Classification of Single and Combined Power Quality Disturbances Using Stockwell Transform, ReliefF Feature Selection Method and Multilayer Perceptron Algorithm

Authors : Düzgün AKMAZ

Pages : 13-23

Doi:10.46572/naturengs.1033182

View : 14 | Download : 8

Publication Date : 2022-06-30

Article Type : Research Paper

Abstract :: In this study, a method based on Stockwell transform insert ignore into journalissuearticles values(ST);, ReliefF feature selection method and Multilayer Perceptron Algorithm insert ignore into journalissuearticles values(MPA); algorithm was developed for classification of Power Quality insert ignore into journalissuearticles values(PQ); disturbance signals. In the method, firstly, ST was applied to different PQ signals to obtain classification features. A total of 30 different classification features were obtained by taking different entropy values of the matrix obtained after ST and different entropy values of the PQ signals. The use of all of the classification features obtained causes the method to be complicated and the training/testing times to be prolonged. Therefore, so as to determine the effective ones among the classification features and to ensure high classification success with less classification features, ReliefF feature selection method was used in this study. PQ disturbances were classified by using 8 different classification features determined by ReliefF feature selection method and MPA. The simulation results show that the method provides a high classification success in a shorter training/testing time. At the same time, simulation results have shown that the method was successful on testing data with noise levels of 35 dB and above after only one training.
Keywords : Classification, Multilayer perceptron algorithm, Power quality, Relief feature selection, S transform