Ca

BD

a

ARRAA

KBRPKB

1

ctosap

aobacrcpt

h0

Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229

Contents lists available at ScienceDirect

Journal of Analytical and Applied Pyrolysis

journa l h om epage: www.elsev ier .com/ locate / jaap

oupled reactor and particle model of biomass drying and pyrolysis in bubbling fluidized bed reactor

ijan Hejazi ∗, John R. Grace, Xiaotao Bi, Andrés Mahecha-Botero1

epartment of Chemical and Biological Engineering, University of British Columbia, 2360 East Mall, Vancouver, BC, Canada, V6T 1Z3

r t i c l e i n f o

rticle history:eceived 23 January 2016eceived in revised form 16 July 2016ccepted 4 August 2016vailable online 5 August 2016

eywords:iomass particleenewable energyyrolysisinetic modelingubbling fluidized bed reactor

a b s t r a c t

A comprehensive single particle model of biomass drying and pyrolysis is coupled with an ideal reac-tor model of a bubbling fluidized bed with nitrogen as the fluidizing gas. To predict biomass pyrolysisproducts yield and composition under steady-state operation, a two-step biomass pyrolysis kinetic mech-anism is adopted with primary pyrolysis described by two different reaction kinetic schemes: (a) Threeparallel first-order reactions producing primary pyrolysis products (non-condensable gas, tar/bio-oil andchar), and (b) a detailed solid state kinetics mechanism based on multiple reactions of lignin, cellulose, andhemicellulose contents of two specific types of biomass (poplar and lodgepole pine). Secondary pyrolysisis modeled by two parallel reactions describing homogeneous thermal cracking of tar to non-condensablegas and char. In addition to the yields of pyrolysis products often modeled as lumped species, this modeladdresses an existing gap of knowledge in predicting the proportions of major compounds in the pyrol-ysis gas based on a few simplifying assumptions and CHO elemental balances. This predictive model is auseful tool to relate biomass pyrolysis products yield and composition to process operating parameterssuch as biomass ultimate analysis, reactor temperature, gas residence time, mean solids residence timeinside the reactor, as well as biomass particle size and moisture content. Model predictions for the twokinetic schemes are in good agreement with available experimental data from the literature.

© 2016 Elsevier B.V. All rights reserved.

. Introduction

Emission of enormous amounts of greenhouse gases to the atmosphere as a byproduct of burning fossil fuels is causing unfavorablelimate change. To move towards a sustainable energy future and to combat the greenhouse effect, developing alternative technologieshat make use of renewable energy sources is of crucial importance. The use of the energy contained in biomass as the best form of storagef solar energy is rapidly growing worldwide. Biomass pyrolysis is the thermo-chemical decomposition of solid biomass into a carbon-richolid (char) and volatile matter (bio-oil and non-condensable gases) by heating in the absence of oxygen. The volatile products of pyrolysisre considered as carbon neutral and renewable fuels, offsetting fossil fuel consumption. The generated bio-char can be used for heatingurposes or incorporated into soils to provide stable storage for carbon over a long time [1].

Fluidized bed reactors are among the mature technologies for thermal treatment of carbonaceous feedstock due to their effective heatnd mass transfer, temperature uniformity, high solid flow rates and flexibility with regards to fuel quality [2]. Enhancing the feasibilityf biomass pyrolysis in fluidized bed reactors is important from both scientific and industrial points of view. Upon entering a fluidizeded reactor, biomass particles are exposed to high heating rates that cause rapid increase of particle temperature, evaporation of moisturend devolatilization into volatile matter that constitutes more than 80% of the original particle mass. The rest remains in the form of solidhar (predominantly carbon). The sensitivity of pyrolysis gas and tar compositions to key process parameters such as biomass properties,eactor temperature, gas residence time, mean solids residence time inside the reactor, as well as biomass particle size and moisture

ontent, have mostly been overlooked in the literature. For instance, many researchers have assumed instantaneous pyrolysis of biomassarticles when modeling biomass gasifiers [3–8]. This assumption limits the numerical solution to after the pyrolysis stage, possibly leadingo under-prediction of methane in the syngas [9].

∗ Corresponding author.E-mail addresses: bijanhejazi@gmail.com, bhejazi@chbe.ubc.ca (B. Hejazi).

1 Present Address: NORAM Engineering, 200 Granville Street, Suite 1800, Vancouver, BC, Canada, V6C 1S4.

ttp://dx.doi.org/10.1016/j.jaap.2016.08.002165-2370/© 2016 Elsevier B.V. All rights reserved.

214 B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229

Nomenclature

A Bubbling bed cross-sectional area, m2

Cp Specific heat capacity, J/kg Kdb Bubble size, mdpore Pore diameter, mdp,sand Sand particle diameter, mDi Molecular diffusivity of the i th species, m2/sDeff,i Effective diffusivity of the i th species, m2/sE Activation energy, kJ/molE(t) Solids residence time distribution function, s−1

Ej Activation energy of the j th reaction, kJ/molg Acceleration of gravity, 9.81 m2/shmass,i Mass transfer coefficient of the ith species, m/shoverall Overall heat transfer coefficient, W/m2 Kk0j Pre-exponential factor of the j th reaction, s−1

kj Arrhenius-type kinetic rate constant of the j th reaction, s−1

k Thermal conductivity, W/m Kkcond Molecular thermal conductivity of gas and solid phases, W/m Kkeff Effective thermal conductivity, W/m Kkrad Radiative conductivity of solid particle, W/m KLbed Dense bubbling bed height, mm Mass, kg·

m Mass flow rate, kg/sMP Biomass particle hold-up inside reactor, kgMW Molecular weight, g/moln Particle shape parameter, n = 0 for flat, n = 1 for cylinder, n = 2 for sphereNor Number of holes in distributor plate, −Nu Nusslet number, −P Pressure, PaPr Prandtl number, −qflux Source heat flux, W/m2

Q Volumetric flow rate, m3/sr Radial coordinate inside the particle, mR Universal ideal gas constant, 8.314 J/mol KRe Reynolds number, −Rp Spherical biomass particle radius, mSc Schmidt number, −Sh Sherwood number, −t Time, sT Temperature, Ku Gas-phase velocity inside particle pores, m/sU Superficial gas velocity, m/sUmf Superficial gas velocity at minimum fluidization, m/swC ,wH ,wO Carbon, hydrogen and oxygen mass fractions, −XB Dry-ash-free biomass conversion, −XP Moist biomass particle conversion, −XP Average biomass particle conversion inside the fluidized bed, −Y Average yield at reactor exit, kg/kg moist biomassz Axial coordinate along the reactor height, m

Greekˇg Permeability of gas phase, −�i Stoichiometric coefficient of i th species in tar cracking reaction, −�Hrxn,j Heat of j th reaction, kJ/kgε Bed voidage at height z/porosity, −ε Average bubbling bed voidage, −� Viscosity, kg/m s� Density, kg/m3

�i Density of ith species inside particle pores, kg/m3

�i,bulk Bulk density of ith species inside reactor, kg/m3

�P Average biomass particle density inside bubbling bed, −

B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229 215

� Stefan-Boltzmann constant, 5.75 × 10−8 W/(m2·K4)� Particle tortuosity,-�g Total gas residence time inside reactor, s�s Mean solids residence time, sω Emissivity, −

Subscriptsb BubbleB BiomassB0 Virgin biomassC Char, carbonCf Final charg Gas phaseG Non-condensable gasG2 Secondary gasH Hydrogeni Species numberj Reaction number, element numberM MoistureM0 Initial moistureO OxygenP ParticleS Solid

cbtdttatb

kawpsIcttahp

2

sd

T Tar (Bio-oil)V Water vapor

For almost all engineering applications, researchers resort to fixed values or temperature-dependent experimental correlations for theomposition of the pyrolysis gas [10–14]. In those gasification and pyrolysis kinetic models in which Arrhenius-type correlations haveeen adopted to capture the temperature dependency of pyrolysis reactions [15–17], the effect of other process parameters is often notaken into account. Note that biomass particle size, for instance, can change the pyrolysis product distributions (yield and composition)ue to the change in the heating rate of biomass particles inside the reactor. For almost all gasification and pyrolysis models developed inhe literature, the values adopted for pyrolysis gas and tar compositions do not satisfy elemental balances, and in the few cases in whichhey do (e.g. [15]), information on bio-oil composition as a model input parameter is missing. Moreover, a subset of tar compounds (e.g.cetic acid (CH3COOH), acetone (CH3COCH3), ethanol (CH3CH2OH), phenol (C6H5OH), naphthalene (C10H8), etc. [9,15,18,19]) is consideredhat cannot fully represent the myriad of low- and high-molecular-weight organic compounds detected in the tar (bio-oil) generated fromiomass pyrolysis.

