Icing thickness prediction of overhead transmission lines base on combined kernel function SVM

Abstract

Light icing of overhead transmission lines would cause huge hazards such as conductor galloping and line over load, while heavy ice would cause serious faults such as disconnection and tower falling down, resulting in paralysis of the grid. With the prediction of the icing thickness on transmission lines which can preclude the contingency caused by icing. This paper studies two types of kernel function (KF), the radial basis function (RBF) KF and polynomial (Poly) KF. The two kinds of KFs are applied to the support vector machine (SVM) to predict the icing thickness and then the results were compared. Based on the characteristics of RBF KF and Poly KF, a combined KF of weighting sum RBF KF and Poly KF is proposed. From the experimental results, the prediction performance of combined KF SVM is superior to single KF SVMs.