Lignin has great potential use as a feedstock for thermochemical processes to produce biofuels or other aromatic chemicals. It is a highly cross-linked amorphous polymer accounting for up to 40 wt% of dry biomass, which significantly complicates the understanding of molecular or mechanistic scale chemical reaction networks governing conversion. Empirical and global models of lignin pyrolysis using overall apparent kinetics derived from TGA studies are well established and often adequately predict yields of lumped products within a small range of experimental conditions. The main limitations of these models are their inability to predict the composition and distribution of product tars and gases, or to cover a wide range of heating rates and temperatures with one set of kinetic parameters.  A fully mechanistic model of lignin pyrolysis would be of great interest for these reasons, but developing such a model has proved challenging due to the random structure of lignin polymers and complex kinetics involved. <br/><br/>As a step toward a more mechanistic model, we explore the use of a semi-detailed kinetic model for lignin pyrolysis to predict the compositions of gases and tar components from the fast pyrolysis of lignin over a wide range of conditions.  The model combines elementary and lumped reaction steps to track approximately 100 real molecules, radicals, lumped heavy species, and functional groups linked to the degrading polymer. The sensitivity of permanent gases and tar precursor molecules to experimental conditions is examined, and using this model we are able to explore new mechanistic details about the evolution of functional groups. In the first part of the talk we will present the model as well as a tool for looking a high dimensional sensitivity analysis of kinetic modeling data. <br/><br/>The second part of the talk will present efforts to address a standing challenge in using mechanistic modeling schemes within more complicated numerical optimizations (e.g., spatiotemporal reactor models like fluidized beds or global optimization schemes that are limited on the solution time of the kinetic model). We introduce a scheme using neural networks to enable fast solution of the kinetic model on timescales up to 4 orders of magnitude faster than a stiff ODE solver.  Implementation of the model for lignin pyrolysis and results will be discussed.

Data science tools for biomass pyrolysis – application to thermochemical conversion of lignin

253rd National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.