Objectives and background
Obtaining accurate data of surface solar irradiance in real-time is essential to initiate operational photovoltaic (PV) production forecasts. It is also useful for PV production monitoring: it enables to detect a decrease of the performance ratio due to sudden system failure or soiling effects.
Solar irradiance retrieval from geostationary meteorological satellites is a widely used process to obtain solar irradiance maps, time-series or real-time data with satisfactory accuracy, at world-scale and for a reasonable price. Its main advantage is to avoid the use of a meteorological instrument whose purchase, installation and maintenance can be very expensive, requiring specific skills.
In this work, we assess the performance of SunSat liveTM. This operational service is able to provide global and direct irradiance using five satellite covering about 70 % of the Earth surface excluding polar areas. It has been implemented and assessed on the most recent geostationary satellites available in 2018, including GOES-16 and GOES-17.
Method
A specific computer architecture has been set up in order to receive satellite data in real time. Parallel computing permit to obtain solar irradiance on required pixels in less than 2 minutes. Real-time weather services have been solicited to obtain water vapour concentration and aerosol optical depth forecast that have been reprocessed to be usable as real-time products. Time series of satellite-derived irradiance data have been then routinely generated and stored into a dedicated database.
About one hundred time series of global and direct irradiance measurements have been collected across the world for the period 2017-2018. Co-located time series of satellite-derived data have been also generated (with a shorter period for GOES-17 due to the delayed data release).
Comparisons have been computed between satellite-derived data and collocated measurements in global and direct irradiance. Site adaptation performance has been also assessed using 2017 dataset of neighbouring stations in a similar climatic area.
Principal findings
After site adaptation, the results shows an average bias error around 2 % for the pixels processed using Himawari-8 satellite and -1 % for those of MSG IODC. Coefficient correlation is superior to 0.95 in all cases. The effects of the higher resolution delivered by third generation satellites (GOES-16 and -17, Himawari-8) has been also investigated.
Conclusion
The service SunSatTM is able to provide solar irradiance in real-time, using very recent satellites, with a good accuracy compared to the state-of-the-art. Using 12 months of data for site-adaptation enables to deliver more accurate results. The automated procedure of site adaptation enables to benefit this bias reduction on the fly.