Chemical Engineering Science 60 (2005) 4469–4484

www.elsevier.com/locate/ces

Three-dimensional modeling of a circulating fluidized bedgasifier for sewage sludge

I. Petersen, J. Werther∗

Technical University Hamburg-Harburg, Denickestr. 15 D-21073 Hamburg, Germany

Received 6 February 2004; received in revised form 1 December 2004; accepted 4 February 2005Available online 5 May 2005

Abstract

A three-dimensional model of a circulating fluidized bed gasifier was developed, which uses continuous radial profiles of velocities andsolids hold-up with regard to the description of fluid mechanics. A complex reaction network of sewage sludge gasification is included inthe model. In the simulation calculations the influence of the axial location and the number of feeding points was examined for gasifiersof different scales. It was found that due to the very fast decomposition of the volatiles and the high volatile content in the sewage sludge,lateral mixing of the gas around the feeding port is not complete, and plumes with high pyrolysis gas concentrations are formed. Bettermixing was found to be achieved by a larger number of fuel feeding ports which are recommended in the case of high-volatile fuels andare shown to be necessary for larger bed widths.� 2005 Elsevier Ltd. All rights reserved.

Keywords:Fluidization; Modeling; Population balance; Gasification; Circulating fluidized bed; 3D-simulation; Devolatilization

1. Introduction

If modeling is to be used for scale-up or as a predic-tion tool for the operation of a gasifier a one- or even two-dimensional (2D) model might not be sufficient becausethese do not fully account for local effects. This is well-known for fluidized bed combustion. For example,Knoebiget al. (1997, 1998)measured combustion in two circulatingfluidized beds of different scales. They stated that althoughexperiments in the smaller scale may indeed be a valuabletool for the investigation of combustion phenomena the pos-sibility of scale-up predictions from small-scale experimentsis limited. The deviations of the measured axial gas compo-sition profiles were caused by three-dimensional (3D) effectsin the large-scale combustor, e.g. the axial oxygen concen-tration profiles were strongly influenced by the incompletepenetration of secondary air into the large-scale combustor.

∗ Corresponding author. Tel.: +49 40 42878 3239;fax:+49 40 42878 2678.

E-mail address:werther@tu-harburg.de(J. Werther).

0009-2509/$ - see front matter� 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.ces.2005.02.058

During co-combustion of wood chips in a CFB combustor,Lyngfelt and Leckner (1999)detected local oxygen-depletedregions, which persisted as streamers all the way up in thecombustion chamber until the cyclone inlet. Similar effectsmay occur in fluidized bed gasification since the reactivity ofbiomass material is very high. Plumes of tar or hydrocarboncompounds formed during devolatilization in the vicinity ofthe feeding port may also survive in streamers up to the topof the riser. Such effects can only be accounted for by 3Dmodeling which is the subject of the present work.

A first step in the direction of 3D modeling of CFB gasi-fiers is the work performed byKersten (2002)andKerstenet al. (2002). They modeled a CFB wood gasifier assumingrotational symmetry. So their model is essentially 2D. Inparticular, such an approach is not able to handle the prob-lem of spatial distribution of fuel, which is introduced at oneor more discrete points on the wall of the gasifier. However,the latter authors abandon the often used core-annulus ap-proach for the description of the CFB fluid dynamics anduse instead a continuous model of radial profiles of gas andsolids velocities and solids hold-up.

4470 I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484

Fig. 1. Reaction network for devolatilization and gasification of biomass. The network has been validated and numerical values of the kinetic rateconstants were determined byPetersen and Werther (2005).

The same basic approach is followed here. Sewage sludgewas chosen as a fuel in this work for several reasons. It is animportant renewable bioenergy source, because the amountof sewage sludge production is immense. For example, inGermany about 3,000,000 t (dry mass) of sewage sludge arecurrently produced per year. The thermal treatment of thesludge is a problem in many industrialized countries sincedisposal is considered to be undesirable due to ecologi-cal reasons. The present work addresses the gasification ofbiomass in large-scale industrial-size gasifiers.

2. Theory

3D modeling of multiphase flow reactors with complex re-action networks is still a challenge and easily exceeds avail-able computing capacities and tolerable computing times.The present model therefore largely simplifies the fluid dy-namics of the circulating fluidized bed by only making thedistinction between gas and solid phases and by assumingcontinuous radial profiles of velocities and solids volumeconcentrations.

The mass balances are described by convection–diffusionequations. As in previous work (Kersten et al., 2002) thegas and solids velocity profiles were assumed to have sim-ilar parabolic shapes. Kersten et al. described these shapeswith a power law relation fromGodfroy et al. (1999). Thisapproach was adopted here with some slight modifications.The model considers a CFB riser with square cross section.It describes the unsteady state, and is isothermal. The modelaccounts for the devolatilization and the gasification reac-tions, which are schematically shown inFig. 1. The speciesdistribution which is obtained after the devolatilization step,has been determined for the case of sewage sludge. Thevalidation of the corresponding kinetic rate expressions hasbeen described elsewhere (Petersen and Werther, 2005). The

mass balance equations for the gas in Cartesian coordinatescan be written in the 3D form for the speciesi:

�((1 − cv)cg,i)

�t−

Dg,hor

�2((1 − cv)cg,i)

�x2

+Dg,hor�2((1 − cv)cg,i)

�y2

+

�((1 − cv)uxcg,i)

�x

+�((1 − cv)uycg,i)

�y

+�((1 − cv)uzcg,i)

�z

=

(1 − cv)

∑j (g.g)

�i,j rj

+cv∑

j (g.s)�i,j rj

+n′′′feed,i .

(1)

Dispersion in the axial direction is neglected because of thedominating effect of convection. At the bottom the gas hasthe inlet concentration,cg,i,IN , and it is assumed that thereis no transfer of gas across the walls,

x = 0, Xt : �cg,i�x

= 0; y = 0, Yt : �cg,i�y

= 0;

z = 0 : cg,i = cg,i,IN . (2)

For the solid phase the conservation equation reads

�s�(cvXs,i)

�t− �s

��x

(Dx

�(cv Xs,i)

�x

)

+ ��y

(Dy

�(cv Xs,i)

�y

)

+ ��z

(Dz

�(cv Xs,i)

�z

)

+ �s

�(cv vx Xs,i)

�x

+�(cv vy Xs,i)

�y

+�(cv vz Xs,i)

�z

=

cvMi

∑j (g−s)

�i,j rj

+m′′′feed+m′′′return.

(3)

The solid phase componentsXs,i are carbon(i=1), and ashand inert material(i = 2).

I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484 4471

x

y

z Feed

Return

to Cyclone

Wall layer

Velocity profile

cv

z

Hbz

cv,bz

Axial solids volumeconcentration profile

Fig. 2. Illustration of the model CFB (left), wall layer and velocity profile (center), and axial profile of the cross-sectional average solids volumeconcentration (right).

There is also no transfer of solids mass across the wallsor across the distributor:

x = 0, Xt : �Xs,i

�x= 0; y = 0, Yt : �Xs,i

�y= 0;

z = 0 : Xs,i = 0; z = Ht : �Xs,i

�z= 0. (4)

The cross-sectional average solids volume concentration,cv,is constant in the bottom zone and is denoted by,cv,bz. Atthe transition from the bottom zone to the upper dilute zonethere is a strong decrease in solids volume concentration,which is often described by the definition of a separate splashzone. In the present work no such splash zone is considered.Instead, the solids volume concentration at this transition isassumed to be only half of the bottom zone concentration.

cv(z) = cv,ez + (0.5cv,bz − cv,ez)

× exp(−�cv(z − Hbz))z>Hbz. (5)

The exponential decay follows a suggestion byKunii andLevenspiel (1991).

