A hierarchical state estimation and control framework for monitoring and dissolved oxygen regulation in bioprocesses

The integration of state estimation and control is a promising approach to overcome challenges related to unavailable or noisy online measurements and plant-model mismatch. Extended Kalman filter (EKF) and moving horizon estimator (MHE) are widely used methods that have complementary features. EKF provides fast estimation and MHE optimal performance. In this paper, a novel hierarchical EKF/MHE approach combined with a dynamic matrix controller (DMC), denoted as EKF/MHE–DMC, is proposed for process monitoring and dissolved oxygen control in airlift bioreactors. The approach is successfully tested in simulated cultivations of Escherichia coli for recombinant protein production, considering specific scenarios of step set point tracking, step disturbance rejection, plant-model mismatch, and measurement noise. Results also show that, given a model that describes the measured dissolved oxygen precisely, as assumed in this study for the in silico experiments, the EKF/MHE–DMC approach is able to estimate the cell, protein, substrate, and dissolved oxygen concentrations based only on the measurement of the latter, reducing the estimation error by 93.8% when compared to a benchmark case employing EKF and DMC. The general structure of the proposed EKF/MHE–DMC framework could be adapted for implementation on other relevant bioprocess systems employing their derived process models.