Data-Driven Stochastic-Robust Planning for Resilient Hydrogen-Electricity System With Progressive Hedging Decoupling

Given the severe damage that typhoons can inflict on power systems, it is of utmost urgency to develop proactive pre-disaster planning to enhance power system resilience. To this end, considering the high efficiency and long-term storage capabilities of hydrogen energy, a resilience-oriented planning model of the hydrogen-electricity integrated energy system (H-EIES) is constructed in this work. In this resilience-oriented planning model, hardening power lines and constructing hydrogen-to-power stations (H2Ps) are jointly planned to reduce load shedding during typhoons. A tri-layer stochastic-robust optimization (SRO) approach is proposed to find the optimal construction strategy under uncertainties, where the first layer determines the planning strategy, the second layer identifies the worst outage contingency, and the third layer ensures the optimal operation. However, the uncertainty of typhoons presents significant challenges to the feasibility of the proposed SRO approach. Therefore, this work leverages the Wasserstein generative adversarial network with gradient penalty (WGAN-GP) and spectral clustering method to quantify the typhoon uncertainty. Additionally, a novel nested column-and-constraint generation with progressive hedging algorithm (C&CG-PHA) is elaborately designed to solve the complex tri-layer and multi-scenario coupled optimization problem. Validating by the real-life typhoon data, numerical experiments are carried out and indicate the economic efficiency and robustness of the proposed SRO approach, where the system's load shedding losses are reduced by 95.7%. Also, it has been proved that the C&CG-PHA can notably accelerate the solution process and exhibit excellent scalability.