An algorithm for the classification of power quality disturbance signals using a tunable-Q-factor wavelet transform and ensemble learning methodology

Abstract

The automatic categorization of power quality disturbances (PQDs) presents a formidable task for utility and industry professionals. This study introduces a novel technique for the automatic classification of PQDs that involves employing an ensemble classification model combined with Tunable-Q wavelet transform (TQWT) analysis. The TQWT has been observed to accurately capture the sparsity in PQD signals within the time-scale domain. The methodology under consideration commences by implementing the TQWT to effectively assess the behavior and sparsity of PQD signals. The considered signals are decomposed through the use of TQWT into a set of sub-bands that possess limited bandwidth. This decomposition process enables enhanced feature extraction capabilities. In order to assess the efficacy of classification algorithms, a set of three distinct statistical features (variance, skewness, and kurtosis) is obtained from the TQWT decomposed signals, which are utilized as the training vector. The present study employs random forest and AdaBoost algorithm-based classifiers to effectively classify twelve frequently occurring PQD signals. An investigation was conducted regarding several power signal disturbances, each associated with varying levels of noise. The objective was to demonstrate the effectiveness of a proposed classification scheme when operating under noisy conditions. A comparative analysis of the classification accuracies using previously proposed techniques reveals a distinctly enhanced performance. The research results illustrate that the proposed ensemble classification model exhibits exceptional performance in terms of classification accuracy, achieving a perfect score of 100%. These findings further highlight the superior recognition performance of the random forest classifier within the ensemble.