Biomass-derived pyrolysis oil is a promising renewable alternative to fossil fuels; however, its environmental impacts are highly sensitive to both feedstock characteristics and process conditions (Bridgwater, 2012). Traditional modeling approaches, which require detailed case-by-case analysis, are labor-intensive and hinder rapid evaluation of new bioenergy pathways. We developed a machine learning (ML) framework that leverages a dataset of 53 diverse biomass feedstocks to predict key process conditions such as grinding and drying energy requirements, bio-oil yield, and pyrolysis temperature for both pretreatment and thermochemical conversion. Characteristics that were collected from the literature data such as C%, O%, H%, S%, N%, Ash%, bio-oil composition, etc. were used to trail the ML model. The model achieves strong predictive performance for bio-oil yield and moisture content after drying (R² up to 0.85 and 0.71, respectively), while initial prediction of pyrolysis temperature improved from R² = 0.04 to 0.34 by including additional characteristics directly related to the pyrolysis engineering equations. Predicted process conditions are coupled with engineering equations used to estimate life cycle inventory data that was subsequently incorporated into the Research and Development Greenhouse gases, Regulated Emissions, and Energy use in Technologies (R&D GREET 2024) model to estimate the GHG emissions from feedstock pretreatment to conversion. The production of rice husk bio-oil was used as a case study. The ML framework estimated the GHG emissions of rice husk bio-oil at 75.83 g CO₂e/kg, representing a 13% lower value compared to the 86.97 g CO₂e/kg obtained from the literature. This approach demonstrates the value of integrating data-driven predictions with process engineering and environmental assessment, supporting more efficient and robust evaluation of emerging thermochemical technologies.
References:
- Argonne National Laboratory. (2024). GREET® Model [Computer software]
- Bridgwater, A. V. (2012). Review of fast pyrolysis of biomass and product upgrading. Biomass and Bioenergy, 38, 68–94. https://doi.org/10.1016/j.biombioe.2011.01.048
