Transformer fault diagnosis based on feature parameter preference and ICOA optimization CNN

Abstract To improve the accuracy of fault diagnosis for power transformers, an RFRFE-ICOA-CNN intelligent fault diagnosis method for power transformers based on Dissolved Gas Analysis (DGA) in oil is proposed. First, to address the issue that manually selecting fault feature parameters may lead to the omission of some key features, and that multi-dimensional raw fault data can increase the difficulty of transformer fault diagnosis, a method combining the Recursive Feature Elimination (RFRFE) algorithm with Random Forest is proposed for optimal selection of fault feature parameters. Next, the Improved Coati Optimization Algorithm (ICOA) is introduced to optimize the hyperparameters of the Convolutional Neural Network (CNN), such as learning rate, kernel size and number, and the number of neurons in the fully connected layer, in order to improve the accuracy of the model’s diagnostic results. Finally, through case studies, the performance of the established RFRFE-ICOA-CNN method is evaluated, and the effectiveness of the proposed method for transformer fault diagnosis is validated.