A Novel Electric Field Feature Set of Transmission Line - Tower Air Gaps and Its Application for Insulation Strength Prediction

Abstract

Air gap is the main form of external insulation of transmission lines, whose discharge characteristics are closely associated with the spatial electric field distribution between electrodes. This paper defines 83 features to characterize the electric field distribution of transmission tower gaps, which are determined by the original data of hypothetical interelectrode paths and equipotential rings obtained by finite element method. A machine learning model constructed by support vector machine and trained by electric field feature set of 500kV, ±660 kV, and 750 kV transmission line - tower air gaps was used to calculate the switching impulse discharge voltages of ±800 kV and 1000 kV air gaps. The predicted results with different input feature subsets, which were selected by Pearson correlation coefficient, was analyzed and compared with experimental values. The results show that the minimum mean absolute percentage error is only 3.4 % after feature selection. This method is useful to predict long air gap discharge voltage with engineering gap configurations.