Evaluation of Strategies for Enhancing Production of Low-Viscosity Liquids From Tight/Shale Reservoirs

Transcription

1 SPE MS Evaluation of Strategies for Enhancing Production of Low-Viscosity Liquids From Tight/Shale Reservoirs George J. Moridis, Lawrence Berkeley National Laboratory; Thomas A. Blasingame, Texas A&M University Copyright 2014, Society of Petroleum Engineers This paper was prepared for presentation at the SPE Latin American and Caribbean Petroleum Engineering Conference held in Maracaibo, Venezuela, May This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright. Abstract Production of low-viscosity liquids (including condensates) from tight reservoirs (such as shales) is severely restricted by the ultra low-permeability of such formations, limiting production to a very small fraction (usually less than 5 percent) of the liquids-in-place. In this study, which is part of a wider investigation, we evaluate by means of numerical simulation several possible strategies to enhance low-viscosity liquids production from such reservoirs. These strategies include (a) physical displacement processes, (b) non-thermal processes to reduce the viscosity and the critical saturation of the liquids, (c) thermal processes, (d) enhanced reservoir stimulation, (e) novel well configurations and (f) combinations thereof. The objectives of this effort are to (1) to remove from further consideration potential production strategies that hold limited (if any) promise, and (2) to identify production strategies that appear to have potential for further study and development. We first determine the baseline production performance of such reservoirs corresponding to several reference production regimes that involve minimal or no reservoir stimulation, standard displacement fluids (H2O or CH4), standard well configurations and no thermal treatment. We then evaluate the efficiency of several production strategies: (a) traditional continuous gas flooding using parallel horizontal wells and using the currently abundant shale gas, (b) water-alternating-gas (WAG) flooding, (c) huffand-puff injection/production strategies using lean gas/rich gas in a traditional (single) horizontal well with multiple fractures, (d) flooding using appropriate gases (e.g., CO2, N2, CH4) using appropriate well configurations (mainly horizontal), with the viscosity reduction resulting from the gas dissolution into the liquids, and (e) thermal processes, in which the viscosity reduction is achieved by heating, possibly to the point of liquid vaporization and transport through the matrix to the production wells as a gas. Our study includes an analysis of the sensitivity of the liquids production to the main parameters defining each of the strategies listed above in an effort to determine the critically important parameters and factors that control the production performance and efficiency. Introduction Background Gas production from tight gas/shale gas reservoirs over the last decade has met with spectacular success with the advent of advanced reservoir stimulation techniques (mainly hydraulic fracturing), to the extent

2 2 SPE MS that shale gas is now among the main contributors to US hydrocarbon production. This remarkable success has not been matched by similar progress in the production of (relatively) low-viscosity liquid hydrocarbons (including condensates) because of the significant challenges to liquid flow posed by the ultra-low permeability (and the correspondingly high capillary pressures and irreducible liquids saturations) of such reservoirs. These difficulties have limited liquids production to a very low fraction (usually 5%) of the resources-in-place. Increasing the recovery of liquids from these ultra-low permeability systems even by percent over its current very low levels (to a level that is still low in absolute terms, but very significant in relative (hence, economic) terms) not only will increase production and earnings, but will have considerable wider economic implications, as the enhanced recovery will affect reserves and the valuation of companies. Our effort aims to address this issue by using numerical simulation to investigate a wide range of possible strategies for improved liquids production from tight/shale reservoirs. In essence, this is an attempt to provide a baseline mechanistic study that deploys some state-of-the-art tools in reservoir modeling. If successful in identifying interactions, processes and methods that can increase production by as little as percent over the current low recovery rates (usually 5% or less), the impact in the industry can be significant and potentially dramatic (despite a recovery that may remain low in absolute terms) not only because of an increase in hydrocarbon recovery, but also because this can allow an increase in reserve estimation and company valuation, with considerable economic consequences. In short, the extremely low oil recovery of these systems warrants the identification of key processes and interactions, and the development of strategies to achieve maximum recovery. The current effort aims to address these issues. Objectives The main problem that this study addresses is the extremely low liquid hydrocarbon production from light/tight oil systems. The overall objective of the effort (of which the study reported in this part is a part) is to to identify, analyze and quantify by means of numerical simulation the dominant interactions (fluid-fluid, fluid-media, fluid-fractures, etc.) that underlie and control the processes involved in the production (by various methods) of low-viscosity hydrocarbons from tight (shale) media, and to use the gained insights to increase/optimize production. The specific objectives are: To identify and quantitatively describe mechanisms that control fluid flow and the various system interactions in tight media, and to quantitatively describe the behavior of the several fluids involved in the production process through the extremely small pore space of tight media, leading to promising strategies for enhanced recovery. To identify production strategies that hold limited (if any) promise, and remove them from further consideration potential. To identify and focus on strategies that appear to have potential for significant enhancement of hydrocarbon production (in terms of maximization of both production and recovery), to numerically evaluate their large-scale and long-term performance, to identify the dominant mechanism and processes, and evaluate the relative importance of the properties, conditions and parameters that control them. Based on insights provided by the various analyses, to propose (if possible) novel approaches as new methods for enhanced production of low-viscosity fluids from tight/shale oil reservoirs. The studies in this paper focus on scoping calculations that aim to determine the baseline performance of various key production methods that are considered standard in the attempt to produce low-viscosity liquids from tight- and shale-oil reservoirs. These include recovery using displacement methods, viscosity reduction methods (thermal processes and those caused by the dissolution of appropriate substances, such as CO 2 ), enhanced reservoir stimulation, and combinations thereof.

