Machine learning - driven solar forecasting in dust-prone regions for sustainable energy systems

• We present a data-driven framework that enhances solar energy forecasting by incorporating dust-related parameters into machine learning models, overcoming traditional limitations that often neglect dust storms, and improving accuracy under high dust load conditions.f • Our study analyzes dust patterns in dust-prone regions, emphasizing the diurnal and seasonal effects on solar power output, contributing to better energy system modeling. • We compare several machine learning models, including LSTM-GRU hybrids, to capture the complex relationship between dust and solar irradiance, with results validated by rigorous statistical tests. • This forecasting framework is adaptable to various dust-prone and polluted regions globally, aiding energy operators, researchers, and policymakers in enhancing solar forecasting accuracy and grid stability. • In line with the UAE's Vision 2050, our research advances solar energy predictability, facilitates renewable energy integration, and strengthens climate resilience. This research focuses on improving solar energy forecasting in dust-affected regions such as the UAE, where frequent dust storms reduce photovoltaic (PV) efficiency by scattering and absorbing sunlight. Many existing models overlook the impact of dust events, leading to inaccurate forecasts during such conditions. To address this, the study develops machine learning models—including LSTM, GRU, and hybrid LSTM-GRU architectures—that incorporate solar, weather, and dust-related features. The models were evaluated across multiple forecasting horizons (1, 6, 12, and 24 hours), demonstrating that including dust-related variables significantly enhances prediction accuracy, particularly for short-term forecasts. Temporal and seasonal analyses revealed that dust events, most frequent in the late afternoon and early spring, correlate with substantial drops in solar power output. The LSTM model consistently outperformed the others, achieving a Mean Absolute Error (MAE) of 0.018034 for a 1-hour horizon when dust features were included. Statistical tests confirmed that dust events significantly affect forecasting accuracy, reinforcing the importance of dust-related features for reliable predictions. This research contributes to optimizing PV power generation in challenging environments, supporting sustainable energy systems and decarbonization efforts. It also offers insights for further model refinement and the inclusion of additional environmental variables.