- Email: [email protected]
Simulation of biomass-steam gasification in fluidized bed reactors: Model setup, comparisons and preliminary predictions
Accepted Manuscript Simulation of biomass-steam gasification in fluidized bed reactors: Model setup, comparisons and preliminary predictions Linbo Yan, C. Jim Lim, Guangxi Yue, Boshu He, John R. Grace PII: DOI: Reference:
S0960-8524(16)31356-6 http://dx.doi.org/10.1016/j.biortech.2016.09.089 BITE 17112
To appear in:
Bioresource Technology
Received Date: Revised Date: Accepted Date:
13 July 2016 19 September 2016 20 September 2016
Please cite this article as: Yan, L., Jim Lim, C., Yue, G., He, B., Grace, J.R., Simulation of biomass-steam gasification in fluidized bed reactors: Model setup, comparisons and preliminary predictions, Bioresource Technology (2016), doi: http://dx.doi.org/10.1016/j.biortech.2016.09.089
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Simulation of biomass-steam gasification in fluidized bed reactors: Model setup, comparisons and preliminary predictions Linbo Yana, b, c, C. Jim Limb, Guangxi Yuec, Boshu Hea,∗ , John R. Graceb a
Institute of Combustion and Thermal Systems, School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University, Beijing 100044, China b Department of Chemical and Biological Engineering, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z3, c
Department of Thermal Engineering, Key Laboratory for Thermal Science and Power Engineering of Ministry of Education, Tsinghua University, Beijing 100084, China
Abstract: A user-defined solver integrating the solid-gas surface reactions and the multi-phase particle-in-cell (MP-PIC) approach is built based on the OpenFOAM software. The solver is tested against experiments. Then, biomass-steam gasification in a dual fluidized bed (DFB) gasifier is preliminarily predicted. It is found that the predictions agree well with the experimental results. The bed material circulation loop in the DFB can form automatically and the bed height is about 1 m. The voidage gradually increases along the height of the bed zone in the bubbling fluidized bed (BFB) of the DFB. The U-bend and cyclone can separate the syngas in the BFB and the flue gas in the circulating fluidized bed. The concentration of the gasification products is relatively higher in the conical transition section, and the dry and nitrogen-free syngas at the BFB outlet is predicted to be composed of 55% H2, 20% CO, 20% CO2 and 5% CH4. Keywords: Biomass-steam Gasification; Dual Fluidized Bed (DFB); Multiphase Particle in Cell (MP-PIC); OpenFOAM; Simulation
1 Introduction As the fourth largest energy resource on earth, biomass is relatively clean and carbon-neutral compared to coal (Qin et al., 2012). However, its low calorific value makes ∗
Corresponding author. E-mail: [email protected] (B.S. He)
the direct use of biomass uneconomic (Yan et al., 2016). One promising solution for this issue is the dual fluidized bed (DFB) gasification technology, which converts biomass into high quality syngas with high conversion of char (Barisano et al., 2016). However, the DFB gasification technology is very complex due to the pressure balance, heat balance and material balance between the two beds, as well as other factors like the turbulence, chemical reactions, solid-gas interactions and solid-solid interactions. It is therefore very important to fully understand the phenomena inside the DFB before practical design and application of such a gasifier. The traditional way is to set up a lab or pilot scale experimental device and try to understand the characteristics during its operation (Shrestha et al., 2016; Göransson et al., 2015). This method is effective, but expensive and time-consuming. In addition, it can expose the operators to unexpected hazards. Another way is to model the DFB gasifier using some semi-empirical models and correlations (Mahecha-Botero et al., 2009; Grace, 1982; Abba et al., 2003; Sit and Grace, 1981; Kunii and Levenspiel, 1991). This method is very efficient and can usually give referential predictions. However, because the modeling is usually zero- or one-dimensional, many simplifying assumptions are required and limited information can be obtained from the results (Epstein and Grace, 2011). Recently, with the rapid development of computer science, the computational fluid dynamics (CFD) technology has been widely used in simulations of hydrodynamics and reaction kinetics for different reactors, and some commercial and license-free software packages have been developed for the simulation of dense particulate flows and reactions in fluidized bed reactors, either using the Eulerian-Eulerian (EE) (Shah et al., 2015) or the Eulerian-Lagrangian (EL) approach (Saidi et al., 2015). The EE approach treats the discrete particles as continuum phase and solves the governing equations in Eulerian coordinates for both the fluid and particles. Since there is no need to track the discrete particle and the inter-particle collisions are calculated based on the kinetic theory, the EE approach is relatively efficient and can be used for large scale fluidized bed simulations (Feng et al., 2012). However, there are also some limitations. First, the detailed information for the particles is missed. Second, the particle size distribution (PSD) and its effects are hard to be considered, although coupling the EE method and the population balance model (PBM) can predict the PSD of some particles like crystals which are going through the nucleation and
growth process (Bhutani et al., 2016). Third, the development of kinetic-theory closures is very challenging especially for the polydisperse particle group (Garzo et al., 2007). Many commercial CFD packages like Fluent, CFX, STAR-CD, STAR-CCM and COMSOL can now simulate the fluidized bed reactors with the EE approach, so can some open source software packages like MFIX and OpenFOAM. Unlike the EE approach, the EL approach tracks particles in the Lagrangian coordinate. The classic EL approach for fluidized bed simulation is the discrete element method (DEM) which tracks every particle and calculates the inter-particle collisions directly (Fan et al., 2016). Thus, the DEM model is relatively accurate, but very computationally intensive since there will be billions of particles in the large scale fluidized bed gasifiers and the inter-particle collision frequency is very high in the dense bed zone. Thereby, it is unrealizable to use the DEM model for pilot or industrial scale fluidized bed simulations. The professional commercial software, EDEM, is an advanced DEM software package for simulation and analysis of particulate solids. The function of this software can be greatly extended by coupling with Fluent (Chen et al., 2016). Open source software packages like MFIX and OpenFOAM are also capable of doing simulations with the DEM model. Another EL approach, usually considered as a hybrid EE/EL approach, is called the multi-phase particle-in-cell (MP-PIC) approach (Andrews and O'Rourke, 1996; Snider, 2001). This approach uses the parcel concept to greatly reduce the total number of the computational particles because one parcel can represent several real particles of identical size, velocity, position, temperature and composition. In addition, inter-particle collisions in the MP-PIC approach are allowed for by using a continuum description of the particle stress. Unlike the EE method, the solid stress is not calculated with the kinetic theory but with an empirical correlation (Snider, 2001). These unique characteristics make the MP-PIC approach attractive for large-scale fluidized bed simulations. The commercial software packages using MP-PIC for computational particle fluid dynamics (CPFD) simulations are mainly Barracuda and Arena-flow, while non-commercial ones are mainly MFIX and OpenFOAM. As aforementioned, DFBs are promising for biomass steam gasification, but numerical simulations for their hydrodynamics and reaction kinetics are rarely reported, especially those capable of tracking the discrete particles. Cheng et al (Cheng et al., 2016) used the EE
approach to implement a two-dimensional simulation for the biomass gasification with CO2 in a fluidized bed gasifier with the commercial software, FLUENT. Sensitivity analyses were done to determine the optimal operating conditions. Li et al (Li et al., 2010) used the EE approach to numerically investigate the change of volumetric flow in fluidized bed reactors based on MFIX, with the effects of bubble size and shape factor analyzed. Gerber & Oevermann (Gerber and Oevermann, 2014) applied the two-dimensional DEM model to simulate wood gasification in a charcoal fluidized bed. Their simulation results showed good agreement with experimental data. Liu et al (Liu et al., 2015) used the MP-PIC approach to implement the simulation for biomass steam gasification in a DFB gasifier based on Barracuda, and compared their results with the experimental data. The commercial software package, Barracuda, is expert for fluidized bed simulations and its calculation is robust and fast. However, alternative physical models that deal with turbulence, turbulence-reaction interaction, and radiation are currently very limited. Since this software is completely packaged, with few user-defined models allowed to be embedded, its flexibility is also limited. In addition, the meshing method and the complexity of the computational domain are also restricted because the staggered grid concept is chosen for the discretization of the governing equations. Thus, we turned to other more flexible way to simulate fluidized bed reactors. Due to the complexity of multiphase flow with chemical reactions, it can take a long time for one to build a numerical program for the corresponding simulations because many details need to be carefully considered and the program must be bug free. One flexible and efficient way is to adopt the open source packages, with which one can add custom sub-models and modify the original models freely to implement some specific simulations. In this work, OpenFOAM (OpenFOAM®, 2015) was adopted as the basic simulation tool because it facilitates the use of many alternative physical models, standard solvers, standard utilities and libraries. Moreover, it is compatible with most popular grid generation software and other commercial CFD packages. However, in the current versions of OpenFOAM, there is no standard solver for the simulation of the fluidized bed gasification processes. Therefore, three things are done and they are the major contributions of this work. First, a user-defined solver is built based on the standard solvers and libraries in OpenFOAM
for fluidized bed gasification simulation with the MP-PIC approach. Then, to test the reliability and accuracy of this user-defined solver, the model predictions are compared with a series of experimental data. Finally, the solver is used to simulate the hydrodynamics and the reaction kinetics of an existing pilot scale DFB gasifier at UBC for biomass steam gasification. The research is intended to be of reference value for the simulation and operation of other DFB gasifiers.
2 Materials and Methods 2.1 General methods for building the user-defined solver For the simulation of the gas-solid flow, there is usually the one-way coupling, the two-way coupling and the four-way coupling approaches. The one-way coupling approach only considers the effect of gas on the solid particle. The two-way coupling considers the interaction between the gas and the particles. The four-way coupling not only considers the interaction between the gas and particles, but also considers the inter-particle collisions. For the dense particulate flow simulation, the four-way coupling approach is required. All the three approaches (EE, DEM and MP-PIC) mentioned in section 1 belongs to the four-way coupling approach. In OpenFOAM 3.0.0, there are some standard two-way coupling EL solvers like “coalChemistryFoam” and “reactingParcelFoam” which allow for the gas-solid interaction and the heterogeneous reactions but cannot address the inter-particle collisions. There is also a four-way coupling standard MP-PIC solver titled “MPPICFoam” for the isothermal dense particulate flow simulation where the inter-particle collisions are calculated while the heterogeneous reactions are not. In OpenFOAM, the solver and the library are separated, and the functions of a solver mostly depend on the libraries it called. Since OpenFOAM is an open source package written in C++, it is possible to combine the standard libraries used in the two-way EL solver and the four-way MP-PIC solver by class inheritance to generate a new user-defined library which allows for both particle surface reactions and inter-particle collisions. Then, the standard two-way EL solver is modified to allow for the voidage in the governing equations and to link the user-defined library so that it can be recompiled as a user-defined solver capable of the gasification simulation in fluidized bed
reactors. In OpenFOAM, the surface reactions are defined in the libraries while the homogeneous reactions are defined in the specific cases. Since specific surface reactions need to be considered for the biomass steam gasification, the surface reaction sub-model in the user-defined library also needs to be modified. When the user-defined library and solver are ready, a case for a specific simulation can be then set up. The initial and boundary conditions, the properties for the gaseous and solid phase, the homogeneous and heterogeneous reaction kinetic parameters, the turbulence models, the reaction models and the radiation models all need to be determined. Then, detailed solution schemes for the transient term, the gradient term, the divergence term, the Laplacian term, the interpolation scheme and the surface normal gradient schemes all need to be identified. In addition, the multigrid parameters, the iteration tolerances and the relaxation factors need to be defined. Once these issues are resolved, a simulation can be implemented. Details of these determinations will be elaborated in the following sections.
