Efficient Machine Learning Models for Solar Radiation Prediction Using Ensemble Techniques: A Case Study in Low-Rainfall Arid Climates

Solar energy is a green energy source that can be used to generate electricity, making it one of the oldest existing renewable energy sources. However, solar energy values are not always available in many areas far from urban cities, due to the high cost of purchasing and maintaining measuring instruments. In these areas, the only way to know the values of solar radiation is to predict these values. To address this problem, we have designed an efficient model capable of forecasting hourly solar radiation a climate with a combination of arid and humid conditions using real-time data from meteorological variables measured by a Meteorological Station located in a university area isolated from the central city. For the development of our model, 11 meteorological variables were collected from January to December 2023. From this, dimensionality reduction was carried out using the Principal Component Analysis (PCA) technique, obtaining the 8 most appropriate representative variables for the prediction of solar radiation using different Machine Learning (ML) models. To ensure the reliability of our model, we compared the performance metrics of several ML models. Finally, ensemble techniques were applied to enhance prediction accuracy by combining various ML algorithms. The best results were obtained using the Stacking Regressor (SR), achieving an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i><sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.92, RMSE of 47.30 W/<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">m</i><sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, and MAE of 18.27 W/<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">m</i><sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, demonstrating high predictive performance in the target climate region.