Adaptive inferential predictive control for fermentation bioreactors

An adaptive multirate inferential predictive control strategy is presented for discrete-time systems and applied to a model of a biological fermentation reactor. The output variable to be controlled is available at low sampling rates, due to sensor delay and dynamics. Other auxiliary signals are then measured at higher sampling rates and an identification procedure is used to estimate the output at every instant when a direct value is not available. These estimates are then used for control calculations according to the long-range adaptive control algorithm MUSMAR. The simulation results show the controller´s ability to control plants with unmodeled dynamics or nonminimum phase zeros