Abstract:The operation status of the transformer is directly related to the safe and stable operation of the entire power system, and it is of great practical significance to effectively diagnose the transformer. Dissolved Gas Analysis (DGA) in power transformer oil has become an effective support data for oil-immersed transformer fault diagnosis. Based on the DGA data, this paper first summarizes the advantages and disadvantages of conventional IEC ratio method. Several effective improvements have been summarized for their boundary problems. Secondly, this paper summarizes the application of artificial neural network, support vector machine, rough set, fuzzy mathematics, extreme learning machine, Bayesian network, clustering, artificial immune and petri network in transformer fault diagnosis. The inherent problems summarize their respective optimization methods. Finally, this paper introduces a comprehensive diagnostic method based on evidence theory, analyzes its advantages over a single intelligent algorithm, and introduces the application of some other methods in transformer fault diagnosis. It is concluded that the integrated diagnostic approach to i-nformation fusion can better diagnose transformer faults than a single intelligent approach.