The distinction between bottom and upper-dilute zone isalso necessary for the calculation of the wall layer thickness.In contrast to previous authors (Kersten, 2002; Godfroy etal., 1999), who assumed the parabolic velocity profile to havezero velocity at the wall, in the present model the velocitiesnear the wall are assumed to be negative in order to takethe backmixing of gas and solids and the internal circulationinto account. Zero velocity is therefore attained at a certaindistance from the wall, which is expressed by a wall layerthickness. This wall layer thickness,�, is taken fromvan derDrift et al. (2001):

�(z) = 0.55Dt

(uges(z)Dt�g

�g

)−0.22(Ht

Dt

)0.21

×(

1 − z

Ht

)0.73

, z>Hbz. (6)

This latter equation has been determined in the upper dilutezone. In the bottom zone the wall layer thickness is reduced

from the top of the bottom zone down linearly to zero atzero height (Schlichthaerle and Werther, 2001).

�(z) = 0.55Dt

(uges(Hbz)Dt�g

�g

)−0.22(Ht

Dt

)0.21

×(

1 − Hbz

Ht

)0.73z

Hbz

, 0�z�Hbz. (7)

Fig. 2 shows a scheme of the model CFB with square crosssection and illustrates the relationship between the wall layerthickness, radial velocity profile, and the axial profile of thecross-sectional average solids volume concentration.

For a radial solids volume concentration distribution in acylindrical vessel, one can think of the following power lawdescription, having a minimum concentration,cvl , at r = 0on the central axis of the riser and increasing towards thewall:

cv(r, z) = cvd(z)

(r

Rt

)n+ cvl(z), (8)

where,cvd , and,cvl , denote the solids volume concentra-tions of dense and lean phase, respectively. To translate thisequation from cylindrical coordinates in the cylindrical ar-rangement to Cartesian coordinates in the presently consid-ered square-shaped riser anx′ andy′ coordinate system isintroduced which has its origin in the middle of the cross-section and therefore counts towards the wall. Instead of ra-diusRt of the cylindrical column, we have half of the widthof the square cross-section,Xt/2.

cv(x′, z) = cvd(z)

x′Xt

2

n

+ cvl(z). (9)

This equation is only valid on an eighth of the square cross-section, which is illustrated inFig. 3. On this eighth thesolids volume concentration profile will be calculated andthen mirrored to all the other sections to get the whole hor-izontal profile.

4472 I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484

x’

y’

z

Xt

2Xt

z

x’

2Xt0

y’

looked at eighth: 0 x’ 2Xt

and 0 y’ x’

quarter: 0 x’ 2Xt

and 0 y’2Yt

≤ ≤

≤ ≤

≤ ≤

≤≤

Fig. 3. The calculation of the lateral velocity distribution in a square cross-section.

The minimum solids volume concentration in the centerof the riser,cvl(z), may be related to the cross-sectionalaverage concentration,cv, (Hartge et al., 2002).

cvl(z) = kcvl cv(z). (10)

The constantkcvl has values in the range from 0.6 to 0.8(Hartge et al., 2002). It is chosen to bekcvl = 0.65.

Then in Eq. (9) the last missing parameter to calculateeach local solids volume concentration,cv(x

′, z), is the

average solids volume concentration at a particular height,cvd(z), which is determined by integrating the differentialmass balance for the solids volume concentration on eachlevel.

1

2

(Xt

2

)2

cv(z)

=∫ Xt

2

0

∫ x′

0

cvd(z)

x′Xt

2

n

+ cvl(z)

dy′ dx′. (11)

After integration we obtain

cvd(z) = n + 2

2(1 − kcvl)cv(z). (12)

The parameter,n, is according toHartge et al. (2002)as-sumed to be equal to 5.

The velocity profiles for the gas and solid phases, respec-tively, are calculated in the same manner. First we need anaverage velocity for each level in analogy to the averagesolids volume concentration,cv(z), then the power law re-lation for the eighth of the cross-section is formulated inwhich the missing parameter is then calculated from the

integral balance. For the mean average superficial gas ve-locity in the riser, we have the fluidizing velocity at the bot-tom as a starting point. The only change to the superficialvelocity, which occurs with increasing height above the dis-tributor, is the increase or decrease due to gas volume pro-duction or reduction by the reactions, i.e., by homogeneousgas phase reactions (indexg–g) or heterogeneous gas–solidreactions (indexg–s). For all reactions the average gas pro-duction on each level is calculated and added to the super-ficial gas velocity,

�u0

�z= RT

P

∑ncellsk=1

((1 − cvk )

∑i

∑j (g.g) �i,j,krj,k + cvk

∑i

∑j (g.s)�i,j,krj,k

)ncells

(13)

with the boundary condition

u0(z = 0) = ufl. (14)

This procedure, of course a simplification, is considered tobe necessary in order to avoid a distortion of the lateralupward velocity profile as a consequence of locally varyingreaction conditions. However, the local mass balances arenevertheless fulfilled by introducing lateral convective flows,Eq. (19).

The shape of the velocity profile is inversely proportionalto the solids volume concentration, the latter being small atthe center and increasing towards the wall. The velocity, onthe contrary, will be large at the center of the riser cross-section and will decrease in the direction of the wall. At theboundary of the wall layer it will be zero, and inside thewall layer the velocity will be negative. Therefore insteadof, r/Rt , we have,r/(Rt − �), in the following relationshipto account for the negative velocity at the wall. This is amodification of the original equation fromGodfroy et al.(1999),

uz(r, z) = umax(z)

(1 −

(r

R − �(z)

)n). (15)

I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484 4473

The transformation to Cartesian coordinates is the same asfor the solids volume concentration:

uz(x′, z) = umax(z)

1 −

x′Xt

2− �(z)

n . (16)

The only unknown parameter in Eq. (16) is the maximum gasvelocity, which is found in the middle of the cross section ateach height and will be calculated from the integral balance:

1

2

(Xt

2

)2

u0(z) =∫ Xt

2

0

∫ x′

0(1 − cv(x

′, z))umax(z)

×1 −

x′Xt

2− �(z)

ndy′ dx′. (17)

The exponent was chosen to ben = 5 as in Kersten’s work(2002) and is also consistent with that already taken forthe solids volume concentration profile. After integration ofthis equation the following equation for the calculation ofumax(z) is obtained:

umax(z) = u0(z)

Xt

2Xt

2− �(z)

n (

cv(z)(n + 2)2 − n2kcvl

2(n + 1)(n + 2)− 2

n + 2

)+ 1 − cv(z)

(18)

Now the gas velocity in thez-directionuz(x, y, z) is knownat each position. To calculate the horizontal gas velocitiesux(x, y, z) anduy(x, y, z), the continuity equation is used:

�((1 − cv)ux)

�x+ �((1 − cv)uy)

�y+ �((1 − cv)uz)

�z

= RT

P

(1 − cv)

∑i

∑j (g.g)

�i,j rj

+cv∑i

∑j (g.s)

�i,j rj + n′′′feed,i

. (19)

But as there is onlyoneequation fortwo unknowns, a re-placement of variables has to be made to rectify this situa-tion. A variable,�g, is therefore defined as,

��g

�x= (1 − cv)ux , (20)

��g

�y= (1 − cv)uy , (21)

replacing theux anduy and transforming Eq. (19) into anequation of the Poisson-type, which can be solved numer-ically. With the two derivatives of,�g, Eq. (19) is now of

second order:

− �2�g

�x2 − �2�g

�y2

= +�((1 − cv)uz)

�z− RT

P

(1 − cv)

∑i

∑j (g.g)

�i,j rj

+cv∑i

∑j (g.s)

�i,j rj + n′′′feed,i

. (22)

It can be solved with the boundary conditions of no velocityacross the wall (no gas flow through the wall):

x = 0, Xt : ��g

�x= 0; y = 0, Yt : ��g

�y= 0. (23)

