Detecting Open-Circuit Faults in Power Electronic Converters Using Continuous Wavelet Transform and Convolutional Neural Networks for Simultaneous Charging Systems

Simultaneous charging is one of the key factors for advancing electric vehicles, aiming primarily to reduce waiting times at charging stations. The reliability and safety of power electronics converters used in simultaneous charging have attracted considerable attention from researchers due to their susceptibility to failure. Conventional methods for detecting open-circuit faults (OCFs) in power electronic switches face challenges, such as weak feature extraction capabilities and the need for numerous sensors, which result in reduced accuracy in identifying faulty switches. This study proposes a strategy for detecting OCFs in a bidirectional isolated dual-active-bridge (DAB) DC-DC converter to enhance system reliability. The proposed strategy leverages average voltage values to promptly identify faulty switches without additional hardware cost. Additionally, the continuous wavelet transform (CWT) is employed to extract features by transforming the voltage signals into two-dimensional time–frequency representations. These representations are then processed using AlexNet, a well-established deep learning network, to localize faults across different switches. The proposed CWT-CNN strategy effectively reduces the number of required sensors, thereby lowering the overall cost. Simulation results demonstrate that this approach achieves high accuracy in detecting OCFs across a range of operating conditions.