Bioresource Technology 101 (2010) 41694178

Contents lists available at ScienceDirect

Bioresource Technology
journal homepage: www.elsevier.com/locate/biortech

Co-ring straw with coal in a swirl-stabilized dual-feed burner: Modelling

and experimental validation
Chungen Yin a,*, Sren K. Kr a, Lasse Rosendahl a, Sren L. Hvid b
a
b

Department of Energy Technology, Aalborg University, 9220 Aalborg East, Denmark

DONG Energy, Kraftvrksvej 53, 7000 Fredericia, Denmark

a r t i c l e

i n f o

Article history:
Received 9 October 2009
Received in revised form 7 January 2010
Accepted 9 January 2010
Available online 1 February 2010
Keywords:
Biomass
Co-combustion
Low-NOx burner
CFD
Swirl

a b s t r a c t
This paper presents a comprehensive computational uid dynamics (CFD) modelling study of co-ring
wheat straw with coal in a 150 kW swirl-stabilized dual-feed burner ow reactor, in which the pulverized straw particles (mean diameter of 451 lm) and coal particles (mean diameter of 110.4 lm) are independently fed into the burner through two concentric injection tubes, i.e., the centre and annular tubes,
respectively. Multiple simulations are performed, using three meshes, two global reaction mechanisms
for homogeneous combustion, two turbulent combustion models, and two models for fuel particle conversion. It is found that for pulverized biomass particles of a few hundred microns in diameter the intraparticle heat and mass transfer is a secondary issue at most in their conversion, and the global four-step
mechanism of Jones and Lindstedt may be better used in modelling volatiles combustion. The baseline
CFD models show a good agreement with the measured maps of main species in the reactor. The straw
particles, less affected by the swirling secondary air jet due to the large fuel/air jet momentum and large
particle response time, travels in a nearly straight line and penetrate through the oxygen-lean core zone;
whilst the coal particles are signicantly affected by secondary air jet and swirled into the oxygen-rich
outer radius with increased residence time (in average, 8.1 s for coal particles vs. 5.2 s for straw particles
in the 3 m high reactor). Therefore, a remarkable difference in the overall burnout of the two fuels is predicted: about 93% for coal char vs. 73% for straw char. As the conclusion, a reliable modelling methodology for pulverized biomass/coal co-ring and some useful co-ring design considerations are suggested.
2010 Elsevier Ltd. All rights reserved.

1. Introduction
The worldwide concern with the limited availability of fossil
fuels and global warming due to the emission of CO2 and other
greenhouse gases has spurred interest in using biomass for energy
production (Yin et al., 2008). Co-ring biomass with coal in existing coal-red power plants represents an attractive option, to
achieve pressing near-term targets for signicant increase of the
share of renewable energy sources in energy system and reduction
in CO2 emissions. Co-ring demonstrations have been performed
in about 150 power plants globally over the past decade, either
as pilot tests or in commercial use, which provide a valuable collection of information that spans the range of major coal combustion
technologies, important fuel types, feeding congurations, and coring levels. The results show that there are no major technical
obstacles to implement co-ring. The potential problems appear
manageable with judicious consideration of fuels, design and operating conditions of burners and boilers (McKendry, 2002; Robinson
et al., 2003; Baxter, 2005; Hansson et al., 2009).
* Corresponding author. Tel.: +45 30622577.
E-mail address: chy@iet.aau.dk (C. Yin).
0960-8524/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2010.01.018

