Risk-aware GAN driven distributionally robust optimization for electric-thermal-hydrogen integrated energy system considering V2G and demand response

The widespread integration of renewable energy exposes Electric-Thermal-Hydrogen Integrated Energy Systems (ETH-IES) to multi-dimensional uncertainties involving source, load, and price. Additionally, the short-term severe fluctuations of renewable energy introduce extreme instability and operational risks to the system. Existing optimization methods often struggle to balance economic efficiency with robustness against extreme tail risk events. To address these challenges, this study proposes a two-stage Distributionally Robust Optimization (DRO) model driven by a Risk-Aware Generative Adversarial Network (RA-GAN). First, a physical coordination framework integrating Vehicle-to-Grid (V2G) and Demand Response (DR) is established. This mechanism leverages the massive battery resources of electric vehicles as a fast-response buffer to smooth short-term renewable fluctuations and reduce system operating costs. Second, an RA-GAN model incorporating a risk evaluator is developed to capture specific tail-risk features. These features are mapped to the geometric boundaries of the DRO ambiguity set, achieving precise data-driven boundary calibration. Furthermore, adjustable budget parameters are introduced to allow flexible regulation of conservatism based on risk preferences. Case studies indicate that the proposed RA-GAN significantly improves the identification of high-risk scenarios compared to traditional generative models. Economic analysis demonstrates that the coordination of V2G and DR reduces total system costs by 15.93% compared to the baseline. The proposed framework effectively balances system robustness and economy, outperforming traditional stochastic and robust optimization methods.