摘要
提出了一种基于对角回归神经网络的PID控制器结构,给出了PID参数在线自整定的学习控制算法。为检验控制效果同时还使用了静态BP网络来整定PID参数,并在Matlab环境下,分别建立了基于对角回归神经网络和BP网络的液位实时控制系统。实际的控制效果说明,基于动态网络的PID控制器工作稳定,具有较好的鲁棒性。
A new type of adaptive PID controller using diagonal recurrent neural network (DRNN)is presented. An on-line learning algorithm based on PID parameter self-tuning method is given. In order to verify the performance of the proposed approach, a control method that PID parameters are automatically adjusted by back-propagation (BP) algorithm is also introduced. Two real-time level control systems are devised on the basis of DRNN and BP networks using Matlab. The experimental results indicate that the PID controller based on dynamic neural network possesses satisfactory stability and robustness.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2004年第6期777-778,810,共3页
Systems Engineering and Electronics
基金
国家自然科学基金(60274020
69974017)
河北省自然科学基金(602621)
广西省自然科学基金(0135065)资助课题
