled with desired concentration of proppant. It is also shown that small but high conductivity fractures (required for success of the proposed method) can be generated by using currently available proppants and fluid systems. The design considerations for various liner systems that can combine fracturing process and SAGD phase are discussed and a few new designs are proposed. Finally, a step by step roadmap for field implementation of this novel method is presented. iii

4 Acknowledgements I would like to express my deepest gratitude to my advisor, Dr. Antonin Settari for his guidance, patience, and his dedication in mentoring his students. I was very lucky and honored to have him as my mentor during my PhD studies, and I hope I can continue learning from him both personally and professionally for years to come. I am very grateful to the members of my supervisory committee Drs. Thomas Boone and Gordon Moore for their valuable advice and helpful suggestions. I would also like to thank Drs. Raj Mehta, Larry Lines and Alireza Nouri for serving in my candidacy exam and PhD defense committees. I would like to acknowledge and thank the former management and colleagues at Penn West Petroleum Ltd., Bruce Stewart, Don Wood, Bob Shepherd, Dave Middleton, Lucas Law and Kimberley Vincent for their support and encouragements throughout my studies. I would like to thank Dr. Mehran Pooladi-Darvish for all the insight he gave me throughout the years. I am also very grateful to my friends at the University of Calgary, Drs. Hassan Hassanzadeh, Hussein Sheikha, Amir Shahbazi and Rahman Khaledi for their moral and intellectual support. I d like to acknowledge the former members of Dr. Settari s research group at the University of Calgary, Drs. Mohammad Nassir, and Somayeh Goodarzi, Mr. Vivek Swami and Mr. Arshad Isalm for many interesting discussions and their support. Special thanks to Mr. Dale Walters of Taurus Reservoir Solutions for his valuable advice on different occasions. I d also thank Mr. Javed Iqbal of Trican Well Services for his help. iv

5 I d like to acknowledge the generous support of Computer Modelling Group, Taurus Reservoir Solutions, and Barree & Associates LLC for providing the software packages. I owe who I am and anything I have achieved to my parents, Dr. Mohammad R. Saidi and Sedigheh Moini who instilled the value of education in me. I will be indebted to them forever. I d like to thank my brothers Hamid and Saeed for our never ending arguments and also encouragements! Also, I am grateful for the support of my in-laws, Saeed and Pejman Tehranian and Malaktaj Vahedyar. I would also like to remember my late uncle, Dr. Ali M. Saidi who inspired me to pursue reservoir engineering and research. Lastly, I cannot find the right words to thank my partner in life, Sara Tehranian, who supported, encouraged and nurtured me from the very first step of my graduate work to the very last word of my thesis, literally! A heartfelt thank you for always being on my side Sara. v

6 Dedication Dedicated to my parents Mohammad R. Saidi and Sedigheh Moini, and to my wife Sara Tehranian. In memory of my uncle, Dr. Ali M. Saidi. vi

7 Table of Contents Abstract... ii Acknowledgements... iv Dedication... vi Table of Contents... vii List of Tables...x List of Figures and Illustrations... xi List of Symbols, Abbreviations and Nomenclature...xv CHAPTER 1: INTRODUCTION Background Statement of the problem Suggested novel approach: Application of multi-stage fracturing of horizontal wells in SAGD technology Scope of work...6 CHAPTER 2: APPLICATION OF MSFHW IN SAGD OPERATIONS - RESERVOIR ENGINEERING AND RECOVERY PROCESS Background Analytical modelling approach Drainage from fractures with finite conductivity Other limitations of the analytical modelling approach for SAGD with MSFHW Numerical simulation SAGD simulation cases Comparing numerical simulations with analytical solutions Discussion of relative permeability in the fracture blocks Summary...37 CHAPTER 3: SENSITIVITY ANALYSIS OF DESIGN VARIABLES, AND DEVELOPMENT OF AN EXTENDED DIMENSIONLESS FRACTURE CONDUCTIVITY TERM Introduction Well pairs start-up strategy Fracture orientation and spacing Fracture half-length Fracture permeability Dimensionless fracture conductivity Recovery factor vs. pore volume injected (RF-PVI) Summary...64 CHAPTER 4: SOME GEOMECHANICAL ASPECTS OF HYDRAULIC FRACTURING IN OIL SANDS Introduction...66 vii

8 4.2 Stress regimes in the Alberta oil sands and its implications in hydraulic fracturing Hydraulic fracturing trials in the Alberta oil sands In-situ oil sands projects with hydraulic fracturing in shallow depths Muskeg River Pilot [38] Steepbank Pilot [39] Mobil Oil Canada Athabasca Pilot [40] Kearl Lake Pilot [37] In-situ oil sands projects with hydraulic fracturing in greater depths Grigoire Lake pilots [40] Surmont CSS pilot [45] Pony Creek pilot fracture test [40] Numac fracture test in Surmont [30] Mechanical properties of oil sands and shales Hydraulic fracturing crossing sand-shale interfaces Examining the fracture propagation criterion in typical oil sand formations Pore pressure Stress gradients Interface properties Tensile strength Examining the fracture crossing criterion using a coupled numerical model Fracture initiation from a horizontal well bore Summary CHAPTER 5: FRACTURING DESIGN, AND STRESS STATE DURING SAGD OPERATIONS Introduction Proppant selection Proppant concentration Proppant pack conductivity loss Conductivity loss from the fracturing fluid Thermo-elasticity effect on propped fractures Proppant transport in fracturing jobs Design considerations in wellbore completion Summary CHAPTER 6: CONCLUSIONS AND REMENDATIONS Conclusions Recommendations for future research Recommendations for a field trial BIBLIOGRAPHY APPENDIX I: A SAMPLE CMG-STARS INPUT DATA FILE FOR THE RESERVOIR MODEL viii

9 APPENDIX II: A SAMPLE GEOSIM INPUT DATA FILE FOR THE FTRACTURING THROUGH INTERFACE MODEL ix

10 List of Tables Table 2.1 SAGD model properties Table 2.2 SAGD model additional properties for the interbedded cases Table 2.3 Estimated SAGD peak oil rates from analytical equations (all in m 3 /day) Table 2.4 Estimation of average block relative permeability for an extreme case difference Table 3.1 Summary of the SAGD sensitivity cases Table 4.1 Some typical shales and oil sands mechanical properties Table 4.2 Initial reservoir pressure of various oil sand reservoirs in Alberta Table 4.3 Examples of mechanical properties of various rock interfaces Table 4.4 Tensile strength of various rock samples Table 4.5 Extended Renshaw and Pollard criterion test, base case Table 4.6 Estimate of fracture initiation pressure for an openhole horizontal wellbore x

11 List of Figures and Illustrations Figure 1.1 Core photo showing an inclined heterolithic stratification (IHS) from Long Lake SAGD project [5]... 3 Figure 1.2 A sample core photo from upper D Valley facies, Tucker Lake SAGD project [6]... 3 Figure 1.3 Irregular development of steam chamber due to presence of impermeable shale layers... 4 Figure 2.1 A SAGD pair with multi-stage transverse hydraulic fractures on the injector in a marginal pay... 9 Figure 2.2 SAGD Schematic in: A) a continuous pay, B) a stratified pay with a vertical permeability channel [15] Figure 2.3 SAGD Schematic in a stratified pay with multiple vertical longitudinal fractures along the injector, 2D cross section along the wells Figure 2.4 SAGD Schematic in a stratified pay with multiple vertical transverse fractures, 2D cross section along the wells Figure 2.5 The illustration of a gas-liquid density drive in a counter current flow Figure 2.6 Grid fracture planes for the SAGD cases Figure 2.7 Oil production rates for SAGD Cases 1 to Figure 2.8 Steam injection rates for SAGD Cases 1 to Figure 2.9 Oil recovery factors for SAGD Cases 1 to Figure 2.10 Instantaneous Steam to Oil ratios (ISOR) for SAGD Cases 1 to Figure 2.11 Cumulative Steam to Oil ratios (CSOR) for SAGD Cases 1 to Figure 2.12 Instantaneous Steam to Oil ratios (ISOR) vs. Recovery Factor for SAGD Cases 1 to Figure 2.13 Snap shot of oil saturation for SAGD base case, SAGD in an interbedded formation and that with longitudinal fractures, time = 1,280 days Figure 2.14 Comparison of the analytical peak rate estimates and numerical results for the base cases xi

12 Figure 2.15 Comparison of the analytical peak rate estimates times 0.8 with the numerical results Figure 3.1 Sensitivity to circulation time (30, 90 and 150 days), 16 longitudinal fractures, x f =15 m Figure 3.2 Sensitivity to circulation time (30, 90 and 150 days), 16 transverse fractures, x f = 15 m Figure 3.3 Horizontal stress trajectories projected across the WCSB [24] Figure 3.4 Sensitivity of the fracture orientation for various fracture densities (4, 8 and 16), x f = 15 m Figure 3.5 Sensitivity to fracture half-length, longitudinal cases (x f = 5, 10, 15 and 20 m) Figure 3.6 Sensitivity to fracture half-length, Transverse cases (x f = 5, 10, 15 and 20 m) Figure 3.7 Sensitivity to fracture permeability, longitudinal cases (k f between 400 to 19,800 Darcy) Figure 3.8 Sensitivity to fracture permeability, transverse cases (k f = 50, 100, 200 and 400 Darcy) Figure 3.9 Plateau-Peak oil rate vs. fracture half-length for multiple C fdt s Figure 3.10 Plateau-Peak oil rate vs. fracture half-length for longitudinal cases with C fdt values Figure 3.11 Plateau-Peak oil rate vs. fracture half-length for transverse cases with C fdt values Figure 3.12 Plateau-Peak oil rate vs. fracture permeability for longitudinal cases with C fdt values Figure 3.13 Plateau-Peak oil rate vs. fracture permeability for transverse cases with C fdt values Figure 3.14 Plateau-peak rate ratio to clean sand SAGD rate as a function of total dimensionless fracture conductivity (C fdt ) Figure 3.15 Recovery factor vs. pore volume injected (RF-PVI); Longitudinal fracture cases for 2,000 day simulation runs xii

13 Figure 3.16 Recovery factor vs. pore volume injected (RF-PVI); Transverse fracture cases for 2,000 day simulation runs Figure 3.17 Recovery factor vs. pore volume injected (RF-PVI) for a subset of cases with low C fdt values Figure 3.18 Steam-oil ratio vs. recovery factor for longitudinal cases (including extended runs) Figure 3.19 Steam-oil ratio vs. recovery factor for transverse cases (including extended runs) Figure 4.1 Depiction of the vertical and minimum horizontal stresses in a) A tectonically relaxed area, b) In an area with a geological loading history such as glacial ice cap [33] Figure 4.2 Stress measurements in Alberta oil sands (the slope of the line is 24 kpa/m) [34] Figure 4.3 GLISP pilot well configuration and fracture geometry ( )[29] Figure 4.4 Pressure vs. time in a single well hydraulic fracture test followed by steam injection in Surmont area, Gulf Canada Resources, 1979 [47] Figure 4.5 Left: Location of the Numac pilot wells and surface uplift after the second grout fracture [30]; Right: Scaled model of the five cored wells from left to right E3, E2, F1, E4, and E1. Solid black lines represent grout filled fractures, dashed lines show hairline fractures, injection points are marked by arrows [49] Figure Potential outcomes of a propagating fracture to an interface [62] Figure 4.7 Simple criterion for crack re-initiation [69] Figure 4.8 Schematic explanation of the Renshaw and Pollard crossing criterion (2D) Figure 4.9 Horizontal stress estimation for a tectonically relaxed area using Equation ( 4.5) Figure 4.10 Interface crossing check, sensitivity to the friction coefficient for base case data (Table 4.5) Figure 4.11 Interface crossing check, sensitivity to the interface cohesion for base case data (Table 4.5) Figure 4.12 Interface crossing check, sensitivity to shale tensile strength for base case data (Table 4.5) xiii

14 Figure 4.13 Interface crossing check, sensitivity to σv'/σh' ratio for base case data (Table 4.5) Figure 4.14 Interface crossing check, sensitivity to horizontal stress contrast for base case data (Table 4.5) Figure 4.15 Vertical permeability multipliers as a function of effective stress (log scales) Figure 4.16 Snapshots of vertical permeability multiplier in the fracture plane; Figure 4.17 Snapshots of effective stress normal to the fracture plane; Figure 4.18 Effective stress normal to the fracture plane versus time for the first block in the shale layer; Top: Base case, no stress contrast, Bottom: Case with stress contrast at the interface Figure 4.19 Shear stress state planar view; Left: Base case (before and after fracture crossed the interface), Right: Case with stress contrast at the interface (fracture arrested below the interface) Figure 4.20 Expected fracture orientation from a horizontal wellbore; Left: Horizontal well parallel to maximum horizontal stress, Right: Horizontal well perpendicular to maximum horizontal stress Figure 5.1 Normal effective stress acting on a propped fracture Figure 5.2 Fracture permeability of various proppants at 10 kg/m 2 concentration vs. stress [86] Figure 5.3 Proppant embedment at elevated closure stress (not to scale) Figure 5.4 Schematic temperature profile around a propped fracture under steam injection Figure 5.5 Fracture job modelling standard plot Figure 5.6 Proppant concentration in the fracture as predicted by fracture model Figure 5.7 An openhole high temperature stimulation system with sand control (courtesy of Packers Plus Energy Services Inc.) xiv

15 List of Symbols, Abbreviations and Nomenclature Symbol Definition a L Volumetric thermal expansion coefficient of formation rock (1/ o C) BH Brinnel hardness index C f Fracture conductivity (D.m or m 3 ) C fd C fdt Dimensionless fracture conductivity Total dimensionless fracture conductivity C prop Proppant concentration in the fracture (kg/m 2 ) E Young s modulus (MPa) g Gravity (m/sec 2 ) h Formation pay thickness (m) k Formation absolute permeability (D or m 2 ) k f Fracture permeability (m 2 or D) k rg k rliq k ro k rw k rwf Gas relative permeability Liquid relative permeability Oil relative permeability Water relative permeability True fracture water relative permeability k v Effective vertical permeability (m 2, md or D) k block Average absolute permeability of fracture grid block (D or m 2 ) k rw block L m M n P br P p PVI q 1 Average water relative permeability of fracture grid block Horizontal well length (m) Viscosity exponent in the Butler equation Number of fracture stages Number of sand layers above the first shale bed Break down pressure of the formation (kpa or MPa) Pore pressure (kpa) Pore volume injected Plateau-peak SAGD rate in clean sand (m 3 /day) xv

16 q i RF RPC S g Various plateau-peak flow rates (m 3 /day) Recovery factor Renshaw and Pollard criterion Gas saturation S o S o S w SOR Oil saturation Displaceable oil saturation Water saturation Steam to oil ratio T Temperature ( o C) T o v w f Tensile strength of rock (kpa) Poisson s ratio Fracture width (m or cm) w fp Propped fracture width (cm or m) x f Fracture half length (m) x Fracture grid block width, matrix portion (m) α Biot constant α Thermal diffusivity coefficient (m/sec 2 ) β Poroelastic parameter μ Coefficient of friction equal to tan(φ) μ g Gas phase viscosity (cp) μ liq Liquid phase viscosity (cp) μ o Oil phase viscosity (cp) θ s Kinematic viscosity of oil at steam temperature (m/sec 2 ) ρ prop Proppant grain density (kg/m 3 ) ρ Density difference between liquid and gas phases (kg/m 3 ) ρ liq Acting density difference for liquid drainage (kg/m 3 ) ρ g Acting density difference for steam rise (kg/m 3 ) φ prop σ h Friction angle Porosity Proppant pack porosity Minimum horizontal stress (kpa) xvi

17 σ h σ h σ H σ v σ v Effective minimum horizontal stress (kpa) Change in horizontal stress (kpa) Maximum horizontal stress (kpa) Vertical stress (kpa) Effective vertical stress (kpa) σ xx (max) Maximum effective tensile stress in the upper interface material (kpa) σ xz σ xz (max) Shear stress at the interface (kpa) Maximum shear stress at the interface before slippage (kpa) σ zz Effective normal (vertical) stress acting on the interface (kpa) τ o Cohesion of the sand-shale interface (kpa) xvii

18 CHAPTER 1: INTRODUCTION 1.1 Background In-situ thermal recovery methods have been widely used for heavy oil and bitumen reservoirs exploitation. The first commercial steam injection project was started in the 1960s in California s San Joaquin Valley [1]. First steam injection trials in Canada started in Cold Lake and Peace River areas [2]. Before the 1990 s, all in-situ steam operations were based on vertical well development. The advent of horizontal drilling technology in the 1990 s helped unlock a vast portion of Canadian heavy oil and bitumen resources. In particular, by 2014 steam assisted gravity drainage (SAGD) technology and its variations, as well cyclic steam stimulation (CSS) technology has become the basis for the production of 1.2 million barrels per day in Canada [3]. Many high quality SAGD candidate pools with extensive and clean sand body and good vertical permeability have already been identified. Some are already developed (pools such as Christina Lake and McKay River) and some are in the planning and development stages. There are also many marginal pools identified either in the vicinity of high quality deposits or independently in which, SADG process at its current state may not be efficient. Developing lower quality oil sand assets will require trying different new technologies, and the proposed research is focused on one such promising idea. 1

19 1.2 Statement of the problem In the SAGD operation and its variations, two parallel horizontal wells are drilled with about 5 m depth offset. After the initial steam circulation warm-up phase, steam is injected in the top well and oil and water are withdrawn from the bottom well. The success of a SAGD operation relies on the vertical development of the steam chamber and effective drainage of the hot water-oil emulsion to the production well which is dictated by the geology of the pool. To be more specific, the key parameter for success is the existence of clean sand without major shale barriers and with a fairly good vertical permeability along the wellbores that extends to the top of the formation. As mentioned above, many of the SAGD sweet spots have already been either developed or are under planning and development.. However, there are many lower quality oil sand deposits with non-ideal vertical permeability or lack of continuous sand body, which may not deliver commercially viable SAGD operations using the classical well pair configuration. In geological terms, two similar geological facies can be highlighted in Athabasca and in Cold Lake oil sands regions. In the Athabasca oil sands region, a particular form of interbedded pays known as inclined heterolithic stratification (IHS) is characterized by interbedded sandstone and siltstone/shale units deposited with a slight inclination as a result of episodic depositional energy fluctuations. The siltstone or shale layers may act as impermeable lenses within associated reservoir deposits [4]. Figure 1.1 shows an example of IHS pay from Long Lake SAGD project in NE Alberta. Within the lower cretaceous formations in the Athabasca oil sands region, IHS beds are believed to be tidal point bar deposits. 2

20 Figure 1.1 Core photo showing an inclined heterolithic stratification (IHS) from Long Lake SAGD project [5] In the Cold Lake oil sands region, the deposits are typically tidal and fluvial in nature with areas containing common to abundant sand/shale interbeds. Tucker Lake SAGD project could contain an example of a marginal SAGD pool with a sand/shale interbedded pay zone. The Tucker Lake SAGD project had an expected output of 30,000 bopd but only achieved 7,000 bod five years after the start-up [6]. The lower than expected production rate can be partially attributed to low effective vertical permeability of the upper D Valley facies which contain frequent thin shale stringers as shown in Figure 1.2. Figure 1.2 A sample core photo from upper D Valley facies, Tucker Lake SAGD project [6] 3

