神经网络控制器在未建模型的非线性系统中的应用

Application of Neural Network Controller in Unfounded Model Nonlinear System

应用神经网络对未建模型的非线性随机系统进行控制。采用Spall[1] 提出的同步扰动随机逼近的算法 ,通过系统的输出误差对神经网络控制器进行训练。因为系统模型未知 ,本文在Spall[1] 的基础上采用了隐层能自动增加的神经网络 ,通过实时的对神经网络的结构进行控制 ,可以实现对未建模型的非线性随机系统的最优控制。

Use Neural Network to control unfounded model nonlinear random system.Adopt the arithmetic of synchronization disturbance random approach presented by Spall to train the Neural Network Controller by the output error of system.Based on Spall,the Neural Network that concealed level can automatically increase is adopted to control the Neural Network structure and it can realize optimization control of unfounded nonlinear random system.