Monitoramento de para-raios de óxido de zinco com base na medição da corrente de fuga total.

Abstract

The majority of the metal-oxide surge arrester (MOSA) monitoring techniques is based on total leakage current (TLC) decomposition on their capacitive and resistive components. However, these techniques are subject to some financial, technical and practical limitations, which can difficult their employment. In this work, a new monitoring technique for MOSA is proposed. The technique is based on the feature extraction of measured TLC signals and on the analysis of these characteristics by a special kind of artificial neural networks (ANN) called Self-Organizing Maps (SOM). On technique implementation and validation, it was carried out, in laboratory, Maximum Continuous Operating Voltage (MCOV) tests in station class surge arresters with different rated voltages, aiming to measure the TLC signals. On the lab tests, it was developed a methodology to create artificial MOSA operating conditions, with the purpose of to evaluate the technique capacity on distinguish the MOSA condition only from the TLC analysis. Altogether, they were created six different kind of MOSA operating conditions, one of them corresponding to good condition, and the others, corresponding to bad conditions, where, the evaluated MOSA could be with: superficial pollution, displacement along the active column, non-uniform voltage distribution, varistor degradation or internal humidity. On the MOSA operating condition classification process were obtained hit ratios greater than 98.5%, when the MOSA was classified as defective or non defective. Hit ratios greater than 97.7% were reached on the identification of the MOSA defect. The results show that it is possible to perform the MOSA monitoring and diagnosis from only the TLC.