2.2 Gas-phase governing equations The main Reynolds-averaged governing equations in the solver for fluidized bed gasification include the mass conservative equation, the energy balance equation, the momentum balance equations, the species transport equations and the equation of state (Xie et al., 2012). The k-ε two-equation model that allows for the volume fraction of the gaseous phase is proper for the dense particulate flow and is used to calculate the turbulence viscosity (Ku et al., 2015). The P-1 radiation model which takes into account the contribution of both the gaseous phase and pulverized fuel is used to calculate heat source term with respect to radiative heat transfer, and the laminar finite rate model is chosen as the reaction model in this work. The form of the mass conservative equation is: ∂ (θ g ρ g ) ∂t
+ ∇ ⋅ (θ g ρ g u g ) = δ m s
(1)
where, θg, ρg, u g and t denote the gas volume fraction, the gas density, the gas velocity and the residence time, respectively. δ m s is the mass source of the gaseous phase. The momentum balance equation can be written as:
∂ (θ g ρ g u g ) ∂t
+ ∇ ⋅ (θ g ρ g u g u g ) = −θ g ∇p + F + θ g ρ g g + ∇ ⋅ (θ gτ g )
(2)
where, p is the gas thermodynamic pressure; F denotes the interphase momentum transfer rate per unit volume; g is the gravitational acceleration and τg is the gas stress tensor given by: ∂ui ∂u j + ∂x j ∂xi
τ g ,ij = µ g
2 ∂u − µ g δ ij k ∂xk 3
(3)
where, µg is the effective shear viscosity of the gas, allowing for contributions from both the laminar and turbulent viscosities. δij is the unit tensor. The energy balance equation with respect to enthalpy can be cast into:
∂ (θ g ρ g hg ) ∂t
∂p + ∇ ⋅ (θ g ρ g u g hg ) = ∇ ⋅ ( λgθ g ∇Tg ) + θ g + u g ⋅∇p + φ + q D + Q + S h + S h, p ∂t
(4)
where, hg, Tg and λg denote the gas enthalpy, the gas temperature and the gas mixture thermal conductivity, respectively; ϕ and q D denotes the viscous dissipation and enthalpy diffusion;
Q is the energy source due to radiation; Sh,p denotes the enthalpy source term due to homogeneous reactions; Sh,p denotes the enthalpy source due to heterogeneous reactions. The species transport equation can be written as:
∂ (θ g ρ gYg ,i ) ∂t
+ ∇ ⋅ (θ g ρ g ugYg ,i ) = ∇ ⋅ ( ρ g Dθ g ∇Yg ,i ) + δ m i ,chem + δ m p ,i ,chem
(5)
where, Yg,i denotes the mass fraction of species i; D is the effective mass diffusivity; δ m i ,chem is the net generation rate of species i due to homogeneous chemical reactions;
δ m p ,i ,chem denotes net generation of species i due to the heterogeneous reactions. The governing equations for k and ε are as follows:
∂ (θ g ρ g k ) ∂t ∂ (θ g ρ g ε ) ∂t
µ + ∇ ⋅ (θ g ρ g u g k ) = ∇ ⋅ θ g µ g + t ∇k + θ g Gk − θ g ρ g ε σ k
(6)
µ ε + ∇ ⋅ (θ g ρ g u g ε ) = ∇ ⋅ θ g µ g + t ∇ε + θ g (Cε 1Gk − Cε 2 ρ gε ) σε k
(7)
where, µt is the turbulent viscosity; the constants Cε1=1.44, Cε2=1.92,σk=1.0 and σε=1.3. The parallel simulation is implemented in a 4-core Inter i7-4810 CPU to accelerate the
calculation speed. The thermodynamic properties of the gas mixture are calculated with the JANAF thermodynamic coefficients. The transport properties are calculated using the Sutherland model. The unsteady continuous phase governing equations are discretized in the Eulerian coordinate with the finite volume scheme. The Euler scheme is used to discrete the transient term. The Gauss linear scheme is used to discrete the gradient term. The Gauss linear limited scheme is used to discrete the Laplacian term. For the divergence terms in most of the governing equations, different Gauss integration based schemes including the Gauss upwind unlimited scheme, the Gauss limited linear scheme etc. are used. For the divergence terms in the species transport equations and the enthalpy balance equations, the limited linear scheme is used. The linear scheme is used to interpolate values from the cell center to the face center. The limited scheme is used for the surface normal gradient terms. The PIMPLE method (Yan et al., 2016; Robertson et al., 2015), which combines the SIMPLE (Patanka, 1980) and PISO methods (Issa, 1986), is used as the solution algorithm in this solver. For the pressure correction equation, the relax factor is set as 0.5, while for other equations, the relax factors are set as 0.7. The final residuals of all these equations during iterations are controlled to be smaller than 10-6. For all the simulations, grid sensitivities are implemented to make sure that the simulation results are independent of the grid size.
2.3 Solid-phase governing equations The particle movement is addressed by the MP-PIC method. In MP-PIC, the particle dynamics are calculated by solving a transport equation of the particle distribution function (PDF), f, which is a function of particle spatial position xs, particle velocity u s, particle mass ms and time t (Andrews and O'Rourke, 1996; Snider, 2001). The particle distribution can be updated by solving this governing equation, and the local averaged particle parameters can be calculated by integrating f over the particle mass (or particle volume multiplied by particle density) and velocity. Based on the PDF equation, the continuum particle mass and momentum equations on the Eulerian grid can be derived. After the grid equations are solved, the local gas velocities, gas pressure gradients and the solids stress gradients are interpolated back to the particle positions and the particle velocities can then be explicitly updated. Note
that in some commercial software like Barracuda, the PDF is also related to the particle temperature (Liu, 2015). In this work, the MP-PIC method is only used to calculate the particle movement so the temperature is not included in f. The calculations of heat and mass transfer between the particles and the continuous phase are just like those in the two-way coupling EL solver. The transport equation for f is (O’Rourke and Snider, 2010): ∂f ∂ ( fu s ) ∂ ( fA) f D − f + + = ∂t ∂x ∂u s τD
(8)
where, A and us denote the particle acceleration and velocity, respectively; fD is the PDF for the local mass-averaged particle velocity; τD is the collision damping time. The particle acceleration is calculated by: A=
dus ∇p ∇τ s = Ds ( u g − us ) − − + g + Fs dt ρs θs ρs
(9)
where, Ds is the interphase momentum transfer coefficient; θs and ρs are the solids volume fraction and solid density, respectively; τs denotes the particle contact normal stress; Fs is the particle friction per unit mass. For Ds, the most widely used drag model is combining the Ergun, and Wen and Yu equations, with the Ergun equation used when θs exceeds 0.2 (Bokkers et al., 2004). θ s2 µ g θs ρg u g − us 150 2 2 + 1.75 θg d p θgd p Ds = 3 C θ s ρ g u − u θ −2.65 g s g 4 d d p
24 0.687 Re (1 + 0.15 Re p ) Cd = p 0.44
Re p = θ g ρ g d p ug − us / µ g
θ g <0.8 (10)
θ g ≥ 0.8
Re p < 1000
(11)
Re p ≥ 1000
(12)
where, Re p is the particle Reynolds number; dp is the particle diameter. The particle volume fraction is defined as:
θ s = 1 − θ g = ∫∫ f
ms
ρs
dms dus
(13)
The particle normal stress (pressure) can be expressed as: Psθ sβ τs = δ max θ cp − θ s , γ (1 − θ s )
(14)
where, Ps, β and γ are constants; θcp is the close-packed particle volume fraction; and δ is the Kronecker delta (Snider et al., 1998). The interphase momentum exchange rate, F, in equation (2) is calculated as: ∇p F = ∫∫ fms Ds ( u g − u s ) − dm du ρ s s s
(15)
The discrete particles and the continuous gas phase are couple via the inter-phase source terms, such as δ m s , F, S h, p and δ m p ,i ,chem .
2.4 Gasification kinetics The biomass steam gasification in DFBs is decomposed into three processes: evaporation, devolatilization, and the homogeneous and heterogeneous reactions. The moisture evaporation rate is calculated as: dms = N i As M w dt
(16)
where, As is the particle surface area; Mw is the liquid (H2O in this work) molar weight; Ni denotes the molar flux of vapor calculated by: N i = kc ( Ci ,s − Ci ,∞ )
(17)
where, kc is the mass transfer coefficient; Ci,s denotes vapor concentration at the particle surface; Ci,∞ is the vapor concentration in the bulk gas. kc, Ci,s and Ci,∞ are calculated with the equations: Sh AB =
Ci ,s =
kc d p Di ,m
1/3 = 2.0 + 0.6 Re1/2 p Sc
psat (Tp ) RTp
Ci ,∞ = X i
p RT∞
(18)
(19)
(20)
where, ShAB is Sherwood number; Di,m denotes the diffusion coefficient of vapor in the bulk;
Sc is the Schmidt number; psat is the saturation pressure at a specific temperature; Xi is the local bulk mole fraction of species i; p is the local absolute pressure; and T∞ is the local bulk temperature of the gas. The heat transferred due to moisture evaporation is calculated as:
hevap =
dms h fg dt
(21)
where, hfg is the latent heat of the liquid (Ranz and Marshall, 1952; Ranz and Marshall, 1952). The moisture evaporation process is followed by volatile pyrolysis, which is modeled using a single rate correlation in this work (Ku et al., 2014).
1.2 E 08 dmdevol = −5 E 06 exp − mdevol dt RTp
(22)
where, mdevol is the residual volatile mass; R is the ideal gas constant, 8314.5 J/(kmol·K). Char is assumed to be carbon, and the volatile composition is calculated based on the element balance. After pyrolysis, char gasification occurs. Since the DFB gasifier includes a combustor and a gasifier, reaction kinetics for the gasification and combustion of volatiles and char should be considered. As listed in Tables 1 and 2, three heterogeneous reactions and six homogeneous reactions are allowed for in this work. The shrinking core model is used to reflect the effect of char core shrinkage on the heterogeneous reaction rate. The kinetic-diffusion mechanism is used for the heterogeneous reaction rates calculation, with the overall reaction rate given (Abani and Ghoniem, 2013; Ku et al., 2015) by: rc,i = pi , g
rdiff ,i rkin ,i
(23)
rdiff ,i + rkin,i
− Ei βi rkin ,i = AT i p exp RTp rdiff ,i
0.5 (Tg + Tp ) = Ci dp
(24) 0.75
(25)
where, rc,i is the net reaction rate for species i; rdiff,i and rkin,i denote the diffusion rate and the kinetic rate, respectively; Tp and Tg are the particle and gas temperatures, respectively; Ai, β
and Ei are the typical Arrhenius parameters; Ci is the mass diffusion constant and is set as 5.0E-12 s/K0.75. For the heterogeneous char oxidation, a different correlation is used to calculate the kinetic rate (Brown et al., 1988):
(
rkin ,O 2 = Tp AO 2 + BO 2 Tp
)
(26)
Note that the shrinking core model is used only for the heterogeneous reaction rate calculations. For the hydrodynamic calculation, the particle diameter is assumed to be constant.
3 Results and Discussions 3.1 Comparison of simulation results with experimental data The simulation results from the user-defined solver in OpenFOAM are compared with experimental data to test its reliability and accuracy. First, the isothermal simulation was implemented for a single-bubble-injection (SBI) fluidized bed with monodisperse bed material particles and the predicted hydrodynamics are compared with both the experimental results and the predictions from literature based on the DEM approach. Then, simulations were carried out for biomass steam gasification in a spouted bed gasifier under different operation conditions, and the predictions are compared with the corresponding experimental data.