In this modeling study, a more fundamental modeling approach is taken to improve the previous research, address some gaps innowledge and shed light on areas of research that deserve further scrutiny. By coupling the single particle model of biomass dryingnd pyrolysis with an ideal reactor model of a bubbling fluidized bed and constraining the coupled single particle and reactor modelith CHO elemental balances, biomass drying and pyrolysis products yield, as well as the composition of principal components in theyrolysis gas are predicted as functions of the most important operating parameters, such as reactor temperature and biomass particleize. To address the uncertainties in tar measurement and analysis, we treat tar containing a mixture of carbon, hydrogen and oxygen.n addition to kinetic parameters and fuel-related and process-related parameters, the only other model input parameters are tar andhar elemental compositions, which are straightforward to obtain experimentally from ultimate and proximate analyses. Our objective iso build a detailed kinetic model of drying and pyrolysis of biomass particles in a bubbling fluidized bed (BFB) reactor with nitrogen ashe fluidizing gas. Another advantage of developing an independent kinetic model of biomass drying and pyrolysis is that it can be testedgainst experimental data to provide the basis for developing a seamless kinetic model of a BFB gasifier by incorporating homogeneous andeterogeneous biomass gasification reactions in the presence of an oxidizing agent such as air or steam. Taking into account the drying andyrolysis steps contributes to building kinetic models capable of more realistically evaluating the overall performance of biomass gasifiers.

. Model development and simplifying assumptions

A comprehensive coupled particle and reactor model is developed to predict biomass drying and pyrolysis products yield and compo-ition generated by bubbling fluidized bed reactor under steady-state operating conditions. To develop a sufficiently complex model forrying and pyrolysis of biomass particles in a bubbling fluidized bed (BFB) reactor, the following simplifying assumptions are adopted:

1 The reactor operates under steady-state conditions.2 The freeboard region of the BFB is ignored.3 Solids entrainment is ignored.

4 Uniform temperature and atmospheric pressure are assumed throughout the dense bed.5 The solids in the BFB reactor are perfectly mixed, and there is an even distribution of pyrolysis products throughout the dense bed

height.6 Plug flow of gases and volatiles is assumed inside the reactor.

216 B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229

Fig. 1. Kinetic mechanism of biomass pyrolysis and evaporation of moisture [22,23].

Table 1Kinetic parameters for biomass drying and pyrolysis.

Rxn. (j) k0j (1/s) Ej (kJ/mol) Hrxn,j (kJ/kg) Ref.

1 4.38 × 109 152.7 64 [24,26]2 1.08 × 1010 148 64 [24,26]3 3.27 × 106 111.7 64 [24,26]4 4.28 × 106 107.5 −42 [25,23]

1

1111

2

edmvbic

wc

bbtklkfcfi

wt

5 1 × 105 108 −42 [27,23]6 5.13 × 1010 88 2440 [28]

7 Local thermal equilibrium between gas and solid phases inside biomass particles.8 Ideal gas law for all volatiles (gas and bio-oil) released from biomass pyrolysis.9 Particle size and shape do not change during the degradation process, i.e. shrinkage, cracking and fragmentation are neglected.0 Pyrolysis involves a two-step process where primary and secondary pyrolysis are described by three parallel heterogeneous reactions

and a homogeneous tar cracking reaction, respectively.1 First-order chemical reactions with constant heats of reaction are assumed for drying, primary and secondary pyrolysis.2 The sand particles are inert and do not catalyze reactions.3 The ash plays a negligible role in catalyzing reactions.4 For elemental balances, each lumped species, including unreacted biomass, char, non-condensable gas and tar, is treated as a homoge-

neous mixture of carbon, hydrogen and oxygen. Other elements, including nitrogen and sulfur, are neglected.

.1. Particle model description

During drying and pyrolysis, biomass particles act as a porous medium allowing outflow of water vapor and volatiles into the reactornvironment. To study transport phenomena with reaction kinetics inside a biomass particle, a single particle model is developed forrying and pyrolysis in which temperature and concentration gradients inside the particle of changing thermo-physical properties areodeled, based on conservation of mass, energy and momentum. For small enough particles (<1 mm in diameter), the effect of particle

olume shrinkage during thermo-degradation is negligible [20], and the pyrolysis of single unchanging-size biomass particles is describedy the Progressive Conversion Model (PCM) [21], in which a constant biomass particle size is assumed, while its density decreases due to

ncreasing porosity during the course of the reaction. If mB0 denotes the initial mass of a dry biomass particle and mCf is the mass of finalhar in the completely-converted particle, the dry-ash-free biomass conversion is defined as:

XB = (mB0 − (mB + mC )) /(

mB0 − mCf

)(1)

here mB and mC are the instantaneous masses of biomass and char in the solid particle, respectively. The dry-ash-free biomass conversionan also be written in terms of mass densities of biomass and char at constant particle volume as:

XB = (�B0 − (�B + �C )) /(

�B0 − �Cf

)(2)

Fig. 1 illustrates the generic two-step biomass pyrolysis reaction mechanism originally proposed by Shafizadeh et al. [22], where dryiomass is converted through three competitive paths to non-condensable gas, tar and char during the primary pyrolysis. This is followedy secondary pyrolysis at higher temperatures in which thermal cracking of tar generates more gas and char. Parallel to biomass pyrolysis,he evaporation of biomass particle moisture content to water vapour is expressed as an independent chemical reaction [23]. The selectedinetic parameters for primary pyrolysis are adopted from Di Blasi et al. [24], who examined the kinetics of primary degradation of thinayers of beech wood powder (150 �m) in a tube furnace in the 300–435 ◦C temperature range with high heating rates (1000 K/min) underinetically-controlled and isothermal conditions. The kinetic parameters for thermal cracking of tar to non-condensable gas are adaptedrom Liden et al., 1988 [25] who correlated experimental data from a flash pyrolysis pilot plant which used whole tree poplar wood andlean aspen-poplar wood in a fluidized bed reactor in the temperature range 500–650 ◦C, also extendable to 400–700 ◦C with caution. Forrst-order reactions, all reaction rate constants are expressed in Arrhenius form:

kj = k0j exp(−Ej/R.T

)(3)

ith the kinetic parameters and heats of reaction are summarized in Table 1. Note that although heats of reaction are commonly assumedo be constant in the literature, it is also possible to more accurately formulate them as functions of temperature [29].

((v

t

bfa

w

td

a

wt

t

pt

wg

B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229 217

In the particle model, a total of seven species are included. For simplicity, the densities of the solid phase species, i.e. biomass (B), charC) and moisture (M), are defined per total particle volume, while the densities of the gas phase species, i.e. gas (G), tar (T), water vapourV) and inert gas (N2), are defined per pore volume inside the particle. For the solid phase, conservation of mass with constant particleolume gives:(

∂�B/∂t)

= − (k1 + k2 + k3) .�B (4)(∂�C/∂t

)= k3.�B + k5. (εP.�T ) (5)(

∂�M/∂t)

= −k6.�M (6)

The second term on the right hand side of Eq. (5) is for secondary char generation through repolymerization of tar within the pores ofhe particle. The particle porosity, εP , is approximated as a linear function of biomass conversion:

εP = εB0. (1 − XB) + εCf .XB (7)

As the two-step pyrolysis kinetic model is reported to have limitations with regards to decomposition kinetics of different types ofiomass with varying properties [30], we also apply a correlation from Mochulski [31], who developed a solid state kinetics mechanismor multiple reactions of biomass pyrolysis based on comprehensive experimental analysis of two specific types of woody biomass (poplarnd lodgepole pine) as a function of their lignin, cellulose, and hemicellulose contents:

(∂XB/∂t

)=

[2.74 × 1012exp

(−172310/ (R.TP)

)× 3

2(1 − XB)

2⁄3(

1 − (1 − XB)1⁄3

)−1

× 0.06

]

+[1.77 × 108exp

(−113500/ (R.TP)

)× (1 − XB) × (−0.12 × cellulose + 0.72 × hemicellulose + 0.67 × lignin)

]+

[4.06 × 108exp

(−135000/ (R.TP)

)× 2(1 − XB)

1⁄2 × (1.05 × cellulose + 0.62 × lignin)

] (8)

here, from differentiation of both sides of Eq. (2) with respect to time:(∂XB/∂t

)= −

((∂�B/∂t

)+

(∂�C/∂t

))/(

�B0 − �Cf

)(9)

Note that this kinetic model gives the global rate of biomass conversion (i.e. primary pyrolysis) without providing information abouthe distribution of pyrolysis products. Therefore, two additional adjustable parameters are required to fully predict the pyrolysis productsistribution in terms of char, tar and gas:

Biomass → YCChar + YT,maxTar + YGGas (10)

As a first approximation, we use the fixed-carbon fraction (FC) obtained from proximate analysis to estimate the yield of char generations: (

∂�C/∂t)

= −FC.(∂�B/∂t

)(11)

here FC = 10 wt% according to Wagenaar et al. [32], who pyrolysed pine in a rotating cone reactor. Upon combination of Eqs. (9) and (11),he overall rate of biomass consumption is obtained as:(

∂�B/∂t)

= −(∂XB/∂t

).(

�B0 − �Cf

)/ (1 − FC) (12)

As the second adjustable parameter, ultimate tar yield is set as YT,max = 0.703 from reference [25] where the kinetic parameters forhermal cracking of tar to non-condensable gas are adopted.

Assuming a one-dimensional and time-dependent domain and accounting for convective and diffusive transport phenomena of gashase species inside the pores of the particle, we formulate the gas phase mass balance equations as a function of radial coordinate (r) andime (t):

∂ (εP.�i) /∂t +(

1⁄rn)

.∂(

rnεP.�i.u)

/∂r =(

1⁄rn)

.∂(

rnεP.Deff,i.(∂�i/∂r

))/∂r +

⎧⎪⎪⎪⎪⎨⎪⎪⎪⎪⎩

i = G :(∂�G/∂t

)primary

+ k4. (εP.�T )

i = T :(∂�T /∂t

)primary

− (k4 + k5) . (εP.�T )

i = V : k6.�M

i = N2 : 0

(13)

here n is 0 for a flat plate or a slab-shaped particle, 1 for a cylindrical particle, and 2 for a spherical particle, and the rates of tar and gaseneration from primary pyrolysis are calculated for the two kinetic schemes as:

(∂�G/∂t

)primary

={

k1.�B Two − step kinetic model

−(

1 − YC − YT,max)

.(∂�B/∂t

)Solid − state kinetic model

(14)

(∂�T /∂t

)primary

={

k2.�B Two − step kinetic model

−YT,max.(∂�B/∂t

)Solid − state kinetic model

(15)

2

(

wf

w

a

a

oi

w

bt

w

cr

w

wg

18 B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229

Given the overall gas-phase density,

�g =∑

i=G, T, V, N2

�i (16)

the overall mass balance equation for the gas phase inside the particle is obtained as:

∂(

εP.�g

)/∂t +

(1⁄rn

).∂

(rnεP.�g.u

)/∂r =

{(k1 + k2)�B − k5. (εP.�T ) + k6.�M Two − step kinetic model

− (1 − YC ) .(∂�B/∂t

)− k5. (εP.�T ) + k6.�M Solid − state kinetic model

(17)

In Eq. (13), Deff,i, the effective diffusion coefficient of gaseous species i is obtained from Di the molecular diffusivity of the ith speciesneglecting the contribution of Knudsen diffusion):

Deff,i = (εP.Di) /� (18)

here � is the tortuosity (taken as 1) and εP is the porosity of the particle. In Eq. (17), u is the one-dimensional gas-phase velocity calculatedrom Darcy’s law inside a porous medium as:

u = −(

ˇg,P/�g

).(∂P/∂r

)(19)

ith the dynamic viscosity of the gas phase calculated as:

�g =∑

i=G, T, V, N2

(�i.�i) /�g (20)

nd the permeability of the gas phase approximated by a linear function of conversion:

ˇg,P = ˇg,B0. (1 − XB) + ˇg,Cf .XB (21)

Large variations in permeability of solid biomass and char along and across the grain direction have been reported [33–35]. In order toccount for particle anisotropy, these physical properties are averaged across and along the grain direction [21,36] as:

ˇg,B0 = (ˇg,B0,along + ˇg,B0,across)/2 (22)

ˇCf = (ˇg,Cf,along + ˇg,Cf,across)/2 (23)

Note that while ideal gas behaviour can be assumed for the pyrolysis gas, this assumption may not be suitable for tar (bio-oil) consistingf high-molecular-weight condensable hydrocarbons. Nevertheless, in the absence of a better model, the ideal gas law is commonly usedn the literature. Assuming that the gas and solid phases are at local thermal equilibrium, the local instantaneous pressure is:

P =(

�g.R.TP

)/MWg (24)

here the average molecular weight of the gas phase is calculated as:

MWg =

⎛⎝ ∑

i=G, T, V, N2

(�i/

(�g.MWi

))⎞⎠−1

(25)

The particle temperature TP is obtained as a function of time and radial coordinate from an energy balance on a single particle. For porousiomass particles, the energy balance is given by combining the gas and solid phases in a single conservation equation [21], expressed inerms of species specific heat capacities (Cp) and densities:

(�BCp,B0 + �CCp,Cf + �MCp,M + εP.�gCp,g).(∂TP/∂t) + (1⁄rn).∂(rnεP.�gCp,g.TP.u)/∂r = (1⁄rn).∂(rnkeff .(∂TP/∂t))/∂r

−{(k1.�B.)�Hrxn,1 + (k2.�B).�Hrxn,2 + (k3.�B).�Hrxn,3 + (k4.εP�T ).�Hrxn,4 + (k5.εP�T ).�Hrxn,5 + (k6.�M).�Hrxn,6}(26)

here the heat capacity of the gas mixture is:

Cp,g =∑

i=G,T,V,N2

(�i.Cp,i)/�g (27)

Heat transfer inside the porous structure of biomass particle takes place through complex interaction between the transfer mechanisms:onduction, convection, and radiation [35]. The effective thermal conductivity due to the contribution of both molecular conductivity andadiative thermal conductivity, accounting for radiation in the pore system [21,33,35] is:

keff = kcond + krad (28)

here krad can be expressed as a function of third power of particle temperature [35]:

krad = 4.ωP.εP.�.TP3.dpore/ (1 − εP) (29)

here � = 5.67 × 10−8 W/(

m2.K4)

is the Stefan-Boltzmann constant. kcond is dependent on the thermal conductivities of the solid andas phases [21,33,35] as:

kcond = kP + εP.kg (30)

w

c

bf

a

wtstacad

w

w

sthwo

B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229 219

ith the solid phase thermal conductivity [21,33,35]:

kP = kB0. (1 − XB) + kCf .XB (31)

The pore diameter and particle surface emissivity are estimated as:

dpore = dpore,B0. (1 − XB) + dpore,Cf .XB (32)

ωP = ωB0. (1 − XB) + ωCf .XB (33)

The thermal conductivity of wood varies with the direction with respect to the grain, as well as with temperature, density, and moistureontent. As with the permeability, the thermal conductivities of biomass and char are averaged across and along the grain direction [36]:

kB0 = (kB0,along + kB0,across)/2 (34)

kCf =(

kCf,along + kCf,across

)/2 (35)

After specifying initial and boundary conditions, the coupled partial differential equations describing mass, energy and momentumalances are solved for lumped species densities inside the particle, as well as particle temperature, gas phase pressure and velocity asunctions of one-dimensional radial coordinate (r) and time (t). At time zero:

�B (t = 0, r) = �B0 ; �M (t = 0, r) = �M0 ; �C (t = 0, r) = 0 (36)

