Noise reduction on partial discharge data with wavelet analysis and appropriate thresholding

One of the major challenges of partial discharge (PD) measurements is separation of PD signals from different type of noises resulting from measurement circuit or surrounding environment. White noise is the most common noise component that couples the PD data in practice. De-noising with wavelet shrinkage method is capable of separating the noise component to some extent, but thresholding is a key factor which is directly related to distortion of PD waveform and quality of de-noising process. Although there are applications in literature for PD noise separation, selection of thresholding rule and thresholding function, which affects the evaluation of PD characteristics, is still challenging. In this paper, by using the simulative noisy PD pulses, different thresholding rules and thresholding functions with various combinations have been studied in order to achieve the best thresholding scheme by way of observing the Signal to Noise Ratio (SNR), Mean Square Error (MSE) and Magnitude Error (ME). Results show that level dependent thresholding rule together with combined thresholding function gives the best result for de-noising white noise coupled to the PD signal.