Predicting daily mean solar power using machine learning regression techniques

Daily mean solar irradiance is the most critical parameter in sizing the installation of solar power generation units. The average solar irradiation on a specific location can help predict the amount of electricity that will be generated through solar panels and an accurate forecast can help in calculating the size of the system, return on investment (ROI) and system load measurements. To predict the mean solar irradiation Wh/m
²
various regression algorithms have been used in conjunction with various parameters related to solar irradiance. In this paper we present a comparative analysis of forecasting through artificial neural networks (ANN) against the standard regression algorithms. Furthermore, we show that incorporation of azimuth and zenith parameters in the model significantly improves the performance.