基于神经网络的造纸过程定量水分系统预测函数控制

The predictive functional control for basis weight and moisture system in papermaking process based on neural network

针对造纸过程中的定量水分控制系统具有大时滞、大惯性、强耦合及多变量等特性,提出了一种新的多变量系统的预测函数控制方法。首先给出了PID神经网络解耦控制方法,将多变量系统变成单变量系统;然后根据预测函数控制原理对单变量系统设计控制器,并结合Smith预估思想,有效地克服了时滞对控制系统稳定性及控制品质的影响;最后通过仿真,证明该控制方法使系统具有在线计算量小、跟踪快速和精度高等特点,能够实现复杂多变量系统的有效控制。

Considering the features of basis weight and moisture control system in papermaking process,such as large time delay,big inertia,strong coupling and multivariable,it proposes a new predictive functional control method for multivariable systems. In order to change multivariable system into a single variable system,it presents PID neural network decoupling control method. Based on the basic principle of predictive functional control and these decoupled single variable systems,it designs a controller. Combining Smith prediction,this control can overcome the delay influence on stability and performance index effectively. The simulation results show that the proposed method with less calculation online,fast tracking and high control accuracy can realize effective control for complex multivariable systems.