Islanding Detection Using Image Categorisation with HOG Features by Support Vector Machine

A novel approach for detecting islanding based on image classification is described in this research utilizing support vector machines (SVM). The growing demand for power drives the growth of Distributed Generation (DG). One of the key issues associated with excessive penetration of DG sources is islanding. Images are utilized to generate the histogram of oriented gradient (HOG) features in order to distinguish between non-islanding and islanding events. The suggestive approach generates spectrogram visuals from time-series signals collected from the point of common coupling. These images are used to extract features using an oriented gradient histogram, which is then used as a feature vector input for the support vector machine classifier during training and testing. It derives the spectrogram image using the rate of change of negative sequence voltage (ROCONSV) parameter. To assess the detection accuracy of the suggestive technique, various kernels are used. The effectiveness of the SVM classifier is evaluated using 5, 10, 15 and 20-fold cross-validation. The classification results show that identification of islanding using a support vector machine classifier and histogram of oriented gradient feature achieved great results with 100% accuracy and a validation error of 0.155.