The percentage of fluctuating energy sources in our power supply is growing at a steady pace. Precise chronological forecasts of power feed-in are thus playing an increasingly important role. Wind, photovoltaics and hydroelectric power feed-in forecasts are already essential parameters in the management of transfer and distribution grids.
ZSW researches application-oriented technologies that raise the accuracy of forecasts of the complex interaction between weather model forecasts, satellite data and meteorological measurements and the historic and future wind, solar and hydroelectric power yields of a specific site or region.
The latest algorithms for machine learning methods play an important role in this context. They are capable of significantly improving physically calculated wind, solar irradiation and output forecasts by employing long-term measurement series that help them learn the interaction patterns between the numerous parameters that cannot be covered by physical models and apply the results to the respective current weather situation. Systematic deviations that generally occur in physical models are automatically corrected.