Identification of Power Quality Disturbance Signals with Stockwell Transform, ONE-R Feature Selection Method and Random Forest Algorithm

In this study, a method was applied to classify the Power Quality (PQ) distortion signals. In the method, a 2- dimensional Time-Frequency (TF) matrix was obtained by applying the Stockwell transform (S-transform-ST) to different PQ signals. Then, classification features were obtained with different entropy values of the obtained matrix and PQ signals. However, using all of the obtained classification features causes the performing method to become complicated. Therefore, in this study, the ONE-R feature selection method was used to determine the effective features in classification, to ensure high classification success with fewer features and a simpler structure. PQ disturbances were classified with the determined features and Random Forest (RF) algorithm. The simulation results showed that the method was successful. At the same time, the simulation results showed that the applied method, once trained, is also effective in the classification of different PQ disturbances with noise levels of 35 dB and above.