Abstract:As one of the main equipment in the power system,it is extremely important to diagnose transformer faults in advance.However,it is difficult to establish an accurate and perfect condition assessment model for power transformers by traditional technology,and then artificial intelligence is gradually coming into the public view as a new idea and method.This paper summarizes and systematically analyzes various traditional and existing transformer fault diagnosis methods at home and abroad,and lists in detail the latest advances in intelligent fault diagnosis technology based on dissolved gas analysis technology in oil or inspection images of power equipment.Various deep learning algorithms applied to transformer fault diagnosis are studied,such as deep neural network,sparse restricted Boltzmann machine,and deep confidence.The final results of various diagnostic techniques are compared.