摘要
“双碳”目标下大规模新能源并网以及电力电子设备的大量投入运行使得系统的惯性降低,对系统的稳定运行造成了影响。传统的暂态稳定分析具有建模困难、计算效率不高、容易受到不确定因素的干扰等不足。近几年,强化学习发展迅速,深度强化学习结合了深度学习与强化学习的优势,可应用于大规模场景及信息有限的场合,学习海量高维、不确定数据求解决策问题。该文首先介绍了深度强化学习的概况,接着对深度强化学习在电力系统暂态稳定控制决策方面的已有研究成果进行总结和概括,从系统预防控制、系统紧急控制、系统恢复控制3个方面分析了深度强化学习算法在电力系统暂态稳定控制决策中的研究现状和优势,并就这一研究方向中存在的问题作了深入探讨,最后对深度强化学习未来技术发展和实际应用进行了展望。
Under the carbon peak and neutrality targets,the large-scale grid connection of renewable energy and the op-eration of high proportion of power electronic equipment have reduced the inertia of the system and impacted the stable operation of the power system.Traditional transient stability analysis has some shortcomings,such as difficulty in model-ing,low calculation efficiency,and being easy to be disturbed by uncertain factors.In recent years,reinforcement learning has developed rapidly.Deep reinforcement learning combines the advantages of deep learning and reinforcement learning,and it can learn a large number of high-dimensional and uncertain data to solve decision-making problems in large-scale scenes with limited information.This paper first summarizes deep reinforcement learning.Next,the existing research re-sults of reinforcement learning in power system transient stability control decision-making are summarized.Then,this paper analyzes the research status and advantages of deep reinforcement learning algorithm in power system transient sta-bility control decision-making from three aspects of system preventive control,system emergency control,and system recovery control,and the existing problems in this research direction are discussed in depth.Finally,the prospects in fu-ture technical developments and practical applications of deep reinforcement learning are put forward.
作者
江昌旭
刘晨曦
林铮
林俊杰
JIANG Changxu;LIU Chenxi;LIN Zheng;LIN Junjie(College of Electrical Engineering and Automation,Fuzhou University,Fuzhou 350108,China;Fujian Smart Electrical Engineering Technology Research Center,Fuzhou 350108,China)
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2023年第12期5171-5186,共16页
High Voltage Engineering
基金
福建省自然科学基金(2022J05125,2021J05134)。
关键词
强化学习
深度强化学习
暂态稳定
稳定控制决策
预防控制
紧急控制
恢复控制
reinforcement learning
deep reinforcement learning
transient stability
stable control decisions
preventive control
emergency control
recovery control
