Using machine learning to predict biochar yield and carbon content: Enhancing efficiency and sustainability in biomass conversion
XU Qing-sheng, Long Du, Rui Deng
Abstract
The production of biochar through pyrolysis of biomass is expected to reduce dependence on traditional energy sources and mitigate global warming. However, current predictive models for biochar yield and composition are computationally intensive, complex, and lack accuracy for extrapolative scenarios. This study utilized machine learning to develop predictive models for biochar yield and carbon content based on pyrolysis data from lignocellulosic biomass. Assessing the importance of input features revealed their significant role in predicting biochar properties. The findings indicate that eXtreme Gradient Boosting (XGBoost) algorithms can accurately forecast biochar yield and carbon content based on biomass characteristics and pyrolysis conditions. This research contributes new insights into understanding biomass pyrolysis and enhancing biochar production efficiency.