CFD studies on biomass gasification in a pilotscale dual fluidizedbed system

Parametric studies were conducted to investigate the impact of gasifier temperature, steam to biomass ratio S/B, and air supply to the combustor on the producer gas composition.. Accordi

CFD studies on biomass gasification in a pilot-scale

dual fluidized-bed system

Hui Liu, Robert J Cattolica*, Reinhard Seiser

Department of Mechanical and Aerospace Engineering, University of California, San Diego, 9500 Gilman Drive, La

Jolla, CA, 92093, USA

a r t i c l e i n f o

Article history:

Received 21 December 2015

Received in revised form

2 April 2016

Accepted 28 April 2016

Available online xxx

Keywords:

Biomass

Gasification

Fluidization

Circulation rate

CFD

Pilot-scale

a b s t r a c t

A comprehensive CFD (Computational Fluid Dynamics) model using the MP-PIC (Multi-phase Particle-In-Cell) method was developed to simulate a pilot-scale (6 tons/day, 1 MW th) dual fluidized-bed biomass gasification system In this model the particulate phase was described with the blended acceleration model The momentum, mass, and energy transport equations were integrated with the kinetics of heterogeneous biomass and char reactions and homogeneous gas-phase reactions to predict the particle circulation, pro-ducer gas composition, and reactor temperature The simulation results were compared with experimental data from the pilot-scale gasification system to validate the model at different operating conditions Parametric studies were conducted to investigate the impact of gasifier temperature, steam to biomass ratio (S/B), and air supply to the combustor on the producer gas composition The studies showed that increasing gasifier temperature and steam to biomass ratio (S/B) promoted syngas (COþ H2) production and increased hydrogen content in producer gas The effect of air supply was minor, because for the dual fluidized-bed system air was not directly involved in biomass gasification

© 2016 Hydrogen Energy Publications LLC Published by Elsevier Ltd All rights reserved

Introduction

Currently, energy and chemical industries rely primarily on

fossil fuels Hydrogen as a clean energy source with the high

energy density can become an alternative to fossil fuels[1,2];

however, the current hydrogen production also depends on

fossil fuels and most of hydrogen is produced from natural gas

reforming and coal gasification[3,4]

Biomass as a renewable energy source can be used to

produce hydrogen through biomass gasification [5,6] Two

types of technologies such as fixed-bed and fluidized-bed are

mainly used for biomass gasification Fixed-bed biomass

gasifiers are mostly preferable for small-scale syngas

production with regard to the simple process and low capital investment[7]; however, due to the insufficient gas-particle contact, biomass gasification process in fixed bed reactors is slow and the tar content in the producer gas is relatively high [8e10] Therefore, fixed-bed gasifiers are not suitable for large-scale syngas production and are only preferable for small size plants with the capacity of up to 1.5 MW th; comparatively, the capacities of atmospheric bubbling fluidized-bed gasifiers can

be up to 25 MW th [11] In addition, fluidized-bed gasifiers demonstrate good tolerance to particle sizes The gasesolid mixing is more efficient and less tar is generated in fluidized-bed gasifiers[12,13]

In a conventional single-reactor fluidized-bed gasifier, air and biomass are fed to the same reactor (gasifier) which has

* Corresponding author Tel.: þ1 858 5342984

E-mail address:rjcat@ucsd.edu(R.J Cattolica)

Available online at www.sciencedirect.com

ScienceDirect

http://dx.doi.org/10.1016/j.ijhydene.2016.04.205

0360-3199/© 2016 Hydrogen Energy Publications LLC Published by Elsevier Ltd All rights reserved

some disadvantages After the oxygen gas in air is consumed,

the remaining nitrogen gas in air mixes with producer gas and

dramatically dilutes the producer gas concentration

Conse-quently, the heating value of producer gas generated in such a

single-reactor fluidized-bed systems is low[14,15]

