A Fault Diagnosis Method of Electronic Current Transformer Based on Deep Learning

In order to improve the fault diagnosis efficiency of electronic current transformers under strong electromagnetic disturbance, a deep learning-based method fot fault diagnosis is proposed. First, carry out comprehensive laboratory simulation tests of equipment failure process, stably repeat the failure process of electronic current transformers, classify waveforms obtained from electromagnetic compatibility tests of different electromagnetic interference sources, and establish data sets. Secondly, based on the causal relationship between the characteristic quantities of strong electromagnetic disturbance and the fault characteristic quantities of electronic current transformers, A fault diagnosis model utilizing a convolutional neural network driven by image data is constructed, with PP-YOLOv2 chosen as the detection model. Finally, in the process of model training, compare different learning rate decay strategies. The strategy that provided the optimal mean average precision (mAP) was selected, leading to an impressive performance with the model achieving 72f/s. The fault diagnosis model can diagnose the cause of EMC failure and trace the type of transient electromagnetic disturbance that caused the fault. Through the construction and testing of the electronic transformer fault diagnosis model, the corresponding relationship between the electromagnetic protection level and the fault characteristic quantity can be quantitatively evaluated to make the diagnosis process intelligent, which can effectively diagnose the cause of the fault of the field equipment, and is conducive to supporting the fault of the electronic transformer Mechanism analysis and fault diagnosis work.