3.2 Simulation for single bubble injection The settings for this simulation are listed in Table 3. They are the same as those reported previously in literature (Bokkers et al., 2004; Deen et al., 2006). The background fluidization velocity is 1.7 m/s. The injection has a velocity of 20 m/s and lasts 150 ms. The computational domain is adapted with 675 hexahedra cell. The dimensions and meshes are shown in Figures1 (a1) and (a2). The software, Gambit, is used to generate mesh for all the cases in this work, and the mesh is then converted to the form that can be read by OpenFOAM through the “fluentMeshToFoam” command. The experimental pattern of the
single bubble at 150 ms and 200 ms after injection (Bokkers et al., 2004), the simulation results from the DEM model in the literature (Deen et al., 2006) and the simulation results from this work at the same times are compared in Figures 1 (b) and (c). From Figures 1 (b) and (c), it can be seen that the size and shape of the bubble, and the interface between the dense bed and the freeboard predicted in this work are similar to those predicted in the literature with the DEM model (Deen et al., 2006) and close to the experimental results (Bokkers et al., 2004), indicating that the user defined solver is reliable in the prediction of particle dynamics using the MP-PIC method.
3.3 Simulation of biomass steam gasification in a spouted bed After the comparison for the isothermal particulate flow dynamics, the simulation predictions for the biomass steam gasification in a spouted bed are compared with the experimental data reported in literature (Ku et al., 2015; Song et al., 2012). The spouted bed is a rectangular bed, with a cross-section of 40×230 mm2 and a height of 1500 mm. A 60 o conical distributor with a injector of 20 mm i.d. is mounted at the bottom of the spouted bed. The biomass feed rate is 3.0 kg/h, and the steam-to-biomass ratios, Rsb, in the simulations are 0.8 and 1.2, respectively. Argon is the pneumatic conveying medium for biomass particles and is injected at a flow rate of 1.4 m3/h. The Sauter mean diameter and density of the biomass particle are 1.5 mm and 470 kg/m3, respectively. The Sauter mean diameter and the density of the bed material are 350 µm and 2500 kg/m3, respectively. The reactor temperatures, T, in this work are set as 820 oC and 870 oC according to the experiments. The bed material specific heat is 1500 J/(kg·K), and that of biomass is 860 J/(kg·K). A total number of 200,000 parcels are considered for the bed material, with each parcel representing 5,000 real particles. The as-received basis mass fractions of moisture, ash, volatile and fixed carbon of the biomass used in the experiments and the simulation (Ku et al., 2015; Song et al., 2012) are 11.89%, 1.56%, 71.78% and 14.77%, respectively. The dry-ash-free basis mass fractions of C, H, O, and N & S of the biomass are 46.29%, 6.48%, 46.08% and 1.15%, respectively. Three simulations were carried out for biomass steam gasification in the spouted bed
under three different operating conditions. The dimensions and the meshes of the spouted bed, and the comparisons with the corresponding experimental data are depicted in Figures 2 (a)-(e). From Figures 2 (c)-(e), it is found that the simulation results agree well with the experimental data, indicating that the user-defined solver in OpenFOAM is capable of giving reliable predictions for biomass steam gasification for the cases tested. The deviations between the predictions and the experimental results are mainly caused by the chemical reaction rate calculations. Although many different kinetic triplets, including the reaction mechanism, the frequency factor and the Arrhenius activation energy, have been evaluated and those gave the best predictions are chosen in this work, the direct use of the kinetic parameters given in the literature may still cause some errors. Actually, for one specific prediction, the kinetic parameters can be calibrated so as to give very good comparison results.
3.4 Simulation of biomass steam gasification in a dual bed After the comparisons, the user-defined solver based on OpenFOAM is finally used to predict the biomass steam gasification in the DFB gasifier at UBC. The dual bed includes a bubbling fluidized bed (BFB) serving as the gasifier, a circulating fluidized bed (CFB) serving as the combustor, a U-bend connecting the BFB and the CFB, a cyclone and a downcomer. The BFB consists of a dense bed section with an i.d. of 0.279 m and a height of 1.37 m, a freeboard section with an i.d. of 0.635 m and a height 0.762 m, and a 30o conical transition section with a height of 0.305 m. Three N2 injectors are deployed in the U-bend to keep local fluidization of the particles, and transfer them to the CFB side. The CFB height is 6.0 m with an i.d. of 0.102 m. Biomass gasification takes place in the BFB, while the bed material is heated in the CFB, with the high-temperature flue gas mainly generated from methane combustion. The dimensions and meshes adapted for the DFB are shown in Figures 3 (a) and (b). In order to improve the mesh quality, the stair step mesh method is used for this dual bed gasifier. The interval size of the grid is set as 2 cm and there are totally 29983
hexahedrons. The biomass composition in this simulation is the same as that used in the spouted bed aforementioned. The properties of the inert bed material are the same as those used in the monodisperse fluidized bed above, except that the Sauter mean diameter is 200 µm. The total mass of the bed material is 105.8 kg. The total number of parcels for the bed
material is 400,000, with each parcel representing 25,000 real particles. The boundary conditions for the DFB gasifier are listed in Table 4. The particle distribution (PD) and the voidage at different times are shown in Figure 4. The species mass fractions of H2, CO, CO2, CH4 and H2O at different times are shown in Figures 5 (a)-(e), respectively. The evolution of species volume fraction of the dry and nitrogen-free syngas are shown in Figure 6. Figure 4 shows the histories of the particle distribution (the left one) and the corresponding voidage (the right one) distribution from the initial state. The voidage is the volume fraction of the gaseous phase and is represented by the symbol, alpha. The legend in Figure 4 indicates the value of this voidage (alpha) at different times. The blue particles are the silica sand particles while the red ones are the biomass particles. From Figure 4, it can be seen that the bed material circulation loop can be formed automatically from the initial state due to pressure balance between the BFB and the CFB. This is because the DFB works like a communicating vessel. When it runs stably, a stable pressure loop can be built. This pressure loop ensures that the bed materials are circulating between the BFB and CFB with specific flux. At the initial state of the simulation, because there are no particles in the CFB side, the pressure in the BFB side is much higher than that in the CFB side, and the particles are flowing from the BFB to the CFB due to pressure driving. When the particles reached the CFB chamber, the pressure in the CFB side begin to increase. The flue gas from of the bottom of the CFB then pushes them to rise along the CFB and finally go to the BFB side through the cyclone and downcomer. Then again, the bed material particles are flowing from the BFB side to the CFB side due to pressure driving. This is actually a dynamic balance of pressures in the DFB system. The bed height is about 1 m according to the dimension shown in Figure 3. The volume fraction of the bed materials is the highest in the dense bed zone of the BFB. The voidage gradually increases along the z-coordinate in the bed zone, indicating a gradual increase in the bubble size. Very few biomass particles appear in the freeboard and
no silica sand particles reach there, because the sand is much denser than the biomass particle. Between the dense bed zone and the freeboard zone, there is a splash zone, which is a typical fluid structure of the bubbling fluidized bed. It can also be seen that the particles tend to accumulate at the bottom of the cyclone and in the middle elbow zone of the downcomer, which can work like a loop seal to prevent the gas in the BFB from ascending along the downcomer. From Figures 5 (a)-(d), it can be seen that as the gasification process goes on, BFB is gradually filled with the syngas species. The mass fractions of H2, CO and CH4 in the conical section are relatively high. This is because the biomass particles tend to concentrate there due to their low density. From Figure 5 (c), it can be seen that the carbon dioxide concentration in the CFB is high. This is because methane is burnt before entering the CFB, and some of the biomass char particles also react with the residual oxygen there. This can be seen from the boundary conditions listed in Table 4. From Figure 5 (e), it can be seen that the steam concentration near the distributor is the highest and then decreases obviously a little farther from the distributor. This is because the steam is diluted by both the gasification products and the nitrogen from the feeding point. It also shows in Figures 5 (a)-(e) that the U-bend and cyclone can well prevent the syngas in the BFB and the flue gas in the CFB from mixing because little amount of syngas can be detected in the CFB side and no CO2 in the CFB goes to the BFB side, which is essential for the safe operation of the DFB and for the high quality yield of the syngas production. From Figure 6, it can be seen that the volume fractions of H2, CO, CO2 and CH4 in the dry and nitrogen-free syngas are around 55%, 20%, 20% and 5%. Since in the initial condition, the temperatures of the bed material and the BFB wall are set as 850oC, just like the stable operating temperature, the composition of the syngas does not change much as time advances.
4 Conclusions A user-defined solver based on OpenFOAM is built, and its reliability for biomass steam gasification in fluidized bed is proved by comparison with experiments. The dense bed
zone, splash zone and freeboard in the BFB can be detected, as well as the particle circulation loop. The dense bed of BFB is about 1 m and the voidage gradually increases along the height. The syngas and flue gas can be well separated. The dry and nitrogen-free syngas at the BFB outlet is composed of 55% H2, 20% CO, 20% CO2 and 5% CH4 under the given operating condition.
Acknowledgments The authors gratefully acknowledge financial support from the National Natural Science Foundation of China (NSFC, 51576014) and the China Postdoctoral Science Foundation (2015M570096).
References [1] Abani, N., Ghoniem, A.F., 2013. Large eddy simulations of coal gasification in an entrained flow gasifier. Fuel 104, 664-680. [2] Abba, I.A., Grace, J.R., Bi, H.T., Thompson, M.L., 2003. Spanning the flow regimes: generic fluidized-bed reactor model. AIChE J. 49, 1838-1848. [3] Andrews, M.J., O'Rourke, P.J., 1996. The Multiphase Particle-in-Cell (MP-PIC) Method for Dense Particle Flows. Int. J. Multiphase Flow 22, 379-402. [4] Barisano, D., Canneto, G., Alvino, F.N.E., Pinto, G., Villone, A., Carnevale, M., Valerio, V., Battafarano, A., Braccio, G., 2016. Steam/oxygen biomass gasification at pilot scale in an internally circulating bubbling fluidized bed reactor. Fuel Process. Technol. 141, 74-81. [5] Bhutani, G., Brito-Parada, P.R., Cilliers, J.J., 2016. Polydispersed flow modelling using population balances in an adaptive mesh finite element framework. Comput. Chem. Eng. 87, 208-225. [6] Bokkers, G.A., Annaland, M.S., Kuipers, J.A.M., 2004. Mixing and segregation in a bidisperse gas–solid fluidised bed: a numerical and experimental study. Powder Technol. 140, 176-186. [7] Brown, B.W., Smoot, L.D., Smith, P.J, Hedman, P.O., 1988. Measurement and prediction of entrained-flow gasification processes. AIChE J. 34, 435-446. [8] Chen, J.K., Wang, Y.S., Li, X.F., He, R.Y., Han, S., Chen, Y.L., 2015. Reprint of “Erosion prediction
of liquid-particle two-phase flow in pipeline elbows via CFD–DEM coupling method”. Powder Technol. 282, 25-31. [9] Cheng, Y.P., Thow, Z.H, Wang, C.H., 2016. Biomass gasification with CO2 in a fluidized bed. Powder Technol. 296, 87-101. [10] Deen, N.G., Annaland, M.V.S., Kuipers, J.A.M., 2006. Detailed computational and experimental fluid dynamics of fluidized beds. Appl. Math Model 30, 1459-1471. [11] Epstein, N., Grace, J.R., 2011. Spouted and spout-fluidized beds. New York: Cambridge University Press. [12] Fan, H.J., Mei, D.F., Tian, F.G., Cui, X., Zhang, M.C., 2016. DEM simulation of different particle ejection mechanisms in a fluidized bed with and without cohesive interparticle forces. Powder Technol. 288, 228-240. [13] Feng, Y.Q., Swenser-Smith, T.S., Witt, P.J., Doblin, C., Lim, S., Schwarz, M.P., 2012. CFD modeling of gas–solid flow in an internally circulating fluidized bed. Powder Technol. 219, 78-85. [14] Fluent, Inc., 2006. FLUENT6.3-User’s Guide; Fluent Inc., New York. [15] Garzo, V., Dufty, J.W., Hrenya, C.M., 2007. Enskog theory for polydisperse granular mixtures. I. Navier-Stokes order transport. Physical Review E 76, 1-27. [16] Gerber, S., Oevermann, M., 2014. A two dimensional Euler–Lagrangian model of wood gasification in a charcoal bed – Part I: model description and base scenario. Fuel 115, 385-400. [17] Gómez-Barea, A., Leckner, B., 2010. Modeling of biomass gasification in fluidized bed. Prog. Energy Combust Sci. 36, 444-509. [18] Göransson, K., Söderlind, U., Engstrand, P., Zhang, W., 2015. An experimental study on catalytic bed materials in a biomass dual fluidised bed gasifier. Renewable Energy 81, 251-261. [19] Grace, J.R., 1982. Fluidized-bed hydrodynamics. In Handbook of Multiphase Systems; Hetsoni G, Ed.; Washington, D.C.: Hemisphere. [20] Issa, R.I., 1986. Solution of Implicitly Discretized Fluid Flow Equations by Operator Splitting. J. Comput. Phys. 62, 40-65. [21] Ku, X.K., Li, T., Løvås, T., 2015. CFD-DEM simulation of biomass gasification with steam in a fluidized bed reactor. Chem. Eng. Sci. 122, 270-283. [22] Ku, X.K., Li, T., Løvås, T., 2014. Eulerian-Lagrangian simulation of biomass gasification behavior in a high-temperature entrained-flow reactor. Energy Fuels 28, 5184-5196.