This issue can be eliminated by the use of a dual

fluidized-bed gasification system As shown inFig 1, a dual

bed system generally consists of two reactors: a

fluidized-bed gasifier and a fluidized-fluidized-bed combustor In the process

biomass and steam are fed to the gasifier while air is only

supplied to the combustor Char is generated from biomass

pyrolysis in the gasifier Char is entrained by the bed material

and is delivered from the gasifier to the combustor Char

re-acts with air in the combustor to release heat The combustion

heat is absorbed by the bed material and is returned to the

gasifier through the bed material circulation within the dual

fluidized-bed system Since air is only present in the

combustor and the combustor is separated from the gasifier,

the producer gas in the gasifier is free of nitrogen Accordingly,

the producer gas in dual fluidized-bed systems will have a

higher heating value than the producer gas generated in a

single-reactor fluidized-bed system

The dual fluidized-bed system as depicted inFig 1contains

a gasifier and a combustor These two reactors can be either a

bubbling fluidized-bed (BFB) or a pneumatic transport bed/

riser reactor There are mainly four configurations of dual

fluidized-bed gasifiers: a BFB gasifier and a BFB combustor, a

BFB gasifier and a riser-combustor, a riser-gasifier and a BFB

combustor, and a riser-gasifier and a riser-combustor[16] A

dual fluidized-bed system with a gasifier and a

riser-combustor was used to gasify waste and biomass in 1970s A

dual fluidized-bed system with a BFB gasifier and a

riser-combustor was developed in 1990s and a biomass

gasifica-tion plant using this technology was built and has operated in

Austria since 2002[17,18] A dual fluidized-bed system with a

riser-gasifier and a BFB combustor was adopted by Ebara Co.,

Ltd in 2003[19] Among these configurations, the combination

of BFB gasifier and riser combustor was considered to be

optimal for large-scale biomass gasification with regard to

efficient particle circulation, high fuel conversion, and low tar

generation[20]

The design and scale-up of dual fluidized-bed gasifiers are challenging and have in the past depended on empirical scaling formulas, especially for dual fluidized-bed gasifiers The interaction between reactors and cyclone separators re-quires sophisticated analysis [15,21] In recent years, CFD (computational fluid dynamics) has proved to be a powerful tool for the simulation of gas-particle system and numerous CFD models were developed to simulate fluidized-bed reactors [22e25] Currently, three main methods have been applied for the CFD modeling of fluidized-bed gasifiers: the Euler-ianeEulerian (EE) approach, the EulerianeLagrangian (EL) approach, and the hybrid EulerianeLagrangian approach Compared with the EE and EL approaches, the Multiphase

Euler-ianeLagrangian approach can provide both the required ac-curacy and efficiency The MP-PIC method was initially developed by Harlow et al.[26]for single-phase flows and then was improved significantly by O'Rourke et al.[27]for multi-phase flows In the MP-PIC method an isotropic stress term is applied in the particle acceleration equation to calculate par-ticle interactions Since this solid stress term is defined by a function of solid volume fraction, the trajectory of each par-ticle is not needed, which saves the computation time

computational-efficient method and can also be applied to simulate dense phase flows[28,29]

For the EE approach, particle sizes in a particulate phase must be set to the same value In contrast, particles in the MP-PIC method can have different diameters by the use of particle size distribution function (PSD) Additionally, the calculations

of momentum, mass, and energy transfer for the particulate phase in the MP-PIC method are implemented on individual particles or numerical particle parcels Thus, if there are sol-idegas reactions occurring on particles, each particle size can change in accordance with solid species generation or con-sumption by the reactions

Shi et al.[30]established a hydrodynamic model using the MP-PIC method to investigate the effect of particle size Two size distribution functions, the Gaussian and Lognormal size distributions, were applied to simulate the particulate phase

in a circulating fluidized-bed (CFB) riser The study showed that the PSDs had significant impact on the flow pattern in the lower region of the riser Wang et al.[31]developed another hydrodynamic MP-PIC model to study a binary PSD case and a polydisperse case The predicted flow pattern and particle velocity in both of cases showed good agreement with experimental data

In comparison with the EL approach that can be only applied to simulate small-scale fluidized-bed systems, the MP-PIC method is capable of simulating the full-loop of large-scale fluidized-bed gasifiers Wang et al.[32]built a hydrody-namic MP-PIC model to simulate the full-loop of a CFB system including a riser, a cyclone separator, and a loop-seal Their model successfully predicted the particle circulation and the pressure distribution in the full-loop of CFB system Jiang et al [33]also conducted a hydrodynamic study to simulate the full-loop of a CFB system including a riser, six cyclone separators, and six loop-seals The predicted solid circulation, pressure, and velocity profiles were validated with experimental data

