A new approach to forecast solar irradiances using WRF and libRadtran models, validated with MERRA-2 reanalysis data and pyranometer measures

Abstract

Solar energy is playing an increasingly important role in the production of electricity through photovoltaic and Concentrated Solar Power (CSP) systems. This type of energy requires an efficient storage system that will allow its use when the sunset or during cloudy weather. Hence, 0-48 hours forecast lead time is crucial for electrical network managers to anticipate energy deficits during periods of cloud occurrence. This study presents a new approach to forecast solar irradiances using a Numerical Weather Prediction (NWP) model outputs, coupled with a Radiative Transfer operator, namely libRadtran. The NWP model used in this study is the WRF-Solar configuration of the Weather Research and Forecasting (WRF), designed for solar energy in order to forecast solar irradiances. The other parameters forecasted by WRF like the vertical profiles of cloud liquid water (CLW), cloud ice water (IWC), clouds fraction, relative humidity, and temperature will be used as input for libRadtran to compute off-line the solar irradiances. Both forecasted irradiances were first compared with the Modern-Era Retrospective Analysis for Research and Applications Version-2 (MERRA-2) data, then validated a second time by ground measurements for three locations in Algeria for the years 2016 and 2017. The results clearly show that the solar irradiances forecasted directly by WRF-Solar underestimate the MERRA-2 data for high solar zenith angles. With regard to the proposed approach, there is an improvement in the forecast of the solar irradiances during clear sky days and the rRMSE of the calculated Global Tilted Irradiance (GTI) reaches 12% in winter and increases to 22% in summer.