Study On The Denoising Algorithm Of Partial Discharge

Partial discharge (PD) is generally caused due to the internal or insulation insulatorsurface local electric field focus particularly, usually this discharge is a nonstationaryaperiodic signal with the duration less than1ps,following the sound, light, heat and chemicalreactions or other phenomenen,the frequency ranges from tens of kiloshertz to hundreds ofmegahertz, the PD detection systems are often vulnerable to strong external interferences,andsometimes the PD signals are submerged in noises(white noise for example) completely.Sothe signals acquired must be preprocessed to obtain the reliable PD.information,The paperpresent the wavelet transform(WT) analysis and the empirical mode decomposition(EMD)analysis to fulfil automatic noises reduction: By selecting an optimal wavelet using minimumdescription length (MDL)criterion and doing signal’s WT,processing the high-frequencycomponent using the spatial correlation based on the differences of the wavelet coefficients indifferent resolutions,PD signals are then reconstructed with modified detailedcoefficients;And according to the characteristics of PD,this paper has also present anothermethod called EMD-MDL using EMD and the MDL criterion to extract PD signals fromexcessive noises, firstly PD signals are decomposed into Intrinsic Mode Functions (IMF) byEMD,and then have the IMFs threshold processing to put forward part of the high frequencynoise interference,then obtain the key IMFs by MDL criterion,finally the key IMFs arereconstructed to achieve de-noised PD,the experimental results illustrate that the methodpresented in this paper are efficient and feasible and outperforms other general method of PDnoise reduction.