A Hybrid Online Model for Short-Term Solar Irradiance Forecasting using Minimal Data

Solar photovoltaic (PV) systems have emerged as a prominent renewable energy source (RES) especially in Asia. However, the intermittent and unpredictable nature of solar irradiance poses a significant challenge. Hence, accurate forecasting of solar irradiance is essential to ensure reliable and efficient operation solar PV systems. In this paper, a hybrid model is developed for short-term solar irradiance forecasting. Offline and online data are obtained from a local remote solar PV condition monitoring (CM) system, where real-time environmental details as are used as inputs to achieve results on the proposed online prediction model. Different algorithms were used to perform univariate single step irradiance forecasting based on only historical irradiance data, and subsequently multivariate inputs and time features for the proposed online prediction model were used. Based on the results, it is clear that online method outperforms the offline prediction results, with about 25% improvement on RMSE performance metrics.