Machine learning prediction of contents of oxygenated components in bio-oil using extreme gradient boosting method under different pyrolysis conditions

This work aims to develop a prediction model for the contents of oxygenated components in bio-oil based on machine learning according to different pyrolysis conditions and biomass characteristics. The prediction model was constructed using the extreme gradient boosting (XGB) method, and the prediction accuracy was evaluated using the test dataset. The partial dependence analysis (PDA) method was used to derive the pattern of influence of each input feature individually or in combination on the output variable. The results show that the prediction models constructed from biomass ultimate analysis and pyrolysis conditions can predict the contents of oxygenated components in bio-oil more accurately than the models constructed from biomass proximate analysis. Moderate C and O contents, higher H content of biomass, lower flow rate, and higher pyrolysis temperature can improve bio-oil quality.