CloudCV 10-Second Sky Images and Irradiance Dataset

The CloudCV 10-Second Sky Image and Irradiance Dataset contains sky images and irradiance measurements recorded every 10 seconds during daylight hours for 90 days between September 5th to December 3rd, 2019. The dataset was collected at the National Renewable Energy Laboratory (NREL) Solar Radiation Research Laboratory (SRRL) mesa-top campus in Golden, Colorado, USA. The instruments used include an ELP 180 degree Fisheye Lens Wide Angle USB Camera webcam and a co-located LICOR LI200 pyranometer. Historical datasets containing other measurements from co-located instruments with overlapping time period may be available from the NREL Baseline Measurement System. The purpose of collecting this dataset was to train very short term solar irradiance forecasting models from sequential image data of cloud cover.

A more detailed description of the CloudCV sky imager and some preliminary analysis conducted with the dataset can be found in NREL Technical Report NREL/TP-2C00-77999

The code repository contains the sky imager's firmware, as well as analysis code and Jupyter notebooks used to perform a preliminary analysis of the short term cloud cover prediction using a linear advection method based on optical flow. The repository is available on Github under NREL Software Record SWR-24-119.