Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=gcst20

Download by: [Clemson University] Date: 28 September 2015, At: 05:58

Combustion Science and Technology

ISSN: 0010-2202 (Print) 1563-521X (Online) Journal homepage: http://www.tandfonline.com/loi/gcst20

Tar Yield and Collection from the Pyrolysis of LargeBiomass Particles

R. S. Miller & J. Bellan

To cite this article: R. S. Miller & J. Bellan (1997) Tar Yield and Collection from the Pyrolysisof Large Biomass Particles, Combustion Science and Technology, 127:1-6, 97-118, DOI:10.1080/00102209708935689

To link to this article: http://dx.doi.org/10.1080/00102209708935689

Published online: 06 Apr 2007.

Submit your article to this journal

Article views: 76

View related articles

Citing articles: 6 View citing articles

Combust. Sci. and Tech., 1997. Vol. 127, pp.97-118Reprints available directly from the publisherPhotocopying permitted by license only

© 1997 OPA (Overseas Publishers Association)Amsterdam B.V. Published under license

under the Gordon and Breach SciencePublishers imprint.

Printed in India.

Tar Yield and Collection 'from thePyrolysis of Large Biomass Particles

R. S. MILLER and J. BELLAN

Jet Propulsion Laboratory, California Institute of Technology,4800 Oak Grove Drive, Pasadena, CA 91109-8099

(Received2 July 1996; In final form 17 February1997)

Tar yield collection from the pyrolysis of relatively large particles of biomass are investigatedusing the model of Miller and Bellan (1997).A variety of feedstocks are considered by varying theratios of cellulose, hemicellulose and lignin within the biomass. Effectsof secondary tar reactions,quenching, temperature, particle size and carrier gas are assessed. Secondary tar reactionsoccuring in both the particle's interior and the exterior boundary layer strongly reduce thepotential amount of tar available for collection compared to the maximum given by kineticpredictions. The primary effect of these reactions is the existence of an optimal reactor tempera­ture range for maximizing tar yields. This range is a function of both the quenching location andthe initial particle size. For rapid qvenching near the particle surface, tar collection is maximizedat high temperatures for small particles, and at low temperatures for large particles. For delayedquenching, low temperatures slow the secondary reactions and provide larger tar yields for allparticle sizes investigated. Tar yields are also dependent on the choice of the inert carrier gas;primarily due to changes in heat capacity. A sensitivity study is performed in order to assess theinfluence of the biomass apparent density, thermal conductivity, heat capacity and primary heatsof reaction.

Keywords: Biomass; modeling; porous particle; pyrolysis; tar; wood

1. INTRODUCTION

The efficient extraction of condensable tar vapors from commercial biomasspyrolysis reactors is relevant to many technologies. Diebold and Powers(1988)discussed the potential commercial applications of biomass tars for usein resins and adhesives. More recently, Chornet et al. (1994) addressed thefeasibility of fast pyrolysis processes for the large scale harvesting of tar oilsfrom crude biomass for hydrogen fuel production. Chornet et al. (1994)

suggested that it may be possible to obtain up to 75% of weight conversion to

97

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

98 R. S. MILLER AND 1. BELLAN

liquid yields and described techniques for conversion to hydrogen. Althoughthe usefulness of tar oils is well accepted, optimal reactor designs for oilextraction are not established. Fluidized bed and entrained flow reactors havereceived major attention (Scott et al., 1988) due to both their potential forscaling to commercially relevant capacities, and to widespread previoushydrodynamic studies (e.g., Lim et al., 1995). Other reactor designs underinvestigation include the rotating cone (Wagenaar et al., 1994) and thecylindrical vortex (Diebold and Powers, 1988) reactors both of which usecentrifugal acceleration of biomass particles against a heated wall for rapidheat transfer. Reactors designed for tar collection generally operate at moder­ate temperatures (- 800K) as secondary tar decomposition reactions areobserved to dominate at larger temperatures. It is also recognized that rapidquenching of the pyrolysis vapors can be used to minimize the extent ofsecondary reactions in order to improve the oil yields. However, resultsobtained for the cylindrical vortex reactor are so far limited to bench scalereactors and greater understanding of collection methodology is necessarybefore a commercially successful scaleup is attempted.

Most of the previous scientific investigations of biomass pyrolysis arelimited to the study of residual mass as related to charcoal production (see e.g.,Di Blasi, 1993b for a recent review). Liquid and/or gas yields have not onlybeen studied much less, but the studies have also been restricted primarily tovery small, kinetically controlled particle sizes. Thurner and Mann (1981)present experimental measurements from the pyrolysis of oak particles withmean size < 1mm at relatively mild temperatures < 675K. Reported tar yieldsat intermediate times show peak values of :::; 50% by mass whereas at latertimes, the yields are observed to decrease, indicating the presence of secondrytar decomposition reactions. Scott and Piskorz (1982) and Scott et al. (1988)studied maple particles with mean size -lOO~m and report maximum taryields> 80% by mass at moderate reactor temperatures of :::; 775K. Theselarge yields were the result of the small particle size and rapid quenching of thepyrolysis vapors. Liden et al. (1988) developed a kinetics model incorporatingsecondary tar reactions to explain these results. The proposed reaction schemesuggests that primary tar yields increase monotonically with increasingreaction temperatures, but that competing secondary tar decompositiondominates at large temperatures, therefore resulting in an optimal tempera­ture range for collection.

