Islanding Detection Using Pattern Recognition by HOG Features with Ensemble Classifier

This study presents a novel approach for detecting islanding based upon image categorisation using Ensemble Classifier is proposed. Images are used to derive the histogram of oriented gradient (HOG) characteristics so as to identify non-islanding as well as islanding events. The ever-increasing need for power encourages the growth of Distributed Generation (DG). Islanding is one of the key challenges associated with excessive penetration of DG sources and can cause harm to the clients and their equipment. The islanding needs to be detected in two seconds, according to the IEEE 1547 DG interconnection rules, and the DG must be turned off. In the suggestive technique, spectrogram images are acquired through the time-series signals that are derived from the point of common coupling (PCC). The features are extracted from these images using the histogram of oriented gradient (HOG). The resulting feature vector is then utilised as an input for both the training and testing processes of the ensemble classifier. For deriving the spectrogram image, the rate of change of voltage is utilised as a parameter. In order to evaluate the performance of the ensemble classifier, 5, 10, 20, and 25-fold cross-validations are performed. The findings from the classification analysis indicate that the utilisation of an ensemble classifier in conjunction with the histogram of oriented gradient feature for image classification yielded remarkable outcomes, including an accuracy of 97.45% and a validation error of 0.188.