Two-stage multi-timescale optimal scheduling for electricity-hydrogen coupling systems based on scenario approach and deep reinforcement learning

In electricity-hydrogen coupling systems， fluctuations in wind and solar power output， as well as the different timescales for electricity and hydrogen energy dispatch， pose significant challenges for economic and efficient system scheduling. To address these challenges， this paper， using scenario approach and deep reinforcement learning （DRL）， proposes a two-stage multi-timescale optimal scheduling method for electricity-hydrogen coupling systems considering uncertainties of wind and solar power generation. First， the operational characteristics of energy storage devices， including electrical and hydrogen storage devices， are analyzed， and a two-stage optimal scheduling framework for electricity-hydrogen coupling systems is designed. Next， with the uncertainties of wind and solar power generation considered， long-term and short-term timescale optimal models are developed. The long-term timescale model aims to maximize the systems’ energy self-balance by generating typical wind and solar output scenarios using Latin hypercube sampling （LHS） for scenario generation and reduction， followed by optimization. The short-term model focuses on minimizing the systems’ operational costs and is solved using the deep deterministic policy gradient （DDPG） algorithm. Finally， case study simulations demonstrate that the proposed method effectively facilitates hydrogen energy shifting， smooths fluctuations in wind and solar output， verifying the method’s effectiveness.