Synthetic generation of hourly photovoltaic power forecasts using Bayesian Regularised Neural Networks

Accurate day-ahead photovoltaic (PV) power output forecasting techniques are important for grid and plant operators. In this study, a methodology for PV power forecasting is proposed based on a synthetic generation of multiple Bayesian regularised neural network (BRNN) models. Specifically, different BRNN models were developed for every hour of each day within the power production period. The final PV power production forecasting was provided by combining the outputs of all the developed models. The accuracy of the developed model was experimentally validated by deploying it to a test system installed in Cyprus. The day ahead power production forecasting performance accuracy of the proposed methodology in terms of normalised root mean square error and mean absolute percentage error was 5.80% and 4.49%, respectively.