互联电网CPS调节指令动态最优分配Q–学习算法.pdfVIP

第30 卷 第7 期 中 国 电 机 工 程 学 报 Vol.30 No.7 Mar.5, 2010 62 2010 年3 月5 日 Proceedings of the CSEE ©2010 Chin.Soc.for Elec.Eng. 0258-8013 (2010) 07-0062-08 TM 71 A 470 40 文章编号： 中图分类号： 文献标志码： 学科分类号： ⋅ 互联电网CPS 调节指令动态最优分配Q– 学习算法 余涛，王宇名，刘前进 (华南理工大学电力学院，广东省 广州市 510640) Q-learning-based Dynamic Optimal Allocation Algorithm for CPS Order of Interconnected Power Grids YU Tao, WANG Yu-ming, LIU Qian-jin (College of Electric Power, South China of Technology, Guangzhou 510640, Guangdong Province, China) ABSTRACT: The dynamic optimization of automatic 度，提高AGC 系统调节能力。遗传算法和工程实用方法在 generation control (AGC) order allocation based on the NERC’s 标准两区域模型及南方电网模型的仿真研究显示，Q–学习 control performance standard (CPS) is a problem on stochastic 有效提高了系统的适应性、鲁棒性和CPS 考核合格率。 optimization in the AGC system for the interconnected power 关键词：Q–学习；随机最优；离散时间马尔可夫决策过程； system. The CPS order allocation was discretized and viewed as 控制性能标准；自动发电控制 a discrete time Markov decision process (DTMDP). The dynamic control method based on Q-learning was proposed. 0 引言 Reward functions in Q-learning were designed based on different 自动发电控制(automatic generation control ， optimization objectives. Thermal and hydro units were integrated, AGC)应用于现代大电网频率控制和互联电网之间 with the regulating margin for hydro units being considered, toimprove the regul