A parallel implementation via the OpenMP API for numerical simulation of natural gas reservoirs

Abstract

Over the past decades, the oil and gas industry has increasingly used resources to reduce computational costs in reservoir numerical simulations. Realistic case studies generally lead to the solution of systems of nonlinear algebraic equations that require significant computational processing and memory efforts. We obtain such equations from a process of discretization of partial differential equations used in reservoir flow modeling. An example of a technique applied to high-performance implementations that leads to reduced processing time is the OpenMP Application Programming Interface (API), based on shared memory usage and threads. In this work, OpenMP is used to improve the performance of a numerical flow simulator in natural gas reservoirs. Jacobi, Gauss-Seidel, and SOR stationary methods for the solution of algebraic equation systems were parallelized and compared. We choose the SOR method to be applied in studies involving the variation of the number of computational mesh volumes and the flow characteristics. In all simulations performed, gains were obtained with parallelization when compared to the performance of serial versions, reaching maximum speedup values higher than 5 in some cases.