Submerged membrane bioreactors are increasingly applied for wastewater treamentbut requires a tight control of the membrane fouling so as to ensure safe and efficient operation.The objective of this paper is to design a nonlinear model predictive control to minimize theirreversible resistance while keeping the trans-membrane pressure, which is a good indicatorof membrane fouling, at an acceptable level. To this end, the manipulated variables are thepermeate flow and the air scouring flow, which allows the material layer formed on the membrane(in short the “cake”) to be detached. The NMPC structure is tested in simulation consideringa detailed simulator as the reference process, and a reduced-order model as the predictor. Theresults show that the process can be regulated until the irreversible resistance takes the mainrole in the fouling resistance. When this state is reached, a chemical cleaning is required, or alarger trans-membrane pressure has to be accommodated.