Economic model-predictive control of membrane bioreactors for wastewater treatment

The optimization of membrane bioreactor (MBR) operation by nonlinear modelpredictive control (NMPC) with activated sludge models is investigated in this work. To this end, different variants of NMPC are applied to simulation models of singletrain MBR systems in order to demonstrate the technical feasibility of this concept and assess the economic impact of NMPC to full-scale systems. The application of economic NMPC to the model of a single-train MBR system under nominal conditions shows that the electricity cost of single-train MBR systems can be reduced by up to 7–10% and carbon dosage costs by up to 15%, compared to well-tuned conventional control. The economic potential of NMPC is shown to be greatest for control applications with tight effluent limits and a large number of control actuators. The feasibility of economic and ecological NMPC for full-scale MBR is tested by simulation with realistic measurement feedback from on-line measurement. Simultaneous optimization of economic and ecological plant performance is realized by a hybrid discrete-continuous NMPC algorithm which considers economic and ecological control objectives. The algorithm calculates optimal switching times between the objectives together with the optimal control inputs, satisfying optimality conditions defined for the overall control problem. It is shown that the chosen NMPC approach performs robustly when combined with a well-tuned Extended Kalman Filter and a fast trajectory-tracking NMPC despite imperfect state estimator performance and inaccurate disturbance predictions. Directions for future research are pointed out in a discussion of the process control challenges for large-scale MBR and the challenges which need to be addressed to implement this technology in industrial practice.