Fed-batch bioreactor control using a multi-scale model

The fed-batch mode of bioreactor operation is amongst the most commonly used in the fermentation industry, and fed-batch bioreactors are difficult to control due to their nonlinear and dynamic behavior. In this work a multi-scale model is used that describes the dynamic behavior of aerobic fed-batch cultures of Saccharomyces cerevisiae. Open-loop and closed-loop control issues are addressed for bioreactor optimization and control with the goal of maximizing the ethanol concentration at the end of the batch. Nonlinear optimization techniques are used off-line to obtain an open-loop optimal substrate profile, and a shrinking horizon nonlinear quadratic dynamic matrix control (SHNLQDMC) algorithm is used for online closed-loop control. Particular emphasis is placed upon disturbance compensation techniques for closed-loop operation. The controller achieves an admirable nominal performance finishing within 1% of the final open-loop optimal ethanol concentration. For disturbance compensation it was found that a closed-loop re-optimization strategy improved the final ethanol concentration by 40% compared to the case when the open-loop optimal profile was employed. Finally, it was observed that reducing the sample time considerably improved the controller performance and made the system more robust to disturbances.