摘要
本文提出了一种基于BP神经网络的非线性广义预测学习控制器，它由一个BP网络构成．在整个学习与控制过程中，首先根据被控对象的输出与BP网络的学习输出之间的误差来修改网络的权值，以逐步建立被控对象的合理的多步预报模型;然后，根据网络的多步预报输出序列与设定值序列的偏差修改控制律．学习过程与控制过程交替进行．仿真结果证实了该控制器的有效性，为实现非线性系统的控制提供了一条可行途径．
In this paper, a BPN - based nonlinear generalized predictive learning controller is introduced. It is made up of a single BP neural network. During the whole learning and controlling process, in each sample period, we at first revise the values of the weights of the network to create a reasonable multistep output prediction model by the error between the BPN's productions and the plant's outputs, then the control law is revised by the error between the outputs sequence of the multistep prediction neural network model and the setpoint sequence of the system. The simulation experiments show the validity of this controller. This method of designing nonlinear controller gives an available way to synthesis the controller of nonlinear system.
出处
《南开大学学报（自然科学版）》
CAS
CSCD
北大核心
1997年第4期52-56,共5页
Acta Scientiarum Naturalium Universitatis Nankaiensis
基金
天津市青年科学基金
关键词
非线性控制
BP网络
广义预测控制
学习控制
nonlinear control
BP neural network
generalized predictive controll learning control
