Transient Simulations of Spouted Fluidized Bed for Coal-Direct

Article pubs.acs.org/EF

Transient Simulations of Spouted Fluidized Bed for Coal-Direct Chemical Looping Combustion Zheming Zhang,† Ling Zhou,†,‡ and Ramesh Agarwal*,† †

Department of Mechanical Engineering & Materials Science, Washington University in St. Louis, Missouri 63130, United States Research Center of Fluid Machinery Engineering & Technology, Jiangsu University, Zhenjiang 212013, China

‡

S Supporting Information *

ABSTRACT: Chemical-looping combustion holds significant promise as one of the next generation combustion technology for high-efficiency low-cost carbon capture from fossil fuel power plants. For thorough understanding of the chemical-looping combustion process and its successful implementation in CLC based industrial scale power plants, the development of highfidelity modeling and simulation tools becomes essential for analysis and evaluation of efficient and cost-effective designs. In this paper, multiphase flow simulations of coal-direct chemical-looping combustion process are performed using ANSYS Fluent CFD code. The details of solid−gas two-phase hydrodynamics in the CLC process are investigated by employing the Lagrangian particle-tracking approach called the discrete element method (DEM) for the movement and interaction of solid coal particles moving inside the gaseous medium created due to the combustion of coal particles with an oxidizer. The CFD/DEM simulations show excellent agreement with the experimental results obtained in a laboratory scale fuel reactor in cold flow conditions. More importantly, simulations provide important insights for making changes in fuel reactor configuration design that have resulted in significantly enhanced performance.

1. INTRODUCTION To avoid the “unequivocal” warming of the global climate system,1 high-efficiency carbon capture technologies are needed to address the reduction of carbon emissions from power generation using fossil fuels. Postcombustion treatment, oxy-fuel combustion, and precombustion treatment have been proposed for this purpose. Although these technologies have demonstrated success in reducing the CO2 emissions from the flue stream of the power plants, at the same time, they are also associated with high energy penalty.2−4 Recently, a relatively new technology known as the chemicallooping combustion (CLC) is showing good promise for highefficiency low-cost carbon capture. Instead of having air to support the combustion process, an oxygen compound (metal or non-metal based) is used as an oxidizer in the fuel reactor; thus, the fuel is chemically combusted by the metal oxide than the oxygen present in air in case of standard power plants. Due to the absence of air in the fuel reactor, the combustion products are not diluted by other gases (e.g., N2) resulting in high purity of CO2 stream available at outlet of the fuel reactor. Research has shown that the energy cost of solid circulation (which is the only energy cost of separation) in CLC is considerably smaller (approximately 0.3% of total energy released by the combustion process5) compared to other precombustion technologies (e.g., oxy-fuel combustion in which the oxygen separation process consumes nearly 15% of the electricity generation2). Therefore, CLC holds significant promise as a next generation combustion technology due to its potential to allow near zero CO2 emissions with minimal effect on the overall efficiency of the power plant.6 Additionally, several energy and exergy analysis of CLC systems suggest that power plant efficiencies greater than 50% can be achieved along with nearly complete CO2 capture.7−11 In recent years, the use of coal for CLC process has gained increasing interest since coal is likely to maintain its dominant role in the power generation sector in the foreseeable future. © 2014 American Chemical Society

Additionally, other solid fuels such as biomass also hold potential for use in CLC. The utilization of coal in a CLC process can be achieved by two pathways. The coal can be first gasified in a standalone gasifier to subsequently introduce the freshly converted syngas to the fuel reactor. From the perspective of CLC process, such a scenario is essentially identical to the one that uses gaseous fuel. The CLC process utilizing gaseous fuel has been widely studied both numerically and experimentally.5,12,13 On the other hand, if solid coal (pulverized) is directly fed into the CLC system, such a process is known as the coal-direct chemical-looping combustion or CD-CLC. The CD-CLC concept eliminates the necessity of a gasification chamber and therefore reduces the complexity of the entire power generation system. In the case of CD-CLC, there have been two proposed options as to how the metal oxide will participate in the coal combustion. One of these options is to make the coal gasification occur in an identical reactor by the oxygen carrier wherein the coal is gasified in situ by H2O and/or CO2 supplied as a fluidization agent.14 Such a process is known as the in situ gasification chemical-looping combustion or iG-CLC. The other option is to utilize special oxygen carrier which releases gaseous oxygen under reactor conditions to sustain the combustion of solid coal in the fuel reactor. Mattison et al.15 have proposed the so-called chemical-looping combustion with oxygen uncoupling (CLOU) as a variant of CD-CLC process. Two major concerns about the CD-CLC process arise from the solid/gas mixing in the fuel reactor and the appearance of agglomeration between oxygen carrier and the coal ash. It has been identified that the fuel conversion rate in the iG-CLC Received: December 23, 2013 Revised: January 26, 2014 Published: January 28, 2014 1548

