基于一维卷积神经网络的低压万能式断路器附件故障诊断  

Fault Diagnosis of Accessories for the Low Voltage Conventional Circuit Breaker Based on One-Dimensional Convolutional Neural Network

作　　者：孙曙光 李勤 杜太行 崔景瑞 王景芹[2] Sun Shuguang;Li Qin;Du Taihang;Cui Jingrui;Wang Jingqin(School of Artificial Intelligence Hebei University of Technology,Tianjin 300130 China;State Key Laboratory of Reliability and Intelligence of Electrical Equipment Hebei University of Technology,Tianjin 300130 China)

机构地区：[1]河北工业大学人工智能与数据科学学院,天津300130 [2]省部共建电工装备可靠性与智能化国家重点实验室(河北工业大学),天津300130

出　　处：《电工技术学报》2020年第12期2562-2573,共12页Transactions of China Electrotechnical Society

基　　金：河北省教育厅科研资助项目(ZD2016108)。

摘　　要：由于低压万能式断路器分合闸附件的线圈回路采用交流供电方式,因此线圈回路合闸相位的随机性会导致同一运行状态下电流信号存在差异。利用传统的智能故障诊断方法可能会造成电流信号故障特征提取不准确,导致故障识别率降低。针对此问题,提出一种基于第一层宽卷积核自适应一维深度卷积神经网络(AW-1DCNN)的故障诊断算法。相较于传统智能诊断方法中存在人工特征提取与故障分类两个阶段,该方法将两者合二为一。首先,考虑到分合闸线圈电流信号的特点,采用一维卷积神经网络模型,并将模型的第一层卷积层的卷积核设为宽卷积核来扩大感受野区域;然后,利用特征提取层对电流信号进行自适应特征提取;最后,利用Softmax分类器输出故障诊断结果。实验结果表明,该算法不仅能对不同相位下同一故障进行有效识别,而且在泛化实验中仍能保持较高的故障识别率,能够有效地克服合闸相位变化对故障诊断结果的影响。Since AC power supply is adopted in the coil circuit of switching accessories for low voltage conventional circuit breaker,the randomness of the closing phase angle of the coil circuit may cause the difference of current signals under the same operating state.Using the traditional intelligent fault diagnosis method may lead to inaccurate fault feature extraction of current signal,and then result in lower fault identification rate.To solve this problem,an intelligent fault diagnosis algorithm based on adaptive one-dimensional deep convolutional neural network with wide first-layer kernel(AW-1DCNN)is proposed.Compared with the traditional intelligent fault diagnosis method including two stages of manual feature extraction and fault classification,the proposed method combines these two stages into one.Firstly,considering the characteristics of the current signal of the switching coil,a one-dimensional convolutional neural network model is adopted,and the convolution kernel for the first convolutional layer of the model is set as a wide convolution kernel to expand the receptive region.Secondly,the feature extraction layer is used to complete the adaptive feature extraction of the current signal.Finally,the fault diagnosis results are output by the Softmax classifier.The experimental results demonstrate that the proposed algorithm can not only effectively identify the same fault at different phase angles,but also maintain a high fault identification rate in the generalization experiment,which effectively overcomes the influence of closing phase angle on fault diagnosis results.

关 键 词：万能式断路器 分合闸附件 线圈电流 一维深度卷积神经网络 故障诊断 

分 类 号：TM561[电气工程—电器]