As shown above, the MP-PIC models are capable of presenting Fig 1 e Schematic of dual fluidized-bed gasifier

more practical and valuable information for the design and

scale up of fluidized-bed gasifiers than the EL approach

Previous models were primarily developed for

hydrody-namic studies without considering chemical reactions and

were mainly focused on single-reactor fluidized-bed systems

A complete CFD model of biomass gasification in a dual

fluidized-bed system including reaction kinetics is rarely seen

Additionally, previous analyses of the single-reactor systems

can't be applied directly to the dual fluidized-bed systems,

considering the significant difference in the configurations of

two fluidized-bed systems It is necessary to develop a CFD

model of dual fluidized-bed gasifier to facilitate the design and

scale-up of such gasifiers

In this work, a three-dimensional CFD model using the

MP-PIC method is built to simulate a pilot-scale dual

fluidized-bed biomass gasifier with the capacity of 6 tons/day

and 1 MW th This simulated dual fluidized-bed system

in-cludes a BFB gasifier, a riser-combustor, a cyclone separator,

and a loop-seal In this model, the gas phase is described by

the Large Eddy Simulation (LES) while the particulate phase

is described by the blended particle acceleration equation

The momentum, mass, and energy transport equations are

integrated with the kinetics of gasesolid and gasegas

re-actions to simulate biomass gasification in the dual

fluidized-bed system The simulation results such as

compared with experimental data to validate the model at

different operating conditions The effects of important

operating parameters such as gasifier temperature, steam to

biomass ratio, and air supply to the combustor are also

analyzed in this work

Governing equations

Gas phase

The continuity equation and the momentum transport

equation for the gas phase are as follows:

v
agrg



agrgug



v
agrgug



agrgugug



¼ Vp þ F þ agrggþ V$tg (2) where dmp is the mass production from the gasesolid

re-actions, tgis the stress tensor of the gas phase and is

calcu-lated using the following equations[34]:

tg¼
mlamþ meddy

vug;i

vxj þvuvxg;j i



2 3

mlamþ meddy



dijvuk

vxk

(3)

S ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffi2SijSij

q

(5)

Sij¼12



vui

vxjþvuj

vxi



(6)

D ¼p3ffiffiffiffiV

(7) The species transport equations is shown as follows:

v
agrgYn



agrgugYn



¼ V$
agm

ScVYn



þ dmn; react: (8) The energy transport equation is applied to solve for the temperature of the gas phase, as shown below:

v
agrgE

agrgugE

¼ agvp

vtþ agug$Vp  V



ag



kmol

þ keddy

VTg

þ Sinterþ qdiffþ Q (9)

keddy¼Cpmeddy

Prt

(10) where Sinteris the energy exchange between the gas and par-ticulate phases, qdiff is the enthalpy diffusion, and Q is the energy source by chemical reactions

Particulate phase

The dual fluidized-bed gasification system contains two types

of particles, biomass and bed material The solid movement and solid mixing in the binary-particle system are described

by the blended particle acceleration equation The equation is shown as follows[35]:

dup

dt ¼ Dp



ug up

Vpr

p þ X þ g þup up

2tD

(11) where X is the modified acceleration due to the contact force between particles tD is the damping time due to inelastic particle collisions and is defined as follows:

1

tD¼ 16ffiffiffiffiffiffi 3p

p ass

r32

h ¼1þ ep

s2¼r1

pas∭ fmup up

2

as¼ ∭ fmr

p

rp¼ 1

up¼ 1

g0ðasÞ ¼ as;cp

as;cp as

(18)

r32¼∭ fr3dmpdupdTp

∭ fr2dmpdupdTp

(19) where s is the mass-weighted particle velocity variance, r32is the Sauter mean radius, g0(as) is the radial distribution

function, and ep is the particleeparticle restitution

coef-ficient.X is defined in the following equations:

rpas

vtp

vxiþ g1ðasÞ

"

D

ug ~up

 Dug ~up

rp

1

rp

! vp

vxi

#

(20)

max acp ap

; ε1 ap

g1ðasÞ ¼

0 if as¼ 0

D¼∭ fmDdmpdupdTp

rpas

(23)

~up¼∭ fmDupdmpdupdTp

where tpis the isotropic solid contact stress, Ps, b, and ε are the

model constants, g1(as) is the blending function, D is the

average particle drag coefficient, and~upis the drag-averaged

particle velocity

The interphase momentum transfer, F, is defined by:

