Forecast of photovoltaic production using NN for the realization of a nautonomous PV system within a tertiary building

Abstract

The installation of a solar system meets a specific need on a given site. It should only be considered after a study of photovoltaic (PV) system production in order to have a fairly precise idea of the return on investment of the installation. This study aimed to forecast PV Production using Neural network (NN) for the realization of a stand-alone PV installation within a tertiary building, Department of electronics at Amar Telidji, Laghouat . We set out to forecast PV production in the short and long terms using 10 months (hourly steps) weather data. MATLAB programs were needed to treat the data for training. The first step was the calculation of the energy needs of the location. Then we did the sizing with two methods: by calculation and with the PVsys (a software that offers a detailed sizing reports and simulation) using PVgis ( offers sunshine maps irradiation in kWh / m² ).Next , We enriched our work with an economical features. Finally, We used NN to forecast PV production. The results obtained show that NN are a very effective means for PV forecasting, in the short and even in the long term with a fairly low error rate.