Partial Discharge Spectrogram Data Augmentation based on Generative Adversarial Networks

High voltage insulators are critical elements in power distribution and transmission systems, since their failure may represent losses in power quality and system reliability, leading to faults which bring problems to both consumer and electric utilities. Insulators in operation are subjected to all kinds of pollution which may cause the formation of dry bands when in presence of moisture, which leads to the formation of partial discharges which, in turn, causes degradation and aging of the surface material. Partial discharges origin can be identified by the pattern of their pulses. Machine learning algorithms are one of the most useful and efficient solutions for pattern recognition and classification tasks. In this paper, a methodology is proposed for data augmentation of the training set for an artificial neural network by means of generative adversarial networks.