Some results on a SMES device optimization benchmark problem

Abstract

The scope of this paper is to throw some light on the difficulties occurring when optimizing in a discrete parameter space. In the first section of this paper some general considerations on the optimization of a SMES device in a discrete space are made. In the second part a pattern search technique as a representative for deterministic, an evolution strategy and a simulated annealing algorithm as members of the class of stochastic methods are presented and used to optimize the problem. The effect of varying the discretization is investigated for the pattern search method. In the next section some features of the benchmark problem are introduced and finally the optimization results are given.