3 SPE MS 3 The Simulation Code and the Underlying Fundamental Equations of Flow and Transport The numerical simulation code We used the TOUGH MultiCompPhase code - i.e., the TOUGH core simulator (Moridis et al., 2008; Zhang et al., 2008) with the MultiCompPhase equation of state - to conduct the numerical studies in this paper. This code (hereafter referred to as T MCP) can model all the known processes involved in the non-isothermal, multi-component, two-phase flow of fluids and heat through porous and/or fractured geologic media. By appropriate domain discretization and property assignment, the T MCP code can describe flow through both conventional and unconventional (fractured tight- and shale-gas and -oil) reservoirs. The simulator can be used isothermally if scoping calculations indicate show conclusively limited thermal response to gas production. T MCP is a compositional simulator, and its formulation accounts for heat and up to 13 mass components (H 2 O, up to 10 gases, and up to two oil components) that are partitioned between four possible phases: gas, aqueous, organic and liquid CO2. The T MCP can describe all the possible thermodynamic states (phase combinations) of the system. Because of code is its fully implicit formulation that is based on the Jacobian, it can handle all possible phase changes, state transitions, strong nonlinearities and steep solution surfaces that may arise. The mass components A non-isothermal fractured tight-gas or shale-gas system can be fully described by the appropriate mass balance equations and an energy balance equation. The following components (and the corresponding indicators used in the subsequent equations), corresponding to the number of equations, are considered: g i the various gaseous components (compounds) i constituting the natural gas (i 1,...,N G ) O j the oil components, i.e., a light-oil component (j 1) and a heavier-oil component (j 2) w water heat (treated as a pseudo-component) The mass and energy balance equations Following Pruess et al. (1999), mass and heat balance considerations in every subdomain (gridblock) into which the simulation domain is been subdivided by the integral finite difference method dictates that (1) where: V, V n volume, volume of subdomain n [m 3 ]; M mass accumulation term of component [kg m 3 ]; A, n surface area, surface area of subdomain n [m 2 ]; F flow vector of component [kg m 2 s 1 ]; n inward unit normal vector; q source/sink term of component [kg m 3 s 1 ]; t time [s]. Mass accumulation terms Assuming a maximum of four-phase ( A: aqueous, G: gas, O: organic and C: liquid CO 2 ) flow conditions, the mass accumulation terms M for the mass components in equation (1) are given by (2) where porosity [dimensionless]; density of phase [kg m 3 ];

4 4 SPE MS S saturation of phase [dimensionless]; mass fraction of component w, g i, o j in phase [kg/kg] R rock density [kg m 3 ]; mass of sorbed gaseous component g i per unit mass of rock [kg/kg] 1 for shales and 0 for non-sorbing tight-gas system (usually devoid of substantial organic carbon) The first term in equation (2) describes fluid mass stored in the pores, and the second the mass of Y g i S gaseous components sorbed onto the organic carbon (mainly kerogen) content of the matrix of the porous medium. Although gas desorption from kerogenic media has been studied extensively in coalbed methane reservoirs, and several analytic/semi-analytic models have been developed for such reservoirs (Clarkson and Bustin, 1999), the sorptive properties of shale are not necessarily analogous to coal (Schettler and Parmely, 1991). The most commonly used empirical model describing sorption onto organic carbon in shales is analogous to that used in coalbed methane and follows the Langmuir isotherm that, for a single-component gas, is described by (3) The m L term equation (3) describes the total mass storage of component g i at infinite pressure (kg of gas/kg of matrix material), p L is the pressure at which half of this mass is stored (Pa), and k L is a kinetic constant of the Langmuir sorption (1/s). In most studies applications, an instantaneous equilibrium is assumed to exist between the sorbed and the free gas, i.e., there is no transient lag between pressure changes and the corresponding sorption/desorption responses and the equilibrium model of Langmuir sorption is assumed to be valid (Figure 6). Although this appears to be a good approximation in shales (Gao et al., 1994) because of the very low permeability of the matrix (onto which the various gas components are sorbed), the subject has not been fully investigated. For multi-component gas, equation (3) becomes (4) where B gi is the Langmuir constant of component g i in 1/Pa (Pan et al., 2008), and y gi is the dimensionless mole fraction of the gas component. Heat accumulation terms The heat accumulation term includes contributions from the rock matrix and all the phases, and is given by the equation (5) where: C R heat capacity of the dry rock [J kg 1 K 1 ]; U specific internal energy of phase [J kg 1 ]; u gi specific internal energy of sorbed component g i [J kg 1 ]; T temperature [K]

5 SPE MS 5 T 0 a reference temperature [K] The specific internal energy of the gaseous phase is a very strong function of composition, is related to the specific enthalpy of the gas phase H G, and is given by (6) where is the specific internal energy of component in the gaseous phase, and U G, dep is the specific internal energy departure of the gas mixture [J kg 1 ]. The internal energy of the aqueous phase accounts for the effects of gas and oil solution, and is estimated from (7) where and are the specific internal energies of H 2 O, of the gas component and of the oil component, respectively, at the conditions prevailing in the aqueous phase, and and are the specific internal energies of dissolution of the various gas and oil components in water. The internal energies of the organic phase and of the liquid-co 2 phases are determined in an entirely analogous manner. Fluid flux terms The mass fluxes of water and gaseous components include contributions from the aqueous and gaseous phases, i.e., (8) The mass flux of component in phase incorporates advection, dispersion and diffusion contributions, and is given by (9) where and b is the Klinkenberg [1941] b-factor accounting for gas slippage effects [Pa], the term is the diffusive mass flux of component in the phase [kg m 2 s 1 ], the term is the dispersive mass flux of component in the phase [kg m 2 s 1 ], and v is described by Darcy s law as (10) k rock intrinsic permeability [m 2 ]; k r relative permeability of phase [dimensionless]; viscosity of phase [Pa s]; P pressure of phase [Pa]; g gravitational acceleration vector [m s 2 ]. The aqueous pressure P A ( A) is given by (11) where P G is the gas pressure [Pa], P cgw is the gas-water capillary pressure [Pa], and are the partial pressures [Pa] of the water vapor, of the oil components, and of the natural gas components in the gas phase, respectively. Similarly, the pressures of the other phases can be determined in a manner similar to that in equation (11) by using the gas pressure and the various capillary pressures between the

