摘要
针对传统潮流计算难以满足日趋复杂的电网架构,且计算收敛效果不佳等问题,提出了一种基于深度强化学习的电网潮流分析方法。详细论述了电网传统潮流计算的方法及计算不收敛的影响因素,利用融合深度学习与深度双Q网络(DDQN)自动调节网络参数,使潮流计算趋于收敛。最终根据收敛后的潮流值分析电网运行状态,并采取转移负荷等措施完成电网调度,保证系统稳定运行。基于PSASP平台对所提方法进行实验论证,结果表明DDQN具有较好的收敛特性,且当系统配置理想时,仅有3%的潮流计算不收敛。同时,经过优化后的电网潮流,轻载变压器由12%上升至19%,部分变电站改由双回路供电,使电网更加稳定可靠。
In order to solve the problem that traditional power flow calculation is difficult to meet the increasingly complex power grid architecture and the convergence effect is poor,a power flow analysis method based on deep reinforcement learning is proposed.The traditional method of power flow calculation and the influence factors of non convergence are discussed in detail,the Deep Double Q Network(DDQN),which combines deep learning is used to automatically adjust the network parameters,so that the power flow calculation tends to converge.Finally,according to the converged power flow value,the operation state of the power grid is analyzed,and measures such as load transfer are taken to complete the power grid dispatching to ensure the stable operation of the system.Based on PSASP platform,the experimental results show that DDQN has good convergence characteristics,and only 3% of the power flow calculation does not converge when the system configuration is ideal.At the same time,after the optimization of power flow,the light load transformer increased from 12% to 19%,and part of the substation is changed to double circuit power supply,so the power grid is more stable and reliable.
作者
王申
李文臣
贾培浩
黄彦浩
章姝俊
WANG Shen;LI Wenchen;JIA Peihao;HUANG Yanhao;ZHANG Shujun(Beijing Jiuzhang Yunji Technology Co.,Ltd.,Beijing 100085,China;China Electric Power Research Institute Co.,Ltd.,Beijing 100192,China;State Grid Zhejiang Electric Power Co.,Ltd.,Hangzhou 310007,China)
出处
《电子设计工程》
2022年第16期46-50,共5页
Electronic Design Engineering
基金
国家电网有限公司科技项目(5100-202055003A-0-0-00)。
关键词
电网潮流分析
深度学习
深度双Q网络
计算收敛
电网调度
传统潮流计算
power flow analysis
deep learning
deep double Q network
calculation convergence
power grid dispatching
traditional power flow calculation