In analogy to the gas velocity, the local solids velocities inx-, y-, andz-direction can be calculated. The mean averagesolids velocity (in analogy tou0) depends on the solids cir-culation rateGs , which can be expressed by (Colakyan andLevenspiel, 1984):

Gs = 0.011m

s�s

(1 − vt

uges

)2

, z�Hreturn. (24)

For the terminal velocity,�t , in Eq. (24) an expression byMartin (1980)is used.

vt = 18�gdp

√√√√√1 + 1

9

√√√√gd3p(�s − �g)

�2g�g

− 1

2

. (25)

The horizontal profile of the axial solids velocity can beexpressed by the following power law expression:

vz(x′, z) = vmax(z)

1 −

x′Xt

2− �(z)

n ,

z�Hreturn. (26)

With the maximum velocity from the integral mass balanceon an eighth of the riser cross-section:

1

2

(Xt

2

)2

Gs(z) = �s

∫ Xt

2

0

∫ x′

0cv(x

′, z)vmax(z)

×1 −

x′Xt

2− �(z)

n dy′ dx′. (27)

4474 I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484

After integration,vmax(z), becomes

vmax(z) = Gs(z)

�s cv(z)

× 1

1 −

Xt

2Xt

2− �(z)

n (

(n + 2)2 − n2kcvl

2(n + 1)(n + 2)

) ,

z�Hreturn. (28)

As the solids circulation rate is zero below the solids returnleg (z<Hreturn), a different procedure was adopted for de-termining the maximum solids velocity. It is assumed thatvmax obtained at,z = Hreturn, is decreasing linearly to zerovelocity at the bottom of the CFB.

vl(z) = vmax(Hreturn)z

Hreturn, 0�z<Hreturn. (29)

Eq. (26) changes therefore to

vz(x′, z) = vl(z) − vd(z)

x′Xt

2− �(z)

n

,

0�z<Hreturn (30)

and,vd(z), is now achieved from the integral mass balancewith the consideration thatGs is zero below the solids returnpoint atz = Hreturn.

∫ Xt

2

0

∫ x′

0cv(x

′, z)

×vl(z) − vd(z)

x′Xt

2− �(z)

n dy′ dx′ = 0. (31)

After integration we obtain

vd(z) = vl(z)

Xt

2Xt

2− �(z)

n

(n + 2)2 − n2kcvl

2(n + 1)(n + 2)

. (32)

For the horizontal solid velocities the same procedure is usedas for the gas, i.evx andvy are replaced in the continuityequation by a variable,�s , with the following definition:

��s

�x= cvvx , (33)

��s

�y= cvvy . (34)

The result is again a Poisson-equation:

−�2�s

�x2 − �2�s

�y2 = �(cvvz)�z

+ m′′′feed

�scv. (35)

…

feedv

feed

feedholeA

Grid mesh

Xc K vfb �.. .

m

Fig. 4. The transfer of feed particles into the fluid bed.

The same boundary conditions are valid as for the gas phaseonly, no solids flow through the wall. The feeding is real-ized in local source terms and is explained in the followingparagraphs.

The feeding is described by a model in which the parti-cles entering the bed are transported by convection with theentrance velocity. The particles are believed to flow a certaindistance, and are then carried away by the motion of the bedparticles, which is a sink term for the feed mass balance.For each speciesi (ash, carbon, volatiles, water) of the solidphase the mass balance becomes:

�scv�Xs,i,feed

�t+ �svfeed

�(cvXs,i,feed)

�x+ m′′′

feed= 0. (36)

In Eq. (36) the feed velocity depends on the fuel mass flowrate and the cross-sectional area of the feeding port,

mfeed= �svfeedcvAfeed. (37)

The sink term in Eq. (36) can be calculated from

m′′′feed= Kfb�scvXs,i,feed, (38)

where theKfb is the transfer constant for the transfer of thefuel feed particles to the bed particles. This volume basedmass flow rate is also used in the overall mass balance forthe solids, Eq. (3). However, it must be taken into accountthat the area of the feeding port and the grid space are notthe same:

m′′′feed= Kfb�scvXs,i,feed

Afeed

�y �z. (39)

The feeding model is also illustrated inFig. 4.For the volatile and moisture components of the fuel par-

ticles a further calculation is necessary. The feed materialwhich is transferred into the bed is assumed to be in thegaseous state. However, its release into the bed occurs in thevicinity of the feeding port, taking account of the finite dry-ing and devolatilization time of the fuel particles. The molarflow of volatiles and vapor in the feed,n′′′

feed,i , in (1), is in ac-cordance with Eq. (39) the mass flow rate based on volume

I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484 4475

divided by the molar mass of volatiles or water, respectively.

n′′′feed,v = 1

MvolKfb�scvXs,vol,feed

Afeed

�y �z, (40)

n′′′feed,w = 1

MwaterKfb�scvXs,water,feed

Afeed

�y �z. (41)

For each of the gaseous components the,n′′′feed,v, has to be

multiplied by the stoichiometric coefficient from the de-volatilization equation as given inPetersen and Werther(2005) to obtain then′′′

feed,i in Eq. (1). Thisn′′′feed,i is also

the last term on the RHS of Eq. (19). The devolatilizationequation for sewage sludge was found to be described bythe same authors as

CvcHvhOvoSvsNvn

→ vsH2S+ 1

2vnN2 + �COvoCO+ (1 − �CO)

2voCO2,

+[(1 − 2�C2H4 − 6�tar)vc − �CO + 1

2vo

]CH4

+[

1

2vh − 2(1 − �C2H4 − 4.5�tar)vc + (�CO + 1)

× vo − vs

]H2 + �C2H4vcC2H4 + �tarvcC6H6. (42)

The splitting factors in Eq. (42) were for the sludge investi-gated obtained asCO = 0.3, C2H4 = 0.1, andtar = 0.005,respectively.

In analogy to the feeding, the solids return flow is mod-eled. The only difference is the solids circulation rate,Gs ,instead of the fuel feed rate. Therefore the velocity,vreturn,in the balance

�scv�Xs,i,return

�t+ �svreturn

�(cvXs,i,return)

�x+ m′′′

return= 0 (43)

is calculated from

vreturn= GsAriser

�scvAreturn. (44)

Areturnis the cross-sectional area of the return line opening inthe wall. A transition constant,Krb, is used for the transitionof the return flow particles to the bed, depending on thepenetration depth of the returning particles,

m′′′return= Krb�scvXs,i,return (45)

has to be considered as a source term in the mass,m′′′return,

balance for the bed solids, Eq. (3). For the use in Eq. (3)again the grid space has to be taken into account.

m′′′return= Krb�scvXs,i,return

Areturn

�y�z. (46)

Thus, the mass balances and closure laws are defined. Theonly missing parameters are the values of the dispersioncoefficients in the mass balance equation.

For the gas phase no axial dispersion was taken into ac-count because of the strong convectional effect due to thehigh gas velocities. In the horizontal direction constant dis-persion coefficients were assumed in the bottom and upperdilute zone, respectively. For the upper dilute zoneLuecke etal. (2004)in their validation of a 3D CFB combustor modelfound good agreement between their calculations and localgas concentration measurements in an industrial boiler for aPeclet numberPeg,hor,ud = 150,

Peg,hor = u0Dhyd

Dg,hor. (47)

In Eq. (47)Dhyd is the hydraulic diameter of the riser, whichis defined as four times the cross-sectional area divided bythe circumference of the riser, and is identical to the widthin case of a square cross-section. It should be noted thatEq. (47) yields numerical values ofDg,hor,ud which are ingood agreement with measurements of lateral gas dispersionby Sternéus et al. (2000)in an industrial CFB boiler.

