Dealing with rainfall forecast uncertainties in real-time flood control along the Demer river

Abstract

Real-time Model Predictive Control (MPC) of hydraulic structures strongly reduces flood consequences under ideal circumstances. The performance of such flood control may, however, be significantly affected by uncertainties. This research quantifies the influence of rainfall forecast uncertainties and related uncertainties in the catchment rainfall-runoff discharges on the control performance for the Herk river case study in Belgium. To limit the model computational times, a fast conceptual model is applied. It is calibrated to a full hydrodynamic river model. A Reduced Genetic Algorithm is used as optimization method. Next to the analysis of the impact of the rainfall forecast uncertainties on the control performance, a Multiple Model Predictive Control (MMPC) approach is tested to reduce this impact. Results show that the deterministic MPC-RGA outperforms the MMPC and that it is inherently robust against rainfall forecast uncertainties due to its receding horizon strategy.