dx.doi.org/10.1021/ef402521x | Energy Fuels 2014, 28, 1548−1560

Energy & Fuels

Article

process is limited by the char gasification step.16 Loss of reactivity due to agglomeration has also been reported when the ash concentration is high.17 To address such concerns, the configuration of spouted fluidized bed for the fuel reactor has been proposed for CD-CLC. Instead of evenly distributing the gas feeding at the inlet, a high speed center jet is assigned. The presence of the center jet can lead to high circulation rates of solid particles which enhance the solid−gas mixing for stronger gasification process. The enhanced particle circulation can also help to rub off the deposit of ash on oxygen carrier to reduce agglomeration.18 The successful design and optimization of a CD-CLC system requires the development of credible simulation models for multiphase fluid flow, which include solid particle dynamics and chemical kinetics. Considering physical phenomenon alone, there have been two approaches to address the simulation of solid−gas coupled multiphase flow. One is to assume that the particulate phase is in continuum and thus can be treated as a secondary but heavier “gas” phase. Because the fluid-based mass and momentum equations are solved for both phases within the Eulerian framework, this approach is also known as the multiphase Eulerian method. In the second approach, the particulate phase is modeled at individual particle level tracking their movement using the Lagrangian framework and is coupled with the fluid field for interphase mass and momentum exchange. It should be noted that the interparticle collision is described using a deterministic model, discussed later in the paper. Because particles are treated as discrete elements, this approach is also known as the discrete-element method (DEM). The typical interconnected fluidized bed configuration of CD-CLC and the presence of solid fuel in it require the capability of accurate capture of the solids circulation and separation in the system. Therefore, the consideration of solid−gas two-way coupling and solid−solid interaction becomes important for such simulations. In this paper, a series of transient simulations of quasi-3D spouted fluidized bed apparatus for CD-CLC process are performed using the CFD/DEM coupled model. Unsteady tracking of individual solid particles and their interaction with each other as well as the ambient fluid is enabled to provide the most accurate and realistic representation of the multiphase flow field. The modeling of chemical reactions is not considered in this paper; it will be addressed in future work. The CFD/DEM simulation approach has been used in various applications which require the modeling of particle/fluid interaction;19−23 however, its application in the context of close-loop spouted fluidized bed system has so far been quite limited and preliminary. By establishing the validated (validated against the laboratory scale experimental data) accurate cold-flow models of CD-CLC configurations, it is expected that this work should help in gaining deeper insights and provide best practices for modeling and design of the industrial scale CD-CLC process.

the commercial simulation package ANSYS Fluent, release version 14.5.24,25 The simulations are carried out on a Dell workstation (quad-core Intel Xeon CPU @ 3 GHz and 8 G of memory) running Windows 64-bit operating system. All particles are traced individually as parcels with the actual particle diameter. Fluid Equations. To accommodate the presence of solid particles, the fluid equations of mass and momentum balance are slightly modified. The concept of fluid porosity εf, defined as the ratio of gas volume to the entire volume of the computational cell, is introduced in the equations. Let ρg, pg, ug, and g denote the density of the fluid, pressure of the fluid, velocity vector of the fluid field, and gravitational constant vector, respectively. The volume-averaged continuity and Navier−Stokes equations can be written as ∂ (εf ρf ) + ∇(εf ρf u f ) = 0 ∂t ∂ (εf ρf u f ) + ∇(εf ρf u f u f ) = − εf ∇pg − ∇(εf τf ) + εf ρf g − Sp ∂t (1)

where Sp = βsg (uf − us) is the additional source term that is determined by the presence of solid particles contributing to the drag force between the two phases. The introduction of Sp enables the two-way coupling of fluid and solids. For Newtonian fluid such as air or gaseous CO2, the viscous shear tensor of fluid τf can be described as ⎛ 2 ⎞ τf = μf (∇u f + ∇uTf ) + ⎜λf + μf ⎟(∇u f )I ⎝ 3 ⎠

(2)

Particle Equations. In the Eulerian−Lagrangian modeling approach, the solids are treated as individual particles. The prediction of the trajectory of the solid particles is obtained by integration the Newton’s equation of motion which is constructed in the Lagrangian framework. The forces acting on a particle balance the inertia of the particle, as given below: ms

du s = dt

∑ Fi=FGra + FBuo + FDrag + FPre + FSaf

+ FMag + Fcon

(3)

In eq 3, FGra and Fbuo are the volume forces due to gravity and buoyancy respectively; FDrag, Fpre, FSaff, and FMag are hydrodynamic forces due to the moment exchange between the particles and the surrounding fluid, namely, the drag force due to fluid viscosity, pressure force due to pressure gradient, Saffman lift force due to interparticle friction, and Magnus force due to the spinning motion of the particles respectively; Fcon is the shortrange contact force between the particles. In this work, the drag force FDrag is calculated using the drag model of Syamlal−O’Brian described in the next section. FPre and FMag are dropped from eq 3 in the calculation without significant effect on accuracy, considering the large difference in particle and gas density and negligible rotational motion of the particles. There are several ways to numerically construct the contact force Fcon between the particles. The soft-ball model based on Hooke’s law has been used in this work25 in which the contact force Fcon is decoupled in its normal and tangential components. The normal contact force is modeled as

2. MODELING APPROACH In the coupled DEM and CFD approach to multiphase simulation of CD-CLC process, the fluid motion is obtained by solving the incompressible continuity and Navier−Stokes equations. The motion of particulate solids is modeled by the Newtonian equation of motion. By introducing the corresponding source terms related to movement and interaction of solid particles in the fluid dynamics equations and terms due to forces experienced by the solid particles in the Newtonian equations for the motion of the solid particles, the coupling of solid−fluid interaction is achieved as described below. This work employs

