基于DBSCAN与二叉决策树算法的短路故障定位方法研究  

Research on short-circuit fault localization method based on DBSCAN and binary decision tree algorithm

作　　者：陈雨娟 顾涛 CHEN Yujuan;GU Tao(School of Safety Engineering,North China Institute of Science and Technology,Yanjiao 065201,China;School of Computing,North China Institute of Science and Technology,Yanjiao 065201,China)

机构地区：[1]华北科技学院安全工程学院,北京东燕郊065201 [2]华北科技学院计算机学院,北京东燕郊065201

出　　处：《华北科技学院学报》2024年第3期42-49,共8页Journal of North China Institute of Science and Technology

基　　金：中央高校基本科研业务费资助项目(3142015024);河北省物联网监控工程技术研究中心基金项目(3142016020)。

摘　　要：随着人工智能技术在配电网中不断扩大,配电网故障判断的速度和准确率得到了显著提升。然而,当监测系统发生多起短路故障报警时,往往会伴随大量的衍生短路故障的报警,影响现场人员对真实短路位置的判断。为了提高配电网在线监测及诊断能力,同时过滤掉衍生报警,本研究提出了一种基于DBSCAN和二叉决策树算法的短路故障定位推理机。以时间为密度对线路报警信息进行聚类,并根据配电网线路的拓扑结构设计相应的推理树,以此开发出故障定位推理机,最终实现对配电网短路故障的快速诊断定位,准确率可达97%以上。With the continuous expansion of artificial intelligence technology in the distribution network,the speed and accuracy of fault diagnosis in the distribution network have been significantly improved.However,when multiple short circuit fault alarms occur in the monitoring system,it often accompanies a large number of derivative short-circuit fault alarms,affecting field personnel′s judgment on the real location of short circuits.To improve the online monitoring and diagnostic capabilities of the distribution network while filtering out de⁃rivative alarms,this study proposes a short-circuit fault location inference machine based on DBSCAN and bi⁃nary decision tree algorithm.Cluster the alarm information of power lines based on time density,and design corresponding inference trees according to the topology structure of power distribution network lines. Based on this, develop a fault localization inference machine, and ultimately achieve rapid diagnosis and localization of short circuit faults in power distribution networks, with an accuracy rate of over 97%.

关 键 词：机器学习 DBSCAN算法 二叉决策树 配电网 故障诊断 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]