F¼ ∭ f

(

mp

"

Dp



ug up

Vpr p

#

þ up

dmp dt

)

The drag coefficient, Dp, is described as follows[36]:

Dp¼6

8Cd

rgju g u pj

rpdp

(26)

Cd¼

8

>

>

<

>

>

:

24a2:65g

Re ; Re < 0:5 24a2:65g

Re



1þ 0:15Re0:687

; 0:5  Re  1000

0:44a2:65

g ; Re > 1000

(27)

Since the MP-PIC method is a Lagrangian-based method,

mass transfer and energy transfer in the model are calculated

on the basis of numerical particles Note that in the MP-PIC

method a numerical particle represents a cluster of real

par-ticles to simplify the computation

The mass and energy transport equations for the

particu-late phase are:

dmp¼ ∭ fdmp

dmp

dt ¼XN

i¼1

dmp;n

dmp;n

dt ¼agMwp;n

rpap

mp

dCp;n

CV

dTp

dt ¼m1

p

kdNu

dp

Ap



Tg Tp

(31) where Cp,nis the concentration of solid species n

Reaction kinetics

In this dual fluidized-bed system, after biomass is fed to the gasifier, moisture is released from biomass and then biomass

is decomposed into char, volatile gases and ash Some char remains in the gasifier to react with gases while the rest of char is transported to the combustor to react with O2 to release heat In this model the heterogeneous reactions such

as biomass drying, pyrolysis, and char gasification and com-bustion are included The homogeneous gas-phase reactions including the water gas shift reaction, steam reforming reac-tion, and gas oxidation reactions are also included

The heterogeneous reactions are defined as discrete par-ticle reactions In this model the calculations of heteroge-neous reactions are implemented on each numerical particle Each numerical particle can have its own size, temperature, and solid species Note that in this model all of heterogeneous and homogeneous reactions are described with global reac-tion schemes, and detailed reacreac-tion mechanisms are not used

It is well-known that detailed reaction kinetics can provide much more accurate predictions than global reaction kinetics; however, detailed reaction kinetics require intense computa-tion For example, as reported by Titova et al.[37], the detailed reaction kinetics of propane combustion required 599 reaction steps and 92 species For a gasification model involving both heterogeneous and homogeneous reactions, thousands of elementary reactions may be required Considering the un-affordable computation cost of detailed reaction schemes, global reaction schemes were adopted for the CFD modeling of biomass gasification[23,38e41]

Biomass drying

The biomass drying rate is described in the following Arrhe-nius equation[42]:

r1¼ 5:13  1010exp



10585

Tp



where mbiois the mass of biomass

Biomass pyrolysis

The one-step global-reaction scheme is used to simulate biomass pyrolysis in which biomass is decomposed into vol-atile gases and char In the experiments tar content was less than 2% of mass fraction due to the high gasifier temperature

of 850 C In this model tar is not included and is assumed to be fully converted to non-condensable gases for simplicity Biomass pyrolysis is modeled as follows:

Biomass/a1COþ a2CO2þ a3H2þ a4CH4þ a5C2H4þ a6C2H6

þ a7Char

R (2) Various methods can be applied to determine the values of stoichiometric coefficients One method is to assign the co-efficient values of pyrolysis products based on pyrolysis experimental results[43,44] The advantage of this method is that all of the coefficient values are directly from experimental measurements However, in such a pyrolysis experiment the heating rate is generally much slower than the heating rate of

