Predicting Ash-Phase Macroelement Retention from Biomass Lignocellulosic Composition: A Machine Learning Approach

The thermochemical behaviour of biomass-derived minerals plays a central role in determining combustion efficiency, slag formation, and post-process residue characteristics. However, the extent to which bulk compositional traits can anticipate element-specific ash outcomes remains insufficiently defined. In this study, 135 biomass samples encompassing diverse species and origins were characterised by cellulose, hemicellulose, lignin, and proximate properties. Following controlled combustion at 600ĝ€¯°C, ash-phase macroelements were quantified by X-ray fluorescence spectroscopy. A support vector regression model was trained to associate compositional variables with post-combustion mineral content. Prediction accuracy varied markedly across elements. Low relative errors (<3%) were achieved for the prediction of sulphur, phosphorus, and magnesium, of which retention appears thermodynamically constrained and structurally linked to early-stage degradation of lignocellulosic content. However, the prediction of potassium and calcium exhibited to be less consistent with the experimental result, which indicates more complex reaction pathways and migration behaviours during the thermal conversion process. The results of this study highlight a mechanistic foundation for predicting macroelements in ash from composition indicators, thereby aiding in biomass selection and ash management for the combustion process for green energy conversions.