Power quality disturbance classification with multi-classification SVM based on MST

Abstract

A power quality disturbance classification method based on modified S-transform (MST) and multi-classification support vector machine (SVM) is proposed in this paper. Firstly, the MST, which introduces two regulatory factors into traditional S-transform and obtains proper time and frequency resolution, is detailed. Then, the time-frequency matrix model is obtained through MST time-frequency analysis on the 7 kinds of common power quality disturbance signals, which include swell, sag, interruption, oscillatory, spike, notch, and harmonics. Furthermore, 11 features of time-domain and frequency-domain are extracted from the matrix model. Finally, the extracted features are sent into multi-class SVM to achieve automatic classification. The simulation results indicate that the proposed method not only can avoid the unchangeable and fixed varying patterns of the window in ST with practicability and adaptability, but also is an effective method for power quality disturbances classification.