混杂模型模拟甘油发酵过程  

Dynamie Process Modelling of Glycerol Fermentation by Differential Hybrid Model

作　　者：张金龙[1,2] 吴杰 李会[3] 

机构地区：[1]青岛科技大学,山东青岛266042 [2]滨化集团,山东滨州256600 [3]滨州职业学院,山东滨州256613

出　　处：《广州化工》2011年第21期88-92,共5页GuangZhou Chemical Industry

摘　　要：以1,3-丙二醇为研究对象,首先建立了甘油发酵的混合半参数模型。混杂半参数模型就是机理与经验模型之间的平衡,特别适用于复杂系统,当然也包括非常复杂的生物系统,可以不通过机理阐明而得到大量数据,近来受到人们的重视。根据建立好的混杂模型,我们利用提前终止法提高了BP神经网络的泛化能力,最终确定了隐含层神经元个数为8,训练函数为traingdx。根据已经建立好了的人工神经网络结构,利用ODE45函数对间歇和批示流加过程混杂模型的微分方程组进行积分求解,最后得到了较好的模拟结果。Taking 1,3-propanedio as an issue,the discrete hybrid model was developed for glycerol fermentation.Hybrid semi parametric made the mechanism and empirical model balanced,was specially suitable for the complicated system and the very complex biological system,may obtain the mass data with no mechanism expounded,which got attention by many people recently.By the established hybrid model,the generalization ability of BP neural network was improved by using early termination method,and ultimately determined the number of hidden layer neurons for 8,the training function traingdx.The artificial neural network structure was established,which can use function ODE45 instructions for hybrid model of the batch process and fed-batch process to the solution of differential equations,and got good simulation results.

关 键 词：1 3-丙二醇 代谢工程 混杂模型 甘油发酵 人工神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]