Model structures for batch and fed-batch ethanol fermentations

摘要

Parameter estimation is one of the most challenging steps while modelling a biological process, due to the inherent complexity of this type of processes and the difficulty in obtaining reliable experimental data. A common modelling practice in this area is to use a batch culture to estimate model parameters and then use these values to simulate and optimize a fed-batch culture. Most of the times this approach is inadequate, resulting in unreliable model predictions. In this study, we analyzed and reparameterized batch and fed-batch kinetic models taken from the literature. A sensitivity analysis applied to the original models showed that most estimated parameters were unidentifiable, unsensitive and not significant for the data of different temperatures. Applying HIPPO, an automatized iterative reparameterization procedure coded in MATLAB, we searched for model structures, for both batch and fed-batch cultures, with identifiable, sensitive and significant estimated parameters. After the analysis of the models, we found a single model structure, suitable for both type of cultures, with two estimable parameters that showed a good fit at all temperatures. Nevertheless, a residual analysis showed that the parameters are positively correlated, indicating that apparently the model needs another degree of freedom.