Machine Learning-Based Prediction of CO2 Emissions from Biomass Solvent Extraction

• Modeling CO 2 Emissions from Biomass Pyrolysis via Solvent Extraction • Ash content, temperature, sulfur, carbon, and oxygen as the most influential predictors • AdaBoost achieved the highest overall accuracy (R² = 0.991, MSE = 1.00, AARE% = 0.89) This study addresses the challenge of accurately predicting carbon dioxide emissions from biomass solvent extraction, a critical refining process strongly influenced by feedstock composition and operational conditions. The objective is to develop and evaluate machine learning models that can reliably estimate CO₂ emissions using physicochemical and process parameters derived from experimental data. A curated dataset comprising 643 records from peer-reviewed publications was assembled, encompassing elemental composition, biochemical composition, and reaction conditions, with CO₂ emissions as the target variable. Eight machine learning models, including Decision Tree, AdaBoost, Random Forest, K-Nearest Neighbors, Ensemble Learning, Convolutional Neural Network, Support Vector Regression, and Multi-Layer Perceptron, were optimized through systematic hyperparameter tuning. Model performance was assessed using 5-fold cross-validation and evaluated with R², Mean Squared Error (MSE), and Average Absolute Relative Error (AARE%). Results indicate that AdaBoost achieved the highest overall accuracy (R² = 0.991, MSE = 1.00, AARE% = 0.89), followed by Ensemble Learning and Decision Tree, while MLP-ANN displayed comparatively poorer generalization. SHAP analysis identified ash content, temperature, sulfur, carbon, and oxygen as the most influential predictors, with ash negatively and temperature positively impacting emissions. These findings suggest that optimizing key compositional and operational variables can materially reduce CO₂ emissions from biomass processing. The developed models provide robust tools for predictive environmental assessment, supporting more sustainable biomass utilization strategies.