基于免疫分裂算子的粒子群优化算法

Particle swarm optimization based on immunity spallation operator

针对粒子群优化(particleswarmoptimization,PSO)易收敛于局部最优的缺点,提出了一种基于免疫分裂算子的PSO。该算法在初始化时,运用正交的思想,使得粒子分布均匀;在进化时,提出了一种基于速度与位置的亲和度,当粒子相似度满足要求时,才对粒子进行免疫变换,并且变异操作只针对性能较差的粒子。这样在保证粒子多样性的基础上减少了运算量提高了收敛速度。在Matlab环境下对Rosenbrock函数、Rastrigrin函数、Griewank函数3个多峰函数进行了仿真验证,实验结果表明,改进的PSO算法能够有效地达到全局最优。

Due to the premature convergence of PSO, PSO based on immunity spallation operator is proposed. Firstly, orthogonal principle is applied to the initialization of particle, so particles are equably distributed. During the evolution, the immunity operation is applied to the particles when they are under the threshold of sinailarity degree which is based on position and velocity. So it is not only avoiding the premature convergence but also having better convergence rate. And it is verified by Rosenbrock function, Rastrigrin function and Grievank function in MATLAB, the result shows that the improved PSO can realize global optimization effectively.