Feature extraction and pattern recognition of multi-source PD signals

Author

Gao, Kai ; Tan, Kexiong ; Li, Fuqi ; Gao, Wensheng

Author_Institution

Dept. of Electr. Eng., Tsinghua Univ., Beijing, China

fYear

2001

fDate

2001

Firstpage

119

Lastpage

122

Abstract

Partial discharge signals from industrial models of stator winding bars were used to compose multisource PD patterns, which were represented by 3-D charts. The tabulated data were calculated and served as characteristic features. Pattern recognition was performed using a back-propagation (BP) network. Successful classification of multi-source PD signals showed that tabulated data were sensitive and stable to reflect different PD patterns. The suitability and robustness of the BP network for classification was further verified by analyzing the untrained patterns´ recognition

Keywords

backpropagation; electric machine analysis computing; feature extraction; insulation testing; machine insulation; partial discharge measurement; pattern classification; stators; 3-D charts; HV power apparatus insulation systems; back-propagation network; feature extraction; industrial models; multi-source PD signals; pattern classification; pattern recognition; stator winding bars; untrained pattern recognition; Bars; Fault location; Feature extraction; Partial discharges; Pattern analysis; Pattern recognition; Power measurement; Robustness; Signal processing; Stator windings;

fLanguage

English

Publisher

ieee

Conference_Titel

Electrical Insulating Materials, 2001. (ISEIM 2001). Proceedings of 2001 International Symposium on

Conference_Location

Himeji

Print_ISBN

4-88686-053-2

Type

conf

DOI

10.1109/ISEIM.2001.973580

Filename

973580