S-transform and mahalanobis distance based approach for classifying power quality disturbances

Abstract

This paper presents the detection and classification of power quality (PQ) events or disturbances employing S-Transform and Mahalanobis Distance (MD) based approach. The proposed method exploits only four features extracted through S-transformation of the voltage signal; then, using these four features, classification is conducted by MD based classifier. Since proposed approach exploits less number of features, less memory space is required for classification and it facilitates to reduce the computational burden to a great extent. Furthermore, MD based classification does not require any off-line training of the PQ events; it only requires to form the feature matrix labeled with PQ event-wise groups. In this paper, classification of several PQ disturbances, such as, voltage sags, swells, harmonics, notch, flicker, transient oscillation, momentary interruption, etc., are considered. The simulation results demonstrate that the proposed approach is very effective and accurate in detecting and classifying PQ events. Moreover, comparative classification performance of MD based classifier with MED (minimum Euclidean distance) and LVQ (learning vector quantization) reveals the superiority of the proposed approach.