21 As illustrated in Figure 1.3, the presence of impermeable shale stringers within the pay zone can impede the development of steam chamber which in turn results in lower production rates and recovery factor from the targeted drainage area. Figure 1.3 Irregular development of steam chamber due to presence of impermeable shale layers In summary, significant problems still exist in the SAGD applications in lower quality assets. The focus of this work is the investigation of a novel method to overcome the adverse effects of low effective vertical permeability due to presence of finer grain intervals or shale laminae in IHS pays. 1.3 Suggested novel approach: Application of multi-stage fracturing of horizontal wells in SAGD technology The Multi-Stage Fracturing of Horizontal Wells (abbreviated as MSFHW) completion technology was first used in tight and shale gas plays. In shale gas context with nano to micro Darcy permeability, vertical wells or un-stimulated horizontal wells would 4

22 not provide sufficient productivity for commercial development. However, MSFHW has revolutionized the gas industry (and more recently light crude industry) by making development of the otherwise inaccessible hydrocarbon resource commercially viable. Nowadays, different variations of the field execution of this technology are available in the industry. If the MSFHW technology can be combined with in-situ steam based recovery methods such as SAGD, some of the problems with development of lower quality assets such as IHS pays can be addressed. The idea of enhancing vertical conductivity of interbedded reservoirs has been proposed before and can be traced back in the literature both for conventional and unconventional reservoirs. Some have proposed using undulating horizontal or deviated wells by crisscrossing shale and tight intervals [7-9] or drilling horizontal well with upward sidetracks (like a half fish bone) [10]. Others have suggested the use of hydraulic fracturing in horizontal wells to increase the vertical permeability immediately above the wells where it is needed the most [11, 12]. All the references noted indicate that the general idea of enhancing vertical permeability is not new. However, what appears to be missing is an integrated process evaluation and design, from reservoir recovery and geomechanics to completion design, and detailed understanding of the mechanisms of flow and the possible benefits to recovery. Other application of MSFHW in SAGD could be in bitumen bearing Carbonate formations such as Grossmont or Debolt in northern Alberta. In the absence of a natural fracture network, hydraulic fractures can break the rock and access the vugs and tunnels. 5

23 This technology is likely most applicable in areas immediately adjacent to the high quality deposits after they have been depleted. At that point, the operators would be looking for to the next best tranche of accessible reservoir to utilize the existing facilities and infrastructures. The application of MSFHW technology in in-situ steam operations will present new technical challenges. The evaluation of the merit of this idea and the anticipated issues as much as they can be foreseen is the focus of this work. 1.4 Scope of work The research presented in this work has two major parts. The first part examines the problem from the flow mechanics and recovery point of view. This work shows that the process has a potential to improve recovery, and therefore it is necessary to examine the problem also from the geomechanics and fracturing point of view. We will begin in Chapter 2 with the discussion of the reservoir engineering and recovery process basics of MSFHW in a SAGD operation using analytical equations and numerical simulation runs. The underlying assumption is that inducing vertical hydraulic fractures with the desired properties is possible (this assumption will be then examined later in Chapters 4 and 5). We will then move on in Chapter 3 to evaluation of some of the parameters that affect the performance of the stimulated SAGD wells through a matrix of sensitivity cases. The sensitivity results will be used in development of a concept of dimensionless fracture conductivity to be used as a guide in first pass design of this process. Some overall 6

24 conclusions about the recovery method will be made, which support the feasibility of the process from the reservoir flow mechanics point of view. The second part of this work covers some topics on geomechanical considerations and fracturing design, in order to establish if the fracturing with the desired geometry and conductivity can be achieved. The geomechanics of fracturing in interbedded sands is discussed in Chapter 4. The problem is studied using recent theories of fractures crossing interfaces as well as by numerical modelling by a coupled flow and geomechanical model. Chapter 5 then deals with some practical aspects of the fracturing and stress related issues that are likely to be encountered. Finally, the conclusions and recommendations for field implementation of this method are discussed in Chapter 6. 7

25 CHAPTER 2: APPLICATION OF MSFHW IN SAGD OPERATIONS - RESERVOIR ENGINEERING AND RECOVERY PROCESS 2.1 Background The success of a SAGD operation relies on the vertical development of the steam chamber and effective drainage of hot water-oil emulsion to the production well. This is strongly dependent on the existence of clean sand with a fairly good vertical permeability along the wellbores that extends to the top of the formation. In many existing SAGD operations, the geological setting is such that the vertical permeability is significantly lower in some intervals along the well. When this type of a well undergoes steam injection, after a period of time the steam chamber develops within the high vertical permeability intervals but it may stop at a vertical permeability barrier if it is sufficiently continuous. The portion of the reservoir above the vertical permeability barrier could remain cold and hence not contribute to the production. However, if such barriers can be targeted and broken with vertical hydraulic fracturing treatments before steaming, it is very likely that the steam can pass through the barriers, effectively heat the otherwise bypassed oil, and eventually drain it. The challenge with the targeted fracturing is that one should know the location of such barriers beforehand. The other approach could be using the multi-stage fracturing of the injector without the exact knowledge of the location of barriers, and open them up at a number of locations. Such technology is available and widely used in shale reservoirs but has not been applied in SAGD to our 8

26 knowledge. Figure 2.1 shows a schematic of a SAGD pair with multi-stage fractures in a marginal pay such as IHS sands. Cap Rock Hydraulic Fractures Shale Stringers Reservoir Pay Figure 2.1 A SAGD pair with multi-stage transverse hydraulic fractures on the injector in a marginal pay The focus in this chapter is to evaluate and determine the reservoir engineering and recovery process basics of MSFHW in a SAGD operation with the underlying assumption that it is possible to induce the hydraulic fractures with the desired properties. This assumption will be then debated in the following chapters. 9

27 2.2 Analytical modelling approach The idea of SAGD was first proposed by Butler and his colleagues. They analytically formulated the SAGD process and its variations in various articles [13, 14]. Their formulation describes the rise of steam chamber and drainage of hot bitumen and water emulsion in a continuous sand body as shown in Figure 2.2-A. Equation ( 2.1) was developed by Butler et al. for a SAGD well pair plateau-peak production rate with no-flow boundary. In the context of SAGD process, plateau-peak rate is the flat portion of production profile when steam chamber height is fully developed and expanding sideways. q 1 = 2L 1.5k v gα S o h mθ s ( 2.1) where L is the well length, k v is the effective vertical permeability, g is the gravity, α is the thermal diffusivity coefficient, is the rock porosity, S o is the displaceable oil saturation, h is the formation pay thickness, θ s is the kinematic viscosity of oil and m is the viscosity exponent. Boone et al. [15] showed that creating a vertical channel along the length of the horizontal well in a stratified pay can, in theory, increase the production rate. They assumed that a stratified pay with a vertical channel or fracture can be approximated by three SAGD well pairs with one third of the pay per well pair (Figure 2.2-B). For example, the plateau-peak rate for one of the reservoir compartments shown in Figure 2.2-B can be written as the following: q 1 3 = 2L 1.5k v gα S o (h 3 ) mθ s = 1 3 q 1 10

28 calculated as: By adding the rates for all three reservoir compartments, the total rate can be q 2 = 3 2L 1.5k v gα S o (h 3) mθ s = 3 q 1 ( 2.2) The buried assumptions in this formulation are that it is possible to create a vertical channel or fracture like the one shown in Figure 2.2-B, and that the fracture has sufficient fluid conductivity both for steam and bitumen. This formulation suggests that MSFHW can theoretically increase the peak SAGD production rate of a compartmentalized reservoir beyond that of a clean reservoir, which is overly optimistic. The assumptions essentially make the fracture channel equivalent to having multiple pairs of SAGD wells as shown on the right part of Figure 2.2-B. Figure 2.2 SAGD Schematic in: A) a continuous pay, B) a stratified pay with a vertical permeability channel [15] The validity of the assumptions and their limits will be discussed later. However, it should be noted that this formulation can only be considered where using the original 11

29 Butler s equation for SAGD can be justified. For example, if the compartments are too thin, neither Butler s equation, or the ones discussed here are not considered to be valid. The above example showcases that it would be more beneficial to focus in increasing the permeability right above the wells as opposed to other methods for enhancing reservoir access areally. It also shows the power of analytical formulations for evaluating the performance potential of different recovery processes. A similar analytical approach is used to estimate the peak SAGD rate in a compartmentalized reservoir with multiple vertical fractures. The cases are summarized in the following section. The main assumption in these estimates is infinite conductivity of the fracture, that is, the fracture can pass any amount of steam from the lower compartments upward and fluids draining from the upper compartments downward. We will now derive analytical estimates for various extensions of the above idea. 1. Base case: Peak SAGD production rate in continuous pay This is the general SAGD rate derived by Butler et al. [13, 14] for a SAGD well pair peak production rate with no-flow boundary as shown in Figure 2.2-A. It has been already given by Equation ( 2.1): q 1 = 2L 1.5k v gα S o h mθ s ( 2.1) 12

30 2. Stratified pay with one longitudinal fracture along the injector and equal pay thickness intervals It is assumed that the vertical connection between the intervals exists along the entire length of the well pair L. Such connection could be thought of as a vertical fracture along the entire wellbore length. Such fracture (in the plane of the well) is termed a longitudinal fracture, whereas a fracture perpendicular to the well is a transverse fracture. This case is an extension of Boone et al. [15] formulation in Equation ( 2.2) for a general case with n equal pay thickness intervals where the sum of the pay interval thicknesses is equal to the pay in the base case, h. The following general formulation is obtained: q 3 = n 2L 1.5k v gα S o (h n) mθ s = n q 1 ( 2.3) 3. Stratified pay with a continuous longitudinal fracture along the injector and various pay thickness intervals This case is similar to the previous one except that the pay thickness in each interval can be different. The general production rate equation can be deduced to be: n q 4 = 2L 1.5k v gα S o h i mθ s 1, n > 1 ( 2.4) This equation is especially useful for cases where the thickness of shale intervals is not negligible when compared to the gross pay like some inclined heterolithic stratification (IHS) reservoirs. 13

31 4. Stratified pay with multiple longitudinal fractures along the injector The connection between the pay intervals will be created by hydraulic fracture stages and not form one continuous fracture along the horizontal well. It is assumed that all fractures have similar half lengths, x f, with n equal pay thickness intervals as illustrated in Figure 2.3. The drainage rate can be calculated by adding the small individual SAGD chamber drainage rate of each fracture in the upper reservoir compartments. The SAGD drainage rate for the base reservoir compartment that includes the wells is calculated using the original Butler s equation without any effect from fractures. The plateau-peak production rate for a SAGD well pair in a stratified pay with multiple longitudinal fractures along the horizontal well can be estimated by Equation ( 2.5) where M is the number of fracture stages. Equation ( 2.5) also shows relationship with the original Butler plateau-peak rate in Equation ( 2.1). The major difference between this case and case 3 is that the pay interval containing the well pair (the lowest interval in Figure 2.2) is drained similarly while the other intervals are only drained in the fractured intervals. In this formulation, the drainage of the upper intervals beyond the fractures length is neglected. 14

32 15 Figure 2.3 SAGD Schematic in a stratified pay with multiple vertical longitudinal fractures along the injector, 2D cross section along the wells s o v f s o v m h S g k L M x n n L n m h S g k L q q L M x n n n q f, 1 n, 1. L x M f ( 2.5) Equation ( 2.5) is poroposed for intervals of equal thickness of (h n ). The following Equation ( 2.6) is a more general case where the net pay intervals have different thicknesses and the shale layer thicknesses are not negligible. n s i o v f s o v m h S g k L M x m h S g k L q ( 2.6)

33 5. Stratified pay with multiple transverse fractures The direction of principal stresses can strongly affect the orientation of hydraulic fractures. In certain situations, the fractures could grow perpendicular to the direction of horizontal wells (which typically is desirable for stimulation of shale gas wells). A cross section of such a case is illustrated in Figure 2.4. It is assumed that all fractures have similar half lengths,x f, with n equal net pay intervals. The drainage rate can be calculated by adding the small individual SAGD chamber drainage rate of each fracture in the upper reservoir compartments. The SAGD drainage rate for the base reservoir compartment that includes the wells is calculated using the original Butler s equation without any effect from fractures. It is assumed that the SAGD chambers in the higher intervals develop in the direction perpendicular to the well pair, and the spacing of the fractures is sufficient such that the chambers do not interfere. Then the plateau-peak production rate for a SAGD well pair in a stratified pay with multiple transverse fractures can be estimated by Equation ( 2.7) where M is the number of fracture stages. Equation ( 2.7) also illustrates relationship with the original Butler peak rate in Equation ( 2.1). From the illustration and the analytical equation, one could see that the pay interval containing the well pair is drained all along, while the other intervals are only drained within the fracture length. Furthermore, fracture in each interval is assumed to behave like a well pair with length of x f perpendicular to the main wells. It is assumed that the drainage of the upper intervals beyond the fractures length can be neglected. 16

34 q 7 2L 1.5k g S h m n 2x M( n 1) 1. 5k g S h v M n 1 1 q x f q1, n 1 n n L o s f v o m n s ( 2.7) Equation ( 2.8) is a more general case where the pay intervals have different thicknesses and the shale layer thicknesses are not negligible. q n L 1.5k v g S oh1 m s 2x f M 1. kv g S ohi m s ( 2.8) Figure 2.4 SAGD Schematic in a stratified pay with multiple vertical transverse fractures, 2D cross section along the wells 2.3 Drainage from fractures with finite conductivity As stated, one of the assumptions in the above analytical SAGD rate estimates was the infinite fracture conductivity. To examine this assumption, the ability of the fracture to conduct liquids under a typical hydrostatic head in SAGD process can be calculated. The flow rate estimates in equations ( 2.5) to ( 2.8) have two terms. The first term is the portion 17

35 of the production coming from the lower part of the reservoir pay that is immediately adjacent to the well pair. The second term is the contribution of the upper parts of the reservoir pay that are separated by the shale layers, and can only flow in through the fractures. If the second term s contribution is less than the total flow that the fractures permit, the infinite conductivity formulations would be valid. Otherwise, the contribution of the pay above the first shale break to the SAGD rate would be limited to the total flow capacity of the fractures. The flow pattern in the steam chamber within the sand intervals has a three dimensional geometry. However, where the fracture channel crosses the shale barrier, the pattern can be pictured by a counter current flow in a vertical fracture slab, in which steam flows upward and oil and steam condensate flow downward. The main driving force in SAGD is the density difference between the gaseous (steam) and liquid (oil and condensed water) phases. In a gravity driven counter current flow, the total density difference is shared by the gaseous (rising) and liquid (draining) phases as illustrated in Figure 2.5. Figure 2.5 The illustration of a gas-liquid density drive in a counter current flow In counter current flow within the fracture at near steady state flow condition (when the steam chamber pressure change is minimal), the volumetric rate of steam rise should be 18

36 equal to the rate of liquid drainage. This equality can be expanded by writing the Darcy Law for a vertical slab: q steam rise = q liqud drainage k f. k rg. w f. 2x f. ρ g. g μ g = k f. k rliq. w f. 2x f. ρ liq. g μ liq k rg. ρ g μ g = k rliq. ρ liq μ liq ( 2.9) where k f and w f are fracture channel permeability and width, k rg and k rliq are the relative permeabilities of the gaseous and liquid phases, μ g and μ liq are the viscosities of the gaseous and liquid phases at reservoir conditions and ρ g and ρ liq are the acting density difference for gaseous and liquid phases as illustrated in Figure 2.5. Equation ( 2.9) can be rearranged for dimensional analysis: ρ liq ρ g = μ liq μ g k rg k rliq ( 2.10) The first term on the right hand side of the Equation ( 2.10) during the steam injection is very large (with liquid viscosity being more than two orders of magnitude larger than steam viscosity). The k rg and k rliq values are changing as the SAGD process advances, but they are in the same order of magnitude. Hence: ρ g ρ liq ρ liq ρ (total) ( 2.11) As a result, the Darcy law can be written for the liquid phase drainage multiplied by the number of propped fractures to estimate the maximum flow capacity of fractures for this gravity driven process: 19

37 q liq frac max = M k f. k rliq. w f. 2x f. ρ. g μ o (T) ( 2.12) Equation ( 2.12) can be extended for three phase relative permeability to estimate the oil flow capacity of fractures: q o frac max = M k f. k ro. w f. 2x f. ρ. g μ o (T) ( 2.13) studies. Equation ( 2.13) will be used later to make plateau-peak rate estimates in some case 2.4 Other limitations of the analytical modelling approach for SAGD with MSFHW There are some limitations in the above analytical approaches. For example, the presence of IHS facies is usually analogous to lower porosity, permeability and oil saturation. As a result, the oil in place of a reservoir with IHS facies would likely be less than that of the same reservoir with clean sand. The other assumption in this analytical approach is that, in the case with multiple fractures, it is assumed that the development of steam chambers in different pay intervals starts at the same time. In reality, however, this may not be the case. Therefore, the peak rate calculated by the analytical formulations could be higher than that of a real operation. As discussed, the above formulation is an estimation of the peak SAGD production rate only and not of the recovery factor. Although in some of the cases the analytically estimated peak rate could be higher than the base case, it does not mean the recovery factor would be similar. Later on, it will be shown that the numerical simulation of the same processes suggests that both recovery factor and peak rate of the fractured cases in a 20

38 layered reservoir (although considerably better than in the un-fractured case), would be always less than that of the clean sand. 2.5 Numerical simulation Numerical modelling of complex recovery processes provides more opportunities in understanding of the mechanics involved in the processes. Many of the simplifying assumptions that limit applicability of analytical solutions can be avoided. At the same time, numerical modelling is more prone to different errors some of which are discussed here. One of the major issues with numerical simulation is the approximation (discretization) error. Numerical solutions are in fact approximations of the exact solutions. For example, in modelling complex processes such as SAGD or CSS which involves fluid flow, heat transfer and sudden phase change all at once, setting up an accurate numerical model could be challenging. In modelling CSS for example, choosing the right grid cell size in such processes is very important since the computed depth of influence of pressure and temperature could be significantly different with different grids. At the same time, while smaller grids keep the discretization errors small, unreasonably small grids will lead to excessive run times without materially changing the results. Another pitfall of numerical simulation is using unrealistic inputs. Inexperienced users may get confused about the realistic range of physical properties required as inputs to the model. Furthermore, there can be many combinations of input parameters that lead to a similar solution of which only a handful can be used to generate representative forecasts. In 21

