This paper focuses on a two-layer control algorithm for operational management of river barrages, which is based on a model-predictive controller in the upper layer and local controllers at each barrage in the lower layer. The incorporation of hydro-power utilization into the objective function of the predictive controller will be discussed in detail. The process description is based on an efficient numerical schema (Godunov-method) for the full Saint-Venant-equations using a coarse spatial discretization. A simulation study for three barrages of the river Moselle shows the benefits of the suggested approach for a real inflow scenario

Rainfall-runoff models are used as an integral part of flood warning systems. Especially in small catchment areas with a fast response to intense rain, an early enough triggering of warnings requires high quality weather forecasts as well as sufficiently accurate models. A conceptual rainfall-runoff model proposed by Lorent/Gevers serves as core routine of a pilot flash flood warning system for the Truse catchment in Thuringia (Germany). A moving horizon state estimator is used in order to enable this model for online application

No abstract

A decision support system (DSS) for optimized operational water management of artificial inland waterways is presented. It will be deployed as part of a supervisory control and data acquisition (SCADA) system of the Mittellandkanal (MLK), a large canal structure in northern Germany, and relies on experience gained from a similar system. The DSS uses a model predictive controller with a 48 h prediction horizon to calculate optimal pump and discharge strategies that will ensure navigable water levels and at the same time minimize operational costs. The internal process model for the model predictive controller is obtained from a numerical integration of the Saint Venant equations using Godunov's method. The initial state needed for an accurate prediction is estimated using moving horizon state estimation (MHE) or unscented Kalman filtering. Additionally, the state estimation methods are used to estimate non-measurable disturbance inflows, which may have a strong impact on the control performance if not compensated for by the model predictive controller. The optimal control strategy is transformed into discrete-valued pump and discharge jobs that account for technical and operational input constraints. Closed-loop simulations with a highresolution hydrodynamic numerical model of the MLK illustrate the ability of the control algorithm to adapt to model uncertainties and non-controllable inputs

No abstract