Parallel Optimizations and Load Balancing for Multiple Layers Numerical Simulation of Reservoir

In numerical reservoir simulations, the improvement of resource utilization and parallel performance is usually resolved by a fast parallel linear system solver and appropriate load-balancing strategy. Presented in this paper is an optimization approach to improve the parallel computing performance of numerical reservoir simulations with multiple layers. It is characteristic of a rapid convergence and the capability of selection of two different domain sizes. A new load-balancing algorithm is also designed and shown to successfully reduce states of load unbalancing in the simulations. The results show that the optimized parallel processing generated by this global optimization can reduce computational time considerably and get higher speedup. This approach can be applied to investigations of numerical simulation of the multi-layer reservoir.