Extracting partial discharge signals from colored noises using chirplet transform

Detecting partial discharge (PD) signals is a widely used method in identifying the existence of incipient transformer faults. The validity of recognition is abated because of noises contained in fault signals. Chirplet transform is a type of parametric time-frequency (T-F) analysis. Rotating and shearing in the T-F plane, chirplet transform can provide skew analytical windows, which is different from wavelet transform. By adaptive chirplet decomposition and reconstruction, pure PD signals were primely approached releasing strong T-F coupling from noises. The results of simulation show that chirplet transform is more effective in eliminating colored noises than wavelet transform. So it provides more reliable basis for analyzing the state of operating transformer.