Optimization of Grading Ring for Resin Impregnated Paper UHVDC Wall Bushing Using Neural Network Method

Abstract

In this paper, a method using neural network for optimizing the grading ring of ±1100kV ultra-high voltage direct current (UHVDC) wall bushing is presented. Firstly, the finite element method (FEM) is applied to calculate the electric field distribution along hollow insulator surface with various pipe diameter, ring diameter and installation position of the grading ring and the optimal goal is set according to the FEM numerical results. Then the neural network model is built and trained with L- M algorithm using 300 sets of data calculated by the method of parametric scanning. Finally, the parameters of grading ring are optimized according to the neural network fitting results. The optimized grading ring uniforms the electric field distribution along the hollow insulator surface. This paper can provide a reference on structural design of UHVDC wall bushing.