Predictive modeling for multifaceted hydrothermal carbonization of biomass

Hydrothermal carbonization is a widely recognized process for converting biomass into biomass charcoal, aimed at reducing and replacing the use of natural resources while also mitigating CO2 emissions responsible for global warming. However, the complexity of the hydrothermal carbonization process presents challenges in elucidating the interplay of different features. Thus, this study employs extreme gradient boosting machine learning algorithm to predict the values of six target outputs namely; yield, calorific value, ash, carbon, hydrogen, and oxygen content of hydrochar. Notably, the study leverages a rich dataset comprising over 1000 data points and 16 input features. The results indicate high correlations, with R-2 ranging from 0.84 to 0.97. Furthermore, the study investigates the impact of input features on all six outputs through the application of Shapley values and SHAP dependence plot techniques, highlighting its novelty and contribution to the field.