Classificação da hidrofobicidade em isoladores elétricos utilizando análise de frequência e RNA

Abstract

Transmission lines transport the energy produced in power plants to distribution centers. In these lines, insulators perform the role of segregating regions of different electrical potential, while accomplishing the mechanical function of supporting the cables. Due to the nature of their function, insulators are exposed to electrical and mechanical stress throughout their life span, in addition to withstanding the wear caused by the environment through solar radiation, moisture, pollution and other weather conditions. Regarding polymeric high voltage insulators, which are the kind of insulator most widely used in the current scenario, these stresses lead to a decrease in their surface hydrophobicity, allowing moisture to accumulate on the insulator, giving rise to a leakage current and increasing the probability of a flashover to occur. In this sense, the present work presents a methodology for evaluating and classifying the surface hydrophobicity of polymeric high voltage insulators based on the method proposed by the Swedish Transmission Research Institute (STRI). The classification is performed automatically, using a Multilayer Perceptron artificial neural network, based on digital image processing, using spatial frequency information. A method for segmenting hydrophobicity images produced under unbalanced lighting conditions and with low contrast is also proposed. Furthermore, an image database containing 1200 hydrophobic surface samples in various stages of degradation was created. Images of an insulating column collected in the environment of a 500 kV substation were also used to validate the proposed method. The results obtained were compared with two other methods in the literature and it could be seen that the methodology developed was able to successfully segment and classify surface hydrophobicity images, obtaining a success rate above 78% for the database used.