Multi-scenario Model Predictive Control of Combined Sewer Overflows in Urban Drainage Networks

Urban drainage networks (UDN) are among the most vital infrastructures within the natural water cycle. The most widely applied Real Time Control (RTC) on these systems is Model Predictive Control (MPC), which typically incorporates transport time delays and the effect of disturbances explicitly in the objectives and constraints. One of the greatest challenges in the control of UDNs is to formulate multiple control criteria regarding operational requirements of the network. Furthermore, MPC faces the challenge of handling uncertainty caused by disturbances, e.g. weather predictions. One way to incorporate the uncertainty in the decision making is to consider multiple scenarios, i.e. to generate different ensembles based on rain forecasts. To this end, we propose a Multi-scenario MPC (MS-MPC) approach, that deals with uncertainty in the expected inflow. First, a generic multiobjective MPC is established which deals with the time delays explicitly in the optimization. Then, this framework is extended to our formulation of the multiple scenario problem. The algorithm is verified through a case study by interfacing a high-fidelity simulator model of a sewer network as virtual reality.