Abstract: A new hybrid particle swarm optimization combined with genetic algorithm was proposed in order to solve complex questions with high dimensions and overcome prematurity and the weak ability of local search. The global solution can not be found because of the bad results of some dimensions, and it is difficult to find all the best value in each dimension using the usual optimization algorithm. Enlightened by genetic algorithm, the improved algorithm can find the best position through evaluating each dimension and picking out the bad ones, adopting mutation and improving it during the process of evolution. Experimental results on several benchmark complex functions with high dimensions show that the algorithm can rapidly converge at high quality solutions.

Key words: genetic algorithm, premature convergence, particle swarm optimization, high dimension optimization problems, roulette wheel selection

摘要： 为解决高维复杂函数的优化问题，克服标准粒子群算法早熟收敛、局部搜索能力弱等缺点，在标准粒子群优化算法中融合了遗传算法的设计思想，提出了一种新颖的混合粒子群算法。高维函数个别维上的差解导致算法最终无法找到全局最优解，而通常的优化算法很难寻找到每一维上的最佳值。受遗传算法思想的启发，在粒子的进化过程中，通过对最优粒子的每一维进行评价，找到导致最终解质量差的维度，对其维上的数据进行变异，进而有针对性地改进，寻找到每一维上的最佳位置。对典型高维复杂函数的仿真表明：算法在求解质量和求解速度两方面都得到了好的结果。

关键词: 遗传算法, 早熟收敛, 粒子群算法, 高维优化问题, 轮盘赌