Unfortunately, the grinding of biomass to kinetically controlled sizes isrelatively expensive and commercial processes necessarily entail relativelylarge, diffusion limited particle pyrolysis (e.g., Antal, 1982). In contrast to tarresulting from particles pyrolyzing in the kinetically controlled region, the tar

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS 99

produced from large particles may exhibit prolonged tar residence times insidethe biomass particle. Thus, secondary tar decomposition in these regions maysignificantly reduce the available tar yields and cannot be overcome byimproved quenching methods. Experimental evidence does indicate that theextent of secodnary cracking can be reduced under near vacuum conditions(Roy et al., 1994); however, such low pressures may not be feasible forcommercial reactors. Effects of secondary tar reactions were observed in thenumerical simulations of a 2.5em thick wood slab at large reactor tempera­tures performed by Di Blasi (1993a) who showed that tar yields collected at theparticle surface are increased by more than a factor of two by turning off thesecondary reactions. These simulations which include secondary reactionsindicate tar yields < 20% by mass; much less than observed for kineticallycontrolled pyrolysis. The Di Blasi model does not, however, consider theexterior particle boundary layer where further reactions occur and hencecannot account for the associated effects inherent in tar collection in realreactors. Furthermore, Miller and Bellan (1996) showed that the kineticsemployed in Di Blasi (1993a) are incapable of predicting experimentallyobserved pyrolysis behavior. In contrast, the new kinetics of Miller and Bellan(1996b) combined with the porous particle model of Miller and Bellan (1996)predicted results which agree favorably with observations from a large varietyof biomass pyrolysis experiments. Their particle pyrolysis model includes fullproperty variations, thermal and mass boundary layer effects, and accountsfor feedstock variations through superimposed cellulose, hemicellulose andlignin kinetics. An important prediction from the model is the existence of anoptimal temperature regime for tar production.

The objective ofthis paper is to present results relevant to tar yield optimiza­tion from biomass pyrolysis of relatively large particles. Efforts are focused onparticle sizes ~ Icm in order to illustrate the effectsof pyrolysis on typical wastewood chips potentially used in commercial applications. Although results areobtained only for spherically symmetric and isolated particles in initiallyquiescent environments, implications for large scale reactor design are discussedwhere appropriate. The macro-particle model of Miller and Bellan (1997) isused to simulate the pyrolysis of a variety of biomass feedstocks. Of particularinterest are the effects of secondary tar reactions, quenching, particle size,reactor temperature and the choice of an inert carrier gas on the tar yield. Inaddition, an analysis is performed illustrating the sensitivity of several impor­tant biomass properties. The paper is organized as follows: Section 2 presents asummary of the pyrolysis model. Section 3 addresses the simulation results for avariety of biomass feedstocks and pyrolysis conditions as related to tar yields.Section 4 is devoted to conclusions and further discussions.

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

100 R. S. MILLER AND J. BELLAN

2. PYROLYSIS MODEL

The macro-particle pyrolysis model has been described in detail in Miller andBellan (1997) and is therefore only summarized here. The kinetics scheme forthe model is based on superimposed cellulose, hemicellulose and ligninreactions. In this manner, any biomass feedstock can be simulated through theknowledge of its initial mass composition with respect to these three primarycomponents. Each of the virgin components undergoes an initial depolyme­rization reaction:

Virgin - K 1 - Active.

This is followed by two primary competing decompositions, to tar:

and to a combination of char and gas:

Active - K 3 - X Char + (1 - X) Gas.

(1)

(2)

(3)

Secondary tar decomposition is also modeled as a single step irreversiblereaction:

Tar -K4 -Gas. (4)

Throughout the paper, 'Tar' is used to refer to the primary high molecularweight pyrolysis condensable sometimes called 'bio-oil' or 'bio-crude' andincludes any enriched oxygen and water vapor content. All reactions aremodeled with first order Arrhenius kinetics. The frequency factors and activa­tion energies for reactions K" K 2 , K 3 and the mass ratio X are all dependenton the particular biomass component, whereas all heats of reaction andsecondary tar decomposition parameters are independent of the source com­ponent. Reaction K I has L!.h 1 = 0, reaction K 2 is endothermic with to.h 2 = 255k.I/kg, and both the char formation and secondary tar reactions are exother­mic with L!.h3 = -20 k.l/kg and L!.h4 = -42 kJ/kg. All remaining parametervalues are provided in Miller and Bellan (1997).

