Monitoramento e diagnóstico de para-raios a ZnO através de redes neurais artificiais.

Abstract

Lightning arresters provide protection for transmission lines and power equipments. The correct operation and maintenance of these equipments is an important task in substations as they can assure the integrity of important equipments, like power transformers. During the occurrence of lightning or switching surges, a voltage level above the equipment insulation level can be reached. An arrester, properly working, is able to limit this voltage level avoiding damages to the protected equipment. A defective arrester may not be able to provide the proper protection level in substations, exposing to damage the equipments it should protect. Besides that, it can be a risk for the integrity of other equipments and nearby people as on of the consequences of a defective arrester is the possibility of its explosion. Routine inspections are the better way to evaluate the integrity and proper operation of arresters. Monitoring of surge arresters is usually conducted by means of current measurement or thermal images acquisition. But, among power companies, there is no model procedure for the monitoring conduction and analysis of the obtained results. Besides that, when some abnormality is detected, the arrester is replaced by a new one and no further study is conducted to evaluate what kind of problem happened to it. This works proposes a method for analysis of ZnO arresters by the usage of artificial neural networks - ANN. The ANN is able to analyse the thermal profile and detect and classify patterns the could be undetected by a simple visual analysis. Besides that, if the ANN is properly trained with a wide range of different possible arrester failures, it can also be able to detect what kind of failure is usually happening to the equipments of a certain plant. It is a useful tool for engineers as they give further information about the problem, aiding in a maintenance decision and also indicating the possible origin of the problem, which allows the proper resolution to be taken as a preventive assessment.