Accelerating Multi-Contingency Simulation Through Physics-Data-Hybrid Mode Decomposition

Fast simulation of power grid dynamics under multi-contingency (e.g.,$N{-}k$events and cascading failures) is crucial for identifying vulnerabilities and supporting real-time decision-making. However, conducting multi-contingency simulations is notably challenging, as conventional numerical integrations are computationally intensive, while learning-based methods struggle to accommodate the extensive variability in topology, parameters, and human or automated operations. This paper develops a Physics-Data-Hybrid Mode Decomposition (HyMoD) approach for fast dynamical simulation under multiple contingencies. Contributions are fourfold: (1) An hybrid automaton-based simulation architecture is developed to depict the multi-contingency process, serving as the foundation for integrating HyMoD with multi-contingency simulation; (2) A HyMoD framework is devised to enable the iteration-free multi-contingency simulation, where physical operators respond to discrete events and the Mori–Zwanzig formalism linearly represents continuous-time dynamics via memory terms; (3) Data-free physical operators are derived, along with a physics-guided correction method that update the HyMoD operator to track the evolving dynamics; (4) On-the-fly updating and monitoring techniques are introduced to update HyMoD operators with high efficiency using incremental data and evaluate the simulation accuracy in real time. Numerical studies on real-scale systems with both synchronous generators and inverter-based resources validate the efficacy of HyMoD under diverse multi-contingency and cascading scenarios. The accuracy, efficiency, and reliability of HyMoD are demonstrated.