Based on measurements byKoenigsdorff et al. (1995)who found that the numerical values of the coefficients oflateral solids dispersion in the upper dilute zone of a pilot-scale CFB riser were more or less equal to the coefficientsof lateral solids dispersion it is assumed here that

Ds,hor,ud = Dg,hor,ud . (48)

For the horizontal solids dispersion in the bottom zone thevalueDs,hor,bz = 0.12 m2/s was adapted from the measure-ments bySchlichthaerle and Werther (2001)in a 0.4 m di-ameter riser. For lateral gas mixing in the bottom zone thefinding bySchlichthaerle et al. (2001)was used that the dis-persion coefficients of lateral solids and gas mixing were ofthe same order of magnitude. They obtained their results ina pilot-scale (0.4 m diameter) CFB riser operated with Gel-dart B type solids under ambient conditions. In the followingcalculations it has been assumed therefore

Dg,hor,bz = Ds,hor,bz = 0.12 m2/s. (49)

As far as the axial dispersion of the solids in CFB risers isconcerned, it is difficult to find quantitative information inthe literature. It is only the work ofMostoufi and Chaouki(2000), which provides numerical values between 0.01 and1 m2/s. Contrary to the gas the axial dispersion of the solidsin the bottom zone will be significant whereas in the upperdilute zone the contribution of the axial solids dispersion tothe species distribution of the gasification processes will bemuch less. As a very first and very crude approximation, alinear variation of the dispersion coefficient,Dz, with thecross-sectional average solids volume concentration is as-sumed,

Ds,ax = D∗cv (50)

with D∗ being arbitrarily chosen as 2 m2/s. However, itshould be noted in this context, that the overall solids mixingin the vertical direction is mostly governed by the large-scale

4476 I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484

solids circulation induced by the parabolic lateral solids ve-locity profile (downward near the wall, upward in the core)and the superimposed axial dispersion is of relatively minorimportance.

It follows for the mass balance, Eq. (3),

�s�(cvXs,i)

�t− �s

Ds,hor�2(cvXs,i)

�x2

+Ds,hor�2(cvXs,i)

�y2

+D∗ ��z

(cv

�(cvXs,i)

�z

)

+ �s

�(cv�xXs,i)

�x

+�(cv�yXs,i)

�y

+�(cv�zXs,i)

�z

=

m′′′

feed+ m′′′return+cvMi∑

j (g.s)�i,j rj .

(51)

Now that all variables and parameters are set the system ofequations can be solved. The model is unsteady state. Theprogramming language is C. At the start of the program allthe global constants have to be set, namely the plant di-mensions (height and width), operating conditions (excessair ratio , fluidizing velocity u0, temperature, etc.), fuelproperties (proximate and ultimate analysis), fluid dynamicparameters (dispersion coefficients, parameters for the axialsolids concentration profile), and—because we are assum-ing an isothermal model—all the constants for the Arrhe-nius equations for the kinetic rate expressions. A differentialtime step has to be chosen as well as the grid spacing for theaxial and horizontal directions. In the axial direction a finergrid spacing is necessary in the lower section. In the initialassignment calculation block, the starting velocity and thesolids volume concentration field are calculated and the ki-netic rate constants are assigned. When the unsteady-statecalculation loop is started, the feeding and return flows arecalculated first based on the “old” solids concentrations ofthe preceding time step. Afterwards the reaction productsare computed based on the “old” time step’s gas and solidsconcentrations. By feeding and by reaction the amount ofgas changes; therefore in the fluid dynamics submodel thenew solids volume concentration and the velocity profileare calculated based on the newly determined gas flow. Allthese parameters from the submodel are used as constantsin the mass balance block where only the new gaseous andsolids concentrations are calculated. To make sure that thecalculating delivers reliable results, a mass check is donethereafter. To prevent the computer from calculating withtoo small values, a limit of 10−10 was set. Computed con-centrations being lower than this limit are set to zero. Eachcell is examined to see whether the sum of gaseous and solidspecies exceeds 100%. The calculation ends when the totalcomputation time is achieved.

3. Results and discussion

3.1. Validation of fluid dynamics

As the model used in this work to three-dimensionallysimulate the fluid dynamics in a circulating fluidized bedis completely new, the results of the computation had tobe validated with measured values before looking at thecalculated gasification results.

Measured data were taken from two cold-model facilitieslocated at TUHH. One is a circulating fluidized bed withcircular cross-section of 400 mm diameter and the other isa CFB test facility with a rectangular riser cross section of1000 mm× 300 mm. Detailed descriptions of the two facil-ities have been given elsewhere (Lackermeier and Werther,2002; Schlichthaerle and Werther, 2001). The measurementsof local solids volume concentrations and solids velocitieswere made with optical fiber probes and capacitance probes,respectively. More details about the measurements may befound in Lackermeier’s (2001)work. Some solids volumeconcentration profiles and some solids velocity profiles mea-sured at different heights are presented together with theprofiles obtained with the present model inFigs. 5and 6,respectively. Although the radialcv-profiles are generallywell described, there are some deviations with the velocityprofiles inFig. 6. The factor 0.55 inWerdermann’s (1992)correlation for the wall layer thickness,� (Eq. (6)) had to beadjusted to 0.34, because otherwise the velocity in the centerand at the wall would have been predicted too high and toolow, respectively. This is due to the shape of the horizontalprofile of the axial velocity assumed in the present model,Eq. (16).

Another check of the practical applicability of the presentfluid dynamics description was made by applying the modelto fluidization experiments in a CFB riser with 0.3×1.0 m2

cross-section which was studied bySchlichthaerle (2000).Quartz sand with a surface mean diameter of 140�m wasfluidized with air at velocities between 3 and 5 m/s. Themeasurements shown inFig. 7 were taken with capacitanceprobes at a superficial fluidizing velocity of 3 m/s with a to-tal riser pressure drop of 7500 Pa. The comparison betweenthe measurements and the calculation shows that the modelis able to describe not only the decrease of the solids volumeconcentration with height but also the change of the lateralconcentration profile. At a height of 0.67 m above the dis-tributor, there is a comparatively smooth decay of the con-centration with increasing distance from the wall whereas ata height of 1.32 m a steep profile is observed which is char-acteristic of the core-annulus structure in the upper dilutezone of circulating fluidized beds.

3.2. Simulation of a large-scale CFB gasifier

For the 3D modeling a riser height of 15 m and a width ofthe square cross-section of 0.5 m were chosen on the basis of

I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484 4477

Fig. 5. Measured and calculated solids volume concentration profiles (data fromLackermeier, 2001).

Fig. 6. Measured and calculated solids velocity profiles (data fromLackermeier, 2001).

the example given bySchaad and Sommer (2002). They de-scribe a waste-water treatment plant which produces about7200 t/a of sewage sludge (dry matter). If an operation of thefacility during 7000 h/a is assumed, approximately 1 t/h ofsludge is to be gasified. The stoichiometric air requirementand the composition of the sewage sludge are taken fromthe sludge studied byPetersen and Werther (2005), Table 1.An excess air ratio of = 0.3 is chosen, because this wasfound to be a good choice for the operation of a sewagesludge gasifier (Petersen and Werther, 2005). A gas velocityof about 6 m/s is typical for circulating fluidized bed com-bustors and the same value is assumed to hold for the presentapplication. Therefore, a fluidizing velocity at the grid levelof 4 m/s was chosen in the model because it was assumedthat the amount of gas increases by roughly 50% during gasi-fication. The width of a gasifier with square cross-section,operated at 4 m/s fluidization velocity with = 0.3 excessair ratio and 1 t/h feed, will be about 0.5 m.This width lookssmall compared to the dimensions of combustion facilitiesof the industrial scale. But it should be borne in mind thatthe advantage of gasification technology (compared to com-bustion) is that the plant can be kept much smaller in cross-section for the same fuel feed rate.