39 many cases, a user should apply engineering judgment to the modelling results to decide if the answers are valid and the model can be used to generate reliable forecast scenarios. SAGD process can be modelled using reservoir simulators with thermal option. The principal equations solved in the thermal reservoir simulations are mass conservation and energy conservation with the fluid flow governed by Darcy law. The flow equations are solved for multi-phase (oleic, aqueous and gaseous phases), multi-mechanisms (convection, diffusion and dispersion), and multi-components. In thermal reservoir simulators, the components are treated as pseudo components that can be defined by lumping individual chemical components. The mass transfer of different pseudo components between the phases is governed by equation of state or a table of K-values which is a function of pressure and temperature. Further details can be found in classic reservoir simulation text books [16, 17]. 2.6 SAGD simulation cases The physical properties for the SAGD simulation cases in this work are based on the reservoir and fluid properties of the Underground Testing Facility (UTF) project [18]. UTF project was one of the first SAGD field trials funded by the government of Alberta s AOSTRA program in conjunction with a number of industry partners. The goal was to demonstrate and prove the SAGD technology and learn the challenges involved in the field scale implementation. Table 2.1 summarizes the base case properties for the SAGD runs. A sample simulation data file is presented in Appendix I. 22

40 Table 2.1 SAGD model properties Model dimensions (m) 70 x 800 x 25 Oil saturation 0.83 Well spacing 70 m Initial pressure (datum) 1,450 kpa Well offset 5 m Injection pressure 2,600 kpa Porosity 35% Production BHP 2,250 kpa Horizontal permeability 3,000-5,000 md Viscosity (res. temp.) 5.0e6 o C Vertical permeability 2,000-3,000 md Viscosity (steam temp.) 7 o C Formation compressibility 7.0E-6 1/kPa Oil compressibility 6.84E-7 1/kPa Avg. thermal conductivity 1.469E5 J/m.day.C Rock heat capacity 2.39E6 J/m 3.C For the SAGD numerical investigation, it was decided to use CMG STARS which is a thermal reservoir simulation package and is suitable for SAGD simulation. The model was setup using only dead oil with no light hydrocarbon component and convection mass transfer only. The heat transfer considers both convection and conduction. The details of the model PVT, relative permeabilities and thermal parameters are found in Appendix I. The grid sizes are explained later. To start the investigation, four simulation runs were set up to demonstrate the performance difference between a case with clean continuous sand (base case), a case with interbedded pay, followed by two cases that are stimulated by transverse and longitudinal hydraulic fractures. The well pair length in all cases is assumed to be equal to the length of the model with production well drilled 2 m above the reservoir base. Table 2.2 shows the additional definitions for the simulation cases. 23

41 Table 2.2 SAGD model additional properties for the interbedded cases Case 1 (Red): SAGD base case Case 2 (Blue): SAGD with shale layers Case 3 (Green): SAGD with shale layers and 16 transverse fractures Case 4 (Magenta): SAGD with shale layers and 16 longitudinal fractures Vertical permeability of Shale Layers 10 md Shale layer thickness 2 x 2 m layers Fracture half-length 15 m Fracture permeability 4,800 D Fracture width 1 cm The first question to be raised is the range of fracture permeability that should be assumed for the purpose of a comparative modelling. The feasibility of this process will be shown to rely upon very high fracture conductivity. This issue will be discussed in detail in the sensitivity cases in Chapter 3. For now, it is assumed that the base case fracture has a permeability of 4,800 D and fracture width of 1 cm which results in a fracture block permeability of 50 D for a one meter wide block, and 100 D for a half a meter wide block. Later in the sensitivity analysis, limiting cases with fracture permeabilities of 400 to 19,800 Darcy are presented and discussed. The upper end of this fracture permeability range is arguably ambitious to achieve. However, various proppant choices that can deliver the required conductivity are available in the market and will be reviewed in Chapter 5: Chapter 5. Most of these proppants are course grain sandstones or manufactured ceramics which have very high conductivity values at low confining stresses. The Base case consists of a SAGD well pair with 800 m length and 5 m offset between the injector and producer. The grid cells are elongated in the direction of horizontal wells with 1 m x 1 m cross sections (in the plane perpendicular to the wells). 24

42 Parallel to the direction of horizontal wells, the grid cell sizes are picked to accommodate the fractures. In Case 3, to accommodate accurate flow to/from transverse fractures gradual increase in grid size was used to and away from fractures (0.5 m, 1 m, 2 m, 5 m, m, 20 m) with the rest of grids being 20 m. In Case 4, the gird sizes are a mix of 5 m and 10 m cells such that fractures with various half lengths (5 m, 10 m, 15 m and 20 m) can be inserted in the model. Figure 2.6 Grid fracture planes for the SAGD cases 25

43 For Cases 2, 3 and 4, it was assumed that there are two 2 m-thick shale layers present, one right above the injector and the second one, 9 m above the injector. Figure 2.6 illustrates the SAGD grid and fracture planes for different cases. The process starts with a steam circulation phase in both the injector and producer (90 days for Cases 1 and 2, and 150 days for Cases 3 and 4). After the warm up, the injector is put under continuous steam injection at 2,600 kpa with maximum a steam injection rate of 600 tonnes per day with 90% quality. The bottom well is produced at 2,250 kpa with maximum liquid rate of 900 m 3 per day. There is also a steam trap control that limits the steam production to 5 tonnes per day. Figure 2.7 to Figure 2.12 show the oil production rate, steam injection rate, recovery factor and instantaneous and cumulative steam to oil ratios (ISOR and CSOR) of the SAGD Cases 1 to 4 as well as ISOR vs. recovery factor. As expected, the performance of Case 1 is better than all the other cases. It reaches peak oil production within a year after start of steam injection and recovery factor of 80% while maintaining the CSOR of 2.2. These numbers are on par with best performing pads in the leading SAGD operations in Alberta. Figure 2.12 shows that the energy efficiency of the stimulated cases, although not as good, is comparable with that of the clean sand case. Case 2 has a notably less favorable performance in terms of oil rate, recovery factor and SORs due to presence of shale layers. However, stimulation of Case 2 with hydraulic fractures (Cases 3 and 4) significantly improves the performance of SAGD process as steam can now reach the otherwise inaccessible oil. This is evident from the simulation results and will be discussed in more details. 26

44 Cases 3 and 4 have more steam leak off during the circulation period through the hydraulic fractures. As a result, a longer steam circulation time is needed to establish oil mobility between the injector and producer (150 days instead of 90 day). More details for start-up strategy for stimulated cases will be discussed later. Evaluating the cases 3 and 4 shows that the times to reach the plateau steam and oil rates are longer than the base case and oil recovery increase is more gradual. Also, the peak oil rates or the stimulated cases are lower than in the base case. This is in contrast with the analytical modelling results in which the peak oil was in fact larger than the base case. Later in this chapter, analytical rate estimates will be compared with the simulation cases and the differences will be discussed. Figure 2.7 Oil production rates for SAGD Cases 1 to 4 27

45 Figure 2.8 Steam injection rates for SAGD Cases 1 to 4 Figure 2.9 Oil recovery factors for SAGD Cases 1 to 4 28

46 Figure 2.10 Instantaneous Steam to Oil ratios (ISOR) for SAGD Cases 1 to 4 Figure 2.11 Cumulative Steam to Oil ratios (CSOR) for SAGD Cases 1 to 4 29

47 Figure 2.12 Instantaneous Steam to Oil ratios (ISOR) vs. Recovery Factor for SAGD Cases 1 to 4 Another issue that was noted from the sensitivity runs (shown in Chapter 3) is the fact that a certain minimum number of fractures or a minimum of enhanced permeability volume is required to compensate for the existence of the shale layers. For example, if Cases 3 and 4 were forecasted with half of the fracture contact, it would take years of steam injection before a reasonable oil production could be achieved. This issue will be addressed with more details in the sensitivity cases in Chapter 3. Steam injection rate in SAGD operation is somewhat determined by the capability of the reservoir to take the steam. That is, since the governing flow mechanism is gravity flow in constant steam chamber pressure, the injection volume is strongly related to the oil and water withdrawal volumes. As a result, for a reservoir with significant permeability 30

48 barriers, both injection and withdrawal rates are lower than a case without barriers. This provides an opportunity to potentially develop RF-PVI (recovery factor vs. pore volume injected) type curves for different settings where rates and times do not skew the comparisons. Figure 2.13 shows the comparison of the drainage of oil between the base SAGD case, Case 2 and Case 4. While the OIP of all cases are similar, the low permeability of shale layers in interbedded cases results in a very slow drainage of them which in turn reduces the recovery factor at the same point in time. SAGD Clean Sand (Base Case) SAGD with Shale (unfractured) Case 2 SAGD with Shale (longitudinal fractures) Case 4 Figure 2.13 Snap shot of oil saturation for SAGD base case, SAGD in an interbedded formation and that with longitudinal fractures, time = 1,280 days 31

49 These simulation cases clearly illustrate the potential of using MSFHW technique in SAGD operation when significant vertical permeability barriers are present. It is worth noting that in reality we will likely never have true longitudinal or transverse fractures, and these configurations are essentially the limiting orientations. 2.7 Comparing numerical simulations with analytical solutions Earlier in this chapter, some analytical solutions were derived to predict the plateaupeak oil production for fractured stimulated SAGD wells based on the Butler analytical solution if effective permeability to oil is used in the equations. Figure 2.14 shows the comparison between the analytical estimates from Equations ( 2.1), ( 2.5) and ( 2.7) with numerical Cases 1, 3 and 4. To do this, the Butler s solution in Equation ( 2.1) was matched to base SAGD case (clean reservoir with no shale layer) using viscosity exponent m = 1.1. Then, Equations ( 2.5) and ( 2.7) were used to estimate the peak rate for the two base fractured cases. Table 2.3 shows the same comparison for various sensitivity cases that will be presented in Chapter 3. The assumption of infinite permeability of the fracture in the analytical formulation can be examined here. Equation ( 2.13) provides an estimate of the maximum rate transmissible through fractures under gravity drainage head. Table 2.3 compares the analytical rate estimates with that predicted by Equation ( 2.13) using the reservoir data from Table 2.1 and fracture properties from Table 2.2. Table 2.3 shows that the infinite conductivity assumption in the analytical formulation is likely valid. However, if the fracture permeability is low enough (in this case below 1,000 D for 16 fracture stages each 32

50 with 15m half length) the assumption of infinite conductivity is violated, and the analytical formulations cannot be used. Table 2.3 Estimated SAGD peak oil rates from analytical equations (all in m 3 /day) Clean Sand Untreated Shale-Sand Shale-Sand with 16 longitudinal fractures Shale-Sand with 16 transverse fractures Net contribution from fractures Net contribution from fractures Maximum oil rate transmissible from fractures under gravity drainage: Q frac max, k f = 4,800 D, k ro = Q frac max, k f = 2,400 D, k ro = Q frac max, k f = 1,200 D, k ro = Q frac max, k f = 1,000 D, k ro = Figure 2.14 shows the comparison of analytical peak rate and numerical simulation results. The analytical peak rates over-estimate the numerical results. Analytical solutions by nature are simplified, and usually cannot include all the physics involved in such complex processes. A major issue that can be highlighted is an implied assumption in analytical formulation that all the equivalent pseudo well pairs (as shown in Figure 2.2) start to inject/produce at the same time. In reality, however, there is a time lag for upper sand layers to start contributing to the drainage of the SAGD scheme. If a factor of 0.8 is applied to the analytical rate estimates, an acceptable match to the numerical results can be obtained as can be seen in Figure This correction factor can be used as a rule of thumb to improve the predictive capability of the analytical estimates. However, more combinations of reservoir and fracture propertied should be considered before this rule of thumb can be adopted for general use. 33

51 Figure 2.14 Comparison of the analytical peak rate estimates and numerical results for the base cases Figure 2.15 Comparison of the analytical peak rate estimates times 0.8 with the numerical results 34

52 2.8 Discussion of relative permeability in the fracture blocks In a propped fracture, the grains are presumably fairly uniform. This makes the capillary forces to be fairly low. As a result, the relative permeability curves in the fracture itself are almost linear, meaning that the relative permeability of a phase is equal to the saturation of that phase [19]. If the gaseous phase flow velocity is high, the inertial losses may cause deviation from the Darcy s law which reduces the phase mobility [20]. In absence of modelling non-darcy flow, this effect can be captured by reducing relative permeability. However, with the driving force of the flow in SAGD being the gravity head, the inertial losses in a fractured SAGD process should be negligible. In the numerical models in this work, it was assumed that the relative permeability in the fracture is similar to that of formation. This assumption could be somewhat inaccurate. However, since the fractures are treated as part of the larger grid block, the weighted average dampens the effect. The following is a more detailed analysis of the problem. Consider a single-phase flow in a block with a fracture. If matrix permeability is k, fracture permeability is k f, the dimension of the matrix part of the block is x and fracture width is w f, then the average permeability of the block, k block, can be defined by: k block = k. x + k f. w f x + w f ( 2.14) If we now consider multi-phase flow, Equation ( 2.15) can be used to calculate the true average fracture block relative permeability. 35

53 k rw block = k. x. k rw + k f. w f. k rwf k block ( x + w f ) ( 2.15) As an example, the calculation of the true average relative permeability of the block for the data of the SAGD base case is shown in Table 2.4. At water saturation of 0.5, the reservoir and fracture (linear) relative permeability are at their maximum difference. However, the true average relative permeability of the block is changed only to 0.13 from the current At all other saturation values, the difference would be even smaller. After the fracturing treatment, the aqueous phase saturation in the fracture is likely to be close to 1.0. In a typical SAGD project, there could be a considerable length of time between the completion operation and the first steam, sometimes up to a year. Depending on the reservoir conditions, such as presence of mobile water and reservoir fluid mobility, the cold bitumen could slowly squeeze into the fractures and push the aqueous phase back. In the simulation cases, the saturation in the fracture gird blocks was initialized similar to the rest of the reservoir. To calculate the real relative permeability of the fracture grid block at the beginning of SAGD circulation, we can use Equation ( 2.15) with k rw (S wi ) and k rwf (S wif ) where k rw (S wi ) < k rwf (S wif ). This means for the early injection times, the true effective permeability of the block is larger than what the simulator calculates. As time goes by, and depending on what portion of the fracture has been receiving steam or draining bitumen, the difference between the two calculated values is reduced. Another factor that affects the early time effective permeability is the amount of gels and crosslinking additives in the fracturing fluid. It is expected that all these materials degrade after steam is introduced to the fractures. Overall, there are many complexities 36

54 associates with the early stages and start-up that are not being considered. However, it is reasonable to assume that they will not impact long term performance results, pending the successful initiation of the counter current flow in the fractures. This would be an area for future study. Table 2.4 Estimation of average block relative permeability for an extreme case difference k (D) 4.0 k rw res. (S w = 0.5) 0.1 x matrix (m) 0.99 k f (D) 4,800 k rw f (S w = 0.5) 0.5 w f (m) 0.01 k block (D) 52 x block (m) 1.0 k rw block (S w = 0.5) Summary In this chapter, the basics of the recovery process for using multi-stage fracturing in SAGD operations were discussed. Some analytical equations to estimate the peak oil rate of the process were derived based on the original Butler equation. The assumption of infinite vertical permeability of the fracture was also examined and it was shown that if the fracture permeability is reasonably high to allow counter current gravity flow capacity that significantly exceeds the capacity for the reservoir fluids delivery, this assumption is valid. Some numerical simulation cases were created to show the framework of the more detailed modelling and performance prediction of this process. Comparison of the analytical rate estimates and numerical simulation results showed that the analytical equations overestimate the peak rate. It is proposed to use a factor of 0.8 to analytical rate estimates 37

55 of the stimulated cases to correct for the simplifying assumptions. The corrected estimate can be used as a quick first pass guess for the peak rates. Using similar relative permeability for fracture and matrix could introduce a minor level of inaccuracy to the results, although it is not expected to change the conclusions. Overall the results presented in this chapter have shown that the multi-stage fracture stimulation appears to be a promising novel method to improve SAGD recovery from formations with interbedded shales. Consequently, there is merit to do a more detailed sensitivity analysis of the process and study the geomechanical issues in the following chapters. 38

56 CHAPTER 3: SENSITIVITY ANALYSIS OF DESIGN VARIABLES, AND DEVELOPMENT OF AN EXTENDED DIMENSIONLESS FRACTURE CONDUCTIVITY TERM 3.1 Introduction In the proposed method, there are several parameters that can affect the performance of the stimulated SAGD wells which are related to specification of the fracturing rather than the reservoir properties. In this chapter a sensitivity analysis of some of these important parameters is presented. We will start with the start-up strategy of the well pairs in terms of pre-heating and circulation time. Then the orientation and density of the fractures will be taken as variables. Another variable is the size of the fracture job which in turn determines the target fracture length. We will then discuss the target permeability of the fractures and its importance in effective drainage of stimulated SAGD wells. The sensitivity results will then be used in development of a dimensionless fracture conductivity to be used as a guide in first pass analysis of how much total conductivity is required for this enhanced SAGD process. All results in this chapter are based on numerical simulations using the same type of model as in Chapter Well pairs start-up strategy Starting up a SAGD well pair is one of the most critical parts of the operation. Steam is circulated in both injector and producer for a period of time, usually between 30 39

57 to 90 days [21, 22], to warm up the cold bitumen between the well pair mostly under conductive heating. The circulation is stopped using some of the following indications: - A certain temperature is achieved between the well pairs; - A large portion of the injected steam leaks off into the formation and does not return to the surface as a condensate; - The circulation return contains a certain oil cut; - The pressure differential between the injector and producer suggests strong hydraulic communication between the well pair. Premature well communication in a section of the well could cause the rest of the well not to receive enough steam and hence remain cold. This may cause reduction in the productivity and sweep efficiency of the well pair [21]. In a scenario where the wells have been stimulated prior to start-up and very high conductivity is present in the fractured intervals, the significance of the circulation period becomes even more important, since well communication is likely to be initially achieved in the vicinity of fracture intervals. Hence, it is important to assure that conduction heating warms up the area between the well pair regardless of fractures locations, even if hydraulic communication is achieved between the well pairs. To study the effect of circulation time in the performance of stimulated well pairs, three cases were considered with 30, 90 and 150 days of circulation, using the data for the fractured base SAGD cases (Cases 3 and 4 of Table 2.2). Figure 3.1 and Figure 3.2 illustrate the sensitivity results for the circulation time for the base cases of longitudinal and transverse fracture orientation, respectively. The graphs 40

58 indicate that longer circulation time is needed for better performance of a stimulated SAGD well pair. For a traditional SAGD in clean sand, a shorter circulation time would have been sufficient to establish proper well-long communication (see Case 1 in Figure 2.7). In the cases with stimulated well pairs, ending circulation time early lowers the production contribution from the un-stimulated portion of the wells. In addition, some fractures are likely to get more steam earlier on the circulation period, causing a delay on contribution of other fractures in the wellbore, which in turn prolongs the time required to reach peak production. Figure 3.1 Sensitivity to circulation time (30, 90 and 150 days), 16 longitudinal fractures, x f =15 m 41

