Multiscale Simulation of Different-Sized Methanol-to-Olefins Fluidized Bed Reactors with Consideration of Coke Distribution | AIChE

Multiscale Simulation of Different-Sized Methanol-to-Olefins Fluidized Bed Reactors with Consideration of Coke Distribution


Lu, B. - Presenter, Dalian National Laboratory for Clean Energy
Li, H., Dalian National Laboratory for Clean Energy and National Engineering Laboratory for MTO, Dalian Institute of Chemical Physics, Chinese Academy of Sciences
Ye, M., Dalian Institute of Chemical Physics, Chinese Academy of Sciences
Wang, W., University of Chinese Academy of Sciences

simulation of different-sized methanol-to-olefins fluidized bed reactors with consideration
of coke distribution

Lu1*, Hua Li2, Mao Ye2, Wei Wang1

Key Laboratory of Multiphase Complex Systems, Institute of Process Engineering,
Chinese Academy of Sciences, Beijing 100190, China.*Corresponding author:

Engineering Laboratory for MTO, Dalian Institute of Chemical Physics, Chinese
Academy of Sciences, Dalian 116023, China

1. Introduction

The methanol to
olefins (MTO) process is an important route to produce light olefins and has
attracted much attention in recent years. In China, more than 10 industrial
units using the dimethyl ether or methanol to olefins (DMTO) technique have
been successfully operated for producing light olefins. The main unit of DMTO
process follows the design of modern fluid catalytic cracking process, but was modified
due to the unique features of MTO reactions. For example, as the ethylene selectivity
is closely related to the coke deposition on SAPO-34 catalyst, a densely
fluidized bed is hence preferred to achieve the desired coke content. Understanding
of hydrodynamics and related reaction features can help us to further optimize
the process. Computational Fluid Dynamics (CFD) is a powerful tool to help
understand the complex hydrodynamics and reactions. However, reactor scale-up
is always involved with variation of flow structures and reaction behaviors, accurate
simulations of a series of MTO reactors ranging from the lab scale to
industrial scale pose bigger challenges to CFD modeling.

2. Methods

Fig.1 presents
geometries of four different-sized reactors and boundary settings. Two-fluid
model (TFM) with kinetic theory for granular flow and a seven lumped reaction
network are employed with FluentÒ15 as the solver. EMMS/bubbling drag model[1]
is employed for the micro-, pilot- and demo-scale reactors operating under low
gas velocity, and its newly developed two-step version[2] which was
found suitable for coarse-grid simulation is used for the simulation of an
industrial reactor. A reactor model developed in our previous study[3]
is used to provide initial distribution of the coke content. Both gas and solid
phases are treated as mixtures, nine species (CO2, CH4, C2H4,
C3H6, C3H8, C4H8,
C5H10, H2O and methanol) for the gas mixture,
and two species (coke and catalyst) for solid phase. The catalyst particles
belong to Geldart type A particles (dp=97mm and rp=1500 kg/m3). More details can refer to our previous
studies [4-5].

˵Ã÷: Figure 1

Schematic diagram of four different-sized DMTO


3. Results and discussion

As shown in Fig.
2a, a relatively uniform flow structure is found in the micro-scale fluidized
bed. As the gas velocity increases, flow structures evolve to have a heterogeneous
distribution in larger DMTO reactors, such as typical bubbling fluidization in pilot-scale
reactor, very diffuse bed surface in demo-scale reactor, particle clusters in
the commercial reactor. The predicted pressure drops of reaction zones for four
reactors agree well with the experiment, showing a reasonable choice of drag
models. The predicted mass fractions of products for both micro-scale and
pilot-scale reactors are in good agreement with experiment, but large
discrepancy is found between simulation and measurement for both demo-scale and
commercial reactors (particularly ethylene is over-predicted).

The experiment
found that there exists a wide distribution of coke content in large MTO
reactors, but the above TFM simulations predict uniform coke distribution. As
the TFM simulation averages different coke contents in a computational cell, thus
fails to predict ethylene which is sensitive to the coke content. We introduced
the population balance model (PBM) to describe the evolvement of coke content
and combined it with TFM. The catalysts with different coke contents are denoted
by a group of solid species. The simulation of 2D demo-scale reactor shows the
ethylene prediction is greatly improved (Fig. 2b).

˵Ã÷: Fig5_vf_fixed_pilot_demo_ind2    ˵Ã÷: PBM_massfraction


  (a) snapshots of solids concentration of
four DMTO reactors as well as mass fractions of gaseous products; (b) mass
fraction of gaseous product by combining PBM and TFM.

4. Conclusions

Combination of TFM and EMMS drags
successfully captures flow structures in different-sized DMTO reactors, but
shows growing discrepancy in predicting product concentration. The possible
reason is the poor prediction in coke distribution. Integrating PBM into TFM enables
to describe change of the coke distribution, thus improving the prediction in
light olefins.


K. Hong, Z. Shi, W. Wang, J. Li. Chem. Eng. Sci.
2013, 99: 191-202.

H. Luo, B. Lu, J. Zhang, H. Wu, W. Wang. Chem.
Eng. J. 2017, 326: 47-57.

B. Lu, H. Luo, H. Li, W. Wang, M. Ye, Z. Liu, J.
Li. Chem. Eng. Sci. 2016, 143: 341-350.

B. Lu, J. Zhang, H. Luo, W. Wang, H. Li, M. Ye,
Z. Liu, J. Li. Chem. Eng. Sci. 2017, 171: 244-255.

J. Zhang, B. Lu, F. Chen, H. Li, M. Ye, W. Wei.
Chem. Eng. Sci. 2018, 189:212-220