The improved entropy method and the grey correlation analysis method of transformer fault diagnosis is proposed to solve the problem of incomplete data information and low accuracy in fault diagnosis. Firstly, the method standardizes the sample data of transformer fault, then it use the improved entropy weight method to determine the index weights of transformer fault diagnosis, then by calculating the grey Euclid weighted correlation degree to determine the transformer fault types. The example analysis results show that the validity and feasibility of the method. Comparing with the traditional entropy weight method, the index weight distribution is more reasonable, and it can overcome the deficiency of the traditional grey correlation analysis indicators of parity, and the fault diagnosis accuracy is significantly higher than the IEC three ratio method and general grey relation analysis.