The porous particle model incorporates all property variations, is validboth inside and outside the particle, and employs a fully transient momentumequation in contrast to the traditional use of the empirical Darcy's Law. Thederivation of the model has been addressed previously in Miller and Bellan

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS 101

(1997) and only the final general form of the equations (in sphericallysymmetric coordinates) is presented here:

where

op. .~=s.at s.r s

+ ~Jl'ffn _ 4ll,rrU3 r ,2 '

P (g )-p = eg L Y,jM, RT,

(5)

(6)

(7)

(8)

(9)

(10)

s

Pg= efJg, P,.,= (1- e)fJ,." e = 1 - LP,.JfJ,." (11)

g

1/ =e"YI/(1]reff ~ ,t"', (J2)

and

(13)

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

102 R. S. MILLER AND J. BELLAN

All properties are provided in Miller and Bellan (1997). Previous comparisonswith a large variety of experiments for different biomass feedstocks werefavorable, particularly for small particle sizes ( < lcm). Deviations from experi­mental observations for large particles were associated with differencs in theparticle geometry between the experiment and model, and with the identicalset of properties used for all biomass components and types. Thus, the onlydifference in the calculations for different feedstocks is in the mass composi­tions of cellulose, hemicellulose and lignin. The lack of experimental datacharacterizing biomass components and types is addressed here through asensitivity study.

3. RESULTS

The configuration considered is that of a single isolated and sphericallysymmetric biomass particle in an initially quiescent environment composed ofa super-heated and inert carrier gas. The outer boundary of the computationaldomain is chosen to be at RR = IORp•o, the thermal radius is RTIRp•o = 5 (RT isdefined such that the temperature is held constant at T= TR for all positionsr;;;' R T ) and 96 numerical grid points are used for all simulations. The entiredomain r';; RR is resolved in the simulations in order to account for secondarytar reactions and their effects on pyrolysis evolution. All simulations are forconstant reactor temperatures, atmospheric pressure, and a uniform initialparticle temperature Tp,o = 300K. Boundary conditions, the numericalmethod, and a discussion of effects of the above parameters are provided inMiller and Bellan (1996). The carrier gas is chosen to be steam (unlessotherwise noted) due to its attractiveness for commercial applications (seeTab. I for carrier gas properties used in this work). All simulations areterminated when the particle mass achieves 99.9% conversion (denoted as theconversion time, tJ

In biomass reactors, the pyrolysis vapors are typically quenched in order tocool the tar oils and thereby minimize the extent of secondary tar decomposi­tion, However, it is expected that some decomposition will occur within theboundary layer exterior to the particle. This process was well understood by

TABLE I Properly values for slam and nitrogen. Values are for T= 800K andp= 100kPa

Species ),[&] D[~]

H,ON,

18.01628.013

2.20 2.9 x 10-' 7.8 x 10-' 1.1 X 10- 4

0.8246 3.58 x 10-' 5.63 x 10-' 8.52 X 10- 4

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS 103

Scott et ai. (1988)who used rapid quenching to maximize tar yields. In order toaddress the effects of quenching in the current simulations, an effective"harvest radius" (RH ) is employed to define the maximum normalized "tarcollection available at location RH:

J'<41!R2 .p Y. udtTar Collection == 0 H 9 la' ,

J;;'41!r 2 p, (t = O)'dr(15)

where all variables in the numerator are evaluated at r = RH , and the denomi­nator is the initial particle mass. With this definition, the tar collectioncorresponds to the normalized mass of tar passing through the radial positionRII(R p •o ';;; RII,;;; RR) integrated over the entire duration of the simulation.Thus, the tar collection at small RII corresponds to rapid quenching with'minimal secondary tar reactions in the boundary layer, whereas large RH

values indicate minimal quenching conditions. The derived tar collection isnot an exact representation of achievable yields as tar remaining inside theregion r < RH (in particular inside the particle) at the final simulation timet = t, is not included. The value of studying the radial dependence of the tarcollection is to provide insight into the relative effects of quenching secondarytar reactions in the boundary layer upon attainable tar yields during pyrolysis.

Figure 1 presents the variation in biomass pyrolysis behavior for a varietyoffeedstocks. The conversion is defined as (1-% char formed/IOO)and the tarcollection is defined by Eq. (15). The simulated particles have initial sizeRp •o = 0.5cm, the reactor temperature is TR = 900K and the tar collection isevaluated at the particle surface (RH = Rp •o)' All mass compositions of thebiomass components are provided in Table II. Initial heating of the particlefrom room temperature requires ~ 60s before significant pyrolysis occurs.The plots are all extended to t = 250 s, past the conversion time, in order toclarify the differences among feedstocks. The conversion variable is observedto behave similarly to the surface tar collection. High lignin content biomass(e.g.,olive husk) tends to produce large char yields and therefore low relative

TABLE II Biomass compositions by mass used in this study. All extractive and ash content areincluded with the hemicellulose

Biomass Cellulose Hemicellulose Lignin Source

Beech 0.48 0.28 0.24 Maschio et al. (1994)Maple 0.40 0.38 0.22 Mok el al. (1992)Oak 0.35 0.40 0.25 SERI (1979)Olive husk 0.22 0.33 0.45 Maschio el al. (1994)Pine 0.50 0.27 0.23 Wald & Braslaw (1985)Wheat straw 0.42 0.42 0.16 Chum ez al. (1994)

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

FIGURE 1 Temporal evolutions of the (a) conversion, and (b) tar collection at the surface(RII/Rp,o= I), for various biomass feedstocks, The particle conditions are: Rp,o =0,5 cm andT.=900K,

tar collections. In contrast, the largest tar collections are produced by the highcellulose content beech and pine woods. Significant deviations in final charyields are observed for the various feedstocks ranging over approximately10% of the initial particle mass. However, the absolute variation in tar yield isnot strongly dependent on the particular biomass (approximately 3% range)and may be smaller than experimental uncertainty for tar measurements. Thisis due to the large particle size which allows considerable secondary reactionsto occur within the particle. Note that the tar yields are in all cases < 20% bymass, much less than the sub-millimeter sized particles used in the experimentsof Scott et al. (1988).The remaining discussions are primarily for maple which

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS 105

represents a typical hardwood, and/or for olive husk representing a high­lignin content feedstock.

