摘要
本文提出了一种基于神经网络的显式自校正控制方案:用二级神经网络NNR(神经网络调节器)和NNI(神经网络辨识器)构成自校正控制器.NNR与受控对象构成一种反馈控制结构.NNI建立受控对象的模型,并预报对象的输出,以此为基础修改NNR的参数. 调节器NNR采用面向实时控制的,具有传统的PID控制机理的二层神经网络.其参数采用随机逼近法或梯度法实时修改.NNI采用单隐层神经网络,对受控对象进行实时辨识,其辨识算法为递推预报误差法(RPE).该算法克服了BP算法收敛速度慢、鲁棒性差的缺点,具有收敛速度快、预报精度高等优点.本文将RPE算法引入自校正控制,并对相关问题作了深入的研究.运用本文提出的控制方案,对一典型的工业过程进行了数字仿真研究.结果证实了本文所提出的基于神经网络的显式自校正控制在克服对象参数时变及较大的随机干扰方面的有效性.
A Neural Network based explict self-tuning control scheme is proposed in this paper. The self-tuning controller is composed of the NNR (neural network regulator) and the NNI (neural network identifier). A feedback control configuration is formed using the NNR. and the NNR's parameter is modified on the basis of the predicted output of the NNI which models the plant on line.
The NNR has the structure of 2-layered neural network and has a mechanism similar to that of PID controller. The NNI is made up of a network with one hidden layer. The RPE(recursive prediction error) algorithm is used to identify the plant on line and some related problems are probed. Numerical simulation results indicate that the proposed scheme has the ability to eliminate parameter perturbation and random disturbance effectively.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1997年第1期15-20,共6页
Pattern Recognition and Artificial Intelligence