Identification Techniques for a Recombinant Fed-Batch Fermentation for Ethanol Production

Optimization and control algorithms rely on mathematical models. Nonlinear models are necessary to describe biological system behavior. Two input/output modeling approaches are described for a fed-batch fermentation for ethanol production using recombinant Escherichia coli. Volterra series shows a big improvement when an autoregressive term is added, leading to a more general nonlinear model such as NARX. Neural network models are also used for identification of the fed-batch fermentation experiments. In these experiments, neural networks outperform the estimation and prediction of Volterra series. Special attention is given to the design of experiments and modeling techniques.