3.1. Effects of Secondary Reactions and Quenching

The dramatic limitations imposed on the potential tar collection by finiteparticle sizes and secondary reactions are i1Iustrated in Figure 2 where thekinetic limits of char and tar production for both maple and olive husk(secondary reactions turned off) are plotted. Theoretically, almost the entireparticle mass can be converted into tar at large temperatures. However,several factors present during realistic pyrolysis strongly reduce these poten­tial yields. First, the secondary reactions which occur both inside and outsidethe particle convert tar to gas at rates which increase exponentially withtemperature. Large particles are characterized by longer tar residence times inthe particle's interior, hence more time for secondary reactions to occur.Second, endorthermic reactions and diffusion limitations result in "effectivepyrolysis temperatures" significantly below the reactor temperature for finiteparticle sizes. Therefore, real particles pyrolyze in a relatively low temperaturerange in which tar production is reduced. This latter effectis discussed in detailin Miller and Bellan (1996) and Miller and Bellan (1997).

An additional constraint on the actual tar collection in a reactor is introdu­ced by the method of pyrolysis vapor quenching. As pyrolysis proceeds,pressure builds within the particle resulting in tars and gases injected into theparticle's surroundings which is typically at temperatures large enough toinduce further secondary tar reactions. It is therefore desirable to quench these

1.0

0.8

0 0•6

§j>0 0.4

0.2

TAR

800 1000TEMPERATORE [K]

1200

FIGURE 2 Kinetic yield limits as a function of temperature for both char and tar collectionwith secondary reactions turned off. The solid line is for maple and the dotted is for olive husk.

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

106 R. S. MILLER AND J. BELLAN

reactions as rapidly as possible in order to maximize the actual collected taroils. Figure 3 depicts the normalized tar collection evolution for variousharvest locations. The results for maple are under the same conditions as thoseof Figure I. The maximal tar collection is at the particle surface, correspond­ing to a perfectly efficient quenching process. As quenching becomes lessefficient, the tar collection is strongly reduced. Under the present conditions,only negligible tar yields are observed for RH ;;, SRp •o' In addition, themaximal tar yield is less than 20% of the initial particle mass; this issignificantly less than the kinetic limits shown in Figure 2. These results clearlyindicate the importance of the quenching process for reactors which aim atmaximizing tar collection.

3.2. Effects of the Reactor Temperature and the Initial Particle Size

Several effects of the reactor temperature are addressed in Figure 4 whichdepicts both the final char yield and the final tar collection magnitudes forboth maple and olive husk. The curves are produced using results at t = t, fromeleven different simulations for each feedstock. The initial particle size isR p •o = O.Scm. As discussed above, the observed variation in char yield amongthe feedstocks is much larger than for the tar collection; the reduced differencebetween tar collection for different feedstocks is due to secondary reactions.Note also that the final char yield dependence on temperature is markedlyreduced from the kinetic limtis observed in Figure 2 due to endothermicityand associated effective pyrolysis temperature effects. In fact, a comparison

0.201.0

,,--- 1.25

~--1.5

~--1.75

.__--2.0

._--3.04.0

Z00.15

~0.10ou..:;:;0.05

0.000 50 100 150 200 250 300TIME [s]

FIGURE 3 Temporal evolution of the tar collection obtained from maple at various of R"IR .The particle conditions are: R 0 = 0.5 em and Tv = 900K. p.O

p.

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS

0.5

107

0.4cr....l~0.3

><0::<0.2::t •u

0.1

...............~............. 01' h k_ ~.Y..:_ ~.s. _-- Maple

(a)

0.0600 800 1000

TEMPERATURE [K)1200

~O.151.06

~:l 0.101.250

u0::<0.05 1.5E-

2.0

O.ORoo 800 1000 1200 1400(b) TEMPERATURE [K]

FIGURE 4 Temperature dependence of the (a)char yield, and (b) tar collection at various valuesof RHIR

p•o' for both maple and olive husk. The initial particle size is R

p•o = O.5cm.

with the char yield limits in Figure 2 suggests that the effective pyrolysistemperature is < 650K for all reactor temperatures considered for this particlesize. This behavior is in agreement with previous observations of Narayan andAntal (1996), Miller and Bellan (1996), Miller and Bellan (1997), and helpsto explain the markedly reduced tar yields obtained for large particle sizes(as discussed above).

