A hybrid predictive and uncertainty-aware framework for renewable hydrogen system efficiency assessment and sustainable energy planning

Abstract Renewable hydrogen is emerging as a critical element in the transition toward low-carbon and sustainable energy systems; however, accurately evaluating system efficiency remains challenging due to complex, nonlinear interactions among technical and environmental factors. This study proposes a hybrid predictive and uncertainty-aware framework for assessing renewable hydrogen system efficiency and supporting sustainable energy planning. The framework integrates Elastic Net and Extra Trees regression models enhanced through meta-heuristic optimization using Sparrow Search and Firefly algorithms to improve predictive performance. Recursive Feature Elimination is applied to identify the most influential variables governing system behavior, while Dempster–Shafer Theory is incorporated to model uncertainty and construct a robust ensemble prediction strategy. The proposed approach is evaluated using a dataset of 2,535 renewable hydrogen system records with spatial attributes derived from documented geographical coordinates. Experimental results demonstrate notable improvements in predictive accuracy, with a 12% increase in R² and a 15% reduction in MAE compared to baseline approaches. Sensitivity analysis indicates that photovoltaic power generation and electrolysis efficiency are the most influential factors affecting hydrogen system performance. Beyond improved prediction accuracy, the framework provides interpretable insights that can support infrastructure development, investment planning, and energy policy design. The proposed methodology demonstrates how hybrid predictive modeling combined with uncertainty quantification can enhance decision-making processes in renewable hydrogen systems and contribute to more reliable low-carbon energy transitions.