For the input parameters the same axial profile ofthe cross-sectional average solids volume concentration

(Eq. (5)) was chosen as was obtained byLackermeier (2001)for the 400 mm diameter riser leading tocv,bz=0.25, cv,ez=0.0024 and�cv = 1.3 m−1. The feeding position is locatedat 1.5 m above the distributor in the base case design. A gridmesh of 10 cells in both,x andy horizontal direction, and24 cells in axialz-direction was chosen.

A sensitivity study (Petersen, 2004) showed that a smallergrid spacing of 0.0325 and 0.025 m, respectively, had no vis-ible influence on the calculated gas concentration distribu-tions. The same is valid for the time step chosen for the cal-culation which turned out to have no effect on the calculatedresults when it was varied between 10−4 and 5× 10−3 s.Therefore a time step of 10−3 s was applied in the follow-ing simulations. With this value the calculation of 500 s realtime which was necessary to attain steady state conditionstook about 3–4 h on a PC with Intel Pentium 4 processor.

As a first test of the 3D model, a run was started with op-eration parameters used byPetersen and Werther (2005)intheir experimental work. The feeding height in this case was2.5 m above the distributor. The temperature in the experi-ments was 1023 K, and the superficial gas velocity at the topof the riserufl was 5 m/s at an excess air ratio of = 0.3.The measured axial gas concentration profiles (Petersen andWerther, 2005) and the results of the present model are com-pared inFig. 8. It should be noted here, that the CFB riser

4478 I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0 0.05 0.1 0.15 0.2

distance from the wall, cm

c v, -

h = 132 cm

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 0.05 0.1 0.15 0.2

distance from the wall, cm

c v, -

h = 67 cm

Fig. 7. Measured and calculated solids volume concentration profiles(capacitance measurements bySchlichthaerle (2000)in a CFD riser with0.3 m× 1.0 m cross section,ufl = 3 m/s,�priser = 7500 Pa).

Table 1Proximate and ultimate analysis of the dried sewage sludge

Proximate analysisWater wt% (raw) 7.3Ash wt% (raw) 42.1Combustiblesa wt% (raw) 50.6Volatiles wt% (waf) 83.4

Ultimate analysisC wt% (waf) 50.5H wt% (waf) 6.6Oa wt% (waf) 34.5N wt% (waf) 7.1S wt% (waf) 1.2

Lower heating value (LHV) MJ/kg (raw) 10.0

aBy difference.

used in the experiments had a diameter of 0.1 m only whereasthe calculations were made for a 0.5 m square cross-section.The simulated axial gas concentration profiles, as shown inFig. 8 have been calculated as flow-averages,

ci =∑

(1 − cv)uci∑(1 − cv)u

. (52)

The overall description of the measurements by the simu-lations is quite good. CO2 is a typical combustion product,

which is mostly generated in the bottom zone close to thedistributor. All other gases are strongly involved in the gasifi-cation and devolatilization reactions, which proceed mainlyin the bottom and splash zones. The feed is introduced at2.5 m above the distributor, which is responsible for the stepincrease of the H2, H2O, CO, CH4 and C2H4 concentrations.Judging from the profiles shown inFig. 8 the gasificationreactions seem to be more or less complete at a fairly shortdistance above the feed point.

Some test-runs were started to check the possibilities ofpredicting the gas quality with the present model. The pa-rameters excess air ratio, temperature, and total riser heightwere systematically varied. At first, simulations with vary-ing excess air ratio from = 0.15 to = 0.4 were con-ducted. The results for a temperature of 1023 K are shown inFig. 9. As can be seen, the lower heating value LHV de-creases with increasing excess air ratio. The concentrationsof the combustibles strongly decrease, whereas the carbondioxide content slightly increases with increasing excess airratio . This is in agreement with experimental findings inthe literature (Narvaéz et al., 1996; van der Drift et al., 2001).

Another check of the applicability of the kinetics and thereaction network is the modeling of a set of runs at differenttemperatures. Results are shown inFig. 10for temperaturesranging from 973 to 1223 K. The influence of temperaturechange is strong, due primarily to the exponential increase ofthe kinetic rate constants, which have Arrhenius-type char-acteristics. Especially the combustion reaction of the charcarbon is influenced. As can be seen fromFig. 10, the hydro-carbon content and also the concentration of hydrogen donot change significantly above 1123 K, however, the carbonmonoxide concentration strongly increases. The combustionreaction of char is competing with the combustion of hy-drogen. The increase in the reaction rate for the carbon ox-idation is higher than the one for the hydrogen combustionin the kinetics applied here and therefore this reaction willbe favored to consume available oxygen. As a consequence,at higher temperatures a higher carbon monoxide content isachieved. The decrease in hydrogen content with increasingtemperature is slowed down because of its coupling withCO via the water–gas shift reaction. The further increase ofCO may be in part the result of the Boudouard reaction.

It must be noted here that in the calculations the temper-ature in the reactor could be chosen independently of theexcess air ratio. In reality this is not true. The excess airratio determines the temperature in the riser and thereforevery high temperatures might not be reached with moder-ate excess air ratios. However, high temperatures might alsonot be desirable because of the sintering problems with thesewage sludge ash.

The results of modeling provide insights into the questionof necessary total riser height. No significant changes of thegas composition were detected above approximately 6 m (cf.Fig. 8). In another series of simulation runs the riser heightwas varied from 15 m down to 6 m. The resulting exit gascompositions were calculated at a temperature of 1023 K. As

I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484 4479

0

2

4

6

8

10

12

14

16

18

0 2 4 6 8 10 12 14 16 18

Co

nce

ntr

atio

n, v

ol-

% r

aw

O2

CO2

CO

H2

CH4

C2H4

H2O

Feed top of riser

after cyclone

H2O

CO2

CO

CH4

C2H4

H2

Height, m

Fig. 8. Simulation results of cross-sectional average gas concentrations for sludge gasification in the modeled square cross-sectional plant of 0.5m widthand measurement values from the experiment at = 0.3, T = 1023 K, hfeed= 2.5 m, andutop = 5 m/s (measurements taken fromPetersen and Werther,2005).

0

2

4

6

8

10

12

14

16

18

0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45

λ λ , -

Co

nce

ntr

atio

n o

f C

om

po

nen

t, v

ol-

%H

eati

ng

Val

ue,

MJ/

m³

T = 1023 K

C2H4

C6H6 LHV

CH4

CO

H2O

CO2

H2

Fig. 9. Variation of the excess air ratio (operating parameters:ufl = 4 m/s, hfeed= 1.5 m andT = 1023 K).

0

2

4

6

8

10

12

14

16

18

923 973 1023 1073 1123 1173 1223 1273

Temperature, K

Co

nce

ntr

atio

n o

f C

om

po

nen

t, v

ol-

%

Hea

tin

g V

alu

e, M

J/m

³

C2H4

CH4

LHV

CO

H2O

H2

CO2

Fig. 10. Variation of the temperature (operating parameters:ufl = 4 m/s, hfeed= 1.5 m, and = 0.3).

can be seen inFig. 11, there are only slight changes in theexit gas composition when the total riser height is decreasedfrom 15 to 6 m. The lower heating value is nearly unaffected,because the increase in carbon monoxide concentration iscompensated by the decrease in hydrogen content.

0

2

4

6

8

10

12

14

16

18

6 7 8 9 10 11 12 13 14 15

Total Riser Height H t , m

Co

nce

ntr

atio

n o

f C

om

po

nen

t, v

ol-

%

Hea

tin

g V

alu

e, M

J/m

³

T = 1023 K

CO

H2O

H2

CO2

C6H6

C2H4

LHV

CH4

Fig. 11. Variation of the total riser height from 6 to 15 m (operatingparameters:ufl = 4 m/s, hfeed= 1.5 m, T = 1023 K, and = 0.3).