59 Figure 3.2 Sensitivity to circulation time (30, 90 and 150 days), 16 transverse fractures, x f = 15 m During the startup strategy, it is very important to avoid inducing convective heating as much as possible and only rely on conductive heating between the well pair to achieve a fairly uniform temperature profile along the wellbore. The circulation rate in the above simulation cases is set at cold water equivalent of 50 m 3 /day/well. Applying higher circulation rates would certainly affect (shorten) the time of conversion to SAGD. However, it would also increase the chance of localized heating nearby the fractures. 3.3 Fracture orientation and spacing One of the important factors in the performance of a stimulated SAGD well pair is the orientation of the fractures. The limiting orientations here are considered to be 42

60 longitudinal fractures (parallel to the wells) and transverse fractures (perpendicular to the wells). The direction of fractures is primarily determined by the direction of minimum stress in a formation. In a stress state in which the minimum stress is horizontal, stimulating wells drilled parallel to minimum horizontal stress are likely to generate transverse fractures, and fractures in wells drilled perpendicular to the minimum horizontal stress would likely be longitudinal fractures. Figure 3.3 shows the stress map in Western Canadian Sedimentary Basin (WCSB). In general terms, the minimum horizontal stress in WCSB is parallel to the Canadian Rocky Mountains ranges. There might be local variation in the stress direction due geological conditions (such as stress changes in Peace River Arch [23]) or oilfield operations induced factors (such as reservoir depletion or water flooding) that may change the direction of the minimum stress locally. In some cases, the reservoir deposition is in the form of narrow channels which dictates the drilling direction of horizontal wells for development. In these cases, the fractures direction is dictated both by drilling direction and in-situ stresses, and can be between the parallel and transverse cases. In such situations the connection of the body of the fracture with the well can be complex and cause undesirable, excessive pressure drop (near-well tortuosity). 43

61 Figure 3.3 Horizontal stress trajectories projected across the WCSB [24] Figure 3.4 shows the sensitivity cases for the effect of fracture orientation (longitudinal vs. transverse) on the SAGD profiles for cases with various fracturing densities. The sensitivity results indicate that picking the more effective orientation depends on the fracture density. For cases with lower fracture density (in this case 4 and 8 fracture stages for an 800 m well), the transverse orientation shows better results. This can be understood by looking back at Figure 2.3 and Figure 2.4 schematics. When the fracture stages are far enough apart, transverse orientation case forms independent steam chambers perpendicular to the well. These chambers cover a larger portion of the reservoir when compared to that of longitudinal orientation. However, if the fracture density is high enough, the volume of the heated reservoir is roughly the same for both orientations, although the longitudinal 44

62 direction has a faster ramp-up time due to more effective near wellbore heating in early times. Figure 3.4 Sensitivity of the fracture orientation for various fracture densities (4, 8 and 16), x f = 15 m Fracture spacing is another optimization factor. Naturally, more fracture stages per unit well length means higher injectivity and productivity. However, the cost also increases and there are operational limitations on how close the fractures can be planned. In the longitudinal orientation for example, planning fracture stages too closely could become practically impossible, since the fractures could interfere or coalesce with each other and negatively affect the shape and size of one another. Within the realm of technical possibility, what determines optimized number of fracture stages is the project economics. 45

63 Adding more stages beyond a certain point is merely a recovery acceleration practice. One can make an economic case for additional stages by incorporating the cost per stages and the accelerated recovery type curve and find the optimal number of stages for a given reservoir [25]. 3.4 Fracture half-length The next sensitivity analysis was performed on the length of the fracture stages in both longitudinal and transverse directions. As discussed earlier, the main goal of fracture stimulation of the wells is to increase the vertical permeability. Therefore, it was assumed that the fractures break down and open up the interbedded shale layer all the way to the top of the formation, i.e. fracture height and width are set the same and the only varying parameter is fracture half-length. Figure 3.5 shows four cases with longitudinal fracture half-length of 5, 10, 15 and 20 m (with 16 fracture stages in each case). It can be perceived that larger fractures offer better well pair performance. However, when comparing cases with 15 m and 20 m fracture half-lengths, the production improvement is marginal. This means that at some point, larger fracture would not offer improved performance since the smaller fracture are capable of forming similar size steam chambers and draining the hot liquid to the production well. Furthermore, there is a physical limit on the size of fracture in the longitudinal direction. For example, for the cases shown in Figure 3.5 with an 800 m well length, placing 16 longitudinal fractures with half-length of 20 m would put the tips of two adjacent fractures at only 10 m. 46

64 Figure 3.5 Sensitivity to fracture half-length, longitudinal cases (x f = 5, 10, 15 and 20 m) Figure 3.6 shows four cases of transverse fractures with half-lengths of 5, 10, 15 and 20 m (with 16 fracture stages in each case). Unlike the longitudinal case, the early time productions of all four cases are essentially identical. This can be explained by picturing the development of steam chambers around each fracture in early times, which is not affected by fractures lengths. However, the cases with larger fractures accelerate the growth of steam chamber away from the wellbore while drainage for the cases with shorter fractures has to flow through paths with original reservoir conductivity before getting to the fractures. Similar to the longitudinal case, the significance of performance improvement is reduced when comparing cases with 15 m and 20 m fracture half-length in late times. In 47

65 comparing 15 m to 20 m fracture half-length for example, the former has enough capability to grow the steam chamber and drain the hot liquid. There is also a physical limiting factor for the transverse fractures half-length which is imposed by well spacing between the adjacent well pairs. For example, if the well spacing is 70 m, a 20 m fracture half-length means that the fracture tips from the two neighbor wells may only be 15 m apart. Figure 3.6 Sensitivity to fracture half-length, Transverse cases (x f = 5, 10, 15 and 20 m) 3.5 Fracture permeability One of the most important factors that determine the effectiveness of a hydraulic fracturing job is fracture conductivity C f which is the product of fracture permeability k f, and the fracture aperture w f : 48

66 C f = k f. w f ( 3.1) The concept of dimensionless fracture conductivity will be expanded later. If we assume a constant fracture aperture, fracture permeability can be used for sensitivity analysis on fracture conductivity. The underlying factors that determine the fracture permeability are the type, size and concentration of the proppant and its interaction with the formation under different confining stresses. The proppant selection will be discussed in the subsequent chapters. Once again, sensitivity runs were set up for longitudinal and transverse fractures. In the longitudinal runs, the fracture grid is 1 m wide, while in transverse run it is 0.5 m. As a result, the fracture block average permeabilities in the transverse runs are twice as those of the longitudinal runs to generate equivalent conductivities. Figure 3.7 shows the sensitivity results for longitudinal cases. Similar to the conclusions of fracture half-length sensitivity, it can be seen that increased fracture permeability improves the production rates. However, it does not improve the performance of the well pair beyond a certain point. This finding is very useful in designing proper fracture treatments that provide efficient levels of total drainage capacity. 49

67 Figure 3.7 Sensitivity to fracture permeability, longitudinal cases (k f between 400 to 19,800 Darcy) Figure 3.8 shows the sensitivity results for transverse cases. The conclusion is similar to that of longitudinal cases on the optimized fracture permeability target. Interestingly, the results for longitudinal and transverse fractures match each other, meaning that fractures with the same conductivity in both directions result in roughly similar production performance. That is to say, if the fracture conductivity is high enough, fracture direction is not that important in improving the production. 50

68 Figure 3.8 Sensitivity to fracture permeability, transverse cases (k f = 50, 100, 200 and 400 Darcy) 3.6 Dimensionless fracture conductivity So far, some major factors in a fracturing stimulation were reviewed. While evaluation of these factors individually is helpful in providing a better understanding of the recovery mechanism, they are very subjective to the reservoir, wellbore and fracturing specifications. For example, the well length in all the above cases was fixed at 800 m and an optimized case can be picked from the above result. If some of the parameters change, a whole new study should be initiated to pick the best fracturing design. To avoid that, it is worthwhile to examine the concept of dimensionless fracture conductivity, C fd as defined in Equation ( 3.2) [26] and see if some general design guideline can be inferred. 51

69 C fd = k f. w f k. x f ( 3.2) All of the above sensitivity cases are summarized in Table 3.1. Economides et al. showed that in higher permeability formation, the ideal C fd is 1.6 [26]. Table 3.1 Summary of the SAGD sensitivity cases S n t v t a k (D) w (m) k (D) x (m) L (m) a. (m) L a. C C t 500 P ak at at (m 3 /d) (m 3 /d) 2000 day nal t al rate t. (m3/d) SAGD_SHLBR_16stage-Xf=10m-Longitudinal_150dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=10m-Longitudinal_150dstrup_kf=9800D SAGD_SHLBR_16stage-Xf=10m-Longitudinal_150dstrup_kf=19800D SAGD_SHLBR_16stage-Xf=10m-Longitudinal_150dstrup_kf=2300D SAGD_SHLBR_16stage-Xf=10m-Longitudinal_150dstrup_kf=1050D SAGD_SHLBR_16stage-Xf=10m-Longitudinal_150dstrup_kf=400D SAGD_SHLBR_16stage-Xf=10m-Transverse_150dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=10m-Transverse_150dstrup_kf=9800D SAGD_SHLBR_16stage-Xf=10m-Transverse_150dstrup_kf=19800D SAGD_SHLBR_16stage-Xf=10m-Transverse_150dstrup_kf=2300D SAGD_SHLBR_16stage-Xf=10m-Transverse_150dstrup_kf=1050D SAGD_SHLBR_16stage-Xf=10m-Transverse_150dstrup_kf=400D SAGD_SHLBR_16stage-Xf=15m-Longitudinal_150dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=15m-Longitudinal_150dstrup_kf=9800D SAGD_SHLBR_16stage-Xf=15m-Longitudinal_150dstrup_kf=19800D SAGD_SHLBR_16stage-Xf=15m-Longitudinal_150dstrup_kf=2300D SAGD_SHLBR_16stage-Xf=15m-Longitudinal_150dstrup_kf=1050D SAGD_SHLBR_16stage-Xf=15m-Longitudinal_150dstrup_kf=400D SAGD_SHLBR_16stage-Xf=15m-Longitudinal_30dstrup_kf=4800D NA 30 SAGD_SHLBR_16stage-Xf=15m-Longitudinal_90dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=15m-Transverse_150dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=15m-Transverse_150dstrup_kf=9800D SAGD_SHLBR_16stage-Xf=15m-Transverse_150dstrup_kf=19800D SAGD_SHLBR_16stage-Xf=15m-Transverse_150dstrup_kf=2300D SAGD_SHLBR_16stage-Xf=15m-Transverse_150dstrup_kf=1050D SAGD_SHLBR_16stage-Xf=15m-Transverse_150dstrup_kf=400D SAGD_SHLBR_16stage-Xf=15m-Transverse_30dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=15m-Transverse_90dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=20m-Longitudinal_150dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=20m-Longitudinal_150dstrup_kf=9800D SAGD_SHLBR_16stage-Xf=20m-Longitudinal_150dstrup_kf=19800D SAGD_SHLBR_16stage-Xf=20m-Longitudinal_150dstrup_kf=2300D SAGD_SHLBR_16stage-Xf=20m-Longitudinal_150dstrup_kf=1050D SAGD_SHLBR_16stage-Xf=20m-Longitudinal_150dstrup_kf=400D SAGD_SHLBR_16stage-Xf=20m-Transverse_150dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=20m-Transverse_150dstrup_kf=9800D SAGD_SHLBR_16stage-Xf=20m-Transverse_150dstrup_kf=19800D SAGD_SHLBR_16stage-Xf=20m-Transverse_150dstrup_kf=2300D SAGD_SHLBR_16stage-Xf=20m-Transverse_150dstrup_kf=1050D SAGD_SHLBR_16stage-Xf=20m-Transverse_150dstrup_kf=400D SAGD_SHLBR_16stage-Xf=5m-Longitudinal_150dstrup_kf=4800D SAGD_SHLBR_16stage-Xf=5m-Transverse_150dstrup_kf=4800D SAGD_SHLBR_04stage-Xf=15m-Longitudinal_150dstrup_kf=4800D SAGD_SHLBR_04stage-Xf=15m-Transverse_150dstrup_kf=4800D SAGD_SHLBR_08stage-Xf=15m-Longitudinal_150dstrup_kf=4800D SAGD_SHLBR_08stage-Xf=15m-Transverse_150dstrup_kf=4800D The concept of dimensionless fracture conductivity can certainly help a first pass fracturing design criteria based on the reservoir properties. However, the original definition of the C fd does not include any parameter for the number of fractures. As a result, the cases 52

70 with less number of fracture stages (4 and 8) do not scale properly. To address this problem, it is proposed to add another dimensionless term to the C fd to represent the number of fractures. The new dimensionless number will be called total dimensionless fracture conductivity, or C fdt. C fdt k f. w k. x f f L frac spacing ( 3.3) Table 3.1 includes all sensitivity cases. The peak oil rate vs. fracture half-length for all cases is plotted in Figure 3.9 labeled with total dimensionless fracture conductivity values. Note that longitudinal cases labels are on the left hand side of the data markers and transverse cases labels are on the right. For more clarity Figure 3.10 and Figure 3.11 are presented, depicting longitudinal and transverse cases individually. A general observation that can be made from these figures is that for C fdt >10 and x f >15 m, the total conductivity is large enough to achieve an appropriate drainage capacity. Higher conductivity does certainly improve the efficiency of the process; however, the improvement is not substantial. Also, by comparing C fd and C fdt in Table 3.1, and the graphs, one can see the cases with 4 and 8 fractures have similar C fd values while C fdt captures the difference in total conductivity with 16 fracture cases. Once again, the individual plots (Figure 3.10 and Figure 3.11) show that the longitudinal fracturing cases are not very sensitive in larger C fdt values whereas larger C fdt values in transverse fracturing cases show a more visible improvement in peak oil rate. 53

71 Peak oil rate, Flat portion (m3/d) Peak oil rate, Flat portion (m3/d) C fdt Longitudinal (left side lables) and Transverse (right side lables) Cases CfDt=79.2 CfDt=39.2 CfDt=12.8 CfDt=52.8 CfDt=26.1 CfDt=52.8 CfDt=26.1 CfDt=12.8 CfDt=39.6 CfDt=19.6 CfDt=9.6 CfDt=4.6 CfDt=39.6 CfDt=19.6 CfDt= CfDt=39.2 CfDt=79.2 CfDt=19.2 CfDt=19.2 CfDt= CfDt=6.13 CfDt= CfDt=9.2 CfDt=9.2 CfDt=6.4 CfDt=2.8 CfDt=2.1 CfDt=2.1 CfDt=6.4 CfDt=2.8 CfDt=1.06 CfDt=4.2 CfDt=4.2 CfDt=0.8 CfDt=3.2 CfDt=0.8 CfDt=1.06 CfDt=1.6 CfDt= Fracture half length, Xf (m) Figure 3.9 Plateau-Peak oil rate vs. fracture half-length for multiple C fdt s C fdt Longitudinal Cases CfDt=12.8 CfDt=52.8 CfDt=26.1 CfDt=39.6 CfDt=19.6 CfDt=9.6 CfDt= CfDt=39.2 CfDt=79.2 CfDt= CfDt= CfDt=9.2 CfDt=2.8 CfDt=2.1 CfDt=6.4 CfDt=4.2 CfDt=1.06 CfDt=0.8 CfDt= Fracture half length, Xf (m) Figure 3.10 Plateau-Peak oil rate vs. fracture half-length for longitudinal cases with C fdt values 54

72 Plateau-Peak oil rate, Flat portion (m3/d) C fdt Transverse Cases CfDt=79.2 CfDt=39.2 CfDt=52.8 CfDt=26.1 CfDt=12.8 CfDt=39.6 CfDt=19.6 CfDt= CfDt=19.2 CfDt= CfDt= CfDt=9.2 CfDt=6.4 CfDt=2.1 CfDt=2.8 CfDt=4.2 CfDt=3.2 CfDt=1.06 CfDt=0.8 CfDt= Fracture half length, X f (m) Figure 3.11 Plateau-Peak oil rate vs. fracture half-length for transverse cases with C fdt values Another way of presenting this data set is to plot the peak rate against the fracture permeability as shown in Figure 3.12 and Figure 3.13 (noting that only the cases with 16 fracture stages are plotted). These figures offer an enhanced picture of the trends within the large matrix of sensitivity cases concerning fracture permeability and half-length. In the longitudinal cases (Figure 3.12), longer fractures clearly offer improved peak rates, by approaching an ideal case of one continuous fracture along the full well length. In contrast, for the transverse cases (Figure 3.13) the peak rate improvement for 20 m fracture half-length is marginal. In both fracture directions, there is an inflection point in higher permeabilities. Increased fracture permeability above 4800 D does not lead to material gain in the peak rate. 55

73 Plateau-Peak oil rate, Flat portion (m3/d) Plateau-Peak oil rate, Flat portion (m3/d) C fdt Longitudinal Cases CfDt=9.6 CfDt=19.6 CfDt= CfDt=4.6 CfDt=12.8 CfDt=26.1 CfDt= CfDt=6.13 CfDt=19.2 CfDt=39.2 CfDt= CfDt=2.8 CfDt=9.2 X f = 20 m CfDt=2.1 X f = 15 m 120 CfDt=1.06 X f = 10 m CfDt=0.8 CfDt=1.6 CfDt= Fracture Permeability, k f (D) Figure 3.12 Plateau-Peak oil rate vs. fracture permeability for longitudinal cases with C fdt values C fdt Transverse Cases CfDt=26.1 CfDt=19.6 CfDt=52.8 CfDt= CfDt=9.6 CfDt=12.8 CfDt=39.2 CfDt= CfDt=4.6 CfDt=6.13 CfDt= CfDt=9.2 CfDt=2.1 X CfDt=2.8 f = 20 m X f = 15m 120 CfDt=4.2 X CfDt=0.8 f = 10 m CfDt=1.06 CfDt= Fracture Permeability, k f (D) Figure 3.13 Plateau-Peak oil rate vs. fracture permeability for transverse cases with C fdt values 56

74 Plateau-Peak Rate Ratio to that of Clean Sand The observed behavior can also be explained by a relative plateau-peak rate of the stimulated cases to the SAGD rate in clean sand with unrestricted access to the full reservoir pay as a function of the a non-dimensional fluid flow capacity as shown in Figure The flatter portion of the curve corresponds to C fdt values larger than 10 for x f >15m. Hence, the previous conclusion can be maintained, i.e., for C fdt >10 and x f >15m, the total conductivity is large enough to achieve adequate drainage capacity Longitudinal Fracture Cases, Xf = 10 m Transverse Fracture Cases, Xf = 10 m Longitudinal Fracture Cases, Xf = 15 m Transverse Fracture Cases, Xf = 15 m Longitudinal Fracture Cases, Xf = 20 m Transverse Fracture Cases, Xf = 20 m Total Dimensionless Fracture Conductivity ( C fdt ) Figure 3.14 Plateau-peak rate ratio to clean sand SAGD rate as a function of total dimensionless fracture conductivity (C fdt ) The type curves presented in Figure 3.14 can be presented with other dimensionless groups as well. For example, for longitudinal fractures the normalized rates can be grouped based on fracture spacing over half-length to represent the fraction of the horizontal well length 57