6 6 SPE MS gas and the phase in question. The gas solubility in the aqueous phase is related to law, through Henry s (12) where H g i H g i (T) [Pa] is a temperature-dependent factor akin to Henry s constant. Note that it is possible to determine the from the equality of fugacities in the aqueous and the gas phase. Although this approach provides a more accurate solution, the difference provides a small increase in accuracy that cannot justify the significantly larger execution time. The dissolution of the oil components in water, and of water in the oil components is described by the parametric relationships discussed by Falta et al. (1995) and found in Poling et al. (2001). The solubility of gases in oil components of the organic phase, as well as all thermophysical properties of the oil components and of the organic phase above, at and below the bubble point, are described using the relationships recommended by McCain et al. (2011). The solubility of gases in the liquid CO 2 -phase and the corresponding thermophysical properties are computed using simple parametric equations, but there is very little information on the subject. The term is the diffusive mass flux of component in the phase [kg m 2 s 1 ] that is described by (13) where is the multicomponent molecular diffusion coefficient of component in the gas phase in the absence of a porous medium [m 2 s 1 ], and is the tortuosity of phase [dimensionless] computed from the Millington and Quirk [1961] model. The diffusive mass fluxes of the water vapor, the oil vapors, and the gases in the gas phase are related through the relationship of Bird et al. [1960] (14) which ensures that the total diffusive mass flux of the gas phase is zero with respect to the mass average velocity when summed over the components. Then the total gas phase mass flux is the product of the velocity and density of the gas phase. The term is the dispersive mass flux [kg m 2 s 1 ] of component in phase ( A,O,C) that is described by (15a) where is the dispersion tensor of component in phase, (15b) and are the transverse and longitudinal dispersivities of component in phase [m], is the length of the velocity vector, and all other terms are as previously defined. In media with nano-scale pores (such as shales), the Klinkenberg b-factor is computed using the method of Florence et al. (2007) and Freeman et al. (2009) from the equation (16) where K n is the Knudsen diffusion number (dimensionless), which accounts for the effects of the mean free path of gas molecules being on the same order as the pore dimensions of the porous media and characterizes the deviation from continuum flow. Knudsen diffusion can be very important in porous

7 SPE MS 7 media with very small pores (on the order of a few micrometers or smaller) and at low pressures. It is computed from (Freeman et al., 2009b) as (17) with M being the molecular weight and T the temperature ( K). The term is determined from Karniadakis and Beskok (2001) as (18) from Karniadakis and Beskok (2001). Non-Darcy flow If the flow is non-darcian (e.g., in the case of fast gas flow in fractures), then equation (9) still applies, but v is now computed from the solution of the quadratic equation (19) in which T is the turbulence correction factor (Katz et al., 1959). The quadratic equation in (19) is the general momentum-balance Forchheimer equation (Forchheimer, 1901; Wattenbarger and Ramey, 1968), and incorporates laminar, inertial and turbulent effects. Its solution is (20) and v from equation (20) is then used in the equations (9) of flow. Heat flux The heat flux accounts for conduction, advection and radiative heat transfer, and is given by Where a representative thermal conductivity of the fluid-impregnated rock [W m 1 K 1 ]; h specific enthalpy of phase [J kg 1 ]; f radiance emittance factor [dimensionless]; 0 Stefan-Boltzmann constant [ Jm 2 K 4 ]. The specific enthalpy of the gas phase is computed as (21) (22) where is the specific enthalpy of component in the gaseous phase, and H dep is the specific enthalpy departure of the gas mixture [J kg 1 ]. The specific enthalpy of the aqueous phase is estimated from (23) where and are the specific enthalpies [J/kg] of H 2 O, of the gas component and of the oil component, respectively, at the conditions prevailing in the aqueous phase, and and are the specific enthalpies of dissolution of the various gas and oil components in water. The enhalpies of the organic and of the liquid-co2 phases are computed in a manner that entirely analogous to that of equation (23).

