Feature extraction of UHF PD signals by wavelet packet based MRSD analysis For defect identification in Gas Insulated Systems

In condition monitoring of Gas Insulated Systems (GIS) the upcoming failure modes can be diagnosed well ahead by analyzing the Partial Discharge (PD) activity. In recent times, Ultra High Frequency (UHF) technique is recognised as effective tool for online PD analysis. In this paper, a simple and effective wavelet packet based multi-resolution signal decomposition feature extraction scheme is proposed to estimate the spectral energy distribution of UHF signal for PD source location. Further, the relationship between critical frequency components of UHF PD signal generated by different type of defects is investigated. The optimized model is ascertained for common onsite defects called protrusion on electrode, freely moving metallic particle, surface discharge in epoxy spacer, floating particle sitting on the epoxy spacer of compressed SF6 GIS under HVAC. The four defects are modeled in laboratory environment and defect identification with the proposed technique is verified with the minimum success rate of 95% irrespective of physical conditions like, gas pressure, PD severity or defect location.