EGU General Assembly 2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Inter-comparison of PV power simulations from seven gridded irradiance data sets 

Darragh Kenny1 and Stephanie Fiedler1,2
Darragh Kenny and Stephanie Fiedler
  • 1University of Cologne, Institute of Geophysics and Meteorology, Meteorology, Cologne, Germany (
  • 2Hans-Ertel-Center for Weather Research, Climate Monitoring and Diagnostic, Bonn/ Cologne, Germany (

Accurate irradiance data is necessary for model estimates of expected photovoltaic (PV) power production. Such data is freely available from reanalysis and satellite products with a high temporal and spatial resolution, including locations without ground-based measurements. Gridded irradiance data is therefore used for the characterization of solar resources at specific locations and larger areas, e.g. by power system modellers. Past assessments of irradiance data for PV modelling often relied on the evaluation of global horizontal irradiance (QGHI). However, the direct and diffuse irradiance components as well as differences in seasonal characteristics can strongly affect the PV capacity factors (C) potentially leading to larger biases in C than for QGHI. We therefore systematically assess differences in QGHI, direct and diffuse horizontal irradiance (Qdir  and Qdif) and quantify the subsequent bias propagation from individual radiation components to C in a contemporary PV power model. Our PV model simulations use seven different gridded irradiance data sets, namely ERA5, COSMO-REA6, COSMO-REA6pp, COSMO-REA2, CAMS radiation service, SARAH-2 and CERES Syn1Deg. All data sets provide Qdir and Qdif as separate time series spanning seven to 43 years and with a temporal resolution of 15 minutes to one hour. The results are compared against seven years of simulations based on reference measurements from 30 weather stations of the German Weather Service. We compute metrics characterizing biases in seasonal and annual spatial means, day-to-day variability and extremes in C, considering single stations and a simulated PV fleet. Our results highlight biases of -1.4 % (COSMO-REA6) to +8.2 % (ERA5) in annual and spatial means of C at single stations across Germany, while the bias in QGHI is -3 % for COSMO-REA6 to +3.6 % for ERA5. We also show the bias on days of very low PV production, relevant for extreme event analysis: The days within the lowest ten percent of daily PV production in a PV fleet show a bias of +70.2 % in ERA5, while it is only +4 % in the post-processed COSMO-REA6 data (COSMO-REA6pp). SARAH-2 and COSMO-REA6pp outperform the other products for many metrics, but also cause some biases in C. For instance, SARAH-2 yields good results in summer, but overestimates C in winter by 16 % averaged across all stations. COSMO-REA6pp represents day-to-day variability in C of a simulated PV fleet very well and has a relatively small bias of +0.5 % in the annual spatial means, but this is partly due to compensating biases from individual stations. Our results suggest that gridded irradiance data should be used with caution for site assessments and should ideally be complemented by local measurements. For power system modellers, our results may provide guidance for the quantification of uncertainties caused by gridded irradiance data.



Kenny, D., and Fiedler, S., in press, Which gridded irradiance data is best for modelling photovoltaic power production in Germany?, Solar Energy.

How to cite: Kenny, D. and Fiedler, S.: Inter-comparison of PV power simulations from seven gridded irradiance data sets , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5888,, 2022.