pyrolysis in a gasifier where both of biomass pyrolysis and

combustion occur Since biomass pyrolysis is dramatically

influenced by the heating rate, the product composition from

slow pyrolysis may not be the same as the product

composi-tion of fast pyrolysis in a gasifier [45e47] Additionally, in

Reaction-2, on the left (reactant) side the contents of C, H, and

O in biomass were determined in the ultimate analysis while

on the right (product) side the contents of these elements were

measured in the pyrolysis experiment Since these contents

were measured in two different experiment using different

methods, the contents of C, H, and O in the reactant may not

be the same as those in the products in Reaction-2 Therefore,

the elemental balance may not be strictly followed in this

pyrolysis modeling approach

In this model another method is used to calculate the

co-efficients to achieve better elemental mass balances The

values ofa1a2,a3, anda7for the major species, CO, CO2, H2,

and Char, are calculated from the proximate and ultimate

analysis data, as proposed by other researchers[48e50] The

values ofa4,a5, anda6for the minor species such as CH4, C2H4,

C19.8166H24.524O11.8501is defined as biomass, based on the ratios

of C/H/O in the ultimate analysis of biomass sample The

co-efficient values are shown in Table 1 It is seen that the

elemental mass balances are applied The contents of C, H,

and O in the products of biomass pyrolysis agree well with the

elements measured in the ultimate analysis of biomass Note

that due to the strict elemental mass balances, changes in any

defined species of pyrolysis can cause changes in the

coeffi-cient values of other species For example, if only CH4

and C2H4 are defined as minor species, rather than three

species of CH4, C2H4, and C2H6as in the current model, the

coefficient values of other species need to change to maintain

the elemental mass balances[51]

Char combustion and gasification

In the dual fluidized-bed system, some char is transported

from the gasifier to the combustor and reacts with air in the

combustor to release combustion heat Char combustion is

defined as follows[52]:

r3¼ 8:68  106acTexp



29160

Tp



The pre-exponential factor of 8:68  106 is adjusted to fit

experimental data

The reactions between char, CO2, H2O, and H2are consid-ered to be reversible reactions and are described in the for-ward and reverse reactions:

The rates of the reactions are calculated in the following equations:

r4f¼ 1:272mcTexp



22645

Tp



r4r¼ 1:044  104mcT2exp



2363

Tp  20:92



r5f¼ 1:272mcTexp

22645

Tp



r5r¼ 1:044  104mcT2exp

6319

Tp  17:29



r6f¼ 1:368  103mcTexp



8078

Tp  7:087



r6r¼ 0:151mcT0:5exp



13578

Tp  0:372



The reaction kinetics were originally proposed by Syamlal and Bissett[53]and then were adjusted by Snider et al.[54]to suit the purpose of particle-chemistry modeling in the MP-PIC method

Homogeneous gas-phase reactions

Water gas shift reaction, steam reforming reaction, and oxidation reactions of CO, H2, CH4, C2H4, C2H6, and C3H8are included in this model The kinetics of homogeneous re-actions are shown inTable 2

Experiment setup and model settings The data used to validate the CFD model are from the opera-tion of a pilot-scale dual fluidized-bed gasifier with the ca-pacity of 6 tons/day and 1 MW th at Woodland Biomass Research Center (WBRC), located in Woodland, California,

fluidized-bed system are demonstrated inFig 2(b)

Almond prunings were used as biomass feedstock in the experiments and the properties of almond prunings are dis-played inTable 3 As seen inFig 2(c), this dual fluidized-bed system included a gasifier, a combustor, a cyclone separator, and a loop-seal In the experiments biomass was discharged constantly from a storage cart on a weight scale Biomass was transported by a bucket elevator to a couple of biomass hop-pers and then was distributed continuously to the gasifier with a screw conveyor Steam was generated in a steam generator and was superheated over 330 C in a series of heat

Table 1 e Pyrolysis product coefficients

exchangers Then, the superheated steam was injected at the

bottom of the gasifier through six nozzles Some of char

generated from biomass pyrolysis was transported from the

gasifier to the combustor The air was preheated above 290 C

in a series of heat exchangers and was supplied to the combustor at three different locations as the 1st, 2nd, and 3rd air supplies Char in the combustor burned with air to release the combustion heat Propane and an additional amount of air

Table 2 e Homogeneous reaction kinetics [55e59]

Homogeneous reactions

Water gas shift reaction

COþ H2O/CO2þ H2(R7)

r7¼ 2:75asexp



10079

T g



½CO½H2O

the pre-exponential factor, 2.75, is adjusted to fit experimental data Steam reforming reaction

