Mathematical modelling of fed-batch fermentation processes for amino acid production

The unstructured dynamic models of the fed-batch fermentation processes for microbiological L-leucine production have been discussed. Generalised stoichiometric equations describing the proposed process are proved by autoregressive models. Non-linear dynamic models based on the stoichiometric equations are presented. The structure and parameters of the specific rates are selected using linear- and non-linear regressions as a first step of the structural and parametric identification. The dissolved oxygen concentration in the culture liquid which is an important variable for this process is included in the models of specific rates. The initial estimation of the parameters of dynamic models and their confidence intervals is obtained by non-linear regression procedure. The optimal parameters of the dynamic models are obtained as a second step of the structural and parametric identification. The optimisation procedure is performed by the direct search method and the least squares criterion is applied. The adequacy of the proposed models is proved through simulation research.