Abstract
Solar radiation plays a critical role in the carbon sequestration processes of terrestrial ecosystems, making it a key factor in environmental sustainability among various renewable energy sources. This study integrates two advanced signal processing techniques-empirical mode decomposition (EMD) and ensemble empirical mode decomposition (EEMD)-with machine learning (ML) algorithms, including multilayer perceptron (MLP), random forest regression (RFR), support vector regression (SVR), and ridge regression, to forecast long-term solar radiation. Meteorological data spanning 13 years (2000-2012) from seven locations across India (Bhubaneswar, Chennai, Delhi, Hyderabad, Nagpur, Patna, and Trivandrum) were used for training and testing. The optimal model was identified based on performance metrics, including the highest linear correlation coefficient (R), and the lowest mean absolute error (MAE) and root mean square error (RMSE). The results indicate that EEMD integrated with ML algorithms consistently outperformed EMD-based approaches. Among the ML models evaluated, EEMD integrated with MLP achieved the best performance across all locations, with RMSE = 0.332, MAE = 0.26, and R(2) = 0.99. Furthermore, a comparative analysis with previous studies demonstrated that the proposed approach provides superior accuracy, underscoring its efficacy in solar radiation forecasting.