|Theme||Visible||Selectable||Appearance||Zoom Range (now: 0)|
Summary Modeling the dynamic fluid behavior of low-salinity waterflooding (LSWF) at the reservoir scale is a challenge that requires a coarse-grid simulation to enable prediction in a feasible time scale. However, evidence shows that using low-resolution models will result in a considerable mismatch compared with an equivalent fine-scale model with the potential of strong, numerically induced pulses and other dispersion-related effects. This work examines two new upscaling methods that have been applied to improve the accuracy of predictions in a heterogeneous reservoir where viscous crossflow takes place. We apply two approaches to upscaling to bring the flow prediction closer to being exact. In the first method, we shift the effective-salinity range for the coarse model using algorithms that we have developed to correct for numerical dispersion and associated effects. The second upscaling method uses appropriately derived pseudorelative permeability curves. The shape of these new curves is designed using a modified fractional-flow analysis of LSWF that captures the relationship between dispersion and the waterfront velocities. This second approach removes the need for explicit simulation of salinity transport to model oil displacement. We applied these approaches in layered models and for permeability distributed as a correlated random field. Upscaling by shifting the effective-salinity range of the coarse-grid model gave a good match to the fine-scale scenario, while considerable mismatch was observed for upscaling of the absolute permeability alone. For highly coarsened models, this method of upscaling reduced the appearance of numerically induced pulses. On the other hand, upscaling by using a single (pseudo)relative permeability produced more robust results with a very promising match to the fine-scale scenario. These methods of upscaling showed promising results when they were used to scale up fully communicating and noncommunicating layers as well as models with randomly correlated permeability. Unlike documented methods in the literature, these newly derived methods take into account the substantial effects of numerical dispersion and effective concentration on fluid dynamics using mathematical tools. The methods could be applied for other models where the phase mobilities change as a result of an injected solute, such as surfactant flooding and alkaline flooding. Usually these models use two sets of relative permeability and switch from one to another as a function of the concentration of the solute.
Summary Numerical fidelity is required when using simulations to predict enhanced-oil-recovery (EOR) processes. In this paper, we investigate the conditions that lead to numerical errors when simulating low-salinity (LS) waterflooding (LSWF). We also examine how to achieve more accurate simulation results by scaling up the flow behavior in an effective manner. An implicit finite-difference numerical solver was used to simulate LSWF. The accuracy of the numerical solution has been examined as a function of changing the length of the grid cell and the timestep. Previously we have shown that numerical dispersion induces a physical retardation such that the LS front slows down while the formation water front speeds up. We also report for the first time that pulses can be generated as numerical artifacts in coarsely gridded simulations of LSWF. These effects reflect the interaction of dispersion, the effective-salinity range, and the use of upstream weighting during calculation, and can corrupt predictions of flow behavior. The effect of the size of the timestep was analyzed with respect to the Courant condition, traditionally related to explicit numerical schemes and also numerical stability conditions. We also investigated some of the nonlinear elements of the simulation model, such as the differences between the concentrations of connate water salinity and the injected brine, effective-salinity-concentration range, and the net mobility change on fluids through changing the salinity. We report that to avoid pulses it is necessary, but not sufficient, to meet the Courant condition relating timestep size to cell size. We have also developed two approaches that can be used to scale up simulations of LSWF and tackle the numerical problems. The first method is dependent on a mathematical relationship between the fractional flow, effective-salinity range, and the Péclet number and treats the effective-salinity range as a pseudofunction. The second method establishes an unconventional proxy method equivalent to pseudorelative permeabilities. A single table of pseudorelative permeability data can be used for a waterflood instead of two tables, as is usual for LSWF. This is a novel approach that removes the need for relative permeability interpolation during the simulation. Overall, by avoiding numerical errors, we help engineers to more efficiently and accurately assess the potential for improving oil recovery using LSWF and thus optimize field development. We also avoid the numerical pulses inherent in the traditional LSWF model.