Multivariable Adaptive Control of a Simulated Continuous Yeast Fermentation

Abstract

The performance of a multivariable adaptive controller has been tested on a simulated, continuous, glucose-limited yeast fermentation. The controller is a LQ type controller with iterative solution of the Riccati equation combined with identification of a linear ARMAX-model. Two controlled (biomass and ethanol) and two manipulated variables (dilution rate and concentration of glucose in the feed) are used. The controller has been tested for setpoint tracking and disturbance rejection and compared to a classical dual PI control. The simulation results show, that the multivariable approach has better performance than the multiloop approach and this is ascribed to the ability of the multivariable controller to handle couplings between the process variables. The multivariable controller is especially well suited for on-line optimizing control.