A machine learning-assisted prediction of potential biochar yield subjected to the physicochemical properties of biomass

The thermochemical process of slow pyrolysis is commonly employed to convert organic waste into biochar. This study utilised Pearson correlation for feature selection and applied seven machine learning models, such as Random Forest Regression, Decision Tree Regression, Bagging Regression, Extreme Gradient Boost Regression, Support Vector Regression, Gradient Boost Regression, and Extra Tree Regressor, to predict biochar yield based on 13 selected variables. Partial dependence analysis provided insights into the relationships between feature variables. RFR, GBR, XGBR, BaggingR and ETR demonstrated superior performance, achieving R² values above 0.90 for the test set and over 0.90 for the training set. Partial dependence plots revealed that pyrolysis conditions have a more substantial influence on the biochar yield than biomass components. Interestingly, temperature was found to be a crucial element in maximising the biochar yield. This study signifies the effectiveness of integrating approaches based on machine learning frameworks to optimise thermochemical conversion processes, demonstrating their potential for efficient pyrolysis testing and predictive process modelling, thereby promoting sustainable biochar production.