The incoming radiation at the site of the PV system has to be known to simulate the expected energy yield. This information is derived currently by measurements of the geostationary weather satellite Meteosat-7 (It will be changed to Meteosat-8 in the future.). These data have a high temporal resolution and a better spatial coverage than meteorological ground stations. The Meteosat images are received half-hourly operational at the University of Oldenburg. The Heliosat method (Hammer, 1999) is used to determine ground irradiance for a pixel with a resolution of 3 km x 5 km above the site of a PV system. Validations with ground stations showed that the error of this method can reach up to 30% for hourly values and lies between 5 and 10 % for monthly irradiance values (Hammer, 1999). These errors are strongly dependent on the weather situation (i. e. variability) and on the sun elevation angle. Under cloudy sky conditions the error increases while only small errors occur on days with clear skies. Larger errors also occur for low solar elevation angles and result in more deviations during wintertime. The quality assessment on these derived irradiance data and its dependencies is rather complex. Further details and explanations on the quality assessment will be published soon.

One further development to improve the irradiance calculation and to reduce the error is the introduction of a new model to calculate diffuse irradiance (Mueller et al, 2004).

Another way for improving the accuracy is the combination of the satellite data with ground measurements of irradiance with the geostatistical method kriging-of-the-differences. The error can be reduced this way to 3-4% for monthly values and also for hourly data (Betcke and Beyer, 2004).

These errors within the derived irradiance data have to be considered in the error detection routine. The exact knowledge of these errors and its dependencies is an absolutely necessary prerequisite for a variable definition of allowed differences between the expected and the real energy yield and therefore, for the exact determination of the occured malfunction.