Detection and Feature Extraction of Single Power Quality Disturbances Based on Discrete Wavelet Transform, Energy Distribution and RMS Extraction Methods

Monitoring of power quality disturbances (PQD) in power systems is crucial in determining their causes and avoid equipment damages. In this work, a MatlabR algorithm was implemented to detect and extract the distinctive features of seven simple power quality disturbances (sag, swell, interruption, flicker, harmonics, oscillatory transient, and notch). Firstly, a database was generated with the seven types of disturbances designed from their mathematical models. The decomposition of the signals was subsequently performed using the Discrete Wavelet Transform (DWT) through the Multi-Resolution Analysis (MRA) with six levels of details. The sampling frequency was varied to identify the useful features that, with energy distribution and RMS extraction methods, serve as input to classifiers to distinguish disturbances. Three classifiers were considered to demonstrate the effectiveness of the algorithm to identify the useful features, Probabilistic Neural Network (PNN), k-nearest neighbors (k-nn) and Multilayer Feed Forward Neural Network (MLFF). With the proposed method, it was possible to detect the seven simple disturbances analyzed, maintaining a balance between simplicity, robustness, and efficiency, which will have an impact on guaranteeing a lower processing cost and can be used in real-time applications.