神经网络控制系统的仿真研究

A Study on Computer Simulation of Neural Networks for Control Systems

本文提出了一种基于神经网络预测控制新方法，并叙述了将其用于非性线系统控制的稳定性。计算机仿真结果表明，基于神经网络预测控制新方法响应速度快，鲁棒性好，可实现无静态余差，控制算法采用多值预估控制等价于单值预估控制的思想，使程序大为精练，占用内存少且计算速度快，可望能在非线性控制系统中在线实现。

in this paper a new predictive control method based on neural networks is proposed,and its stability properties are described while dealing with nonlinear control systems. The results from the computer simulation indicate that its response speed and robustness are excellent and it can lead to no static error. In terms of the algorithm,the idea that the multivalued predictive control is equivalent to the singlevalued one is adopted,resulting in succinct programme,less memories and faster running speed. It is expected that the new method can be on-line realized in nonlinear control systems.