Paper presented to the 3rd Southern African Solar Energy Conference, South Africa, 11-13 May, 2015.
Some small territories, like islands and isolated areas, actually experience a high penetration rate of PV inside a small electricity grid. Moreover, the high amplitude fluctuations of PV outputs can destabilize the grid stability. In order to avoid the risk of blackout, some countries set up regulatory limits of PV integration. In this context, the forecasting of the PV output is necessary for the supply-demand balance and for the increase of the penetration rate of PV. Previous works on this topic were mainly done for large-scale continental grids. Due to the small scale of the climatic phenomena, forecasting the solar irradiance in insular territories addresses new issues. In order to cope with specific plant operations, forecasts must be provided with different granularities and horizons. In this work, we will focus on day ahead forecasts with an hourly granularity. Dayahead forecasts are produced for scheduling of resources and commitment of units of production. This paper presents a comparison of two post processing models. A Model Output Statistics (MOS) and an Artificial Neural Network (ANN) are applied to the IFS (Integrated Forecast System) forecasts for the insular site of Saint-Pierre in Reunion Island. The small scale of the climatic phenomena requires to set up these post processing methods differently than in the continental areas.