A Virtual Sky Imager Testbed for Solar Energy Forecasting

Benjamin Kurtz, Felipe Mejia, Jan Kleissl

Final Article in Solar EnergyPreprint PDF


Whole sky imagers are commonly used for forecasting irradiance available for solar energy production, but validation of the forecast models used is difficult due to sparse reference data. We document the use of Large Eddy Simulations (LES) and a 3D Radiative Transfer Model to produce virtual clouds, sky images, and radiation measurements, which permit comprehensive validation of the sky imager forecast. We then use this virtual testbed to investigate the primary sources of sky imager forecast error on a cumulus cloud scene. The largest source of nowcast (0-minute-ahead forecast) errors is the converging-ray geometry implied by use of a camera, while longer-term forecasts suffer from overly-simplistic assumptions about cloud evolution. We expect to use these findings to focus future algorithm development, and the virtual testbed to evaluate our progress.

Online Content

As supporting material for the graphs shown in the journal article, we are pleased to provide virtual sky images and simulated ground-level flux data for the day of simulations shown in the paper. LES inputs and outputs can potentially be provided upon request, however the outputs in particular are quite large (over 70GB for each day of simulation time).

For our internal book-keeping purposes, we assigned a calendar date to each simulation we ran. These dates were assigned arbitrarily and should not be inferred to have any particular meaning. The simulation results presented in the paper were assigned a calendar date of 2017-04-23, and are currently the only results available for download. Additional simulations were run, but as those results have not been thoroughly validated, they are not provided at this time.

Users of this data will want to note that SHDOM simulations (producing virtual images and ground-level fluxes) were run with a constant solar position, in order to emphasize that trends in our results are due to changing cloud conditions rather than changes in solar geometry.

Fluxes at ground level were computed for red, green, and blue channels, with wavelengths of 670 nm, 550 nm, and 450 nm respectively, the same wavelengths used to generate the color channels for the images. Fluxes are normalized so that ToA "DNI" for each color channel is 1 (i.e. ToA flux is equal to cosine of solar zenith angle). X and Y coordinates in the flux files are in meters. Virtual sky images use an equisolid-angle projection.

2017-04-23 Image Data (413 MB)

2017-04-23 Flux Data (118 MB)