神经网络自适应逆控制的仿真研究

Simulation Study on Neural Network Based Adaptive Inverse Control

将神经网络引入逆控制，提出了一种基于神经网络的自适应逆控制。该控制结构主要有两个子神经网络组成，其中一个用于对系统进行辨识，另一个子网络实现对模型的逆作为控制器，从而构成自适应的逆控制。将其应用于热工系统中并进行大量仿真研究，结果表明针对不同的热工对象，该控制系统都能有效的克服扰动，适应环境及参数的变化，表现出良好的鲁棒性和控制精度。

This paper introduces an adaptive inverse neural network-based control architecture. There are two three-layer back-propagation neural networks in it, one is applied to identify the controlled object, other one approximates the plant inverse transfer function. The on-line training method using inverse control network is described in detail. Extensive simulation studies on large inertia objects with long time-delay in the power plants have shown that the controller performs very well and can easily be accomplished on-line.