Layer | Fill | Outline |
---|
Map layers
Theme | Visible | Selectable | Appearance | Zoom Range (now: 0) |
---|
Fill | Stroke |
---|---|
Collaborating Authors
Results
Summary Dispersion (or local mixing) degrades miscibility in miscible-flood displacements by interfering with the transfer of intermediate components that develop miscibility. Dispersion, however, also can improve oil recovery by increasing sweep efficiency. Either way, dispersion is an important factor in understanding miscible-flood performance. This paper investigates longitudinal and transverse local mixing in a finite-difference compositional simulator at different scales (both fine and coarse scale) using a 2D convection-dispersion model. All simulations were of constant-mobility and -density, first-contact miscible flow. The model allows for variations of velocity in both directions. We analyzed local (gridblock) concentration profiles for various miscible-displacement models with different scales of heterogeneity and permeability autocorrelation lengths. To infer dispersivity, we fitted an analytical 2D convection-dispersion model to the local concentration profile to determine local longitudinal and transverse dispersivities simultaneously. Streamlines of simulation models were traced using the algorithm proposed by Pollock (1988). To our knowledge, this is the first systematic attempt to numerically study local transverse dispersivity. The results show that transverse mixing, which is usually neglected in the 1D convection-dispersion model of dispersion, is significant when the flow direction changes locally as a result of heterogeneity. The computed streamlines, which highlight the variation in flow directions, agree with the computed transverse-dispersivity trends. We find that both transverse and longitudinal dispersion can grow with travel distance and that there are several instances in which transverse dispersion is the larger of the two. Often, the variations in the streamlines are suppressed (homogenized) during upscaling. This paper gives a quantitative and systematic procedure to estimate the degree of transverse mixing (dispersivity) in any model. We conclude that local mixing, including transverse mixing, should be considered when upscaling a fine-scale model for miscible displacement to ensure proper preservation of fine-scale sweep and displacement efficiency and ultimate oil recovery for miscible-displacement simulations.
- Europe > Romania > Black Sea > Muridava Block > D-T Field (0.98)
- North America > United States > Mississippi > Anna Field (0.89)
- Europe > United Kingdom > North Sea > Southern North Sea > Southern Gas Basin > Sole Pit Basin > Block 49/22 > Victor Field > Leman Sandstone Formation (0.89)
- Europe > United Kingdom > North Sea > Southern North Sea > Southern Gas Basin > Sole Pit Basin > Block 49/17 > Victor Field > Leman Sandstone Formation (0.89)
Summary Inaccurate modeling of reservoir mixing by using large gridblocks in compositional simulation can affect recoveries significantly in miscible gasfloods and lead to inaccurate predictions of recovery performance. Reservoir mixing or dispersion is caused by diffusion of particles across streamlines; mixing can be enhanced significantly if the surface area of contact between the reservoir and injected fluid is increased as fluids propagate through the reservoir. A common way to convert geological models into simulation models is to upscale permeabilities on the basis of reservoir heterogeneity. Upscaling affects the degree of mixing that is modeled, but the importance of reservoir mixing in upscaling is largely ignored. This paper shows how to estimate the level of mixing in a reservoir and how to incorporate mixing into the upscaling procedure. We derive the key scaling groups for first-contact miscible (FCM) flow and show how they have an impact on reservoir mixing. Heterogeneities are assumed to dominate the flow regime so that gravity effects are negligible. We examine only local mixing, not apparent mixing caused by variations in streamline path lengths (convective spreading). Local mixing is important because it affects the strength of the injected fluid and can cause an otherwise multicontact miscible (MCM) flood to become immiscible. More than 1,000 2D numerical simulations are carried out using experimental design to estimate dispersivity as a function of the derived scaling groups. We show that reservoir mixing is enhanced owing to fluid propagation through heterogeneous media. Because mixing is dependent on heterogeneities, upscaling is an iterative process in which the level of mixing in both the longitudinal and transverse directions must be matched from the fine to the coarse scale. The most important groups that affect mixing are the mobility ratio, dispersion number, correlation lengths, and the Dykstra-Parson coefficient. Large dispersion numbers yield greater dispersivities away from the injection well. We show through simulations of both FCM and MCM floods that gridblock size can be increased significantly when reservoir mixing is large. Heterogeneous reservoirs with large longitudinal correlation lengths can be upscaled to larger gridblocks than reservoirs with random permeability fields. This paper shows how to determine a priori the maximum gridblock size allowed in both the x- and z-directions to predict the oil recovery from miscible gasfloods accurately.
