Comparison of methods of recognition and classification of insulation defects of power equipment

The paper presents the results of comparison of methods for recognition and classification of insulation defects from partial discharges depending on the number of extracted signal features. For this purpose, the built-in application MATLAB Classification Learner and Keras library were used. The results were compared using the learning algorithms Support Vector Machine, Nearest Neighbour Classifiers and Convolutional Neural Network for four types of partial discharge defects.