A very high-order flux reconstruction approach coupled to the MPFA-QL finite volume method for the numerical simulation of oil-water flows in 2D petroleum reservoirs

Abstract

In recent years, high-order methods for the discretization of the advection dominated saturation equation, and that results from the modelling of the 2D Oil-Water displacement through porous formations, have received renewed attention from the CFD community, as a suitable alternative to the traditional first-order upwind (FOU) finite volume (FV) scheme, which is the standard in industry. A key aspect of high-order methods applied to numerical simulation of petroleum reservoirs is the lower computational cost for a comparable degree of accuracy against their lower-order counterparts. In this context, in the present paper, for the first time in literature, we propose the use of the very high order Flux Reconstruction (FR) formulation for the discretization of the advective terms in the saturation equation using unstructured quadrilateral meshes. To solve the elliptic pressure equation, we use the recently developed Multipoint Flux Approximation method with a Quasi-Local Stencil (MPFA-QL). This method is very robust and able to handle highly heterogeneous and anisotropic permeability tensors. The system of equations is solved in a segregated manner by using the classical Implicit Pressure Explicit Saturation (IMPES) procedure. To properly couple the cell-centered MPFA-QL method with the FR formulation, it is necessary to obtain an adequate velocity reconstruction throughout the control volumes of the mesh. As the cell-centered finite volume method naturally delivers fluxes across control surfaces of the primal grid, then, we use a reconstruction operator based on the first-order Brezzi–Douglas–Marini (BDM) space, and the Piola transformation, to get the complete knowledge of the velocity field throughout the domain. To eliminate spurious oscillations around the discontinuities, which are typical in high-order methods, we have used a hierarchical multidimensional limiting process (hMLP). To compute fluxes over the control surfaces we have used the Roe-E approximate Riemann solver. Finally, through some representative examples, we show the accuracy, efficiency and robustness of our new high order numerical methodology to model oil-water displacements in highly heterogeneous and anisotropic petroleum reservoirs.