Precision and accuracy on solar resource estimates is required for solar plant bankability assessment. Precise estimates can be generated for anywhere in the world with the fast and yet physical radiative transfer code SMART-G [1]. The direct and diffuse fractions of the solar radiation reaching the surface are both explicitly computed. Concentrating solar plants (CSPs) mainly collect the direct fraction, and solar resource in CSPs is usually estimated with computation of the direct normal irradiance (DNI). Even for photovoltaïc installations, DNI can be computed to estimate the impact of potential shadows, while the diffuse fraction is separately computed to estimate the impact of horizon, especially for tilted panels in contrasted topography.
Comparisons were made between computations and observations made at Lille, France, in 2011 and 2016. Clear-sky DNI was computed at 1-hour time sampling, using the aerosol properties and the water vapor content provided by the AERONET station of Lille. The clear-sky conditions provide the largest values of DNI, which are however highly sensitive to the aerosols. The agreement was satisfying, with correlation coefficients between observed and computed clear-sky DNI of 0.971 in 2011 and 0.990 in 2016, and with root mean square differences (RMSD) of 35 W/m2.
Strict cloud-screening procedure was applied to generate the observation-computation coincidence pairs, by both exploiting Level 2.0 AERONET data and applying further cloud-screening tests on the DNI observation data set. Most cloud influences are expected to be avoided while keeping the aerosol influence, as the aerosol optical thickness (AOT) varied between 0.02 and 0.64, with a mean annual AOT of 0.17±0.11 in 2011, and a mean annual AOT of 0.11±0.09. Consequently, the annual average of the clear-sky DNI was 692±137 W/m2 in 2011 and 756±134 W/m2 in 2016.
Comparisons in DNI also inform about the appropriateness in the input data of the radiative transfer code, as DNI is more sensitive to atmospheric extinction than the global irradiance. Clear-sky DNI was also computed with input data extracted from the global MERRA-2 data set. The RMSD increased by ~65%, and the correlation coefficient decreased to 0.91-0.93. Such a score remains competitive in regards to operational state-of-the-art solar resource data sets.
We expect to further improve the agreement between computation and observation by simulating the fraction of the diffuse radiation contributing effectively to the measurement of DNI. Indeed a pyrheliometer measures not only the solar radiation coming directly from the sun, and not interacting with the atmosphere (strict DNI [2]), but also the circumsolar radiation scattered by the atmosphere in a solid angle of a few degrees around the solar direction, which generates the solar aureole. The annual average of the computed strict clear-sky DNI under estimated the observation by ~1% in 2011 and by ~4% in 2016 (with both AERONET and MERRA-2 input data), when the circumsolar radiation is expected to increase strict DNI by a few % [2].