一个与powell算法相结合的改进微粒群算法(格式修改完2).docVIP

摘 要 本文提出了一个Powell方法和改进微粒群算法相结合的混合算法。Powell搜索法是由Powell于1964年首先提出的解无约束最优化问题的一种直接搜索法。它的优点是：计算简单、具有快速的收敛性且不需要计算导数，但却不能保证收敛到全局最优解，容易陷入局部最优。微粒群算法(particle swarm optimization, PSO)是一种基于群体智慧的优化算法，在寻找全局最优解时，不需要局部信息，并且很少陷入到局部最优，但缺点在于局部搜索能力较弱，收敛速度较慢，计算成本相对较高。改进PSO算法重点在于探索即对于全局的开发；而Powell搜索法重点在于局部的开发以便达到更加精细的结果。因此，本文将改进PSO算法与局部寻优能力极强的Powell算法相结合，提出了一个混合算法。对实验用例的测试表明所提出的混合算法收敛速度大大加快，收敛精度高，能有效地降低计算成本，且只需要很少的迭代次数，优于PSO算法、GCPSO算法和NM-PSO算法。 关键词：Powell直接搜索法；改进微粒群算法；Powell-PSO; 混合算法 Abstract In this paper we propose a hybrid algorithm combined Powell algorithm with improved particle swarm optimization. Powell search method is a direct search method，which was first proposed in 1964 by Powell for unconstrained optimization problems. Its advantages are: the calculation is simple, fast convergence rate and not requiring to compute the derivative, while can not guaranteed to converge to the global best solution and is very easy to be trapped in the local best. However, the particle swarm optimization (PSO), based on the colony wisdom, is an optimizing algorithm, which does not need local information and is less likely to be trapped in local best when searching for the global optimizing solution. But its disadvantage is in that the capacity is weak in the local search，the convergence speed is slow and the computational cost is comparatively high. PSO algorithm emphasizes on exploration, namely, exploring on the global, while the Powell search method emphasizes on the exploration of local so as to attain more subtle results. Therefore, in this paper we combine the PSO algorithm with the Powell algorithm which has powerful local searching capacity and propose a hybrid algorithm. The test on the experiment cases indicates that the proposed algorithm has much faster convergence speed, higher convergence accuracy, and can reduce the computational cost effectively and need only a few iterations. It is better than the PSO algorithm, GCPSO algorithm and NM-PSO algorithm. Keywords: Powell direct se