e-Article
Detection of super-high-frequency partial discharge by using neural networks.
Document Type
Article
Author
Source
Subject
*RADIO frequency discharges
*ARTIFICIAL neural networks
*OSCILLOSCOPES
*ELECTRIC substations
*ETHERNET
*RADIO frequency
*DETECTORS
*
*
*
*
*
*
Language
ISSN
1354-2575
Abstract
A system has been developed for the detection of super-high-frequency (SHF) partial discharge (PD) at frequencies up to 6 GHz. The system consists of three antennas for capturing PDs and a fast digital oscilloscope for sampling data. One of the antennas is a disk-cone antenna with frequency range below 710 MHz. The other two half TEM horn antennas have been designed and constructed for the frequency range 716 MHz – 5 GHz. To extend the frequency range up to 6 GHz, a methodology has been developed by compensating amplitude-response to frequency-magnitude. The compensation is realised by using multilayer feed-forward neural networks to equalise on amplitude-response. A direct sampling method is used to log the captured PD data. This PD detection system has been implemented to measure PDs at a 400 kV electrical substation (Strathaven, Scottish Power Ltd). [ABSTRACT FROM AUTHOR]