The most striking feature of Figure 4 is the presence of maximum tarcollection magnitudes for moderate reactor temperatures when the quenchingis realtively rapid (RH ~ 1.5Rp •o)' As was described above, although largerprimary tar production occurs for increasing temperatures (see Fig. 2),corres­ponding increases in the secondary reaction rates are in competition with this

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

108 R. S. MILLER AND J. BELLAN

process. Therefore, an optimal pyrolysis temperatue for tar collection mayexist, depending on the efficiency of the quenching process. As quenchingbecomes less efficient, the optimal pyrolysis temperature decreases and tarcollection becomes a decreasing function of TR • The existence of optimaltemperatures for tar collection is confirmed by past maple wood pyrolysisobservations of Scott and Piskorz (1982) and Scott et al. (1988). Although inthese experiments, in which only very small particle sizes (~ IOOllm) wereinvestigated, the optimal reactor temperature was observed to be ;:;, 775Kwhich is in good agreement with the current results.

The initial particle size is also expected to influence tar collection frombiomass pyrolysis by altering the tar residence time inside the particle. Thiseffect is highlighted in Figure 5 for three different reactor temperatures andthree different harvest radii. In all cases, tar collection is observed to decreasewith increasing particle size at constant temperature. This result is indepen­dent of the harvest radius and is due to an increase in the tar residence timeinside of the particle. For collection locations relatively far from the particlesurface (Rn ~ 1.5Rp •o), the tar collection decreases with increasing reactortemperature within the range considered. This is again due to the correspond­ing increased secondary tar decomposition for large temperatures.

The most interesting feature of Figure 5 is the observed crossing points forthe tar collection for small harvest radii, corresponding to rapid quenching. Inthese cases (Figs. 5a, 5b) the optimal reactor temperature for tar production isa function of the particle size. For small particles ~ 1mm, tar residence timesare decreased and the tar collection is governed predominantly by the primaryproduction reaction; therefore, maximal collections are observed at large

_700K....... 900K.......•....... 1lOOK

0.3

:;.,..~::., .....

"' '-.t.:.:.:.:.:.:: :..:.:.:

(a)

0.0 '0.0 0.2 0.4 0.6 0.8 1.0

INITIAL PARTICLE RADIUS [em]

FIGURE 5 Tar collection for maple as a function of the initial particle size for various values ofT.; (a) RI//R

p•o = I, (b) RI//R

p•o = 1.25, and (c) RI//R

p•o = 1.5.

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS 109

0.3

zoE=: 0.2

~u~ 0.1..:E-<

___ 700K,,-_."_ 900K................ 1lOOK

' ..' .."" ........ ........

.....................

(b)

0.0 '-:'----..,....,.............,..........~~~......,..~..,...0.0 0.2 0.4 0.6 0.8 1.0

INIDAL PARTICLE RADIUS [em]

0.3

zoE=: 0.2

~8~ 0.1;:;

___ 700K""••• " 900K········.··,,···1 lOOK

.-.: ....'-......-.•.

'"'. ......... ~ -v,

....•.•......•.•....•..............:.:.:.:.:.:"..".:.:.:.:.:

(c)

0.0 ..,...--..,.....,...~,........~...........~..........~......0.0 0.2 0.4 0.6 0.8 1.0

INIDAL PARTICLE RADIUS [em]

FIGURE 5 (Continued).

reactor temperatures. However, as the particle size increases, along with tarresidence times, secondary tar decomposition dominates at high temperatures.This results in desirable lower reactor temperatures for tar collection.

3.3. Effects of the Carrier Gas

All of the above results were obtained using steam as an inert gas; however, inmany experiments nitrogen is used to purge tar from the particles and make itavailable for collection (e.g., Scott et al., 1988). Figure 6 illustrates the effect ofthe carrier gas by comparing the pyrolysis of maple wood using both steamand nitrogen as inert gases, under otherwise identical conditions. The initialparticle size is R p •o = 0.5 cm and the reactor temperature is TR = 900K.

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

110 R. S. MILLER AND J. BELLAN

(a)

800

700

~600

/I

!;j 500

400

0.2 0.4 0.6tlte

0.8 1.0

0.25

60.20

~ 0.15...l...l

80.10c::-e.... 0.05

0.000 50(b)

1.0

1.5

2.0

250

FIGURE 6 Effects of altering the carrier gas on the pyrolysis of a maple particle withR =0.5cm and T. =900K. (a) The mass averaged particle temperature as a function ofn6r~alized time. (b) Temporal evolution of the tar collection at various values of Ru/R o:

p.

Figure 6a depicts the mass averaged particle temperature:

(16)

as a function of normalized time. The mass averaging is with respect to thesolid phase species, neglecting the char contribution, and this parameter hasbeen discussed previously in Miller and Bellan (1996) and Miller and Bellan([997). The primary difference between the two carrier gases affecting thepresent results is in the magnitude of the specific heat capacity (see Tab. I)

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS 111

which is more than 2.5 times larger for steam. The associated larger specificinternal energy of the steam results in an increased "thermal mass" for heatexchange with the particle. Therefore, the maple sample in the steam environ­ment is heated to, and reacts at, higher temperatures than the same sample in anitrogen environment (Fig. 6a). As discussed above, the larger reaction tem­perature causes an associated decrease in the tar collection when rapidquenching is performed. The small reversal in trends observed at RH = 2Rp •o isrelated to the far field temperature exterior to the particle. Thus, care must betaken in extrapolating any results to different reactor conditions. Neverthe­less, the primary influence by the inert gas on pyrolysis is through its heatcapacity and corresponding changes in the effective reaction temperature.

