Application of Artificial Neural Networks to Predict Daily Solar Radiation in Sokoto

Abstract

This paper presents an application of Artificial Neural Networks (ANNs) to predict daily solar radiation in Sokoto (lat. 13° 03’N, log. 5° 14’E). The mean daily data for sunshine hours, air temperature, and relative humidity data, along with day number and month number for period of 3years were selected as the input variables to the ANN models. The ANN models and regression models based on Hargreaves-Samani and Angstrom Prescott approaches were tested for the study area. The ANN model indicated a reasonably strong predictive power, where mean bias error (MBE), root mean square error (RMSE), mean percentage error (MPE), and R2 values were found to be 0.063, 0.164, -2.489%, and 0.965 respectively for the training, and 0.103, 0.288, -4.177% and 0.959 respectively for the testing. These results showed that artificial neural network could give reasonably good estimation of global solar radiation in the study area and other locations with similar climatic factors.