关键词: 深度强化学习, 优化调度, 综合能源系统, 软演员-评论家算法, 可再生能源-制氢-制甲醇一体化站

Abstract: Producing hydrogen from renewable energy and synthesizing it into methanol is an effective approach for green hydrogen substitution in the chemical industry. The core facility of this pathway is the renewable energy hydrogen methanol integrated station (REHMIS). This study designs a novel renewable energy hydrogen methanol integrated station integrated energy system (REHMIS-IES) topology, which satisfies both the methanol demand of REHMIS and the multi-energy demand of its supporting buildings. This study proposes a strict-constraint soft actor-critic (SC-SAC) framework to develop an efficient operating strategy for the REHMIS-IES. The mathematical model is formulated as a Markov decision process, and a state constraint mechanism (SCM) is introduced to prevent large fluctuations in the energy storage system. The SC-SAC algorithm embeds the trained Q-network and action constraints into a mixed-integer linear programming (MILP) model, ensuring that scheduling decisions satisfy all operational constraints. Simulation results under multiple scenarios show that the proposed system can effectively reduce operating costs while meeting multi-energy demands. The SC-SAC algorithm reduces system energy imbalance by at least 16.2% and operating costs by no less than 11.7%, outperforming other deep reinforcement learning algorithms.

Key words: Deep reinforcement learning, Scheduling, Integrated energy system, Soft actor-critic, Renewable energy-hydrogen-methanol integrated station