Waste water treatment plants with submerged membrane modules form an interesting alternative to common waste water plants where the separation of particulate substances from the purified water is achieved by large clarifiers. The advantages of the membrane bio reactor comprise better waste degradation resulting from higher biomass concentration, improved effluent quality even under strong feed fluctuations and a more compact plant layout due to the absence of the clarifier. However, operational costs of this type of plant currently exceed the operational costs of common waste water treatment plants. Therefore, the presented work formulates a model based on-line optimization scheme in order to guarantee effluent quality under varying feed scenarios while simultaneously minimizing the operational costs. Due to the varying amount and concentration of waste in the inflow, the complex biological transformation steps and necessary simplifications a model based optimization approach requires the combined estimation and reconstruction of the model states and disturbances and also relies on a good predictive model describing the main phenomenon in the waste water treatment plant. This work presents a decentralized optimization based control framework that exploits different time scale and results in two separate optimization problems for the biological waste degradation and the operation of the membrane modules. The objective for the biological part is to guaranty effluent constraints for certain species, while minimizing the operational costs. However, under storm water scenarios it may be necessary to change the operational strategy in order to cope with the amount of waste water that has to be processed. Under this conditions the control strategy has to guarantee the constraints on the effluent concentration and the throughput through the plant withoutreaching a state where off-spec waste leaves the plant. This situation is possible if several optimization constraints act simultaneously and thus reduce the flexibility of the plant. Therefore, in the presented work the control strategy is selected on-line based on the current state of the plant and on a prediction of the plant behavior. The proposed control frameworkdetermines the optimal sequence and duration for the control strategy based on actual plant conditions and state predictions and computes the optimal control moves. It also determines the permeate flow that has to be reached by the control of the membrane modules.The objective for the membrane modules is to realize the permeate flow determined by the control strategy applied for the biological part of the plant while simultaneously minimizing the operational costs for the membrane modules. The model used for this optimization is based on simple measurements available at the membrane module. A second objective is to calculate the maximum possible flow that can be realized under the current membraneblocking. This information is needed to give an upper limit for the optimization of the biological part.The developed model based optimization framework is realized in a realtime environment and tested in simulation studies. The model used for the simulation studies is developed from literature and presents a highly detailed model including biological waste degradation based on the Activated Sludge Model No. 3, climate induced temperature variations, cake formation, the effect of the aeration on the cake formation and the pore blocking resultingfrom fouling.The presented results show, that the control strategy can be adopted based on the surrounding scenario. The main savings for the operational costs of this type of plant can be achieved byreducing the aeration of the membrane modules.