•  
  •  
 

Keywords

transformer；fault diagnosis；dissolved gas in oil；ratio method；multi‑scale convolutional neural networks；antagonistic generation network

Abstract

In order to improve the diagnostic accuracy of transformer fault, a transformer fault diagnosis method based on the multi‑scale convolutional neural network model is proposed. Firstly, two multi‑scale convolutional modules are designed on the basis of the 1DCNN structure, and the overall structure of the transformer fault identification model is constructed. Secondly, to handle the problem of less sample features, the feature expansion method based on the ratio method is adopted to enhance the sample features from 5 dimensions to 25 dimensions. To solve the small sample size of faults and uneven distribution of sample numbers between faults, a sample number enhancement method based on adversarial generation network is adopted, and a large number of simulated samples are generated. Finally, the modified dataset was used to train and test the designed model. The results show that the average accuracy of the model is 93.24%, and the model performs well compared with the relevant mainstream methods under different datasets.

DOI

10.19781/j.issn.1673-9140.2023.04.011

First Page

104

Last Page

112

Download

Share

COinS