Enhancing predictive models for steam gasification: A comparative study of stoichiometric, equilibrium, data-driven, and hybrid approaches

Steam gasification offers a pathway to generate synthesis gas (syngas) rich in hydrogen (H 2 ), a crucial element in efforts to decarbonize and mitigate greenhouse gas emissions. However, the intricate web of reactions involved in the process demands predictive tools to enable its large-scale application. While models based on stoichiometry, chemical equilibrium, and data algorithms have made strides, previous works lack comprehensive comparative studies on their efficacy and adaptability. This study addresses this gap by developing and juxtaposing four models: stoichiometric, equilibrium-based, data-driven, and a hybrid approach to forecast steam gasification products against experimental data gleaned from a systematic literature review. Among these models, the hybrid variant emerges as the most accurate in predicting syngas composition, boasting an average root mean square error (RMSE) of 5.63 and an average R 2 of 0.59. Moreover, it yields predictions for tar, char, and gas with respective RMSEs of 42.79 g/Nm 3 syngas, 72.99 g/kg biomass, and 0.33 Nm 3 syngas/kg biomass. Notably, the robust validation process of this model enhances its versatility while maintaining commendable prediction accuracy compared to the existing literature. Future enhancements could entail integrating advanced kinetic and equilibrium expressions and incorporating fresh experimental data into the training phases of data-driven models. • Steam gasification was modeled by using different modeling techniques. • Stoichiometric, equilibrium, and machine learning models were validated. • A hybrid model combining the three approaches showed improved performance. • Hybrid model is accurate in a wide range of feedstocks and configurations. • The model predicts accurately gas composition, tar, and char yields.