Synergistic valorization of biomass-plastic waste: machine learning optimized oil production via catalytic co-pyrolysis with life cycle sustainability analysis

초록

Catalytic co-pyrolysis of biomass and plastics represents a highly efficient approach for oil production. However, unraveling the oil evolution mechanism proves challenging owing to the intricate interplay of multiple factors. Integrated machine learning algorithms were employed to optimize the preparation parameters and screen catalysts for the co-pyrolysis of biomass and plastics. Additionally, a life cycle assessment was conducted to evaluate the environmental implications associated with the utilization of waste biomass and plastics. This study reveals that feedstock elemental composition-particularly hydrogen (H) and sulfur (S) content-and blending strategy are the dominant factors controlling energy conversion (oil) yields in catalytic co-pyrolysis of biomass and plastic, surpassing the influence of temperature and catalyst properties. High H content increases liquid yield, while S and moisture decrease solid residue. Machine learning models, especially Random Forest (RF), show good predictive capability for oil yield (R-2 > 0.90). Principal Component Analysis (PCA) and Shapely Additive Explanation (SHAP) analyses confirm feedstock chemistry as the primary driver of product distribution. Life cycle assessment indicates significant environmental burdens from pyrolysis and combustion phases, yet using waste feedstocks reduces CO2 and environmental impacts by 25% and 38%, respectively, compared to petroleum-derived oil.

키워드