We investigate effects of catalyst activity, catalyst particle shape (sphere, slab, and hollow cylinder), size (i.e., diffusion length), catalyst distribution (uniform vs eggshell type distribution ...for a spherical particle), and process conditions (temperature, pressure, syngas composition, and conversion level) on catalyst effectiveness factor and methane selectivity inside the catalyst pellet. In numerical simulations we utilize kinetic parameters for CO consumption rate and CH4 formation rate determined from experiments with a highly active Co/Re/γ-Al2O3 catalyst. It is found that the use of small spherical particles (0.2–0.5 mm) or eggshell distribution for larger spherical particles with catalyst layer thickness less than approximately 0.13 mm is needed to avoid negative impact of diffusional limitations on CH4 selectivity under typical Fischer–Tropsch synthesis operating conditions. For monolith reactors with wash-coated catalyst, diffusional limitations can be avoided by using a catalyst layer thickness less than 0.11 mm at base case conditions (473 K, 25 bar, and H2/CO molar ratio of 2).
Predictions of vapor liquid equilibrium for Fischer-Tropsch mixtures were compared using the classical Soave-Redlich-Kwong (SRK) and cubic-plus-association (CPA) equations of state. The performance ...of the two equations of state was evaluated based on comparison with results from eight sets of experimental runs in which different process conditions (pressure, reactants feed ratio, space velocity) were used. Flash calculations were used to determine the phase split at defined process conditions, whereas the phase equilibrium was defined utilizing the concept of equal fugacities in the vapor and the liquid phase for all components. A total of 75 components were considered in the reaction mixture: CO, H2, H2O, CO2, C1-C57 paraffins and C2-C15 olefins. All calculations were performed in MATLAB. The results showed that both equations of state had similar performance regarding the hydrocarbons, whereas CPA gave better results with inorganic components and SRK with prediction of the composition of the liquid phase. Computational time for CPA was substantially (100 times with the CPU used) higher than that for SRK. Overall, the use of CPA did not improve VLE prediction for FTS systems significantly enough to be recommended for use in FTS reactor models.
nema