基于K-Means聚类算法的直流电网换流器故障自动化检测系统  

Automatic Detection System for Converter Faults in DC Power Grid Based on K-Means Clustering Algorithm

作　　者：翁子韵 WENG Ziyun(State Grid Heilongjiang Electric Power Co.,Ltd.Shuangyashan Power Supply Company,Shuangyashan 155100,China)

机构地区：[1]国网黑龙江省电力有限公司双鸭山供电公司,双鸭山155100

出　　处：《自动化与仪表》2025年第4期86-90,共5页Automation & Instrumentation

摘　　要：直流电网换流器结构复杂、监测信号较多,为了自动从大量监测信号中筛选关键特征,准确识别电网换流器故障,设计基于K-Means聚类算法的直流电网换流器故障自动化检测系统。采集的各线路电压信号,采用改进主成分分析法将高维的监测信号数据降维成少数几个主成分,作为反映线路电压信号的主要特征;通过改进K-Means聚类算法对所提取信号主成分特征进行分组归类,实现电网换流器故障信号分类检测。经测试,此系统对直流电网换流器单极故障、双极故障样本进行聚类识别后,识别结果的误差平方和最大值仅为0.02,可实现高精度的直流电网换流器故障自动化检测。The DC power grid converter has a complex structure and many monitoring signals,in order to automatically screen key features from a large number of monitoring signals and accurately identify faults in power grid converters,a DC power grid converter fault automatic detection system based on K-Means clustering algorithm is designed.The collected voltage signals of each line are reduced into a few principal components using an improved principal component analysis method to reflect the main characteristics of the line voltage signal.By improving the K-Means clustering algorithm to group and classify the extracted signal principal component features,the fault signal classification detection of power grid inverters is achieved.After testing,it was found that this system can cluster and identify samples of single pole and bipolar faults in DC power grid converters.The maximum sum of squared errors in the identification results is only 0.02,which can achieve high-precision automatic detection of DC power grid converter faults.

关 键 词：主成分分析法 K-MEANS聚类算法 直流电网 换流器 故障检测 自动化 

分 类 号：TM46[电气工程—电器] TP274[自动化与计算机技术—检测技术与自动化装置]