Model predictive control of anaerobic co-digestion: Experimental validation of a robust tube-based scheme e

• Model predictive control improves biomethane production. • Diet and setpoints are optimized offline based on a high-fidelity model. • An Extended Kalman filter is designed based on a reduced-order model. • Tube-based NMPC is robust against modeling uncertainties and unknown disturbances. • ∼ 17% production gain achieved in a real bench-scale reactor under safety constraints (e.g., VFA). Modeling, optimization and control tools may play a key role in improving the techno-economic performance of biomethane production via the anaerobic co-digestion process, helping to deal with the challenges that such plants will face in the next decades. Due to the intrinsic complexity of the process and associated safety considerations, industrial practice typically adopts conservative loading conditions that remain well below the maximum production potential. The objective is to safely modify the input diet to track improvements in biomethane flow rate. As model-based and predictive control techniques are of interest due to their inherent ability to optimize a cost function while ensuring constraint satisfaction, this study presents the experimental validation of a tube-based nonlinear model predictive controller (NMPC) for the anaerobic co-digestion of agro-industrial feedstocks. The NMPC framework employs a predictor based on a simple two-stage anaerobic digestion model extended with the hydrolytic step of multiple co-feedstocks, whereas the setpoints are set by an offline optimization carried out with a high-fidelity model. The controller is combined with an Extended Kalman Filter (EKF) which uses only biogas flow and composition data to estimate the unmeasurable states. A successful application of the proposed control scheme to a real bench-scale reactor is presented and benchmarked against a previously validated PI-based strategy.