基于深度强化学习的微电网复合储能协调控制方法  

A Coordinated Control Method for Hybrid Energy Storage System in Microgrid Based on Deep Reinforcement Learning

作　　者：张自东 邱才明 张东霞 徐舒玮 贺兴[1] ZHANG Zidong;QIU Caiming;ZHANG Dongxia;XU Shuwei;HE Xing(Research Center for Big Data Engineering and Technologies (Shanghai Jiao Tong University),Minhang District,Shanghai200240,China;China Electric Power Research Institute,Haidian District,Beijing 100192,China)

机构地区：[1]大数据工程技术研究中心(上海交通大学),上海市闵行区200240 [2]中国电力科学研究院有限公司,北京市海淀区100192

出　　处：《电网技术》2019年第6期1914-1921,共8页Power System Technology

基　　金：国家重点研发计划项目(2018YFF0214705)~~

摘　　要：在微电网中,由于负荷和电源功率波动较大、各种不确定因素复杂,通常需要增加储能系统以保证供需实时平衡、并提高可再生能源的利用率。该文构建了一个孤岛运行的包含光伏发电、储氢装置(长期储能)、蓄电池(短期储能)的复合能源系统,并将复合储能系统的协调控制转化为序列决策问题,并采用深度强化学习方法加以解决。在发电量、负荷等多种因素变化的情况下,复合能量协调控制是一个复杂的优化决策问题,不同方案可能会影响系统供电稳定、利用效率和经济效益。为此该文设计了适合解决该类问题的深度强化学习框架和神经网络模型,经过足够的训练后能够实现实时在线决策控制,避免了传统算法在这方面的不足,同时,对于不同时刻、天气、季节的场景均能有效处理。结果表明了深度强化学习在复合储能协调控制问题中的可行性和有效性,具有较强的学术意义和工程价值,并可以用于处理相似的问题。In microgrid, due to fluctuations in load and power supply and other uncertain factors, it is necessary to allocate energy storage systems to ensure real-time supply and demand balance and improve the utilization rate of renewable energy. In this paper, a hybrid energy system consisting of photovoltaic(PV) system, hydrogen(long-term energy storage devices) and battery(short-term energy storage devices) is constructed. Coordinated control of the hybrid energy storage system can be formulated as sequential decision-making problem to be solved effectively with deep reinforcement learning(DRL) method. Affected by multiple factors such as real-time power generation and load change, hybrid energy coordinated control is a complex optimization decision problem, where different solutions may affect system power supply stability, utilization efficiency and economic benefits. A deep reinforcement learning framework and a neural network model are designed to solve such problems. Real-time online decision and control can be realized with sufficient training, which compensates the shortcomings of traditional methods in this respect. At the same time, it can be effectively processed for different moments, weather and seasonal scenarios. The results show feasibility and effectiveness of deep reinforcement learning in the hybrid energy coordination control, making this paper both of academic significance and engineering value, and this method can be adopted to solve similar problems.

关 键 词：微电网 深度强化学习 能量协调控制 光伏发电 孤岛系统 

分 类 号：TM721[电气工程—电力系统及自动化]