Model Predictive Control with SWMM

Real time control (RTC) is particularly promising in large, flat and heterogeneous sewer systems with a high in-line storage volume. For the simulation of such systems with major backwater effects the use of hydrodynamic routing models is indicated but for model predictive control (MPC) they are generally regarded as infeasible because they are time consuming in computation and therefore not practical for receding horizon applications.

The aim of this work is to test the application of hydrodynamic flow routing calculations for MPC. It focuses on the question whether it is possible to overcome the time restrictions to find acceptable regulator settings. For calculations the software BlueM.MPC was developed which enables MPC simulations with the hydrodynamic sewer network model SWMM (Rossman, 2009). BlueM.MPC is a supplementation to the generic software framework BlueM.Opt (Muschalla, Fröhlich, u. a., 2009), which provides various optimization algorithms and is able to integrate any text-based simulation software. It offers consideration of different time horizons in order to take into account the time span required to evaluate the optimization objectives (prediction horizon), the time span for which system input is known in advance (forecast horizon) and the time span for which regulating set points have to be optimized. For the formulation of control objectives parameters representing flow and water quality conditions can be used.

The generated MPC framework keeps modules for optimization and flow simulation separated leading to a text-based parameter optimization procedure. In order to reduce calculation times different alternatives were analysed. The results which will be presented include comparisons of different optimization algorithms taking into account local and global methods as well as the analysis of possibilities to simplify the sewer network model.