SREF: A data-driven stacked ensemble framework for solar radiation prediction using meteorological features

The accurate forecasting of the solar radiation is critical in the effective planning, performance, and reliability of the solar energy systems. Nevertheless, most current models offer difficulty of expressing sophisticated non-linear correlations and time dependence in meteorological data, thereby reducing the accuracy of predictions made. This paper suggests a new ensemble-based method named the Solar Radiation Ensemble Forecaster (SREF) that is constructed on the interaction of machine learning and deep learning methods to enhance the prediction. The SREF framework uses the base learners of Random Forest, Gradient Boosting and Multilayer Perceptron and the meta-learner is Ridge Regression to efficiently combine the different learning patterns. Moreover, deep learning models of time-series (LSTM, BiLSTM, and ConvLSTM) are also used to compare the sequential prediction performance. The analysis of the methodology is based on the Solar Prediction dataset of 32,686 hourly meteorological measurements in Alabama, USA. Experimental findings indicate SREF performance is much better and with a Mean Absolute Error (MAE) of 47.57 W/m 2 and R 2 of 0.94 compared with individual models and the current literature. The proposed SREF model offers a cost-effective, scalable, and accurate solution for solar radiation forecasting. Practical Application This study introduces a solar radiation forecasting framework, SREF, that is (1) scalable and lightweight for built environment professionals to optimize photovoltaic system performance and (2) to use in energy planning. The framework integrates ensemble models with domain informed feature engineering to provide high prediction accuracy at limited computational expense. Therefore, the growing need for this type of solver is particularly valuable for real time applications in smart cities, building energy management systems and in resource constrained environments. A reproducible pipeline enables better grid integration strategies, promotes sustainable design and informs adaptive energy dispatch to allow better, data driven decision making in the built environment.