Parallelization of a reservoir simulator

Abstract

Numerical reservoir simulation demands very high computing performance. Vectorization has been widely used as a performance enhancing technique. Parallel computing can also be used to satisfy the increasing requirements of more sophisticated models. This paper presents the implementation of a novel method of parallelizing an already vectorized reservoir simulator, which has the significant advantage of reducing not only the elapsed CPU-time, but also the total CPU-time used. The method relies on the physical independence of isolated geological structures, coupled through the wells, a feature present in many of the hydrocarbon reservoirs found in the North Sea. The vectorized and parallelized code has already been in production a period of time. Results are shown for a realistic case.