Abstract: In order to overcome the problems of premature convergence frequently in Particle Swarm Optimization（PSO）, a new PSO is proposed. After the update of the particle velocity and position, two positions from set of the current personal best position are closed at random. A new position is produced by the quadratic interpolation given through three positions, i.e., global best position and two other positions. The current personal best position and the global best position are updated by comparing with the new position. Simulation experimental results of six classic benchmark functions indicate that the new algorithm greatly improves the searching efficiency and the convergence rate of PSO.

Key words: Particle Swarm Optimization, quadratic interpolation, convergence speech, global optimal

摘要： 为了克服粒子群优化算法容易早熟的问题，提出了一种新的粒子群优化算法。算法在进行速度和位置更新后，随机选取两个个体历史最好位置（不含全局最好位置）与全局最好位置，利用二次插值产生新的位置，并与当前个体历史最好位置相比较，更新当前个体历史最好位置和全局历史最好位置。对6个经典测试函数进行数值实验，结果表明该算法提高了算法的寻优能力和收敛速度。

关键词: 粒子群优化算法, 二次插值, 收敛速度, 全局最优