Improving Solar Power Forecasting Using Air Quality and Weather Data with Ensemble Learning

Solar power generation is non-continuous and very much dependent on atmospheric conditions such as weather variability and pollution levels. Forecasting solar power accurately is important for effective energy planning, grid stability, and large-scale integration of PV systems. Traditional forecasting approaches (sensor-based) are effective, yet costly, depending on the location. This study proposes a data-driven solar power forecasting framework that integrates weather parameters and Air Quality Index (AQI) indicators using ensemble machine learning models. The study focuses on existing methodologies that use meteorological variables, air pollutant concentrations (PM2.5, PM10, NO2, SO2, CO2, O3) and solar irradiance indicators for solar power forecasting. This paper reviews machine learning approaches, including regression models, ensemble learning, and stacked architectures, aligning with prediction accuracy, scalability, and practical deployment. This paper also emphasizes the impact of integrating AQI parameters into solar prediction models and demonstrates how effectively the combined weather-AQI features enhance the accuracy of solar power predictions. This study suggests employing a stacked ensemble model based on tree-based learners to align with the trends identified in the literature. The review concludes by identifying key research gaps and future directions, highlighting the potential of AQI-aware, data-driven models for cost-effective, scalable, and reliable solar power forecasting in smart energy systems.