基于强化学习理论的输电网扩展规划方法  

Transmission Expansion Planning Based on Reinforcement Learning

作　　者：王渝红[1,2] 胡胜杰 宋雨妍 江栗 沈力 WANG Yuhong;HU Shengjie;SONG Yuyan;JIANG Li;SHEN Li(College of Electrical Engineering,Sichuan University,Chengdu 610065,Sichuan Province,China;Key Laboratory of Intelligent Electric Power Grid of Sichuan Province(Sichuan University),Chengdu 610065,Sichuan Province,China;State Grid Southwest Branch Corporation,Chengdu 610041,Sichuan Province,China)

机构地区：[1]四川大学电气工程学院,四川省成都市610065 [2]智能电网四川省重点实验室(四川大学),四川省成都市610065 [3]国家电网公司西南分部,四川省成都市610041

出　　处：《电网技术》2021年第7期2829-2838,共10页Power System Technology

基　　金：国家电网西南分部科技项目(SGSW0000GHJS 1900117)。

摘　　要：该文将人工智能扩展至传统输电网规划中,提出基于强化学习理论的输电网扩展规划方法,以带自适应学习因子的多步回溯α-Q(λ)算法进行求解。基于数据库与蒙特卡洛法,并计及输电可靠性成本建立了扩展规划模型,设计自适应学习因子的多步回溯Q(λ)算法,利用强化学习智能体以最大累积奖励为目标,结合输电网扩展规划特性,将混合整数规划模型转换为算法的智能体与环境,用以模拟规划人员对电网的规划过程。在Garver-6与IEEE 24-RTS系统中验证该文所提方法的有效性,并与其他智能算法进行比较。By applying the artificial intelligence to the traditional transmission expansion planning,a transmission expansion planning method using a-Q(λ)algorithm with adaptive learning factor is proposed based on reinforcement learning.With the help of the prepared database and the Monte Carlo method,the transmission expansion planning model is constructed by considering the reliability cost in the optimal object function.Combining the characteristics of the transmission network,the multi-step backtrackingα-Q(λ)algorithm with adaptive learning factor is designed.Then the mixed integer planning model is transformed into the agent and environment of in theα-Q(λ)algorithm to simulate the planning process of a power grid.The validity of the proposed method is verified by Garver-6 and IEEE 24-RTS system,and the comparison with other intelligent algorithms is shown.

关 键 词：输电网扩展规划 强化学习 多步回溯Q(λ)算法 自适应学习因子 

分 类 号：TM721[电气工程—电力系统及自动化]