Intra-day Solar Irradiance Forecasting Based on Artificial Neural Networks

Abstract

Accurate solar irradiance forecasting is important for improving forecasting precision of photovoltaic (PV) power. In this study, an intra-day (i.e. 1 to 6 hours ahead) machine learning model based on an artificial neural network (ANN) was implemented for forecasting the intra-day incident solar irradiance (GI). The methodology included the implementation of the optimal ANN topology which was trained and validated on historical yearly datasets. The forecasting results demonstrated a normalised root mean square error (nRMSE) in the range of 4.23% to 9.51%. The lowest nRMSE of 4.23% was achieved for the hour-ahead forecast while the highest nRMSE of 9.51% was observed when forecasting at a horizon of 6 hours ahead. Finally, the mean absolute percentage error (MAPE) varied from 4.10% to 8.19% for the 1 hour to 6 hours ahead forecasts respectively.