[23] Kunii, D., Levenspiel, O., 1991. Fluidization Engineering, 2nd ed., Boston: Butterworth-Heinemann. [24] Li, T.W., Mahecha-Botero, A., Grace, J.R., 2010. Computational Fluid Dynamic Investigation of Change of Volumetric Flow in Fluidized-Bed Reactors. Ind. Eng. Chem. Res. 45, 6780-6789. [25] Liu, H., Cattolica, R.J., Seiser, R., Liao, C.H., 2015. Three-dimensional full-loop simulation of a dual fluidized-bed biomass gasifier. Appl. Energy 160, 489-501. [26] Mahecha-Botero, A., Grace, J.R., Elnashaie, S.S.E.H., Lim, C.J., 2009. Advances in modeling of fluidized-bed catalytic reactors: A comprehensive review. Chem. Eng. Commun. 196, 1375-1405. [27] OpenFOAM®, 2015. The Open Source CFD Toolbox, Ver. 3.0.0. ESI Group, Paris, France. [28] O’Rourke, P.J., Snider, D.M., 2010. An improved collision damping time for MP-PIC calculations of dense particle flows with applications to polydisperse sedimenting beds and colliding particle jets. Chem. Eng. Sci. 65, 6014-6028. [29] Patankar, S.V., 1980. Numerical Heat Transfer and Fluid Flow. Taylor & Francis. [30] Qin, K., Jensen, P.A., Lin, W., Jensen, A.D., 2012. Biomass gasification behavior in an entrained flow reactor: Gas product distribution and soot formation. Energy Fuels 26, 5992-6002. [31] Ranz, W.E., Marshall, W.R.J., 1952. Evaporation from Drops, Part I. Chem. Eng. Prog. 48: 141-146. [32] Ranz, W.E., Marshall, W.R.J., 1952. Evaporation from Drops, Part II. Chem. Eng. Prog. 48, 173-180. [33] Robertson, E., Choudhury, V., Bhushan, S., Walters, D.K., 2015. Validation of OpenFOAM numerical methods and turbulence models for incompressible bluff body flows. Comput. Fluids 123, 122-145; [34] Saidi, M., Tabrizi, H.B., Grace, J.R., Lim, C.J., 2015. Hydrodynamic investigation of gas-solid flow in rectangular spout-fluid bed using CFD-DEM modeling. Powder Technol. 284, 355-364. [35] Shah, S., Myöhänen, K., Kallio, S., Ritvanen, J., Hyppänen, T., 2015. CFD modeling of gas–solids flow in a large scale circulating fluidized bed furnace. Powder Technol. 274, 239-249. [36] Shrestha, S., Ali, B.S., Jan, B.M., Lim, M.T., Sheikh, K.E., 2016. Hydrodynamic properties of a cold model of dual fluidized bed gasifier: A modeling and experimental investigation. Chem. Eng. Res. Des. 109, 791-805. [37] Sit, S.P., Grace, J.R., 1981. Effect of bubble interaction on interphase mass transfer in gas fluidized beds. Chem. Eng. Sci. 36, 327-335. [38] Snider, D.M., 2001. An Incompressible Three-Dimensional Multiphase Particle-in-Cell Model for Dense Particle Flows. J. Comput. Phys. 170, 523-549.
[39] Snider, D.M., O’Rourke, P.J., Andrews, M.J., 1998. Sediment flow in inclined vessels calculated using a multiphase particle-in-cell model for dense particle flows. Int. J. Multiphase Flow 24, 1359-1382. [40] Song T, Wu JH, Shen LH, Xiao J. Experimental investigation on hydrogen production from biomass gasification in interconnected fluidized beds. Biomass Bioenergy 2012; 36: 258-267. [41] Umeki, K., Yamamoto, K., Namioka, T., Yoshikawa, K., 2010. High temperature steam-only gasification of woody biomass. Appl. Energy 87, 791-798. [42] Wen, C.Y., Chaung, T.Z., 1979. Entrainment coal gasification modeling. Ind. Eng. Chem. Process Des. Dev. 18, 684-695. [43] Xie, J., Zhong, W.Q., Jin, B.S., Shao, Y.J., Liu, H., 2012. Simulation on gasification of forestry residues in fluidized beds by Eulerian–Lagrangian approach. Bioresour. Technol. 121, 36-46. [44] Yan, L.B., Yue, G.X., He, B.S., 2016. Thermodynamic analyses of a biomass-coal co-gasification power generation system. Bioresour. Technol. 205, 133-141.
Figure captions: Figure 1. Comparisons with results from experiments and literature at 150 ms and 200 ms Figure 2. Meshes of the spouted bed and comparisons with experimental results Figure 3. Dimensions and meshes of the dual fluidized bed gasifier Figure 4. Particle (left) and voidage (right) distributions at different times Figure 5. Species mass fractions at 3, 6, 9 and 12 s Figure 6. Evolution of the dry nitrogen-free syngas composition
c 15
m
1c m
m 5c 1.
(a1)Dimensions of the SBI fluidized bed
(a2)Meshes of the SBI fluidized bed
(b1 ) Results from experiments at
(b2) Predictions from literature at
(b3) Predictions from this work at
150 ms (Bokkers et al., 2004)
150 ms (Deen et al., 2006)
150 ms
(c1) Results from experiments at
(c2) Predictions from literature at
(c3) Predictions from this work at
200 ms (Bokkers et al., 2004)
200 ms (Deen et al., 2006)
200 ms
Figure 1 Comparisons with results from experiments and literature at 150 ms and 200 ms
(a) Dimensions of the spouted bed 0.6
exp cal
0.6
0.5
exp cal
0.5
T=1093 K, Rsb=1.2 Mass fraction
0.4
0.3
0.2
T=1143 K, Rsb=1.2
0.4 0.3 0.2 0.1
0.1
0.0
0.0
H2
CO
CO2
CH4
H2
CO
Species
CO2
CH4
Species
(c) Comparisons at T=1093 K and Rsb=1.2
(d) Comparisons at T=1143 K and Rsb=1.2
0.6
exp cal
0.5
Mass fraction
Mass fraction
(b) Meshes of the spouted bed
T=1143 K, Rsb=1.2
0.4 0.3 0.2 0.1 0.0 H2
CO
CO2
CH4
Species
(e) Comparisons at T=1093 K and Rsb=0.8 Figure 2 Meshes of the spouted bed and comparisons with experimental results
Ø1 0. 19 cm
(a) Dimensions of the DFB gasifier
(b) Meshes of the DFB gasifier
Figure 3 Dimensions and meshes of the DFB gasifier
(a) 0 s
(b) 3 s
(c) 6 s
(d) 9 s
Figure 4 Particle (left) and voidage (right) distributions at different times
(e) 12 s
(a) H2 mass fraction at 3, 6, 9 and 12 s
(b) CO mass fraction at 3, 6, 9 and 12 s
(c) CO2 mass fraction at 3, 6, 9 and 12 s
(d) CH4 mass fraction at 3, 6, 9 and 12 s
(e) H2O mass fraction at 3, 6, 9 and 12 s Figure 5 Species mass fractions at 3, 6, 9 and 12 s
0.6
Species volume fraction
0.5
H2 CO CO2
0.4
CH4 0.3
0.2
0.1
0.0 2
3
4
5
6
7
8
9
10
11
time (s) Figure 6 Evolution of the dry nitrogen-free syngas composition
12
Table captions: Table 1 Heterogeneous reactions and corresponding kinetic parameters Table 2 Homogeneous reactions and corresponding kinetic correlations Table 3 Settings for the single bubble injected into a monodisperse two-dimensional fluidized bed Table 4 Boundary conditions for the DFB gasifier
Table 1 Heterogeneous reactions and corresponding kinetic parameters (Ku et al., 2015; Brown et al., 1988) Reactions
Kinetic parameters AO2 = −1.68 × 10
−2
s/(m ⋅ K) , BO2 = 1.32 × 10−5 s/(m ⋅ K)
R1
C+0.5O2 → CO
R2
C+H 2 O → CO+H 2
AH 2O = 45.6 s/(m ⋅ K) , EH2 O = 4.37 × 107 J/kmol , β H 2O = 1.0
R3
C+CO 2 → 2CO
ACO2 = 8.3 s/(m ⋅ K) , ECO2 = 4.37 × 10 7 J/kmol , β CO2 = 1.0
R4 R5
Table 2 Homogeneous reactions and corresponding kinetic correlations Reactions Kinetic correlations (kmol/m3/s) Ref. r4 = 0.312 exp(−30000 / (1.987Tg ))CCH CH 4 +H 2 O → CO+3H 2 (Wen and Chaung, 1979) r5 = 2.5 × 108 exp( −16597.5 / Tg )CCO CH O CO+H 2 O → CO 2 +H 2 (Umeki et al, 1996) 4
2
9
r6 = 9.43 × 10 exp(−20563.51/ Tg )CCO2 CH 2
R6
CO2 +H2 → CO+H2 O
R7
CO+0.5O 2 → CO 2
r7 = 1.0 ×1010 exp(−15154.25 / Tg )CCO CO0.52 CH0.52 O
R8
H 2 +0.5O2 → H 2 O
r8 = 2.2 × 109 exp( −13109.63 / Tg )CH 2 CO2
R9
CH 4 +2O2 → CO 2 +2H 2 O
0.2 r9 = 2.119 × 1011 exp(−24379.1 / Tg )CCH CO1.32 4
(Umeki et al, 1996) (Gómez-Barea Leckner, 2010) (Gómez-Barea Leckner, 2010) (Fluent, 2006)
and and
Table 3 Settings for the single bubble injected into a monodisperse two-dimensional fluidized bed Bed dimensions Number of cells Reactor width, x = 0.15 m Reactor thickness, y = 0.015 m Reactor height, z = 0.45 m
NX = 15 NY = 1 NZ = 45
Parameters
Value
Background velocity Central jet (1 cm width) pulse velocity Central jet pulse duration Particle density Particle diameter Drag model
1.7 m/s 20 m/s 150 ms 2526 kg/m3 2.5 mm Ergun/Wen & Yu
Table 4 Boundary conditions for the DFB gasifier Parameter Value Biomass feed rate Steam-to-biomass mass ratio Steam inlet temperature BFB wall temperature Other wall conditions Bed material initial temperature CFB inlet gas temperature CFB inlet gas composition CFB inlet gas flow rate U-bend N2 flow rate BFB N2 purging velocity
10 kg/h 1.0 600 oC 850 oC Adiabatic 850 oC 1227 oC 8.37% O2, 73.99% N2, 9.70% CO2, 7.94% H2O 155.5 kg/h 15.29 m3 2.0 m/s
Highlights of the manuscript
A user-defined solver for biomass steam gasification in fluidized bed is built. The solver is tested against experimental data in literature and works well. Biomass steam gasification in the DFB at UBC is preliminary predicted. Flue gas and syngas can be well separated by the U-bend and cyclone. Dry N2-free syngas is composed of 55% H2, 20% CO, 20% CO2 and 5% CH4.