Fncon = (kδ + γ(u12e))e

(4)

where k is the spring constant of the pair of particles for collision, δ is the overlap of the particle pair for collision, γ is the damping 1549


Energy & Fuels

Article

where τp is the particle relaxation time defined as

coefficient, u12 is the relative velocity between the collision pair, and e is the unit vector. Neglecting the particle’s rotational motion during collision, the tangential frictional force between the two colliding particles can be evaluated as a fraction of the normal contact force, which gives n Ftcon = μFcon

τp =

where μ is the friction coefficient described as vr < vglide

− μ = μglide if

vr ≥ vglide

18μf

(10)

In eq 9, f is the drag coefficient, and it can be modeled by numerous empirical equations. To name a few, there are spherical,26 nonspherical,27 Gidaspow,28 Wen−Yu,29 and Syamlal−O’Brian30 drag laws. In the this paper, the Syamlal− O’Brian drag law is used, which gives

(5)

− μ = μstick + (μstick − μglide )(vr /vgilde − 2)(vr /vglide) if

ρp ds2

f=

where μstick is the sticking friction coefficient, μglide is the gliding friction coefficient, vglide is the gliding velocity, and vr is the relative velocity of the collision pair. Detailed description of the friction model can be found in the ANSYS Fluent Theory Guide.25 A schematic showing how the contact force of a collision pair is evaluated is shown in Figure 1.

C DRe pαf 2 24vr,p

and

⎛ C D = ⎜⎜0.63 + ⎝

⎞2 4.8 ⎟ Re p/vr,p ⎠⎟

(11)

where vr,p is the terminal velocity correction for the particulate phase defined as vr,p = 0.5(A − 0.06Re +

(0.06Re)2 + 0.12Re(2B − A) + A2 )

(12)

In eq 12, the coefficients A and B have the following values: A=

Solid−Gas Momentum Exchange. For solid−gas multiphase flow, momentum exchange is enabled by considering the drag force. The momentum transfer from the continuous gas phase to the discrete solid phase is computed by examining the change in momentum of a particle as it passes through each control volume in the CFD model. In the momentum equation, the drag forces exerted on the particle by the ambient moving gas is modeled as (6)

where u is the gas velocity vector, up is the particle velocity vector, and FD is the drag coefficient. The drag coefficient can be modeled as FD =

18μ C DRe p ρp d p2 24

(7)

where μ is the molecular viscosity of the gas, ρ is the fluid density, ρp is the density of the particle, and dp is the particle diameter. CD and Re respectively are the drag coefficient and Reynolds number of the solid particle based on the diameter. The relative Reynolds number Rep is modeled as Re p =

ρf d p|u f − u p| μ

1.28 ⎧ if αf ≤ 0.85 ⎪ 0.8αf ⎨ B= ⎪ 2.65 if αf > 0.85 ⎩ αf

(13)

3. COLD FLOW SIMULATIONS OF SPOUTED FLUIDIZED BED The introduction of spouted fluidized bed in the context of coaldirect CLC has generated a great deal of interest in recent years because of its several technical benefits. First, the introduction of high velocity jet is likely to create strong mixing of solids and gas avoiding loss of reactivity due to the ash agglomeration with the oxygen carrier.17 Second, the oxygen carrier needs to be made into particles with relatively large diameter to make it spoutable, and the relatively large size of oxygen carrier particles is likely to result in easier ash separation from the recirculating oxygen carrier. A significant differentiation of a spouted fluidized bed from others (e.g., a bubbling fluidized bed) is the intense particle− particle and particle−wall collision. Gryczka et al.’s work on spouted bed with 2 mm-diameter particles has suggested that accurate numerical representation of particle dynamics is not likely to be achieved using the multiphase granular solid phase approximation due to “the inadequacies of the continuum model”.31 Such inaccuracy has been identified to originate from the nonphysical closure for the continuum model such as the frictional solids viscosity or the solids pressure. To address such inaccuracy, CFD/DEM coupled simulations using Lagrangian− Eulerian modeling approach, which has stronger physical foundation, has been introduced in recent literature. To validate the CFD/DEM simulations using Lagrangian− Eulerian approach, cold flow simulations employing the discrete element method (DEM) coupled with hydrodynamics solver are first conducted. The modeling and simulation is conducted for the laboratory-scale quasi-3D experimental CLC fuel reactor at Darmstadt University of Technology (TU-Darmstadt).32 For the cold flow simulations, only the hydrodynamics and particle motions are considered without chemical kinetics. The CFD/ DEM modeling approach allows the consideration of each particle in the system individually representing its temperature, composition, position, velocity, and its interaction with other

Figure 1. Schematic of particle collision model for DEM.

FDrag = FD(u f − u p)

αf4.14 ,

(8)

Recalling the momentum equation for the gas phase, the two-way coupling between the gas and the solids is enabled by introducing an additional source term on the right-hand-side of eq 1 as Sp = βsg(uf − up). βsg is the multiphase exchange coefficient and is modeled in the following fashion: αpρp f βsg = τp (9) 1550


Energy & Fuels

Article

Figure 2. Test rig of the spouted fluidized bed apparatus at Darmstadt University of Technology (left) and CFD model (right).

