GPU-Based Batch LU-Factorization Solver for Concurrent Analysis of Massive Power Flows

Abstract

In many power system applications such as N-x static security analysis and Monte-Carlo-simulation-based probabilistic power flow (PF) analysis, it is a very time consuming task to analyze massive number of power flows (PF) on identical or similar network topology. This letter presents a novel GPU-accelerated batch LU-factorization solver that achieves higher level of parallelism and better memory-access efficiency through packaging massive number of LU-factorization tasks to formulate a new larger-scale problem. The proposed solver can achieve up to 76 times speedup when compared to KLU library and lays a critical foundation for massive-PFs-solving applications.