Geoweather

In the ‘Geoweather’ scaling project, satellite-based radiation forecasts and radiation data were further developed on the basis of our algorithms for the Meteosat satellites used in Europe. Using machine learning methods, PV forecasting algorithms were adapted to different meteorological and regulatory conditions for various regions worldwide, with a focus on the curtailment of PV systems. These include, in particular, the aspects of soiling and aerosol pollution. The (semi-) automatic analysis of existing historical and current high-resolution multispectral earth observation data has also opened up a new data source that can be used to optimise existing forecasting models. PV power forecasting models for PV systems with frequent curtailment could therefore be improved by training machine learning methods on satellite-based solar irradiance data from geostationary weather satellites instead of ground measurement data. Furthermore, high-resolution multispectral satellite images from low-earth orbit satellites were used to identify pollution on PV systems and correlations with PV performance. The aim of the project was to licence the IPR generated here to a start-up to be founded in the field of ‘PV forecasting’. Fraunhofer ISE and the EMI's ‘Geoanalyses’ group complement each other in the fields of PV and radiation forecasting as well as earth observation from remote sensing data. The start-up ‘Horizon’, which is to be spun out of Fraunhofer ISE, is therefore in active negotiations with the Fraunhofer Gesellschaft to draw up a licence agreement for the relevant software packages, as well as in active talks with various potential investors to initially finance the start-up's operational business. The software packages to be licensed contain the improvements in the forecasting model achieved through the Geoweather project and are to be used in future for operational PV power forecasts for large PV power plants.