Research On Signal Extraction And Pattern Recognition Of PD Based On Virtual Instruments

Based on the analysis of domestic and foreign signal extraction and pattern recognition in large-scale generators insulation on-line monitoring, this paper developed a set of partial discharge data acquisition and identification system based on virtual instruments. This system uses the work mode of up-side machine and down-side machine, the up-side machine is used to signal acquisition and preprocessing, the down-side machine is used to extraction of features and identification of the type of discharge. In signal extraction, the upper and lower thresholds for two-ways approach is proposed based on wavelet thresholding denoising method, and is effective to reduce the energy of interference pulse caused by making and breaking of the generator SCR devices; in pattern recognition, through building two-dimensional and three-dimensional spectrum, six statistical operators and 320-dimensional three-dimensional tabulated data are extracted, and MATLAB software and LabVIEW calling DLL are used to achieve BP neural network classifier design, and practice shows the latter is more applicable to the actual works.