- Europe > United Kingdom > North Sea > Northern North Sea > East Shetland Basin > Block 211/7a > Magnus Field > Kimmeridge Formation > Magnus Formation (0.94)
- Europe > United Kingdom > North Sea > Northern North Sea > East Shetland Basin > Block 211/7a > Magnus Field > Kimmeridge Formation > Lower Kimmeridge Clay Formation (0.94)
- Europe > United Kingdom > North Sea > Northern North Sea > East Shetland Basin > Block 211/12a > Magnus Field > Kimmeridge Formation > Magnus Formation (0.94)
- (8 more...)
Summary Equations of state (EOSs) are typically tuned to black-oil pressure/volume/temperature (PVT) data such as constant volume-depletion, constant-composition-expansion, differential-liberation, and separator tests. Other PVT data more appropriate for gas injection could include multicontact and swelling tests and slimtube tests. The standard method of tuning, however, does not typically incorporate important displacement parameters, such as the minimum miscibility pressure (MMP), minimum miscibility enrichment (MME), or the likely compositions that result in a reservoir from condensing-vaporizing (CV) displacements. This paper demonstrates an improved reservoir-fluid-characterization procedure for miscible gas floods that can represent the interaction of flow and phase behavior more accurately. We demonstrate the approach for two displacements, an 11-component CO2 flood and a 12-component enriched-gas flood. The method-of-characteristic (MOC) theory is used to determine the MME (or MMP) of both lumped and unlumped models. The results show that by tuning to the calculated MME/MMP, fewer pseudocomponents are required to characterize the fluid than with conventional tuning methods. For the cases studied, fluids lumped to as few as four or five pseudocomponents can provide a good match to the composition profiles and oil recoveries of the unlumped models. Introduction Gas injection into oil reservoirs results in complex interactions of flow with phase behavior that often are not modeled accurately by black-oil simulation. This is especially true for miscible or nearly miscible floods in which significant mass transfer occurs between the hydrocarbon phases. Such floods are modeled best by compositional simulation. A significant disadvantage of compositional simulation, however, is that it is more computationally intensive than black-oil simulation. The primary reason for the increased computational time is the result of solving repeated flash calculations with cubic EOSs. The use of fewer pseudocomponents could reduce the flash computation time, but fewer components results in poor fluid characterizations and reduced accuracy. Reservoir oils typically are subjected to standard black-oil PVT experiments that give volumetric behavior for recovery predictions from conventional methods, such as waterflooding. These experiments include constant-volume-depletion, differential-liberation, constant-composition-expansion, and separator tests. Standard PVT experiments, however, do not provide sufficient phase-behavior data in the range of compositions that result from mixing of gas with resident oil. For gas floods, multicontact experiments, along with swelling tests and slimtube experiments, are sometimes performed (Pedersen et al. 1989). Most gasfloods, such as those with CO2 and enriched-gas injection, however, have features of both condensing and vaporizing drives (Zick 1986; Stalkup 1987; Johns et al. 1993). Miscibility in these CV drives is developed in the transition zone between the condensing and vaporizing regions at an equilibrium tie line, known as the crossover tie line (Johns et al. 1993, 2002; Johns and Orr 1996). Multicontact tests attempt to mimic the composition paths that result from either vaporizing or condensing drives, but not both. Thus, these tests do not provide sufficient PVT data in the compositional range of interest, especially in the transition zone near the critical region in which miscibility is developed in CV drives. Slimtube tests can and should be used to tune an EOS by matching the experimental recoveries with 1D compositional simulations (Shanin and Kremesec 1992). Slimtube tests, however, are expensive and time-consuming to obtain, and their recoveries can be affected by dispersion and relative permeabilities (Johns et al. 1994; Solano et al. 2001). Slimtube tests are not always available, and even if they are, it would be helpful to have a method that is not dependent on the level of dispersion or relative permeability parameters, and one that is very fast so that regression of the MMP/MME is possible. Recent research has shown how to calculate the dispersion-free MMP/MME from an EOS by MOC (Jessen et al. 1998; Wang and Orr 2002; Yuan 2003; Yuan and Johns 2005). EOS are used to predict the compositions and volumetric behavior that result when oil and gas mix in the reservoir. These EOS fluid characterizations must be tuned to match the PVT behavior of the original reservoir fluid. The process of tuning an EOS involves:selection of the pseudocomponents, determination of EOS properties for the pseudocomponents, and adjustment of pseudocomponent EOS properties by regression to the PVT data. The fluid characterizations that result from the lumping and tuning process are dependent on the method used and the experimental PVT data available (Pedersen et al. 1989). Often the tuning process involves iteration and subjectivity concerning which parameters to regress and the number of pseudocomponents to use. The usual approach is first to lump the