3.4. Sensitivity Analysis

The above particle pyrolysis calculations were performed by using the same setof properties for all biomass species. The only differences between feedstocks istheir composition in terms of mass percentages of cellulose, hemicellulose andlignin. In their current form, the calculations are made for identical propertiesand heats of reaction for all three of the primary biomass components (cellulose,hemicellulose and lignin). In addition, these properties and heats of reaction areassumed to be independent of the source feedstock. This is the result of very'limited data available regarding these properties, making it impossible to tailortheir values for each component and/or feedstock. Miller and Bellan (1997)suggested that observed deviations between the model predictions and experi­mental results for large (> lcm) particles may be attributed to the lack ofspecificity in property values particularly to the assumed feedstock indepen­dence. In order to determine the importance of these effects on tar collection, asensitivity study of several of the assumed biomass thermal properties is made.In particular, the biomass apparent density, thermal conductivity, heat capacityand primary heats of reaction are investigated.

Figure 7 compares the effects of the initial apparent density of the biomasson the temporal evolution of the tar collection at several harvest radii. Allsimulations are for pine wood with Rp •o = O.5cm and TR = 900K 'and the tarcollection values are for various harvest radii, RH/Rp •o' The base value for thecalculations is P". = 650kg/m3 which was taken from experimental hardwoodmeasurements made by Koufopanoset al. (1991). However, this value may betoo large for softwoods and other biomass feedstock. For example, Pyle andZaror (1984) presented experimental measurements in the range [450kg/m 3,

550kg/m3] for pine wood. In fact, the deviations found in Miller and Bellan

(1997) for large particles were for pine wood pyrolysis comparisons. These

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

112

0.20

~ 0.15

~:::l 0.10

8~

~ 0.05

R. S. MILLER AND J. BELLAN

1.0

1.5

300

FIGURE 7 Temporal evolution of the tar collection at various of RII/Rp•o for pine. The initialapparent densities of the biomass are Po=650kg/m' (solid line) and Po=450kg/m' (dotted line).The particle conditions are: Rp •o = O.5cm and TR = 900K.

deviations were attributed primarily to the predicted pyrolysis initializationbeing longer than observation. Figure 7 shows that these deviations may beattributed to the apparent density. Lower values of the density result in earlierpyrolysis initialization; however, the final tar yields are not altered significant­ly. Therefore, care should be taken in prescribing the apparent density forsoftwoods and other feedstocks, particularly when temporal evolutions are ofinterest.

Figure 8 presents the temporal evolution of the tar collection from maplewood as a function of both the biomass thermal conductivity and heatcapacity. In all cases, the initial particle size is Rp•o =0.5cm, TR = 900K, andthe tar collection is at the particle surfce (RH = Rp •o)' Three values for eachparameter are chosen, including the base value. Figure 8a shows that neither a25% increase Of decrease in the thermal conductivity significantly alters thetar collection. Increases in the conductivity allow larger particle/reactiontemperatures and result in larger tar yields at the particle surface (see Fig. Sa).On the other hand, tar collection is particularly sensitive to changes in the heatcapacity by as little as ± 10% (Fig. 8b). Large heat capacities initially causedelayed heating of the particle due to increased thermal inertia. However, atlater times during the reaction, the larger heat capacities result in an increasedresistance to endothermic cooling and therefore are characterized by a largerreaction temperature. These competing effects cause the curve crossing atintermediate times in Figure 8b and result in the increasing final tar collec­tions with heat capacity.

The final parameters for sensitivity investigation are the heats of reaction forthe primary pathways K 2 and K 3. Effects of these parameters on the tar

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS 113

0.20

Z00.15!=:&J:J 0.100U<>:~ 0.05

0.000 50 100 150 201l 250

(a) TIME [s]

0.25

Z 0.200

~ 0.15...J...J0U 0.10<>:-cf-o 0.05

0.000 50(b)

FIGURE 8 Temporal evolution of the tar collection at RIl/Rp,o = I for maple; (a) varying thebiomass thermal conductivity ±25%, and (b) varying the biomass heat capacity ± 10%,The particle conditions are Rp•o=0.5cm and T. =900K, and the base values are.lw = 1.256 X 10- 4 kl/ms- K and C., =2.3 kJ/kg- K.

collection evolution are depicted in Figure 9 for the same particle conditionsas in Figure 8. In both cases, the heat of reaction is altered by ± 20%, Thesensitivity of the endothermic heat of reaction is not.substantial as observed inFigure 9a. Increasing the magnitude of !'>h 2 decreases the particle and reactiontemperature; this correspondingly decreases the tar collection at the surface ofthe particle as discussed above, The three curves for various values of !'>h3 (charproduction reaction) almost completely overlap (Fig. 9b). This is because athigh temperatures the primary tar reaction dominates over the char produc-:tion reaction; particularly when weighted by the heat of reaction. At lower

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

114 R.S. MILLER AND J. BELLAN

0.20

80.15

6[.LI

:l 0.100U

~0.05

0.00. 0 50 100 150 200 250

(0) TIME [s)

0.20

80.15

~:l 0.100ue<:~0.05

0.000 50 250(b)

