Inverse Reinforcement Learning-Based Asynchronous Filtering for SMIB Power Systems With Stochastic Mode Switching

This paper investigates the asynchronous filtering problem for single-machine infinite bus (SMIB) power systems subject to stochastic transmission line faults. The system is modeled as a discrete-time Markov jump system (MJS) to capture the random switching behavior induced by transmission line faults. To address the asynchrony between the system modes and the filter operation, a hidden Markov model (HMM) is adopted. The filtering problem is reformulated as a regulation problem by introducing a quadratic performance index based on output estimation errors, offering a filtering-based alternative to control strategies. To solve the associated coupled algebraic Riccati equations (CAREs), an inverse reinforcement learning (IRL)–based algorithm is developed, which enables model-free filtering without requiring prior knowledge of the system dynamics or transition probabilities. The convergence of the proposed algorithm is rigorously analyzed, and a numerical example based on an SMIB power system with stochastic faults is provided to validate its effectiveness.