Parallel Simulation of Full-Field Polymer Flooding

摘要

In recent years, polymer flooding has become a mature technology to improve the reservoir recovery rate and has been successfully implemented in many oil fields. Because of high operating costs of a polymer flooding project and the impact to the environment, the polymer injected per unit volume should be tracked for each polymer flooding pattern. Hence, the full-field reservoir simulations with fine-scale grids are needed to capture fine-scale phenomena and to optimize the process. To meet these objectives, a polymer module has been developed in our in-house parallel black oil simulator, which allows parallel simulations using clusters and supercomputers. With our parallel simulator, the elapsed time of full-field simulations with millions of grid blocks can be reduced from days to hours or minutes, and even for a much larger model with hundreds of millions ofgrid blocks, the simulations can be finished in practical time. In order to guarantee the computational efficiency and the parallel scalability, an inexact Newton method is applied and a new CPR (Constrained Pressure Residual)-type preconditioner is designed. In this paper, a SPE10-based polymer flooding case is tested, the performance of nonlinear and linear solvers are robust, and an encouraging parallel scalability is obtained. The second case is a full-field polymer flooding case, for which a coarse grid model anda fine grid model are both used. The results show the differences of the oil production rate and the water cut, which illustrates the accuracy of fine-grid simulations. For the fine grid model, up to 1024 CPU cores are employed, and an excellent speedup is achieved.