The above discussion is concerned with cross-sectionalaverage gas concentrations. Of particular interest is the dis-tribution of the concentrations of gaseous components lo-cally around the feeding port and over the whole cross-section at the feeding level. InFig. 12, representations aregiven of the distribution of the main gaseous componentsin gasification (CH4, CO, H2) over the cross-section. Thecorresponding distribution in the vertical direction on thex, z-plane is given inFig. 13.

The penetration depth of the devolatilization zone isclearly visible inFigs. 12and13 with strong concentrationgradients in the horizontal direction. Due to the parabolicvelocity profile, the gaseous components move downwardsat the wall and upwards in the center. Mixing in the horizon-tal direction is slower than the axial mixing or dispersionalong the riser height, and the gaseous components de-volatilized at the feeding height are transported in streamersto higher regions. On their way they obviously further react,and are distributed by mixing.

4480 I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484

0 0.15 0.3 0.450

0.1

0.2

0.3

0.4

0.5

x, m

12-168-124-80-4

CH4

Feed

0 0.15 0.3 0.450

0.1

0.2

0.3

0.4

0.5

x, m

11.5-149-11.56.5-94-6.5

CO

0 0.1 0.2 0.3 0.4 0.50

0.1

0.2

0.3

0.4

0.5

x, m

y, m

20-2515-2010-155-10

H2

Fig. 12. Concentration distribution for CH4, CO, and H2 on the feeding level atz= 1.5 m (operating parameters:ufl = 4 m/s, hfeed= 1.5 m, T = 1023 K,and = 0.3).

0 0.50

2.5

5

7.5

10

12.5

15

14-189-145-90-5

CH4

Hei

ght a

bove

dis

trib

utor

, m

x, m

Feed

0 0.50

2.5

5

7.5

10

12.5

15

12-159-126-93-6

CO

Hei

ght a

bove

dis

trib

utor

, m

x, m0 0.5

0

2.5

5

7.5

10

12.5

15

21-2814-217-140-7

H2

Hei

ght a

bove

dis

trib

utor

, m

x, m

Fig. 13. Concentration distribution for CH4, CO, and H2 on thex, z-planeat y=Yt /2 (operating parameters:ufl =4 m/s, hfeed=1.5 m, T =1023 K,and = 0.3. The numbers in the legend denote vol %).

In Fig. 13 incomplete mixing is visible up to about10–12 m. Due to the slow gasification reactions in a gasifier,one might be inclined to assume complete mixing of theparticles before the gasification reactions are completed.For the heterogeneous gas–solid gasification reactions thiscould be true, but not for the very fast devolatilization. Asbiomass and especially sewage sludge contain high amountsof volatiles, lateral mixing of the gas from pyrolysis will beincomplete in larger facilities, as is demonstrated here for ariser diameter of 0.5 m already.

Critical parameters in the present model are certainly thecoefficients of horizontal dispersion since they are decisivefor the formation of lateral concentration profiles. Sensitivitystudies were therefore carried out for the coefficients of gasand solids dispersion in the bottom zone and in the upperdilute zone.Fig. 14shows CH4-concentration distributionswhich have been calculated with higher and smaller values,respectively, for the mixing coefficients in the bottom zonecompared to the calculation shown inFig. 13. It is obviousthat the main characteristics of the concentration distributionfield remain unchanged: we see a steep profile at the feedlevel which just reaches a depth of about 0.3 m from the

0 0.50

2.5

5

7.5

10

12.5

15

x, m

F

R

1.25

2.50

3.75

5.00

6.25

0 0.50

2.5

5

7.5

10

12.5

15

17.5-18.7516.25-17.515-16.2513.75-1512.5-13.7511.25-12.510-11.258.75-107.5-8.756.25-7.55-6.253.75-52.5-3.751.25-2.50-1.25

CH4

Hei

ght a

bove

dis

trib

utor

, m

x, m

F

R

1.25

2.50

3.75

5.00

Fig. 14. Sensitivity analysis: influence of the horizontal dispersion coef-ficient in the bottom zone (Dg,hor,bz = Ds,hor,bz = 0.2 m2/s (left) and

0.06 m2/s (right); operating parameters ofFig. 13; the numbers in thelegend denote vol %, F= feed,R = return of ash).

feed point in the horizontal direction and which indicates alocally high pyrolysis activity near the feed point.

Basically the same result is obtained when the lateralmixing coefficients for gas and solids in the upper dilutezone are varied (Fig. 15). A variation of the numerical valueDg,hor,ud =Ds,hor,ud = 0.02 m2/s used in the simulation ofFig. 13 by ±50% yields some change in the concentrationprofiles—with a higher mixing coefficient the equalizationof the profile is reached faster with increasing height abovethe feed level—but again the penetration depth of the feedzone remains more or less unchanged. The general conclu-sion from these calculations is that near the feed point therapid devolatilization creates lateral gas concentration pro-files which persist more or less over a long distance abovethe feed level.

The lesson to be learnt from these simulations is that inthe CFB gasifier both the feed and the pyrolysis/gasificationproducts are not automatically distributed evenly, neitherin the volume nor over the cross-section. Although in thepresent calculation there seems to be little effect of the riserheight—which will equalize concentration distributions—onthe product quality and species distribution (cf.Fig. 11) it

I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484 4481

0 0.50

2.5

5

7.5

10

12.5

15

x, m

FR

1.25

2.50

3.75

5.00

0 0.50

2.5

5

7.5

10

12.5

15

17.5-18.7516.25-17.515-16.2513.75-1512.5-13.7511.25-12.510-11.258.75-107.5-8.756.25-7.55-6.253.75-52.5-3.751.25-2.50-1.25

CH4

Hei

ght a

bove

dis

trib

utor

, m

x, m

FR

1.25

2.50

3.75

5.00

6.25

7.5

Fig. 15. Sensitivity analysis: influence of the horizontal dispersion coef-ficient in the upper dilute zone (Dg,hor,ud =Ds,hor,ud = 0.04 m2/s (left)

and 0.01 m2/s (right); operating parameters ofFig. 13; the numbers inthe legend denote vol %, F= feed,R = return of ash).

should be pointed out here that the present calculation isisothermal and therefore is not able to describe importantconsequences of the locally acting feed. These are the tem-perature effects. Devolatilization and combustion reactionsare equally fast. This means that fresh fuel particles whichare heated up near the feed point will be transported down-ward in the wall region and will be contacted with oxygennear the distributor. Locally burning volatiles will then leadto locally high heating rates for the bed which bears the riskof hot spot formation near the feed point. This hot spot for-mation which may lead to local ash melting is a major riskin the thermal processing of biomass. The 3D concentra-tion distributions calculated with the present model are be-lieved to be indicative of the risk—the flatter the profiles thesmaller the risk—and in the following several means will bestudied to reduce the steepness of the concentration profiles.

In order to examine the influence of the number of feedingports on the product gas distribution and mixing, anotherseries of simulation runs were investigated. The fuel wasadded on the two opposite sides of the riser on the feedinglevel, such that half of the total fuel amount was fed onthe left side, and the other half was fed from the right side.The solids flow from the downcomer also returned to theriser from the right side. For the CH4 content we obtain theconcentration distributions shown inFig. 16.

The concentration distribution is obviously strongly in-fluenced by the feed level relative to the solids return level.The returning solids enter the riser at 0.75 m above the dis-tributor plate. InFig. 16 the two left drawings show situa-tions where the fuel is fed either below or at the solids returnlevel. The solids return mass flow is much higher than theentering fuel feed flow. Therefore the horizontal velocitiesin this region are higher, and no plumes of volatiles developin the upper part of the riser. When feeding above the level

of the solids return nearly symmetrical profiles of CH4 arevisible. Comparing the results from the modeling with twofeeding ports to the results with one single feeding position,the effect is clear: complete mixing is achieved much fasterwith two feeding ports and there are no streamers survivinguntil the top of the riser.