8 8 SPE MS Source and sink terms is described by In sinks with specified mass production rate, withdrawal of the mass component (23) where q is the production rate of the phase [kg m 3 ]. For a prescribed production rate, the phase flow rate q is determined from the phase mobility at the location of the sink. For source terms (well injection), the addition of a mass component occurs at desired rates qˆ. The contribution of the injected or produced fluids to the heat balance of the system are described by (24) where q is the rate of heat addition or withdrawal in the course of injection or production, respectively (W/kg). P- and T-dependence of and k The effect of pressure change on the porosity of the matrix is described by two options. The first involves the standard exponential equation (25) where T is the thermal expansivity of the porous medium (1/K) and p is the pore compressibility (1/Pa), which can be either a fixed number or a function of pressure (Moridis et al., 2008). A second option describes the p-dependence of as a polynomial function of p. The - k relationship in the matrixis described by the general expression of Rutqvist and Tsang (2003) as: (26) where is an empirical permeability reduction factor that ranges between 5 (for soft unconsolidated media) and 29 (for lithified, highly consolidated media). Note that the equations described here are rather simple and apply to matrix and k changes when the changes in p and T are relatively small. These equations are not applicable when large pressure and temperature changes occur in the matrix, cannot describe the creation of new (secondary) fractures and cannot describe the evolution of the characteristics of primary and secondary fractures (e.g., aperture, permeability, extent, surface area) over time as the fluid pressures, the temperatures, the fluid saturations and the stresses change. For such problems, it is necessary to use the T M model (Kim and Moridis, 2013) that couples the flow-and-thermal-process T RW simulator discussed here with the ROCMECH geomechanical code. This coupled model accounts for the effect of changing fluid pressures, saturations, stresses, and temperatures on the geomechanical regime and provides an accurate description of the evolution of and k over the entire spectrum of p and T covered during the simulation. Fractured System Description and Subdomain Representation Subdomains and fractures The fractured system in producing tight- and shale-gas reservoirs can be described as a set of interacting subdomains (Moridis et al. 2010). These include the following: Subdomain 1 (S1): The original (i.e., in its undisturbed state prior to the initiation of production operations) reservoir rock, which may be naturally fractured and may be characterized by distinct sets of fractures, each one with its own properties (aperture, length, orientation, density, etc.). The original fractures in S1 are hereafter referred to as native fractures (NF).

9 SPE MS 9 Subdomain 2 (S2): The fractures or fractured network created during the reservoir stimulation (e.g., by hydro-fracturing the reservoir rock). These artificial fractures penetrate Subdomain 1, increase the surface area from which can be produced, and may intercept the natural fractures of Subdomain 1, thus providing access of gas in these fractures to the well. The fractures of Subdomain 1 are expected to be the dominant pathways of flow to the well and are referred to as primary fractures (PF). Subdomain 3 (S3): This is the subdomain defined by the stress-release fractures that are induced by changes in the geomechanical status of the rock in the vicinity of the PF following the reservoir stimulation process. Their orientation is a function of the stress distribution and geomechanical propeties of the rock, but tend to occur on planes that are roughly perpendicular (ar at an angle) to the PF. Such fractures are referred to as secondary fractures (SF): they penetrate Subdomain 1 mainly adjacent PFs (i.e., they occur within the fracture spacing, and, in their upper limit, they can cover it entirely), and intercepts native and primary fractures, thus increasing the flow area and, consequently, production. Subdomain 4 (S4): This is the subdomain defined by the stress-release fractures that are induced by changes in the geomechanical status of the rock in the immediate vicinity of the wellbore because of drilling. S4 is expected to have a roughly cylindrical shape centered around the wellbore axis, and to be characterized by a limited radius (i.e., short fracture length), high fracture density and small aperture. Thus, S4 is expected to represent a small fraction of the overall system volume, but may be important to flow as they can increase significantly the flow area, in addition to being directly connected to S1 and S2, and possibly intersecting fractures in S3. The fractures in S4 are hereafter referred to as radial fractures (RF). Thus, S1 is that natural (at discovery) state of the system. Drilling and well installations may inevitably cause S4 to form. S2 is the result of reservoir stimulation activities (and the only one over which the operator can exert control), while S3 is a direct byproduct of them. S1 cannot provide sufficiently high production rates without stimulation in tight- and shale-gas reservoirs because of very low permeability. For the same fracture characteristics (except fracture length), production is maximized when the cumulative size of S2, S3 and S4 is maximized. Types of fractured systems Based on the properties and characteristics of S1 and the occurrence (or absence) of S3, there are four possible types of producing tight- or shale-gas reservoir systems. These are depicted in Figures 1 to 4 (all of which involve a horizontal well configuration), and are listed below in order of increasing complexity (in terms of description, simulation and analysis): 1. Type I (Figure 1): This is characterized by (a) the absence of NF fractures in S1, which now involves the unfractured matrix, and (b) the absence of the S3 subdomain. This is the simplest and least productive system, as it is characterized by the minimum surface area and flow pathways for path production. It is possible to further simplify it by assuming absence of the S4 subdomain and of the RF. 2. Type II (Figure 2): Unlike Type I, Type II systems feature the S3 subdomain and SF. This is expected to yield higher gas production because of increased surface area and more pathways to flow. SF can extend along the entire length of the fracture spacing (i.e., d sf d f, in which case production is expected to be maximized). 3. Type III (Figure 3): The difference between Type I and Type III systems is the occurrence of the native fractures (NF) in S1. Such a system is expected to have higher gas production than Type I systems. Comparison of its productivity to Type II is not straightforward, because production is

10 10 SPE MS Figure 1 Stencil of a Type I system involving a horizontal well in a tight- or shale-gas reservoir (Moridis et al., 2010). Figure 2 Stencil of a Type II system involving a horizontal well in a tight- or shale-gas reservoir (Moridis et al., 2010).

