Application Of Neural Networks For Predicitng The Optimal Sizing Parameters Of Stand-Alone Photovoltaic Systems

Abstract

A database of daily global solar radiation obtained from Nigeria Meteorological Agency and National Aeronautics and Space Administration (NASA) for 45 locations in Nigeria were used to determine the optimal sizing parameters of a stand-alone photovoltaic system. Analytical technique was used for determining the optimal parameters (Aopt, Eopt). Subsequently, an artificial neural network was used for the prediction of the optimal parameters in remote areas based only on geographical coordinates; for this, 41 couples of Aopt and Eopt were used for the training of the network and 4 couples were used for testing of the model. The results show that the unknown validation sizing parameters produce estimation with mean bias error (MBE) of 0:046m 2 , root mean square error (RMSE) of 0:046m 2 and mean percentage error (MPE) of -1.262% for optimal Photovoltaic (PV) array area, and MBE of -0.078kWh, RMSE of 0.085kWh and 0.749% for optimal battery storage capacity. This study thus demonstrates the potential of Artificial Neural Network (ANN) to predict optimal sizing parameters of PVsystems in these locations.