Development, application and evaluation of a computational tool for management high voltage break disconnector based on self-organizing maps and image processing

This work has the objective of developing, analysing and applying a new tool for management the status of break disconnectors in high voltage substations from digital images. This tool uses a non-supervised kind of artificial neural network using the Kohonen learning algorithm, known as a self-organizing maps. In order to develop the proposed tool, C/C++ programming language, provided with easily used interfaces, is used. In order to obtain the results, three environments are considered: one for laboratory simulation and two pilot projects installed in the Fortaleza II/CHESF substation. These pilots are used for 230 kV EV-2000 type and 500 kV semi-pantographic type break disconnector management tests. The results prove the developed system’s efficiency, because it is able to detect 100% of open and closed identification situations. However, the neural network utilised for management break disconnectors has become suitable for installation in high voltage substations in order to support the maintenance team in safely handling these disconnectors.