Support Vector Regression Based S-transform for Prediction of Single and Multiple Power Quality Disturbances

This paper presents a novel approach using Support Vector Regression (SVR) based S-transform to predict the classes of single and multiple power quality disturbances in a three-phase industrial power system. Most of the power quality disturbances recorded in an industrial power system are non-stationary and comprise of multiple power quality disturbances that coexist together for only a short duration in time due to the contribution of the network impedances and types of customers’ connected loads. The ability to detect and predict all the types of power quality disturbances encrypted in a voltage signal is vital in the analyses on the causes of the power quality disturbances and in the identification of incipient fault in the networks. In this paper, the performances of two types of SVR based S-transform, the non-linear radial basis function (RBF) SVR based S-transform and the multilayer perceptron (MLP) SVR based S-transform, were compared for their abilities in making prediction for the classes of single and multiple power quality disturbances. The results for the analyses of 651 numbers of single and multiple voltage disturbances gave prediction accuracies of 86.1% (MLP SVR) and 93.9% (RBF SVR) respectively. Keywords: Power Quality, Power Quality Prediction, S-transform, SVM, SV