Robust Adaptive Optimal Fault-Tolerant Control Scheme for Multi-Machine Power Systems via the Stabilizing Two-Stage Policy Iteration Framework

For a class of multi-machine power systems (MMPSs) with unknown dynamics and actuator faults, two model-free optimal fault-tolerant (FT) control frameworks based on policy iteration (PI) are proposed. First, the actuator failure is modeled within the interconnected MMPSs. Building on this foundation, a stable FT-PI framework is developed that eliminates the conventional dependence on initial admissible controllers through a proposed two-stage implementation. Additionally, an enhanced Q-learning equivalent framework is introduced, which further relaxes the conditions on the initial stabilizing policy typically required by traditional integral Q-learning methods. Both of these data-driven algorithms dynamically solve the FT continuous algebraic Riccati equation (FT-CARE) without requiring knowledge of the system dynamics and actuator fault information. Meanwhile, the convergence and stability of the algorithms are rigorously proved. Finally, the effectiveness of the proposed algorithms is verified through a three-machine power system.