Multiobjective optimisation by self-adapting Pareto genetic algorithms for electrical system design

In this paper, Pareto Genetic Algorithms are applied to solve multiobjective optimisation problems. In particular, a recent version of the nondominated sorting genetic algorithm (NSGA-II) is presented. A self-adaptive recombination scheme is used for crossover operators to improve the algorithm efficiency. Tests on mathematical functions of various difficulties are carried out to show the robustness of self adaptation. Finally, the self-adaptive NSGA-II is applied to the optimal design of an electrical system based on a inverter - permanent magnet motor - reducer - load association. It allows to reduce the global losses and weight in the system and help the designer to understand couplings and interactions between design variables in relation to technological constraints and objectives