Machine learning-guided prediction of energy storage performance of carbon cathode materials for zinc-ion hybrid capacitors

Aqueous zinc-ion hybrid capacitors (ZIHCs) have emerged as promising candidates for energy storage systems due to their excellent performance and environmental advantages. A key challenge in enhancing the energy storage capability of ZIHCs lies in the design of high-performance carbon cathodes. The current advancement of computational techniques, particularly machine learning (ML) and deep learning (DL), has provided highly promising strategies for material design and performance prediction. In this work, we investigated three ML models and three DL models to predict the specific capacitance (Cs) of carbon cathode-based ZIHCs. Among these, LightGBM demonstrated remarkable prediction accuracy, achieving an exceptionally high coefficient of determination (R²) of 0.986 and a remarkably low root mean square error (RMSE) of 4.88 mAh g⁻¹. Furthermore, Shapley Additive exPlanation (SHAP) and partial dependence plots (PDP) were applied to identify key carbon material properties, that is, pore volume, nitrogen content, and specific surface area as critical factors influencing Cs. Importantly, biomass and metal organic framework (MOF) derived carbon materials were synthesized as cathodes for ZIHCs to validate the reliability of the ML results. This work innovatively integrates ML techniques to predict the energy storage performance of ZIHCs, offering practical guidance for the design and optimization of carbon cathode materials in advanced ZIHCs.