Irradiance maps applied for the performance assessment of PV-Systems- a case study for the German federal state of Saxony

For the estimation of the expected annual energy gain and the month by month check of the system performance, methods based on irradiance maps published by weather services, both general or dedicated to solar energy application, are in use. Examples for this type of information for Germany are annual and monthly irradiance maps published by the German weather service DWD or the data bank of hourly irradiance data with continuous spatial coverage prepared by the University of Oldenburg and offered as commercial service by the company Meteocontrol. The DWD maps are derived from both, satellite images and ground measurements and are given with a spatial resolution of 1km x 1km. The Oldenburg data bank is basically derived from satellite data only (resolution ~ 5km x 5km), although an option for amelioration by ground station data exists.
To assess the validity of these data sets for the aforementioned tasks, a case study for a region covering the German federal state of Saxony is performed, using data for the years 2004-2005.
Saxony is characterised by a wide variety of topographical features, ranging from flat relief to highly structured mountainous terrain. For this region an independent high quality ground data set given by a dense network operated for agro-meteorological purposes is available. An assessment of the end use accuracy of the irradiance data can by done via a set of monthly yield data of grid connected PVsystems.
The comparison of information on radiation sums is on one hand performed by the monthly analysis of the bias and the rms-error for the data bank versus the ground station data. On the other hand, the quality of the information on the spatial structure of the radiation field that is extractable from the data banks is analysed via the inspection of the a co- and cross-variograms.
For the assessment of the end use accuracy of the data, procedures to derive an estimation of the system yield from irradiance data as developed by the PVSAT2 project are applied for the Oldenburg data set.
A first impression of the reliability of the map data is given by the inter comparison of the annual maps for the year 2004. This shows deviations of up to 40 kWh/m², which presents an relative error of about 4%. Besides a systematic bias of about 20 kWh/m² in favour the Oldenburg data, remarkable differences in the spatial structure of the irradiance field concerning the homogeneity/inhomogeneity within some regions occur.
Thus, to pin down the reliability of the content of the maps it has to be compared to the ground data.
For the Oldenburg data it can be stated that even having a quite low bias for each station on an annual scale, monthly sums can deviate by up to 40%, especially for the winter months. This finding is coherent with the results of the PVSAT project, indicating the general increase of the uncertainties of satellite based estimates in months with low radiation sums.
Concerning the annual end use accuracy, a typical deviation of ? 5% for the estimated annual energy gain of a set PV-systems - chosen for their excellent yield figures - can be stated. Regarding the fact, that the system simulation is base on data sheet information rather than system specific parameters, this result is quite satisfactory. However, as expected remarkable relative deviation for the winter
months are hidden by these annual figures which are dominated by the PV gain in the summer months.
Using the respective results of for sets of both 2004 and 2005 general conclusions on the uncertainties connected to the use of state of the art irradiance information in view of PV application can be draw.