GPU-Accelerated Sparse LU Factorization for Concurrent Analysis of Large-Scale Power Systems

Analyzing a massive number of Power Flow (PF) equations even on almost identical or similar network topology is a highly time-consuming process for large-scale power systems. The major computation time is hoarded by the iterative linear solver to solve nonlinear equations to achieve convergence. This is a paramount concern for any PF analysis methods, such as Newton-Raphson (NR) method, Gauss-Seidel (GS) method, Fast Decoupled (FD) method, etc. The computation time increases by increasing the size and complexity of the system. This paper presents a GPU-accelerated sparse matrix solver that is fast to be implemented in the real-time and concurrent analysis of large-scale power systems. The achieved results show a computational gain of over six times compared to MATPOWER, which is a well-known tool for PF analysis of power systems.