<p>Accurate solar radiation forecasting is essential for the efficient planning, operation, and integration of solar energy systems into modern power grids. This study presents the first comprehensive benchmark analysis of solar radiation forecasting in Egypt, utilizing an unprecedented 12-year meteorological dataset collected from 56 monitoring locations distributed across diverse climatic zones. Leveraging the PyCaret automated machine learning framework, a wide array of regression models was systematically evaluated, encompassing ensemble methods, linear regressors, and boosting algorithms. The experimental results, assessed through extensive 10-fold cross-validation, identified ensemble tree-based models particularly Extra Trees Regressor, Random Forest Regressor, and LightGBM as the most effective predictors. The Extra Trees Regressor achieved the best overall performance, recording the lowest error metrics (MAE: 2.0190, RMSLE: 0.0195, MAPE: 0.0128) and an exceptional R² score of 0.9996, while maintaining efficient training times. The models demonstrated strong generalization capabilities across Egypt’s diverse regions, as confirmed by residual analysis and prediction error plots, underscoring their reliability for real-world forecasting tasks. Furthermore, the study highlights the deployment readiness of these models through automated workflows and lightweight pipelines, enabling seamless integration into cloud platforms, IoT devices, and solar energy management systems. By providing a transparent, reproducible, and scalable benchmark, this work not only reinforces the suitability of ensemble learning approaches for high-fidelity solar radiation prediction in Egypt but also lays the groundwork for future research in data-scarce regions, addressing model robustness, continuous learning, and hybrid modeling approaches.</p> Graphical Abstract <p></p> <p><b>Graphical Abstract Description</b>: This study presents a comprehensive, nationwide framework for high-accuracy solar radiation forecasting across Egypt using automated machine learning techniques. The graphical abstract illustrates the full modeling pipeline, beginning with the collection and preparation of a 12-year meteorological dataset from 56 diverse locations, capturing key atmospheric and irradiance parameters. The dataset is processed through PyCaret’s AutoML system, which automates key tasks including data preprocessing, feature transformation, model comparison, and hyperparameter tuning. The visual representation highlights the evaluation of a wide range of regression models, with ensemble learning methods, particularly Extra Trees, Random Forest, and LightGBM, demonstrating the highest predictive accuracy and reliability across spatial and temporal dimensions. These results are validated using 10-fold cross-validation, residual analysis, and hold-out testing for both geographic and seasonal variation. The final segment of the graphical abstract illustrates the deployment phase, showcasing how the optimized models can be integrated into real-time solar forecasting applications. These include smart grid systems, cloud-based platforms, and IoT devices designed for energy management in both urban and remote environments. The graphical abstract efficiently conveys the study’s core contribution: a scalable, automated, and reproducible machine learning workflow that transforms raw meteorological data into accurate solar radiation forecasts. This framework provides critical support for renewable energy planning and operational decision-making, particularly in data-scarce or climate-sensitive regions.</p>