A very serious issue affecting high-voltage lines and substations is the phenomenon of surface discharge on insulators, which can lead to potential service interruptions. This article presents a new approach to detecting surface discharges based on a deep learning and computer vision framework: the Video Discharge Extractor. The research was conducted in active collaboration with TERNA (the Italian TSO). The framework automatically processes videos collected from the LANPRIS insulator aging and monitoring station camera.

The artificial intelligence system, based on an object detection approach using convolutional neural networks, has been enhanced and integrated into an image processing chain to maximize the information extracted from monitoring videos. Such a framework could be a valuable tool for the TSO to process videos for monitoring the aging status of insulators and detecting any surface discharge phenomena. The developed framework is described in detail, showcasing the potential of such an approach in image analysis within the LANPRIS experimentation context and beyond.