A Comparative Study of Ensemble Learning Models for Accurate Solar Irradiance Forecasting

Accurate estimation of solar radiation is very important in energy planning and the integration of renewable energy infrastructures into power grids for most places where the data may be sparse. This paper investigates advanced machine learning technologies specifically, ensemble methods such as XGBoost, LightGBM, CatBoost, NGBoost, AdaBoost, Gradient Boosting, Random Forest, and SGD Boost—to improve the prediction of solar radiation. It shows the performance of the above-mentioned models to forecast solar irradiance from meteorological factors using data from the Hawaii Space Exploration Analog and Simulation (HI-SEAS) weather station. It has been shown that ensemble learning boosts the forecast accuracy considerably, especially for ensemble learners optimized for categorical data and probabilistic predictions. This shows how machine learning can be potential and support the global shift to sustainable energy with reliable forecasts of solar radiation in data-scarce regions and offering valuable insights for optimizing renewable energy deployment.