�G (t = 0, r) = �T (t = 0, r) = �V (t = 0, r) = �N2 (t = 0, r) = 0 (37)

TP (t = 0, r) = 25oC ; P (t = 0, r) = 1atm ; u (t = 0, r) = 0 (38)

When a biomass particle is decomposing, the boundary conditions are given in terms of pressure, heat and mass fluxes at the centernd surface of the particle. From symmetry at the particle center:(

∂�i/∂r)

|t, r=0

= 0 (i = G, T , V, N2) (39)(∂TP/∂r

)|t, r=0

= 0 (40)(∂P/∂r

)|t, r=0

= 0 (41)

Given reactor conditions at the particle surface:

−Deff,i.(∂�i/∂r

)|t, r=RP

= hmass,i.(

�i (t, r = RP) − �i,bulk

)(i = G, T , V, N2) (42)

−keff .(∂TP/∂r

)|t, r=RP

= hconv. (TP (t, r = RP) − Tbulk) +ωP.�.(

TP4 (t, r = RP) − T4

wall

)+qflux (43)

P (t, r = RP) = 1 atm (44)

here hmass,i and �i,bulk are the mass transfer coefficient and bulk density of the ith species inside the reactor, hconv is the convective heatransfer coefficient, Tbulk the bulk gas temperature inside the reactor, Twall reactor wall temperature and qflux is the heat flux from a heatource, such as a furnace or arc lamp (not considered in our case). As a first approximation, the bulk densities of gas phase species (except forhe fluidizing medium N2) are assumed to be negligible compared to species densities at the surface of the particles i.e. �i,bulk � �i (t, r = RP)nd the reactor wall temperature is approximated the same as the bulk gas temperature i.e. Twall

∼= Tbulk. The convective heat transferoefficient between the bulk gaseous phase and the biomass particle in the dense bed is calculated using correlations for forced convectionround submerged objects [37]. Agarwal et al. [38] provided an equation to calculate the convection coefficient in the dense bed forevolatilization of spherical coal particles (diameter dp) in fluidized beds:

Nu = hconv.dp/kg ={

0.03.Re1.3p Rep < 100

2 + 0.6.Pr1/3.Re1/2p Rep > 100

(45)

here the Reynolds and Prandtl numbers are based on bulk gas physical properties:

Rep = �bulk.Ubulk.dp/�bulk (46)

Pr = �bulk.CP bulk/kbulk (47)

Through analogies, mass transfer coefficients for each species can be approximated from the Sherwood number and Frössling equation:

Sh = hmass,i.dp/Deff,i = 2 + 0.6.Sc1/3.Re1/2p (48)

here the Schmidt number is:

Sc = �bulk/(

�bulk.Deff,i

)(49)

The thermo-physical and transport properties of the biomass, final char and gas-phase species are summarized in Table 2. For a givenet of process conditions, the final char density is calculated iteratively from the particle model. Koch [39] and Grønli et al. [35] correlated

he specific heat of pine and spruce as linear functions of temperature. Consistent with [35], we use the data from [40] to correlate theeat capacity of the solid and gas species and dynamic viscosity of the gas phase species as functions of temperature. Constant moleculareight, thermal conductivity and molecular diffusivity of the gas phase are assumed with values given in Table 2. More accurate estimation

f thermo-physical properties is possible after determining the compositions of tar and gas from elemental balances.

220 B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229

Table 2Thermo-physical and transport properties of gas and solid species.

Property Ref. Property Ref.

Cp,Pine

[J/kg.K

]= −91.2 + 4.4 × TP [K] [35] MWG = 38 g/mole [35]

Cp,Spruce

[J/kg.K

]= 1500 + TP MWT = 110 g/mole

Cp,Cf

[J/kg.K

]= 420 + 2.09 × TP − 6.85 × 10−4 × TP

2 �G

[kg/m.s

]= 7.85 × 10−6 + 3.18 × 10−8 × TP [35]

Cp,G

[J/kg.K

]= 770 + 0.629 × TP − 1.91 × 10−4 × TP

2 �T

[kg/m.s

]= −3.73 × 10−7 + 2.62 × 10−8 × TP

Cp,T

[J/kg.K

]= −100 + 4.4 × TP − 1.57 × 10−3 × TP

2 �V

[kg/m.s

]= −1.47 × 10−6 + 3.78 × 10−8 × TP

Cp,V

[J/kg.K

]= 1670 + 0.64 × TP �air

[kg/m.s

]= 9.12 × 10−6 + 3.27 × 10−8 × TP

Cp,air

[J/kg.K

]= 950 + 0.188 × TP ˇg,B0,along = 10−11m2 [36]

dpore,B0 = 5 × 10−5m [35] ˇg,B0,across = 10−14m2

dpore,Cf = 10−4m ˇg,Cf,along = 5 × 10−11m2

Dg = 10−6m2/s [23] ˇg,Cf,across = 5 × 10−12m2

kB0,along = 0.255W/m.K [36] εB0 = 0.4 [21]kB0,across = 0.1046W/m.K εCf = 0.91k = 0.105W/m.K �B0 = 570 kg/m3 [41]

at

2

sb

wopfl

w

i

ta

w

Cf,along

kCf,across = 0.071W/m.K ωB0 = 0.6 [34]kg = 2.577 × 10−2W/m.K ωCf = 1

The mass balance equations are solved using the MTALAB PDE solver (PDEPE) which solves initial-boundary value problems of parabolicnd elliptic type with one space and one time variable. Based on a specified grid size, the solver converts the PDEs to ODEs and, for a givenime step, the ODEs are integrated with respect to time, using the MATLAB differential-algebraic equation solver.

.2. Coupled reactor and particle model

In the present work, the gas and particle flows are treated separately. While perfect mixing provides a reasonable representation ofolid mixing inside a bubbling fluidized bed reactor, gas flow can be close to plug flow. A solid particle is an aggregate that behaves as aatch reactor inside the fluidized bed reactor. Let us define the conversion of a single biomass particle of unchanging size as:

XP = ((mB0 + mM0) − (mB + mM + mC )) /(

mB0 + mM0 − mCf

)(50)

here mB0 and mM0 denote the initial biomass and moisture contents of un-reacted particles, mB, mM and mC are the instantaneous massesf biomass, moisture and char in the solid particle, and mCf is mass of final char in the completely-converted particle. During drying andyrolysis of a biomass particle (radius RP), some water vapor, gas and tar accumulate within the pores of the particle, whereas the restows to the reactor environment at the particle surface. From conservation of mass over a single particle at any given time:

[Initial particle mass] = [Instantaneous particle mass]

+[Cumulative mass of volatiles released from the surface of the particle from time zero to “t′′] (51)

In mathematical terms:

mB0 + mM0 =∫ Rp

0

�P.4�r2dr +∫ t

0

((∂mT /∂t

)|r=Rp +

(∂mG/∂t

)|r=Rp +

(∂mV /∂t

)|r=Rp

). dt (52)

here the instantaneous particle bulk density �P is calculated from:

�P = �B + �M + �C + εP. (�T + �G + �V ) (53)

The second term on the right side of Eq. (52) is obtained from boundary conditions of PDEs describing gas-phase species mass balancesn the single particle model as:(

∂mT /∂t)

|r=RP

=(

4�R2P

).(εP.�T .u) |r=RP

;(∂mG/∂t

)|r=RP

=(

4�R2P

).(εP.�G.u) |r=RP

;(∂mV /∂t

)|r=RP

=(

4�R2P

).(εP.�V .u) |r=RP

(54)

The conversions of reactants (dry biomass and moisture) in a solid particle depends on their duration of stay in the reactor. However,he length of stay is not the same for all particles, and the average particle conversion of solid particles leaving the reactor is calculated in

manner similar to a macrofluid (non-coalescing droplets of very viscous liquids) [42] as:

XP =∞∫0

(XP) individual

particle

.E (t) dt (55)

here we utilize the residence time distribution function, E(t), for perfectly-mixed solids:

E (t) = 1�s

. exp(−t⁄�s

)(56)

e(

r

m

w

v

w

pfl

wocg

w

i

B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229 221

Given the reactor solids residence time distribution function, the change in the average mass of biomass particles inside the reactor isqual to the cumulative mass of volatiles released from the surface of the same particles into the reactor environment. Combining Eqs.52) and (55) gives:

(mB0 + mM0) −∞∫0

⎛⎝ Rp∫

0

�P.4�r2dr

⎞⎠ .E (t) dt =

∞∫0

⎛⎝ t∫

0

((∂mT /∂t

)|r=Rp

+(∂mG/∂t

)|r=Rp

+(∂mV /∂t

)|r=Rp

).dt

⎞⎠ .E (t) dt (57)

The average biomass particle density �P , as well as the biomass drying and pyrolysis product yields of different species leaving theeactor, are obtained as:

�P =∞∫0

⎛⎝ Rp∫

0

(�B + �M + �C ) .4�r2dr

⎞⎠ .E (t) dt/

(4⁄3�R3

P

)(58)

Y i =∞∫0

⎛⎝ Rp∫

0

�i.4�r2dr

⎞⎠ .E(t).dt/ (mB0 + mM0) (i = B, C, M) (59)

Y i =∞∫0

⎛⎝ Rp∫

0

(εP.�i).4�r2dr +t∫0

(∂mi

∂t)|

r=Rp

.dt

⎞⎠ .E(t).dt/ (mB0 + mM0) (i = T, G, V) (60)

The mean solids residence time inside the BFB is obtained from the mass hold-up of biomass particles inside the BFB divided by theass flow rate of particles leaving the reactor:

�s = MP/mP,out (61)

here mP,out is the mass flow rate of biomass particles leaving the reactor.

(1 − ε) .A.Lbed =(

MP/ �P

)+

(Msand/�sand

)(62)

The average bubbling bed voidage and height are obtained iteratively:For given sand inventory Msand, average voidage ε, cross-sectional area A and height Lbed of the of dense bubbling bed, the total solids

olume hold-up is:

ε =

⎛⎝Lbed∫

0

ε.dz

⎞⎠/Lbed (63)

here local bed voidage (ε) and bubble size (db) are estimated from two-phase correlations as a function of height, z, in the bed [43,44]:

ε = 1 −(

1 − εmf

)/(

1 +(

U − Umf

)/(

0.711√

g.db

))(64)

db = 0.54g−0.2(

U − Umf

)0.4(

z + 4√

A/Nor

)0.8(65)

As shown in the schematic diagram of the dense bubbling fluidized bed of Fig. 2, we assume a uniform distribution of drying and pyrolysisroducts throughout the entire dense bed height, while homogeneous tar cracking also occurs along the reactor. For one-dimensional plugow of gas phase, the mass balances at height z of the bed for a cell size of �z are:

.mT (z) − .

mT (z + �z) + YT .( .

mB,in + .mM,in

).(

�z/Lbed

)− k4.

( .mT (z) .��g

)= 0 (66)

.mG (z) − .

mG (z + �z) + YG.( .

mB,in + .mM,in

).(

�z/Lbed

)+ k4.

( .mT (z) .��g

)= 0 (67)

.mV (z) − .

mV (z + �z) + YV .( .

mB,in + .mM,in

).(

�z/Lbed

)= 0 (68)

here, as discussed above, YT ,YG and YV denote the average yields of tar, non-condensable gas, and water vapour released from the surfacef biomass particles into the reactor and, as denoted in Fig. 2, YB,YC and YM are the average yields of unreacted-biomass, char and moistureontent of solid particles exiting the reactor, respectively. In addition, the gas residence time in each cell is calculated as the volume of theas phase divided by its volumetric flow rate:

��g = (ε.A.�z) /(

QT + QG + QV + QN2

)= (ε.A.�z) / (U.A) (69)

here Qi=G, T, V, N2are the total volumetric flow rates of gas phase species.

The superficial gas velocity (U) inside the fluidized bed is related to the uniform operating conditions (i.e. T & P) of the reactor by thedeal gas law as:

U = R.T / P.A .(( .

/MW)

+( .

z /MW)

+( .

z /MW)

+( .

z /MW))

(70)

( ) ( ) mN2 N2 mT ( ) T mG ( ) G mV ( ) H2O

Substituting for the differential gas residence time from Eq. (69) and dividing both sides of Eqs. (66) to (68) by �z and letting �z → 0:

d.

mT /dz = Y T .( .

mB,in + .mM,in

)/Lbed − k4.

( .mT .ε/U

)B.C. :

.mT (z = 0) = 0 (71)

222 B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229

Fm

fl

ip

2

Ac

t

w

wcpf

c

ig. 2. Schematic diagram of a bubbling fluidized bed reactor assuming a uniform distribution of drying and pyrolysis products throughout the entire dense bed heightodeled as perfectly-mixed solids and plug flow of gases.

d.

mG/dz = YG.( .

mB,in + .mM,in

)/Lbed + k4.

( .mT .ε/U

)B.C. :

.mG (z = 0) = 0 (72)

d.

mV /dz = YV .( .

mB,in + .mM,in

)/Lbed B.C. :

.mV (z = 0) = 0 (73)

With the aid of MATLAB ODE solver (ODE45), the above ODEs are solved numerically to give the axial profiles of vapor, tar and gas massow rates along the reactor, as well as the profile of superficial gas velocity.

Note that an individual biomass particle may change its size due to shrinkage, cracking and fragmentation [20,21]. Furthermore, flu-dization of biomass particles with relatively fine silica sand particles inside the BFB may cause severe attrition, also changing the biomassarticle size distribution. In these cases, a population balance model may be required.

.3. Elemental balance on coupled particle and reactor model

In this study, an attempt is made to estimate major compounds composition in pyrolysis gas, as well as the tar elemental composition.ssuming that each lumped species is a homogeneous mixture of carbon, hydrogen and oxygen, the coupled particle and reactor model isonstrained with CHO elemental balances. An average overall mass balance on biomass particles inside the BFB gives:

1 = YB + YC + YM + YV + YT + YG (74)

For given steady-state operating conditions of the reactor, average elemental compositions are assigned to each lumped species andhe particle mass balance is broken down into three elemental balances:(

mB0/ (mB0 + mM0))

.wj,B +(

mM0/ (mB0 + mM0))

.wj,H2O = YB.wj,B + YC .wj,C +(

YM + YV

)wj,H2O + YT .wj,T + YG.wj,G

(j = C, H, O)(75)

here the elemental composition of dry-ash-free biomass (wj,B) is obtained from its ultimate analysis. For moisture/water vapour:

wC,H2O = 0, wH,H2O = 2/18, wO,H2O = 16/18 (76)

The elemental composition of char (wj,C ) is approximated from an empirical formula (CH0.2526O0.0236) from the literature [45].Solving Eq. (75) for average elemental composition of tar released from the surface of the average particle (wj,T ) gives:

wj,T =((

mB0/ (mB0 + mM0) − YB

).wj,B +

(mM0/ (mB0 + mM0) − YM − YV

).wj,H2O − YC .wj,C − YG.wj,G

)/YT

(j = C, H, O)(77)

Note that the kinetic constants for tar (bio-oil) yields reported by Di Blasi et al. [24] include both organic compounds and pyrolyticater produced from dehydration of the chemical structure of dry biomass. Although the tar derived from the pyrolysis of biomass is a

omplex mixture of low- and high-molecular-weight oxygenated compounds, such as carboxylic acids, aldehydes, ketones, alcohols andhenols [46], its dry-basis elemental composition is a very weak function of temperature [14]. As a model input parameter, we adopt theollowing empirical mass ratios for the organic constituents of tar [14]:

wC,T,organic = 1.14.wC,B, wH,T,organic = 1.13.wH,B, wO,T,organic = 1 − wC,T,organic − wH,T,organic (78)

¯

If YH2O,pyr denotes the average pyrolytic water yield released from the surface of the particles inside the reactor, the overall tar (bio-oil)omposition is related to its organic part elemental composition as:

wj,T =((

Y T − YH2O,pyr

).wj,T,organic + YH2O,pyr .wj,H2O

)/YT (j = C, H, O) (79)

B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229 223

Table 3Relative products yield from thermal cracking of tar [15,47].

