A Novel Approach for Solar Radiation Prediction Using Artificial Neural Networks

This article presents a novel solar radiation prediction approach using artificial neural networks. The developed model predicts three meteorological variables using sunshine ratio, day number, and location coordinates. These meteorological variables are solar energy, ambient temperature, and relative humidity. However, three statistical values are used to evaluate the proposed model. These statistical values are mean absolute percentage error, mean bias error, and root mean square error. Based on the results, the developed model predicts accurately the three meteorological variables. The mean absolute percentage error, root mean square error, and mean bias error in predicting solar radiation are 1.3%, 5.8 (1.8%), and 0.9 (0.3%), respectively. While the mean absolute percentage error, root mean square error, and mean bias error values for ambient temperature prediction are 1.3%, 0.4 (1.7%), and 0.1 (0.4%). In addition, the mean absolute percentage error, root mean square error, and mean bias error values in relative humidity prediction are 3.2%, 3.2, and 0.2.