Application of life cycle assessment and machine learning for the production and environmental sustainability assessment of hydrothermal bio-oil

Abstract

The hydrothermal bio-oil (HBO) production from biomass conversion can achieve sustainable and low-carbon development. It is always time-consuming and labor-intensive to quantitative relationship between influential variables and bio-oil yield and environmental sustainability impact in the hydrothermal conditions. Machine learning was used to predict bio-oil yield. Life cycle assessment (LCA) is further conducted to assess its environmental sustainability effect. The results demonstrated that gradient boosting decision tree regression (GBDT) have the most optimal prediction performance for the HBO yield (Training R2 = 0.97, Testing R2 = 0.92, RMSE = 0.05, MAE = 0.03). Lipid content is the most significant influential factor for HBO yield. LCA result further suggested that 1 kg of bio-oil production can cause 0.02 kg ep of SO2, 2.05 kg ep of CO2, and 0.01 kg ep of NOx emission, and environmental sustainability assessment of HBO is exhibited. This study provides meaningful insights to ML model prediction performance improvement and carbon footprint of HBO.