Biomass fast pyrolysis prediction model through data-based prediction models coupling with CPFD simulation

Biomass fast pyrolysis has been traditionally investigated through experiments or numerical simulations including hydrodynamics, transport phenomena of heat and mass, and chemical reactions. Instead of traditional methods, this study utilized data-based prediction algorithms to model a biomass fast pyrolyzer using a spouted fluidized bed, to predict the yields of major products. We used the labeled dataset generated by the computational particle fluid dynamics (CPFD) simulation to train the data-based prediction models, and used the reaction temperature and gas residence time as inputs. Eight data-based prediction models, including machine learning and deep learning, were selected and compared. The selected data-based prediction methods indicated better agreement with the CPFD product yields than those of the lumped process models. This study provides new guidelines for modeling fast pyrolysis reactions using CPFD and data-based prediction methods.

Automated summary

This study utilized data-based prediction algorithms to model a biomass fast pyrolyzer and predict the yields of major products. Eight data-based prediction models were compared, indicating better agreement with experimental results compared to traditional lumped process models. The study provides new guidelines for modeling fast pyrolysis reactions using data-based prediction methods.