Optimal design of superconducting generator using genetic algorithm and simulated annealing

In the 12-year Japanese national project (the so-called Super-GM), R&D on a 70 MW class of superconducting generator model has been successfully finished as the first stage of verifying electrical features in the electric power system and to propose future projects. However, it has been known that its design method was carried out by trial and error. Hence, based on some design parameters of the Super-GM model-A machine, optimal designs of the superconducting generator (SCG) using a genetic algorithm and simulated annealing have been individually carried out for the purpose of improving its energy efficiency and/or specific power density. The results of optimal design by two such approaches as well as multiobjective optimal design by a min-max approach are compared. In addition, the influence of some machine parameters on performance of the SCG is evaluated. To optimise the energy efficiency and specific power density, its loss and volume are defined as objective functions, respectively, subject to some electrical and mechanical constraints. In the multiobjective optimal design, the min-max approach is utilised to find the best compromise solution between the optima of loss and volume. It is clarified that the design approaches developed are effective and reasonable to optimise the energy efficiency and specific power density of the SCG, referring to design parameters of the Super-GM model-A machine.