original fluid analysis to as few as 12 to 15 components and pseudocomponents. This EOS model is tuned to match the available PVT data, and it can be lumped into fewer pseudocomponents as needed. There are several methods for lumping components into pseudocomponents and determining their EOS properties (Danesh 1998; Pedersen and Christensen 2006). The simplest methods assign pseudocomponents on the basis of component mole fractions (Cotterman and Prausnitz 1985), mass fractions (Pedersen et al. 1985), ranges in molecular weights (Whitson 1983), and K -values (Li et al. 1985; Newley and Merrill 1991) pore-complex methods include the statistical approach of Mehra et al. (1982). The method used in this research is that of Newley and Merrill (1991), which is based on K-values at some selected feed composition. We use this method because analytical theory has demonstrated that components within the reservoir are chromatographically separated by their K-values (Orr 2007). Several regression procedures have been suggested for tuning EOS characterizations (Hong 1982; Fong et al. 1992; Khan et al. 1992; Liu 1999; Zurrita and McCain 2002). The selection of parameters to tune to match a set of PVT data is more of an art than an exact science. Adjusting too many parameters could result in poor PVT predictions away from the range of the measured PVT data. Jhaveri and Youngren (1984) recommend classifying PVT experimental data into volumetric and compositional data. Preselected EOS parameters are adjusted to match the compositional data first, and, then, volumetric data are matched by adjusting the volumetric-shift parameters. Pedersen and Christensen (2006) showed that fluid characterizations can predict fluid properties better when most binary-interaction parameters (BIPs) between hydrocarbon components are zero. Typically, the parameters associated with the heaviest pseudocomponents are adjusted by up to 10% to match the compositional data because these components have properties with the largest measurement uncertainties (Danesh 1998; Christensen 1999; Pedersen and Christensen 2006). This paper presents a method to improve fluid characterizations that can account for the complex composition paths that result from a CV process. Such a method can be used to reduce the number of required pseudocomponents for use in compositional simulation. The proposed method is based on matching all available PVT data and the analytical calculation of MMP/MME from the lumped EOS models to the original unlumped fluid characterizations. The lumping and tuning procedure is demonstrated for 11-component and 12-component oil displacements by gas using the Peng-Robinson EOS (Peng and Robinson 1976).
- Europe (0.93)
- North America > United States > Texas (0.49)
A Screening Model for CO2 Flooding and Storage in Gulf Coast Reservoirs Based on Dimensionless Groups
Wood, Derek J. (The University of Texas at Austin) | Lake, Larry W. (The University of Texas at Austin) | Johns, Russell T. (The University of Texas at Austin) | Nunez, Vanessa (The University of Texas at Austin)
Summary Concerns over global warming have led to interest in removing greenhouse gases, specifically CO2, from the atmosphere. Sequestration of CO2 in oil reservoirs as part of enhanced oil recovery (EOR) projects is one method that is being considered. This paper first presents the scaling groups necessary to describe CO2 flooding for a typical line-drive pattern and then uses these groups in a Box-Behnken experimental design to create a screening model most applicable to candidate Gulf Coast reservoirs (Box and Behnken 1960). By generating oil recovery and CO2 storage curves, the model estimates the cumulative oil recovery and CO2 storage potential for a given reservoir. Past screening models—Rivas et al. (1992) and Diaz et al. (1996)—focused only on oil recovery and simply assigned qualitative rankings to reservoirs. Models that did include quantitative results, including CO2 Prophet (Dobitz and Prieditis 1994) and the CO2 Predictive Model (Paul et al. 1984), did not include the effects of dip. This model focuses on both oil recovery and CO2 storage potential, produces quantitative results for each, and includes the effects of dip. This model quickly estimates the oil recovery and CO2 storage potential for a reservoir. Operators can quickly screen large databases of reservoirs to identify the best candidates for CO2 flooding and storage. The scaling groups also provide the basis for future models that may be more specific to other regions. The results show that continuous CO2 flooding can be fully described using 10 dimensionless groups: aspect ratio, dip angle group, water and CO2 mobility ratios, buoyancy number, dimensionless injection and producing pressures, residual oil saturation to water and gas, and initial oil saturation. The effects of capillary forces and dispersion were secondary effects in this model and were not included in the scaling. Dimensionless oil recovery was effectively modeled with the dimensionless oil breakthrough time and the dimensionless recovery at three different dimensionless times, while CO2 storage potential was calculated only at the final dimensionless time. The reservoir-specific parameters discussed above were calculated from response surface fits. The scaling does not work as well at small buoyancy numbers; however, it is effective in the range of values typical of Gulf Coast reservoirs. Introduction CO2 flooding is a popular EOR technique; however, it has not heretofore been scaled for