Table 1. Key Modeling Parameters for Simulation of the TU-Darmstadt Experiment param. avg. diam. of particles avg. density of particles mass load of particle in bed inlet velocity background gas velocity primary phase secondary phase outlet boundary condition inlet boundary condition drag law particle collision law spring constant coefficient of restitution friction coefficient numerical schemes time step size

value 2.5 mm 2500 kg/m3 ∼0.7 kg 22.69 m/s 0.85 m/s air glass beads pressure outlet with Pout,gague = 0 Pa velocity inlet, central jet velocity of 22.69 m/s, background flow velocity of 0.85 m/s Syamlal−O′Brian Spring−Dashpot 410 kN/m 0.97

Figure 4. Time variation of bed height.

particles and walls. The overall behavior of the system is given by the interaction of all individual particles among themselves and with the surrounding gas phase. The TU-Darmstadt experiments are performed with a spouted fluidized bed similar to that in the test rig used by Link.33 It is a Plexiglas model with 100 cm in height, 15 cm in width, and 2 cm in depth. A high speed air jet is supplied through a centrally placed nozzle with dimension of 1 cm (W) × 2 cm (D), while low speed background air flow is

0.1−0.5 based on relative velocity phase coupled SIMPLE, second-order upwind for momentum equation, QUICK in volume fraction, second-order implicit in time particle: 5 × 10−5, Fluid: 5 × 10−4

Figure 3. Particle distribution and velocity in the reactor for the first 300 ms of jet injection (Top: Fluent CFD/DEM. Bottom: TU-Darmstadt experiment). 1551


Energy & Fuels

Article

They are generally very hard to get fluidized. Once the bubbles are formed, they tend to rise more slowly than the rest of the gas percolating through the emulsion and coalesce rapidly. According to Link’s work,33 it is necessary to introduce background gas flow to prefluidize the bed so that prominent gas bubbles can form when high-velocity central jet is supplied. Therefore, in TU-Darmstadt experiment, background gas flow is introduced by the two side panels beside the center nozzle. The interaction between particles and injected air, which is the twoway exchange of momentum, is accounted by the SyamlalO’Brian drag law given by eq 11. In the simulation, a total of 36 500 particles are released with a small offset in z-direction (Figure 2 (right)). Because of gravity, they start to experience free fall. The aggregate kinetic energy of the particles is monitored. When the aggregate kinetic energy of the particles becomes sufficiently small, the bed is supposed to be at nearly static condition and the simulation time is set as zero. Such an initialization is consequential, since it ensures the random packing of the particles in the static bed. For the continuous gas phase, the velocity inlet boundary condition is applied at the bottom panels of the domain, with 22.69 m/s gas velocity at the center panel and 0.85 m/s gas velocity at the side panels. Pressure outlet boundary condition with zero gauge pressure is applied to the top surface of the domain. Standard nonslip boundary condition is applied to the lateral walls. For the discrete particulate phase, all surrounding surfaces are treated as reflecting walls with reconstitution coefficient of 0.97. To ensure numerical accuracy, second-order discretization in both space and time is applied. Due to large value of the spring constant, the particle tracking time step is set sufficiently small to capture the key particle collisions. The key numerical parameters are summarized in Table 1.

Figure 5. Pressure at various “z” from the bottom.

introduced through the two side panels. The top of the domain is open to the atmosphere. A high speed camera is used to record the distribution of the particles with 1 ms resolution in time. The change of static pressure at four different height locations of z = 2 cm, 12 cm, 22 cm, and 40 cm is monitored by pressure sensors. The experiment is initially loaded with 36 500 glass beads with an average diameter of 2.5 mm and a density of 2500 kg/m3. The TU-Darmstadt experimental apparatus is shown in Figure 2 (left). The computational domain shown in Figure 2 (right) consists of 11 000 hexahedral cells. The continuum fluid (air) is considered as the primary phase and the discrete particulate solid (the glass beads) is considered as the secondary phase. The glass beads used in the TU-Darmstadt experiment fall into the category of group D particles based on their size and density.

Figure 6. Geometry outline with five pressure taps (left), mesh (middle), and frame-wire (right) of the complete CD-CLC configuration. 1552


Energy & Fuels

Article

Figure 7. Particle distributions and velocity magnitudes (in color) in the complete CD-CLC configuration for the first 800 ms of jet injection.

Figure 8. Streamlines due to variation in static pressure (in color) in the complete CD-CLC configuration for the first 800 ms of jet injection.

As the high-velocity gas enters the static bed from the central nozzle, it suddenly experiences great resistance due to the presence of the particles. Due to the momentum conservation, the static gas pressure near the central nozzle suddenly builds up and is immediately felt by the lowest pressure transducer as shown in Figure 5. Such pressure buildup can be explained by the consideration of the solid pressure, which is the additional pressure felt by a surface within the bed due to the presence of the solid column above it. Through both viscous friction and pressure gradient, the gas partially passes its kinetic energy to the

The simulation is carried out on a Dell workstation with quadcore Intel Xeon CPU. For each run from 0 to 500 ms simulation time, it requires about 48 h of CPU time. The simulation results showing the particle distribution and velocity and its comparison with experimental observation for the first 300 ms since the inception of the jet injection are presented in Figure 3. The simulation results of the average bed height and its comparison with the experiment are presented in Figure 4. Transient pressure response at the four monitoring locations of z = 2 cm, 12 cm, 22 cm, and 40 cm is shown in Figure 5. 1553


Energy & Fuels

Article

Figure 9. Static pressure in the CD-CLC system of Figure 6 at five pressure taps at 400 ms (left) and 800 ms (right).

