A novel approach for reduction of electric field stress in vacuum interrupter chamber using advanced soft computing algorithms

This investigation presents a new algorithm for optimized internal shape design of high voltage equipments like a vacuum interrupter (VI) chamber, regarding to electric field intensity distribution pattern. At first a commercial vacuum chamber is modeled using Solid-Works and Comsol-Multiphysics softwares. The model design parameters are changed randomly to generate a new structure which is analyzed using finite element (FE) method to obtain 8 regions with maximum electric field intensity. This loop continues some times to arrange a data bank from input parameters vector and coupled output matrix for 8 internal electric field intensity components. The next stage is to train a suitable general regression neural network (GRNN) for every electric field output, separately, to form an objective functions with 8 objective components. Regarding the constrains and inequality conditions specified, a multiobjective genetic algorithm (GA) is arranged to run. A set of non-dominated points, called Pareto Front, are specified by this optimization algorithm which are edited and clustered in five sets using k-mean statistical function. In every set the point with smallest distance from the center point of that set is elite for final modeling and validation check. So the final optimized point is specified after F.E. modeling. This algorithm is studied in the case that the main shield is considered float. In continue we analyze the electric field reduction percentage achieved, which causes lower electric field stress on the insulators especially on the contact surfaces region. With respect to Fowler-Nordheim (FN) equation lower electric field causes lower field emission currents especially at the insulation recovery interval which increases the endurable transient recovery voltage (TRV) level of the chamber. Furthermore and in the future works the maximum size reduction of a vacuum interrupter chamber, or other equipment, can be performed with small change in the introduced algorithm.