Interpretable surrogate-assisted scheduling of solar thermochemical hydrogen production in hydrogen refueling-integrated energy systems

Abstract

Solar thermochemical hydrogen production converts intermittent solar energy into storable clean fuel, but dynamic control of the process remains a challenge. This study proposes an interpretable surrogate-assisted scheduling framework for meeting users’ diverse energy needs in a hydrogen-centered integrated energy system with CeO 2 -based solar thermochemical hydrogen production and multi-energy supply. An improved ensemble extreme learning machine is developed to predict the energy-saving rate, CO 2 emissions, and operating cost, and bootstrap-derived uncertainty estimation is incorporated into the optimization process to reduce computational effort required for objective-function estimation. The SHapley Additive exPlanations (SHAP) method is employed to quantify the contributions of hourly CeO 2 flow rates and ICE power outputs to system performance, thereby identifying critical scheduling periods and decision variables. The results demonstrate that the proposed framework effectively coordinates intermittent solar hydrogen production, dynamic refueling demand, and multi-energy supply, providing a practical approach to meeting diverse energy needs.