基于深度强化学习的氢能综合能源系统优化调度方法  

Optimal Scheduling Method for Integrated Energy Systems with Hydrogen Based on Deep Reinforcement Learning

作　　者：张磊[1] 吴红斌[1] 何叶 徐斌 张明星 丁明[1] ZHANG Lei;WU Hongbin;HE Ye;XU Bin;ZHANG Mingxing;DING Ming(Anhui Province Key Laboratory of Renewable Energy Utilization and Energy Saving(Hefei University of Technology),Hefei 230009,China;Electric Power Research Institute of State Grid Anhui Electric Power Co.,Ltd.,Hefei 230601,China;Lu’an Power Supply Company of State Grid Anhui Electric Power Co.,Ltd.,Lu’an 237000,China)

机构地区：[1]新能源利用与节能安徽省重点实验室(合肥工业大学),安徽省合肥市230009 [2]国网安徽省电力有限公司电力科学研究院,安徽省合肥市230601 [3]国网安徽省电力有限公司六安供电公司,安徽省六安市237000

出　　处：《电力系统自动化》2024年第16期132-141,共10页Automation of Electric Power Systems

基　　金：国家自然科学基金区域创新发展联合基金资助项目(U19A20106);安徽省自然科学基金资助项目(2108085UD05);安徽省科技重大专项计划资助项目(202203f07020003)。

摘　　要：为实现碳减排目标,氢能与综合能源系统的结合成为最具潜力的发展方向之一。针对当前氢能综合能源系统调度策略灵活性不足、复杂系统多目标优化求解困难等问题,提出一种基于深度强化学习的氢能综合能源系统优化调度方法。首先,采用耦合设备的变工况模型,构建风-光-氢-冷-热-电综合能源系统,拓展设备联合供能空间。其次,考虑系统运行成本、碳排放量、系统自供给平衡度和新能源利用率,基于最优解距离构建多目标优化模型,激发智能体探索性。然后,通过时序片段表征优化深度强化学习算法,增强了智能体对系统状态变化的估计精度。最后,在源荷实测数据的基础上设计仿真算例。结果表明,所提方法可以有效提高氢能综合能源系统调度的灵活性,充分挖掘氢能的碳减排潜力,实现调度经济性和环保性的双重优化。In order to achieve carbon reduction targets,the combination of hydrogen energy and integrated energy systems has become one of the most potential development directions.Aiming at the problems such as the insufficient flexibility of scheduling strategy of hydrogen integrated energy system and difficulty in solving multi-objective optimization of complex systems,an optimal scheduling method for hydrogen integrated energy systems based on deep reinforcement learning is proposed.First,the variable operation condition model of coupled equipment is used to construct a wind-solar-hydrogen-cooling-heat-electricity integrated energy system,and expand the joint energy supply space of equipment.Secondly,considering the system operation cost,carbon emissions,system self-supply balance and renewable energy utilization rate,a multi-objective optimization model is built based on the optimal solution distance to stimulate the exploration of the agent.Then,the deep reinforcement learning algorithm is optimized by time segment characterization to enhance the estimation accuracy of the system state change.Finally,a simulation case is designed based on the measured data of the source and load.The results show that the proposed method can effectively improve the scheduling flexibility of the hydrogen integrated energy system,fully tap the carbon emission reduction potential of hydrogen energy,and realize the dual optimization of scheduling economy and environmental protection.

关 键 词：综合能源系统 氢能 优化调度 深度强化学习 多目标优化 可再生能源 

分 类 号：TK01[动力工程及工程热物理] TM73[电气工程—电力系统及自动化]