Model-based automatic tuning of a filtration control system for submerged anaerobic membrane bioreactors (AnMBR)

This paper describes a model-based method to optimise filtration in submerged AnMBRs. The method is applied to an advanced knowledge-based control system and considers three statistical methods: (1) sensitivity analysis (Morris screening method) to identify an input subset for the advanced controller; (2) Monte Carlo method (trajectory-based random sampling) to find suitable initial values for the control inputs; and (3) optimisation algorithm (performing as a supervisory controller) to re-calibrate these control inputs in order to minimise plant operating costs. The model-based supervisory controller proposed allowed filtration to be optimised with low computational demands (about 5min). Energy savings of up to 25% were achieved when using gas sparging to scour membranes. Downtime for physical cleaning was about 2.4% of operating time. The operating cost of the AnMBR system after implementing the proposed supervisory controller was about 0.045/m3, 53.3% of which were energy costs.