Research on Energy Management Decision‐Making Methods for Hydrogen–Electricity Coupling in Multi‐Microgrids Under the Virtual Power Plant Model

The coordinated operation of multiple microgrids under the virtual power plant model is crucial for integrating renewables and hydrogen. However, the system faces challenges from multi‐timescale coupling: minute‐level electrical fluctuations contrast with hour‐level hydrogen dynamics. This hinders traditional methods from balancing real‐time needs with long‐term optimisation. To address this, this paper proposes an energy management method fusing long short‐term memory (LSTM) with deep deterministic policy gradient (DDPG). By leveraging LSTM to track historical trajectories, the method explicitly models the fast–slow timescale coupling between minute‐level electrical fluctuations and hour‐level hydrogen storage dynamics, effectively resolving partial observability. Furthermore, a price‐guided cooperative mechanism is incorporated into the reward design to achieve cross‐microgrid energy complementarity. Case studies demonstrate the method's superiority in achieving game equilibrium. Compared to traditional DDPG, the proposed strategy increases system net revenue by 8.18%, an 11.23% improvement over pre‐optimisation levels. These results, further corroborated by sensitivity analyses under varying price spreads, validate the algorithm's value in enhancing economics, smoothing fluctuations, and reducing policy oscillations.