11 SPE MS Figure 3 Stencil of a Type III system involving a horizontal well in a tight- or shale-gas reservoir (Moridis et al., 2010). Figure 4 Stencil of a Type IV system involving a horizontal well in a tight- or shale-gas reservoir (Moridis et al., 2010). 11

12 12 SPE MS controlled by the characteristics of the fractures, and the relative contributions of NF and SF cannot be determined a priori. 4. Type IV (Figure 4): This system involves all 4 subdomains (S1 to S4) and all fracture types (NF, PF, SF and RF). This is the most complex system to describe, simulate and analyze. All other features and characteristics being equal, this is clearly the system with the highest production potential because of its maximum (compared to the other types) surface area and the largest number of flow pathways to the production well. Note that Figures 1 to 4 describe the basic stencils of the four types, i.e., the smallest possible repeatable or repetitive element (fraction of the domain) that is necessary and sufficient to fully characterize the system and describe its behavior during production. A schematic describing the full system and the extraction of the stencils appears in Figure 5 (Freeman et al., 2009). Implicit in the selection of these stencils are the assumptions of (a) symmetry about the vertical plane that passes by the horizontal well cenetrline (indicating an assumption of homogeneous property distributions about this plane of symmetry), (b) symmetry about the horizontal plane that passes by the centerline of the well, indicating an assumption of minimal gravitational effects in the overall system performance (a reasonable approximation in deep, thin reservoirs), and (b) negligible tow-to-heel pressure differences during production (a valid approximation in relatively short horizontal wells). Thick reservoirs may necessitate accounting for gravitational effects (especially at higher pressures), in which case the symmetric lower half of the stencil/domain (i.e., that below the horizontal plane at z 0) will need to be included in the analysis. Significant differences between tow and heel pressures, and/or significant heterogeneity in a substantial part of the reservoir may require consideration of the entire horizontal well system. Application of these observations to vertical wells is entirely analogous. The Simulation Approach and Specifics Geometry and discretization of the simulated system The simulated systems are the Cartesian 3D ones described by the general horizontal well systems in the stencils of Figures 1 to 4. A more specific representation is provided in Figures 5 and 6, which also indicate the relative position of a second horizontal well (not always used) for injection of fluids and/or heat in displacement and thermal processes. Additionally, Figures 5 and 6 include non-flowing parts of the domain (in the overburden and underburden) that need to be considered in non-isothermal processes. The main dimensions of the simulated system are shown in Table 1. We employed very high definition grids with mm-size resolution in the vicinity of the PF in S2, and in the vicinity of the confluence of the matrix in S1 with the PF and SF in subdomains S2 and S3. The surface area of the Cartesian system at the well was corrected to reflect its cylindrical geometry. The discretization resulted in a large number of elements (gridblocks) that varied between a minimum of about 800,000 for a simple Type I reservoir with no RF, to a maximum of about 3,000,000 for the most complex problem of Type IV. The number of equations we solved varied according to the number of components used in a particular study. All studies were conducted non-isothermally. Because even the smallest grids and the fewest equations still resulted in enormous solution matrices (Jacobian) with very large computational requirements, we used the distributed-memory, parallel version of the code (Zhang et al., 2008) in all the simulations discussed in this paper. System properties The hydraulic and thermal properties of the various subdomains (S1 to S4), as well as the initial conditions and the system fracture properties and characteristics are listed in Table 1. Note that the flow properties and consitions used in the simulations were typical of Eagle Ford shale oil reservoirs. We assumed that the initial aqueous, gas and organic saturations (S A, S G and S O, respectively) were distributed uniformly in the matrix. The initial S A was kept at a level slightly below the very high irreducible water saturation S ira that is normally expected in such tight (low-permeability) formations

13 SPE MS 13 Figure 5 Detailed stencil of the tight/shel reservoir investigated in this study View A. Figure 6 Detailed stencil of the tight/shel reservoir investigated in this study View B.

14 14 SPE MS Table 1 Properties and conditions of the reference case (Type I) Parameter Value Initial pressure P Pa (2900 psi) Initial temperature T 60 C Bottomhole pressure P w Pa (1450 psia) Oil composition 100% n-octane Initial saturations in the domain S O 0.7, S A 0.3 Intrinsic matrix permeability k x k y k z m 2 ( 10 5 md) Matrix porosity 0.05 Fracture spacing X f 30 m Fracture aperture W f m Fracture porosity f 0.60 Formation height 10 m Well elevation above reservoir base 1 m Well length 1800 m (5900 ft) Heating well temperature T w 95 C Grain density R 2600 kg/m 3 Dry thermal conductivity k RD 0.5 W/m/K Wet thermal conductivity k RW 3.1 W/m/K Composite thermal conductivity model 16 k C k RD (S 1/2 A S 1/2 H )(k RW k RD ) Capillary pressure model 14,23 S ira P Pa Relative permeability k ro (S O *) n Model 17 k rg (S G *) n S O * (S O S iro )/(1 S ira ) S G * (S G S irg )/(1 S ira ) EPM model n 4 S iro 0.20 S ira 0.60 because phase interference and the effect on flow and gas production were not a primary focus of this study. Thus, the reasonable relative permeability and capillary pressure relationships and parameters did not have a significant (if any) impact on the results because they remain practically constant during the simulation. Initial and boundary conditions, and well description The no-flow conditions (of fluids and heat) that were applied at the stencil outer boundaries (at at x d f /2, z 0, see Figures 1 to 5) implied symmetry at these locations (see earlier discussion). No-flow conditions were also applied at the upper boundary of the reservoir at z h/2 (indicating the presence of an impermeable bounding formation). The initialization process included a gravity and temperature equilibration, thus accounting for gravitational and geothermal gradient effects. Reservoir fluids were produced by imposing a constant bottomhole pressure P w regime at the gridblocks corresponding to the radius of the lower horizontal well (the lower one in Figures 5 and 6). Thus, the well is treated as an internal boundary. The total production rate Q was determined by summing the flows between the boundary gridblocks that represented the well and the corresponding adjacent gridblocks in the formation that were in contact with the well along its entire length in the stencil. No gridblocks internal to the well were considered, and no lateral flows within the well. By imposing a constant P w Pa ( 2903 psia), a thermal conductivity k 0 W/m/K, and a realistic (though unimportant) constant temperature T w at this internal

