http://lattes.cnpq.br/4637812678569774; RIBEIRO, Girlene Lima.
Resumo:
This research presents a methodology for the improvement of the classification of polymeric insulators by using thermographic measurements and UV radiation in combination with Digital Image Processing (DIP) and Artificial Neural Networks (ANNs). The methodology is based on the analysis of the occurrence of corona discharges and temperature variations along the insulator in order to classify their stage of degradation. Each insulator was subjected to the 133 kV phase-to-ground voltage over a period of 30 minutes, in order to cause heating and corona discharges in the insulators. The experiments were performed using a corona detector for UV measurement and the temperature data were acquired using a thermal imager. The images acquired by the monitoring instruments during the tests were subjected to digital image processing to extract information of pixel density, persistence of discharges and relative distances from the discharge areas to the insulator. From information obtained through infrared (temperature) images descriptive statistics and Fisher's discriminant test were applied to present objective parameters with high level of separability to the classification system. The classification system used ANN to determine the insulators degradation state. The classification was performed in individual and in combination ways, with vectors formed by UV and infrared attributes. The developed system helped on the decision making, concerning to the necessity of intervention or not to the insulators. The classification of the insulators, in an individual way, obtained accuracy for temperature of 80.00% and UV 74.05%. The classification of the isolators, combined (UV and infrared), obtained an average accuracy of 92.58%, evidencing the improvement in the classification.