EUBCE 2025 - Georgiana BELE - AI-Driven Optimization of Fast Pyrolysis: Maximizing Bio-Oil and Biochar

AI-Driven Optimization of Fast Pyrolysis: Maximizing Bio-Oil and Biochar

Short Introductive summary

Pyrolysis processes receive an increased interest in the industrial decarbonization context, due to the simultaneous production of biochar and bio-oil that can substitute products in hard-to abate industries. This study addressed the optimization of pyrolysis process performances towards industrial decarbonization through the identification of operating parameters that favour emissions reduction using resulting bioproducts. It is based on the development of an advanced AI-assisted tool integrating both predictive and descriptive machine learning (ML) techniques to characterize process yields to gas, liquid and solid products of fast pyrolysis, as well as biochar and bio-oil energy densities. The tool is trained on diverse experimental datasets from different pyrolysis setups, designed to address the variability and complexity inherent in data on emerging technologies. By leveraging an ensemble of ML models, it overcomes the limitations of single-method approaches, providing robust predictions and interpretable insights into the relationships between process parameters, feedstock characteristics, and yield outputs.

M. Moubarak, Polytechnique Montreal, Montreal, CANADA
E. Ibrahim, Natural Resources Canada, Varennes, CANADA
A. Ragab, Natural Resources Canada, Varennes, CANADA
M. Benali, Natural Resources Canada, Varennes, CANADA
G. Bele, Natural Resources Canada, Varennes, CANADA
C. Diffo Téguia, Natural Resources Canada, Varennes, CANADA
M.S. Ouali, Polytechnique Montreal, Montreal, CANADA