Adaptive Development of Parallel Power System Dynamic Simulation Application in Python
Cong Wang, Liwei Wang, Shuangshuang Jin · Chronicle of computing · 2022
Due to its intensive computational demands for real-time operation and diagnosis, large-scale power system dynamic simulation requires high-performance computing technologies to accelerate its computation on advanced computing platforms. In this paper, leveraging high-level Python and its parallel scientific computing libraries, three parallel power system dynamic simulation applications are adaptively developed using native MPI for Python on CPU, PETSc for Python on CPU, and CuPy on GPU with dedicated data manipulation strategies and implementations, respectively. Their computational performance is compared using different sizes of testing systems and indicates that: 1) MPI and PETSc can make a decent performance for small and moderate-size systems on limited CPU resources, and 2) GPU has better potential in speeding up dynamic simulation for larger and more complex systems. The results demonstrate Python's suitability in parallelizing power system modeling and simulation with fast computational performance and easy development.