Figure 10. Geometry outline with five pressure taps (left), mesh (middle), and frame-wire (right) of the modified CD-CLC configuration.

300 ms, respectively. Figures 3−5 clearly demonstrate that the Fluent CFD/DEM solver is capable of accurately capturing the dynamics of the solid−gas interaction in the fuel reactor of a CD-CLC configuration. Excellent agreement between the calculated expansion of bed height with the experiment is obtained including its expansion rate and finally the prediction of bubble burst. These results show the technical advantage of the Eulerian−Lagrangian modeling approach in modeling the spouted fluidized bed. One thing worth noting is that the computed transient pressure response shown in Figure 5 is likely to be underestimated due to the relatively coarse mesh employed in the flow field. A technical limitation of the Fluent CFD/DEM solver has prevented further refinement of the mesh due to the implementation of volume fraction on the computational cells.

ambient particles, forming a notable gas bubble originating from the central nozzle. With kinetic energy passed to the ambient particles, the particles start to move upward making room for the growth of the gas bubble as depicted in Figure 3. The gas bubble keeps a steady growth rate as seen from Figures 3 and 4 for about 450 ms, during which the particles above the gas bubble keep getting accelerated while the initial pressure buildup is gradually relieved. By about 450 ms, the spouted bed height expansion reaches its maximum value and begins to collapse. In Figure 5, it is seen that the initial pressure buildup is immediately felt at both locations of z = 2 and 12 cm suggesting the rapid propagation of pressure wave. However, due to the momentum exchange from gas to solids, such pressure buildup is not felt at z = 22 and 40 cm until about 80 and 1554


Energy & Fuels

Article

4. COLD FLOW SIMULATION OF A COMPETE CD-CLC SYSTEM UTILIZING SPOUTED FLUIDIZED BED AS FUEL REACTOR With the successful numerical simulation of the TU-Darmstadt experiment in the previous section, we consider the simulation of bed expansion, particle separation, and oxygen carrier recirculation in a complete three-dimensional CD-CLC model. Based on the Plexiglas test rig of the TU-Darmstadt experiment for the fuel reactor, necessary peripherals such as cyclone separator and loop seal are added to form a complete recirculating CD-CLC configuration. The geometry and the computational model of such a configuration are shown in Figure 6. In the 3D CD-CLC configuration shown in Figure 6, the fuel reactor is where the coal particles get combusted by the oxygen carrier. The reduced oxygen carrier is then transported to the cyclone for separation from the flue gas and then transferred into the down-comer. Necessary gas supply is introduced from the bottom of the loopseal to aerate the particles deposition and enable its recirculation back into the fuel reactor. To ensure the adequacy of particles deposition in the system for its recirculation, 16 000 additional particles are deposited into the downcomer and the loopseal during the initialization stage of simulation. The velocity of the central jet is increased to 40 m/s so that sufficient gas momentum can be transferred to the particles for them to reach the top of the reactor. The aeration velocity in the loopseal is set at 1 m/s. All other parameters in the simulation remain unchanged from the quasi-3D study of the TU-Darmstadt experiment reported in the previous section. With the same computer hardware platform, each run from 0 to 800 ms simulation time requires about 96 h of CPU time to complete. The particle velocity and distributions in this CD-CLC configuration for the first 800 ms are examined and are presented in Figure 7. As shown in Figure 7, a prominent and continuously growing bubble can be observed to form for the first 480 ms since inception of the high-velocity central jet injection. Due to the presence of the particle recirculation duct on one side, the bubble becomes slightly asymmetric from 80 to 400 ms as can be seen in Figure 7. From 480 ms and beyond, as the top portion of the spouted bed becomes detached from the gas bubble (can be seen from the location of the zero particle velocity regime of the bubble near the walls in Figure 7), which indicates the bubble bursting, the pressure buildup at the lower portion of the bed vanishes and results into the particles falling back into the fluidized bed. In addition, recirculation of small number of particles back into the fuel reactor is observed from about 60 to 360 ms. The leading front of the spouted bed reaches the ceiling of the fuel reactor at about 480 ms. Upon reaching the ceiling, the particles are slammed into the cyclone through the connecting duct, as can be seen from 520 to 600 ms. From 640 to 800 ms, the gas bubble in the fuel reactor completely breaks and the initial pressure buildup vanishes. Due to the absence of pressure buildup, particles at the top portion of the fuel reactor begin to fall back to the bottom. In the meantime, particles that are pneumatically transported into the cyclone get separated from the flue stream and begin to deposit in the downcomer. On the other hand, the high velocity jet makes an unsteady pathway for air to rapidly pass through the dense bed regime in the fuel reactor. Such pathway is very undesirable once formed, since no new gas bubbles can get initiated due to its presence. The upward spreading of the particles becomes very weak since most of the air bypasses the fluidized bed through the pathway, and it becomes impossible for the particles to reach the top and participate in

