一种基于非线性系统的动态感知系数的自适应粒子群优化算法
A Dynamic Perception Coefficients Self-tuning Particle Swarm Optimization Algorithm Based on Nonlinear Systems
Abstract
在优化非线性复杂系统问题中,智能算法已成为一种重要手段.提出了一种基于动态感知系数的自适应粒子群优化(particle swarm; optimization,PSO)算法(self-tuning; PSO,SPSO),将PSO算法的感知系数与神经网络算法结合,并于在线学习训练过程中动态调整感知系数,改善了PSO算法的计算效率以及全局收敛效率; .进一步将2个相互关联的神经网络比例积分微分(proportion integration; differentiation,PID)神经网络及SPSO神经网络结合起来,使其能有效解决非线性控制模型的问题.为了验证该算法,引入了4个仿真例; 及2种PSO优化算法传统PSO(conventional PSO,CPSO)和修正PSO(modified; PSO,MPSO),来比较SPSO算法在解决控制问题中的非线性复杂系统的高效性,结果显示SPSO算法有较好的全局收敛性能、收敛速度以及较强的鲁棒; 性. We propose a dynamic perception-coefficients algorithm,called; self-tuning particle swarm optimization (SPSO) in this paper, which is; used for determining optimal nonlinear systems.The theory of SPSO; integrates conventional PSO (CPSO) with neural network (NN) and trains; perception coefficients with the gradient descent algorithm to improve; computational efficiency,rate of convergence and global convergence.Then; we combine proportional-integral-derivative (PID) NN with SPSO to; optimize complex nonlinear systems.This paper presents four examples and; another two optimization algorithms CPSO modified PSO(MPSO) to help; estimate performances of the proposed SPSO.The result demonstrates that; the SPSO exhibits great performances in global convergence, rate of; convergence and strong robustness while optimizing complex nonlinear; systems.