Detailed CFD-DEM Simulation of Biomass Gasification in a Fluidized Bed Reactor

Ostermeier, P., Techinal University of Munich
Fischer, F., Techinal University of Munich
Fendt, S., Techinal University of Munich
DeYoung, S., Techinal University of Munich
Spliethoff, H., Techinal University of Munich
Gas-solid fluidized beds play an important role in many industrial operations. Nevertheless, there is still a lack of knowledge of the processes inside the bed which impedes proper designing and upscaling of fluidized bed operations. In this work, the biomass gasification process with steam in a fluidized bed reactor is investigated with a detailed CFD-DEM approach.

The numerical model in the Eulerian-Lagrangian framework treats the gas phase as a continuum and describes the particle interactions with the discrete element method (DEM). The non-spherical shape of the particles is accounted for in the momentum exchange calculation with the gas phase. The considered systems consists of steam (ideal gas, temperature 800 °C) as the fluidization gas, entering the bottom region of the complex three dimensional reactor geometry through inclined nozzles in an inner duct at the center axis. The wet wood pellets (sphere equivalent diameter 6.1 mm, particle density 1160 kg/m3, sphericity factor 0.86) are fed into the reactor in the freeboard region together with a nitrogen purge gas stream (ideal gas, 80 °C).

Two different operation modes are investigated. For the start-up procedure, the reactor is filled with 2.565 kg of sand particles (50-450 µm, particle density 2210 kg/m3, sphericity factor 0.86, coarse-graining 1.5e5 parcels). The inert material is allowed to settle for three seconds and is fluidized for another three seconds until quasi steady-state is achieved. Afterwards, the wet wood pellets are added for 35 seconds at a rate of one pellet every two seconds and undergo the processes of heating, drying, pyrolysis and coke conversion. From the simulation results and literature correlations, the average time intervals for pyrolysis and coke conversion, as well as the respective particle properties (temperature, diameter, density) are estimated. With the knowledge of the average composition of reactive material in the reactor, variable operating points can be simulated.

The second case investigated is after approximately 50 hours of operating time. Additionally to the start-up procedure, 1.5 kg of inert coke residue (100-4000 µm, particle density 400 kg/m3, sphericity factor 0.86, coarse-graining 5e4 parcels) is added with the sand particles. Again, the inert material is allowed to settle for three seconds and is fluidized for another three seconds. Subsequently, the average reactive material obtained from the start-up simulations is patched into the bed region and wet wood pellets are fed to the freeboard region.

The following phenomena are implemented in the numerical model: gas-solid momentum exchange, solids collisional behavior, turbulence, heat and mass transfer, particle shrinkage and change in material properties, pyrolysis, and homogeneous and heterogeneous chemical reactions. Here, the focus is on the influence of properly modeling turbulence and heat transfer.

Simulation results are analyzed both qualitatively and quantitatively. The particle flow fields and the mixing between sand, wood, and residue are investigated for laminar and turbulent flow and for different filling of the reactor. Furthermore, the heating and conversion of the wood pellets is examined for different heat transfer models. Bed pressure drop and product gas composition are compared to experimental data to verify the numerical model. For this purpose, literature correlations for the wood pellet conversion are considered as well.

The results show that the average heating rate for the wood pellets is similar with and without turbulence model. However, the pellets in the simulation with turbulence exhibit a more repeatable heating and thus conversion behavior (smaller spread in the temperatures and properties of the different pellets) than in the laminar case. On the other hand, the choice of the heat transfer model to calculate the heat exchange between the different phases has a higher influence than the turbulence model. The user-defined approach including the Gunn-correlation (recommended for dense fluidized beds) results in faster heating of the particles than the correlation according to Ranz and Marshall (the default model in ANSYS Fluent). Finally, the ranges for the bed pressure drop, the gas composition, and the conversion time obtained with the numerical simulation agree well with experimental data and literature correlations. This indicates that the proposed model can make a significant contribution towards understanding and improving the internal processes in fluidized bed reactors for biomass gasification and combustion.