oa Nowcasting of Solar Irradiance with Autorregressive Models in Qatar

Solar radiation in near-real time is of high importance for the integration of PV electricity into the Grid and operation of solar plants. Solar radiation is characterized by a high temporal and spatial variability mainly due to the effect of clouds. In the case of solar energy plants with storage energy system and CSP plants in general, its management and operation need reliable predictions of solar irradiance in the very short term or nowcasting. Solar forecasts are also needed in on-grid applications of solar energy where sudden changes in solar irradiance can trigger unacceptable voltage deviation, if not properly managed. Solar radiation forecasting beyond six hours and up to several days ahead is based on Numerical Weather Prediction (NWP) models for supplying hourly or daily forecasted values. However, statistical and machine learning techniques have shown to be effective for solar forecasting at higher-resolution timescales (e.g. minutes). The aim of this work is to show the effectiveness of autoregressive models for the prediction of solar radiation in the short term (nowcasting) in the Middle East. The temporal steps of the predictions are 1, 5 and 10 minutes, and the temporal horizon is 15 steps ahead, so that the ensuing predictions includes 15, 45 and 150 minute ahead forecasts, respectively. We describe the models and show validation results for one year of ground measured global horizontal irradiance in Doha (Qatar).