15 SPE MS 15 Figure 7 Subgridding in the method of Multiple INteracting Continua (MINC) (from Pruess et al., 1999). boundary, a realistic constant-p condition regime was implemented, while avoiding any non-physical temperature distribution in the well vicinity when the simulations were conducted non-isothermally (the large advective flows into the gridblocks representing the well from its immediate neighbors eliminated any unrealistic heat transfer effects that could have resulted from an incorrect k and/or T w ). Numerical treatment of the fractured subdomains The primary (hydraulically-induced) fractures in subdomain S2 as treated as discrete fractures, i.e., they are described as a porous medium with its own distinct flow properties. This approach allows seamless interaction with other fractures and the matrix in the remaining subdomains of the system. The flow bewteen fractures and matrix in the remaining subdomains is more complex, and is described by the method of Multiple INteracting Continua (MINC) proposed by Pruess and Narasimhan (1982; 1985). In MINC, resolution of these gradients is achieved by appropriate subgridding of the matrix blocks, as shown in Figure 7. The MINC concept is based on the notion that changes in fluid pressures, temperatures, phase compositions, etc., due to the presence of sinks and sources (production and injection wells) will propagate rapidly through the fracture system, while invading the tight matrix blocks only slowly. Therefore, changes in matrix conditions will (locally) be controlled by the distance from the fractures. Fluid and heat flow from the fractures into the matrix blocks, or from the matrix blocks into the fractures, can then be modeled by means of one-dimensional strings of nested grid blocks, as shown in Figure 7. In general it is not necessary to explicitly consider subgrids in all of the matrix blocks separately. Within a certain reservoir grid block, all fractures will be lumped into continuum # 1, all matrix material within a certain distance from the fractures will be lumped into continuum # 2, matrix material at larger distance becomes continuum # 3, and so on. Quantitatively, the subgridding defines nested shells that are specified by means of a set of volume fractions V Fj into which the primary porous medium grid blocks are partitioned. The information on fracturing (spacing, number of sets, shape of matrix blocks) required for this is provided by a proximity function which expresses, for a given reservoir domain V o, the total fraction of matrix material within a distance l from the fractures. If only two continua are specified (one for fractures, one for matrix), the MINC approach reduces to the conventional double-porosity method. Full details can be found in Pruess (1983).

16 16 SPE MS Figure 8 Flow connections in the dual permeability model. Global flow occurs between both fracture (F) and matrix (M) grid blocks. In addition there is F-M interporosity flow (from Pruess et al., 1999). The MINC-method can easily describe global matrix-matrix flow. Figure 8 shows the most general approach (implemented in this study), often referred to as dual permeability, in which global flow occurs in both fracture and matrix continua, and matrix-matrix flow is also considered. In this case, two continua are specified (as in the case of the double-porosity approach), but the difference is the number of connections, i.e., there is the extra matrix-matrix connection. For any given fractured reservoir flow problem, selection of the most appropriate gridding scheme must be based on a careful consideration of the physical and geometric conditions of flow. The MINC approach is not applicable to systems in which fracturing is so sparse that the fractures cannot be approximated as a continuum. In our study, the proximity function is a logartithmic function of distance from the fracture face. The subdomains S1 to S4 are subdivided into gridblocks using variations of the MINC concept. The main MINC approach used in this study is limited to the dual permeability method, using N NS 1 nested shells to describe the matrix continua (as opposed to the standard N NS 1 of the original description of the concept). A problem that quickly arises with increasingly more complex MINC construncts and increasing number of nested shells is the rapidly increasingly number of gridblocks and connections, resulting in a rapidly increasing size of the Jacobian matrix and of the number of non-zero entries in it. Thus, accounting for fracture and matrix in lieu of the conventional single gridbocks (Effective Continuum Model ECM) increases the number of equations by a factor of (N NS 1), while the number of connections increases from 1 in ECM to 2N NS 1 for the dual porosity model, to 3N NS 1 for a dual permeability model, to 5N NS 2 for a triple-continuum model (Wu et al., 2009). The problem becomes much more complex when accounting for the effect of fractures that intersect more than one of the S1 to S4 subdomains, as additional connections (describing the extra flows) need to be added (and a special code was developed to identify and mathematically describe such connections). Because of the very large size of the Jacobian matrices and the complexity of describing the flow between connections, our study was limited to evaluation of the dual permeability model with a single matrix shell, and did not delve at this time into the more complex triple-continuum model (Wu et al., 2009). This will be a future activity. Results and Discussion The reference cases We consider two reference cases. The first case (Case R) involves only the native S1 system, and represents the unstimulated formation. The second reference case (Case RF) involves a stimulated system

