Photovoltaic (PV), Forecasting, Cloud Index, Surface Solar Irradiance, Meteosat
Summary / Abstract:
At intraday horizon, photovoltaic (PV) electricity production forecasting is crucial for a massive and safe integration of this energy into the grid. PV production is mainly driven by incident solar irradiance at ground level. This parameter cannot be accurately determined as a numerical weather prediction model (NWP) output at this time-horizon. Forecast methods using images from meteorological geostationary satellites are a proven alternative. Such methods consist in assessing cloud motion and in extrapolating cloud data from current satellite images to predict future cloud patterns. Two approaches have been implemented to derive cloud motion, using respectively image processing techniques and wind data provided by a NWP model. Comparative benefits of the two approaches have not fully been quantified. Therefore, we propose a benchmark of these two approaches. We implemented these algorithms and compared them with ground measurements. Results shows that cloud motion vectors derived from satellite are more pertinent within a short time horizon. After three hours, external wind data permit to reduce the bias between forecasts and in-situ measurements.