Hourly solar radiation forecasting on SAURAN network datasets using deep learning method: La Reunion and Durban cases study

International audience The purpose of this article is to describe the data pretreatment (smoothing, normalizing) and present hourly forecasting method using XGBoost deep learning tool on the global horizontal irradiance (GHI). This method will be applied on two sites with different typical meteorological profiles. An estimation of prediction skills will be given and discussed against classical persistence model.