The ever-increasing level of geological details and complexity of reservoir models have created computational difficulties for reservoir simulators so that upscaling of properties such as permeability have become common in petroleum industry. In an upscaling an averaging technique is implemented to capture equivalent values for coarser scales, however in presence of heterogeneity and subgrid complex features the averaging associates with numerical dispersion errors. Alternatively researchers developed algorithms that either involves enhancing computational speed and memory capacity or tries to solve a large scale problem in decomposable parts within efficient methods known as Multiscale. As an alternative to multiscale approach, we developed modified version of an upscaling downscaling method that mimics the behaviour of a reference solution of large model by addition of a post-processing step called nested- gridding downscaling to a correctional iterative upscaling method. In doing so, we devised spatial and temporal adaptivity in both upscaling and downscaling to remedy the shortcomings of a conventional upscaling method. Although not so exact as recently developed multiscale methods in convergence to fine solution, the comparable simplicity and robustness of the algorithm makes it desirable for practical simulation of flow where the main important output is recovery curves of engaging fluids in porous media. For comparison purposes we use pressure-solver upscaling method. We examined the performance of the methods on water flooding of highly heterogeneous porous media. By increasing the accuracy of upscaling, saturation errors decrease and consequently the production curves approach to the reference fine model. The implementation of adaptivity confirms the numerical efficiency compared to fine scale simulation, however, the time saving of conventional upscaling is several order higher than the scheme of interest.
展开▼