One could now think of feeding the fuel into the down-comer and let the fuel particles enter the riser with the returnflow, because then mixing of the fuel is better. But it shouldbe remembered that there is a competition for the oxygenbetween the solid carbon and the gaseous components. Itis desirable that the oxygen in the plant reacts preferablywith the solid carbon for a good carbon conversion and notwith the volatiles and gasification products. In the bottomzone a carbon-combustion region should therefore develop,which more or less completely consumes the oxygen fromthe fluidizing air. In the circulating fluidized bed the return-ing solids contain the unreacted carbon. Therefore it is ad-visable to let the solids return transport the unreacted carboninto the bottom zone as close to the distributor as possible inorder to establish the carbon combustion zone. By heatingthe particles this combustion zone creates the necessary heatfor the gasification zone on top of it. The fuel should be in-troduced therefore on top of the combustion zone. As notedabove the present model is not able to describe temperatureeffects. The high CH4 concentrations which are observednear the two feed points when the sludge is fed below thesolids return level will most probably result in the forma-tion of hot spots. This is another reason why feeding nearthe distributor should be avoided.

The 3D model may also be used to simulate risers of largerdiameters. For the base case design with 0.5× 0.5 m2 crosssection inFig. 13 the CH4 is mixed over the whole crosssection on its way to the top. When the scale of the CFBis further increased, the areas of incomplete mixing will in-crease (Fig. 17). The corresponding plots for the simulationruns with two feeding ports are shown inFig. 18. Even withfeeding from both sides the mixing is incomplete up to thetop. But it can be seen that with an additional feeding portthe concentration gradients are less steep. Especially for thesimulation with a riser width of 2 m, the improvement isobvious.

Since the excess air ratio is the same in these simulationruns, the total heating value of the gas mixture at the riserexit is nearly unaffected. However, it should be borne inmind that the steep lateral profiles of the gas concentrationsalso imply similarly steep profiles of the tar concentrations.The effect of the existence of tar streamers in large scalebiomass gasifiers is not clear. However, it can be stated al-ready that for cases, where the tar is to be catalytically re-acted (e.g.Corella et al., 1988; Olivares et al., 1997) theexistence of such tar plumes will decrease the efficiency ofthe tar conversion since not enough catalyst will be avail-able in the plumes to effect complete reaction of the tar. Onthe other hand, far away from the plumes, the catalyst is notefficiently used due to the low tar concentrations.

4482 I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484

0 0.50

2.5

5

7.5

10

12.5

15

x, m

F R

F

1.25

0 - 1.25

0 0.50

2.5

5

7.5

10

12.5

15

x, m

FRF

1.25

1.25

2.50

3.75

0 0.50

2.5

5

7.5

10

12.5

15

16.25-17.515-16.2513.75-1512.5-13.7511.25-12.510-11.258.75-107.5-8.756.25-7.55-6.253.75-52.5-3.751.25-2.50-1.25R

F

CH4

Hei

ght a

bove

dis

trib

utor

, m

x, m

F

5.0

0 - 1.25

2.50 - 3.75

3.75

3.75

Fig. 16. CH4-concentration profiles for three simulation runs for feeding below (left), at the same height (middle) and above (right) the solids return(operating parameters:ufl = 4 m/s, T = 1023 K, and = 0.3. The numbers in the legend denote vol %, F= feed,R = return of ash).

0 0.5 10

2.5

5

7.5

10

12.5

15

x, m

FR

1.25

2.5

5

7.5

10

12.5

0 0.5 1 1.5 20

2.5

5

7.5

10

12.5

15

x, m

Hei

gh

t ab

ove

dis

trib

uto

r, m

17.5-18.7516.25-17.515-16.2513.75-1512.5-13.7511.25-12.510-11.258.75-107.5-8.756.25-7.55-6.253.75-52.5-3.751.25-2.50-1.25

CH4

FR

1.25

2.5

5

7.5

10

12.5

15

17.5

Fig. 17. Scale-up: influence of riser width with feeding from one side (ufl = 4 m/s, T = 1023 K, and = 0.3; solids are returned on the right-hand side;numbers in the legend denote vol %, F= feed,R = return of ash).

0 0.5 10

2.5

5

7.5

10

12.5

15

x, m

F

R

1.25

2.5

5

7.5

10

F

2.5

5

7.5

0 0.5 1 1.5 20

2.5

5

7.5

10

12.5

15

x, m

Hei

gh

t ab

ove

dis

trib

uto

r, m

17.5-18.7516.25-17.515-16.2513.75-1512.5-13.7511.25-12.510-11.258.75-107.5-8.756.25-7.55-6.253.75-52.5-3.751.25-2.50-1.25

CH4

FR

1.25

2.55

7.5

10

12.5

15

1.252.5

5

7.5

10

Fig. 18. Influence of riser width with feed divided on both sides (operating conditions ofFig. 17, F = feed,R = return of ash).

I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484 4483

4. Conclusions

In order to be able to account for local effects in scale-up simulations, a 3D model for circulating fluidized bedgasification was developed. The new model contains the fluiddynamic characteristics of the circulating fluidized bed withgas and solids upflow in the center and downflow at thewall. The horizontal profile for the velocity of both, gas andsolids, and for the solids volume concentration, respectively,is modeled as a parabolic profile. The fluid dynamics fromthe 3D model were validated with available measurementdata from cold flow CFB units, which were taken from theliterature. A reaction network for gasification is includedwith kinetic rate expressions taken from literature, whichwere verified for sewage sludge gasification in previous work(Petersen and Werther, 2005).

The proposed model was used to calculate gasification ofsewage sludge in a CFB gasifier at an excess air ratio of = 0.3. The bed temperature and fluidizing velocity were1023 K and 4 m/s, respectively. The square cross-section wasvaried between 0.5 and 2 m. The influence of the axial loca-tion and the number of feeding points was examined. Fromthe validation calculations and the simulations of the largerscale gasifier for sewage sludge the following conclusionscan be drawn:

• Due to the very fast release of the volatiles and the highvolatile content in the sewage sludge, mixing of the gasaround the feeding port is not complete, and plumeswith high amounts of pyrolysis gas concentrations areformed.

• If the sewage sludge is fed below the level of the solidsreturn into the bottom zone pyrolysis gas will be con-sumed by the fluidizing air and the risk of a hot spot nearthe fuel feed point arises.

• In general, the uneven distribution of the pyrolysis prod-ucts over the cross-sectional area of the gasifier bearsthe risk of hot spot formation near the feed inlet withsubsequent risk of ash agglomeration or even melting.

• Better mixing is achieved by increasing the number offuel feeding ports. This is advisable in general for thecase of high-volatile fuels, and is shown to be necessaryfor larger bed widths.

• For the prediction of operation characteristics of CFBgasifiers of larger scales, the 3D model suggested hereshould be a valuable tool.