S0960-8524(16)31356-6 http://dx.doi.org/10.1016/j.biortech.2016.09.089 BITE 17112
To appear in:
Bioresource Technology
Received Date: Revised Date: Accepted Date:
13 July 2016 19 September 2016 20 September 2016
Please cite this article as: Yan, L., Jim Lim, C., Yue, G., He, B., Grace, J.R., Simulation of biomass-steam gasification in fluidized bed reactors: Model setup, comparisons and preliminary predictions, Bioresource Technology (2016), doi: http://dx.doi.org/10.1016/j.biortech.2016.09.089
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Simulation of biomass-steam gasification in fluidized bed reactors: Model setup, comparisons and preliminary predictions Linbo Yana, b, c, C. Jim Limb, Guangxi Yuec, Boshu Hea,∗ , John R. Graceb a
Institute of Combustion and Thermal Systems, School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University, Beijing 100044, China b Department of Chemical and Biological Engineering, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z3, c
Department of Thermal Engineering, Key Laboratory for Thermal Science and Power Engineering of Ministry of Education, Tsinghua University, Beijing 100084, China
Abstract: A user-defined solver integrating the solid-gas surface reactions and the multi-phase particle-in-cell (MP-PIC) approach is built based on the OpenFOAM software. The solver is tested against experiments. Then, biomass-steam gasification in a dual fluidized bed (DFB) gasifier is preliminarily predicted. It is found that the predictions agree well with the experimental results. The bed material circulation loop in the DFB can form automatically and the bed height is about 1 m. The voidage gradually increases along the height of the bed zone in the bubbling fluidized bed (BFB) of the DFB. The U-bend and cyclone can separate the syngas in the BFB and the flue gas in the circulating fluidized bed. The concentration of the gasification products is relatively higher in the conical transition section, and the dry and nitrogen-free syngas at the BFB outlet is predicted to be composed of 55% H2, 20% CO, 20% CO2 and 5% CH4. Keywords: Biomass-steam Gasification; Dual Fluidized Bed (DFB); Multiphase Particle in Cell (MP-PIC); OpenFOAM; Simulation
1 Introduction As the fourth largest energy resource on earth, biomass is relatively clean and carbon-neutral compared to coal (Qin et al., 2012). However, its low calorific value makes ∗
Corresponding author. E-mail: [email protected] (B.S. He)
the direct use of biomass uneconomic (Yan et al., 2016). One promising solution for this issue is the dual fluidized bed (DFB) gasification technology, which converts biomass into high quality syngas with high conversion of char (Barisano et al., 2016). However, the DFB gasification technology is very complex due to the pressure balance, heat balance and material balance between the two beds, as well as other factors like the turbulence, chemical reactions, solid-gas interactions and solid-solid interactions. It is therefore very important to fully understand the phenomena inside the DFB before practical design and application of such a gasifier. The traditional way is to set up a lab or pilot scale experimental device and try to understand the characteristics during its operation (Shrestha et al., 2016; Göransson et al., 2015). This method is effective, but expensive and time-consuming. In addition, it can expose the operators to unexpected hazards. Another way is to model the DFB gasifier using some semi-empirical models and correlations (Mahecha-Botero et al., 2009; Grace, 1982; Abba et al., 2003; Sit and Grace, 1981; Kunii and Levenspiel, 1991). This method is very efficient and can usually give referential predictions. However, because the modeling is usually zero- or one-dimensional, many simplifying assumptions are required and limited information can be obtained from the results (Epstein and Grace, 2011). Recently, with the rapid development of computer science, the computational fluid dynamics (CFD) technology has been widely used in simulations of hydrodynamics and reaction kinetics for different reactors, and some commercial and license-free software packages have been developed for the simulation of dense particulate flows and reactions in fluidized bed reactors, either using the Eulerian-Eulerian (EE) (Shah et al., 2015) or the Eulerian-Lagrangian (EL) approach (Saidi et al., 2015). The EE approach treats the discrete particles as continuum phase and solves the governing equations in Eulerian coordinates for both the fluid and particles. Since there is no need to track the discrete particle and the inter-particle collisions are calculated based on the kinetic theory, the EE approach is relatively efficient and can be used for large scale fluidized bed simulations (Feng et al., 2012). However, there are also some limitations. First, the detailed information for the particles is missed. Second, the particle size distribution (PSD) and its effects are hard to be considered, although coupling the EE method and the population balance model (PBM) can predict the PSD of some particles like crystals which are going through the nucleation and
growth process (Bhutani et al., 2016). Third, the development of kinetic-theory closures is very challenging especially for the polydisperse particle group (Garzo et al., 2007). Many commercial CFD packages like Fluent, CFX, STAR-CD, STAR-CCM and COMSOL can now simulate the fluidized bed reactors with the EE approach, so can some open source software packages like MFIX and OpenFOAM. Unlike the EE approach, the EL approach tracks particles in the Lagrangian coordinate. The classic EL approach for fluidized bed simulation is the discrete element method (DEM) which tracks every particle and calculates the inter-particle collisions directly (Fan et al., 2016). Thus, the DEM model is relatively accurate, but very computationally intensive since there will be billions of particles in the large scale fluidized bed gasifiers and the inter-particle collision frequency is very high in the dense bed zone. Thereby, it is unrealizable to use the DEM model for pilot or industrial scale fluidized bed simulations. The professional commercial software, EDEM, is an advanced DEM software package for simulation and analysis of particulate solids. The function of this software can be greatly extended by coupling with Fluent (Chen et al., 2016). Open source software packages like MFIX and OpenFOAM are also capable of doing simulations with the DEM model. Another EL approach, usually considered as a hybrid EE/EL approach, is called the multi-phase particle-in-cell (MP-PIC) approach (Andrews and O'Rourke, 1996; Snider, 2001). This approach uses the parcel concept to greatly reduce the total number of the computational particles because one parcel can represent several real particles of identical size, velocity, position, temperature and composition. In addition, inter-particle collisions in the MP-PIC approach are allowed for by using a continuum description of the particle stress. Unlike the EE method, the solid stress is not calculated with the kinetic theory but with an empirical correlation (Snider, 2001). These unique characteristics make the MP-PIC approach attractive for large-scale fluidized bed simulations. The commercial software packages using MP-PIC for computational particle fluid dynamics (CPFD) simulations are mainly Barracuda and Arena-flow, while non-commercial ones are mainly MFIX and OpenFOAM. As aforementioned, DFBs are promising for biomass steam gasification, but numerical simulations for their hydrodynamics and reaction kinetics are rarely reported, especially those capable of tracking the discrete particles. Cheng et al (Cheng et al., 2016) used the EE
approach to implement a two-dimensional simulation for the biomass gasification with CO2 in a fluidized bed gasifier with the commercial software, FLUENT. Sensitivity analyses were done to determine the optimal operating conditions. Li et al (Li et al., 2010) used the EE approach to numerically investigate the change of volumetric flow in fluidized bed reactors based on MFIX, with the effects of bubble size and shape factor analyzed. Gerber & Oevermann (Gerber and Oevermann, 2014) applied the two-dimensional DEM model to simulate wood gasification in a charcoal fluidized bed. Their simulation results showed good agreement with experimental data. Liu et al (Liu et al., 2015) used the MP-PIC approach to implement the simulation for biomass steam gasification in a DFB gasifier based on Barracuda, and compared their results with the experimental data. The commercial software package, Barracuda, is expert for fluidized bed simulations and its calculation is robust and fast. However, alternative physical models that deal with turbulence, turbulence-reaction interaction, and radiation are currently very limited. Since this software is completely packaged, with few user-defined models allowed to be embedded, its flexibility is also limited. In addition, the meshing method and the complexity of the computational domain are also restricted because the staggered grid concept is chosen for the discretization of the governing equations. Thus, we turned to other more flexible way to simulate fluidized bed reactors. Due to the complexity of multiphase flow with chemical reactions, it can take a long time for one to build a numerical program for the corresponding simulations because many details need to be carefully considered and the program must be bug free. One flexible and efficient way is to adopt the open source packages, with which one can add custom sub-models and modify the original models freely to implement some specific simulations. In this work, OpenFOAM (OpenFOAM®, 2015) was adopted as the basic simulation tool because it facilitates the use of many alternative physical models, standard solvers, standard utilities and libraries. Moreover, it is compatible with most popular grid generation software and other commercial CFD packages. However, in the current versions of OpenFOAM, there is no standard solver for the simulation of the fluidized bed gasification processes. Therefore, three things are done and they are the major contributions of this work. First, a user-defined solver is built based on the standard solvers and libraries in OpenFOAM
for fluidized bed gasification simulation with the MP-PIC approach. Then, to test the reliability and accuracy of this user-defined solver, the model predictions are compared with a series of experimental data. Finally, the solver is used to simulate the hydrodynamics and the reaction kinetics of an existing pilot scale DFB gasifier at UBC for biomass steam gasification. The research is intended to be of reference value for the simulation and operation of other DFB gasifiers.
2 Materials and Methods 2.1 General methods for building the user-defined solver For the simulation of the gas-solid flow, there is usually the one-way coupling, the two-way coupling and the four-way coupling approaches. The one-way coupling approach only considers the effect of gas on the solid particle. The two-way coupling considers the interaction between the gas and the particles. The four-way coupling not only considers the interaction between the gas and particles, but also considers the inter-particle collisions. For the dense particulate flow simulation, the four-way coupling approach is required. All the three approaches (EE, DEM and MP-PIC) mentioned in section 1 belongs to the four-way coupling approach. In OpenFOAM 3.0.0, there are some standard two-way coupling EL solvers like “coalChemistryFoam” and “reactingParcelFoam” which allow for the gas-solid interaction and the heterogeneous reactions but cannot address the inter-particle collisions. There is also a four-way coupling standard MP-PIC solver titled “MPPICFoam” for the isothermal dense particulate flow simulation where the inter-particle collisions are calculated while the heterogeneous reactions are not. In OpenFOAM, the solver and the library are separated, and the functions of a solver mostly depend on the libraries it called. Since OpenFOAM is an open source package written in C++, it is possible to combine the standard libraries used in the two-way EL solver and the four-way MP-PIC solver by class inheritance to generate a new user-defined library which allows for both particle surface reactions and inter-particle collisions. Then, the standard two-way EL solver is modified to allow for the voidage in the governing equations and to link the user-defined library so that it can be recompiled as a user-defined solver capable of the gasification simulation in fluidized bed
reactors. In OpenFOAM, the surface reactions are defined in the libraries while the homogeneous reactions are defined in the specific cases. Since specific surface reactions need to be considered for the biomass steam gasification, the surface reaction sub-model in the user-defined library also needs to be modified. When the user-defined library and solver are ready, a case for a specific simulation can be then set up. The initial and boundary conditions, the properties for the gaseous and solid phase, the homogeneous and heterogeneous reaction kinetic parameters, the turbulence models, the reaction models and the radiation models all need to be determined. Then, detailed solution schemes for the transient term, the gradient term, the divergence term, the Laplacian term, the interpolation scheme and the surface normal gradient schemes all need to be identified. In addition, the multigrid parameters, the iteration tolerances and the relaxation factors need to be defined. Once these issues are resolved, a simulation can be implemented. Details of these determinations will be elaborated in the following sections.
