No abstract

The automatic management of barrage controlled river segments tries to ensure easy and safe navigation and at the same time suppress large water discharge variations. The predominant local PI based control solution used in Moselle river barrages can only guarantee a steady water level at the cost of possibly amplifying inflow disturbances. To balance these contradicting goals, a model predictive controller (MPC) based feed-forward control method is proposed to retrofit the already established local PI-control. The propsed method does not require feedback from the plant, eliminating unwanted interactions with already installed controllers. To achieve real time operation on industrial control hardware, the MPC utilizes simplified river models. Two approaches are compared in this paper: Fully linearized and discretized Saint-Venant equations and an analytical approximate solution of the Hayami equations are used as discharge prediction models for the MPC. The satisfying performance of the proposed approaches is demonstrated using a verified numerical simulation model and is compared to the performance of the existing PI control solution.

Along the German and Luxembourgian part of the Moselle River, eleven distributed local water level and discharge controllers ensure safe navigation by guaranteeing a water level within a specified tolerance and by reducing variations in the river discharge. The current control scheme is based on gain scheduled PI control with a feed forward disturbance compensation element. Both were parametrized using a 1D Saint-Venant model. Due to advancements in control strategies and processing power, the current scheme will be upgraded by adding a model predictive feed forward component (MPFFC) which improves local control and links the isolated local controllers to coordinate their efforts. The authors want to report on this process from an operator's point of view and share their insights from the ongoing testing procedure prior to actual service. A prototype implementation was deployed on the target hardware and linked to the data acquisition system to verify real time operation. The logged results are then verified using a simulation model.