Optimizing hydrogen-rich gas production by steam gasification with integrated CaO-based adsorbent materials for CO2 capture: Machine learning approach

The sorption-enhanced steam gasification of biomass with an integrated carbon dioxide (CO 2 ) capture is a promising process for hydrogen production. By using machine learning (ML) approaches to reduce the amount of CO 2 , this study aimed for prediction and optimization of gaseous products with a higher concentration of hydrogen. ML schemes are applied to hydrogen-rich syngas produced through calcium oxide-based adsorbent. Four predictive techniques are applied on the intrinsic constituents of biomass, adsorbents properties, steam gasification ratio, and temperature to predict the concentrations of hydrogen and CO 2 . The accuracies of ML models demonstrated high feasibility of ML to predict the hydrogen and CO 2 with R-squared (R 2 ) of 0.92; and 6.77 to 7.44 vol% of root-mean-square error (RMSE), respectively. Support vector machine (SVM) is optimized by tuning training data size and radial basis kernel ( rbf ) function. Also, the single and multi-objective(s) genetic algorithm approaches optimized the value of hydrogen concentration by Max f max ( H 2 ) by ∼84 and 88 vol%, respectively. Sensitivity analysis showed the fixed carbon/volatile matter, oxygen content, adsorbent/steam to biomass ratios, and gasification temperature in the range of 7–20 vol% of mean absolute percentage error (MAPE) on hydrogen content. The optimized input sets for the ML modelling procedure improved the hydrogen concentration by within 5 vol%. The results indicate a high proficiency of the ML models in accurate prediction of hydrogen gas in the gasification process. • Sorption-enhanced steam gasification of biomass evaluated by ML models. • ML models KNN, SVM, RF, and DT are used to predict hydrogen and CO 2 concentrations. • SVM and RF exhibited 0.91 to 0.92 of R 2 ; and 6.77 to 7.44 vol% of RMSEs. • Sensitivity analysis and genetic algorithm maximize hydrogen concentration by CO 2 reduction.