基于改进粒子群算法的Volterra模型参数辨识

Parametric identification of nonlinear volterra model based on improved PSO algorithm

针对非线性系统Volterra泛函级数模型,结合混沌优化策略和种群多样性控制思想,提出了一种改进粒子群算法,并应用于Volterra模型参数的辨识,将非线性系统的辨识问题转化为高维参数空间上的优化问题。利用混沌序列增加初始种群的多样性,通过构建动态子群以进行协作寻优,且各子群采用不同的参数自适应调整策略,并定义算法收敛性测度以对精英粒子进行合理的混沌变异,避免了算法早熟收敛,提高了算法的寻优速度和寻优精度。仿真实验中,将该方法与基于标准粒子群算法、遗传算法、量子粒子群算法的Volterra模型参数辨识方法相比较,验证了该辨识方法的有效性和鲁棒性。

By combining the particle swarm optimization （PSO）with the chaotic optimization strategy and the control idea of population diversity,an improved particle swarm optimization （IPSO） algorithm was proposed for parametric identification of nonlinear Volterra series model.The basic idea of the method was that the problem of nonlinear system identification was converted into an optimization problem in a high-dimensional parameter space.The chaotic optimization strategy was employed to increase the diversity of the initial population.By building dynamic subgroups,the optimization was realized through the collaboration of sub groups.The different adaptive adjustment strategies for control parameters of IPSO were used in subgroups.The convergence measure of IPSO was defined to perform the chaotic mutation operation reasonably.So the premature convergence was avoided,the speed and accuracy of IPSO were improved.In simulation tests,the proposed method was compared with Volterra model identification methods based on standard PSO algorithm,the genetic algorithm and the quantum-behaved PSO,respectively.Its effectiveness and robustness were verified.