17 SPE MS 17 Figure 9 Performance of the reference Cases R and RF, and effect of matrix permeability on the rate of oil mass production Q. with hydraulically-induced fractures, i.e., it comprises both S1 and S2 subdomains. The dimensions of these two systems are listed in Table 1. Figure 9 shows the mass rate Q of light oil (having the properties of n-octane) production in Cases R and RF, and Figure 10 shows the corresponding cumulative mass M of produced light oil during the same period. Additionally, Figures 9 and 10 include the evolution of Q and M for various levels of matrix permeability. The effect even of a single fracture (the primary fracture in Case RF) on Q and M is shown to be rather dramatic, increasing their early levels by at least an order of magnitude over that in Case R. This holds true for every level of matrix permeability. As expected, higher matrix permeability corresponds with higher early Q and M, although this cannot be expected to last during the entire duration of production because of the amount of oild in the system is fixed and the same in all cases. Gas displacement processes Figure 11 shows the Q corresponding to a continuous gas drive obtained by maintaining the upper well (see Figures 5 and 6) at a constant bottomhole pressure P w Pa in both Cases R and RF. Given the very low permeability of the formation, a constant P w is a reasonable approach. We used 2 different gases: CH 4 and N 2. Despite expectations of a better performance of the CH 4 drive (because of its significant dissolution in the light oil, and the corresponding expected decrease in viscosity), the results for the two gases were practically the same. This was attributed to the fact that the invading CH 4 is acting mainly as a displacement agent, and its advancing front does not have the opportunity (at least, within the time frame of this study) to effect a change in viscosity and, consequently, an increase in production. The results in Figure 11 indicate that gas displacement does have a positive impact by increasing Q. Its effects become evident earlier in case RF because of the light oil displacement into the hydraulic fracture in the vicinity of the injection well. The increase in Q becomes evident later in the non-fractured system of Case R, but the effect of displacement is more pronounced (in a relative sense) than that in Case R. However, what is interesting is that the gas displacement process can at a later time compensate (at least

18 18 SPE MS Figure 11 Effect of a displacement process (gas drive using CH4 and N2) on Q. No discernible difference is observed between the production for CH4 and N2 drives. Figure 12 Effect of heating on Q. The impact of heating beginning at the time of production is minimal.

19 SPE MS 19 Figure 13 Effect of the presence of native fractures (NF) or similarly-acting secondary fractures on Q. The presence of fractures has the most pronounced positive effect on production.

20 20 SPE MS Figure 10 Performance of the reference Cases R and RF, and effect of matrix permeability on the cumulative mass of produced oil M. partially) for the absence of the primary fracture. However, there is evidence that the positive effects of displacement are accompanied by an undesirable change in the relative permeability of the oil once the gas front reaches the production well. The longer-term behavior is still under investigation. Thermal processes Here we investigate the effect of heating to effect a reduction in the viscosity of the light oil, and the overall effect of heating on gas production. We have two heating cases in each of the systems: Case R-H1 involves heating that begins at the time of production, while Case R-H2 involves heating for an entire year prior to the beginning of production. Cases RF-H1 and RF-H2 are analogous. Heating is accomplished by flowing a fluid at a constant temperature T H 95 o C in the upper well (see Figures 5 and 6). Figure 12 shows that the positive effects of heating are limited in Cases R-H1 and R-H2, and they appear quite late in the production process. This was expected because of the reliance on the very slow method of heat conduction through the matrix (an unavoidable effect of the low permability), which does not allow fast heat transfer from the upper well toward to the rest of the formation. The situation is significantly different in Cases R-H2 and RF-H2. In these cases the increase in Q is significant, i.e., by a factor between 2 and 3 over the non-heated cases. However, this has to be considered vis-à-vis the economic implications of continuous heating for an entire year without production. The heating positive effect on Q is more pronounced in the presence of the primary fracture because of the ability of flow into both the fracture and the well. Additionally, the heating effect cannot compensate for the absence of a fracture in the unfractured system, the Q of which is lower than that for the fractured system (the reversal at later times is caused by the exhaustion of the light oil resource). Native and/or secondary fracturing Here we investigate the effect of the presence of native fractures (NF) or secondary fractures on Q. Addition of such fractures to Case RF turns it from its original Type I status into a Type III system. Case R is now enhanced with a native fracture system. In both cases, the NFs are represented as a 3D system

21 SPE MS 21 with a fracture density of 1 per m, and an aperture of 0.1 mm, and the domain is simulated using the MIC concept. The results in Figure 13 indicate that the addition of the NFs has a dramatic effect on production, which incrases (initially, before the onset of exhaustion of the resource) by several times over the reference Cases R and RF. The impact is more pronounced in Case R. Conclusions The following conclusions can be drawn from this study: X The presence of fractures (hydraulically-induced or native) is by far the most important factor in increasing the production of light oil from tight- and shale reservoirs. The importance of these fractures easily outstrips every other method we investigated (at least within the limited time frame of this study). X Heating that begins at the time of production appears to have a very limited positive effect on production of light oil from tight- and shale reservoirs. Early heating before the onset of production is quite beneficial, but the economics of heating for a prolonged period (on the order of a year) have to be considered. X Gas drives have a positive impact, but this is not very pronounced and appears later in the production process. The benefits of viscosity reduction through CH 4 dissolution in the light oil (vs. N 2 -drive) are not evident during the time frame of this study, possibly because the advancing CH 4 front acts only as a displacement agent without sufficient penetration into the light oil. Acknowledgments The authors wish to thank the Crisman Institute for Petroleum Research at Texas A&M University for partial support of this work. Nomenclature A surface area, surface area of subdomain n, m 2 b Klinkenberg parameter, Pa B gi Langmuir constant of natural gas component g i, 1/Pa c t Total compressibility, 1/Pa Multicomponent molecular diffusion coefficient of component in the gas phase, m 2 s 1 Dispersion tensor of component in phase f Radiance emittance factor, dimensionless F Flux vector of component, kgm 2 s 1, g Gravitational acceleration vector, m s 2 h Specific enthalpy (component), J/kg H Specific enthalpy (phase), J/kg H g i Temperature-dependent Henry s factor describing the solution of gas g i (i 1,...,N G ) in H 2 O, Pa Diffusive mass flux of component in the phase, kgm 2 s 1 k Intrinsic permeability, m 2 k r Relative permeability of phase, dimensionless K n Knudsen diffusion number, dimensionless m L Langmuir mass, kg/pa (equation (3)) M Cumulative mass of produced oil, kg M W Molecular weight of a component, kg/mol n Inward unit normal vector Number of components of natural gas N G