i �i wi,G2

CO 0.5633 0.7222CO2 0.1109 0.1422H2 0.0173 0.0222

pci

s

(

w

ye

ty

we

c

CH4 0.0884 0.1133Tarinert 0.22 0Sum 1 1

Solving for pyrolytic water yield and cancelling YT , we obtain the following relationships for the elemental compositions of tar/bio-oil:

wC,T − wC,T,organic

wC,H2O − wC,T,organic= wH,T − wH,T,organic

wH,H2O − wH,T,organic= wO,T − wO,T,organic

wO,H2O − wO,T,organic(80)

The average elemental composition of non-condensable gas released from the surface of an average particle (wj,G) can be restricted bylacing additional constraints on its composition. It is customary to assume that non-condensable pyrolysis gas is composed of four majorompounds: H2, CO, CO2 and CH4, with other light hydrocarbons (e.g. C2-C3) lumped into methane. The average gas elemental compositions then related to its composition:

wC,G =(

12/28)

.wCO,G +(

12/44)

.wCO2,G +(

12/16)

.wCH4,G (81)

wH,G =(

4/16)

.wCH4,G + wH2,G (82)

wO,G =(

16/28)

.wCO,G +(

32/44)

.wCO2,G (83)

Experimental evidence [14,23] shows that the hydrogen content of the non-condensable gas is almost negligible. Hence, we set:

wH2,G∼= 0 (84)

Given that each compound composition in pyrolysis gas must also fall between zero and unity, the following constraints are obtained:⎡⎢⎣

0 < wH,G < 0.25

0.2727 + 1.9091.wH,G < wC,G < 0.4286 + 1.9091.wH,G

0.5714 − 2.2857.wH,G < wO,G < 0.7273 − 2.2857.wH,G

⎤⎥⎦ (85)

Simultaneous solution of Eqs. (77) and (79) subject to the above constraints gives a range of acceptable results for tar and gas compo-itions, as well as pyrolytic water yield released from the surface of the average particle at different reactor temperatures.

On the other hand, according to the widely adopted reaction stoichiometry of Boronson et al., 1989 [47] for thermal cracking of tarbio-oil) in the temperature range 500–800 ◦C:

Tar → �COCO + �CO2 CO2 + �H2 H2 + �CH4 CH4 + �TarinertTarinert (86)

here the mass-based stoichiometric coefficients (�i) are listed in Table 3.As the amount of tar (bio-oil) cracked is known from reaction kinetics, the yield of inert tar is assigned a zero value and the associated

ield is loaded on the remaining gases [15]. Therefore, the compositions of the pyrolysis gas generated from tar thermal cracking arestimated:

wCO,G2 = �CO/(

�CO + �CO2 + �H2 + �CH4

)= 0.7222 (87)

wCO2,G2 = �CO2 /(

�CO + �CO2 + �H2 + �CH4

)= 0.1422 (88)

wH2,G2 = �H2 /(

�CO + �CO2 + �H2 + �CH4

)= 0.0222 (89)

wCH4,G2 = �CH4 /(

�CO + �CO2 + �H2 + �CH4

)= 0.1133 (90)

As illustrated in Fig. 2, the tar (bio-oil) yield released from the surface of the average particle into the reactor is thermally cracked alonghe reactor. Therefore, the decrease in tar (bio-oil) yield due to homogeneous tar cracking must be equal to the increase in pyrolysis gasield as:

YT .( .

mB,in + .mM,in

)− .

mT,out = .mG,out − YG.

( .mB,in + .

mM,in

)(91)

here.

mT,out and.

mG,out are the mass flow rates of tar (bio-oil) and dry-nitrogen-free pyrolysis gas leaving the reactor. Breaking down thisquation to CHO elemental balances:

YT .( .

mB,in + .mM,in

).wj,T − .

mT,out.wj,T,out = .mG,out.wj,G,out − YG.

( .mB,in + .

mM,in

).wj,G =

( .mG,out − YG.

( .mB,in + .

mM,in

)).wj,G2

(j = C, H, O)(92)

Given the average elemental composition of tar and gas released from the surface of the average particle (wj,T ) and compounds

omposition in secondary pyrolysis gas (wj,G2

), the composition of the pyrolysis gas leaving the reactor is estimated as:

wj,G,out =YG.

( .mB,in + .

mM,in

).wj,G +

( .mG,out − YG.

( .mB,in + .

mM,in

)).wj,G2

.mG,out

(j = C, H, O) (93)

224 B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229

Solving for compounds composition:⎡⎢⎣

wCO,G,out

wCO2,G,out

wCH4,G,out

⎤⎥⎦ =

⎡⎣ 12⁄28 12⁄44 12⁄16

0 0 4⁄1616⁄28 32⁄44 0

⎤⎦

−1

.

⎡⎢⎣

wC,G,out

wH,G,out − wH2,G,out

wO,G,out

⎤⎥⎦ (94)

3. Results and discussion

The following results have been generated for the single particle model and the coupled particle and reactor model. We test the powerof the coupled particle and reactor model by comparing its predictions with an independent set of experimental data from the literature[41].

3.1. Predictions of single particle model

For the single particle model, drying and pyrolysis of spherical biomass particles of two different sizes are studied at two differentreactor temperatures in an atmospheric bubbling fluidized bed of silica sand, with nitrogen as the fluidizing gas. For spherical biomassparticles of 5 mm and 500 �m diameters and a reactor temperature of 600 ◦C, the predicted change in particle conversion and temperatureupon entering the reactor are illustrated in Figs. 3 and 4 as functions of time and radial position inside the biomass particle. As seen, thepredicted effect of particle diameter on the biomass pyrolysis rate is significant. The time required for complete conversion and/or heat-upto reactor temperature of a single particle is predicted to be approximately one minute for a 5 mm diameter particle and less than 5 s for abiomass particle of 500 �m diameter.

Figs. 5 and 6 plot the total particle bulk density (sum of un-reacted biomass, bound moisture and char) and volatiles density (sum oftar, pyrolysis gas and water vapour) as functions of time and radial position inside the biomass particle, respectively. While the first peakof volatiles concentration inside the particle is due to accumulation of water vapour from drying, the second peak originates from theaccumulation of volatiles generated by primary and secondary pyrolysis. Due to evolution of volatiles, a negative pressure gradient forms

Fig. 3. Single spherical biomass particle conversion predicted as a function of time and radial position inside particle: (a) dp = 5 mm, (b) dp = 500 �m, Reactor tempera-ture = 600 ◦C, biomass moisture content = 9 wt%.

Fig. 4. Single spherical biomass particle temperature predicted as a function of time and radial position inside particle: (a) dp = 5 mm, (b) dp = 500 �m, Reactor tempera-ture = 600 ◦C, biomass moisture content = 9 wt%.

B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229 225

Fig. 5. Biomass particle density as a function of time and radial position inside spherical particle: (a) dp = 5 mm, (b) dp = 500 �m, Reactor temperature = 600 ◦C, biomassmoisture content = 9 wt%.

Fc

isi

tptit

3

mbt1

etoadto

pm

ig. 6. Volatiles density as a function of time and radial position inside spherical particle: (a) dp = 5 mm, (b) dp = 500 �m, reactor temperature = 600 ◦C, biomass moistureontent = 9 wt%.

nside the biomass particle, but it disappears with time due to the release of volatiles at the particle surface. Comparing Fig. 6(a) and (b), theecond peak of volatiles density is considerably greater for the larger particle, demonstrating greater accumulation of pyrolysis productsnside the larger particle due to higher internal mass transfer resistances.

In Fig. 7, the predicted drying and pyrolysis product yields are compared for the particle diameters 5 mm and 500 �m, and two reactoremperatures, 600 ◦C and 700 ◦C. As expected, smaller particles and higher reactor temperatures lead to much faster conversion of a biomassarticle. However, for long enough reaction times, the final product yields do not change appreciably with either particle size, or reactoremperature. Therefore, for a reactor with large enough mean solids residence time (longer than the time for complete conversion ofndividual particles), most biomass particles are completely converted, and the product distribution from the reactor becomes insensitiveo biomass particle size distribution and reactor temperature.

