Improvements and validation of nextSENSE solar energy nowcasting and short-term forecasting system

Abstract

Solar energy is one of the main sources of renewable energy nowadays. Since there is a strong dependence of solar power generation on the presence of clouds and aerosols, operational nowcasting and short-term forecasting of solar resources are essential for its integration into the grid. The aim of this study is the assessment of the downwelling surface solar irradiation (DSSI) estimates from the nextSENSE operational service. This service uses as input earth observational data for clouds (EUMETSAT), aerosols (Copernicus Atmosphere Monitoring Service - CAMS) and other important atmospheric parameters to the fast radiative transfer model (RTM) techniques (look-up table – LUT and multi-parametric equations) in order to derive DSSI in real time over Europe and North Africa in high spatial resolution (5 km at sub-satellite point), every 15 min. Recent modifications relative to the older versions are: (i) the use of multi-parametric equations to obtain the effect of clouds from cloud optical thickness (COT) instead of using Artificial Intelligence techniques, and (ii) the use of more detailed LUT. Forecasted DSSI values are also produced up to 3-hours ahead with a 15-min time step by applying a cloud motion vector (CMV) technique to the COT product based on Meteosat second generation (MSG) satellite data. The new modeled (nowcasted and forecasted) DSSI values were validated against ground-based global horizontal irradiance measurements from pyranometers operating at the Baseline Surface Radiation Network (BSRN) stations and at two additional stations, these of Athens and Thessaloniki, Greece, for the year 2017. The nextSENSE forecasted DSSI values were also benchmarked against the smart-persistence forecast method. The performance of the modeled DSSI values were assessed for different cloud conditions in terms of real cloud modification factor (CMF) values derived by ground-based measurements in conjunction with a clear sky model. Additionally, the effects of aerosol related inputs for estimating DSSI were quantified by comparing the utilized CAMS aerosol optical depth (AOD) forecasts against surface retrievals of the AERONET network. Acknowledgements This study was funded by the European Commission project EuroGEO e-shape (grant agreement No 820852).