Testing new ensemble forecasts of solar irradiance

Abstract

The increment of solar energy production requires an accurate estimation of surface solar irradiance. A forecast of surface solar irradiance allows estimate the energy production, i.e., to minimize the fluctuations in the electric grid supply. In this work a numerical weather forecast model provides surface solar radiation estimations over a coastal region with changeable weather and typically partially cloudy days. As direct model output over specific locations is not always accurate, an ensemble forecast with four members (including time and spatial combinations) is proposed and validated against measurements from three different weather stations in that region, along 1-year. From this validation, statistics show that the best forecast is achieved with the ensemble member obtained by the linear combination of the solar radiation forecasted for the corresponding time (H), one hour before (H-1), and one hour after (H+1). No improvement is obtained by spatial solar radiation combinations. Also, this ensemble forecast provides an estimation of the solar radiation uncertainty, which can be useful to provide flexible energy production forecasts.