GPU-Accelerated Power Flow Analysis

In recent years, the increasing complexity of modern power systems has made power flow analysis increasingly important. The matrix-vectorized model for power flow analysis based on the Newton-Raphson framework involves numerous complex operations during computation, such as indexing operations, which significantly degrade performance in a GPU environment. This paper presents a novel GPU-accelerated method for power flow computation aimed at improving computational efficiency and memory management. To further improve the efficiency of sparse matrix indexing on GPUs, we introduce two novel indexing techniques: spGEMM and kernel indexing. The spGEMM method achieves slicing of sparse matrices by utilizing matrix multiplication in place of indexing operations. The kernel indexing approach employs a kernel function in combination with the storage principles of the compressed sparse row (CSR) format to perform matrix indexing, thereby enabling the slicing operation. Simulation results conducted with varying numbers of buses demonstrate that our proposed method significantly outperforms traditional CPU-based approaches in terms of computation speed and scalability. The findings indicate that the proposed indexing techniques effectively improve the speed of power flow analysis and optimize its memory management.