基于神经网络的10kV配网高压断路器机械故障诊断方法  

Mechanical fault diagnosis method of 10kV distribution network high voltage circuit breaker based on neural network

作　　者：骆佳樑 刘晓 LUO Jialiang;LIU Xiao(State Grid Taicang Electric Power Supply Company,Suzhou 215400,China)

机构地区：[1]国家电网国网太仓市供电公司,江苏苏州215400

出　　处：《中国高新科技》2024年第16期117-119,共3页

摘　　要：由于高压断路器故障特征量为多源异构数据,单一故障特征无法准确表征设备的故障特性,导致故障诊断精度较低。为此,提出基于神经网络的10kV配网高压断路器机械故障诊断方法。通过采集设备的振动信号,对其进行分段和离散余弦变换处理,获得信号频带能量熵,结合幅频梯度选取机械故障特征量,结合傅里叶变换方法确定故障特征量的优先级,并构造故障特征空间,进而将单一故障特征进行融合。以此为依据,计算测试样本相对于某一类故障类别的匹配度,从而判定样本的所属故障类型。实验结果表明,所提方法得到的设备机械故障类别与实际故障类型完全一致,故障诊断精度较高。Because the fault characteristics of high-voltage circuit breakers are multi-source heterogeneous data,a single fault characteristic can not accurately characterize the fault characteristics of equipment,resulting in low fault diagnosis accuracy.Therefore,a mechanical fault diagnosis method of 10kV distribution network high voltage circuit breaker based on neural network is proposed.The vibration signal of the equipment is collected and processed by subsection and discrete cosine transform,and the energy entropy of the signal frequency band is obtained.The mechanical fault feature quantity is selected by combining amplitude-frequency gradient,and the priority of the fault feature quantity is determined by combining Fourier transform method,and the fault feature space is constructed,and then the single fault feature is fused.Based on this,the matching degree of the test sample relative to a certain fault category is calculated,so as to determine the fault type of the sample.The experimental results show that the mechanical fault types of equipment obtained by the proposed method are completely consistent with the actual fault types,and the fault diagnosis accuracy is high.

关 键 词：故障特征 故障信号幅频 神经元 机械故障 故障匹配度 

分 类 号：TH165.3[机械工程—机械制造及自动化]