Notation

A Area, m2

Afeed cross-sectional area of feeding port, m2

c Concentration, mol/m3, kg/m3

cv, cv local, average solids volume concentration, re-spectively, dimensionless

d, Dt diameter, total diameter, mdp particle diameter, m, mm,�mD dispersion coefficient, m2/sGs solids circulation rate, kg/(m2s)Hbz height of bottom zone, mHt total height, mHfeed height of feeding chute above the distributor, mHreturn height of return leg above the distributor, mkcvl parameter for the calculation of thecv-

profile(kcvl = 0.65), dimensionlessKfb transfer constant for transition of the feed to the

bed defined by Eq. (38), 1/sKrb transfer constant for the returning particles into

the bed defined by Eq. (45), 1/sLst stoichiometric air requirement, kg air/kg fuelLHV lower heating value, J/m3, J/kgm mass flow, kg/sm′′ mass flow based on an area, kg/(m2s)m′′′ mass flow based on a volume, kg/(m3s)M molar mass, kg/moln exponent for the velocity profile, dimensionlessn molar flow, mol/sn′′′ molar flow based on a volume, mol/(m3s)P system pressure, PaPe Peclet number, dimensionlessr reaction rate, mol/(m3s)r radius, mRt half of (inner) riser diameter, mR ideal gas constant, J/(mol K)S surface area, m2

t time, ST temperature, Ku velocity of gas, m/su0 cross-sectional average superficial gas velocity,

m/sufl superficial cross-sectional average fluidizing ve-

locity at top of riser, m/sv velocity of solids, m/svi mole fraction of componenti in volatiles,

mol/molV volume flow, m3/sx, y, z Cartesian coordinates, mXs,i mass fraction of solid speciesi, dimensionlessXt totalwidth inx-direction, mYt total width iny-direction, m

Greek letters

�cv coefficient for the fitting of thecv, m−1

splitting factor, dimensionless excess air ratio, dimensionless�i,j stoichiometric coefficient of gas speciesi in re-

action j, dimensionless�g cinematic viscosity of gas, m2/s� density, kg/m3

� variable (potential flow field), 1/s

4484 I. Petersen, J. Werther / Chemical Engineering Science 60 (2005) 4469–4484

Indices

bz bottom zoneez exit zoneg gasg–g homogeneous gas phase reactiong–s heterogeneous gas–solid reactionhor horizontali gas speciesj reaction numbermax maximummf minimum fluidization stater radials solid phase

References

Colakyan, M., Levenspiel, O., 1984. Elutriation from fluidized beds.Powder Technology 38, 223–232.

Corella, J., Herguido, J., Gonzalez-Saiz, J., Alday, F.J., Rodriguez-Trujillo,J.L., 1988. Fluidized bed steam gasification of biomass with dolomiteand with a commercial FCC catalyst. In: Bridgwater, A.V., Kuester, P.(Eds.), Research in Thermochemical Biomass Conversion.

Godfroy, L., Patience, G.S., Chaouki, J., 1999. Radial hydrodynamics inrisers. Industrial Engineering and Chemistry Research 38, 81–89.

Hartge, E.-U., Werther, J., Wiesendorf, V., 2002. The influence of scale onsolids flow and solids concentration at the wall of a CFB combustor.In: Grace, Zhu, de Lasa (Eds.), Circulating Fluidized Bed TechnologyVII, Canadian Society of Chemical Engineers, Ottawa, pp. 325–332.

Kersten, S.R.A., 2002. Biomass gasification in circulating fluidized beds.Dissertation Twente University, Twente University Press, Enschede,Netherlands.

Kersten, S.R.A., Venderbosch, R.H., Prins, W., van Swaaij, W.P.M., 2002.Gas–solid contacting in a riser reactor explained from radial gasvelocity and voidage distributions. Proceedings of ISCRE 17 MS #0505.

Knoebig, T., Werther, J.,Amand, L.-E., Leckner, B., 1997. Aremeasurements in small-scale units representative of the performance oflarge-scale combustors with circulating fluidized beds? VDI Berichte1314—Wirbelschichtfeuerungen: Erfahrungen und Perspektiven. VDIVerlag, Duesseldorf, pp. 281–296.

Knoebig, T., Werther, J.,Amand, L.-E., Leckner, B., 1998. Comparisonof large- and small-scale circulating fluidized bed combustors withrespect to pollutant formation and reduction for different fuels. Fuel77, 1635–1642.

Koenigsdorff, R., Fischer, M., Werther, J., 1995. High-temperature heattransport in radiatively heated circulating fluidized bed and pneumatictransport gas-particle flows. Proceedings of the Second InternationalConference on Multiphase Flow 1995 Kyoto/3–7 April 1995, vol. 3,fb2, pp. 23–31.

Kunii, D., Levenspiel, O., 1991. Fluidization Engineering. second ed.Butterworth-Heinemann, Boston.

Lackermeier, U.B., 2001. Modeling the fluid mechanics of circulatingfluidized beds on the basis of flow visualization and probemeasurements. Dissertation Technische Universitaet Hamburg,Harburg, Shaker Verlag, Aachen.

Lackermeier, U., Werther, J., 2002. Flow phenomena in the exit zone ofa circulating fluidized bed. Chemical Engineering and Processing 41,771–783.

Luecke, K., Hartge, E.-U., Werther, J., 2004. A 3D model of combustionin large-scale circulating fluidized bed boilers. International Journal ofChemical Reactor Engineering 2, A11.

Lyngfelt, A., Leckner, B., 1999. Combustion of wood-chips in circulatingfluidized bed boilers—NO and CO emissions as functions oftemperature and air-staging. Fuel 78, 1065–1072.

Martin, H., 1980. Waerme- und Stoffuebertragung in der Wirbelschicht.Chemical Engineering Technology 52, 199–209.

Mostoufi, N., Chaouki, J., 2000. On the axial movement of solids ingas–solid fluidized beds. Transactions of IChemE 78, 911–920.

Narvaéz, I., Orío, A., Aznar, M.P., Corella, J., 1996. Biomass gasificationwith air in an atmospheric bubbling fluidized bed. Effect of sixoperational variables on the quality of the produced raw gas. Industrialand Engineering Chemistry Research 35, 2110–2120.

Olivares, A., Aznar, M.P., Caballero, M.A., Gil, J., Francés, E., Corella,J., 1997. Biomass gasification: produced gas upgrading by in-bed useof dolomite. Industrial and Engineering Chemistry Research 36 (12),5220–5226.

Petersen, I., 2004. Sewage sludge gasification in the circulating fluidizedbed—experiments and modeling. Dissertation Technische UniversitaetHamburg, Harburg, Shaker Verlag, Aachen.

Petersen, I., Werther, J., 2005. Experimental investigation and modeling ofgasification of sewage sludge in the circulating fluidized bed. ChemicalEngineering and Proceedings 44 (7), 717–736.

Schaad, A., Sommer, J., 2002. Energetische Nutzung von Klaerschlammim Rahmen eines Konzeptes zur Klaerschlamm-/Holzvergasung. VDI-Wissensforum: Energetische Nutzung von Klaerschlamm, VDI-Verlag,Düsseldorf.

Schlichthaerle, P., 2000. Fluid dynamics and mixing of solids and gas inthe bottom zone of circulating fluidized beds, Disseration TechnischeUniversitaet Hamburg-Harburg, Shaker Verlag, Aachen.

Schlichthaerle, P., Hartge, E.-U., Werther, J., 2001. Interphase masstransfer and gas mixing in the bottom zone of a circulatingfluidized bed. In: Kwauk, M., Li, J., Yang, W.C. (Eds.), FluidizationX—Proceedings of the 10th Engineering Foundation Conference onFluidization. United Engineering Foundation, New York, pp. 549–556.

Schlichthaerle, P., Werther, J., 2001. Solids mixing in the bottom zone ofa circulating fluidized bed. Powder Technology 120, 21–33.

Sternéus, J., Johnsson, F., Leckner, B., 2000. Gas mixing in circulatingfluidised-bed risers. Chemical Engineering Science 55, 129–148.

van der Drift, A., van Doorn, J., Vermeulen, J.W., 2001. Ten residualbiomass fuels for circulating fluidized-bed gasification. Biomass andBioenergy 20, 45–56.

Werdermann, C.C., 1992. Feststoffbewegung und Waermeuebergang inzirkulierenden Wirbelschichten von Kohlekraftwerken. DissertationTechnische Universitaet Hamburg, Harburg, Shaker Verlag, Aachen.