Investigation on surface electric field distribution features related to insulator flashover in SF6 gas

Abstract

The gas-solid interface is generally considered to be the weakest part in gas-insulated systems. The flashover voltage of a gas-solid interface is difficult to predict because it involves an extremely complex physical process affected by multiple factors. In this paper, we employ the support vector regression (SVR) method based on mass data on the features of the electric field that are indicators of the flashover voltage. We cast 48 kinds of epoxy resin post insulators as samples, and 61 features are extracted to characterize the surface electric field distribution and the flashover voltage under standard lightning impulse voltage. Four indicators that may be useful in predicting the flashover voltage were obtained; the maximum electric field along the surface of the insulator (E max ), the normal electric field components of E max (E n_Emax ), the maximum gradient of the surface tangential electric field component (E t_gm ), and the integral of the area in which the E value exceeds x% of E max along the insulator surface (V t60 ). We establish an SVR flashover voltage prediction model with these four features, whose average deviation is 7% in 10,000 predictions. It is expected that this study may provide a reference for the continued study of the gas-solid surface flashover mechanism.