CH4þ H2O/CO þ 3H2(R8)

r8¼ 720asexp



9057

T g



½CH 4 þ½H 2 O

2½H 2 þ½CO



the pre-exponential factor, 720, is adjusted to fit experimental data

CO oxidation



34761

T g



½CO½O20:25½H2O0:5

H2oxidation

H2þ 0:5O2/H2O (R10)

r10¼ 1:0  1014exp



5052

T g



½H2½O2

CH4oxidation

CH4þ 2O2/CO2þ 2H2O (R11)

r11¼ 5:01  1011exp



24417

T g



½CH40:7½O20:8

C2H4oxidation

C2H4þ 3O2/2CO2þ 2H2O (R12)

r12¼ 1:0  1015exp



20808

T g



½C2H4½O2

C2H6oxidation

C2H6þ 3:5O2/2CO2þ 3H2O (R13)

r13¼ 4:4  1011exp



15199

T g



½C2H60:5½O21:25

C3H8oxidation

C3H8þ 5O2/3CO2þ 4H2O (R14)

r14¼ 8:6  1011exp



15000

T g



½C3H80:1½O21:65

Fig 2 e (a): Dual fluidized-bed gasifier at WBRC (b): Dimensions of dual fluidized-bed (c): Flow chart of biomass gasification system (d): Boundary conditions (BC) of CFD model

were fed to the combustor through the startup burner to

provide the additional heat to the system The bed material

absorbed most of the combustion heat and then was

trans-ported from the combustor to the cyclone separator The bed

material was disengaged from the flue gas in the separator

and then fell down to the loop-seal After fluidized by the

steam in the loop-seal, the bed material was carried back to

the gasifier In the experiments a commercial ceramic bed

material (CARBO HSP®30/60) was used in the experiments

Producer gas was sampled by stainless sampling lines from

the gasifier Then, the gas was cleaned in an impinger filled

with biodiesel at 0 C and was analyzed with an Agilent 2000

Micro GC Tar was sampled using the tar protocol according to

EU-CEN/TS 15439 The gravimetric tar was between 8 and 17 g/

mass fraction of less than 2% and they were not included in

the model for simplicity The error of producer gas analysis

was mainly from GC calibration process, GC measurement,

and gas sampling In this study the overall experimental

un-certainties were estimated at 10% for the major contents

including H2, CO, CO2, and CH4and 15% for the minor contents

such as C2H4and C2H6, as reported by Billaud et al.[60]

A 3D CFD model was built in the CFD software, Barracuda

Virtual Reactor® The full-loop of the dual fluidized-bed

sys-tem including a gasifier, a combustor, a cyclone separator, and

a loop-seal was simulated in this model As displayed inFig 2

(d), a particle injection boundary condition (BC) with 25

in-jection points was defined as the biomass inlet A fluid

injec-tion BC with 48 injecinjec-tion points was applied to simulate 6

steam nozzles in the gasifier Note that the arrow direction of

injection BC indicated the flow direction In the model the

steam flow was set in the downward direction, which

simu-lated the effect of the cap of the steam nozzle In the

experi-ments each of steam nozzles was equipped with a cap When

steam was introduced to the gasifier through the nozzles,

steam initially flew upwards and then was diverted and flew

downwards after encountering the caps A fluid injection BC

with 36 injection points was used to simulate the nozzles of

the 1st air supply in the combustor A fluid injection BC with 3

injection points was used to simulate 3 air feeding pipes as the

2nd air supply Another injection BC with 2 injection points

was applied to simulate 2 air feeding pipes as the 3rd air

supply Two fluid injection BCs with 6 injection points each

were defined for two steam feeding pipes in the loop-seal The

propane supply to the combustor was defined by a mass flow

rate BC The outlets of the gasifier and cyclone separator were

defined by two pressure outlet BCs

The simulation was solved with the finite volume method

As shown in Equations(4)e(7), the subgrid-scale turbulence model of large eddy simulation was applied to describe tur-bulent gas flows in the dual fluidized-bed system [61] The partial donor cell differencing scheme, a weighted average of central difference and upwind scheme, was applied to calcu-late the face value of the variables in the transport equations This scheme is shown as follows[62]:

4 ¼u1a1A1r1þ u2a2A2r2

the donor cell property:

Qd¼Q1 if4 > 0

the acceptor cell property:

Qd¼

Q1 if4 > 0

the face property:

Q1¼1

2Qdð1 þ JÞ þ1

a1V1þ a2V2

(44) where a and b are model constants and are defined as 0.2 and 1.0, respectively

