Artificial Neural Network-Based Modeling for Monthly Average Global Solar Radiation Estimation in South East Nigeria

Abstract

Reliable estimation of global solar radiation (GSR) is essential for the proper planning and performance evaluation of solar power systems. This research presents a neural network–based approach for estimating monthly mean GSR across South East Nigeria, covering Enugu, Imo, Ebonyi, Anambra, and Abia States. Twenty years of meteorological records (2005–2025), including air temperature, relative humidity, and wind speed, were utilized for model development. A feedforward multilayer perceptron trained using the backpropagation technique was implemented for the prediction task.
Model evaluation indicates good agreement between predicted and observed values, with a mean absolute percentage error (MAPE) of less than 5%, a coefficient of determination (R²) of 0.95451, and a root mean square error (RMSE) of 0.32 MJ/m²/day. The findings demonstrate that the developed model can effectively estimate solar radiation in tropical locations where measured solar data are scarce. This approach can support informed decision-making in the design and expansion of solar energy projects within the region.