GPU-based JFNG method for power system transient dynamic simulation

To accelerate dynamic security assessments of large scale transmission power systems, efficient transient simulation tools are desperately required, especially the advanced nonlinear equation solvers. Jacobian-free Newton-GMRES(m) [JFNG(m)] is a high efficiency nonlinear equations solver. However, the computation burden of matrix-vector multiplication dramatically increases as the system enlarges. In this paper, the parallelism capability of GPU(Graphic Processing Unit) is fully exploited to speed up the dynamic simulation of large-scale power system. A JFNG-based power system dynamic simulation program is designed and implemented on GPU. In addition, an enhanced precondition method and special data exchange strategies are proposed to make full use of the fine-grained parallelism on GPU, which greatly enhances the efficiency of the simulation.