2.2 Gas-phase governing equations The main Reynolds-averaged governing equations in the solver for fluidized bed gasification include the mass conservative equation, the energy balance equation, the momentum balance equations, the species transport equations and the equation of state (Xie et al., 2012). The k-ε two-equation model that allows for the volume fraction of the gaseous phase is proper for the dense particulate flow and is used to calculate the turbulence viscosity (Ku et al., 2015). The P-1 radiation model which takes into account the contribution of both the gaseous phase and pulverized fuel is used to calculate heat source term with respect to radiative heat transfer, and the laminar finite rate model is chosen as the reaction model in this work. The form of the mass conservative equation is: ∂ (θ g ρ g ) ∂t
+ ∇ ⋅ (θ g ρ g u g ) = δ m s
(1)
where, θg, ρg, u g and t denote the gas volume fraction, the gas density, the gas velocity and the residence time, respectively. δ m s is the mass source of the gaseous phase. The momentum balance equation can be written as:
∂ (θ g ρ g u g ) ∂t
+ ∇ ⋅ (θ g ρ g u g u g ) = −θ g ∇p + F + θ g ρ g g + ∇ ⋅ (θ gτ g )
(2)
where, p is the gas thermodynamic pressure; F denotes the interphase momentum transfer rate per unit volume; g is the gravitational acceleration and τg is the gas stress tensor given by: ∂ui ∂u j + ∂x j ∂xi
τ g ,ij = µ g
2 ∂u − µ g δ ij k ∂xk 3
(3)
where, µg is the effective shear viscosity of the gas, allowing for contributions from both the laminar and turbulent viscosities. δij is the unit tensor. The energy balance equation with respect to enthalpy can be cast into:
∂ (θ g ρ g hg ) ∂t
∂p + ∇ ⋅ (θ g ρ g u g hg ) = ∇ ⋅ ( λgθ g ∇Tg ) + θ g + u g ⋅∇p + φ + q D + Q + S h + S h, p ∂t
(4)
where, hg, Tg and λg denote the gas enthalpy, the gas temperature and the gas mixture thermal conductivity, respectively; ϕ and q D denotes the viscous dissipation and enthalpy diffusion;
Q is the energy source due to radiation; Sh,p denotes the enthalpy source term due to homogeneous reactions; Sh,p denotes the enthalpy source due to heterogeneous reactions. The species transport equation can be written as:
∂ (θ g ρ gYg ,i ) ∂t
+ ∇ ⋅ (θ g ρ g ugYg ,i ) = ∇ ⋅ ( ρ g Dθ g ∇Yg ,i ) + δ m i ,chem + δ m p ,i ,chem
(5)
where, Yg,i denotes the mass fraction of species i; D is the effective mass diffusivity; δ m i ,chem is the net generation rate of species i due to homogeneous chemical reactions;
δ m p ,i ,chem denotes net generation of species i due to the heterogeneous reactions. The governing equations for k and ε are as follows:
∂ (θ g ρ g k ) ∂t ∂ (θ g ρ g ε ) ∂t
µ + ∇ ⋅ (θ g ρ g u g k ) = ∇ ⋅ θ g µ g + t ∇k + θ g Gk − θ g ρ g ε σ k
(6)
µ ε + ∇ ⋅ (θ g ρ g u g ε ) = ∇ ⋅ θ g µ g + t ∇ε + θ g (Cε 1Gk − Cε 2 ρ gε ) σε k
(7)
where, µt is the turbulent viscosity; the constants Cε1=1.44, Cε2=1.92,σk=1.0 and σε=1.3. The parallel simulation is implemented in a 4-core Inter i7-4810 CPU to accelerate the
calculation speed. The thermodynamic properties of the gas mixture are calculated with the JANAF thermodynamic coefficients. The transport properties are calculated using the Sutherland model. The unsteady continuous phase governing equations are discretized in the Eulerian coordinate with the finite volume scheme. The Euler scheme is used to discrete the transient term. The Gauss linear scheme is used to discrete the gradient term. The Gauss linear limited scheme is used to discrete the Laplacian term. For the divergence terms in most of the governing equations, different Gauss integration based schemes including the Gauss upwind unlimited scheme, the Gauss limited linear scheme etc. are used. For the divergence terms in the species transport equations and the enthalpy balance equations, the limited linear scheme is used. The linear scheme is used to interpolate values from the cell center to the face center. The limited scheme is used for the surface normal gradient terms. The PIMPLE method (Yan et al., 2016; Robertson et al., 2015), which combines the SIMPLE (Patanka, 1980) and PISO methods (Issa, 1986), is used as the solution algorithm in this solver. For the pressure correction equation, the relax factor is set as 0.5, while for other equations, the relax factors are set as 0.7. The final residuals of all these equations during iterations are controlled to be smaller than 10-6. For all the simulations, grid sensitivities are implemented to make sure that the simulation results are independent of the grid size.
2.3 Solid-phase governing equations The particle movement is addressed by the MP-PIC method. In MP-PIC, the particle dynamics are calculated by solving a transport equation of the particle distribution function (PDF), f, which is a function of particle spatial position xs, particle velocity u s, particle mass ms and time t (Andrews and O'Rourke, 1996; Snider, 2001). The particle distribution can be updated by solving this governing equation, and the local averaged particle parameters can be calculated by integrating f over the particle mass (or particle volume multiplied by particle density) and velocity. Based on the PDF equation, the continuum particle mass and momentum equations on the Eulerian grid can be derived. After the grid equations are solved, the local gas velocities, gas pressure gradients and the solids stress gradients are interpolated back to the particle positions and the particle velocities can then be explicitly updated. Note
that in some commercial software like Barracuda, the PDF is also related to the particle temperature (Liu, 2015). In this work, the MP-PIC method is only used to calculate the particle movement so the temperature is not included in f. The calculations of heat and mass transfer between the particles and the continuous phase are just like those in the two-way coupling EL solver. The transport equation for f is (O’Rourke and Snider, 2010): ∂f ∂ ( fu s ) ∂ ( fA) f D − f + + = ∂t ∂x ∂u s τD
(8)
where, A and us denote the particle acceleration and velocity, respectively; fD is the PDF for the local mass-averaged particle velocity; τD is the collision damping time. The particle acceleration is calculated by: A=
dus ∇p ∇τ s = Ds ( u g − us ) − − + g + Fs dt ρs θs ρs
(9)
where, Ds is the interphase momentum transfer coefficient; θs and ρs are the solids volume fraction and solid density, respectively; τs denotes the particle contact normal stress; Fs is the particle friction per unit mass. For Ds, the most widely used drag model is combining the Ergun, and Wen and Yu equations, with the Ergun equation used when θs exceeds 0.2 (Bokkers et al., 2004). θ s2 µ g θs ρg u g − us 150 2 2 + 1.75 θg d p θgd p Ds = 3 C θ s ρ g u − u θ −2.65 g s g 4 d d p
24 0.687 Re (1 + 0.15 Re p ) Cd = p 0.44
Re p = θ g ρ g d p ug − us / µ g
θ g <0.8 (10)
θ g ≥ 0.8
Re p < 1000
(11)
Re p ≥ 1000
(12)
where, Re p is the particle Reynolds number; dp is the particle diameter. The particle volume fraction is defined as:
θ s = 1 − θ g = ∫∫ f
ms
ρs
dms dus
(13)
The particle normal stress (pressure) can be expressed as: Psθ sβ τs = δ max θ cp − θ s , γ (1 − θ s )
(14)
where, Ps, β and γ are constants; θcp is the close-packed particle volume fraction; and δ is the Kronecker delta (Snider et al., 1998). The interphase momentum exchange rate, F, in equation (2) is calculated as: ∇p F = ∫∫ fms Ds ( u g − u s ) − dm du ρ s s s
(15)
The discrete particles and the continuous gas phase are couple via the inter-phase source terms, such as δ m s , F, S h, p and δ m p ,i ,chem .
2.4 Gasification kinetics The biomass steam gasification in DFBs is decomposed into three processes: evaporation, devolatilization, and the homogeneous and heterogeneous reactions. The moisture evaporation rate is calculated as: dms = N i As M w dt
(16)
where, As is the particle surface area; Mw is the liquid (H2O in this work) molar weight; Ni denotes the molar flux of vapor calculated by: N i = kc ( Ci ,s − Ci ,∞ )
(17)
where, kc is the mass transfer coefficient; Ci,s denotes vapor concentration at the particle surface; Ci,∞ is the vapor concentration in the bulk gas. kc, Ci,s and Ci,∞ are calculated with the equations: Sh AB =
Ci ,s =
kc d p Di ,m
1/3 = 2.0 + 0.6 Re1/2 p Sc
psat (Tp ) RTp
Ci ,∞ = X i
p RT∞
(18)
(19)
(20)
where, ShAB is Sherwood number; Di,m denotes the diffusion coefficient of vapor in the bulk;
Sc is the Schmidt number; psat is the saturation pressure at a specific temperature; Xi is the local bulk mole fraction of species i; p is the local absolute pressure; and T∞ is the local bulk temperature of the gas. The heat transferred due to moisture evaporation is calculated as:
hevap =
dms h fg dt
(21)
where, hfg is the latent heat of the liquid (Ranz and Marshall, 1952; Ranz and Marshall, 1952). The moisture evaporation process is followed by volatile pyrolysis, which is modeled using a single rate correlation in this work (Ku et al., 2014).
1.2 E 08 dmdevol = −5 E 06 exp − mdevol dt RTp
(22)
where, mdevol is the residual volatile mass; R is the ideal gas constant, 8314.5 J/(kmol·K). Char is assumed to be carbon, and the volatile composition is calculated based on the element balance. After pyrolysis, char gasification occurs. Since the DFB gasifier includes a combustor and a gasifier, reaction kinetics for the gasification and combustion of volatiles and char should be considered. As listed in Tables 1 and 2, three heterogeneous reactions and six homogeneous reactions are allowed for in this work. The shrinking core model is used to reflect the effect of char core shrinkage on the heterogeneous reaction rate. The kinetic-diffusion mechanism is used for the heterogeneous reaction rates calculation, with the overall reaction rate given (Abani and Ghoniem, 2013; Ku et al., 2015) by: rc,i = pi , g
rdiff ,i rkin ,i
(23)
rdiff ,i + rkin,i
− Ei βi rkin ,i = AT i p exp RTp rdiff ,i
0.5 (Tg + Tp ) = Ci dp
(24) 0.75
(25)
where, rc,i is the net reaction rate for species i; rdiff,i and rkin,i denote the diffusion rate and the kinetic rate, respectively; Tp and Tg are the particle and gas temperatures, respectively; Ai, β
and Ei are the typical Arrhenius parameters; Ci is the mass diffusion constant and is set as 5.0E-12 s/K0.75. For the heterogeneous char oxidation, a different correlation is used to calculate the kinetic rate (Brown et al., 1988):
(
rkin ,O 2 = Tp AO 2 + BO 2 Tp
)
(26)
Note that the shrinking core model is used only for the heterogeneous reaction rate calculations. For the hydrodynamic calculation, the particle diameter is assumed to be constant.
3 Results and Discussions 3.1 Comparison of simulation results with experimental data The simulation results from the user-defined solver in OpenFOAM are compared with experimental data to test its reliability and accuracy. First, the isothermal simulation was implemented for a single-bubble-injection (SBI) fluidized bed with monodisperse bed material particles and the predicted hydrodynamics are compared with both the experimental results and the predictions from literature based on the DEM approach. Then, simulations were carried out for biomass steam gasification in a spouted bed gasifier under different operation conditions, and the predictions are compared with the corresponding experimental data.
3.2 Simulation for single bubble injection The settings for this simulation are listed in Table 3. They are the same as those reported previously in literature (Bokkers et al., 2004; Deen et al., 2006). The background fluidization velocity is 1.7 m/s. The injection has a velocity of 20 m/s and lasts 150 ms. The computational domain is adapted with 675 hexahedra cell. The dimensions and meshes are shown in Figures1 (a1) and (a2). The software, Gambit, is used to generate mesh for all the cases in this work, and the mesh is then converted to the form that can be read by OpenFOAM through the “fluentMeshToFoam” command. The experimental pattern of the
single bubble at 150 ms and 200 ms after injection (Bokkers et al., 2004), the simulation results from the DEM model in the literature (Deen et al., 2006) and the simulation results from this work at the same times are compared in Figures 1 (b) and (c). From Figures 1 (b) and (c), it can be seen that the size and shape of the bubble, and the interface between the dense bed and the freeboard predicted in this work are similar to those predicted in the literature with the DEM model (Deen et al., 2006) and close to the experimental results (Bokkers et al., 2004), indicating that the user defined solver is reliable in the prediction of particle dynamics using the MP-PIC method.
