Classification of Power Quality Disturbances with Hilbert-Huang Transform, Genetic Algorithm and Artificial Intelligence/Machine Learning Methods

Abstract

In this study, Hilbert-Huang Transform method and statistical features are obtained to classify Power Quality (PQ) Disturbances. The appropriate features are selected by the Genetic Algorithm (GA) and k-Nearest Neighbor (KNN) classification approach. Models based on Artificial Intelligence and Machine Learning methods are formed and test process is performed by using data from experimental setup. Noisy situations are produced using mathematical equations. In addition, PQ Disturbances data from the experimental setup is also used in this study. Firstly, Empirical Mode Decomposition (EMD) method is applied to the signals. Then, by applying Hilbert transformation (HT), statistical features are extracted. The same procedure is repeated for Ensemble Empirical Mode Decomposition (EEMD). GA + KNN wrapper approach is used to select necessary features from feature subset. PQ Disturbances models are created based on Multilayer Perceptron (MLP) and KNN methods. The performance of EEMD + HT + GA + KNN classification model for 9 single and 9 multiple types of disturbances is 99.15% for synthetic data and 99.02% for experimental data. Compared to the literature, EEMD + HT + GA + KNN method has the ability to distinguish 9 multiple PQ disturbances. The overall performance gives the best performance with a rate of 99.12%.