基于随机森林算法的低压系统漏电检测技术研究  

Research on leakage detection technology of low-voltage power system based on random forest algorithm

作　　者：肖湘奇 肖宇 黄瑞 黄燕娇 贺星 刘谋海 慕静茹 XIAO Xiangqi;XIAO Yu;HUANG Rui;HUANG Yanjiao;HE Xing;LIU Mouhai;MU Jingru(State Grid Hunan Electric Power Co.,Ltd.,Changsha 410004,China;Hunan Provincial Key Laboratory of Intelligent Electrical Measurement and Application Technology,Changsha 410004,China;School of Electrical&Information Engineering,Changsha University of Science&Technology,Changsha 410114,China)

机构地区：[1]国网湖南省电力有限公司,湖南长沙410004 [2]智能电气量测与应用技术湖南省重点实验室,湖南长沙410004 [3]长沙理工大学电气与信息工程学院,湖南长沙410114

出　　处：《电力科学与技术学报》2024年第3期38-47,115,共11页Journal of Electric Power Science And Technology

基　　金：国家电网有限公司总部科技项目(5700-202155204A-0-0-00)。

摘　　要：随着低压配电系统规模及用户需求量迅速增加,用户线路与家庭电气设备漏电故障频发,极易发生人体触电及电气火灾事故。剩余电流保护器作为检测低压漏电故障的常用手段,近年来因线路(或设备)存在对地泄漏电流而频频误动,极大降低了保护设备投运率与可靠性。为此,本文提出一种基于随机森林(random forest,RF)算法的低压配电系统漏电检测技术,为最大程度贴近真实漏电故障场景,充分考虑实际故障场景存在的正常泄漏电流过大、故障邻近支路负荷投切频繁等干扰因素以获得贴近真实故障场景的原始剩余电流数据;通过对原始剩余电流数据进行数据预处理,分析掌握剩余电流的频域与时域特性并利用傅里叶变换算法提取时频域特征,完成低压系统漏电检测模型的建立与训练;在施加多种干扰因素情况下对漏电检测模型进行深度测试,其漏电故障的检测准确率可达99.97%,实现了多种干扰因素情况下的低压配电系统漏电故障检测;最后将支持向量机(support vector machine,SVM)算法、K最近邻(K-nearest neighbor,KNN)算法与本文基于RF算法的漏电故障检测模型的准确性进行对比,以此验证了基于RF算法的低压系统漏电检测模型的准确性与可行性。With the rapid increase in the scale of low-voltage distribution system and user demand,leakage faults of user lines and household electrical equipment occur frequently,which increase the risk of electric shock and electrical fire accidents.Residual current protector is a common method to detect low-voltage leakage fault.In recent years,due to the existence of leakage current to ground of lines(or equipment),it also frequently false operates,which greatly reduces the operation rate and reliability of protective equipment.To overcome these issues,this paper proposes a leakage detection technology for low-voltage distribution system based on random forest(RF)algorithm.In order to closely simulate the real leakage fault scenario,the original residual current data close to the real fault scenario can be obtained by fully considering the interference factors such as excessive normal leakage current and frequent switching of load in the adjacent branch of the fault scenario.Through data preprocessing of the original residual current data,the frequency domain and time domain characteristics of the residual current are analyzed,and the time-frequency characteristics are extracted using the Fourier transform algorithm to complete the establishment and training of the low-voltage system leakage detection model.The leakage detection model is tested under the condition of multiple interference factors,and the results show that the detection accuracy of the leakage fault can reach 99.98%,realizing the leakage fault detection of low-voltage distribution system under the condition of multiple interference factors.Finally,support vector machine(SVM)algorithm,K-nearest neighbor(KNN)algorithm and the leakage fault detection accuracy based on random forest algorithm are compared to verify the accuracy and feasibility of the proposed leakage fault detection model of low-voltage system.

关 键 词：低压系统 漏电检测 随机森林算法 剩余电流 

分 类 号：TM588[电气工程—电器]