基于机理知识与最小二乘支持向量机的诺西肽发酵过程混合建模方法

提出了一种综合先验机理知识与最小二乘支持向量机的发酵过程混合建模方法。机理知识由两部分组成,一部分是表示发酵过程的质量平衡方程;另一部分是简单的过程参数估计模型。采用最小二乘支持向量机模型对这种简单估计模型进行校正,弥补它的不精确性。这种混合模型被应用到诺西肽发酵过程中进行生物量浓度、基质浓度与产物浓度的估计中,结果表明加入先验机理知识的混合模型增加了单纯的支持向量机模型的泛化能力。

Hybrid Modeling for Nosiheptid Fermentation Process Based on Prior Knowledge and Least Square Support Vector Machines

A hybrid model for fermentation process was proposed which combined prior knowledge and least square support vector machines(LS-SVM).Prior knowledge include two parts,one is the mass balances model of process;another is the simple model of the process parameter estimating.The LS-SVM model was used to provide an additive correction to the simple process models,which compensates for inaccuracy in the simple process models.The hybrid model was applied in predicting biomass,substrate and product concentration of Nosiheptide fermentation process.The results show that the hybrid model with prior knowledge enhances generalization capabilities of a pure SVM model.