Accelerating Multi-Objective Optimization for Next-Generation Magnetic Devices

Magnetic components serve as the fundamental elements in power electronic converters, usually representing the bottleneck of multi-objective optimization, due to their extensive design space and challenges in the modeling process. This paper proposes a new interpolation-based optimization algorithm, which significantly reduces the number of designs that need to be evaluated when searching for the optimal design(s), therefore increasing the speed significantly. Two variants of this algorithm, one for defined goal functions with weightings on different objectives, and the other for Pareto front, are illustrated in detail. They are then compared with brute-force sweeping using a design example, which shows that with the proposed algorithm, the same accuracy can be reached with dramatically decreased time needed for the whole optimization process.