神经网络内模控制在CSTR系统仿真中的应用  

Application of Internal Model Control Based on Neural Network for CSTR System Simulation

作　　者：吕淑萍[1] 马亮[1] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《机床与液压》2010年第14期87-89,70,共4页Machine Tool & Hydraulics

摘　　要：针对CSTR系统的非线性和滞后特性,研究基于神经网络的内模控制的结构和性质。在分析CSTR系统特性基础上,采用神经网络内模控制算法完成控制器设计;用BP网络建立CSTR系统模型和控制器,使用LMBP算法对神经网络权值进行在线调整;为验证该设计方案,在MPCE-1000多功能过程控制实验平台下,对CSTR系统进行仿真分析。仿真结果表明:该方法优于传统的PID控制算法,具有良好的动、静态性能,解决了非线性、滞后等环节的影响。The structure and property of internal model control （IMC） based on neural network were studied to overcome the shortcomings of nonlinearity and time-delay of continuous stirred tank reactor （CSTR） system. After analyzing the characteristic of CSTR system, the IMC controller was constructed based on neural network. Both the CSTR system model and the controller were constructed via BP neural network. LMBP algorithm was used to train the weights of the neural network online. In MPCE-1000 multifunction process control experimental platform, the simulation analysis of the CSTR system was carried on to confirm the validity of the designed scheme. The simulation results demonstrate the performance of this algorithm is better than that of PID algorithm. It has good dynamic and static properties, and the impacts of nonlinearity and time-delay are solved.

关 键 词：内模控制 神经网络 连续搅拌釜式反应器 仿真 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]