纸浆浓度控制系统的仿真研究  

Simulation and Research on Pulp Consistency Control System

作　　者：曹露[1] 熊智新[1] 胡慕伊[1] 

机构地区：[1]南京林业大学江苏省制浆造纸科学与技术重点实验室,江苏南京210037

出　　处：《计算机仿真》2012年第6期176-179,183,共5页Computer Simulation

基　　金：江苏省制浆造纸科学与技术重点实验室开放基金项目(200909)

摘　　要：研究纸浆浓度控制问题,针对纸浆浓度控制系统存在的大滞后、非线性和时变性等特点,常规的PID控制器很难达到理想的控制效果。为了改善纸浆浓度控制系统性能,提出了BP神经网络和将神经网络与PID控制规律融为一体的PID神经网络(PIDNN)两种控制方案。通过对纸浆浓度模型辨识和控制问题的分析,应用BP和PIDNN进行了仿真比较研究。结果表明,BP和PIDNN仿真效果都比较理想,但BP网络结构复杂,参数难以调整;用PIDNN方法既具有常规PID控制器结构简单、参数物理意义明确的优点,又具有神经网络自学习、自适应之能力,满足实时控制的要求,对于复杂系统是一种实用而简便的控制方法。Due to the problems of delay, nonlinear and time - varying for pulp consistency, the conventional PID control effects are not ideal. To improve the performances of pulp consistency control system, the BP neural network and a new type of controller called PIDNN were applied in this paper, which coalesces traditional PID and neural net- work together. Through the model identification and control problems analyses of pulp consistency, the BP and PIDNN were used for simulations and contrast research. The results show that the simulation effects with both BP and PIDNN are satisfactory. However, the BP neural network has complex structure and its parameters are difficult to ad- just, while the PIDNN has the advantages of conventional PID for its simple construction and definite physical mean- ing of parameters, and also has good adaptability and strong robustness of neural network. It is an appropriate and simple control method for the complex control system and good accuracy for real - time control.

关 键 词：神经网络 纸浆浓度 系统辨识 控制系统 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]