Real-time Crew Scheduling for Post-Disaster Restoration of Integrated Power and Gas Networks

Urban energy systems face increasing challenges in maintaining reliable operation during extreme events. Most existing restoration models assume complete information, which is unrealistic under real-world uncertainties like unpredictable repair times. Effective restoration strategies should enable quick decisions and can dynamically adapt to uncertain environments and leverage real-time information efficiently. This paper proposes an online restoration strategy for integrated electricity and gas distribution systems (IEGDS), focusing on repair crew dispatch under uncertainty. Using a Markov decision process (MDP) framework, the method dynamically adjusts restoration actions based on real-time system states. A rollout algorithm is proposed to estimate the Q-function of candidate actions through Monte Carlo simulations, enabling online decision-making without relying on model training or complex mathematical programming. Tested on a 13-bus power system and a 7-node natural gas system, the proposed algorithm achieves performance comparable to offline optimization methods while demonstrating high efficiency.