Dynamic neighborhood genetic learning particle swarm optimization for high-power-density electric propulsion motor

Abstract

To maximize the power density of the electric propulsion motor in aerospace application, this paper proposes a novel Dynamic Neighborhood Genetic Learning Particle Swarm Optimization (DNGL-PSO) for the motor design, which can deal with the insufficient population diversity and non-global optimal solution issues. The DNGL-PSO framework is composed of the dynamic neighborhood module and the particle update module. To improve the population diversity, the dynamic neighborhood strategy is first proposed, which combines the local neighborhood exemplar generation mechanism and the shuffling mechanism. The local neighborhood exemplar generation mechanism enlarges the search range of the algorithm in the solution space, thus obtaining high-quality exemplars. Meanwhile, when the global optimal solution cannot update its fitness value, the shuffling mechanism module is triggered to dynamically change the local neighborhood members. The roulette wheel selection operator is introduced into the shuffling mechanism to ensure that particles with larger fitness value are selected with a higher probability and remain in the local neighborhood. Then, the global learning based particle update approach is proposed, which can achieve a good balance between the expansion of the search range in the early stage and the acceleration of local convergence in the later stage. Finally, the optimization design of the electric propulsion motor is conducted to verify the effectiveness of the proposed DNGL-PSO. The simulation results show that the proposed DNGL-PSO has excellent adaptability, optimization efficiency and global optimization capability, while the optimized electric propulsion motor has a high power density of 5.207 kW/kg with the efficiency of 96.12%.