生物发酵过程的在线神经网络逆解耦控制  

Online neural network inverse decoupling control in fermentation process

作　　者：于霜[1] 刘国海[1] 梅从立[1] 江辉[1] 

机构地区：[1]江苏大学电气信息工程学院,江苏镇江202013

出　　处：《计算机与应用化学》2012年第9期1127-1130,共4页Computers and Applied Chemistry

基　　金：镇江市农业科技支撑项目(NY2010017);江苏省高校优势学科建设工程([2011]6);国家中小型创新基金(12C26213202207);江苏省高校研究生创新项目(CXLX11-0588)

摘　　要：生物发酵过程具有非线性、多变量耦合的特点。为提高发酵过程解耦控制的性能,提出了一种基于参考模型的在线神经网络逆解耦控制方法。根据发酵过程的特点,给出发酵过程的数学模型和可逆性分析。进一步构造神经网络逆解耦器与被控系统串联,由参考模型输出与被控系统输出的误差设计神经网络连接权值参数的在线学习方法,将被控系统解耦成伪线性系统,最后设计线性闭环控制器获得高性能跟踪控制。仿真结果表明,所提出的方法很好的实现了多变量生物发酵过程的解耦控制,鲁棒性强且结构简单易于工程实现。The fermentation process is a nonlinear and multi-variable coupled system. To improve the control performance, an online neural network inverse decoupling control based on reference model is proposed. According to the characteristic of fermentation process, the mathematic model and invertibility are given. Construct neural network inverse system which is in series with the original system. Then design the online learning algorithm which is used to adjust the parameters of neural network. Online learning algorithm is designed using the error between the output of reference model and original system. The controlled system is coupled into the pseudo-linear systems. The high performance is achieved by designing linear losed- loop controller. The simulations show that the control method implements the decoupling control of multi-variable fermentation process. The control method has strong robustness, a simple structure and is easy to implement in engineering application.

关 键 词：发酵过程 参考模型 在线神经网络 逆系统 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]