Bagging Ensemble Classifier for Predicting Lightning Flashovers on Distribution Lines

This paper introduces a bagging ensemble classifier, built from support vector machines (SVM), for predicting lightning flashovers on overhead distribution lines (OHL). Support vectors from the underlying SVM give rise to the so-called curve of limiting parameters (CLP), which features prominently in the statistical method of insulation coordination. Proposed machine learning-based approach enables a straightforward derivation of the line's CLP-from simulations or actual measurements data gathered by the lightning location systems-for its subsequent use in insulation coordination studies. It also facilitates computing the risk of insulation flashover. Both these aspects fully endorse statistical approach to the insulation coordination and flashover performance analysis of OHLs.