Switzerland relies on hydropower for approximately two-thirds of its electricity generation, making efficient plant operations essential for grid stability and economic performance. However, hydro operators face constant uncertainty regarding water inflows, market prices, and electricity demand, complicating long-term planning and profit optimization.
Researchers have developed a comprehensive operational framework that addresses this challenge by combining short-to long-term planning horizons—spanning up to one year—using rolling-horizon optimization. Rather than assuming perfect knowledge of future conditions, the model operates with realistic seasonal and sub-seasonal weather forecasts, mirroring actual decision-making environments.
The framework successfully captures hydrological dynamics across multiple timescales, reproducing observed Swiss hydropower operations with reasonable accuracy. Monthly turbine production predictions achieved a Mean Absolute Percentage Error of 18% when using forecast data alone, demonstrating the model's practical utility despite inherent forecast limitations.
A key finding emerged when comparing forecast-based operations against a hypothetical scenario with perfect information. Revenue losses from forecast uncertainty totaled only 6%, while unplanned spillage increased marginally from 0.18% to 0.64% of total inflows. These modest penalties underscore that realistic operational strategies—constrained by available forecasts—perform near-optimally on a system-wide basis.
This research addresses a critical gap in hydropower modeling. Most existing approaches either assume perfect foresight, which is impractical, or oversimplify hydropower representation in grid simulations. By incorporating genuine forecast dependencies, the framework provides grid operators with actionable insights for balancing competing objectives: maximizing revenue, maintaining adequate storage reserves, and meeting real-time demand without excessive spillage.
For Swiss grid operators and energy market participants, this model offers a realistic tool for day-to-day and seasonal planning. The relatively small performance gap between forecast-constrained and perfect-information scenarios validates the robustness of forecast-based decision strategies, supporting more confident operational commitments and market participation.
