基于深度学习的电流互感器隐性故障诊断方法

Diagnosis Method for Hidden Faults in Current Transformers Based on Deep Learning

电流互感器的正常稳定工作对电力保护系统至关重要,但其隐性故障检测一直是保护系统状态监测的难题。针对这一问题,提出基于循环卷积块网络的电流互感器隐性故障诊断模型。首先对电流互感器的电流以及告警信息时序数据进行预处理,然后通过循环卷积块和池化操作对隐性故障数据进行特征提取,通过softmax分类器进行分类识别,并构建了智能隐性故障诊断模型。最后利用某特高压换流站电流互感器采集实际数据,开展故障识别实验,验证了基于深度学习的电流互感器隐性故障诊断方法的有效性。

The normal and stable operation of current transformers is crucial to power protection systems,but their hidden fault detection is a difficult problem for protection system condition monitoring.To address this problem,this paper proposes a current transformer implicit fault diagnosis model based on cyclic convolutional block network.First,the current and alarm information timing data of current transformer are pre-processed.Then the hidden fault data are extracted by circular convolutional block and pooling operation,and classified and identified by softmax classifier.In addition,an intelligent hidden fault diagnosis model is constructed.Finally,the actual data collected from current transformers of a UHV converter station are used to carry out fault identification ex-periments,and the effectiveness of the deep learning-based implicit fault diagnosis method for current transformers is verified.