In order to solve the high misjudgment rate problem of relying on manual experience when using frequency response method to diagnose winding deformation and realize the precise diagnosis of winding deformation of transformer

a kind of deformation diagnosis method of transformer winding based on signal distance degree model and SVM is proposed. In this paper

the signal distance degree model is used to calculate the sub-frequency band mutual complex distance

mutual distance and correlation coefficient

take them as the characteristic and input them into the SVM for fault classification. The genetic algorithm (GA) is used for parameter optimization of the SVM.57 samples with a 5% grade and 90 samples with a 3% grade are simulated respectively by using Pspice. It is found by simulation that the accuracy rates of winding deformation fault classification by using the mutual complex distance as the characteristic are up to 96.296 2% and 97.222 2%

and the classification effect is significantly higher than that of the mutual distance that only considers the amplitude characteristics and traditional correlation coefficients. Its effectiveness is proved through actual winding deformation data of transformer

which provides a new ideas for fault diagnosis of transformer winding.