基于递归神经网络的大滞后非线性系统预测控制

The predictive control of a nonlinear system with time-delaybased on recurrent neural networks

针对带有大滞后的非线性系统,提出了在迭代多步预测的基础上,将系统多步预测输出值进行线性化,在多步预测目标函数下实现系统控制的方法。采用适合于动态系统实时控制的扩展Elman网络,利用训练速度快的阻尼最小二乘法学习网络权值。仿真实验表明了该方法的有效性。

This paper puts forward a novel control strategy for a nonlinear process with pure timedelay.Based on the recursive multistep predictive method,the actual multistep predictive values of the system are linearized.A novel control method is obtained by a minimized multistep predictive target function.In the control process,the extended Elman neural networks are used and the weights of the network are trained by the damped least square method.Simulation experiments show the effectiveness and good performace.