Abstract: Using Particle Swarm Optimization（PSO） to handle complex functions with high-dimension has the problems of low convergence speed and premature convergence. This paper proposes a hybrid particle swarm optimization. It adopts prematurity judge mechanism by the variance of the population’s fitness and puts gene conversion and mutation operator into algorithm. It constructs a new individual and individual gene fitness function, and will adapt to the worst gene mutation value. To reduce the computation of the proposed algorithm, it uses quantum dissipative particle swarm algorithm structure. Experimental results show that compared with particle swarm algorithm which has only one fitness value, it has faster convergence rate. Especially the hybrid particle swarm optimization is of strong ability to avoid being trapped in local minima, and performances are fairly superior to single method.

Key words: Particle Swarm Optimization, multi-strategy mechanism, prematurity machanism

摘要： 针对粒子群优化算法在处理高维复杂函数时存在收敛速度慢、易陷入早熟收敛等缺点，提出了混合粒子群优化算法。它借鉴群体位置方差的早熟判断机制，把基因换位和变异算子引入到算法中，构造出新的个体和个体基因的适应值函数，将适应值最差的基因进行变异。为减少算法计算量，采用耗散的粒子群算法结构。实验表明，该算法比只有一个适应值的粒子群算法具有更快的收敛速度。且具有很强的避免局部极小能力，其性能远远优于单一优化方法。

关键词: 粒子群优化算法, 多策略机制, 早熟机制