Keywords
GIS, characteristic image, convolution neural network(CNN), deep belief network(DBN), model training
Abstract
Gas insulated switchgear (GIS) partial discharge fault type identification is an important basis for fault warning and maintenance planning, and is of great significance for maintaining the safe and stable operation of power equipment. This paper firstly analyzes several common types of GIS faults. Then, in the processing and classification of the spectral envelop signal collected by the UHF sensor, the composite neural network model formed by the fusion of the convolutional neural network (CNN) and the deep belief network (DBN) can quickly realize the extraction of effective feature signals and accurate classification of fault types. Therefore this paper integrates CNN and DBN, establishes the main structure of the composite neural network, and uses this network to identify GIS partial discharge fault types. Finally, the method is verified in simulation experiments. Results show that the accuracy of the composite neural network model to identify faults can reach up to 99%.
DOI
10.19781/j.issn.1673-9140.2021.04.020
First Page
157
Last Page
164
Recommended Citation
Yuan, Wenhai; Liu, Biao; Xu, Hao; Wang; Dong, Xiaoshun; Wang; Zhong, Lipeng; Si, Yufei; and Xia, Xin
(2021)
"Partial discharge fault type identification of GIS based on composite neural network,"
Journal of Electric Power Science and Technology: Vol. 36:
Iss.
4, Article 20.
DOI: 10.19781/j.issn.1673-9140.2021.04.020
Available at:
https://jepst.researchcommons.org/journal/vol36/iss4/20