基于强化学习的串联超级电容器组非能耗均衡方法  

Non⁃energy⁃consumption Equalization Method for Series Super Capacitor Banks Based on Reinforcement Learning

作　　者：宋倩 蓝俊欢 SONG Qian;LAN Junhuan(College of Big Data and Computer Science,Hechi University,Guangxi Hechi 546300,China;Hechi Power Supply Bureau,Guangxi Power Grid Co.,Ltd.,Guangxi Hechi 546300,China)

机构地区：[1]河池学院大数据与计算机学院,广西河池546300 [2]广西电网有限责任公司河池供电局,广西河池546300

出　　处：《电力电容器与无功补偿》2024年第4期91-96,共6页Power Capacitor & Reactive Power Compensation

基　　金：2022年广西高校中青年教师科研基础能力提升项目(2022KY0606);2023年河池学院校级科研平台(2023XJPT001)。

摘　　要：超级电容器作为新兴储能元件的代表,在创新驱动发展战略实施背景下使用日益广泛,为解决串联超级电容器组使用中各单体电压偏差大、利用效率不高的问题,充分利用人工智能技术,提出了一种基于强化学习的串联超级电容器组非能耗均衡方法,将近端策略优化算法(PPO)应用于串联超级电容器组电压均衡中,在Matlab/Simulink仿真平台中搭建了串联超级电容器组和强化学习环境并验证了该算法的有效性,相较于传统的能耗型电压均衡方法,均衡效率较好,超级电容器的损耗较小,实验结果表明了PPO算法可实现串联超级电容器组电压的均衡。As a representative of emerging energy storage components,super capacitor are increasingly used in the context of the implementation of the innovation-driven development strategy.For solving such problems as large voltage deviation and low utilization efficiency of each single unit in the use of series-con-nected super capacitor banks and fully using artificial intelligent technology,a kind of reinforcement learn-ing-based non-energy equalization method for series super capacitor banks is proposed,and the proximal policy optimization operator(PPO)is applied to the voltage equalization of series super capacitor banks.The super capacitor bank and reinforcement learning environment are set up in Matlab/Simulink simulation platform and the effectiveness of the algorithm is verified.The super capacitor,compared to the traditional energy consumption voltage equalization method,has good equalization efficiency and small loss.The experimental results show that the PPO algorithm can equalize the voltage of series super capacitor bank.

关 键 词：强化学习 PPO算法 人工智能 超级电容器 储能 

分 类 号：TM53[电气工程—电器] TP181[自动化与计算机技术—控制理论与控制工程]