基于残差网络回归的直流断路器电寿命预测方法  

Electrical Life Prediction Method of DC Circuit Breaker Based on Residual Network Regression

作　　者：杨诺铭 吴自然[1,2] 刘泽州 吴晨茜 吴桂初 林义刚 YANG Nuoming;WU Ziran;LIU Zezhou;WU Chenxi;WU Guichu;LIN Yigang(Low Voltage Apparatus Technology Research Center of Zhejiang(Wenzhou University),Wenzhou 325035,China;Technology Institute of Wenzhou University in Yueqing,Wenzhou 325699,China;Zhejiang Tianzheng Electric Co.,Ltd.,Wenzhou 325604,China;Zhejiang Juchuang Intelligent Technology Co.,Ltd.,Wenzhou 325005,China)

机构地区：[1]浙江省低压电器工程技术研究中心(温州大学),浙江温州325035 [2]温州大学乐清工业研究院,浙江温州325699 [3]浙江天正电气股份有限公司,浙江温州325604 [4]浙江聚创智能科技股份有限公司,浙江温州325005

出　　处：《电器与能效管理技术》2024年第10期7-18,共12页Electrical & Energy Management Technology

基　　金：浙江省自然科学基金资助(LY23E070001);温州市重大项目资助(ZG2023049);温州市揭榜挂帅项目资助(ZG2023018)。

摘　　要：针对直流电弧分断特性,提出一种预测直流断路器电寿命预测方法。根据GB/T 14048.2—2020进行测试,并获取分断过程中的电压波形。将断弧发生一段时间内的电压信号经过CEEMDAN滤波后,作为网络训练样本。采用触头质量表示电寿命状况,利用线性插值得到触头质量下降曲线,并将其作为网络训练标签。采用Pytorch框架搭建一种残差网络(ResNet)的一维卷积结构预测模型。实验结果表明,所提方法对于触头质量的预测误差<10%,证明了所提方法可以用于预测GB/T 14048.2—2020标准下的电寿命。According to the breaking characteristics of DC arc,a method for predicting the electrical life of DC circuit breaker is proposed.According to the standard GB/T 14048.2—2020,the voltage waveform during the breaking process is obtained.After CEEMDAN filtering,the voltage signal within a period of arc break is used as the network training sample.The contact mass is used to represent the electrical life condition,and the contact mass decline curve is obtained by linear interpolation,which is used as the network training label.A one-dimensional convolutional structure prediction model of residual network(ResNet)is constructed using Pytorch framework.The experimental results show that the prediction error of the prosed method is less than 10%,which proves that the method can be used to predict the electrical life under GB/T 14048.2—2020 standard.

关 键 词：直流断路器 电寿命预测 残差网络 一维卷积 

分 类 号：TM561[电气工程—电器]