A Comparative Study of Signal Processing and Pattern Recognition Approach for Power Quality Disturbance Classification

Abstract

This paper presents an extensive study of the advanced signal processing techniques for the classification of different Power Quality (PQ) disturbances. A detailed study of application of the signal processing techniques like Wavelet Transform (WT) and Wavelet Packet Transform (WPT) is carried out for the said purpose. These techniques are used to extract useful information from the raw signal in different frequency bands and give the time–frequency information. Hence, the statistical features are derived from this information and are used for the classification purpose. The features extracted are given to the Neural Network (NN) for training and subsequently it is tested for an effective classification. Three types of NN classifiers, namely, Multi Layer Feed Forward (MLFF), Probabilistic Neural Network (PNN) and Radial Basis Function (RBF) NN are analyzed for effective classification of PQ disturbances. For real-time implementation, one has to see the structural complexity of NN along with its capability of accurate classification. Hence, these NNs are compared with respect to classification performance and structural complexity. The simulation results show that the PNN offers acceptable classification accuracies.