Very short-term photovoltaic power forecasting using uncertain basis function

Solar photovoltaics (PV), one of the most promising and rapidly developing renewable energy technologies, has evolved towards becoming a main renewable electricity source. It is termed variable energy resources since solar irradiance is intermittent in nature. This variability is a critical factor when predicting the available energy of solar sources. Capital and operational costs associated with solar PV implementation are highly affected when inaccurate predictions are carried out. This paper presents a new forecasting model for solar PV by utilizing historical inter-minute data to outline a short-term probabilistic model of solar. The proposed methodology employs a probabilistic approach to predict short-term solar PV power based on uncertain basis functions. The PV forecasting model is applied to power generation from a 13.5 kW rooftop PV panel installed on the Distributed Energy, Communications, and Controls (DECC) laboratory at Oak Ridge National Laboratory. The results are compared with standard time series approach, which have shown a substantial improvement in the prediction accuracy of the total solar energy produced.