Empirical Mode Decomposition with Probabilistic Neural Network and Comparison with Support Vector Machine for Power Quality Events Classification

The enhanced use of nonlinear loads, power electronic devices and nonrenewable energy sources which are interconnection to the power grid are the major reasons for the Power Quality (PQ) disturbances. Due to these PQ disturbances the equipments are not operated properly may lead to the risk of black out. Hence it is very important task to the power system engineers to recognize the PQ disturbances by studying the distorted waveforms. This paper proposes an approach based on Empirical Mode Decomposition with Support Vector Machine (EMD-SVM) for feature extraction and classification of the PQ disturbances has been implemented. For classification of PQ disturbances various feature vectors which are obtained by EMD feature extraction method are used to train the support vector machines. The kernel and penalty parameters of the support vector machine (SVM) are determined by cross validation. Several PQ events, such as sag, swell, sag with harmonics, swell with harmonics, spike, notch, and harmonics are tested. After feature extraction step SVM is used to classify the PQ disturbance waveforms and the performance of SVM is evaluated with the Probabilistic neural networks (PNN).