(309c) Multi-Scale Fast Pyrolysis Simulation Framework for Varied Biomass Feedstocks

This presentation summarizes the recent work of building an MFiX-based Multi-scale Biomass Fast Pyrolysis Simulation Framework for biomass feedstocks. The MFiX Suite is a family of multiphase CFD software developed at the National Energy Technology Laboratory and is designed for accurate simulation of multiphase reactors like biomass pyrolysis reactors. In simulating multiphase reactors, the aerodynamic drag related to particle density, shape, and size is a critical element of the biomass material attribute as it determines the particle residence time in the reactor. A hybrid drag model has been proposed to address the challenging problem of accurately capturing the drag of diverse biomass feedstocks. In this approach, a machine learning derived mesoscale drag model was used for small-sized biomass and biochar particles, and a macro-scale experiment measured drag model was used for sand and large biomass particles. The simulation accurately predicted the mixing of sand and biomass. Predicted biochar elutriation rate compared well with experimental data measured in a 2.5-inch fluidized bed using video (60 Hz camera) and high-speed differential pressure techniques with seven 100 Hz pressure differential transducers. A detailed biomass pyrolysis kinetic model with 32 heterogeneous reactions and 59 species was implemented in the MFiX suite of software. The reaction model was validated against two experimental pyrolysis data sets that provided detailed data describing chemical component yields. Sensitivity analysis using the simulation framework revealed the effects of critical material attributes on yields of pyrolysis oil. After validating the drag model and pyrolysis kinetics, the influence of particle shapes was considered using a glued-sphere approach where the composite shape of connected spheres was used to resolve the complex biomass shape, intra-particle heat conduction, and intra-particle species distribution. A simple rolling friction model was introduced to reflect the influence of particle shapes during collisions. To increase simulation speed, intra-particle heat and species distributions were then simulated using a machine learning model derived from detailed, particle-resolved simulations. Finally, the simulation framework was used to investigate the pyrolysis of varied biomass feedstocks over a range of operating conditions and scales to help guide pyrolysis reactor design and optimization.