Skip to Main Content
Incorporating weather forecasts in the control of land surface water levels requires predicting the outcome of a control action. In Model-Predictive Control (MPC) such predictions are generated by a model of the water system. The actuators in this specific application typically cause discontinuities in this model. Avoiding complex solving techniques for such hybrid systems, this paper introduces an alternative which enables pure continuous model-predictive control by smoothing the jumps. Though the resulting underlying model is smooth, it is also highly nonlinear. This requires specialized continuous MPC methods, like the DotX Nonlinear Predictive Controller (DNPC), which are able to cope with these nonlinearities. Employing DNPC for this computation yields an approximation to an optimal control prediction. With a final postprocessing step this control prediction can be translated to the original hybrid system. This paper verifies this method by applying DNPC, equiped with an approximated continuous nonlinear model, to a real life hybrid water system.