AGC系统的神经网络自适应控制

An Adaptive Neural Network AGC System

针对板带轧制过程厚度自动控制（ＡＧＣ）问题，提出逆动态自适应神经网络ＡＧＣ（ＶＡＮＮ－ＡＧＣ）系统，并给出适于过程控制的网络激励函数和改进的ＢＰ网络训练方法。仿真结果表明：ＶＡＮＮ－ＡＧＣ系统可自适应地实现快速跟随响应和抗扰调节；新的网络激励函数和训练方法适用于过程控制的需要；

This paper presents an inverse dynamic neural network AGC system, puts forward a network activation function suitable for control, and improves the training method of backpropagation (BP) applying to dynamic process control. The simulation results show that the VANN-AGC system can adaptively realize fast following response and regulating against any disturbance. It is proved that the new network activation function and the improved BP training method are suitable for process control. The stable error of the system is limited in ±5μm.