S_Kohonen网络在断路器故障识别中的应用  

Application of S_Kohonen Network in Fault Identification of Circuit Breakers

作　　者：刘勇业 马宏忠[1] 张利[2] 屈斌[2] LIU Yongye;MA Hongzhong;ZHANG Li;QU Bin(College of Energy and Electrical Engineering,Hohai University,Nanjing 211100,China;Electric Power Research Institute of State Grid Tianjin Electric Power Company,Tianjin 300384,China)

机构地区：[1]河海大学能源与电气学院,南京211100 [2]国网天津市电力公司电力科学研究院,天津300384

出　　处：《高压电器》2020年第3期203-209,共7页High Voltage Apparatus

基　　金：国网天津市电力公司科技项目(KJ19-1-19)。

摘　　要：针对断路器操动机构常见故障问题,提出将改进的Kohonen网络对断路器典型故障进行诊断研究。首先,在某试验基地建立断路器监测实验平台,模拟操动机构典型故障。然后,分析采集到的合闸振动信号,采用3层小波包变换对降噪后的振动信号进行分解,得到各频段能量占总能量的百分比,利用粗糙集理论对特征参数进行知识约简,降低特征参数的维数。最终保留第1-4及第7频段。将保留的5个特征参数归一化处理后组成特征向量输入到S_Kohonen网络,对故障类型进行识别。实验结果表明,故障识别效果较好,且置信度高。Aiming at the issue of common failure problems of circuit breaker,the improved Kohonen network is proposed to diagnose the typical faults of circuit breakers.First,a monitoring experimental platform of circuit breaker is established in a test site,and the typical failure of the operating mechanism is simulated.Then,the closing vibration signal is analyzed and decomposed by three-layer wavelet packet transform after noise reduction to get the percentage of energy in each frequency band in total energy.Then the rough set theory is used to reduce the dimensionality of eigenvectors.At last,the first,second,third,fourth and seventh frequency bands are kept.The five feature parameters that are retained are normalized and then entered into S_Kohonen network to identify the fault type.The experimental result shows that the fault identification effect is better with a high degree of confidence.

关 键 词：断路器 故障模拟 特征向量 粗糙集理论 S_Kohonen网络 故障类型识别 

分 类 号：TM561[电气工程—电器]