75 opened by fractures. For the transverse fractures, the ratio of the height of dominant reservoir compartment to fracture half-length can be used as a dimensionless group. 3.7 Recovery factor vs. pore volume injected (RF-PVI) In classical reservoir engineering, one of the plots that may demonstrate interesting information is the recovery factor versus pore volume injected (or hydrocarbon pore volume injected). Since the parameters are dimensionless, the learning from the plot can be compared between the reservoirs with various shapes and sizes, and also between different recovery processes. Figure 3.15 and Figure 3.16 illustrate the RF-PVI plot for longitudinal and transverse fracture cases. The sensitivity cases of Table 3.1 were only run for 2,000 days. It can be observed that the RF-PVI has a fairly tight band of curves for most cases, which are theoretically capable of achieving recovery factors beyond 80%. The cases that show either lower ultimate recoveries or slower pace of recoveries are identified with thicker line width. 58

76 Figure 3.15 Recovery factor vs. pore volume injected (RF-PVI); Longitudinal fracture cases for 2,000 day simulation runs A closer look reveals that they are the cases with either lower number of fracture stages, or low fracture conductivity which hinders effective drainage from the reservoir pay. Very short circulation time is also hurting the effectiveness of the SAGD recovery process due to poor steam conformance. Another observation is that all the RF-PVI curves are between those of SAGD in clean sand and SAGD in un-stimulated sand-shale sequences. 59

77 Figure 3.16 Recovery factor vs. pore volume injected (RF-PVI); Transverse fracture cases for 2,000 day simulation runs A recovery factor graph (vs. time or PVI) is normally an increasing concave down curve. With that in mind, some of the recovery curves in Figure 3.15 and Figure 3.16 are still on the rise with no sign of concaving. This is due to premature termination of the simulation runs at 2,000 days for simulation cases with lower total fracture conductivity. Time is an implicit piece of information in the RF-PVI plots. That is, while the plot provides useful information on the process efficiency and ultimate recovery factor, it does not highlight the length of time required to achieve the injection or production target, which can in turn change the project economics significantly. To address this issue, a subset of simulation cases with C fdt values less than 10 was run for up to 5,000 days. 60

78 Figure 3.17 shows that even for the low conductivity cases it is possible to achieve very high recovery factors at a longer time span. Figure 3.17 Recovery factor vs. pore volume injected (RF-PVI) for a subset of cases with low C fdt values For most cases presented, it can be observed that at one pore volume of injection, the recovery factor is between 35 to 45%. At PVI=1 (which is almost equivalent to HCPVI=1.2 for the cases presented) the steam oil ratio should inherently be in the range of 2.6 to 3.4, implying a fairly efficient SAGD process even at the lower range of fracture conductivities. In other words, the energy efficiency of the process is fairly good. Another way of analyzing the results in terms of energy efficiency viability is to plot the steam-oil ratio (SOR) vs. recovery factor as shown in Figure 3.18 and Figure

79 It can be seen that the SOR of all cases are in the range of 2.5 and 3.5 until the recovery factor reaches 65-70% at which point the curves start to rise. The only exceptions to this behavior are the cases with short low permeability fractures. Also, there is a rather large drop in SOR just before the last rise, almost in all cases, which also corresponds to steeper slope of recovery factors in RF-PVI plots. This can be explained from oil production and steam injection rates of the sensitivity cases presented earlier in this chapter. At some point of time, the accumulated heat in the reservoir reaches a point when more steam injection results in live steam production in the model. At this point, the steam injection is cut by the model while the existing heat in the reservoir continues to mobilize and drain the oil. Figure 3.18 Steam-oil ratio vs. recovery factor for longitudinal cases (including extended runs) 62

80 An important conclusion can be made from the Figure 3.18 and Figure It can be argued that using MSFHW in SAGD can efficiently recover oil sands from stratified pays provided that the fracturing creates a certain minimum fracture conductivity. Number of stages, longer and more permeable fractures will accelerate the recovery. However, even the lower range of these variables can still offer fairly decent processes in terms of the energy efficiency. Figure 3.19 Steam-oil ratio vs. recovery factor for transverse cases (including extended runs) 63

81 3.8 Summary In this chapter, a matrix of sensitivity cases was presented to show the competing parameters of the fracture properties and their effect on the recovery and production performance of the process. More fracture stages per unit well length means higher injectivity and productivity. However, what determines the optimum number of fracture stages is the project economics. Adding more stages beyond a certain point is merely a recovery acceleration practice. In terms of fracture orientation, it was shown that fractures with the same conductivity in both longitudinal and transverse directions result in roughly similar production performance. That is to say, if the fracture conductivity is high enough, fracture direction is not important. The concept of dimensionless fracture conductivity was extended to include the effect of the number of fractures as well. It was shown that for C fdt >10 and x f >15 m, the total conductivity is large enough to achieve proper drainage capacity. Higher conductivity does improve the production rate; however, the improvement is not substantial. The recovery factor vs. pore volume injected, and SOR vs. recovery factor plots show that the sweep efficiency and ultimate recovery factors of most sensitivity cases are comparable, except for the cases with very low total fracture conductivity. It can be argued that using MSFHW in SAGD can efficiently recover oil from stratified pays as long as sufficient fracture conductivity has been achieved. Number of stages, longer and more permeable fractures will accelerate the recovery. However, even the lower range of these variables can still offer fairly decent processes in terms of the energy efficiency. 64

82 65

83 CHAPTER 4: SOME GEOMECHANICAL ASPECTS OF HYDRAULIC FRACTURING IN OIL SANDS 4.1 Introduction The feasibility of MSFHW relies on successfully creating vertical fractures in the oil sands. In the absence of multiple vertical fractures that are sufficiently propped across the shale layers, the access to the upper sand layers would be hindered. Interestingly, there has been a good number of fracturing field trials in the Alberta oil sands, although most of them predate the advent of horizontal drilling. A select number of these fracturing trials are reviewed in this chapter and the learnings are compared to general stress state in the oil sands in Western Canadian Sedimentary Basin (WCSB). Later on, the mechanical properties of oil sands and shales, as well as sand-shale interfaces are reviewed. One of the most important questions in this work is to study the condition of vertical fracture when approaching a sand-shale interface. A few interface crossing criterion are reviewed. By using the mechanical properties, fracture crossing across a conceptual, yet realistic, sand-shale interface is studied along with its sensitivity to main physical parameters. These results are then compared to results of a numerical simulation using a coupled reservoir-geomechanics-fracturing simulation software. Finally, the condition of the fracture initiation from a horizontal wellbore in different stress regimes is reviewed. 66

84 4.2 Stress regimes in the Alberta oil sands and its implications in hydraulic fracturing In this section, the stress regime in the Alberta oil sands and its relationship with depth is discussed. Generally speaking, the stress regime in Western Canadian Sedimentary Basin (WCSB) is of a compressive nature. On a continental scale, the WCSB lies within the North American Mid-Plate Stress Province. It is characterized by NE to SW trend for maximum horizontal stress [27]. There are indications that maximum principal stress in WCSB is usually in horizontal direction [24]. However, towards the plains and away from the Rocky Mountains, the maximum principal stress could be in vertical direction [28]. In a typical stress regime, especially in the WCSB context, one of principal stresses is usually vertical (or close to vertical) and the other two are horizontal (or close to horizontal). It is accepted that the direction of hydraulic fracture propagation is perpendicular to the direction of the minimum principal stress. As a result, the defining factor in the direction of hydraulic fracture (horizontal extension versus vertical extension with any angle in the horizontal plain) depends on the magnitude of the vertical stress compared to minimum horizontal stress. Equation ( 4.1) summarizes three possibilities for principal stress regimes and the resulting hydraulic fracture directions. V H Vertical Fracture H h V h Vertical Fracture ( 4.1) Horizontal Fracture H h V 67

85 There is a bit of gray area when σ V and σ h are approximately equal. In such a case, the local factors such as pore pressure depletion and wellbore stress concentration become pivotal for the direction of fracture initiation and extension. The author believes that the conditions encountered in the GLISP pilot [29] and the Numac fracture test [30] to be discussed in the next section, are such that the vertical and minimum horizontal stresses are approximately equal. The shallow depth of Alberta oil sands makes the conclusions about the fracture direction anything but straight forward. The stress regime in shallow deposits in Alberta is affected by the recent glacial history of Canada. Alberta was covered by 3-4 km of ice in the last episode of the Wisconsin Glaciation (30,000 to 10,000 years ago) [31] that had profound effects on the today terrain of the region as well as stress conditions [32]. The weight of the ice had elevated vertical and horizontal stresses, but later was lifted from the overburden. Hence, the vertical stress has been significantly reduced, while the horizontal stresses could not relax proportionally since they are bound by other tectonic forces. This condition is depicted in Figure 4.1. The melting of the glacial ice has formed a crossover point between the vertical and minimum horizontal stresses in shallow depths. 68

86 Figure 4.1 Depiction of the vertical and minimum horizontal stresses in a) A tectonically relaxed area, b) In an area with a geological loading history such as glacial ice cap [33] Figure 4.2 Stress measurements in Alberta oil sands (the slope of the line is 24 kpa/m) [34] Bell and Babcock gathered and analyzed a large number of stress measurement throughout Alberta from various sources [34]. The measurement pertaining to the oil sands are illustrated in Figure 4.2. The slope of the dashed line in the figure is 24 kpa/m, suggesting that in any measurement point below the line, the least principal stress is horizontal. However, one could argue that a 24 kpa/m gradient for vertical stress in shallow 69

87 depths is excessive, and a kpa/m range is more appropriate. The measurements between kpa/m fall within the above mentioned gray area in which other factors could determine the fracture orientation. The field observations (cases 5 to 8 at depths of m) reviewed in the next section of this chapter show strong evidence that the crossover point of the vertical and minimum horizontal stress in the Alberta oil sands (particularly in the Athabasca region) is around the 300 m depth and any fracturing project targeted at depths greater than the 350 m has a good chance of developing vertical fractures which is a pre-condition for applicability of MSFHW technology in SAGD. Another factor in Alberta oil sands is that the reservoirs are normally underpressured. To the author s knowledge, the initial pressure in many Alberta oil sand pools is in the range of 65-75% of their corresponding hydrostatic pressure. Furthermore, many of these pools have experienced cap gas depletion in recent decades [35]. The lower than expected initial pore pressure can relax the horizontal stress in the pool and promote vertical fracture growth. It is needless to say that site specific stress and fracturing measurement is required before planning a pilot project of this nature. Next section will review some of the field trials conducted in Alberta oil sands over the years. 4.3 Hydraulic fracturing trials in the Alberta oil sands Many hydraulic fracturing attempts have been made in Alberta since the 1960 s in various pilot projects. Some of them failed and some showed good potential. One can see 70

88 the step by step learnings and improvements that were applied in the future pilots [36, 37]. Some of these attempts predate the advent of horizontal well drilling technology and vertical wells are used in these pilot schemes. The recovery processes were one or a combination of cyclic steam stimulation, steam flooding and in-situ combustion. The key to success of steam flooding and in-situ combustion in an oil sand reservoir with almost immobile bitumen was shown to be establishing effective communication between the injector and the producer to start up the process. That is where the hydraulic fracturing came in as a solution to establish the initial communication. After an area of hot communication between the wells was established, the recovery process would continue to determine the long term success of the process. Also some oil sand deposits are unconsolidated and therefore it was initially questioned if hydraulic fracturing can be induced in such material. 4.4 In-situ oil sands projects with hydraulic fracturing in shallow depths Muskeg River Pilot [38] In the period of 1957 to 1962, Shell Canada Resources implemented a vertical steam flood pilot in the Muskeg River area in Alberta at the overburden depth of about 50 m. The initial communication between the injector and producer was achieved by a hydraulic fracturing job at the base of the oil sand deposit. Due to shallow depth of the deposit, the minimum stress was the vertical stress and hence the fracture was horizontal. With the help of hydraulic fracture, steam was successfully injected below fracture pressure, a fairly large area of hot communication was developed, and hot bitumen was 71

89 produced from the production well. However, severe rock-fluid interaction caused skin damage around the production well which required repeated clean ups. The operation was stopped in Steepbank Pilot [39] In the period of 1966 to 1969, Fina Oil Company implemented a vertical steam drive pilot to recover bitumen from Steepbank lease in Alberta at an approximate depth of 63 m. Similar to the Muskeg River Pilot, the idea was to establish the initial communication between the injector and producer with horizontal hydraulic fractures. Steam injection would then commence to establish hot communication between the wells. Two separate 5-spot patterns were tested and the fracture orientation in both cases was horizontal. In the second pattern, proppant sand was used in the fracturing job at the rate of about 1 m 3 /min which significantly helped the initial communication. Unfortunately, the pilot was halted prematurely due to steam breakouts around some of the wells. It was believed that the steam breakout was due to bad cementing quality outside of the casing. The Steepbank pilot experience showed that creating propped hydraulic fractures in oil sands can be achieved at relatively high injection rates which could overcome the fluid leak-off to the formation Mobil Oil Canada Athabasca Pilot [40] From 1962 to 1965, Mobil Oil Canada Ltd. conducted an in-situ combustion pilot in a 0.9 acre 9-spot pattern in the Athabasca oil sands region at the overburden depth of about 72

90 115 m. To establish the initial communication, the formation was fractured and propped before injection of combustion air was started. Once again, the fracture orientation was horizontal and monitoring the air injection rate and pressure indicated that the fractures remained open allowing the combustion process to be maintained. The production response was promising; however, some wells were damaged by very hot operating conditions. The Mobil Oil Athabasca pilot is another example in which propped fractures were successfully created and remained conductive throughout the process Kearl Lake Pilot [37] In the period of 1981 to 1991, Canterra operated two steam flood pilots in Kearl Lake area along with Tenneco and AOSTRA (Canterra was acquired by Husky Oil in 1988). Initially, the pilot consisted of an enclosed 7-spot steam flood, A-pattern at overburden depth of about 160 m. Similar to the previous cases; the initial communication was to be established by hydraulic fracturing in the bottom of the lower zone with cold water. The intent was to create horizontal fractures and produce a zone of shear failure and dilated porosity (without proppant). The plan was to follow it up by hot water stimulation until heat communication could be established across the lower part of the formation. The results indicated that the initial communication was established successfully. However, communication worked its way up, even above a shale break between lower and mid McMurray formation. Although significant volume of bitumen was produced, after the two McMurray packages started to communicate, a significant volume of steam was lost to outside of the pattern. 73

91 The experience of the A-pattern was used to design a new 9-spot steam flood pilot nearby, called the B-pattern. The goal was to reassure the heated fluids were captured within the pattern. The start-up plan was similar to the A-pattern with a few twists. For example, the fracturing with cold water was done at a very high rate (about 15 m 3 /min) with rapid conversion from cold, to hot water to steam injection. These changes reduced the time to establish strong heat communication from 13 months down to 5 months. Interestingly, although fracturing was mainly horizontal, there were some vertical components that broke the shale layer between the lower and mid-mcmurray formation and resulted in rather fast steam migration to the mid-mcmurray. The depth at which the fracturing jobs were targeted was about 225 m which is deeper than the previous pilots. It should be noted there were more in-situ pilots conducted in shallow overburden depths by various operators since the 1960 s [36, 40, 41]. Some of them did not utilize hydraulic fracturing as part of their process, and in some others, there were not enough technical details available to determine the exact strategies used to establish the initial communication between the wells. 4.5 In-situ oil sands projects with hydraulic fracturing in greater depths Grigoire Lake pilots [40] Amoco ran a series of pilot trials of various recovery processes in Grigoire Lake in the Athabasca oil sands between 1957 and The overburden depth in this project was 180 to 200 m which was deeper than some previous trials, but still was believed to be at 74

92 depth suitable for creating horizontal fracturing. A short chronological review of these pilots is presented in the following. In the years 1958 to 1961 reverse combustion trials were carried out in the pilot based on promising laboratory results. The field trial was done in a half-acre 5-spot pattern. The field test indicated that the reverse combustion could be sustained only if air was injected at above formation fracture pressure. The fracture(s) was believed to be horizontal and inter well communication was obtained as a result, aided also by the fairly short well distance. The operating philosophy was changed to combination of forward combustion and water flooding (COFCAW) during the period of , again in a half-acre 5-spot pattern. The water flooding was meant to carry the combustion heat deeper into the formation. The performance of the process was fairly good. It was estimated that 65% of the bitumen in the pattern was heated of which 50% was produced. In order to verify the COFCAW process in the commercial level, the size of the pattern was expanded to 10 acre patterns in trial. The increased distance was shown to be too large to control the combustion process. In 1976, Amoco and AOSTRA teamed up to build on the previous trials and pilot test more ideas. To start, the COFCAW process was tried in nine 2.5-acre 5-spot patterns [42]. The final pilot test was the Grigoire Lake In-situ Steam Pilot (GLISP) which was executed in [43]. The GLISP pilot was an inverted vertical 4-spot pattern with multiple observation wells, as shown in Figure 4.3. In GLISP Phase A, steam stimulation 75

93 followed by steam flooding was examined. The flood was continued in GLISP Phase B with steam additives and low temperature oxidation trials. Overall, GLISP was one of the most successful in-situ recovery methods tested in Athabasca oil sands with vertical wells. Similar to previous pilots, the initial communication between the wells was created with hydraulic fracturing. A very detailed monitoring plan was put in place to study the fracture mapping and geometry which was the key to the success of the project. The data collection and analysis consisted of tilt meters data, bottom hole pressure, temperature profile, tracer survey and signs of inter-well communication in the injector, producers, and observation wells. The collected information was analyzed independently and as a whole, and many possible fracture geometries were suggested that could explain at least some of the observations [29, 44]. For example, single horizontal fracture, single and multiple vertical fracture(s), T, inverted T, inverted H, U shape and S shape fracture geometries were all assumed as the potential geometries. They were then examined against the pressure response, tilt meter records and tracer readings to see which one of the assumed fracture geometries could fit the data the best. It was suggested that fracture(s) have a complex geometry consisting of vertical and horizontal components. While part of the fracture is horizontal, it was doubtful that a horizontal fracture can be propagated across the full spacing between the injector and the producer. The most plausible fracture geometry was suggested to be a U-shaped fracture with the injection well close to one of the vertical components as depicted in Figure 4.3. The finding was also confirmed with the temperature profiles in the observation wells after steam injection started. The temperature response indicated that heat has risen in the 76

