Evaluation of artificial intelligence techniques in the classification of partial discharges
Tallennettuna:
| Tekijät: | , , , , |
|---|---|
| Aineistotyyppi: | artículo original |
| Tila: | Versión publicada |
| Julkaisupäivä: | 2021 |
| Kuvaus: | The detection of Partial Discharge (PD) signals in the Ultra High Frequency (UHF) range performs identify and classify, in a minimally invasive way, of defects in high voltage equipment, as well as estimating the degree of urgency in carrying out preventive maintenance. In this paper, machine learning techniques were used to perform automatic recognition of patterns obtained from PD UHF signal envelopes. Therefore, an experimental arrangement was designed to emulate different PD sources: an oil vat with flat-tip electrodes, a hydro generator bar, and a potential transformer. From the signals launched in this arrangement, envelopes were generated, from which a series of attributes in the time domain were extracted, such as kurtosis, maximum amplitude, and rise time. Then, the selection of attributes was carried out through an association of algorithms, including k-means, to reduce the dimensionality of the data to increase the efficiency of the classifier algorithm. Finally, a classification of PD signals was performed using an artificial neural network, decision tree, and random forest. The results induced that the attributes extracted from the envelopes were effective in classifying PD signs, with mean accuracy values g reater than 95% when the optimized database was used. |
| Maa: | Portal de Revistas TEC |
| Organisaatio: | Instituto Tecnológico de Costa Rica |
| Repositorio: | Portal de Revistas TEC |
| Kieli: | Español |
| OAI Identifier: | oai:ojs.pkp.sfu.ca:article/6047 |
| Linkit: | https://revistas.tec.ac.cr/index.php/tec_marcha/article/view/6047 |
| Sanahaku: | Machine learning classification data mining partial discharge Aprendizado de máquina classificação descargas parciais extração de atributos |