A Novel Adaptive Anisotropic Grid Framework for Efficient Reservoir Simulation

Abstract We present a novel parallel Adaptive Mesh Refinement (AMR) framework for the effective simulation of reservoir simulation processes. The framework is based on a Cartesian Cell-based Anisotropic Refinement (CCAR) strategy. The development was motivated by compositional and non-isothermal recovery methods, in particular miscible gas injection and in-situ combustion, and particularly suited to these applications. CCAR offers aggressive grid adaptation, which allows quick transition to the very fine local grids necessary to resolve solution fronts and the fine local grids desirable to resolve reservoir heterogeneity. CCAR is computationally efficient because of the underlying Cartesian topology, and requires minimal user intervention in the grid generation process. We use static refinement to generate a fixed base CCAR grid that accurately resolves the flow field, and propose a multi-level local-global upscaling strategy to determine appropriate permeability values for the CCAR grid. Dynamic refinement criteria are formulated to add adaptivity to the base grid where desired for accurate computation of transport. We also present a novel higher order pressure solver on the CCAR grids. The discretization uses a compact stencil, which together with an effective storage scheme leads to computational efficiency. The equations are solved with a parallel Algebraic Multigrid code.

A Novel Adaptive Anisotropic Grid Framework for Efficient Reservoir Simulation Conference Paper