Least Square Support Vector Machine Modelling of Breakdown Voltage of Solid Insulating Materials in the Presence of Voids

Abstract

The least square formulation of support vector machine (SVM) was recently proposed and derived from the statistical learning theory. It is also marked as a new development by learning from examples based on neural networks, radial basis function and splines or other functions. Here least square support vector machine (LS-SVM) is used as a machine learning technique for the prediction of the breakdown voltage of solid insulator. The breakdown voltage is due to partial discharge of five solid insulating materials under ac condition. That has been predicted as a function of four input parameters such as thickness of insulating samples ‘t’, diameter of void ‘d’, the thickness of the void ‘t 1’ and relative permittivity of materials ‘e r ’ by using the LS-SVM model. From experimental studies performed on cylindrical-plane electrode system, the requisite training data is obtained. The voids with different dimension are artificially created. Detailed studies have been carried out to determine the LS-SVM parameters which give the best result. At the completion of training it is found that the LS-SVM model is capable of predicting the breakdown voltage V b = (t, t 1, d, e r ) very efficiently and with a small value of the mean absolute error.