Influence of model uncertainty on real-time flood control performance

Previous research has shown that intelligent control of hydraulic structures can strongly reduce flood consequences, in ideal circumstances. However, uncertainties can significantly impact the performance of real-time flood control strategies. For the Herk river case study in Belgium, this research aims to quantify the influence of the hydraulic model uncertainty. The flood control is for this case conducted by a combination of a Reduced Genetic Algorithm (RGA) and Model Predictive Control (MPC) as optimization method. First, the influence of the initial river model conditions and the length of the prediction horizon on the model accuracy are investigated. Next, the performance of the MPC-RGA technique with and without real-time model updating by means of data assimilation is evaluated. Preliminary results show that even a basic data assimilation technique can compensate for some performance loss due to model uncertainty.