An improved NSGA-III integrating a good point set for multi-objective optimal design of power inductors

This work presents a multi-objective optimization approach for the optimal design of power inductors with consideration of multiple objective functions and complex constraints. To address this problem, we propose an improved version of NSGA-III (I-NSGA-III), in which a good point set, rather than the original random mechanism, is used to generate a uniformly distributed initial population. Meanwhile, based on the structural feature of the pre-designed inductor, a target model of temperature rise is established by Newton’s law of cooling to ensure calculation precision and improve efficiency. In addition, we integrated a fuzzy membership technique to identify the best compromise solution (BCS) from the generated Pareto-optimal solutions (PSs). The proposed approach is validated using a practical pre-designed power inductor, and the results demonstrate its effectiveness and superior performance compared to classical NSGA-III and MOEA/D.