Solar radiation forecasting based on ANN, SVM and a novel hybrid FFA-ANN model: A case study of six cities south of Algeria

Abstract

This study was conducted for six cities in southern Algeria, where the accuracy of three models—support vector machines (SVM), artificial neural networks (ANN) and a novel hybrid firefly algorithm-based model (FFA-ANN)—were investigated when estimating global solar irradiation throughout an eleven-year period, utilizing nine input parameters as input data. The goal of our novel suggested a hybrid FFA-ANN model, where we relied on the optimization Firefly algorithm to enhance the ANN model created. Despite the fact that the ANN and SVM models produced promising results, our suggested FFA-ANN hybrid model outperformed the stand-alone ANN-based model using three statistical factors—correlation coefficient, relative root mean squared error and mean absolute percent error—with the best values of (R = 0.9321, rRMSE = 9.35% and MAPE = 6.29%). The findings demonstrated that FFA-ANN was preferable to the optimized SVM and ANN models when forecasting daily global solar irradiation in all zones. Furthermore, after comparing the combinations, the study’s findings showed that the ANN model depended on: Extraterrestrial solar irradiation (H0), declination and average temperature (Tavg) together with relative humidity (RH) as inputs in order to estimate daily sun radiation. Thus, the findings of this study suggest that in regions with dry climates and other places with comparable climates, the created model may be used to estimate daily global solar radiation whenever data is accessible.