基于神经网络数据分析方法的电网高损耗元件识别技术  

Identification technology of high loss components in power grid based on neural network data analysis method

作　　者：吴悦 梁琛 张光儒 马振祺 张家午 WU Yue;LIANG Chen;ZHANG Guangru;MA Zhenqi;ZHANG Jiawu(Electric Power Research Institute,State Grid Gansu Electric Power Company,Lanzhou 730070,China)

机构地区：[1]国网甘肃省电力公司电力科学研究院,甘肃兰州730070

出　　处：《电子设计工程》2023年第10期83-87,共5页Electronic Design Engineering

基　　金：国网公司电网降损节能技术实验室研究项目(52272220002z)。

摘　　要：针对大部分电网损耗研究不全面且损耗类型识别不准确等问题,提出了一种基于神经网络数据分析方法的电网高损耗元件识别技术。该技术在全面分析线路及变压器损耗的基础上,将台区损耗类型划分成六类,并将台区电表数据输入全连接神经网络模型完成学习分析,从而实现高损耗元件的识别。基于某台区的109块电表数据样本进行实验分析,结果表明所提技术对于无异常损耗电表的识别准确率高达98.5%,且整体识别准确率可达93.4%,同时识别时间仅为0.35 s,优于其他对比技术。In view of the problems that most of the research on power grid loss is not comprehensive and the identification of loss types is not accurate,a high loss component identification technology based on neural network data analysis method is proposed in this paper.Based on the comprehensive analysis of line loss and transformer loss,the station loss types are divided into six categories,and the station meter data are input into the fully connected neural network model for learning and analysis,so as to complete the identification of high loss components.Based on the experimental analysis of 109 electricity meter data samples in a certain station area,the results show that the recognition accuracy of the proposed technology for electricity meters without abnormal loss is as high as 98.5%,the overall recognition accuracy is 93.4%,and the recognition time is 0.35 s,which is better than other comparison technologies.

关 键 词：电网高损耗元件 异常识别 全连接神经网络 线路损耗 变压器损耗 台区数据分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN911[自动化与计算机技术—计算机科学与技术]