FIGURE 9 Temporal evolution of the tar collection at RHIR = I for maple; (a) varying theheal of reaction t.I1, by ±20%, and (b) varying the heat of re~gtion t.I1, ±20%. The particleconditions are R = 0.5cm and T. = 900K, and the base values for heals of reaction aret.I1, = +255kJ/kg"~nd t.I1, = - 20kJ/kg.

reactor temperatures, such as those employed for charcoal production, it isexpected that changes in dh) will have a greater impact than changes in dh 2•

4. CONCLUSIONS

Tar yields and collection from relatively large particles have been predictedusing the detailed biomass pyrolysis model of Miller and Bellan (1997). Thenumerical results were obtained for spherically symmetric particles in initially

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS 115

quiescent environments. Results for a variety of biomass feedstocks, includingseveral hardwoods, softwoods, olive husk and wheat straw indicated that theabsolute deviation among feedstocks is more significant for char yields thanfor the mass of tar oils available for colJectiori. This was attributed toprolonged residence times and secondary tar reactions occurring inside of theparticles.

The results also show that the competing processes of primary tar produc­tion and secondary tar decomposition interact in different ways depending onthe particular set of pyrolysis conditions. In particular, secondary tar reactionscan occur both in the particle's interior and in the exterior boundary layer.Interior reactions become important when the tar residence time is large,whereas outer reactions are primarily important when the pyrolysis vapors arenot rapidly quenched near to the particle surface. In either case, the potentialtar collection is greatly reduced relative to the kinetic limits for realistic finiteparticle sizes. In addition, as the quenching process becomes less efficient dueto the secondary reactions, tar collection is substantially reduced from thatexiting the particle surface:

An important result which is conceptually applicable to reactor design is theexistence of an optimal reactor temperature for maximizing the tar collection;this optimal temperature is the direct result of the competing tar reactionsdiscussed above. For particle sizes ~ Icm, the optimal temperature was ob­served to be ;::; 875K for efficient quenching near the particle surface. This resultis in good agreement with Scott et al.'s(1988) past experiments in a bench scalefluidized bed reactor. Tar collection is a strong function of the initial particlesize; predominantly due to alterations in tar residence times. The optimalreactor conditions for tar production are highly dependent on the particle size:The simulations show that when quenching is efficient, tar collection increasesfor smalJ particles (~lmm)with increasing reactor temperature; however,largerparticles (~lcm) show the opposite behavior with temperature. For poorquenching (far from the particle surface), decreasing tar yields are obtained withincreasing temperature for all particle sizes considered.

In addition to reactor temperature, initial particle size and quenching, thechoice of an inert carrier gas can also affect the pyrolysis evolution. The resultsshow that for efficient quenching processes, the use of nitrogen as a carrier gascan lead to significantly improved tar yields over those obtained with inertsteam. This difference in behavior was attributed primarily to differences in theheat capacities of the two gases.

Finally, results from a sensitivity analysis of several thermo-chemicalproperties of the biomass were presented. Simulations illustrated the sensitivityof tar yields to the apparent density, thermal conductivity, heat capacity and

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

116 R. S. MILLER AND J. BELLAN

primary heats of reaction. Conversion times were significantly decreased byaltering the apparent density from hardwood to softwood measured valuesdue to earlier pyrolysis initialization; however, final tar yields were relativelyunchanged in magnitude. The results also suggested that the pyrolysis yieldsare relatively insensitive to the thermal conductivity and primary heats ofreaction. However, a relatively strong sensitivity to the biomass heat capacitywas found.

Acknowledgements

This research was conducted at the Jet Propulsion Laboratory (JPL) andsponsored by the U.S. Department of Energy (DOE), with Mr. Neil Rossmeissl(DOE Headquarters) and Mr. D. Hooker (DOE Golden Center) serving ascontract monitors, under an agreement with the National Aeronautics andSpace Administration. Computational resources are provided by the supercomputing facility at JPL. The authors wish to thank Dr. Esteban Chornetand Dr. Robert Evans of the National Renewable Energy Laboratory forhelpful discussions.

NOMENCLATURE

C

dDeKMNpr

RRSt

TuXY

Specific hea tCharacteristic pore length scaleMolecular species dilfusivitySpecific internal energyReaction rateMolecular weightTotal number of speciesPressureRadial coordinateRadial positionUniversal gas constantReaction source/sink termTimeTemperatureGas phase velocityChar formation mass ratio for reaction K 3

Gas phase mass fraction

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

TAR YIELD FROM PYROLYSIS OF BIOMASS

Greek Symbols

117

/).h

B

nAJl

PP(J

w

Heat of reactionPorosityDivergence of the velocityThermal conductivityMolecular viscosityApparent densityTrue densityStefan-Boltzmann constantEmissivity

Subscripts and Superscripts

oceff

9H

Rst

Tv

. Initial valueConversionEffectiveGas phaseHarvestSpecies i .ReactorSolid phaseTotal (all species and phases)ThermalConstant volumeExcluding char.

