Reservoir group operation under uncertainty: A Ruhr case study for low flow conditions with ensemble forecast optimization

The increasing frequency and duration of droughts due to climate change pose significant challenges to reservoir operators in maintaining a balance between water supply and demand. While ensuring water supply typically requires high reservoir levels to meet consumer demand throughout the year, other objectives such as flood control and maintaining ecological flows require careful management of water releases. Thus, operators need to optimize their operational schedule, which is a challenging task given the uncertainty of weather forecasts and catchment response to precipitation. In addition, this study investigates the application of ensemble optimization using the Northern Reservoir Group of the Ruhrverband in Germany as a case study. Using the RTC-Tools software, two variants of stochastic optimization are compared: cross-scenario optimization and tree-based optimization. A retrospective ensemble forecasting (i.e., hindcast) approach was used using the twelve years with the lowest total runoff from April to October, representing potential drought conditions. The evaluation is based on the operating ratio used by the Ruhrverband, which reflects the accuracy of reservoir control and is already used in practice, underlining the practical relevance of this study. The results show that both methods improve reservoir control under forecast uncertainty; however, tree-based optimization proves to be more suitable for practical application due to its ability to consider decisions at different time steps and its superior performance in ensuring reliable outflows and water supply.  By emphasizing practical applicability, this work creates robust solutions in the face of uncertainty.