The no-slip boundary condition was applied at walls for the gas phase while the partial-slip boundary condition was implemented for the particulate phase The time-step size was between 1 103and 1 105s and was automatically controlled by the CFL value (Courant-Friedrichs-Lewy Scheme) and the maximum temperature change in a cell: if the CFL value is lower than 0.8, the time-step size is increased; when the CFL value is higher than 1.5, the time-step size is then decreased Additionally, whenever the temperature change in a cell at a time-step exceeds 300 K, the time-step size also decreases Orthogonal structured grids were generated in Barracuda Virtual Reactor®for this CFD model Three computational grids with 216,972, 243,423, and 348,768 cells were compared and the grid of 243,423 cells was finally selected due to its acceptable ac-curacy and affordable computational cost[51] The residuals

of the equations of volume fraction, pressure, velocity, en-ergy were set to 107, 108, 107, and 106as the simulation convergence criterion

The homogeneous gas-phase reactions were calculated with the cell-volume averaged chemistry method and the heterogeneous gasesolid reactions were modeled by the discrete particle chemistry method The simulation was per-formed with the accelerated GPU (graphics processing unit) computing in a computer workstation with a GTX TITAN black graphics card The simulation time for each run was set as

100 s and the final solutions were averaged from 80 to 100 s, which took about 4 days to be completed The detailed model settings are shown inTable 4

Table 3 e Proximate and ultimate analysis of biomass

feedstock

Proximate analysis (wt %)

Ultimate analysis (wt %)

Results and discussion

Particle flow pattern and gas distributions in the dual

fluidized-bed system

The particle circulation in the system between 0 and 60 s is

presented inFig 3 In the figure, particles are fully fluidized

and are entrained by the gases from the combustor to the cyclone separator The particles fall down to the loop-seal and are fluidized by steam Finally, particles are delivered back to the gasifier

It is also seen that the particles accumulate in the loop-seal

at 5 s The accumulation of particles continues to grow in the loop-seal and even reaches the bottom of the cyclone sepa-rator at 20 s After that, the accumulated particles are

Table 4 e Model settings and operating conditions

Fig 3 e Particle circulation and solid build-up in the dual fluidized-bed system (Case 1)

gradually removed from the loop-seal between 30 and 60 s and

are transported to the gasifier The predicted accumulation of

particles in the loop-seal matches the observation during the

plant startup It is mainly caused by the high rate of the initial

solid mass flow from the combustor to the cyclone separator

and loop-seal during startup Initially, most of the particles in

the combustor are quickly carried outside and delivered to the

cyclone separator and loop-seal from 0 to 5 s due to the high

gas velocity in the combustor The rate of the initial solid mass

flow is so high that the particles entrained to the loop-seal

can't be removed quickly and accumulate from 5 to 20 s

Meanwhile, particles are continuously transported from the

gasifier to the combustor through the connection pipe; due to

the small size of the connection pipe, the solid flow rate from

the gasifier to the combustor is much smaller than the initial

solid flow rate from the combustor to the cyclone separator

As a result of the slow incoming solid flow to the combustor,

the solid mass flow rate from the combustor to the cyclone

separator is then decreased Accordingly, the particle

accu-mulation in the loop-seal stops growing and the accumulated

particles are gradually removed from the loop-seal and are

delivered to the gasifier

The sectional views of H2, CO, CO2, CH4, H2O, and O2on the

xz (vertical) and xy (horizontal) planes in Case 1 are shown in

Fig 4 H2, CO, and CH4are generated in the lower left region of

the gasifier The gases are accumulated in the bed of the

gasifier and then penetrate through the bed to reach the

freeboard region, and finally leave the gasifier at the top

During the process, none of the volatile gases such as H2, CO,

and CH4leak to the combustor CO2as a product of biomass

pyrolysis and char combustion is found in both the gasifier and combustor Meanwhile, steam is injected to the bottom region of the gasifier through nozzles Most of steam rises to the freeboard region to react with other gases after perme-ating through the right side of the bed A small amount of steam escapes from the gasifier to the combustor through the

combustor to react with char and propane, and there is no O2 escaping to the gasifier Since no air is present in the gasifier, the producer gas in the dual fluidized-bed system is free of N2 and can have a high heating value The low content of N2in the producer gas can also be beneficial for the downstream processing by saving the cost of nitrogen gas removal in the process of syngas purification

The sectional views on the xy plane at the heights of 1.6, 2.0, 2.6, and 4.0 m show that the gases are unevenly distrib-uted in the bed of the gasifier The volatile gases such as H2,

