Pattern Recognition Tool For Detection and Classification of Power System Transients

Abstract

Multiresolution time-frequency analysis has been an important field in the signal processing especially for developing pattern recognition techniques for health monitoring and diagnostics. In the last two decade, several algorithms have been applied successfully for power quality analysis, condition monitoring and biomedical signal processing. However, in the most of the studies the research has been limited to specific algorithm and application. In this paper, the authors present a Time-Frequency analysis tool developed for the study of Pattern Recognition problem. It was developed using MATLAB GUI with several Time-frequency algorithms such as S-Transform, Hyperbolic S-Transform, and TT Transform for comparative studies. The developed tool has been successfully applied for the pattern recognition of Power System Transients. Simulations were performed to study the efficacy of the tool for various power system transients such as Load Switching, Motor Starting, Lightning, Capacitor Energization and Back-to-Back Switching. A simple rule-based was developed using Time-Frequency features for detection and classification of various power system transients. The tool also maps the power system transients on CBEMA/ITC curve highlighting the nature of disturbance. The simulation results demonstrate that developed pattern recognition tool was efficient and useful for the study of partial discharges and acoustic signals for developing smart health monitoring system.