分布式强化学习在经济调度问题中的应用  

Application of Distributed Reinforcement Learning in Economic Dispatch Problem

作　　者：陈晓玉 周佳玲 CHEN Xiao-yu;ZHOU Jia-ling(College of Science,Liaoning Technical University,Fuxin 123000,China;School of Automation,Nanjing University of Science and Technology,Nanjing 210094,China)

机构地区：[1]辽宁工程技术大学理学院,辽宁阜新123000 [2]南京理工大学自动化学院,江苏南京210094

出　　处：《控制工程》2022年第3期480-485,共6页Control Engineering of China

基　　金：国家自然科学基金青年科学基金资助项目(62003167);江苏省自然科学基金青年基金资助项目(BK20180459)。

摘　　要：研究了智能电网中电力成本函数未知的多区域动态经济调度问题。该问题的目标是配置每个区域在每个时刻的最优发电和购电量,以最小化多个区域的电力成本之和。为了解决电力成本函数未知的多区域动态经济调度问题,提出了基于Q学习的分布式强化学习算法。在分布式强化学习算法中,区域之间基于信息交互,协同寻找满足供需平衡的电力分配,同时每个区域建立局部Q函数寻找最优电力组合。数值仿真验证了算法的有效性。In this paper,the multi-region dynamic economic dispatch problem with unknown cost functions in smart grid is studied.The objective of multi-region dynamic economic dispatch problem is to find the optimal generation and purchase electricity of each region at each time to minimize the sum of electric power cost.In order to solve the multi-region dynamic economic dispatch problem with unknown cost functions,a distributed reinforcement learning algorithm based on Q-learning is proposed.In the distributed reinforcement learning algorithm,the regions cooperate to find the power distribution that meets the balance of supply and demand based on information interaction,and each region establishes a local Q function to find the optimal electricity combination.The effectiveness of the algorithm is verified by numerical simulation.

关 键 词：分布式强化学习 智能电网 动态经济调度问题 一致性协议 

分 类 号：O29[理学—应用数学]