Abstract: As the core equipment of the power system,the operational reliability of transformers is crucial.Traditional fault diagnosis methods have limitations in timeliness and accuracy.Ultraviolet imaging technology provides new possibilities for early identification of transformer faults by capturing ultraviolet signals generated by corona discharge and partial discharge on the surface of the equipment.This paper aims to explore and establish an intelligent diagnosis method for transformer fault types based on ultraviolet imaging technology.The study first conducted an in-depth analysis of the performance characteristics of common transformer fault types (such as insulation aging, partial discharge,winding deformation,etc.) in UV images.On this basis, we focused on the multiprocessing technology of UV images (including noise suppression and image enhancement) and fault feature extraction algorithms (such as texture analysis and morphological processing),and used feature selection methods to optimize the feature set.In order to achieve automatic identification of fault types,a diagnostic model based on deep learning (such as a constitutional neural network) was constructed and optimized,and a variety of machine learning algorithms were used for comparison and verification.Experimental results show that the proposed method can effectively extract key fault information from UV images,and the constructed diagnostic model shows high accuracy and robustness in identifying transformer fault types,significantly improving the efficiency and reliability of fault diagnosis.This method provides strong technical support for intelligent operation and maintenance of power equipment.