22 22 SPE MS p Pressure, Pa p L Langmuir pressure, Pa (equation (3)) Y Langmuir storage, kg/kg q Mass or heat rate of a source or sink, kg/s or W s f Fracture spacing, m S phase saturation k Matrix permeability, m 2 p i Initial reservoir pressure, Pa P w Well bottomhole pressure, Pa t Time, days T Temperature, K or C u Specific internal energy (component), J/kg U Specific internal energy (phase), J/kg v Darcy velocity vector of phase v Length of vector v, m/s V, V n volume, volume of subdomain n, m 3 Dispersive mass flux of component in phase ( A,O,C), kg m 2 s 1 Mass fraction of component in phase Greek Symbols From equation (18) Transverse and longitudinal dispersivities of component in phase, m T Thermal expansivity of the porous medium, 1/K P Medium pore compressibility, 1/Pa T Turbulence correction factor, dimensionless (equation (20)) Permeability reduction factor, dimensionless (equation (26)) n Surface area, surface area of subdomain n, m 2 Mean free path of a gas, m Porosity Viscosity, Pa.s density, kg m 3 0 Stefan-Boltzmann constant Jm 2 K 4 Tortuosity of phase, dimensionless Subscripts and Superscripts Denotes a phase ( G, A, O, C) Heat (treated as a pseudo-component) RD Denotes dry (gas-saturated) thermal consuctivity RW Denotes wet (liquid-saturated) thermal consuctivity Indicates a component g i (i 1,...,N G ), o j (j 1,2), w 0 Denotes initial state A Denotes aqueous phase cgw Gas-water capillary pressure C Denotes liquid CO 2 phase g i Denotes a component of the natural gas (i 1,...,N G ) G Gas phase o j Denotes an oil component of the organic phase (i 1,...,N G )

23 SPE MS 23 O Organic phase r Denotes an oil component of the organic phase (i 1,...,N G ) R Rock w Denotes water component (e.g., w) or well (e.g., P w ) References Anderson, D.M., Nobakht, M., Moghadam, S. and Mattar, L Analysis of Production Data from Fractured Shale Gas Well. Paper SPE presented at the SPE Unconventional Gas Conference held in Pittsburgh, Pennsylvania, February Aguilera, R Flow Units: From Conventional to Tight Gas to Shale Gas Reservoirs. Paper SPE presented at the Trinidad and Tobago Energy Resources Converence held in Port of Spain, Trinidad, June Amini, S., Ilk, D., and Blasingame, T.A Evaluation of the Elliptical Flow Period for Hydraulically-Fractured Wells in Tight Gas Sands Theoretical Aspects and Practical Considerations. Paper SPE presented at the SPE Hydraulic Fracturing Technology Conference, College Station, Texas, January. Brown, M., Ozkan, E., Raghavan, R., Kazemi, H. Practical Solutions for Pressure Transient Responses of Fractured Horizontal Wells in Unconventional Reservoirs. Paper SPE presented at the 2009 SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 4-7 October Bowker, K.A Barnett Shale Gas Production, Fort Worth Basin: Issues and Discussion. AAPG Bulletin 91 (4): Bumb, A.C. and McKee, C.R Gas Well Testing in the Presence of Desorption for Coalbed Methane and Devonian Shale. SPE Formation Evaluation (March 1988): SPE PA. Carlson, E.S. and Mercer, J.C Devonian Shale Gas Production: Mechanisms and Simple Models. JPT (April 1991): SPE PA. Cinco-Ley, H. and Samaniego, F Pressure Transient Analysis for Naturally Fractured Reservoirs. Paper SPE presented at the SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana, September. Crafton, J.W Flowback Performance in Intensely Naturally Fractured Shale Gas Reservoirs. Paper SPE presented at the SPE Unconventional Gas Conference held in Pittsburgh, Pennsylvania, USA, February Currie, S.M., Ilk, D. and Blasingame, T.A Continuous Estimation of Ultimate Recovery. Paper SPE presented at the SPE Unconventional Gas Conference held in Pittsburgh, Pennsylvania, USA, February Curtis, J.B Fractured Shale-Gas Systems. AAPG Bulletin 86 (11): Cox, D.O., Kuuskraa, V.A., and Hansen, J.T Advanced Type Curve Analysis for Low Permeability Gas Reservoirs. Paper SPE presented at the Gas Technology Conference, Calgary, Alberta, 28 April 1 May. De Swaan, A.O Analytic Solutions for Determining Naturally Fractured Reservoir Properties by Well Testing. Society of Petroleum Engineers Journal (June 1976): SPE-5346-PA. Dougherty, E.L. and Chang, J A Methodology to Quickly Estimate the Probably Value of Shale Gas Well. Paper SPE presented at the SPE Western Regional Metting held in Anaheim, California, USA, May Du, C.M., Zhang, X., Lang, Z., Gu, H., Hay, B., Tushingham, K. and Ma, Y.Z Modeling Hydraulic Fracturing Induced Fracture Networks in Shale Gas Reservoirs as a Dual Porosity System. Paper SPE presented at the CPS/SPE International Oil & Gas Conference and Exhibition in China held in Beijing, China, 8-10 June 2010.