.2. Model verification with experimental data

In this section, we test the accuracy of the coupled particle and reactor model by comparing its predictions with independent experi-ental data from the literature. Westerhof et al. [41] studied the effect of temperature on fast pyrolysis of pine in a continuous bench-scale

ubbling fluidized bed, with silica sand as the fluidization medium and nitrogen as the fluidizing gas. The properties of pine and experimen-al operating conditions are summarized in Tables 4 and 5, respectively. Good reproducibility of the experiments with standard deviations.0, 0.9, 1.3 and 1.0 wt% are reported for char, gas, tar and water yields, respectively [41].

As seen in Fig. 8, the pyrolysis product yields predicted by the two-step kinetic model (dashed curves) are in better agreement with thexperimental data than those from the solid-state kinetic mechanism (solid curves). In the latter case, the char yields are under-predicted (athe expense of over-prediction of tar) at temperatures below 400 ◦C where primary pyrolysis is dominant. The limitations in the predictionsf the solid-state kinetic mechanism may be associated with the two adjustable parameters adopted to develop a complete model in thebsence of experimental analyses on pyrolysis products distribution. Discrepancies between the models’ predictions and experimentalata could also originate from the adopted pyrolysis kinetic parameters and/or reaction mechanisms to secondary tar cracking reactionshat are likely to occur in cyclones and other downstream equipment. In addition, the catalytic effect of ash on wood reactivity, as well as

n tar cracking, could lead to smaller tar yields than predicted by the model.

Furthermore, as discussed above, under given steady-state operating conditions [41], the CHO elemental balances on the coupledarticle and reactor model constrained by major compounds composition in pyrolysis gas are solved for all acceptable gas and tar ele-ental compositions, as well as for pyrolytic water yields at different pyrolysis temperatures. For the adopted empirical formulas of char

226 B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229

Fig. 7. Instantaneous drying and pyrolysis product yields for an individual particle: (a) dp = 500 �m, Reactor temperature = 600 ◦C, (b) dp = 5 mm, Reactor temperature = 600 ◦C,(c) dp = 5 mm, Reactor temperature = 700 ◦C, biomass moisture content = 9 wt%.

B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229 227

Table 4Properties of pine [41].

Bio-chemical Cellulose 35composition Hemicellulose 29(wt%, dry) Lignin 28

Ultimate analysis C 46.58(wt%, dry-ash-free) H 6.34

O 46.98N 0.04S 0.06

Alkali metals K 34(mg/kg, dry) Ca 134

Mg 768Total ash 2600

Biomass moisture (as received) 9–10 wt%

Biomass bulk density 570 kg/m3

Particle diameter 1 mm

Table 5Operating conditions of experimental study [41].

Biomass feed rate (as received) 1 kg/h

Initial sand bed inventory 2.1 kgAverage sand particle diameter 250 �mSand particle density 2600 kg/m3

Bed height 0.25 mColumn diameter 0.1 mSuperficial gas velocity (U) 0.14 m/sU/Umf 3.5Gas/vapor residence time of the bed 0.8 sGas/vapor residence time of the cyclones 0.45-0.55 sBiomass particles mean residence time 20–25 minFreeboard temperature 330–580 ◦CGas/vapor in the freeboard outlet/cyclones 400–540 ◦C

Fig. 8. Pyrolysis product yields at reactor exit per mass of dry biomass as a function of reactor temperature: Dashed and solid curves are for the two-step kinetic model andsolid-state kinetic mechanism, respectively. Experimental data points are from Westerhof, et al. [41].

(padftieo

(cpb

CH0.25O0.024) and organic tar (CH1.62O0.56) as model input parameters, the predicted compounds composition in dry and nitrogen-freeyrolysis gas, as well as pyrolytic water yield (per mass of dry biomass), are shown at different reactor temperatures in Fig. 9(a), (b), (c)nd (d) where the dashed and solid curves are for the two-step and solid-state kinetic mechanisms, respectively, and the experimentalata points are from reference [41]. At temperatures above 400 ◦C, major compounds composition in pyrolysis gas are fairly estimatedrom the average elemental compositions based on Eq. (93), and secondary pyrolysis reaction stoichiometry (Eqs. (95)–(98)). However, atemperatures below 400 ◦C, where the CHO elemental balances are weakly constrained by secondary pyrolysis reaction, large variationsn the carbon/oxygen content of pyrolysis gas make it difficult to accurately predict the CO/CO2 ratio without introduction of an additionalxperimental correlation, e.g. [11,14]. Nevertheless, according to experimental data [11,13,23], CO2 is reported to be the major componentf primary pyrolysis gas at temperatures below 400 ◦C.

Note that our model is generic and can handle a variety of biomass species, as long as biomass-specific reaction kinetics parametersTable 1) and tar and char ultimate analyses corresponding to that type of biomass are used as model input parameters. As seen in Fig. 10, thehoice of biomass-specific kinetic parameters significantly affects the model predictions. Sensitivity analyses show that biomass thermo-hysical properties such as conductivity, heat capacity, emissivity, etc. can also somewhat affect the pyrolysis product yield distribution,ut not as significantly as the kinetic parameters.

228 B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229

Fig. 9. Predicted composition of major compounds in dry and nitrogen-free pyrolysis gas and pyrolytic water yield (per mass of dry biomass) as functions of reactortemperature for empirical dry tar (CH1.62O0.56) and char (CH0.25O0.024) formulae. Dashed and solid curves are for the two-step kinetic model and solid-state kinetic mechanism,respectively. Experimental data points are from Westerhof, et al. [41].

Fp

4

sadsa

aagod

p

ig. 10. Pyrolysis product yields at reactor exit per mass of dry biomass as a function of reactor temperature: Dashed and solid curves are for the primary pyrolysis kineticarameters from Di Blasi, et al. [24] and Chan, et al. [23], respectively. Experimental data points are from Westerhof, et al. [41].

. Conclusions

A comprehensive coupled particle and reactor model is developed to predict biomass drying and pyrolysis products yield and compo-ition generated by bubbling fluidized bed reactor under steady-state operating conditions. The time required for complete conversionnd heat-up of a biomass particle to reactor temperature is predicted to decrease significantly with increasing reactor temperature and/orecreasing biomass particle size. Due to evolution and accumulation of volatiles, a negative pressure gradient forms inside the poroustructure of biomass particles which disappears over time with the release of volatiles at the particle surface. For larger particles, greaterccumulation of pyrolysis products is predicted, related to increased internal mass transfer resistances.

The coupled particle and reactor model predictions, based on two different biomass pyrolysis kinetic schemes (a two-step kinetic modelnd a solid-state kinetic mechanism for multiple reactions), are shown to be consistent with available experimental data. With organic tarnd char elemental compositions as model input parameters, elemental balances lead to predictions of the composition of the pyrolysisas at the reactor exit as a function of process operating parameters feedstock ultimate analysis. The model gives reasonable predictionsf the composition of pyrolysis gas at temperatures above 400 ◦C. With increasing reactor temperature, CO and CH4 mass fractions in

ry-nitrogen-free non-condensable gas increase, while the CO2 mass fraction decreases.

Limitations of the current model in predicting pyrolysis products distribution can be largely attributed to inaccurate or incompleteyrolysis kinetics for different types of biomass with varying properties.

