An Open-Circuit Fault Diagnosis Method for Charging Piles Based on Attention Mechanism

This paper proposes a fault diagnosis method based on channel attention mechanism for identifying switching device open-circuit failures of both the Vienna rectifier and resonant LLC Converter in electric vehicle charging piles. First, five-channel original data of the resonant capacitor voltage together with the output voltage and input currents are measured. They are subsequently segmented by a sliding window whose length is set as half fundamental cycle of the mains to obtain samples. Within each channel of each sample, the features are extracted so that different weights can be assigned by using the channel self-attention mechanism. Then, the enhanced feature map is input into a Multi-layer Perceptron (MLP) model for fault diagnosis. This method considers not only the signal measurability but also the possibility of multiple faults occurring in different switches. Experimental results validate the high diagnosis accuracy of the proposed method on the two-stage charging converters.