3.3 Simulation of biomass steam gasification in a spouted bed After the comparison for the isothermal particulate flow dynamics, the simulation predictions for the biomass steam gasification in a spouted bed are compared with the experimental data reported in literature (Ku et al., 2015; Song et al., 2012). The spouted bed is a rectangular bed, with a cross-section of 40×230 mm2 and a height of 1500 mm. A 60 o conical distributor with a injector of 20 mm i.d. is mounted at the bottom of the spouted bed. The biomass feed rate is 3.0 kg/h, and the steam-to-biomass ratios, Rsb, in the simulations are 0.8 and 1.2, respectively. Argon is the pneumatic conveying medium for biomass particles and is injected at a flow rate of 1.4 m3/h. The Sauter mean diameter and density of the biomass particle are 1.5 mm and 470 kg/m3, respectively. The Sauter mean diameter and the density of the bed material are 350 µm and 2500 kg/m3, respectively. The reactor temperatures, T, in this work are set as 820 oC and 870 oC according to the experiments. The bed material specific heat is 1500 J/(kg·K), and that of biomass is 860 J/(kg·K). A total number of 200,000 parcels are considered for the bed material, with each parcel representing 5,000 real particles. The as-received basis mass fractions of moisture, ash, volatile and fixed carbon of the biomass used in the experiments and the simulation (Ku et al., 2015; Song et al., 2012) are 11.89%, 1.56%, 71.78% and 14.77%, respectively. The dry-ash-free basis mass fractions of C, H, O, and N & S of the biomass are 46.29%, 6.48%, 46.08% and 1.15%, respectively. Three simulations were carried out for biomass steam gasification in the spouted bed
under three different operating conditions. The dimensions and the meshes of the spouted bed, and the comparisons with the corresponding experimental data are depicted in Figures 2 (a)-(e). From Figures 2 (c)-(e), it is found that the simulation results agree well with the experimental data, indicating that the user-defined solver in OpenFOAM is capable of giving reliable predictions for biomass steam gasification for the cases tested. The deviations between the predictions and the experimental results are mainly caused by the chemical reaction rate calculations. Although many different kinetic triplets, including the reaction mechanism, the frequency factor and the Arrhenius activation energy, have been evaluated and those gave the best predictions are chosen in this work, the direct use of the kinetic parameters given in the literature may still cause some errors. Actually, for one specific prediction, the kinetic parameters can be calibrated so as to give very good comparison results.
3.4 Simulation of biomass steam gasification in a dual bed After the comparisons, the user-defined solver based on OpenFOAM is finally used to predict the biomass steam gasification in the DFB gasifier at UBC. The dual bed includes a bubbling fluidized bed (BFB) serving as the gasifier, a circulating fluidized bed (CFB) serving as the combustor, a U-bend connecting the BFB and the CFB, a cyclone and a downcomer. The BFB consists of a dense bed section with an i.d. of 0.279 m and a height of 1.37 m, a freeboard section with an i.d. of 0.635 m and a height 0.762 m, and a 30o conical transition section with a height of 0.305 m. Three N2 injectors are deployed in the U-bend to keep local fluidization of the particles, and transfer them to the CFB side. The CFB height is 6.0 m with an i.d. of 0.102 m. Biomass gasification takes place in the BFB, while the bed material is heated in the CFB, with the high-temperature flue gas mainly generated from methane combustion. The dimensions and meshes adapted for the DFB are shown in Figures 3 (a) and (b). In order to improve the mesh quality, the stair step mesh method is used for this dual bed gasifier. The interval size of the grid is set as 2 cm and there are totally 29983
hexahedrons. The biomass composition in this simulation is the same as that used in the spouted bed aforementioned. The properties of the inert bed material are the same as those used in the monodisperse fluidized bed above, except that the Sauter mean diameter is 200 µm. The total mass of the bed material is 105.8 kg. The total number of parcels for the bed
material is 400,000, with each parcel representing 25,000 real particles. The boundary conditions for the DFB gasifier are listed in Table 4. The particle distribution (PD) and the voidage at different times are shown in Figure 4. The species mass fractions of H2, CO, CO2, CH4 and H2O at different times are shown in Figures 5 (a)-(e), respectively. The evolution of species volume fraction of the dry and nitrogen-free syngas are shown in Figure 6. Figure 4 shows the histories of the particle distribution (the left one) and the corresponding voidage (the right one) distribution from the initial state. The voidage is the volume fraction of the gaseous phase and is represented by the symbol, alpha. The legend in Figure 4 indicates the value of this voidage (alpha) at different times. The blue particles are the silica sand particles while the red ones are the biomass particles. From Figure 4, it can be seen that the bed material circulation loop can be formed automatically from the initial state due to pressure balance between the BFB and the CFB. This is because the DFB works like a communicating vessel. When it runs stably, a stable pressure loop can be built. This pressure loop ensures that the bed materials are circulating between the BFB and CFB with specific flux. At the initial state of the simulation, because there are no particles in the CFB side, the pressure in the BFB side is much higher than that in the CFB side, and the particles are flowing from the BFB to the CFB due to pressure driving. When the particles reached the CFB chamber, the pressure in the CFB side begin to increase. The flue gas from of the bottom of the CFB then pushes them to rise along the CFB and finally go to the BFB side through the cyclone and downcomer. Then again, the bed material particles are flowing from the BFB side to the CFB side due to pressure driving. This is actually a dynamic balance of pressures in the DFB system. The bed height is about 1 m according to the dimension shown in Figure 3. The volume fraction of the bed materials is the highest in the dense bed zone of the BFB. The voidage gradually increases along the z-coordinate in the bed zone, indicating a gradual increase in the bubble size. Very few biomass particles appear in the freeboard and
no silica sand particles reach there, because the sand is much denser than the biomass particle. Between the dense bed zone and the freeboard zone, there is a splash zone, which is a typical fluid structure of the bubbling fluidized bed. It can also be seen that the particles tend to accumulate at the bottom of the cyclone and in the middle elbow zone of the downcomer, which can work like a loop seal to prevent the gas in the BFB from ascending along the downcomer. From Figures 5 (a)-(d), it can be seen that as the gasification process goes on, BFB is gradually filled with the syngas species. The mass fractions of H2, CO and CH4 in the conical section are relatively high. This is because the biomass particles tend to concentrate there due to their low density. From Figure 5 (c), it can be seen that the carbon dioxide concentration in the CFB is high. This is because methane is burnt before entering the CFB, and some of the biomass char particles also react with the residual oxygen there. This can be seen from the boundary conditions listed in Table 4. From Figure 5 (e), it can be seen that the steam concentration near the distributor is the highest and then decreases obviously a little farther from the distributor. This is because the steam is diluted by both the gasification products and the nitrogen from the feeding point. It also shows in Figures 5 (a)-(e) that the U-bend and cyclone can well prevent the syngas in the BFB and the flue gas in the CFB from mixing because little amount of syngas can be detected in the CFB side and no CO2 in the CFB goes to the BFB side, which is essential for the safe operation of the DFB and for the high quality yield of the syngas production. From Figure 6, it can be seen that the volume fractions of H2, CO, CO2 and CH4 in the dry and nitrogen-free syngas are around 55%, 20%, 20% and 5%. Since in the initial condition, the temperatures of the bed material and the BFB wall are set as 850oC, just like the stable operating temperature, the composition of the syngas does not change much as time advances.
4 Conclusions A user-defined solver based on OpenFOAM is built, and its reliability for biomass steam gasification in fluidized bed is proved by comparison with experiments. The dense bed
zone, splash zone and freeboard in the BFB can be detected, as well as the particle circulation loop. The dense bed of BFB is about 1 m and the voidage gradually increases along the height. The syngas and flue gas can be well separated. The dry and nitrogen-free syngas at the BFB outlet is composed of 55% H2, 20% CO, 20% CO2 and 5% CH4 under the given operating condition.
Acknowledgments The authors gratefully acknowledge financial support from the National Natural Science Foundation of China (NSFC, 51576014) and the China Postdoctoral Science Foundation (2015M570096).
References [1] Abani, N., Ghoniem, A.F., 2013. Large eddy simulations of coal gasification in an entrained flow gasifier. Fuel 104, 664-680. [2] Abba, I.A., Grace, J.R., Bi, H.T., Thompson, M.L., 2003. Spanning the flow regimes: generic fluidized-bed reactor model. AIChE J. 49, 1838-1848. [3] Andrews, M.J., O'Rourke, P.J., 1996. The Multiphase Particle-in-Cell (MP-PIC) Method for Dense Particle Flows. Int. J. Multiphase Flow 22, 379-402. [4] Barisano, D., Canneto, G., Alvino, F.N.E., Pinto, G., Villone, A., Carnevale, M., Valerio, V., Battafarano, A., Braccio, G., 2016. Steam/oxygen biomass gasification at pilot scale in an internally circulating bubbling fluidized bed reactor. Fuel Process. Technol. 141, 74-81. [5] Bhutani, G., Brito-Parada, P.R., Cilliers, J.J., 2016. Polydispersed flow modelling using population balances in an adaptive mesh finite element framework. Comput. Chem. Eng. 87, 208-225. [6] Bokkers, G.A., Annaland, M.S., Kuipers, J.A.M., 2004. Mixing and segregation in a bidisperse gas–solid fluidised bed: a numerical and experimental study. Powder Technol. 140, 176-186. [7] Brown, B.W., Smoot, L.D., Smith, P.J, Hedman, P.O., 1988. Measurement and prediction of entrained-flow gasification processes. AIChE J. 34, 435-446. [8] Chen, J.K., Wang, Y.S., Li, X.F., He, R.Y., Han, S., Chen, Y.L., 2015. Reprint of “Erosion prediction
of liquid-particle two-phase flow in pipeline elbows via CFD–DEM coupling method”. Powder Technol. 282, 25-31. [9] Cheng, Y.P., Thow, Z.H, Wang, C.H., 2016. Biomass gasification with CO2 in a fluidized bed. Powder Technol. 296, 87-101. [10] Deen, N.G., Annaland, M.V.S., Kuipers, J.A.M., 2006. Detailed computational and experimental fluid dynamics of fluidized beds. Appl. Math Model 30, 1459-1471. [11] Epstein, N., Grace, J.R., 2011. Spouted and spout-fluidized beds. New York: Cambridge University Press. [12] Fan, H.J., Mei, D.F., Tian, F.G., Cui, X., Zhang, M.C., 2016. DEM simulation of different particle ejection mechanisms in a fluidized bed with and without cohesive interparticle forces. Powder Technol. 288, 228-240. [13] Feng, Y.Q., Swenser-Smith, T.S., Witt, P.J., Doblin, C., Lim, S., Schwarz, M.P., 2012. CFD modeling of gas–solid flow in an internally circulating fluidized bed. Powder Technol. 219, 78-85. [14] Fluent, Inc., 2006. FLUENT6.3-User’s Guide; Fluent Inc., New York. [15] Garzo, V., Dufty, J.W., Hrenya, C.M., 2007. Enskog theory for polydisperse granular mixtures. I. Navier-Stokes order transport. Physical Review E 76, 1-27. [16] Gerber, S., Oevermann, M., 2014. A two dimensional Euler–Lagrangian model of wood gasification in a charcoal bed – Part I: model description and base scenario. Fuel 115, 385-400. [17] Gómez-Barea, A., Leckner, B., 2010. Modeling of biomass gasification in fluidized bed. Prog. Energy Combust Sci. 36, 444-509. [18] Göransson, K., Söderlind, U., Engstrand, P., Zhang, W., 2015. An experimental study on catalytic bed materials in a biomass dual fluidised bed gasifier. Renewable Energy 81, 251-261. [19] Grace, J.R., 1982. Fluidized-bed hydrodynamics. In Handbook of Multiphase Systems; Hetsoni G, Ed.; Washington, D.C.: Hemisphere. [20] Issa, R.I., 1986. Solution of Implicitly Discretized Fluid Flow Equations by Operator Splitting. J. Comput. Phys. 62, 40-65. [21] Ku, X.K., Li, T., Løvås, T., 2015. CFD-DEM simulation of biomass gasification with steam in a fluidized bed reactor. Chem. Eng. Sci. 122, 270-283. [22] Ku, X.K., Li, T., Løvås, T., 2014. Eulerian-Lagrangian simulation of biomass gasification behavior in a high-temperature entrained-flow reactor. Energy Fuels 28, 5184-5196.