A

a

R

[[

[[[[[[[

[[[

[[

[[[

[[[[

[[

[[[

[[[[[[[[[[[[

B. Hejazi et al. / Journal of Analytical and Applied Pyrolysis 121 (2016) 213–229 229

cknowledgements

Financial support from BioFuelNet Canada, a network focusing on the development of advanced biofuels and associated bioproductsnd a member of the Networks of Centres of Excellence of Canada program (www.biofuelnet.ca), is gratefully acknowledged.

eferences

[1] P.A. Brownsort, Biomass Pyrolysis Processes: Review of Scope, Control and Variability, M. Sc. Thesis, UK Biochar Research Centre, 2009.[2] P. Ammendola, R. Chirone, F. Miccio, G. Ruoppolo, F. Scala, Devolatilization and attrition behavior of fuel pellets during fluidized-bed gasification, Energy Fuels 25

(2011) 1260–1266.[3] A. Gómez-Barea, B. Leckner, Modeling of biomass gasification in fluidized bed, Prog. Energy Combust. Sci. 36 (2010) 444–509.[4] A. Ergudenler, A.E. Ghaly, F. Hamdullahpur, A.M. Al-Taweel, Mathematical modeling of a fluidized bed gasifier. Part I—model development, Energy Source 19 (1997)

1065–1084.[5] S. Sadaka, A.E. Ghaly, M.A. Sabbah, Two phase biomass air–steam gasification model for fluidized bed reactors: part I—model development, Biomass Bioenergy 22

(2002) 439–462.[6] A.G. Buekens, J.G. Schoeters, et al., Modelling of biomass gasification, in: Overend (Ed.), Fundamentals of Thermochemical Biomass Conversion, Elsevier Applied

Science Publishers, London, 1985, pp. 619–689.[7] D. Fiaschi, M. Michelini, A two-phase one-dimensional biomass gasification kinetics model, Biomass Bioenergy 21 (2001) 121–132.[8] P. Lü, X. Kong, C. Wu, Z. Yuan, L. Ma, J. Chang, Modeling and simulation of biomass air-steam gasification in a fluidized bed, Front. Chem. Eng. Chin. 2 (2008) 209–213.[9] P. Ji, W. Feng, B. Chen, Comprehensive simulation of an intensified process for H2 production from steam gasification of biomass, Ind. Eng. Chem. Res. 48 (2009)

3909–3920.10] H. Jiang, R.V. Morey, A numerical model of fluidized bed biomass gasifier, Biomass Bioenergy 3 (1992) 431–447.11] H. Thunman, F. Niklasson, F. Johnsson, B. Leckner, Composition of volatile gases and thermo-chemical properties of wood for modeling of fixed or fluidized beds,

Energy Fuels 15 (2001) 1488–1497.12] C. Di Blasi, Dynamic behaviour of stratified downdraft gasifiers, Chem. Eng. Sci. 55 (2000) 2931–2944.13] L. Fagbemi, L. Khezami, R. Capart, Pyrolysis products from different biomasses: application to the thermal cracking of tar, Appl. Energy 69 (2001) 293–306.14] D. Neves, H. Thunman, A. Matos, L. Tarelho, A. Gómez-Barea, Characterization and prediction of biomass pyrolysis products, PECS 37 (2011) 611–630.15] P. Kaushal, J. Abedi, A simplified model for biomass pyrolysis in a fluidized bed reactor, JIEC 16 (2010) 748–755.16] P. Kaushal, J. Abedi, N. Mahinpey, A comprehensive mathematical model for biomass gasification in a bubbling fluidized bed reactor, Fuel 89 (2010) 3650–3661.17] R. Radmanesh, J. Chaouki, C. Guy, Biomass gasification in a bubbling fluidized bed reactor: experiments and modeling, AlChE J. 52 (2006) 4258–4272.18] L. Gerun, M. Paraschiv, R. Vijeu, J. Bellettre, M. Tazerout, B. Gøbel, U. Henriksen, Numerical investigation of the partial oxidation in a two-stage downdraft gasifier, Fuel

87 (2008) 1383–1393.19] Z.A. El-Rub, Biomass Char as an In-situ Catalyst for Tar Removal in Gasification Systems, Ph.D Thesis, University of Twente, 2008.20] B.V. Babu, A.S. Chaurasia, Heat transfer and kinetics in the pyrolysis of shrinking biomass particle, Chem. Eng. Sci. 59 (2004) 1999–2012.21] A.K. Sharma, V. Pareek, S. Wanga, Z. Zhangb, H. Yangb, D. Zhangba, A phenomenological model of the mechanisms of lignocellulosic biomass pyrolysis processes,

Comput. Chem. Eng. 60 (2014) 231–241.22] F. Shafizadeh, P.P.S. Chin, Wood technology, in: I.S. Goldstein (Ed.), Chemical Aspects, American Chemical Society Symposium Series, 1977, pp. 57–81.23] W.C.R. Chan, M. Kelbon, B.B. Krieger, Modelling and experimental verification of physical and chemical processes during pyrolysis of a large biomass particle, Fuel 64

(1985) 1505–1513.24] C. Di Blasi, C. Branca, Kinetics of primary product formation from wood pyrolysis, Ind. Eng. Chem. Res. 40 (2001) 5547–5556.25] A.G. Liden, F. Berruti, D.S. Scott, A kinetic model for the production of liquids from the flash pyrolysis of biomass, Chem. Eng. Commun. 65 (1988) 207–221.26] W.C. Park, A. Atreya, H.R. Baum, Experimental and theoretical investigation of heat and mass transfer processes during wood pyrolysis, Combust. Flame 157 (2010)

481–494.27] C. Di Blasi, Analysis of convection and secondary reaction effects within porous solid fuels undergoing pyrolysis, Combust. Sci. Techol. 90 (1993) 315–340.28] K.M. Bryden, M.J. Hagge, Modeling the combined impact of moisture and char shrinkage on the pyrolysis of a biomass particle, Fuel 82 (2003) 1633–1644.29] Y. Haseli, J.A. van Oijen, L.P.H. de Goey, Modeling biomass particle pyrolysis with temperature-dependent heat of reactions, J. Anal. Appl. Pyrol. 90 (2011) 140–154.30] X. Wang, S.R.A. Kersten, W. Prins, W.P.M. van Swaaij, Biomass pyrolysis in a fluidized bed reactor. Part 2: experimental validation of model results, Ind. Eng. Chem. Res.

44 (2005) 8786–8795.31] D. Mochulski, Multiple Reaction Solid State Kinetic Parameter Determination and Its Application to Woody Biomass, M. Sc. Thesis, University of British Columbia, 2015.32] B.M. Wagenaar, W. Prins, W.P.M. van Swaaij, Pyrolysis of biomass in the rotating cone reactor: modelling and experimental justification, Chem. Eng. Sci. 49 (1994)

5109–5126.33] A.M.C. Janse, R.W.J. Westerhout, W. Prins, Modelling of flash pyrolysis of a single wood particle, Chem. Eng. Proc. 39 (2000) 239–252.34] C. Branca, C. Di Blasi, Kinetics of the isothermal degradation of wood in the temperature range 528–708 K, J, Anal. Appl. Pyrol. 67 (2003) 207–219.35] M.G. Grønli, M.C. Melaaen, Mathematical model for wood pyrolysis: comparison of experimental measurements with model predictions, Energy Fuels 14 (2000)

791–800.36] C. Di Blasi, Physico-chemical processes occurring inside a degrading two-dimensional anisotropic porous medium, Int. J. Heat Mass Transfer 41 (1998) 4139–4150.37] W.E. Ranz, W.R. Marshall, Evaporation from drops parts 1 & 2, Chem. Eng. Prog. 48 (1952) 141–146, & 173–180.38] P.K. Agarwal, W.E. Genetti, Y.Y. Lee, Models for devolatilization of coal particles in fluidized beds, Fuel 63 (1984) 1157–1165.39] P. Koch, Specific heat of oven-dry spruce pine wood and bark, Wood Sci. 1 (1969) 203–214.40] K. Raznjevik, Handbook of Thermodynamic Tables and Charts, Hemisphere publishing corporation, McGraw-Hill Book, 1976.41] R.J.M. Westerhof, Refining Fast Pyrolysis of Biomass, University of Twente, Ph.D Thesis, 2011.42] O. Levenspiel, Chemical Reaction Engineering, third ed., John Wiley and Sons, New York, 1999.

43] R. Clift, J.R. Grace, Continuous bubbling and slugging, in: J.F. Davidson, R. Clift, D. Harrison (Eds.), Fluidization, Academic Press, New York, 1985, pp. 73–132.44] R.C. Darton, R.D. Lanauze, J.F. Davidson, D. Harrison, Bubble growth due to coalescence in fluidized-beds, Trans. Inst. Chem. Eng. 55 (1977) 274–280.45] C. Di Blasi, Modeling wood gasification in a countercurrent fixed-bed reactor, AIChE J. 50 (2004) 2306–2319.46] A. Oasmaa, D. Meier, Fast Pyrolysis of Biomass: A Handbook, in: A.V. Bridgwater (Ed.), CPL Press, 2002, p. 23.47] M.L. Boroson, J.B. Howard, J.P. Longwell, W.A. Peters, Product yields and kinetics from the vapor phase cracking of wood pyrolysis tars, AIChE J. 35 (1989) 120–128.