Decoding Hydrogen Demand Patterns: Exploring Key Drivers Using SHAP and Ensemble Learning

Establishing a good hydrogen infrastructure requires precise planning and construction, beginning with the determinations of the future market demands of hydrogen. Navigating through the complexity of hydrogen demand datasets requires accurate forecast models. To overcome this difficulty, this study presents an ensemble regression model that combines Ridge Regression and Linear Regression. The model achieved best set of metrics of Mean Squared Error (MSE) of 5.07888×10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−14</sup>, a Root Mean Squared Error (RMSE) of 2.25×10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">7</sup>, a Mean Absolute Error (MAE) of 8.20602×10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−8</sup>, and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score of 100% on the hydrogen demand dataset with an 80-20 train-test split, demonstrating exceptional robustness and precision in forecasting hydrogen demand across various scenarios. SHapley Additive exPlanations (SHAP) were used to further improve the model’s interpretability by providing indepth explanations of the various elements influencing demand projections. In addition to increasing prediction accuracy, this method facilitates better strategic planning and decisionmaking, which advances the development of a sustainable hydrogen economy and helps mitigate the effects of climate change.