[23] Kunii, D., Levenspiel, O., 1991. Fluidization Engineering, 2nd ed., Boston: Butterworth-Heinemann. [24] Li, T.W., Mahecha-Botero, A., Grace, J.R., 2010. Computational Fluid Dynamic Investigation of Change of Volumetric Flow in Fluidized-Bed Reactors. Ind. Eng. Chem. Res. 45, 6780-6789. [25] Liu, H., Cattolica, R.J., Seiser, R., Liao, C.H., 2015. Three-dimensional full-loop simulation of a dual fluidized-bed biomass gasifier. Appl. Energy 160, 489-501. [26] Mahecha-Botero, A., Grace, J.R., Elnashaie, S.S.E.H., Lim, C.J., 2009. Advances in modeling of fluidized-bed catalytic reactors: A comprehensive review. Chem. Eng. Commun. 196, 1375-1405. [27] OpenFOAM®, 2015. The Open Source CFD Toolbox, Ver. 3.0.0. ESI Group, Paris, France. [28] O’Rourke, P.J., Snider, D.M., 2010. An improved collision damping time for MP-PIC calculations of dense particle flows with applications to polydisperse sedimenting beds and colliding particle jets. Chem. Eng. Sci. 65, 6014-6028. [29] Patankar, S.V., 1980. Numerical Heat Transfer and Fluid Flow. Taylor & Francis. [30] Qin, K., Jensen, P.A., Lin, W., Jensen, A.D., 2012. Biomass gasification behavior in an entrained flow reactor: Gas product distribution and soot formation. Energy Fuels 26, 5992-6002. [31] Ranz, W.E., Marshall, W.R.J., 1952. Evaporation from Drops, Part I. Chem. Eng. Prog. 48: 141-146. [32] Ranz, W.E., Marshall, W.R.J., 1952. Evaporation from Drops, Part II. Chem. Eng. Prog. 48, 173-180. [33] Robertson, E., Choudhury, V., Bhushan, S., Walters, D.K., 2015. Validation of OpenFOAM numerical methods and turbulence models for incompressible bluff body flows. Comput. Fluids 123, 122-145; [34] Saidi, M., Tabrizi, H.B., Grace, J.R., Lim, C.J., 2015. Hydrodynamic investigation of gas-solid flow in rectangular spout-fluid bed using CFD-DEM modeling. Powder Technol. 284, 355-364. [35] Shah, S., Myöhänen, K., Kallio, S., Ritvanen, J., Hyppänen, T., 2015. CFD modeling of gas–solids flow in a large scale circulating fluidized bed furnace. Powder Technol. 274, 239-249. [36] Shrestha, S., Ali, B.S., Jan, B.M., Lim, M.T., Sheikh, K.E., 2016. Hydrodynamic properties of a cold model of dual fluidized bed gasifier: A modeling and experimental investigation. Chem. Eng. Res. Des. 109, 791-805. [37] Sit, S.P., Grace, J.R., 1981. Effect of bubble interaction on interphase mass transfer in gas fluidized beds. Chem. Eng. Sci. 36, 327-335. [38] Snider, D.M., 2001. An Incompressible Three-Dimensional Multiphase Particle-in-Cell Model for Dense Particle Flows. J. Comput. Phys. 170, 523-549.
[39] Snider, D.M., O’Rourke, P.J., Andrews, M.J., 1998. Sediment flow in inclined vessels calculated using a multiphase particle-in-cell model for dense particle flows. Int. J. Multiphase Flow 24, 1359-1382. [40] Song T, Wu JH, Shen LH, Xiao J. Experimental investigation on hydrogen production from biomass gasification in interconnected fluidized beds. Biomass Bioenergy 2012; 36: 258-267. [41] Umeki, K., Yamamoto, K., Namioka, T., Yoshikawa, K., 2010. High temperature steam-only gasification of woody biomass. Appl. Energy 87, 791-798. [42] Wen, C.Y., Chaung, T.Z., 1979. Entrainment coal gasification modeling. Ind. Eng. Chem. Process Des. Dev. 18, 684-695. [43] Xie, J., Zhong, W.Q., Jin, B.S., Shao, Y.J., Liu, H., 2012. Simulation on gasification of forestry residues in fluidized beds by Eulerian–Lagrangian approach. Bioresour. Technol. 121, 36-46. [44] Yan, L.B., Yue, G.X., He, B.S., 2016. Thermodynamic analyses of a biomass-coal co-gasification power generation system. Bioresour. Technol. 205, 133-141.
Figure captions: Figure 1. Comparisons with results from experiments and literature at 150 ms and 200 ms Figure 2. Meshes of the spouted bed and comparisons with experimental results Figure 3. Dimensions and meshes of the dual fluidized bed gasifier Figure 4. Particle (left) and voidage (right) distributions at different times Figure 5. Species mass fractions at 3, 6, 9 and 12 s Figure 6. Evolution of the dry nitrogen-free syngas composition
c 15
m
1c m
m 5c 1.
(a1)Dimensions of the SBI fluidized bed
(a2)Meshes of the SBI fluidized bed
(b1 ) Results from experiments at
(b2) Predictions from literature at
(b3) Predictions from this work at
150 ms (Bokkers et al., 2004)
150 ms (Deen et al., 2006)
150 ms
(c1) Results from experiments at
(c2) Predictions from literature at
(c3) Predictions from this work at
200 ms (Bokkers et al., 2004)
200 ms (Deen et al., 2006)
200 ms
Figure 1 Comparisons with results from experiments and literature at 150 ms and 200 ms
(a) Dimensions of the spouted bed 0.6
exp cal
0.6
0.5
exp cal
0.5
T=1093 K, Rsb=1.2 Mass fraction
0.4
0.3
0.2
T=1143 K, Rsb=1.2
0.4 0.3 0.2 0.1
0.1
0.0
0.0
H2
CO
CO2
CH4
H2
CO
Species
CO2
CH4
Species
(c) Comparisons at T=1093 K and Rsb=1.2
(d) Comparisons at T=1143 K and Rsb=1.2
0.6
exp cal
0.5
Mass fraction
Mass fraction
(b) Meshes of the spouted bed
T=1143 K, Rsb=1.2
0.4 0.3 0.2 0.1 0.0 H2
CO
CO2
CH4
Species
(e) Comparisons at T=1093 K and Rsb=0.8 Figure 2 Meshes of the spouted bed and comparisons with experimental results
Ø1 0. 19 cm
(a) Dimensions of the DFB gasifier
(b) Meshes of the DFB gasifier
Figure 3 Dimensions and meshes of the DFB gasifier
(a) 0 s
(b) 3 s
(c) 6 s
(d) 9 s
Figure 4 Particle (left) and voidage (right) distributions at different times
(e) 12 s
(a) H2 mass fraction at 3, 6, 9 and 12 s
(b) CO mass fraction at 3, 6, 9 and 12 s
(c) CO2 mass fraction at 3, 6, 9 and 12 s
(d) CH4 mass fraction at 3, 6, 9 and 12 s
(e) H2O mass fraction at 3, 6, 9 and 12 s Figure 5 Species mass fractions at 3, 6, 9 and 12 s
0.6
Species volume fraction
0.5
H2 CO CO2
0.4
CH4 0.3
0.2
0.1
0.0 2
3
4
5
6
7
8
9
10
11
time (s) Figure 6 Evolution of the dry nitrogen-free syngas composition
12
Table captions: Table 1 Heterogeneous reactions and corresponding kinetic parameters Table 2 Homogeneous reactions and corresponding kinetic correlations Table 3 Settings for the single bubble injected into a monodisperse two-dimensional fluidized bed Table 4 Boundary conditions for the DFB gasifier
Table 1 Heterogeneous reactions and corresponding kinetic parameters (Ku et al., 2015; Brown et al., 1988) Reactions
Kinetic parameters AO2 = −1.68 × 10
−2
s/(m ⋅ K) , BO2 = 1.32 × 10−5 s/(m ⋅ K)
R1
C+0.5O2 → CO
R2
C+H 2 O → CO+H 2
AH 2O = 45.6 s/(m ⋅ K) , EH2 O = 4.37 × 107 J/kmol , β H 2O = 1.0
R3
C+CO 2 → 2CO
ACO2 = 8.3 s/(m ⋅ K) , ECO2 = 4.37 × 10 7 J/kmol , β CO2 = 1.0
R4 R5
Table 2 Homogeneous reactions and corresponding kinetic correlations Reactions Kinetic correlations (kmol/m3/s) Ref. r4 = 0.312 exp(−30000 / (1.987Tg ))CCH CH 4 +H 2 O → CO+3H 2 (Wen and Chaung, 1979) r5 = 2.5 × 108 exp( −16597.5 / Tg )CCO CH O CO+H 2 O → CO 2 +H 2 (Umeki et al, 1996) 4
2
9
r6 = 9.43 × 10 exp(−20563.51/ Tg )CCO2 CH 2
R6
CO2 +H2 → CO+H2 O
R7
CO+0.5O 2 → CO 2
r7 = 1.0 ×1010 exp(−15154.25 / Tg )CCO CO0.52 CH0.52 O
R8
H 2 +0.5O2 → H 2 O
r8 = 2.2 × 109 exp( −13109.63 / Tg )CH 2 CO2
R9
CH 4 +2O2 → CO 2 +2H 2 O
0.2 r9 = 2.119 × 1011 exp(−24379.1 / Tg )CCH CO1.32 4
(Umeki et al, 1996) (Gómez-Barea Leckner, 2010) (Gómez-Barea Leckner, 2010) (Fluent, 2006)
and and
Table 3 Settings for the single bubble injected into a monodisperse two-dimensional fluidized bed Bed dimensions Number of cells Reactor width, x = 0.15 m Reactor thickness, y = 0.015 m Reactor height, z = 0.45 m
NX = 15 NY = 1 NZ = 45
Parameters
Value
Background velocity Central jet (1 cm width) pulse velocity Central jet pulse duration Particle density Particle diameter Drag model
1.7 m/s 20 m/s 150 ms 2526 kg/m3 2.5 mm Ergun/Wen & Yu
Table 4 Boundary conditions for the DFB gasifier Parameter Value Biomass feed rate Steam-to-biomass mass ratio Steam inlet temperature BFB wall temperature Other wall conditions Bed material initial temperature CFB inlet gas temperature CFB inlet gas composition CFB inlet gas flow rate U-bend N2 flow rate BFB N2 purging velocity
10 kg/h 1.0 600 oC 850 oC Adiabatic 850 oC 1227 oC 8.37% O2, 73.99% N2, 9.70% CO2, 7.94% H2O 155.5 kg/h 15.29 m3 2.0 m/s
Highlights of the manuscript
A user-defined solver for biomass steam gasification in fluidized bed is built. The solver is tested against experimental data in literature and works well. Biomass steam gasification in the DFB at UBC is preliminary predicted. Flue gas and syngas can be well separated by the U-bend and cyclone. Dry N2-free syngas is composed of 55% H2, 20% CO, 20% CO2 and 5% CH4.