Subproject 1: Optimization of weather and power forecast models
The main task in this subproject is to optimize the weather and power forecast models towards improved wind and PV power production forecasts.
The quality of the weather forecasts is strongly dependent on the accuracy of the initial conditions of the forecast. Different studies , e.g. by the National Center for Atmospheric Research (NCAR) in the USA, have demonstrated improvements in weather forecasts by assimilating measured meteorological data directly from the wind turbines and the photovoltaic power plants. In this project, we investigate the potential of introducing power production measurements for improving the description of the initial conditions for the weather forecast of relevant parameters, such as radiation, cloud cover, and wind speed. For this purpose, transformation operators are required to transform the meteorological parameters into power production. These operators are developed by Fraunhofer IWES and integrated into the data assimilation system at the DWD.
The evaluation and optimization of the weather forecast models has until now generally not been dedicated to parameters with high relevance for applications in the energy sector. In this subproject, we investigate to what extent critical errors in the power forecasts can be reduced by optimizing the model physics in the numerical weather prediction models for parameters that are relevant for energy applications (e.g., wind at hub-height and cloud cover).
Based on the optimized weather forecast models, power forecast models will be developed. Firstly, the neural network applied operationally at IWES will be optimized, and secondly, further physical approaches will be tested. The development, analysis, and optimization of the forecast models will focus on time frames of 0-72 h and on various spatial resolutions such as individual power plants, regions, nodes, control areas, and the whole of Germany.
In a further working step, forecasts will be assigned to situations that are especially easy or difficult to forecast. The predictability will be investigated for different situations connected to critical errors in the power forecasts. For this reason, multi-dimensional analyses against observations (e.g., satellite retrievals) will be carried out. Additional information can be gained from also monitoring the success or failure of a forecast during the first hours, which can be of importance when estimating the risks during the following hours. These findings will be integrated into the result-oriented combination of improvements that are acquired in this subproject.
back to "Project"