94 reservoir through a fairly thick shale layer between lower and middle McMurray formations and has reached the production wells on top of the formation. Figure 4.3 GLISP pilot well configuration and fracture geometry ( )[29] Surmont CSS pilot [45] In the summer of 1979, Gulf Canada Resources conducted a cold water fracturing test in a vertical well with 308 m of overburden material. This test was executed prior to commencing steam injection above fracture pressure to the well. This well was part of a six vertical well pattern of a vertical cyclic steam stimulation pilot located in the Surmont lease in the Athabasca oil sands. The surface area around the well was equipped with a matrix of tilt meters to monitor the shape and size of the fracture with time. The overburden stress was estimated to be just above 6,000 kpa (~19.5 kpa/m) at top of the formation. The breakdown was achieved at 9,300 kpa (~30 kpa/m) and then the injection pressure dropped to 5,500-6,000 kpa while the hydraulic fracture was propagating. The propagation pressure and the tilt meter measurement suggested that the 77

95 orientation of the cold water fracture was vertical. After switching to steam injection, the injection pressure surpassed to values larger than the vertical stress to above 7,000 kpa indicating the fracture orientation flipping to horizontal direction. Figure 4.4 shows the pressure response of this fracture test. Analyzing the tilt meter records suggested a penny shaped fracture with a m radius away from the well. The above finding was also confirmed by history matching of the pilot performance using a (then) state of art numerical simulation model [46]. Figure 4.4 Pressure vs. time in a single well hydraulic fracture test followed by steam injection in Surmont area, Gulf Canada Resources, 1979 [47] 78

96 4.5.3 Pony Creek pilot fracture test [40] In the period of 1964 to 1965, Atlantic Richfield (ARCO) performed a detailed field test in Pony Creek area in Athabasca oil sands with overburden depth of m. The goal was to determine if horizontal fractures can be propagated in such depths as a means to establish a communication path between the injector and producer in a vertical steam flood scheme. Prior fracturing experience in Athabasca oil sands was indicating that fractures would be horizontal in shallower depth. ARCO was trying to push the envelope and test deeper intervals with different fracturing strategies designed to create horizontal fractures. For example, cold water vs. gelled liquids was tried, fractures with and without proppant, and fractures with various injection rates (from 15 to 1,200 bbls/hr) and various sizes (220-3,900 bbls) were pumped. Luhning [48] summarized different practices ARCO experimented with, which all resulted in a vertical fracture with cold water. One of the hopeful strategies was to pre-pressurize the formation by injecting cold water at low rates and below breakdown pressure to alter the local stress state and encourage initiation of a horizontal fracture. However, due to very low water injectivity in the oil sand zone, it is believed that the area with increased pore pressure was too small to affect the fracture orientation which remained vertical Numac fracture test in Surmont [30] In 1977, Numac and AOSTRA took on an experiment to hydraulically fracture McMurray Tar Sands at Surmont area of Athabasca oil sands. The main objective was to determine whether a reliable technique can be devised for creating horizontal fractures 79

97 within oil sands to establish initial communication in vertical wells drilled on a commercial spacing. Prior experiments were reviewed to refine this experiment [48] which indicated that a cold water fracturing attempt in depths greater than 365 m would result in a vertical fracture in Athabasca oil sands. The overburden depth in this location was between m which would provide the desired refinement for the fracturing design. In this trial, three hydraulic fractures were pumped at the well W4; one initiated at a depth of 306 m and the other two at a depth of 352 m. An array of observation techniques such as temperature and pressure observation wells, tilt meters and surface elevation survey, as well as injection pressure and temperature were used to monitor the fracture shape and orientation. In the first fracturing job, hot gel with cp viscosity was used at the depth of 352 m. The injection rate was about 0.3 m 3 /hr, with total injection volume of 76 m 3. A second fracture was initiated using green dyed grout at the same depth with 1.0 m 3 /hr injection rate and total injection volume of 254 m 3. The third fracture was initiated at 306 m using red dyed grout. The reason for using grout slurry was to retain a permanent record of the fracture. After the fracturing job, seven post fracture evaluation wells were drilled and cored in an attempt to intersect the grout fractures. The injection pressure response of the first fracture was indicative of a horizontal fracture. However, after a while there was a sudden drop in injection pressure suggesting that the fracture has re-oriented to vertical. The pressure response of the two grout fractures was indicative of a horizontal fracture while the fracture propagation pressure was changing frequently, suggesting a complex geometry. 80

98 It was hoped that the post fracture cores would confirm the findings of the other observation tools, since the core observations are rather undisputable. The location of the observation and post fracture wells is shown in Figure 4.5. No grout was present in wells E1 and E2 indicating that no fractures intersected them in cylindrical spaces. The grout in E3 was red and hence part of the upper grout fracture. The thickness of grout in E3 was characteristic of an elastic deformation. The E3 well was 7.6 m away from the injection well, and the location of the grout was 5.6 m above the injection point. The grout location along with the injection pressure response of the upper grout fracture suggests a bowl shape fracture which started off horizontal, but was on its way to vertical orientation. Figure 4.5 Left: Location of the Numac pilot wells and surface uplift after the second grout fracture [30]; Right: Scaled model of the five cored wells from left to right E3, E2, F1, E4, and E1. Solid black lines represent grout filled fractures, dashed lines show hairline fractures, injection points are marked by arrows [49]. 81

99 Overall, some important conclusions were made from the Numac fracture test. First of all, the fracture behavior in oil sands deviates from hard rock fracture theory to some extent. In the depth range of this experiment, complex fracture geometries consist of unsymmetrical vertical and horizontal fractures, and hair line fractures can co-exist. Also, the pressure response from the first grout fracture as well as surface elevation changes showed that major plastic deformation could occur especially in the vicinity of the injection well. In addition, these tests proved that discrete fractures are created in unconsolidated media when sufficiently high injection rates are used. The case studies presented above indicated that it is indeed possible to achieve a propped hydraulic fracture within the oil sands. Some theoretical and laboratory works, and some field cases [50] found in the literature may argue the contrary. However, the detailed conditions of the experiments or field cases, such as reservoir fluid viscosity, injection rate and managing the leak-off are very important and could make the difference. Another issue that should be discussed here is that northern area of the Athabasca oil sands region is found in sands at shallow depths that favor horizontal fractures or rather than vertical fractures meaning that this method is not applicable in them. Deeper Athabasca sands towards the south, Cold Lake and Peace River deposits are typically deep enough to favor vertical fracture and make this technology applicable. 4.6 Mechanical properties of oil sands and shales A big portion of the Alberta oil sands are made of loosely consolidated sandstones. The dominant mineral in the sandstones is quartz. The cementing material could also be 82

100 quartz or clay minerals or carbonates [51]. At low confining pressure, sandstones typically show non-linear stress-strain behavior. For weak sandstones in which many of the grain contacts are not cemented, the grain contact itself shows a non-linear behavior [52]. This type of uncemented sandstones is common in Alberta oil sands, and their behavior is typical of soil rather than that of rock. Weak sandstones exhibit dilatant behavior at low effective stress condition. The typical friction angle is near 30 o [53] (pp. 125). In the hydraulic fracturing process with high pressure fluid injection, the effective stress at the fracture face is essentially zero. Fluid flow into the rock matrix causes rock failure and/or dilation which in turn increases the porosity and permeability of the region around the fracture [54]. The post failure behavior of the Alberta oil sands is usually characterized by strain-softening ductile behavior [55]. A robust constitutive model for Cold Lake oil sands have been proposed by Wong et al. [56]. As mentioned before, one of the major challenges in in-situ thermal recovery methods is overcoming the flow barriers in the vertical direction. In the context of Alberta oil sands, many types of such barriers can exist. These flow barriers are generally made of very fine sands, or fine clay materials. Some of the more common cases encountered are siltstones, mudstones, and shales. Siltstones are basically very fine grained sandstones with some clay material, but are more compact and highly cemented together. Shales are sedimentary rocks with a large percentage of clay materials in the structure. Shale structure is strongly anisotropic which can be seen through thin bedding planes of lamination. Mudstones are fine textured 83

101 sedimentary rocks similar to shales, but have higher levels of compaction with less obvious bedding planes. In terms of mechanical properties, siltstones are similar to sandstones but with cemented fine grains, hence harder. Shales mechanical properties are more complex and depend on the type of dominant clay mineral(s) (kaolinite, smectite, illite) in the rock. In particular, the amount of absorbed/bond water present within the mineral and its surfaces, and the size of crystals within the material could significantly impact the stiffness of the shales [53] (pp 128). Another typical mechanical property of shales is large values of Poisson s ratio which can be as high as 0.45 in some cases [57]. Some typical mechanical properties of shales and sandstones are summarized in Table 4.1. Table 4.1 Some typical shales and oil sands mechanical properties Rock Type Formation E (MPa) Poisson s Ratio v Reference Shale Colorado group [57] Shale Grand Rapids [57] Shale Clearwater [58] Shale Grand Rapids [59] Shale Joli Fou [60] Sandstone Clearwater [57] Sandstone unknown [61] Sandstone Clearwater [59] Sandstone Clearwater [59] Sandstone Berea [62] Sandstone McMurray [58] Sandstone Grand Rapids [60] 84

102 4.7 Hydraulic fracturing crossing sand-shale interfaces One of the most important questions in this project is to determine whether vertical fractures that propagate through sand body and shale layers of the oil sand zone can be created. In general, when a propagating vertical fracture approaches a sand-shale interface, four potential situations (or a combination of them) may occur: Arrest, Diversion, Jog or Penetration [62] as illustrated in Figure 4.6. Creating vertical fractures through interbedded pays can happen under penetration or jog scenarios in contrast to slip scenarios. Therefore, it is critical to describe the condition under which the fractures can extend into the shale layers, and from shales to sand layers. Figure Potential outcomes of a propagating fracture to an interface [62] 85

103 As described above, a fracture growing towards an interface (bi-material interfaces, weak planes, bedding planes, natural fractures) may either cross the interface or get contained if interfacial slip occurs. The issue of fracture containment in the geological layers has been discussed in several works since the 1970s [63-66]. A range of factors determine the state of fracture approaching an interface such as the rock mechanical properties and the state of principal stresses. Warpinski et al. [67] reviewed some field scale fracturing jobs and showed that the difference in material properties across the interface are not sufficient to determine the fate of fractures at interfaces. They concluded that in-situ stress state is predominantly influencing a fracture crossing an interface, particularly when a strong stress discontinuity is present [67]. Teufel and Clark conducted a series of experiments to determine the conditions that can contain the vertical growth of hydraulic fractures across interfaces in rock specimens [68]. They determined the most important factors to be weak interfacial shear strength of the interface, and sharp increase in minimum horizontal stress in the top layer, with more emphasis on the latter. They also showed that differences in elastic properties of the layers matter only if they cause a difference in the minimum horizontal stress value. Lam and Cleary [69] argued that when a vertical fracture reaches an interface, some slippage would occur first. In this situation, they found that the point of intersection of the hydraulic fracture with the interface has the highest tensile stress after slippage (as opposed to the points at the ends of the slippage zone). The question to be answered is if the crack encountering a stiff adjacent stratum would break through the interface or not. If it does, 86

104 the most likely failure place would be the point of intersection. In their point of view, fracturing across an interface is a re-initiation after slippage. Lam and Cleary developed a numerical code to determine various scenarios that could happen, however they also suggested a simple re-initiation criterion that is valid for many practical situations. The reinitiation condition is illustrated in Figure 4.7 in which T o is the tensile strength of the upper material. This criterion will be evaluated further later on. Figure 4.7 Simple criterion for crack re-initiation [69] Renshaw and Pollard [70] developed a more involved criterion for fracture propagation across frictional interfaces in linear elastic materials. They also conducted experiments similar to that of Teufel and Clark and verified their own proposed criterion. Renshaw and Pollard s criterion was developed for cohesionless interfaces. This is 87

105 arguably not a bad assumption for a hydraulic fracturing scenario in which the fracturing fluid could penetrate between the bedding layers and weaken the cohesion of the interface. Gu et al. re-arranged the Renshaw and Pollard criterion for effective stresses with compressive stress being positive [71]. Elsewhere, Gu and Weng extended the criterion to interfaces with cohesion and for non-orthogonal geometry conditions [72]. The extended Renshaw and Pollard criterion will be used as the basis of analysis in this work since it contains the most important parameters that determine the state of fracturing through an interface. The basis of the Renshaw and Pollard criterion can be explained simply with stress state depicted in Figure 4.8 and mechanical properties of the upper layer and the interface. The competing forces are described below. In this figure, and the following equations σ xx (max), σ xz (max) and σ zz (max) are the maximum magnitudes of the components of the effective stresses and shear stress acting on the frictional interface. For a fracture to propagate across an interface, the induced tensile stress in the upper material should be equal or greater than its tensile strength, i.e. σ xx (max) T o ( 4.2) written as The maximum shear stress that the interface can endure before it slips can be σ xz (max) τ o + μ. σ v ( 4.3) where τ o is the interface cohesion and μ is the coefficient of friction. 88

106 Figure 4.8 Schematic explanation of the Renshaw and Pollard crossing criterion (2D) The extended Renshaw and Pollard criterion (RPC) suggests that a vertical fracture crosses a horizontal interface if the crossing stress ratio defined by the left hand side of Equation ( 4.4) is larger than the right hand side of the Equation ( 4.4): τ o 0.35 μ + σ v T o + σ > μ h 1.06 ( 4.4) where σ v and σ h are vertical effective stress and minimum horizontal effective stress respectively. Please note that the compressive stress is assumed to be positive. The constants in the Equation ( 4.4) are calculated from the maximum magnitudes of the components of effective stress that act along the frictional interface (σ xx (max), σ xz (max) 89

107 and σ zz (max) in Figure 4.8) within the fracture process zone. The fracture process zone is the area within a critical radius in which the stresses are of sufficient magnitude for inelastic deformation [70]. The extended Renshaw and Pollard criterion can be reduced to Lam and Cleary reinitiation criterion if the RHS of the equation ( 4.4) is approximated by 1.0 and the interface is assumed to be cohesionless i.e. τ o = 0. In other words, Lam and Cleary criterion is a simplified form of the RPC. While the extended Renshaw and Pollard approach will be used in this work, there are a few topics that require a discussion. Most hydraulic fracture models, especially the analytical ones, are based on Linear Elastic Fracture Mechanics (LEFM). LEFM usually provides reasonable prediction for hard rocks; but it may not be as reliable for predicting fracture geometry and propagation in unconsolidated formations. This topic has been discussed in detail by Wang and Economides [73]. Another potential problem with the LEFM approach is the issue of scale. Lam and Cleary suggested that crack re-initiation in the second material is influenced also by the presence and size of the material flaws. However, the LEFM approach can represent the flaws only in continuum sense by the value of fracture energy or critical stress intensity factor. This could be less applicable at the scale of usual flaws in rocks. Thiercelin et al. used the non-linear fracture mechanics concept for the interface problems [62]. They state that the condition necessary for fracture propagation across an interface is that the interface must have sufficient shear capacity to develop a tensile stress in the second material equal to the tensile capacity of that 90

108 material. Therefore, their method is conceptually analogous to that of extended Renshaw and Pollard approach. 4.8 Examining the fracture propagation criterion in typical oil sand formations In the previous section, it was established that the governing variables that determine whether a fracture crosses a rock interface are: 1. Pore pressure to calculate effective stresses; 2. Vertical stress at the interface σ v, and horizontal stress just above the interface σ h ; 3. The interface cohesion τ o, and coefficient of friction μ; 4. The tensile strength of the upper material T o ; The value of each of the above parameters is needed to determine the state of the fracture. Below, some discussions and bounds are provided for these parameters Pore pressure The reservoirs in Alberta oil sands are generally under-pressured, most are in the range of 65-75% of their corresponding hydrostatic pressure. Furthermore, many of these pools have experienced cap gas depletion in recent decades [35]. Some typical pore pressures in Alberta oil sand reservoir are summarized in Table 4.2. Table 4.2 Initial reservoir pressure of various oil sand reservoirs in Alberta Region Formation Pool Depth Pressure Gradient* Reference Cold Lake Clearwater Cold Lake m 3,000 kpa 7.1 kpa/m [74] Athabasca McMurray Christina Lake m 2,500 kpa 7.1 kpa/m [22] Peace River Bluesky Carmen Creek m 4,000 kpa 7.0 kpa/m [75] *The pressure gradient calculation is based on the average depth range 91

109 4.8.2 Stress gradients A typical value for the vertical stress is around kpa/m which can be confirmed from density logs. Multiplying this gradient by the depth of formation provides an estimate of the value of vertical stress, which is essentially equivalent to the weight of the overburden. In a tectonically relaxed environment and for linear elastic material, one could estimate the minimum horizontal stress from: σ h = v 1 v (σ v P p ) + P p ( 4.5) where v is the Poisson s ratio of the rock layers and P p is the pore pressure. One of the major sources of horizontal stress contrast between the geological layers is the difference in their Poisson s ratios. Using the data from Table 4.1, let s assume that the Poisson s ratios of sand and shale are 0.25 and 0.35 respectively, pore pressure is about 2,500 kpa, and the vertical stress gradient is 21 kpa/m. The minimum horizontal stress can be then estimated from Equation ( 4.5) for a tectonically relaxed environment, as illustrated in Figure 4.9. It can be observed that a 0.1 difference is Poisson s ratio of the rock material can easily create a 1,000 kpa stress contrast in a typical Alberta oil sand deposit. In the example presented in Figure 4.9, the thickness of the shale layer within the sand is 5 m. It is plausible that a thick and continuous shale layer with a high Poisson s ratio could create such a stress contrast. However, in many cases in interbedded sands within the reservoir, the more prevalent condition is multiple thin shale lenses that are not laterally continuous. It is intuitively unlikely that such thin shale lenses could create sharp horizontal stress contrasts. Therefore, in the absence of field measurement of minimum horizontal stress in 92

110 Depth (m) the interbedded shale layers, it is more plausible to assume that the thin shale layers within sands have a smaller or no stress contrast compared to what would be indicated by their Poisson s ratios difference. There are many field measurements available for the minimum horizontal stress in the Alberta oil sands. Depending on the depth and geographical location, a gradient range between kpa/m has been reported [76] Pore pressure σv σ v σh σ h Stress/Pressure (kpa) Figure 4.9 Horizontal stress estimation for a tectonically relaxed area using Equation ( 4.5) Interface properties The mechanical properties of the interface (interface cohesion and coefficient of friction) can be measured with Direct Shear Tests on the sand-shale contact. Such data is hard to find in the context of oil sands with shale layers. However, sand-shale interface mechanical properties have been reported where the shale layer acts as cap rock on oil sand formation. Smith [59] conducted a series of experiments to measure the sand-shale 93

