基于核岭回归的谷氨酸发酵过程软测量建模  

Soft Sensor for Glutamate Fermentation Process Based on Kernel Ridge Regression

作　　者：郑蓉建 潘丰 

机构地区：[1]江南大学轻工过程先进控制教育部重点实验室,江苏无锡214122 [2]淮阴工学院自动化学院,江苏淮安223003

出　　处：《控制工程》2017年第11期2195-2200,共6页Control Engineering of China

基　　金：国家自然科学基金资助项目(61273131);江苏省普通高校研究生科研创新计划项目(CXZZ12_0741);中央高校基本科研业务费专项资金(JUDCF12034)

摘　　要：针对谷氨酸发酵产物浓度难以在线检测问题,提出一种基于核岭回归(KRR)的谷氨酸浓度软测量方法。该方法利用一致关联度(URD)对输入辅助变量进行相关性分析,降低模型复杂性,提高模型性能。然后,结合KRR软测量模型,实现谷氨酸发酵过程中产物浓度的预测。发酵数据仿真结果验证了所提方法的有效性,有助于对谷氨酸发酵过程优化控制提供及时指导。As product concentration in the glutamate fermentation process is difficult to measure in real time, a soft sensor modeling method based on kernel ridge regression is proposed. Firstly, the correlation analysis of input secondary variables and product concentration are carried out by uniform relational degree （URD） to reduce the model complexity and improve the model performance. Then, the uniform relational degree is combined with kernel ridge regression （KRR） to predict the product concentration of glutamate fermentation. Simulation results show that the proposed method is effective, and can provide effective operation guidances for the control and optimization of the glutamate fermentation process.

关 键 词：谷氨酸 核岭回归 一致关联度 软测量 发酵建模 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]