Fault Diagnosis Method of Three-phase Inverter Based on Time Convolutional Neural Network

Abstract

Abstract: Power electronics converter is widely used in various applications. When the converter failure occurs, the system will be damaged, so it is of great significance to timely detect the fault and apply effective protection. In this paper, the open-circuit fault of the three-phase inverter is taken as the main object. The time convolutional neural network (TCN) model is designed to realize the fault diagnosis strategy of the inverter based on feature extraction. The simulated experimental results show that the neural network model can realize high accuracy fault diagnosis for different inverter circuits. The overall classification performance is good, and the loss is small. It is shown that the time convolutional neural network method can diagnose different IGBT fault intelligently in the inverter. It is proved that the neural network model is effective and feasible for inverter circuit fault diagnosis. Background: Power electronics converter is widely used in various applications. When the converter failure occurs, the system will be damaged, so it is of great significance to timely detect the fault and apply effective protection. Objective: The open-circuit fault of the three-phase inverter is taken as the main object. Method: The time convolutional neural network (TCN) model is designed to realize the fault diagnosis strategy of the inverter based on feature extraction. Results: The simulated experimental results show that the neural network model can realize high accuracy fault diagnosis for different inverter circuits. The overall classification performance is good, and the loss is small. Conclusions: It is shown that the time convolutional neural network method can diagnose different IGBT fault intelligently in the inverter. It has been proved that the neural network model is effective and feasible for inverter circuit fault diagnosis.