Feature parameters extraction of gis partial discharge signal with multifractal detrended fluctuation analysis

Ultra-high frequency (UHF) method is widely used in gas-insulated switchgear (GIS) partial discharge (PD) online monitoring because this technique has excellent anti-interference ability and high sensitivity. GIS PD pattern recognition is based on effective features acquired from UHF PD signals. Therefore, this paper proposes a new feature extraction method that is based on multifractal detrended fluctuation analysis (MFDFA). UHF PD signals of four typical GIS discharge models that were collected in a laboratory were analyzed. In addition, the multifractal feature of these signals was investigated. The single-scale shortcoming of traditional detrended fluctuation analysis and its sensitivity to interference information trends were overcame. Thus, the proposed method was able to effectively characterized the multi-scaling behavior and nonlinear characteristics of UHF PD signals. With the use of the shape and distribution difference of the multifractal spectrum, seven feature parameters with clear physical meanings were extracted as feature quantity for pattern recognition and input to the support vector machine for classification. Results showed that the feature extraction method based on MFDFA could effectively identify four kinds of insulation defects even with strong background noise. The overall average recognition rate exceeded 90%, which is significantly better than that of wavelet packet-based feature extraction.