基于正交试验设计的粒子群优化算法

A PSO algorithm based on orthogonal experimental design

针对粒子群优化(particle swarmopti mization,PSO)算法在进化初期收敛速度快但容易陷入局部最优、在进化后期收敛速度变慢且精度低的缺陷,为了提高粒子群算法的收敛速度和全局寻优能力,提出了基于正交试验设计的粒子群优化(orthogonal-experi mental-design-based PSO)算法.在基本粒子群算法的基础上,算法OE-PSO对当前搜索到的解进行局部寻优,利用正交试验设计对搜索空间的分布均匀性在可行解的领域选择有代表性的解进行测试.算法OE-PSO用搜索到的更好的解在下一次迭代中引导粒子进行搜索,从而获得更快的收敛速度和更精确的解,同时避免局部最优.实验结果表明,算法OE-PSO不但具有较快的收敛速度,而且能够有效提高解的精确性,增强算法的鲁棒性.

In the early stage of evolutional process of particle swarm optimization（PSO）,it converges very fast but trends to a local optimum.But in the later stage,its convergence speed slows down which causes low accuracy of solution.In order to accelerate the convergence speed of PSO and improve its ability of global optimization,this paper advances a particle swarm optimization algorithm OE-PSO based on orthogonal experimental technique.Based on the classical PSO,OE-PSO searches for local optimum in the neighbor area of the obtained solutions by using the method of orthogonal experimental design,OE-PSO can probe the solutions uniformly distributed in the search space and select the better ones.Using those better solutions OE-PSO can guide the particles searching towards the correct direction in the latter iterations so as to speed up the convergence,get more precise solutions and avoid local optimum.The experimental results show that OE-PSO algorithm not only has faster convergence speed,but also can improve the accuracy of solutions effectively and enhance the robustness of the algorithm.