recirculation. This can be explained by the lack of the critical pressure buildup within an air bubble when the injected air experiences the isotropic resistance in a static bed. Additionally, the recirculation of the solid particles appears to stagnate once the bubble in the fuel reactor is formed. Because the formation of the gas bubble and solids recirculation is primarily controlled by the pressure at various locations in the system, for the purpose of providing insights into this phenomenon the streamlines originating from the fuel reactor and the loopseal bottom due to variation in the static pressure are shown in Figure 8. As shown in Figure 8, the streamlines are significantly distorted due to the presence of the solid particles for the first 480 ms. Some particles are pushed upward with the rise of the gas bubble, while other particles are pushed to the sides first and then fall back into the bed due to gravity. The mixing of gas and solids together with the solids circulation within the bed induces vortices as can be seen from 80 to 400 ms. It is expected that the formation of these vortices is favorable for CD-CLC process since they will enhance both the residual time and the exposure of coal particle with the oxygen carrier. Starting from about 480 ms, the vortices no longer exist when the high-velocity jet channeling through the bed is formed. A single vortex still exists at the top corner of the reactor due to its geometry. Pressure variations inside the CD-CLC configuration can also be examined qualitatively by Figure 8 showing streamlines colored by static pressure. From Figure 8, it can be clearly noticed that the pressure initially builds up significantly at the bottom of both the fuel reactor and the loop seal, that is, in the region where fluidization gas is supplied. As the gas bubble forms within the fuel reactor, the initial pressure buildup is gradually released as the solid column height is reduced. When the high-velocity jet

Figure 11. Schematic of the particle flow in a spouted fluidized bed. 1555


Energy & Fuels

Article

Figure 12. Particle distributions and velocity magnitude (in color) in the modified complete CD-CLC configuration for the first 1800 ms of jet injection.

channel forms in the bed after about 480 ms, the pressure buildup appears again as in-bed particle circulation becomes minimal. Additionally, it can be seen that the low pressure always forms in the cyclone as a result of the high gas velocity and opening to the atmosphere. Another important observation is that only little or even no pressure difference can be found between the two ends

of the particle recirculation duct, i.e., at the loop-seal and the lower portion of the fuel reactor where the solid recirculation duct is connected. Quantitative variation in pressure inside the CD-CLC configuration of Figure 6 at five pressure taps is shown in Figure 9 at t = 400 ms and t = 800 ms. In Figure 9, a large drop in pressure can 1556


Energy & Fuels

Article

Figure 13. Streamlines and static pressure (in color) in the complete CD-CLC configuration for the first 1800 ms of jet injection.

pressure difference, the gas flow is not likely to provide enough momentum exchange to the particles to trigger the solids recirculation. The small pressure difference between the downcomer and the fluidized bed, as can been noted from Figures 8 and 9, implies poor recirculation of solid particles from the loopseal to the fuel reactor, which confirms quantitatively the observations made qualitatively from Figure 7.

be noted as the particles move from the fluidized bed (P1) to the top of the fuel reactor (P2) and then to the cyclone (P3). As the particles deposit in the downcomer, the pressure builds up (P4) and exceeds that in the loop seal (P5) and fluidized bed (P1), providing necessary gas flow for particle recirculation. However, the pressure difference between the downcomer and the fluidized bed is very small at both time t = 400 ms and t = 800 ms. With such a small 1557


Energy & Fuels

Article

Figure 14. Static pressure in the CD-CLC system of Figure 10 at five pressure taps at 400 ms (upper left), 800 ms (upper right), 1200 ms (lower left), and 1600 ms (lower right).

5. COLD FLOW SIMULATION OF A MODIFIED COMPLETE CD-CLC SYSTEM UTILIZING SPOUTED FLUIDIZED BED AS FUEL REACTOR The continuous formation of new bubble and solid recirculation from the loopseal to the fuel reactor are the two key factors for successful CD-CLC operation, which, however, have not been satisfactorily achieved in the simulation results for the CD-CLC configuration considered in the previous section. The results of previous section clearly suggest the necessarity of making modifications to the previous CD-CLC configuration to address these two factors for its improved performance. As shown in Figure 10, we introduce three modifications. First, a chute structure is added to the bottom of the fuel reactor. The dimension of the central jet injection nozzle remains unchanged, while the two side panels of the chute are at about 60° angle with respect to the horizontal axis. Second, the loopseal and solid recirculation duct are elevated by 20 mm. Finally, the height of the fuel reactor as well the downcomer is reduced by 250 mm. The reasons behind these modifications are discussed below. The addition of the chute at the bottom is to induce gravitydriven forced particle circulation into the central jet. Such a consideration originates from the critical angle of repose (θ) for particulate materials,34−38 which is defined as the steepest angle of descent or dip of the slope relative to the horizontal plane when the particulate material on the slope face is on the verge of sliding. Due to the existence of θ, a dead zone in which particles are stagnant can be identified in a spouted fluidized bed as discussed by Takeuchi et. al39,40 and shown in Figure 11 (modified from Takeuchi et. al’s work). The existence of particle circulation dead zone is also evident in Figure 7. Therefore, the poor recirculation of solid particles observed in Figure 7 can also