dipping reservoirs. Scaling is done using a process called inspectional analysis. In this process, the equations governing fluid flow in a reservoir are described and then converted into dimensionless equations. For example, the variable z (distance in the vertical direction) can be transformed into a dimensionless variable by dividing by a scalar parameter z1*, which can be set equal to H, the height of the reservoir. This new group z/z1* is dimensionless. These transformations are made until the equations are entirely in dimensionless form. Then, through various assumptions and mathematical manipulations of the equations, dimensionless terms are canceled out and removed until a final group of independent dimensionless groups is extracted from the equations. Using inspectional analysis, Shook et al. (1992) scaled waterfloods for a homogeneous, 2D, cartesian, dipping reservoir with two phases (oleic and aqueous) present and found five necessary dimensionless groups. They are:RL = [Equation] effective aspect ratio Mw = [Equation] mobility ratio (water) Na = [Equation] dip angle group Ng = [Equation] buoyancy number NPc = [Equation] capillary number These groups served as the initial basis for the scaling of CO2 flooding; however, they proved insufficient. This paper presents the additional groups necessary to scale CO2 flooding. The desire to undertake CO2 flooding begets the need to identify economically attractive candidate reservoirs. Comprehensive simulations may be too costly and time-consuming when large databases of reservoirs must be evaluated. This paper presents a model based on the aforementioned dimensionless groups that quickly estimates the oil recovery and CO2 storage potential for candidate reservoirs.
Summary The computational time for conventional flash calculations increases significantly with the number of components, making it impractical for use in many fine-grid compositional simulations and other applications. Previous research to increase flash-calculation speed has been limited to those with zero binary interaction parameters (BIPs) or approximate methods based on an eigenvalue analysis of the binary interaction matrix. Practical flash calculations, however, nearly always have some nonzero BIPs. Further, the accuracy and speed of the eigenvalue methods varies depending on the choice and number of the dominant eigenvalues. This paper presents a new and simple method for significantly increasing the speed of flash calculations for any number of nonzero BIPs. The approach requires the solution of up to six reduced parameters regardless of fluid complexity or the number of components and is based on decomposing the BIPs into two parameters using a simple quadratic expression. The new approach is exact in that the equilibrium-phase compositions for the same BIPs are identical to those with the conventional flash calculation; no eigenvalue analysis is required. Further, the new approach eliminates the Rachford-Rice procedure (1952) and is more robust than the conventional flash-calculation procedure. We demonstrate the new approach for several example fluids and show that speedup by a factor of approximately 3 to 20 is obtained over conventional flash calculations, depending on the number of components. Introduction Gas injection into oil reservoirs results in complex interactions of flow with phase behavior that often are not modeled accurately by black-oil simulation. This is especially true for miscible or nearly miscible floods in which significant mass transfer occurs between the hydrocarbon phases. Such floods are best modeled by compositional simulation. A significant disadvantage of compositional simulation, however, is that it is much more computationally intensive than black-oil simulation. The primary reason for the increased computational time is the result of solving repeated flash calculations with cubic equations of state (EOS). Research has shown that EOS flash calculations can occupy 50 to 70% of total computational time in compositional simulations (Stenby and Wang 1993; Chang 1990). This is also true for other applications, such as multiphase flow in pipelines. The use of fewer pseudocomponents can reduce the flash computation time, but fewer components results in less accuracy (Hong 1982; Liu 2001; Egwuenu et al. 2005). This is especially true in multicontact miscible displacements, in which miscibility is developed by a combined condensing/vaporizing drive process (Zick 1986; Johns et al. 1993; Egwuenu et al. 2005). Fluid characterization by pseudocomponent models can be improved by tuning to the analytical minimum miscibility enrichment or minimum miscibility pressure (Johns et al. 1994), but those models still require significant computational time, even for fewer pseudocomponents. Another way to reduce computation time is to reduce the number of gridblocks. With coarse grids, however, numerical dispersion is large, which may cloud the results (Solano et al. 2001). Ideally, fine grids should be used that better match the level of dispersion found at the field scale. More recently, methods have been examined to find reduced parameters for flash calculations. Michelsen (1982a, 1982b, 1986) significantly increased flash-calculation speed by finding three reduced parameters, regardless of the number of components. His method, however, assumes zero BIPs, which is too restrictive for real fluid characterization. Michelsen also gave a practical method for stability calculations using the tangent plane distance (TPD) (Michelsen 1982b).
