Transformer Defect Identification Method Based on Random Forest and Convolutional Neural Network

Transformers hold great significance for power systems. Regarding the detection of transformer operating states, automatic defect recognition is of paramount importance to the safe operation of the power grid. This paper studies and proposes a transformer defect recognition method based on random forest and convolutional neural networks. The algorithm in this paper first detects the operating state of the transformer from the perspective of operating data through the state assessment method based on random forest and then identifies the appearance defects of the transformer through the convolutional neural network. Finally, the two algorithms jointly determine the operating state and defect types of the transformer. Through experimental analysis and verification, the accuracy of this method in identifying the operating state and defect type of the transformer exceeds 90%, which is helpful to improve the intelligence and automation level of transformer operation and maintenance service.