Hydrodeoxygenation of Biomass Pyrolysis Oil Towards Renewable Bio-Gasoline Production: Economic and Environmental Optimization Based on a Machine Learning Framework

Abstract

Hydrodeoxygenation (HDO) is an essential step to upgrade biomass pyrolysis oil towards renewable transportation fuel production. An extremer degree of pyrolysis and HDO would better enhance renewable transportation fuel properties, but it could also lead to higher cost. This study tries to systematically concerned biomass pyrolysis and HDO together, thus to achieve overall optimization of economic and carbon emission benefits. A machine learning framework consists of both process simulation and optimization was established and comprehensively discussed, by which the pyrolysis-HDO system was optimized economically and environmentally. The results demonstrate that the integrated learning model outperforms other methods in terms of prediction accuracy, with the best model achieving an average error rate of 5.8% and an R2 value of 0.96. SHAP analysis reveals that pyrolysis temperature, hydrogenation temperature, and solvent type are key factors influencing the properties of bio-oil. The combined analysis using four-dimensional bubble charts and the PDP method reveals that the interaction of key process parameters has a synergistic effect on the quality of high-quality bio-oil. Specifically, HHV, C content, and H2 pressure consumption have a significant impact on its economic viability and carbon reduction potential. Finally, PSO demonstrates that optimizing the combination of process parameters can achieve optimal performance while balancing both economic viability and carbon reduction goals. The optimization framework provides insights for balancing quality, energy, and carbon reduction in bio-oil production, guiding sustainable biomass energy processes.

Keywords: Pyrolysis, Hydrodeoxygenation, Integrated learning, Economic benefit, Carbon reduction