Remaining Useful Life Prognosis of Power Semiconductor Devices Based on On-state Resistance Time Series Data and Deep Learning

Power semiconductor devices are the essential core components of modern power electronic equipment. Predicting the remaining useful life (RUL) of power semiconductor devices can help formulate reasonable maintenance strategies for power electronic equipment, thereby improving the availability of energy conversion systems. This article proposes a deep learning method for predicting the RUL of power semiconductor devices, which uses on-state resistance as a critical state parameter. A long short-term memory (LSTM) neural network (NN) model and a convolutional neural network (CNN) model are used to predict the on-state resistance time series data trend. NN models can predict a short section of future data with past data input. The RUL of power semiconductor devices is estimated through curve fitting based on past and predictive data. The proposed method is validated using the power MOSFET accelerated aging dataset from NASA, and the validation results show that the performance of the proposed method is superior to that of existing methods, such as extended Kalman filtering (EKF) and particle filtering (PF).