FERREIRA, T. V.; http://lattes.cnpq.br/9395719025602516; FERREIRA, Tarso Vilela.
Resumen:
This work presents an electrical insulator pollution estimation technique based
on the ultrasonic noise emitted by them, when connected to energized electrodes. In
order to attain this objective, laboratory tests were performed, during which the
ultrasonic noise was digitally registered for further study of the best processing and
attribute extraction method. As a consequence of this study, the Spectral Sub-band
Centroid Energy Vectors (SSCEV) algorithm was obtained, which can be understood as
a spectral compression, capable of selecting the most significant frequency bands of the
noise. Afterwards, the processed audio, changed into SSCEV, constituted a database
which was fed to an Artificial Neural Network, capable of distinguishing with
remarkable precision a SSCEV from a polluted insulator from a SSCEV from a less
polluted insulator. Finally, in order to validate the technique in the field, measurement
campaigns were performed in the substation Campina Grande II, of the São Francisco
Hydroelectric Company. During these campaigns, ultrasonic noise from several
electrical equipments, exposed to different natural pollution degrees, was obtained, and
the processing, based on SSCEV and Artificial Neural Network was once again applied.
As a result, success rates of over 80% were generally obtained by the Artificial Neural
Network.