Hybrid Ensemble Solar Irradiance Forecasting in Limpopo, South Africa: A Regional Assessment for Renewable Energy Integration

Accurate Global Horizontal Irradiance (GHI) forecasting is crucial for optimising solar energy systems, particularly in high solar potential regions like Limpopo, South Africa. This study presents a hybrid ensemble model integrating a gradient boosting framework (LightGBM) with a deep learning architecture (BiLSTM-GRU) to enhance GHI prediction accuracy. Meteorological data were obtained from the South African Weather Services (SAWS), while solar radiation data were sourced from the National Solar Radiation Database (NSRDB) for six key regions: Thohoyandou, Polokwane, Tzaneen, Musina, Mokopane, and Makhado. The methodology involves rigorous data preprocessing, feature engineering, and temporal analysis, including lag-based features and Fourier Transform analysis, to capture complex irradiance patterns. Predictions from the base models are combined using Linear Regression as the meta-learner, ensuring a robust ensemble approach. Experimental results show that the hybrid model outperforms the standalone BiLSTM-GRU model while matching LightGBM’s performance, demonstrating the efficacy of ensemble techniques in solar forecasting.