Fault Diagnosis of Transformer Winding Based on VMD-SVM

Abstract Considering the nonlinear and non-stationary characteristics of transformer vibration acceleration signal obtained from surface tank, variational mode decomposition (VMD) theory is introduced. Simulation analysis shows that VMD has obvious advantages over EMD when needle frequency similar to signal. It effectively avoids two types modal aliasing over-decomposition, accurately reflects source Aiming at problem core parameters support vector machine are difficult determine, Pareto particle swarm method used perform multi-objective parallel optimization obtain optimal parameters. The VMD-SVM fault diagnosis model tested using instance data, compared with other methods. test results show proposed in this paper highest diagnostic accuracy realizes latent power winding. accurate diagnosis.