论文部分内容阅读
多变量预测控制在应用中经常会遇到模型失配的问题,最终导致控制器不能满足控制要求.本文提出了一种模型预测控制(model predictive control,MPC)架构,通过被控对象和预测模型的频率响应误差判断模型是否失配;当模型失配时,首先对被控对象叠加持续激励信号;然后,通过改进的模型自适应辨识方法辨识对象的传递函数模型;最后,经过拉氏逆变换,将传递函数模型转化为FSR(finite step response)模型,重新恢复多变量预测控制.该方法不需要进行离线辨识试验,实现了模型的多变量辨识;辨识的传递函数模型的动态特性更加清晰,便于分析和修改;经过拉氏逆变换得到的FSR模型更加平滑,能够消除因模型误差引起的静差.经过仿真实验,证明了该方法的有效性.
Multivariable predictive control often encounters the problem of model mismatch in application, which eventually leads to the controller can not meet the control requirements.In this paper, we propose a model predictive control (MPC) architecture, through the controlled object and the predictive model When the model is mismatched, a continuous excitation signal is first superimposed on the controlled object. Then, the transfer function model of the object is identified through an improved model adaptive identification method. Finally, after the inverse Laplace transform , The transfer function model is transformed into a finite step response (FSR) model and the multivariable predictive control is reinstated.This method does not require off-line identification experiments to achieve multivariate identification of the model, the dynamic characteristics of the identified transfer function model are clearer, Which can be easily analyzed and modified.The FSR model obtained after inverse Laplace transform is smoother and can eliminate the static error caused by the model error.The simulation results show that the proposed method is effective.