(140e) Effects on FCC Catalyst Residence Time Distributions Through Simulations of Circulating Reactors

Introduction

Catalytic upgrading^3,10 of biomass pyrolysis vapors prior to condensation has recently received attention as a promising strategy for improving the compatibility of bio-oil with conventional processing infrastructure. Riser reactors, staples of the petroleum processing industry, are commonly used for FCC (fuel catalytic cracking) and other processes that share characteristics with the envisioned biomass to liquid fuel pathway.² We are currently undertaking a computational study to assess the operational feasibility of using such reactors with pyrolysis vapors.

Specifically, we examine the effects of mass flow rates and boundary conditions on the solid catalyst residence times.  The flow rates include the ratios of pyrolysis vapor and fluidizing gases (steam, N₂) to solid catalyst. The boundary conditions include inlet positions, diameters, and the outlet configuration. We additionally examine the effects on clustering, temperature, and velocity profiles.

Starting with non-reactive two- and three-dimensional simulations, this study will ultimately incorporate kinetic mechanisms for catalyzed secondary gas phase reactions to study the overall performance of the reactors. Experimental measurements of bulk quantities (temperature, pressure, steady state catalyst weight %) from a pilot scale reactor will be used for validation of the model fidelity. This work forms a baseline for future studies where we will model both the primary pyrolysis as well as catalyzed secondary gas phase reactions to study and improve the overall performance of the reactors.

Goals of Research

The objective of this project is to simulate and study the catalytic upgrading of pyrolysis vapors through validation with the pilot scale FCC DCR (Davison circulating riser) reactor that is arriving at NREL in Feb. 2014.  Through validation of experiments with the DCR, we will better understand and improve the conversion of biomass to liquid transportation fuels, specifically gasoline and diesel precursors.  This conversion is highly dependent on the contact time of the pyrolysis vapor with the FCC catalyst, which is typically of the zeolite type.  Too short of a contact time leads to under conversion and too long leads to excessive coking, deactivation of the catalyst, and/or undesirable products.  In an actual riser, there is a distribution of residence times and unsteady creation, destruction, and flow of particle clusters, particularly prevalent at the walls.  The distribution’s average as well as its shape (skew preferring slower times) is predicted to have an effect on the resulting biofuel products.

Numerical Approach

Gas-particle flows in FCC reactors occur over a wide range of scales. In order to have a simulation that can run for experimental time and length scales, a continuum and filtered equation approach is warranted. The two-fluid model (TFM) equations for gas-particle flows, which have been developed and analyzed extensively over the past five decades, are able to capture their existence in a robust manner.¹ However, to resolve the FCC catalyst particle clusters and bubble-like voids at all length scales, extremely fine spatial resolution is necessary, but impractical for three-dimensional simulations. An engineering tool, the filtering of the two-fluid equations^4,5,6 is used to probe the macro-scale flow characteristics directly, without having to resolve the smaller scale structures. In this approach, the influence of the small-scale structures appears as residual correlations, called subgrid models. These subgrid modifications of the equations have been constructed with constitutive or closure models that are designed to model macro-scale gas-solids interactions, for example particle drag, as well as the partial slip wall effects of the boundary conditions.

With the MFIX TFM (two-fluid) method^8,9, using solids with a particle size distribution of solid catalyst particles (Geldart A type, size 20-140 micrometers with mean of 75), the clustering and residence times of solids and vapors will be determined.  Peregrine, the NREL supercomputer with ~15,000 available cores is utilized for these simulations.  The use of a hybrid model using distributed (domain decomposition) and shared memory (threading of loops) parallelism is important for efficient, three dimensional, multi-component simulations.

Results and Initial Conclusions

    Preliminary simulations for two and three dimensional riser models have been conducted to determine the effect of initial and boundary conditions on the catalyst and vapor residences times.  A preliminary simulation result for a catalyst to feedstock ratio equal to three, with 3 and 1 weight percent respectively (relative to feed) of steam and additional fluidizing nitrogen has been conducted for the cylindrical riser of ~1cm diameter and height ~3m. Using a Cartesian stretched grid of size 96x2440 cells, the simulation was conducted for 50 seconds of operation using 768 cores for one week on Peregrine at NREL.  Lower resolution simulations were also conducted up to 150 seconds. These baseline simulations were first conducted with out the subgrid scale models such that a highly resolved result could be used for a comparison to subsequent filtered results with the subgrid clustering and wall model modifications. By finding the optimal filter size and wall subgrid parameters, coarse grid simulations can be used to obtain the desired bulk properties at a fraction of the computational cost.