Transformer fault identification method based on Markov transition field and CNN-CBAM

Transformers operate in highly complex environments, which poses challenges for fault extraction. To enhance the accuracy of diagnosing transformer operation faults, this paper proposes a method based on Markov Transformation Field (MTF) and combines it with Convolutional Block Attention Module (CBAM) within a Convolutional Neural Network (CNN). In order to address the issue of limited differentiation of raw features across different defect types, MTF is employed to transform the raw data into modal form. This transformation aims to convert one-dimensional data into two-dimensional image data, allowing for an analysis of distribution characteristics specific to various fault types. Subsequently, the features extracted from the image data are utilized as input sources for classifying faulty fragment data using CNN-CBAM in order to obtain recognition results. Compared with traditional methods, this approach demonstrates more pronounced differences in features and achieves 96.4% diagnostic accuracy, thus validating its feasibility.