Evaluation of Perceptron Learning Algorithm for Day-Ahead Photovoltaic Power Forecasting

The output power of a photovoltaic system depends on weather conditions. It’s well established that the power output efficiencies from solar photovoltaic (PV) systems is very much dependent on weather parameters and deposited solid particles on the panel surface. This paper aims to address the demand for a more accurate prediction of their operational performance in order to maximise their contribution to the overall energy mix. This research applies Perceptron Learning Algorithm to forecast the day-head PV generation capacity whilst considering future and historical meteorological parameters and system data. Model evaluation results, for an ensuing 24-hour ahead period of interest, achieved forecast accuracy in excess of 90% compared to the recorded power from an installed campus-based micro-grid array.