CO, CO2, and CH4are accumulated in the left region of the gasifier where biomass enters; steam mainly stays in the right region of the gasifier The non-uniform patterns of the gas distributions may be caused by the feeding locations of biomass and steam in the gasifier Biomass is fed at the left side of the gasifier while steam is injected at the bottom through the nozzles The asymmetrical shape of the gasifier makes the steam injection location closer to the right side of the gasifier The advantage of the right-side steam injection is that steam can act as a sealing gas to avoid the volatile gases escaping from the gasifier to the combustor and prevent O2 leaking from the combustor to the gasifier For a single-reactor fluidized-bed systems, part of the producer gas is burned so

Fig 4 e H2, CH4, CO2, CO, H2O, and O2distributions in the dual fluidized-bed system (Case 1) The horizontal section views show the distributions of H2, CH4, CO2, CO, and H2O (at the heights of 1.6, 2.0, 2.6, and 4.0 m), and O2(at the heights of 0.7, 2.2, 3.0, and 5.0 m)

not all of the producer gas is available to be delivered to the

downstream unit for further processing which is the case for

the dual fluidized-bed system

Simulation results

The simulation time for each test was set as 100 s and the

predicted simulation results were averaged between 80 and

100 sFig 5demonstrates the producer gas composition at the

outlet of the gasifier in 100 s for Case 1 It is seen that the

concentrations of H2, CO, CO2, CH4, C2H4, and C2H6increase

rapidly between 0 and 40 s After that, the process reaches the

steady-state and all of gas concentrations become constant

The similar trends are also observed in Cases 2 and 3

InFig 6(aec),Fig 7(aec), andFig 8(aec) , the predicted gas

compositions and reactor temperatures for Cases 1e3 are

compared with three sets of experimental data Gasifier

temperatures in the bottom region, the bed, and the upper

region were chosen for the gasifier comparison at the heights

of 0.66, 1.12, and 3.05 m Three combustor temperatures as the

bottom, lower, and upper temperatures at the heights of 0.55,

1.83, and 6.4 m were selected for the combustor comparison

The differences between the predicted and measured

concentrations of major contents such as H2, CO, CO2, and CH4

are 1.40%, 3.65%, 11.80%, and 2.70% in Case 1, 2.78%, 1.58%,

8.97%, and 1.83% in Case 2, and 10.63%, 3.86%, 16.67%, and

14.35% in Case 3 The differences in temperature prediction at

6 locations of the gasifier and combustor are 2.11%, 3.24%,

1.41%, 1.19%, 0.37%, and 2.86% in Case 1, 6.04%, 6.53%, 2.57%,

1.64%, 2.56%, and 4.87% in Case 2, and 5.35%, 1.54%,

2.23%,1.49%, 0.12%, and 3.20% in Case 3 As shown above, the

predicted gas compositions and the gasifier and combustor

temperatures are close to the measured values in the

experi-ments Large discrepancies are observed in the predictions of

C2H4, and C2H6 It is mainly because in the model only two

species, C2H4and C2H6, are considered as minor contents, and

other content such as tar (less than 2% mass fraction) is not

included for simplicity

Effect of gasifier temperature

To understand the effect of the gasifier temperature, Case 1 as

a base case and three additional cases were investigated The

propane flow rates were set as 0, 21.9, and 43.2 kg/s for the

additional cases to generate different gasifier temperatures In this study the gasifier temperature is represented by the

“gasifier bed” temperature as described in the previous section

The influence of gasifier temperature on the producer gas composition is demonstrated inFig 9(a) The concentrations

of H2and CO increase and the concentrations of CO2and CH4 decrease as the gasifier temperature increases The predicted trends are consistent with the experimental data The in-crease in temperature generally promotes the endothermic reactions such as char and carbon dioxide reaction (R4), char and steam reaction (R5), and steam reforming reaction (R8) [63,64] Therefore, when the gasifier temperature increases, the concentrations of H2and CO as products of Reactions-4, 5 and 8, increase and the concentrations of CO2and CH4as re-actants of Reactions-4 and 8, decrease

Compared with the single-reactor fluidized-bed systems, the gas composition in this dual fluidized-bed system varies in

a relatively narrow range with gasifier temperature [65,66],

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Time (s)

H2 CO CO2 CH4 C2H4 C2H6

Fig 5 e Producer gas composition in 100 s (Case 1)

Fig 6 e (a): Gas composition comparison (Case 1) (b): Gasifier temperature comparison (Case 1) (c): Combustor temperature comparison (Case 1)