Intelligent Fault Diagnosis of Transformer Based on Infrared Image and Mask RCNN

The safe and stable operation of the transformer is of great significance to the power grid. In order to improve the efficiency of transformer diagnosis, this paper proposes an intelligent transformer fault diagnosis method based on masked RCNN. First, during operation and maintenance, infrared images of transformers are collected and marked to complete the preliminary construction of the data set. To alleviate the imbalance between positive and negative samples, the number of negative samples is expanded by rotation, scaling, clipping and translation, and a GHM loss function is introduced. In order to meet the requirements of real-time detection. Comparative experiments show that, compared with the original model, the improved Mask-RCNN has higher diagnostic accuracy and lower false positive rate. The fault diagnosis rate was 86.72%, and the false positive rate was 7.68%. It can accurately realize intelligent fault diagnosis of transformers, and has certain engineering value for smart grid construction.