In-Disaster Emergency Dispatch with Unit Commitment and Transmission Switching via Approximate Dynamic Programming

Efficient emergency dispatch is critical for minimizing load shedding and maintaining power system resilience during extreme disasters. However, this task faces significant challenges arising from the high stochasticity of component failures, the non-anticipativity of dispatch decisions, and strict real-time computation requirements. To address these issues, this paper proposes an in-disaster emergency dispatch framework based on Approximate Dynamic Programming (ADP). First, this paper establish a discrete-time Markov Decision Process (MDP) to capture the inter-temporal coupling of unit commitment and transmission switching. Second, a Value Function Approximation (VFA) strategy is designed, which employs piecewise linear functions and linear functions to approximate the future values of unit generation and transmission switching status, respectively. This approach effectively incorporates future uncertainties into the current-stage decision model. Finally, a framework combining "pre-disaster offline training and in-disaster online decision-making"is established to ensure efficient decision-making. Case studies indicate that the introduced framework drastically accelerates the online computational process while maintaining high solution quality, thereby effectively enhancing power system resilience.