Optimizing biochar yield and composition prediction with ensemble machine learning models for sustainable production

Abstract

Biochar production from organic waste can reduce fossil fuel reliance and combat climate change, but current models are computationally demanding and have limited accuracy. The study creates four machine learning models using multiple linear regression, decision trees, Adaboost regressors, and bagging regressors, trained on a dataset of pyrolysis tests. The results show that the data-driven models have significantly higher predictive accuracy than existing models, with an R² of up to 0.96. The Bagging Regressor (BR) demonstrated superior efficacy compared over the MLR, AR, and DT models across all eight output parameters, with R2 values of 0.94, 0.93, 0.93, 0.94, 0.95, 0.90, 0.92, and 0.96 for Biochar Yield, Fixed Carbon, Volatile Matter, Ash, and ultimate composition parameters (C, H, O, and N), respectively. The study developed a data-driven model to predict Biochar yield and compositions, enhancing production processes and promoting sustainable farming practices.