Sorption-enhanced steam methane reforming parameter analysis and performance prediction of ensemble learning methods using improved drag model

Abstract

This study innovatively develops an ensemble learning method to predict the hydrogen yield of sorption-enhanced steam methane reforming (SE-SMR). Firstly, an experimental validated computational fluid dynamics model of SE-SMR is constructed based on an improved drag model. The H yield under different parameters is simulated for creating a database. Then, temperature, pressure, gas velocity and steam to carbon ratio are studied to obtain optimal conditions for methane conversion and gas production. Finally, XGBoost ensemble prediction model is employed to predict H yields and compared with AdaBoost and Bagging models. Results show that an increase in gas velocity and pressure leads to a decrease in methane conversion rate with lower hydrogen production efficiency. In addition, the preferred temperature for the reaction is about 998 K, and the S/C of 4 is an economically optimal choice. XGBoost prediction model predicts H yields with the highest accuracy. This study offers directional guidance for future improvements in hydrogen production efficiency.