Classificação de descargas parciais no espectro UHF utilizando condicionadores de sinais e aprendizagem de máquina.

Abstract

This work proposes a methodology for the generation and classification of phase resolved partial discharges (PRPD) patterns based on the use of a printed UHF monopole antenna and signal conditioning in order to reduce hardware requirements. Subsampling and envelope detection techniques were applied for this purpose. In addition, test objects such as a hydrogenerator bar, dielectric disks with internal cavities, potential transformers and tip-tip electrodes immersed in a vat of oil were used to generate partial discharge (PD) signals. To detect and classify partial discharges, use the standard electrical method per IEC 60270 (2000) as a reference. After acquisition of subsampled and conditioned UHF signals, digital signal filtering techniques such as thresholding and wavelet transform were used and peaks of partial download pulses were extracted. For the classification, control techniques and choice of the best attributes were used to train machine learning algorithms, such as multilayer perceptron, support vector machine and decision tree. Accuracies greater than 90% were met, revealing the classification potential of the methodologies proposed in this work.