Very short term forecasting of the Global Horizontal Irradiance using a spatio-temporal autoregressive model

The integration of massive solar energy supply in the existing grids requires an accurate forecast of the solar resources to manage the energetic balance. In this context, we propose a new approach to forecast the Global Horizontal Irradiance at ground level from satellite images and ground based measurements. The training of spatio-temporal multidimensional autoregressive models with HelioClim-3 data along with 15-min averaged GHI times series is tested with respect to a ground based station from the BSRN network. Forecast horizons from 15 min to 1 h provided very promising results validated on a one year ground-based pyranometric data set. The performances have been compared to another similar method from the literature by means of relative metrics. The proposed approach paves the way of the use of satellite-based surface solar irradiance (SSI) estimation as an SSI map nowcasting method that enables to capture spatio-temporal correlation for the improvement of a local SSI forecast. •HelioClim-3 maps time-series allow the forecast of Global Horizontal Irradiance.•Statistical learning of the main meteorological trends leads to 9% RMSE improvement.•Important irradiance variations (clouds) can be forecasted up to 1 h ahead.