On-Line Optimizing Control of a Simulated Continuous Yeast Fermentation

On-line optimizing control of a simulated fermentation is investigated using a non-segregated dynamic model of aerobic glucose limited growth of saccharomyces cerevisiae. The optimization procedure is carried out with an underlying adaptive regulator to stabilize the culture. This stabilization is especially important during the setpoint changes specified by the optimizing routine.

A linear ARMAX model structure is used for the fermentation process with dilution rate as input and biomass as output variable. The parameters of the linear model structure are estimated using a pseudo linear regression method with bandpass filtering of in- and output variables in order to ensure low frequency validity of the estimated model.

An LQ-regulator is used with iterative solution of the Riccati equation. Simulation results illustrate the tuning of the underlying regulator, and the effect of perturbing the manipulated variable with a dither signal. The results demonstrate that an optimum can be reached smoothly in what corresponds to about twice the effective open loop response time.

The results demonstrate the effectiveness of the stabilizing regulator in rejecting relatively short duration disturbances.