Power Forecasting in Photovoltaic System Using Hybrid ANN and Wavelet Transform Based Method

Solar energy is a sustainable, renewable energy which is a part of latest industry standards of operation in line with industry 4.0. Solar power variability leads to fluctuation and uncertainty in Photovoltaic (PV) output power. It is a significant issue with regard to the high penetration of PV power generation. The solar irradiance is affected by weather conditions, and varies with geographical locations. Accurate PV power output forecasting is essential for the planning and scheduling alternate sources of conventional power. In this paper we propose a frequency domain approach for forecasting of short-term PV output power. The wavelet transform allows identification of periodic components with time localization, whereas the Artificial Neural Network (ANN) technique allows us to model the non-linearities in the PV time series. In this paper, PV power data for the city Bareilly, Uttar Pradesh is forecasted. Numerical simulations show that the proposed forecasting method for PV power output, shows a significant increase in accuracy over other similar methods. The root Mean Square Error, Mean Absolute Error for the proposed method are also calculated and compared with state-of-the art methods for PV power forecasting.