be explained by the fact that the solid recirculation duct has been connected to the bottom of the fuel reactor, that is, to the dead zone. Since the particles in the dead zone are stationary regardless of ambient gas flow, high pressure is always present at the bottom of the fuel reactor causing difficulty in transporting the solid particles pneumatically from the loopseal. In contrast, the dead zone can be eliminated by introduction of the bottom chute relieving the pressure in the fuel reactor for better recirculation of solid particles. The elevation of the loop seal and recirculation duct for solid particles is based on similar logic as for adding the bottom chute, that is, to reduce the pressure felt at the connecting point of recirculation duct and fuel reactor for improved solid particles’ recirculation. The reduction in fuel reactor and downcomer height is to compensate for change in geometry when adding the bottom chute, so that sufficient amount of particles can still reach the top of the fuel reactor with unaltered fluidization conditions. The computational cost of simulation for this modified CDCLC configuration shown in Figure 10 is similar to that for the original case of Figure 6. The simulation results of particles distributions and velocity magnitude in color for this modified CD-CLC configuration for first 1800 ms are shown in Figure 12. With the proposed design modifications, the improvement in the CD-CLC system performance is very significant. As can be noted from Figure 12, a gas bubble is formed from 0 ms to about 480 ms shooting a large portion of solid particles into the cyclone. From 520 to 1340 ms, the remaining particles falls back into the bed, while the particles in the cyclone get separated from the flue stream and drop into the downcomer. Encouragingly, the formation of a second gas bubbles is observed from 1380 to 1800 ms when large number of falling particles reach the bed. 1558


Energy & Fuels

Article

achieved with the introduction of proper design modifications. More importantly, the improved system performance of the new CD-CLC system demonstrates the necessity and benefits of developing the validated CFD/DEM capability with high fidelity for future optimization designs of industrial scale CD-CLC systems.

In addition, solid recirculation from the loopseal is also observed in the process along with the elimination of dead zone in the fuel reactor. The recirculation of solid particles can also be assured by examination of the colored streamlines originating from the gas inlet in the fuel reactor and loopseal, as shown in Figure 13. As can be seen from Figure 13, the pressure at the loopseal is mostly greater than that at the fuel reactor (although not always). Such pressure difference implies some intermittent recirculation of solid particles back to the fuel reactor. Similar to the original configuration of Figure 6, for this new configuration of Figure 10, the mixing of gas/solid particles in the fuel reactor as well as their separation in the cyclone is also evident. Figure 14 provides quantitative details of the static pressure at five pressure tap locations (P1−P5) shown in Figure 10 at t = 400, 800, 1200, and 1600 ms. Comparing Figure 14 to Figure 9, that is, the static pressures at P1−P5 in the modified and original CD-CLC configurations, respectively, two significant observations can be made. First, the pressure difference between the downcomer and the fluidized bed (P4−P1) has increased by an order of magnitude. Second, such pressure difference (P4−P1) has taken a greater portion of the total pressure change in the system (P3−P4). Both of these observations confirm the improved solid particles’ recirculation in the newly designed configuration of Figure 10. In addition, it is believed that the elimination of the solids’ circulation dead zone and enhanced solid recirculation from the loopseal makes the jet experience greater solid load in the fuel reactor in the modified fluidized bed configuration, which in turns triggers the continuous formation of gas bubbles in the bed than a clear jet pathway as seen in the original bed configuration. However, more in-depth study of this aspect is needed in future work to provide better understanding of the key mechanism. Nevertheless, the simulation results shown in Figures 12, 13, and 14 clearly demonstrate that the two major concerns in the original CD-CLC configuration of Figure 6that is, failure in forming new gas bubbles and poor recirculation of solid particles have been eliminated in the modified configuration of Figure 10.

■

ASSOCIATED CONTENT

S Supporting Information *

This material is available free of charge via the Internet at http:// pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors gratefully acknowledge the financial support for this work from the Consortium for Clean Coal Utilization (CCCU) at Washington University in St. Louis.

■

REFERENCES

(1) Climate Change 2007: An Assessment of the Intergovernmental Panel on Climate Change; IPCC: Brussels, 2007. (2) Wikipedia. Oxy-fuel Combustion Process Oxy-fuel combustion process, http://en.wikipedia.org/wiki/Oxy-fuel_combustion_process (accessed Dec. 16, 2013). (3) Hong, J.; Chaudhry, G.; Brisson, J. G.; Field, R.; Gazzino, M.; Ghoniem, A. F. Energy 2009, 34, 1332−1340. (4) Hong, J.; Chaudhry, G.; Brisson, J. G.; Field, R.; Gazzino, M.; Ghoniem, A. F. In 34th International Technical Conference on Clean Coal and Fuel Systems, Clearwater, FL, 2009. (5) Lyngfelt, A.; Leckner, B.; Mattisson, T. Chem. Eng. Sci. 2001, 56, 3101−3113. (6) Ishida, M.; Jin, H. Ind. Eng. Chem. Res. 1996, 5885, 2469−2472. (7) Ishida, M.; Zheng, D.; Akehata, T. Energy 1987, 12, 147−154. (8) Ishida, M.; Jin, H.; Okamoto, T. Energy Fuels 1996, 10, 958−963. (9) Wolf, J.; Anheden, M.; Yan, J. In 18th International Pittsburgh Coal Conference, Pittsburgh, PA, 2001. (10) Marion, J. L. In 5th Annual Conference on Carbon Capture and Sequestration, Alexandria, VA, 2006. (11) Andrus, H. E.; Burns, G.; Chiu, J. H.; Lilijedahl, G. N.; Stromberg, P. T.; Thibeault, P. R. Hybrid Combustion-Gasification Chemical Looping Coal Power Technology Development PHASE III. FINAL REPORT; U.S. Department of Energy: Washington DC, 2008. (12) Mahalatkar, K.; Kuhlman, J.; Huckaby, E. D.; O′Brien, T. Chem. Eng. Sci. 2011, 66, 469−479. (13) Dennis, J. S.; Scott, S. A.; Hayhurst, A. N. J. Energy Inst. 2006, 79, 187−190. (14) Cao, Y.; Pan, W.; Green, B.; July, R. V; Re, V.; Recei, M.; December, V. Energy Fuels 2006, 20, 1836−1844. (15) Mattisson, T.; Lyngfelt, A.; Leion, H. Int. J. Greenhouse Gas Control 2009, 3, 11−19. (16) Leion, H.; Lyngfelt, A.; Mattisson, T. Chem. Eng. Res. Des. 2009, 87, 1543−1550. (17) Rubel, A.; Zhang, Y.; Liu, K.; Neathery, J. Oil Gas Sci. Technol. 2011, 66, 291−300. (18) Shen, L.; Wu, J.; Gao, Z.; Xiao, J. Combust. Flame 2009, 156, 1377−1385. (19) Kloss, C.; Goniva, C.; Aichinger, G.; Pirker, S. In Seventh International Conference on CFD in the Minerals and Process Industries; Melbourne, Australia, 2009. (20) Yang, C.; Duan, Y. Chem. Eng. Technol. 2013, 36, 1907−1914.