111 interface properties in Cold Lake area. He estimated the cohesion value is in the range of kpa, and the friction angle φ is about 20 o. Smith conducted his experiments on actual core samples from Cold Lake oil sand pool. One could argue that such samples would be disturbed during core recovery process regardless of how carefully the coring takes place. Smith also conducted the direct shear experiment on the Grand Rapids shale beds and observed higher friction angles. Similar observation has been made on Joli Fou shale in the Cold Lake area [60]. Direct shear tests has also been conducted in sandstonelimestone [62] and mono-lithologic interfaces [68]. Overall, cohesion of kpa and friction coefficient of were observed. Please note that these experiments were conducted on already disjointed samples. Table 4.3 summarizes the interface properties. The Smith s reported values will be used in this work as it is the most relevant available data for the premise of this study. Interface Type Table 4.3 Examples of mechanical properties of various rock interfaces Cohesion τ o (MPa) Coefficient of Friction μ * Reference Sandstone (Clearwater) Shale (Grand Rapids) [59] Shale (Grand Rapids) Shale (Grand Rapids) [59] Shale (Joli Fou) Shale (Joli Fou) [60] Sandstone (Berea) Limestone (Indiana) [62] Sandstone (Berea) Sandstone (Berea) [68] Sandstone (Arizona) Sandstone (Arizona) [68] Sandstone (Tennessee) Sandstone (Tennessee) [68] Limestone (Lueders) Limestone (Lueders) [68] * μ = tan(φ) where φ is the friction angle Tensile strength The tensile strength of the upper material is another parameter that influences the propagation of the fracture through the interface. The most relevant tensile strength 94

112 measurements that were found for shales suggests a kpa range [60]. It is very difficult to measure the tensile strength for unconsolidated sands in the lab since at zero confining stress, a clean sample may not hold intact. Khodaverdian et al. [77] studied hydraulic fracturing in a large radial model in the lab on clean 200-mesh sand. However, linear extrapolation of their data to zero confining stress resulted in unreasonably high tensile strength and was deemed not representative. In the absence of direct measurement, one could argue that unconsolidated sandstone should have a very low tensile strength. More consolidated sandstones could have tensile strengths in the range of 1 to 10 MPa. Table 4.4 summarizes some reported tensile strengths of different rock samples. In the context of Alberta oil sands, rocks with higher tensile strength would be unusual. For this work, it was assumed that the tensile strength of the oil sands is 100 kpa. Having said that, the author believes that the oil sand saturated with heavy bitumen could exhibit larger tensile strength values in the context of hydraulic fracturing. Table 4.4 Tensile strength of various rock samples Rock Type Formation Tensile Strength T o (MPa) Reference Shale Joli Fou 0.20 [60] Shale Westgate 0.15 [60] Poorly consolidated rock 0.35 [61] Sandstone Berea 1.0 [62] Limestone Indiana 3.8 [62] Sandstone Berea 4.5 [68] Sandstone Arizona 8.4 [68] Sandstone Tennessee 10.8 [70] Limestone Lueders 3.5 [70] The values of the above parameters are now used to create a base case to test the extended Renshaw and Pollard criterion (RPC) as presented in Table 4.5. The goal is to 95

113 evaluate the state of a propagating vertical fracture as it reaches a sand-shale interface. Based on the assumed stress state and mechanical properties in Table 4.5, the criterion suggests that the vertical fracture would propagate from sand to shale and from shale to sand. This is a fairly realistic case that can exist in an actual field. However, changing some of the parameters could change the picture. For example, a higher horizontal stress gradient, or higher tensile strength of the shale or sand could predict a fracture arrest scenario. Table 4.5 Extended Renshaw and Pollard criterion test, base case Interface Depth (m) 350 Vertical Stress Gradient (kpa/m) 21 Min. Horizontal Stress Gradient (kpa/m) 16 Sand Min. Horizontal Stress (kpa) σ h ( an ) 5,600 Shale Min. Horizontal Stress (kpa) σ h ( hal ) 5,600 Vertical Stress (kpa) σ v 7,350 Pore Pressure (kpa) P p 2,450 Sand Effective Min. Horizontal Stress (kpa) σ h ( an ) 3,150 Shale Effective Min. Horizontal Stress (kpa) σ h ( hal ) 3,150 Effective Vertical Stress (kpa) σ v 4,900 Coefficient of Friction μ 0.3 Interface Cohesion (kpa) τ o 150 Tensile Strength of Shale (kpa) T o ( hal ) 200 Tensile Strength of Sand (kpa) T o ( an ) 100 RHS of Extended Renshaw and Pollard Criterion, Equation ( 4.4) Crossing Stress Ratio Shale Overlaying Sand Vertical Fracture Crosses from Sand to Shale? Crossing Stress Ratio Sand Overlaying Shale Vertical Fracture Crosses from Shale to Sand? μ τ o μ + σ v T o + σ h 1.61 TRUE μ + σ v 1.66 T o + σ h τ o TRUE 96

114 With this framework, one can easily generate multiple sensitivity cases for the parameters of the extended Renshaw and Pollard criterion (RPC). For easier illustration, the criterion can be re-arranged by defining an RPC ratio by Equation ( 4.6). An RPC ratio of larger than 1.0, indicates the fracture crossing the interface. τ o 0.35 μ + σ v μ > 1 ( 4.6) T o + σ h 1.06 To evaluate the sensitivity of the crossing criterion to governing properties, several cases were reviewed. Initially, the RPC ratio was calculated for changing values of friction coefficient from 0 to 1 as illustrated in Figure The base case data is such that for friction coefficients less than 0.25 (φ = 14 o ) the vertical fracture will be arrested at the interface. The figure also shows that for higher friction coefficients which could be the case for an undisturbed sand-shale interface it is more likely for fracture to cross the interface. Friction coefficient of 1.0 denotes intact rock. 97

115 Figure 4.10 Interface crossing check, sensitivity to the friction coefficient for base case data (Table 4.5) Figure 4.11 shows the RPC ratio for changing values of interface cohesion, keeping the other parameters at base values from Table 4.5. The stress state and mechanical properties in this particular case example are such that the fracture crosses the interface even if the interface is cohesionless. The tensile strength of the upper material works against the fracture propagation into the upper material as depicted in Figure Presence of a strong material layer in the fracture path could deter fracture growth. For the base case data, if a rock with tensile strength larger than 0.7 MPa (like many rock samples in Table 4.4) is in the fracture path, it can cause a fracture arrest or diversion scenario. 98

116 Figure 4.11 Interface crossing check, sensitivity to the interface cohesion for base case data (Table 4.5) Figure 4.12 Interface crossing check, sensitivity to shale tensile strength for base case data (Table 4.5) If the values of cohesion and tensile strengths can be neglected, or be small in comparison to effective stresses, the crossing stress ratio (LHS of the Equation ( 4.4)) can 99

117 be approximated by σ v σ h. Figure 4.13 shows the sensitivity of the RPC ratio to changing values σ v σ h for the base case data. The σ v σ h was altered by changing the minimum horizontal stress gradient in the range of kpa/m; in other words, there was no horizontal stress contrast above and below the sand-shale interface. Figure 4.13 Interface crossing check, sensitivity to σ v σ h ratio for base case data (Table 4.5) Existence of stress contrast across the interface could also affect the fracture path. As explained before, stress contrast could be caused by higher Poisson s ratio of the shale. Figure 4.14 illustrates that for the base case dataset, a 450 kpa stress shift could impede the fracture propagation. Negative stress contrast in the upper layer is also possible if upper material has a lower Poisson s ratio, although this would not negatively affect fracture propagation. 100

118 Figure 4.14 Interface crossing check, sensitivity to horizontal stress contrast for base case data (Table 4.5) The discussions and results presented above show that the extended Renshaw and Pollard criterion can be used to determine the likelihood of fracture propagation from a sand-shale interface. The precondition of such analysis, however, is to have representative geomechanical data for the rock materials and interface which are in many cases site specific. 4.9 Examining the fracture crossing criterion using a coupled numerical model The problem of the fracture at an interface can be also studied numerically. The advantage of such approach is that the poroelastic effects can be included more rigorously. A base case simulation was created in the GEOSIM software. GEOSIM is a coupled flow, stress, deformation and fracture propagation modelling package. The idea was to examine the fracture propagation at the sand-shale interface in a 2-D setting. Linear elastic 101

119 constitutive model without failure was selected using the rock mechanical properties from Table 4.1 and Table 4.5. A 2-D x-z grid was set up that includes one vertical fracture stage pumped from a horizontal well perforated only in the fracture block. The fracture was assumed to propagate in a vertical plane from a single perforation in the center of the model. The well was assumed to be connected to the reservoir grid only at the perforation interval (which is within the fracture plane/grid block). The well model is based on the Peaceman equation although a large PI multiplier was used to minimize the well connection pressure loss, effectively making the well a source for the block. The grid blocks sizes in horizontal direction (perpendicular to the fracture plane) were selected with large grid blocks that gradually get finer towards the fracture (in meters: 10, 5, 3.5, 2.25, 1.5, 1, 0.7, 0.4, 0.3, 0.2, 0.1, 0.05 : fracture block). In the vertical direction, large grid blocks were used outside the zone of interest. Within the target sand and shale, 1 m thick blocks were used which get gradually finer to 0.1 m at the sand shale interface. The boundaries of the model were all fixed (no movement or slip) except for the top of model which is free to move. The reservoir data were taken from Table 2.1 with the porosities set to 0.05 and 0.35 and permeabilities set to 10 and 2000 md for shale and sand respectively. The full input files for this 2D model are provided in Appendix II. The modeling of the fracture propagation in GEOSIM is different from conventional fracturing models and it is described by Ji et al. [78]. It is essentially a simplification of the fully coupled technique of embedding the fracture mechanics in the flow and geomechanical model developed later by Ji et al. [79]. Fracture is allowed to extend in a plane of grid cells, and it propagates into a new cell when the effective stress 102

120 perpendicular to the fracture plane becomes negative (tensile) and exceeds the tensile strength of the material. The permeability in the fracture plane is then increased as a function of effective stress. Additionally, the porosity can be also increased as a function of the stress to model the fracture volume (which was not used in this work). The permeability and porosity of the fracture is combined with those of the background media resulting in multiplier functions which are input in tabular form to the simulator. The shape of the functions can be derived from the consideration of 2-D fracture opening and their permeabilities as functions of net pressure in the fracture [80]. The technique has been validated against the classical fracture solutions [78, 80, 81]. Because the fracture is propagating through a fixed grid, there are grid effects which cause the effective stress in the fractured blocks to be more negative than what would be indicated by the tensile strength of materials. This effect can partially be controlled by the steepness of the multiplier function and by gridding. For modeling the interface crossing problem, the injection pressure was set at 6700 kpa (or 19 kpa/m gradient) which is above the minimum horizontal stress gradient (16 kpa/m). The viscosity of the injection fluid was set at 100 cp to capture the high viscosity of gel to reduce the leak off from the fracture. The fracture propagation path was defined by means of a vertical plane in the center of the model, with the gridding described above. Two tables of vertical permeability multipliers were applied as a function of effective stress in horizontal direction (one for sand and one for shale layers). The permeability multiplier functions are depicted in Figure The multipliers include some values larger than one at positive effective stresses which represent permeability enhancement due to dilation. The 103

121 reason for using two different functions for shale and sand is that the multipliers are applied to the absolute rock permeability, while the fracture permeability should be independent of absolute permeability. Hence, a factor equal to the ratio of absolute permeability of shale to sand (in this case 200) was used to create the shale multiplier function from that of sand for the negative effective stresses. A second simulation case was created with similar specifications to the base case except for a 500 kpa horizontal stress shift (increase) right above the sand-shale interface. This case can be compared with the interface crossing criterion sensitivity to stress contrast (Figure 4.14). Figure 4.15 Vertical permeability multipliers as a function of effective stress (log scales) Figure 4.16 shows the vertical permeability multiplier during hydraulic fracturing in different times for both base case (left) and the case with stress contrast (right) zoomed in around the injection well and the interface. The multipliers only act on the dedicated fracture plane with the grid block width of 5 cm. The high multiplier values are an 104

122 indication of fracture growth. In both cases, the fracture grows vertically in the sand and reaches the interface after a few minutes of injection. In the base case, the fracture grows into the shale after a pause to pressurize. However, in the case with stress contrast, the fracture is arrested at the interface and does not grow into the shale. The result is in agreement with the sensitivity case presented in Figure It should be noted that this comparison is only qualitative. The numerical solution solves a more complex problem (including poroelastic stresses, stress gradient with depth, etc.), but on the other hand it may require more detailed study to ensure numerical accuracy. Quantitative study was beyond the scope of this investigation and it is included in recommendations for further work. 105

123 Figure 4.16 Snapshots of vertical permeability multiplier in the fracture plane; LHS: Base case, no stress contrast, RHS: Case with stress contrast at the interface Figure 4.17 shows the horizontal effective stress (normal to fracture plane) for both cases. The effective stress calculation is mainly affected by poroelastic effect of fluid injection into the fracture plane and leak-off from fracture to the formation. 106

124 Figure 4.17 Snapshots of effective stress normal to the fracture plane; LHS: Base case, no stress contrast RHS: Case with stress contrast at the interface Figure 4.18 is a plot of horizontal effective stress at the first block in the shale on the fracture plane. In the base case (top), the effective stress becomes negative (tensile fracturing condition) after a short time, while in the case with stress contrast (bottom), the 107

125 effective stress remains positive and levels off after about an hour of injection, i.e., it is unlikely that the fracture ever crosses the interface in the latter case. This coupled simulation cases are complex problems since the effect of fluid leak off and hence poroelastic effects are captured. The constitutive model used does not include failure and therefore the modulus of the fracture blocks remains large even after the block practically fails. As the pore pressure builds up in the blocks (whether in the fracture blocks or in the neighbor blocks due to leak off), negative effective stress values would be calculated since the model assumes the fracture still possess the rock mechanical properties which is not accurate. At the same time, the fracture walls displacement is not modelled either. In reality, the fracture walls movement would compress the neighbor blocks and increase their stresses. 108

126 Figure 4.18 Effective stress normal to the fracture plane versus time for the first block in the shale layer; Top: Base case, no stress contrast, Bottom: Case with stress contrast at the interface Figure 4.19 shows the XZ shear stress in base case (left) and the case with stress contrast (right). The base case snapshots show the stress just before, and after the fracture crossed the interface. The case with stress contrast snapshots are at the same times as for the base case, while the fracture is arrested below the interface. It can be observed that while the shear stress is almost similar in both cases when the fracture arrives at the 109

127 interface, the fracture arrest in the latter resulted in higher shear stress concentration at the interface. The other point to note is that the shear stress level predicted by the simulator is likely lower than reality. This is because the only deformation captured in this treatment of fracturing comes from poroelastic deformation of the grid blocks, and the rock deformation due to displacement of the fracture walls was not accounted for. A more sophisticated treatment of fracturing is required to better capture all the physics of dynamic hydraulic fracturing growth. Figure 4.19 Shear stress state planar view; Left: Base case (before and after fracture crossed the interface), Right: Case with stress contrast at the interface (fracture arrested below the interface) 110

128 4.10 Fracture initiation from a horizontal well bore Up to this point, the underlying assumption was that a fracture has been initiated and is growing. It is desired to review the fracture initiation from a horizontal wellbore. Similar to previous chapters, two limiting cases are discussed: a. Horizontal well direction parallel to maximum horizontal stress (Figure 4.20 left) b. Horizontal well perpendicular to maximum horizontal stress (Figure 4.20 right) Figure 4.20 Expected fracture orientation from a horizontal wellbore; Left: Horizontal well parallel to maximum horizontal stress, Right: Horizontal well perpendicular to maximum horizontal stress For an impermeable rock, the break down pressure P br for an openhole horizontal well can be written as: a. Well parallel to σ H : P br = 3σ h σ v P p + T o ( 4.7) b. Well perpendicular to σ H : P br = 3σ H σ v P p + T o ( 4.8) where P p is formation pore pressure. By comparing the two equations, one can quickly see that the initiating transverse fractures (case b) requires a higher break down pressure than that of the longitudinal fractures (case a). 111

129 For a permeable rock, the break down pressure for an openhole horizontal well can be estimated by: a. Well parallel to σ H : P br = 3σ h σ v βp p + T o 2 β ( 4.9) b. Well perpendicular to σ H : P br = 3σ H σ v βp p + T o 2 β ( 4.10) where β is the poroelastic parameter and is defined by: β = α 1 2v 1 v α: Biot constant; v: Drained Poisson s ratio ( 4.11) This solution is taken from the analogous case of a vertical well and therefore assumes that permeability is isotropic (k v = k h ). The above equations were used to make an actual breakdown pressure estimate using the rock properties and stress conditions from previous cases. Table 4.6 summarizes the assumptions and results. As mentioned, the breakdown pressure estimate to achieve transverse fractures in an openhole horizontal well is expected to be higher than that of longitudinal fractures. However, the predicted values in Table 4.6 for transverse fracture breakdown are unreasonably high. As a result, for the case where the horizontal well is perpendicular to maximum horizontal stress, a complex fracture geometry near the wellbore might form before realigning to transverse direction. For fractures pumped through perforations in cemented liner completion, the near-well fracture geometry may be complex and break down pressure would be difficult to predict. However, when the fracture becomes large compared to wellbore, the overall shape of the fracture would be dictated by the far field stresses as shown in Figure The breakdown 112

130 pressure values estimated in Table 4.6 can be used only as upper and lower limits of the actual breakdown pressure in fracturing job design. Wang et al. [82] argued that the behavior of the poorly consolidated sandstones under different loading stress paths could change, especially in shallow depths. For example, during the drilling operation, if the mud pressure exceeds the elastic range of the rock, the above elastic equations used to predict the break down pressure may not be representative. In this case, different estimates based on elastoplastic models should be used. Wang and Dusseault have derived solutions for non-linear media and discussed this topic in detail [83, 84]. Table 4.6 Estimate of fracture initiation pressure for an openhole horizontal wellbore Horizontal well Depth (m) 355 Vertical Stress Gradient (kpa/m) 21 Min. Horizontal Stress Gradient (kpa/m) 16 Max. Horizontal Stress Gradient (kpa/m) 27 Min. Horizontal Stress (kpa) σ h 5,680 Max. Horizontal Stress (kpa) σ H 9,585 Vertical Stress (kpa) σ v 7,455 Pore Pressure (kpa) P p 2,520 Sand Tensile strength (kpa) T o 100 Biot Constant α 1.0 Sand Poisson s ratio v 0.3 Poroelastic constant β 0.57 Break down pressure for impermeable rock: a. Well parallel to σ H (kpa) 7,165 b. Well perpendicular to σ H (kpa) 18,880 Break down pressure for permeable rock: a. Well parallel to σ H (kpa) 5,768 b. Well perpendicular to σ H (kpa) 13,

131 4.11 Summary In this chapter, several aspects of the fracture mechanics important to the proposed MSFHW process were studied. First, a few hydraulic fracture field trials in the Alberta oil sands were reviewed. It was shown that it is possible to initiate and propagate a propped hydraulic fracture in the oil sands. However, to obtain a vertical fracture, the depth of the target formation should be higher than 350 m (suitable for southern parts of Athabasca deposit as well as Cold Lake and Peace River deposits). Site specific mini-frac tests are required to determine the exact stress state. Another important factor is the knowledge of the mechanical properties. A useful guide range was presented based on the data from literature. Similar to the stress state, site specific lab measurement is required to narrow down the prediction for any given project. Next, some interface crossing criteria were reviewed and the extended Renshaw and Pollard criterion (RPC) was used to study fracture propagation for a conceptual, but realistic sand-shale interface. Through sensitivity cases, it was shown that it is possible to induce vertical fracture across sand-shale interfaces within the expected range of properties and stress states. The prediction of the analytical method was also confirmed by numerical simulation conceptually. This finding is of utmost importance, as it is one of the pre-conditions of the feasibility of the project idea. Finally, the condition of the fracture initiation from an openhole horizontal wellbore in different stress regimes was reviewed. 114

