PD pattern recognition based on multi-fractal dimension in GIS

Abstract

This paper designs four types of gas insulated substation (GIS) defect models based on partial discharge (PD) characteristics and its defections. The GIS gray intensity images are constructed based on the mass specimens gathered by the ultra-high frequency and high-speed sampling systems. The multi-fractal dimension is founded on the box-counting dimension and multi-fractal theories. The GIS gray intensity images distillation methods, based on multi-fractal characteristics, is put forward. The box-counting dimension, multi-fractal dimension, and discharge centrobaric characteristics of the PD images are also extracted. The characteristic variables are then classified by the radial basis function (RBF) network. Identified results show that the methods can effectively elevate the discrimination of the four types of defects in GIS.

Keywords

GIS / PD / box-counting dimension / multi-fractal / pattern recognition