非线性系统多步预测控制的复合神经网络实现

提出一种基于神经网络的非线性多步预测控制，采用由线性网络和动态递归神经网络构成的复合神经网络。在此基础上将线性系统的广义预测控制器扩展为非线性系统的多步预测控制器。通过对非线性过程ＣＳＴＲ的仿真表明，该方法的稳定性和鲁棒性明显优于线性ＤＭＣ预测控制。

Nonlinear Multi-step Predictive Control= Using Compound Neural Networks

A new method for nonlinear multi-step predictive control based on neural networks has been carried out. The neural network is a kind of compound one which includes linear part and nonlinear part, the nonlinear part is realized by using dynamic recurrent neural network. Based on the above compound neural network, generalize predictive controller for linear model is extended to nonlinear multi-step predictive controller. For a nonlinear CSTR process, the simulation result show that the obvious improvement of the suggest method over linear DMC method both in stableness and robustness.