6. CONCLUSIONS In this paper, CFD/DEM coupled multiphase flow simulations of coal-direct chemical-looping combustion have been performed using ANSYS Fluent CFD package. The reactor level simulations have shown excellent agreement with the experimental results obtained by the research group at TU-Darmstadt. Based on the insights from reactor level modeling, simulations for complete 3D CD-CLC configuration are performed to determine the cold-flow performance of the system. It is noted from these simulations that the continuous formation of new bubble and solid recirculation from the loopseal to the fuel reactor are the two key factors for successful CD-CLC operation, which, however, were not satisfactorily achieved in the simulation results for the considered CD-CLC configuration used in the experiment at TU-Darmstadt. Therefore, a new configuration was designed to mitigate these two factors, or in other words to enhance the particle recirculation in the fuel reactor as well as in the entire system and to increase the possibility of continuous bubble formation. The simulations on original configuration suggested the importance of eliminating particle circulation dead zone in the spouted bed and the negative impact of pressure imbalance on the solid recirculation. These factors were taken into account in the design of new configuration. The simulations for the modified CD-CLC configuration have confirmed that significant improvements on the system performance can be 1559


Energy & Fuels

Article

(21) Zhu, H. P.; Zhou, Z. Y.; Yang, R. Y.; Yu, A. B. Chem. Eng. Sci. 2008, 63, 5728−5770. (22) Feng, Y.; Yu, A. 7th International Conference on CFD in the Minerals and Process Industries; Melbourne, Australia, 2009. (23) Chu, K. CFD-DEM Simulaiton of Complex Particle-Fluid Flows; The University of New South Wales: Sydney, 2010. (24) ANSYS Fluent User’s Guide; ANSYS Inc.: Canonsburg, PA, 2012; Vol. 15317. (25) ANSYS Fluent Theory Guide; ANSYS Inc.: Canonsburg, PA, 2012; Vol. 15317. (26) Morsi, S. A.; Alexander., A. J. J. Fluid Mech. 1972, 55, 193−208. (27) Levenspiel, A. H.; O.. Powder Technol. 1989, 58, 63−70. (28) Gidaspow, D.; Bezburuah, R.; D., J. Proc. 7th Eng. Foundation Conf. Fluidization 1992, 75−82. (29) Wen, C.-Y; Yu., Y. H. Chem. Eng. Prog. Symp. Ser. 1966, 62, 100− 111. (30) Syamlal, M.; O′Brien., T. J. AIChE Symp. Ser. 1989, 85, 22−31. (31) Gryczka, O.; Heinrich, S.; Deen, N. G.; van Sint Annaland, M.; Kuipers, J. a. M.; Jacob, M.; Mörl, L. Chem. Eng. Sci. 2009, 64, 3352− 3375. (32) Alobaid, F.; Ströhle, J.; Epple, B. Adv. Powder Technol. 2013, 24, 403−415. (33) Link, J. M. Development and Validation of a Discrete Particle Model of a Spout-fluid Bed Granulator; University of Twente: Enschede, Netherlands, 1975. (34) Zhou, Y.; Xu, B.; Yu, A.; Zulli, P. Phys. Rev. E 2001, 64, 021301. (35) Burkalow, A. V. A. N. Geol. Soc. Am. Bull. 1945, 56, 669−707. (36) Samadani, A.; Kudrolli, A. Phys. Rev. E 2001, 64, 051301. (37) Lee, J.; Herrmann, H. J. J. Phys. A: Math. Gen. 1993, 26, 373−383. (38) Miller, R. L.; Byrne, R. J. Sedimentology 1966, 6, 303−314. (39) Takeuchi, S.; Wang, S.; Rhodes, M. Chem. Eng. Sci. 2004, 59, 3495−3504. (40) Takeuchi, S.; Shan Wang, X.; Rhodes, M. J. Chem. Eng. Sci. 2005, 60, 1267−1276.

1560


Transient Simulations of Spouted Fluidized Bed for Coal-Direct

Recommend Documents