GIS partial discharge recognition based on chaos features of the three-dimensional spectra

Fault diagnosis of gas-insulated switchgear (GIS) partial discharge (PD) is significant for the evaluation of GIS operation conditions. Traditional pattern recognition methods are limited to analysis and recognition the characteristics of PD spectra pattern distribution. Lack of a more comprehensive, more profound and more fundamental analysis on PD characteristics, which results in low recognition rate of a certain type of PD. Considering these problems, this paper proposes a GIS PD recognition method based on the chaos theory. Collect PD signals of 100 frequency cycle are collected continuously and an φ-v-n 3D spectra sample F matrix is formed. Taking one column of this matrix as a signal sequence and then conducting chaos analysis, that is, calculating the largest Lyapunov exponent corresponding to the same phase signal sequence and the distribution characteristics of 36 largest Lyapunov exponents in different phases as PD chaotic characteristics are obtained. The experimental results show that the extracted chaos features can reveal the essential of PD. The whole recognition is better and has high rate recognition of air gap defects that the traditional statistical feature recognition method cannot distinguish, which can be added to the recognition system as an auxiliary method of the statistical feature identification method, and further improve the accuracy of recognition. ©, 2015, Chinese Machine Press. All right reserved.