CC BY-NC-ND 4.0Song, WonjuneWonjuneSongPangalos, GeorgGeorgPangalosWiese, NilsNilsWieseSinger, RolandRolandSingerMoon, SeungpilSeungpilMoonPark, ByungkwonByungkwonPark2025-10-082025-10-082025https://publica.fraunhofer.de/handle/publica/497098https://doi.org/10.24406/publica-563510.1016/j.ifacol.2025.08.13410.24406/publica-56352-s2.0-105016676033The increasing complexity of modern power systems, driven by the integration of renewable energy sources and advanced control systems, has highlighted the demand for faster and more accurate power system dynamic simulation frameworks. To this end, this study explores the emerging technique of neural ordinary differential equations (neural ODEs) for two key areas of power system dynamic analysis. In one hand, neural ODEs are leveraged as the coarse operator in the parallel-in-time (Parareal) algorithm, further enhancing the computational performance of power system dynamic simulation. On the other hand, the study examines the integration of neural ODEs with a system identification method to improve the modeling accuracy of complex dynamic systems. Numerical case studies are conducted to demonstrate the effectiveness of the proposed approaches on two distinct systems: the IEEE 14-bus and 145-bus system are used to evaluate the computational performance of the neural ODEs-enhanced Parareal algorithm, while a real-world high-voltage direct current (HVDC) system from Jeju Island, South Korea, serves as the testbed for the dynamic system identification. The results illustrate a great potential of neural ODEs to advance power system dynamic analysis, demonstrating improvements in both computational speed and modeling accuracy.enfalseNeural ODEsparallel-in-time algorithmpower system dynamic simulationsubspace algorithmssystem identificationjournal article