基于一类混合PSO算法的函数优化与模型降阶研究

为了克服传统微粒群优化(PSO)算法容易早熟收敛和陷入局部极小的缺点,通过对PSO算法特点和行为的分析,提出一类有机结合模拟退火(SA)算法和PSO算法的混合算法.混合算法不仅利用PSO的机制进行群体全局搜索,而且利用模拟退火的思想恰当地选择微粒的最好历史位置,保障了群体多样性,并有效平衡了算法的探索和趋化能力,进而改善了算法的优化性能.基于典型复杂函数优化问题和模型降阶问题的仿真结果表明,所提混合算法具有很好的优化质量、搜索效率和鲁棒性.

Study on Function Optimization and Model Reduction Based on a Class of Hybrid PSO Algorithm

To overcome the weaknesses,such as easy to be prematurely convergent and be trapped in local optima for classic particle swarm optimization(PSO) algorithms,a class of hybrid algorithm is proposed by analyzing the features and behaviors of PSO and by reasonably combining simulated annealing(SA) and PSO.By applying PSO to perform population-based global search and by utilizing the idea of SA to suitably select the best historic positions for particles,the optimization performances of the hybrid algorithm can be improved due to the maintenance of swarm diversity and the balance of exploration and exploitation.The simulation results of typical complex function optimization problems and model reduction problem show that the proposed hybrid algorithm has good optimization quality,searching efficiency and robustness.