Video Discharge Extractor: a Deep Learning and Computer Vision-based Framework for Surface Discharges Recognition on HV Lines Insulators

A very serious problem affecting high voltage lines and stations is the phenomenon of surface discharges on insulators, with consequent possible service interruptions. In this paper, it is presented a novel approach in the detection of surface discharges, based on a deep learning and computer vision framework: the Video Discharge Extractor. The research has been carried out in the frame of an active collaboration with TERNA (the Italian TSO). The framework is able to process automatically the videos collected by the camera of the LANPRIS insulators ageing and monitoring station. The artificial intelligent system, based on an object detection approach using a convolutional neural network, has been improved and incorporated into an image processing chain able to maximise the information extracted from the monitoring videos. Such a framework could be a valid instrument for the TSO to process the videos to monitor the state of the ageing of the insulators and to detect any surface discharge phenomena. The framework developed is described in details and it is shown the potential of such an approach to image analysis in the context of the LANPRIS experimentation and not only. This work can act as a paradigm for further applications in the electro-energetic system where there is a need to analyse images and videos.