Summary Local displacement efficiency from CO2 gas injection is highly dependent on the minimum miscibility pressure (MMP). Correlations are sometimes used to estimate the MMP where the injected fluid may or may not contain impurities such as methane. These correlations, however, are based on a limited set of experimental data and, as such, are not widely applicable. They also do not account accurately for the more complex condensing/vaporizing (CV) displacement process. This paper presents new MMP correlations for the displacement of multicomponent oil by CO2 and impure CO2. The approach is to use recently developed analytical theory for MMP calculations from equations of state (EOSs)to generate MMP correlations for displacements by pure and impure CO2. The advantage of this approach is that MMPs for a wide range of temperatures and reservoir fluids can be calculated quickly and accurately without introducing uncertainties associated with slimtube MMPs and other numerical methods. The improved MMP correlations are based solely on the reservoir temperature, the molecular weight of C7+, and the percentage of intermediates (C2-C6) in the oil. The MMPs from the improved correlations are compared to currently used correlations and 41 experimentally measured MMPs. Correlations are also developed for impure-CO2 floods, in which the injection stream may contain upto 40% methane. The new correlations are more accurate for a wider range of conditions than the currently used correlations. Introduction Whorton et al. received a patent in 1952 to improve oil recovery by the injection of CO2. CO2 injection has been ongoing ever since, primarily because CO2 develops multicontact miscibility (MCM) with reservoir fluids at low pressures. There are also potential environmental benefits of CO2 injection in that subsurface sequestration of greenhouse gases has become an important U.S. priority. The MMP is an important optimization parameter in CO2 floods. Recoveries from slimtube experiments often give a slope change at the MMP. Above the MMP, slimtube recoveries (or local displacement efficiencies) typically do not increase significantly with enrichment. Thus, the accurate determination of MMP is important in gasflood design. Pseudoternary diagrams traditionally have been used to explain the behavior of multicontact miscible (MCM) gas-drive processes. Real oil displacements by CO2, however, have recently been shown to have features of both vaporizing and condensing drives. The 2D nature of pseudoternary diagrams often leads to incorrect interpretations, especially for CV drives. Analytical theory has no such restrictions and can be applied for any number of components. The CV process greatly complicates the accurate estimation of MMP in that miscibility is developed not at the leading edge (condensing region) or trailing edge(vaporizing region) of the displacement, but in between the condensing and vaporizing regions. Four primary methods have been used in recent years to determine MMPs for specific fluid displacements: slimtube experiments,10 compositional simulation,12 mixing-cell models, and analytical methods. Each of these methods has advantages and disadvantages. Slimtube experiments use real fluids but are expensive and time consuming to perform and can give misleading results depending on the level of physical dispersion present. Fine-grid compositional simulations and mixing-cell models can suffer from numerical-dispersion effects and are also time consuming to perform. Dispersion-free analytical methods are often very fast, but like simulation and mixing-cell models, they rely on an accurate fluid characterization by an EOS. A variety of correlations for the estimation of the MMP have been developed from regressions of slimtube data. Although less accurate, correlations are quick and easy to use and generally require only a few input parameters. Hence, they are very useful for fast screening of reservoirs for potential CO2flooding. They are also useful when detailed fluid characterizations are not available. One significant disadvantage of current MMP correlations is that the regressions use MMPs from slimtube data, which are themselves uncertain. Some MMP correlations require only the input of reservoir temperature and the API gravity of the reservoir fluid. Other, more-accurate correlations require reservoir temperature and the total C2-C6 content of the reservoir fluid. A few require detailed EOS characterizations. In nearly all of the correlations, the methane content of the oil is assumed to not affect the MMP significantly. Orr et al. show why this is true using analytical theory.
- North America > United States > Texas > Permian Basin > Levelland Field > Wichita-Albany Formation (0.99)
- North America > United States > Texas > Permian Basin > Levelland Field > Strawn Formation (0.99)
- North America > United States > Texas > Permian Basin > Levelland Field > Abo Formation (0.99)