基于粒子群优化算法的非线性系统辨识

Nonlinear system identification based on particle swarm optimization

针对连续非线性系统中单输入单输出Hammerstein模型,由于传统辨识方法对Hammerstein模型中非线性部分具有不易辨识的缺陷,造成辨识精度低,辨识效果差等问题。为此,采用粒子群优化算法对非线性系统进行辨识的方法,将参数辨识问题转换为参数空间上的函数优化问题。为了进一步增强粒子群优化算法的辨识性能,提出采用迭代粒子群对整个参数空间进行搜索得到系统参数的最优估计。通过MATLAB软件进行仿真,仿真实验结果证明:该方法收敛速度较快,辨识得到的参数精度较高。该方法与最小二乘方法相比,计算量小,过程简单,不用计算多重积分,辨识速度较快,辨识精度高。

For the Hammerstein model with single input and single output in the continuous nonlinear system,the traditional identification method has the defect of being difficult to identify the nonlinear part of the Hammerstein model,resulting in low identification accuracy and poor identification effect.Therefore,the problem of parameter identification is transformed into the problem of function optimization in parameter space by using particle swarm optimization to identify nonlinear systems.In order to further enhance the identification performance of PSO,an iterative PSO is proposed to search the whole parameter space to obtain the optimal estimation of system parameters.MATLAB software is used for simulation,and the simulation results prove that the method has a fast convergence speed and high parameter accuracy.Compared with the least square method,this method has the advantages of small computation amount,simple process,no need to calculate multiple integrals,fast identification speed and high identification accuracy.