Breakdown voltage prediction of SF6 gaps based on electric field features and SVM algorithm

Abstract

The breakdown voltages of SF6 gaps in uniform field and non-uniform field are predicted by a new method based on the electric field features and support vector machine (SVM). The finite element method (FEM) is used to calculate the electric field distribution of the SF6 gap. The parameters including the electric field strength, electric field energy, energy density, electric field gradient and scale parameters are used to characterize the electric field distribution of SF6 gaps. The breakdown voltage prediction model is established by SVM, the electric field features and the gas pressure are taken as the input to the SVM model. The output is whether the gap will breakdown under a given applied voltage and a certain gas pressure. Several experimental values of breakdown voltage are set as training samples and the corresponding electric field features are applied to train the SVM model. The improved grid search method is used to search optimal parameters including the penalty coefficient and kernel function parameter. The proposed method is applied to predict the breakdown voltages of coaxial cylinder gaps and rod-plane gaps in SF6. The predicted results coincide with the experimental values very well.