References

Antal, M.1 (1982)Biomass pyrolysis: A reviewofthe literature. part LCarbohydrate pyrolysis. InBoer, K. and Duffie, J. editors, Advances in Solar Energy, pages 61-11 I. American SolarEnergy Society, Boulder, Co;

Chornet, E., Wang, D., Montane, D. and Czenik, S. (1994) Hydrogen production by fast pyroly­sis of biomass and catalytic steam reforming of pyrolysis oil. In Bridgwater, A. V. editor,Advances in Thermochemical Biomass Conversion, 1, pages 246-262. Blackie Academic andProfessional, New York.

Chum, H. L., Johnson, D. K., Agblevor, F. A., Evans, R.1, Hames, B. R., Milne, T. A. andOverend, R. P. (1994)Status of the [EA voluntary standards activity - round robins on wholewood and Iignins. In Bridgwater, A. V.editor, Advances in Thermochemical Biomass Conver­sion, 2, pages 1701-1716. Blackie Academic and Professional, New York.

Di Blasi, C. (l993a) Analysis of convection and secondary reaction effects within porous solidfuels undergoing pyrolysis. Combust. Sci. and Tech., 90, 315-340.

Di Blasi, C. (1993b) Modeling and simulation of combustion processs of charring and non­charring solid fuels. Prog. Energy Combsut. Sci., [9, 71-104.

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015

118 R. S. MILLER AND 1. BELLAN

Diebold, J. P. and Power, A.(1988) Engineering aspects of the vortex pyrolysis reactor to produceprimary pyrolysis oil vapors for use in resins and adhesives. In Bridgwater, A. V. andKuester, 1. L. editors, Research in Thermochemical Biomass Conversion, pages 609-628.Elsevier Applied Science, New York.

Koufopanos, C. A., Papayannakos, N., Maschio, G. and Lucchesi, A. (1991) Modelling of thepyrolysis of biomass particles studies on kinetics, thermal and heat transfer effects. Can. J.Chern. Eng., 69, 907-915.

Liden, A. G., Berruti, F. and Scott, D. S. (1988) A kinetic model forthe production of liquids fromthe flash pyrolysis of biomass. Chern. Eng. Comm., 65, 207-221.

Lim, K. S., Zhu, J. X. and Grace, J. R. (1995) Hydrodynamics of gas-solid fluidization. Fuel,21(Suppl.)(1994) 141-193.

Maschio, G., Lucchesi, A. and Koufopanos, C. A.(1994)Study of kinetic and transfer phenomenain the pyrolysis ofbiomas particles. In Bridgwater, A. V. editor, Advances in ThermochemicalBiomass Conversion, 2, pages 746-759. Blackie Academic and Professional, New York,

Miller, R. S. and Bellan, 1. (1996) Analysis of reaction products and conversion time in thepyrolysis of cellulose and wood particles. Comb. Sci. Tech., 119,331-373.

Miller, R. S. and Bellan, 1.(1997) A generalized biomass pyrolysis model based on superimposedcellulose,hemicellulose and ligninkinetics. Comb. Sci. Tech., in print in this issue.

Mok, W. S., Antal, M. J., Szabo, P., Yarhegyi, G. and Zelei, B. (1992) Formation of charcoalfrom biomass in a sealed reactor. Ind. Eng. Chern. Res., 31,1161-1166.

Narayanan, R. and Antal, M. 1. (1996) Thermal lag, fusion, and the compensation effect duringbiomass pyrolysis. Ind. Eng. Chern. Res., 35(5), 1711-1721.

Pyle, D. L. and Zaror, C. A. (1984) Heat transfer and kinetics in the low temperature pyrolysis ofsolids. Chern. Eng. Sci., 39(1),147-158.

Roy, C; Blanchette, D., de Caumia, B. and Labrecque, B. (1994) Conceptual design and evalua­tion of a biomass vacuum pyrolysis plant. In Bridgwater, A. V. editor, Advances in Ther­mochemical Biomass Conversion, 2, 1165-1186. B1ackie Academic and Professional, NewYork.

Scott, D. S. and Piskorz, J. (1982) The flash pyrolysis of aspen-poplar wood. Can. J. Chern. Eng.,60,666-674.

Scott, D. S., Piskorz, J., Bergougnou, M. A., Graham, R. and Overend, R. P. ( 1988)The role oftemperature in the fast pyrolysis of cellulose and wood. Ind. Eng. Chern. Res., 27, 8-15.

SERI (July 1979) A survey of biomass gasification: Volume II - Principles of gasification.Technical Report TR·33-239, Solar Energy Research Institute, Golden, Colorado.

Thurner; F. and Mann, U. (1981)Kinetic investigation of wood pyrolysis. Ind. Eng. Chern. ProcessDes. Dev., 20, 482-488.

Wagenaar, B. M., Kuipers, J. A. M., Prins, W. and van Swaaij, W. P. M. (1994) The rotating·cone flash pyrolysis reactor. In Bridgwater, A. V. editor, Advances in ThermochemicalBiomass Conversion, 2, 1122-1133. Blackie Academic and Professional, New York.

Ward, S. M. and Braslaw, J. (1985) Experimental weight loss kinetics of wood pyrolysis undervacuum. Comb. and Flame, 61, 261-269.

Dow

nloa

ded

by [

Cle

mso

n U

nive

rsity

] at

05:

58 2

8 Se

ptem

ber

2015