基于广义预测控制的对角递归神经网络控制器

Diagonal Recurrent Neural Networks Controller Based on Generalized Predictive Control

提出一种基于广义预测控制的对角递归神经网络控制器，并给出该神经网络控制器和辨识器的学习率范围．仿真实验表明，所采用的神经网络控制结构适合于带纯时延的未知的非线性被控对象的广义预测控制，同时能有效地改善神经网络学习的收敛性．

Diagonal recurrent neurocontroller based on generalized predictive control is proposed in this paper. The bounds on learning rates for the weights of the neural networks are obtained. The simulation results show that the neural networks control architecture used is capable of generalized predictive control for unknown nonlinear plants with dead time, and can effectivelly improve convergence behavior in training neurocontroller.