频域滤波模型的粒子群优化算法

摘要

当前对于粒子群优化算法(简称基本PSO)的改进主要从控制参数与数学模型入手,但这可能导致陷入局部最小值.针对这个问题,提出一种基于频域滤波模型的PSO算法(简称FPSO).FPSO是对粒子种群多样性进行定量分析,当粒子集中度低于设定阈值时,以当前最优粒子为中心,在一定半径范围内进行傅里叶变换,通过预设的低通滤波器,削弱当前找到的最优值;然后对当前粒子群施加以最优粒子为势能中心的辐射力,所有粒子在滤波范围外部的空间以较大的速度继续搜索.实验分析表明:基于频域滤波模型的PSO算法提升了种群多样性,有效地提高了全局搜索能力,在求解多峰函数问题时解的精度优于带电PSO算法与变异PSO算法.%A new PSO algorithm based on the FFT model (referred to as FPSO algorithm) was proposed to against the problem that the basic PSO algorithm in solving complex multimodal problems is easy to fall into local optimal solution.The supervision conditions of population diversity were added to the basic PSO algorithm that the process of filtering in frequency field was triggered when the population down to a given threshold value.Particle that within the certain radius was conducted Fourier transform.Using the Gaussian low pass filter which have the minimum dispersion weakened the currently found extremism.The particle have suffered from the force of radiation function and then reentered the other space to search.Compared with the CPSO and GA-PSO,the results indicates that the FPSO algorithm has a higher degree of diversification and better performance to solve multi-modal optimization problem.