Machine learning assisted analysis: inorganic catalyzed hydrothermal carbonization to enhance biomass carbon stability.

Abstract

Hydrothermal carbonization (HTC) is an effective method for sustainable waste conversion and carbon fixation. The carbon stability of solid-derived hydrochar (HC) is a key factor in determining the quality of HTC. This study enhanced carbon recovery by 8% using three inorganic layered double hydroxides (LDHs) as catalysts. LDH addition significantly altered dissolved organic matter (DOM) components and structures in HC. The increase in double bond equivalents (DBE-O) indicated carbon skeleton unsaturation, suggesting DOM mainly comprises compounds with higher unsaturation and lower oxidation. Machine learning (ML) has found that DBE-O is closely related to HC carbon content and stability. Feature importance and SHAP analysis have improved the interpretability of the model. In this study, LDH catalyzed HTC to alter the composition and structure of DOM, resulting in improved carbon stability. ML further revealed the mechanism of DBE-O modification of DOM, thereby enhancing carbon recovery and fixation.