Intelligent Classification of Power Quality Events Using a Hybrid Method based on Random Forest, Signal Processing and Feature Selection

Nowadays, due to load sensitivity, the quality of electric power has gained special importance. Therefore, accurate diagnosis of power quality events is essential before any corrective measures. In this paper, a hybrid intelligent method based on random forest, signal processing, and feature selection is presented to distinguish power quality events. In the first stage, by using Fourier transform and S-transform, different features are extracted from the signals of power quality events. Because there are stationary and non-stationary power quality events, the Fourier transform and the S-transform can be effectively used for extraction of potential features by analysis in both the frequency and time-frequency domains. The second step selects more valuable features using the Gram-Schmidt feature selection method and removes redundant features because the large dimensions of the feature vector can negatively impact the algorithm's generalizability. In the third stage, by using random forest, the power quality events are separated from each other based on the selected superior features. The obtained results show that the proposed method can distinguish 15 types of power quality events with a detection accuracy of 99.5%. Moreover, the detection accuracy is about 98.5% despite the presence of noise with a signal-to-noise ratio of 25 dB.