To date, PV yield prediction software is a product used for the most part by utility grade PV power plants for selling onto the next day market. The state-of-the-art in this field comprises different forms of meteorological forecasting that is applied to the PV plant in a simulation program. For a PV simulation program to be effective, the plant configuration must be known and this is commonly not possible for a neighbourhood collection of individual PV systems, each designed by a different installer that optimises different physical realities. Currently, a vast amount of research institutions and organizations are focusing on ways to improve PV production forecasts based on both measured data and numerical weather prediction (NWP) models. In this direction, strategies of post-processing NWP model results with either measured data or with stochastic learning algorithms are used in order to correct systematic deviations.
Along these lines, the project INFORPV has been initiated in order to enhance the accuracy of day and hour-ahead PV production forecasts for both industrial grade and small residential PV systems, aggregating all systems in a distribution area and the addition of a PV grid location software platform that will clearly define the areas of operation for each managed section of the grid. In this respect, the proposed system presents the first worldwide attempt to achieve PV production forecasting accuracies less than 5 % RMSE for single plants and day-ahead predictions less than 4.5 % RMSE for aggregated PV production forecasts while in parallel allowing active plant control for utilities.