Study On Feature Extraction And Mode Recognition Of Partial Discharge Signal In XLPE Cable

Cross-linked polyethylene (XLPE) cable has gradually become the preferred product of power cables in China, for its superior electrical performance and mechanical and physical properties. In order to evaluate insulation condition of XLPE cable, this dissertation chose electromagnetic coupling method to detect its partial discharge signal. First, based on the analysis of the influence between the Rogowski coil current sensor frequency characteristic and its parameters, a suitable current sensor is designed. Then, according to the differences between PD signals and white noise and narrow construct, the characteristic quantity is constructed for de-noising and its denoising capability is proved. Finally, the dissertation establish a multi-class support vector machine classifier, construct a six-dimensional support vector space from statistical operators of partial discharge spectra, and compile LIBSVM package using Liabview.