New results in forecasting of photovoltaic systems output based on solar radiation forecasting

Accurate short term forecasting of photovoltaic (PV) systems output has a great significance for fast development of PV parks in South-East Europe, as well as in the case of Romania. Our approach on solar radiation forecast is based on two methods: autoregressive integrated moving average and artificial neural network. We have analyzed the daily solar irradiation variability and defined four synoptic situations to include the influence of cloudiness changes. Decadal variations of global solar radiation were also considered for long term forecast. The results were obtained using a database from Bucharest/Afumati Meteorological Station. We have developed an accurate forecasting model for a PV system's power output based on solar radiation forecasting results. By using complete datasets and including meteorological parameters such as cloudiness, relative humidity, air temperature, atmospheric pressure, and sunshine duration, as input for our model, we have managed to minimize forecasting errors and to obtain a more accurate forecast of the power output for the analyzed demo PV system.