基于随机森林与长短时记忆神经网络的真空接触器故障诊断方法研究  

Research on Fault Diagnosis Method of Vacuum Contactor Based on RF-LSTM Model

作　　者：袁钰林 郑运鸿 游一民 戴冬云 张有锋 YUAN Yulin;ZHENG Yunhong;YOU Yimin;DAI Dongyun;ZHANG Youfeng(School of Electrical Engineering and Automation,Xiamen University of Technology,Fujian Xiamen 361024,China)

机构地区：[1]厦门理工学院电气工程与自动化学院,福建厦门361024

出　　处：《高压电器》2022年第5期103-111,共9页High Voltage Apparatus

基　　金：国电泉州热电有限公司科技项目(GDQZ-科技-2020-I-05);福建省科技厅引导性项目(2019H0039)。

摘　　要：针对真空接触器的渐发性故障识别准确率不高的现状,提出了一种基于随机森林与长短时记忆神经网络的故障诊断方法。文中分析了某型号12 kV真空接触器在机械保持工作情况下合闸线圈电流信号的故障特征,构建了两层诊断模型,在初步诊断中利用随机森林分类器,识别特征明显的突发性故障,利用长短记忆神经网络模型发掘数据时序特征的特点,识别渐发性故障,在最终诊断中利用证据融合将两者结果融合。文中提出的故障诊断模型有效解决了传统故障诊断方法对渐发性故障识别困难的不足,实验表明,该方法对渐发性故障识别准确率达到了91.1%以上,整体故障识别的准确率达到93.3%以上,该方法具有一定的应用价值。In view of low accuracy of the gradual fault identification of vacuum contactor,a kind of fault diagnosis method based on random forest and long short-term memory neural network is proposed.In this paper,the fault characteristics of the closing coil current signal of certain type of 12 k V vacuum contactor,in case of mechanical holding operation condition,is analyzed and two layers of diagnosis models are constructed. In the preliminary diagnosis,the random forest classifier is used to identify sudden faults with obvious characteristics,and the long-short memory neural network model is used to discover the characteristics of data time series characteristics and identify the gradual faults. In the final diagnosis,the results of both are fused by evidence fusion. The fault diagnosis model proposed in this paper has effectively solved the shortcomings of the traditional fault diagnosis method in identifying the progressive faults. The experiment shows that the accuracy of the method in identifying the progressive faults is over 91.1%,and the overall fault identification rate is up to 93.3%. This method has a definite application value.

关 键 词：真空接触器 电流曲线 随机森林 长短时记忆神经网络 证据融合 故障诊断 

分 类 号：TM572[电气工程—电器] TP18[自动化与计算机技术—控制理论与控制工程]