Parametric Fault Diagnosis of DC-DC Converters in Noisy Environments Based on Temporal Convolutional Networks

To address the problem of low accuracy of parametric fault diagnosis of DC-DC converter in noisy environment, this article develops a fault detection approach utilizing Temporal Convolutional Network. Firstly, the time-domain sequence data of four measurement points in the boost circuit under different faults are collected, and the original signals are used as fault representations without manually extracting the features. The features are automatically learned and extracted by the network, and the fault diagnosis results are obtained at the softmax layer. Experiments show that the Temporal Convolutional Network model proposed in this paper achieves <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 9. 2 8 \%}$</tex> correct rate under <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{6 d b}$</tex> Gaussian noise injection, which is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2. 7 4 \%}$</tex> higher than the correct rate of 1DCNN, effectively proving the superior performance of TCN in noisy environment.