本文讨论使用神经网络(NN)控制一未知的动态过程以得到期望的闭环特性。过程的逆动态由用于反馈控制的神经网络进行学习和描述,即使用一个三层网络通过调整各层间的权值来学习逆动态过程。网络的学习方法基于使过程输入的实际值与估计值,过程的输入值与期望值之间的加权误差为最小。逆动态的学习中使用误差的反向传播过程和一种推广的δ律。 在这项工作中,神经网络用于实现自适应控制,以达到可变过程条件下特定的闭环回路响应。其目的是为现存的自调节控制提供一种神经网络控制方案。在学习过程中,要对相应的权值进行调整,使误差函数在梯度减小方向上减为最小。要实现对闭环回路期望值(设定值)的自动跟踪,有许多可供选择的方案,包括串联控制回路。采用分布控制实时实现。本文叙述并讨论了设定值跟踪的结果。 This article discusses the use of neural networks (NN) to control an unknown dynamic process to achieve the desired closed-loop behavior. The inverse process of the process is learned and described by the neural network used for feedback control, which uses a three-layer network to learn the inverse dynamic process by adjusting the weights between layers. The network learning method is based on minimizing the weighted error between the input value and the expected value of the process based on the actual value and the estimated value input to the process. Inverse dynamic learning uses the error backpropagation process and a generalized law of δ. In this work, neural networks are used to implement adaptive control to achieve a specific closed-loop response under variable process conditions. Its purpose is to provide a neural network control scheme for the existing self-regulating control. In the learning process, the corresponding weights should be adjusted so that the error function is minimized in the gradient decreasing direction. There are many options to automate the tracking of desired values ​​(setpoints) of a closed-loop, including a series control loop. Distributed control real-time implementation. This article describes and discusses the results of setpoint tracking.