多变量系统的神经网络解耦新方法

New Decoupling Method Based on Neural Network for Multivariable System

利用前馈补偿的原理，设计了两种多变量系统的神经网络解耦方法·一种利用神经网络实现前馈补偿，使补偿以后的系统实现解耦，且解耦单变量系统具有原对象主通道的特性·第二种方法将解耦和神经网络逆动态控制结合起来。

Two new neural network decoupling methods for multivariable system were presented by using the principle of feedforward compensation. One is that neural network is used to realize feedforward compensation and to make the compensated system decoupling. The decoupling single variable system has the characteristics of main channel of the plant. The other is that decoupling is combined with neural network inverse control. The output of the plant tracks the variation of the input. The two schemes are also suitable for multivariable nonlinear systems.