Partial discharge fault identification of switchgear based on frequency constrained independent component analysis

In the switchgear partial discharge (PD) detection procedure, in order to improve the effectiveness of the fault identification under the conditions of the narrowband interference and other kinds of noise, which were difficult to eliminate, we developed a new PD identification method. PD mixed signals were acquired by radio frequency current transformers and high-speed data acquisition system. Then, the effective PD signals were separated from the mixed signals by frequency constrained independent component analysis (FCICA). As the influence of other signals' aliasing was overcame, the signal-to-noise ratio was greatly improved. We used the separated signals' power spectra at the two frequency bands of 0.4~1.1 MHz and 1.8~2.7 MHz to construct characteristic vectors of fault identification, by which PD fault was identified by least squares support vector machine (LSSVM).The validity of the developed methodology was verified through the PD experimental platform. Results show that the switchgear PD fault identification rate is up to 98.0% based on the developed method which has good ability of generalization. ©, 2014, Science Press. All right reserved.