Machine learning-driven stochastic bidding for hydrogen Refueling Station-integrated virtual power plants in energy market

Le résumé fourni par la source

Virtual power plants (VPPs) are gaining significance in the energy sector due to their capacity to aggregate distributed energy resources (DERs) and optimize energy trading. However, their effectiveness largely depends on accurately modeling the uncertain parameters influencing optimal bidding strategies. This paper proposes a deep learning-based forecasting method to predict these uncertain parameters, including solar irradiation, temperature, wind speed, market prices, and load demand. A stochastic programming approach is introduced to mitigate forecasting errors and enhance accuracy. Additionally, this research assesses the flexibility of VPPs by mapping the flexible regions to determine their operational capabilities in response to market dynamics. The study also incorporates power-to‑hydrogen (P2H) and hydrogen-to-power (H2P) conversion processes to facilitate the integration of hydrogen fuel cell vehicles (HFCVs) into VPPs, enhancing both technical and economic aspects. A network-aware VPP, connected to generation resources, storage facilities, demand response programming (DRP), vehicle-to-grid technology (V2G), P2H, and H2P, is used to evaluate the proposed method. The problem is formulated as a convex model and solved using the GUROBI optimizer. Results indicate that a hydrogen refueling station can increase profits by approximately 49 % compared to the base case of directly selling surplus generation from renewable energy sources (RESs) to the market, and profits can further increase to roughly 86 % when other DERs are incorporated alongside the hydrogen refueling station.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.