132 CHAPTER 5: FRACTURING DESIGN, AND STRESS STATE DURING SAGD OPERATIONS 5.1 Introduction In the previous chapters, the desired properties of the fractures were discussed. In this chapter, some more detailed issues in design of the fracturing job and the state of the fracture after starting the steam injection will be covered. The most important topic is whether the desired fracture conductivity can be achieved and maintained. A selection of the proppants that could satisfy the conductivity requirement at the range of closure stress of the process, as well as the proppant concentration target will be discussed. After that, the factors that adversely affect the proppant conductivity will be addressed with a particular focus on proppant embedment, as well as a discussion on the effect of increased thermal stress on the proppant pack. Then, a fracture job modelling will be discussed to show that achieving proper proppant placement at target concentrations in the fracture is feasible. The chapter will be completed with a few proposed cemented and openhole liner designs that can be used to incorporate fracturing in SAGD injection wells. 5.2 Proppant selection In the previous chapters, it was shown that a high value of fracture conductivity is required for this process to work. Figure 3.12 and Figure 3.13 suggest that for a 1 cm fracture aperture, fracture permeability of 4,000 D is desired. This is equivalent to a 115

133 fracture conductivity of 40,000 md.m. Achieving such a high fracture conductivity value is challenging, but possible. There is a limited selection of proppants available in the market that can deliver this level of fracture conductivity. Some are sieved coarse grain sands (such as 6/12 Hickory sand) and some are manufactured coarse ceramics beads (such as CarboLite, CarboHSP and CarboProp 6/10 to 8/12). The cost of using ceramic proppants is typically higher, but they provide a higher and more uniform conductivity, with less degradation with increased stress. Figure 5.1 shows the normal effective stress (in this case minimum horizontal effective stress) acting on a propped fracture. After the fracturing job and placement of the proppants in the fracture, the fluid pressure inside the fracture starts to decline due to leakoff, or flow back. The pressure in the fracture is eventually equalized to the formation pore pressure, and the far field stress acts on the fracture wall and locks in the proppants. This closure stress reduces the porosity, permeability and width of the proppant pack. In a typical deep reservoir fracturing job (e.g., shale gas), proppant conductivity may be reduced by 80-90% as a result of closure stress level. In this project, however, the closure effective stress levels are much lower (typically below 10 MPa range) because of the shallow depth of the Alberta oil sands. 116

134 Hydraulic Fracture σ h Sand Shale Figure 5.1 Normal effective stress acting on a propped fracture There are standardized testing methods available to describe the effect of closure stress on proppant conductivity [85]. The proppant suppliers typically use third party laboratories to run the conductivity tests and use the data in the proppant data sheets. Figure 5.2 shows the proppant pack permeability as a function of closure stress for some natural and ceramic proppants that can provide the desired fracture conductivity. The sand proppants typically get packed and crushed at higher stresses because of their nonideal grain shapes, accompanied by severe loss of permeability. The manufactured ceramic proppants are less affected by the increase in the stress level. 117

135 Fracture Permeability (D) 14,000 12,000 10,000 8,000 6,000 4,000 2, Closure Stress (MPa) CarboProp 6/10 (Stim Lab 6.0) CarboHSP 6/10 (Stim Lab 6.0) ForeRCP 6/10 (Stim Lab 7215) CarboLite 8/12 (Stim Lab 99) ForeProp 10/14 (Stim Lab 7215) Brady Type Hickory Sand 6/12 Colorado Silica 8/12 (Stim Lab 6.0) ForeSand 12/18 (Stim Lab 7215) Jordan Sand 10/20 (Stim Lab 6.0) Figure 5.2 Fracture permeability of various proppants at 10 kg/m 2 concentration vs. stress [86] 5.3 Proppant concentration The proppant concentration is defined here as the amount of proppant per unit area of the fracture wall (one side only). This definition, also referred to as proppant coverage, is different from the proppant concentration during the pumping of the job (mass of proppant per unit volume of fluid). Fracture conductivity increases with higher proppant concentration. Achieving high proppant concentration requires placement of multi layers of proppant in the fracture after closure, which in turn requires higher viscosity fracturing fluid to create the fracture width required to transport the proppants. The relationship between the propped fracture width (w fp ) and the proppant concentration (C prop ) can be written as: 118

136 w fp = C prop (1 prop ). ρ prop ( 5.1) where prop is the proppant porosity, and ρ prop the proppant grain density [87](pp. 8-14). As an example, to achieve a 1 cm propped fracture width using an 8/12 CarboLite proppant with an average grain diameter of 2 mm, at least five proppant layers are needed. For this case, the proppant concentration should be about: C prop = w fp. (1 prop ). ρ prop = 0.01 (1 0.42) 2700 = kg m 2 To put a proppant concentration target range of kg/m 2 in perspective, one can consider a typical proppant concentration used in standardized lab measurements which is around 10 kg/m 2 (~2 lb/ft 2 ). In shale fracturing, typical concentrations placed are very low, resulting in fracture conductivities in md-m. A better analog for our application is fracturing high permeability formations in order to overcome near-wellbore formation damage. These fracturing jobs (called frac-packs ) are designed for short, highly conductive fractures. 5.4 Proppant pack conductivity loss After creating a fracture with the desired conductivity, various mechanisms could impair the proppant pack conductivity during the production phase. Fines migration, scaling on the proppants, proppant crushing and proppant embedment are some of the possible impairment mechanisms. Proppant scaling (or diagenesis) is the process of mineral scaling on the proppant pack that effectively reduces the conductivity of the pack. There are various factors that 119

137 affect the scaling, such as the type of the proppant, proppant coating, the ph of the fracturing fluid and the reservoir fluid and most importantly, the formation water mineral composition. It has been shown that scaling tendency is stronger in silica based proppants (sands) compared to ceramic proppants [88, 89]. Use of resin coated sands lessens the scaling tendency, although they are not applicable in a steam based process because of the elevated operational temperatures. Proppant crushing is another conductivity impairment mechanism. The crushed proppant grains could generate fines that can migrate and trap in the pore throats. It has been shown that a 5% fine generation by crushing could reduce the fracture conductivity by as much as 60% [90]. The natural sands typically generate more fines under stress, while ceramic proppant are more resistant to crush. The higher sphericity of ceramic proppants also reduces the chance of fines entrapment in the pack, resulting in less reduction in conductivity [91]. The natural sand grains typically show noticeable crushing in closure stresses larger than 35 MPa (5,000 psi). Ceramic grains are usually tougher than the natural sands. As mentioned before, the range of stresses expected in typical deposits of Alberta oil sands depths is in the range of 5-10 MPa which is far below the typical crushing stress of proppants. However, higher operating temperature could possibly reduce the crushing stress of proppants, again with less effect on ceramics. Overall, the proppant crushing mechanism is not expected to be a major concern in this project. Proppant embedment is another mechanism for conductivity loss that is more important for our purpose. After the hydraulic fracture is successfully placed, the normal stress acting on the fracture walls would push against the proppant grains. This will 120

138 initially pack the proppants tighter and result in conductivity loss as shown in Figure 5.2. In addition to the packing, depending on the hardness of the rock material, the fracture walls may deform around the proppant grains and reduce the propped width. The proppant modulus is usually greater than that of rock material. Hence, the proppant embedment should occur before potential proppant grain crushing. Figure 5.3 illustrates the proppant embedment mechanism. Figure 5.3 Proppant embedment at elevated closure stress (not to scale) The effect of proppant embedment on conductivity loss is more pronounced for soft rock material. Lacy et al. [92] performed a series of embedment experiments on soft rock samples from North Sea, south Texas, New Mexico and Gulf of Mexico. They concluded that proppant embedment can be more significant for formations with Brinnel Hardness (BH) of less than 20 kg/mm 2 or Young s modulus less than 13 GPa. In particular, for some samples with static Young s modulus of less than 2 GPa, the proppant embedment reduced the fracture width by as much as 60% (experiments were run for proppant concentration of 10 kg/m 2 ). A more detailed evaluation of their experimental results indicated that such high embedment level was only observed at high closure stress levels (5,000-9,000 psi), and for the closure stress below 2,000 psi, the fracture width reduction was minimal. Alramahi and Sundberg [93] also studied the fracture conductivity loss by proppant embedment in shale gas formations. They found that the embedment level was greater for the shale formation 121

139 with more clay content. They too observed more embedment for softer shale material (E < 2 GPa). In the lower range of closure stress in this application, even for soft formation rock, one could argue that for a multilayer proppant pack the embedment should mostly be limited to half a proppant monolayer per fracture wall as depicted in Figure 5.3. For a five layer proppant pack, for example, the fracture would only lose 20% of the width. For poorly propped parts of fracture, the loss of conductivity would naturally be greater. Further evaluation should be done by running creep tests for shale and saturated oil sand material against proppant packs at elevated temperatures. 5.5 Conductivity loss from the fracturing fluid Carrying large proppants at high concentration requires a high viscosity fracturing fluid that creates enough drag force to keep the proppant suspended. Producing a high viscosity fluid is achievable with blending high concentration of polymer and crosslink additives to the base fluid which is usually water (creating crosslinked gels). Using high viscosity fluids would also help in obtaining increased fracture width. In bituminous sands, the absolute permeability is very high and where mobile water is present, a very high leak-off rate could be anticipated. Consequently, it is desired to use a fracturing fluid that can produce an effective layer of filter cake on the fracture walls to control the leak off. The inherent effect of using high viscosity crosslinked gels as fracturing fluid is permeability damage in the proppant pack and the fracture walls. Fortunately, the thermal 122

140 degradation points of most of these additives (gel break temperatures) are in the range of o C [94, 95]. Since the fractures are meant to undergo steam injection, it is expected that the gel will eventually break and the lost conductivity to be eventually restored. The lower initial permeability due to gel retention in the fractures is actually desirable. During the SAGD start-up phase where steam is circulated in the wells, a low volume of steam leak-off into the formation is preferable. Low initial steam leak-off volume will help the uniform warm up of the entire horizontal section, which eventually improves the conformance of the well pair. In contrast, the high steam leak-off volume to the fractures from the beginning could result in an uneven development of steam chamber in the sand immediately surrounding the well pair. On the other hand, a possible negative effect of the low initial fracture conductivity may be a potential delay in establishing the SAGD process in the upper reservoir partitions. Under this alternative, if the preference is to have higher fracture conductivity early on, the design of the fracturing fluid break-up should be based on time-dependent chemistry as opposed to high temperature degradation (since the oil sands reservoir temperature is usually low). 5.6 Thermo-elasticity effect on propped fractures After the operation starts and the steam injection commences, the wellbore and the fractures will slowly begin to warm up. The normal stress acting on the fracture plane will start to change accordingly. The operating pressure in SAGD is typically only slightly above the original reservoir pressure and hence, poroelasticity impact is expected to be 123

141 minimal. On the other hand, with temperature rising toward the steam saturation temperature, the stress level will start to increase as illustrated in Figure 5.4. Hydraulic Fracture Sand Shale Figure 5.4 Schematic temperature profile around a propped fracture under steam injection The stress increase caused by the change in temperature can be calculated from Equation ( 5.2): σ h = a L E 1 v T ( 5.2) where a L is the volumetric thermal expansion coefficient of the formation rock, E is the Young s modulus, v is the Poisson s ratio and T is the temperature difference between the steam temperature and original reservoir temperature. As an example, for the conceptual case presented in Section 4.8 (with mechanical properties from Table 4.5) and a typical 124

142 thermal expansion coefficient for porous rock [96](P. 69) the thermal stress increase can be calculated as: σ h ( hal ) = = 3,143 kpa The original effective stress for this example was 3,150 kpa (from Table 4.5) for the formation depth of 350 m. The addition of thermal stress would double that. While this is a significant increase, the total effective stress would still be fairly low concerning the stress effect of proppant conductivity and embedment. If the fractures are un-propped or do not contain enough proppants, the adverse effect of increased thermal stress could be serious. Fina Oil Company tried three inverted 5-spot steam flooding patterns in from a shallow (63 m overburden) Athabasca deposit in Steepbank [36, 39]. They had envisioned establishing communication between wells by hydraulic fracturing to develop a hot horizontal plane of communication such that they can initiate a steam drive recovery process. The first five-spot pattern had the entire pay open to flow. After a hydraulic fracturing job without proppant, some level of communication between the wells was achieved. However, the communication was later lost as the injection and production wells became heated. In the follow up patterns, propped hydraulic fractures were pumped on the bottom of the wells which turned out to maintain the communication between the wells. The pilot was terminated later on due to steam breakthrough to the surface. The loss of hydraulic communication in the first pattern can be attributed in part to the increase of the thermal stress on un-propped fractures after steaming. The same problem has been reported in steam flooding operations in carbonate 125

143 formations in Oman in which natural fractures tend to lose their permeability with steam injection [97]. The mechanical behavior of the oil sands and shales at elevated temperatures could also affect the long term conductivity of the propped fractures. Long term compaction creep tests on Venezuelan oil sand and shale samples by Blair et al. [98] showed that the compaction was larger at elevated temperatures for both oil sand and shale samples. This is particularly important in the evaluation of proppant embedment effect after the start of steam injection. Consequently, site specific compaction tests at steam temperature would be critical. 5.7 Proppant transport in fracturing jobs To achieve the desired long term conductivity, the proppants should be transported deep into the fracture. In Chapter 3, it was shown that a desired fracture half lengths is about 15m with extension to the top of formation. We discussed earlier that the proppants that can produce high range of fracture conductivity have large grain sizes and typically have higher densities than sands (in the case of ceramic proppants). This could make the proppant transport in the fracture more challenging compared to a slick water fracture job in shale gas with finer sand proppants. For this, a high viscosity fracturing fluid is needed to carry the proppants inside the fracture. On the other hand, high viscosity of the fluid results in higher than normal frictional losses inside the pipes and fractures. To see if proper fracture placement with coarse ceramic proppants and high viscosity fluid is possible, a commercial fracture simulator, GOHFER, was used to create 126

144 a fracturing job model. GOHFER is a planar 3-D geometry fracture simulator [99] with a fully coupled fluid/solid transport simulator, but like many other commercial fracture modelling simulators its capability to model fracturing through rock interfaces is limited. Hence, the model results were only used to draw conclusion on proppant transport and estimated concentration rather than on fracture shape and size. To set up the model, the reservoir and mechanical properties were taken from the Geosim model in Section 4.9 with the exception of Young s modulus which assumed to be 1 MPa for both sand and shale layers. For proppant, the CarboProp 6/10, which is a coarse medium density proppant (with specific gravity of 3.25), was selected. The permeabilitystress of this proppant can be found in Figure 5.2. For fracturing fluid, Trican s QLT BRK with apparent viscosity of 1,360 cp was selected. The fracture was a single stage at the true vertical depth of 355 m and measured depth of 700 m with a 20 m perforated interval. The pad volume and proppant volumes were optimized with a few modelling iterations to achieve the desired proppant concentration and placement in the fracture. The final case is presented in Figure 5.5. The pumping starts with an 18 m 3 of pad volume at the rate of 3 m 3 /min. The proppant is then added at the concentration of 180 kg/m 3 and is increased to 260 kg/m 3 stepwise. The final stage is pumped without proppant with the rate of 7 m 3 /min to clean up the injection string and better distribute the proppant in the fracture. The total clean volume injection is 57 m 3 as well as 5,850 kg of proppant. This exercise shows that the desired fractures are rather small (about 6 tonnes of proppant in this case). 127

145 Figure 5.5 Fracture job modelling standard plot Figure 5.6 shows the proppant concentration in the fracture after closure as predicted by the fracture model which is in the desired range of 15 kg/m 2 for the bulk of the fracture volume. It is important to highlight that the proppant was transported fairly well within the fracture (upward from the injection point which is marked by the cross on Figure 5.6) and is placed in the shale intervals at acceptable concentrations. As mentioned previously, the purpose of presenting the above fracturing job model result was to show the possibility of successful transport of coarse and heavy proppants into the fracture while managing the leak off and frictional losses. To better model a fracturing process in the oil sands and shale, one can use a more comprehensive approach that is includes additional physical phenomena of the process [100]. 128

146 Figure 5.6 Proppant concentration in the fracture as predicted by fracture model 5.8 Design considerations in wellbore completion The fracturing jobs in shale gas or tight oil formations are implemented by using both openhole and cemented liner systems. The choice of liner system is usually made based on the reservoir and rock properties as well as cost considerations. In the context of SAGD recovery in fractured horizontal wells, both systems can be used with specific considerations. 129

147 In a cemented liner system, the only pathway for steam injection would be the perforations that connect the wellbore to the fractures. Considering the process schematics in Figure 2.3 and Figure 2.4 and the analytical formulations, it can be argued that the cemented liner system hinders the steam conformance along the length of the injection well. This could result in lower and/or slower oil recovery from the sand package that is immediately surrounding the SAGD well pair. Furthermore, the perforations in the liner cannot hold back the proppants and reservoir sand from getting into the liner, especially during the circulation phase when the net pressure is small. To avoid sand production, a secondary slotted liner should be run in the cemented liner which in turn adds cost and more complexity to the operations. On the other hand, the fracturing process can be targeted more specifically to the perforated intervals with less risk of fracture communication. Furthermore, as discussed in Section 4.10, in a perforated well, the likelihood of developing transverse fractures is higher if the horizontal well is perpendicular to the maximum horizontal stress. In an openhole liner system, the reservoir to wellbore communication is not restricted just to the fractures, and a rather classic steam chamber could be developed in the sand package immediately surrounding the well pairs. Figure 5.7 illustrates a suggested design for an openhole liner system suitable for SAGD application. For openhole systems and as per Table 4.6, the breakdown pressure estimates suggest that the fractures are likely to initiate with a complex geometry if the well is drilled perpendicular to maximum horizontal stress. Also, the point of fracture initiation in this case would be unpredictable, i.e. it will be somewhere between the packers. As a result, the fracture placement would be 130

148 approximate and the chance of fracture communication with adjacent fracture stages would be higher. The sand control in this suggested design can be done by adding annular inflow sand screens to the liner string so that after fracturing, a mechanical sleeve could be shifted to close the fracturing ports and open the screens as illustrated in Figure 5.7. Figure 5.7 An openhole high temperature stimulation system with sand control (courtesy of Packers Plus Energy Services Inc.) An alternate openhole design would be to use a retractable fracturing string with packers placed immediately on both sides of each fracture port. This way, there will be more control on points of fracture initiation. After the fracturing job, the fracturing string can be pulled out and a regular slotted liner can be run back into the well. Overall, the openhole system seems to be more reliable in terms of operations and oil recovery considerations. 131