Among the different combustion technologies, integration of

biomass ring system into a wall-red boiler could be far more
challenging, because burner aerodynamics and fuel properties
have a much greater impact on combustion and emission characteristics in a suspension-red boiler than in a grate-red boiler
or a uidized bed combustor and there are many different burner
designs on the market (Laux et al., 2002; Yin et al., 2004, 2008).
Fundamental combustion studies must be performed in order to
determine combustion behaviour characteristics in controlled laboratory settings and nally to aid in the design and optimization of
the co-ring burners and boilers (Sami et al., 2001). The body of
work available in literature for the fundamental characteristics of
biomass/coal co-ring ames and the interpretation is still severely limited (e.g., Abbas et al., 1994; Spliethoff and Hein, 1998;
Sami et al., 2001; Damstedt et al., 2007; Damstedt, 2007; Chao
et al., 2008; Lu et al., 2008a; Molcan et al., 2009). Valuable knowledge on the effects of biomass addition (including the type, share,
particle size and injection mode of biomass) and burner design and
operation mode on ame characteristics (e.g., ignition, structure,
brightness, oscillation and stability) and emissions are reported.
For instance, bituminous coal and pine sawdust co-ring in a
well-controlled, 0.5 MW down-red dual-feed burner reactor were

4170

C. Yin et al. / Bioresource Technology 101 (2010) 41694178

investigated and it was found that co-ring ratio and fuel injection
mode had signicant effect on ame ignition, combustion aerodynamics, and NOx emissions (Abbas et al., 1994). Comprehensive
experimental studies on coal, biomass and co-ring ames in a
150 kW down-red dual-feed burner ow reactor were conducted
to look into biomass co-ring impacts on ame structure and emissions (Damstedt et al., 2007; Damstedt, 2007).
CFD has been successfully used as a powerful tool to study pulverized biomass/coal co-ring in the past years (e.g., Laux et al.,
2002; Gera et al., 2002; Backreedy et al., 2005; Ma et al., 2007,
2009; Pallars et al., 2009), with focus on different aspects, e.g., effect of particle shape on motion and conversion of biomass particles, new biomass char oxidation model, carbon burnout of large
biomass particles and burner design considerations. In this study,
the straw/coal co-ring ame (Damstedt, 2007) is investigated
numerically, in which the effects of meshes, reaction mechanisms
for homogeneous combustion, turbulent combustion models, intra-particle heat and mass transfer are carefully examined. The
aim of this study is to develop a reliable modelling methodology
for biomass/coal co-ring, to allow an in-depth analysis of co-ring
ames and nally to aid in the design and operation of swirl-stabilized biomass/coal co-red burners.
2. Methods

industrial co-ring pulverized-fuel burners such as those on the

Studstrup station in Denmark and other research burners (Abbas
et al., 1994; Spliethoff and Hein, 1998; Sami et al., 2001), is placed
on the top of the reactor and res the fuels downward into the
reactor. The combustion products pass through an ash separator
and scrubber before being vented into the air.
Table 1 lists the main properties and kinetic data of the fuels
and co-ring condition. In all the nal CFD simulations, 10 groups
of particle sizes are used for both the fuels. For coal particles, the
10 different sizes are 25, 44.4, 63.9, 83.3, 103, 122, 142, 161, 181
and 200 lm. For straw particles, the 10 particle sizes are 50, 156,
261, 367, 472, 578, 683, 789, 894 and 1000 lm. The mass fraction
of each size group can be calculated by using Eq. (1) from the measured RosinRammler size distribution parameters, i.e., mean
diameter dp and spread parameter (n) as given in Table 1,

h
i
Y d exp dp =dp n

where Yd is the mass fraction of particles with size larger than dp.
The swirl number in Table 1, S, dened as the ratio of the axial ux
of angular momentum to the axial ux of axial momentum, is determined using Eq. (2) from the measured torque on a ow straightener mounted in the funnel between swirl block outlet and quarl
inlet (Hvid, 2006),

_2
S  Torque  qpRo Ri =m
2.1. Experimental
Under the same framework of this study, all the experimental
works were performed by Brigham Young University (located in
Provo, Utah, United States) in their 150 kW swirl-stabilized dualfeed burner ow reactor (Damstedt et al., 2007; Damstedt, 2007).
The main measuring effort was to produce high quality, quantitative maps of gas species in the ames. Temperature, velocity and
unburnt carbon in ash were not measured. Here only a very brief
introduction of the experimental efforts is given. Fig. 1 shows a
scaled cross-section view of the inner chamber of the reactor, indicating locations of the burner quarl and access windows, measuring grids, and (x, r) coordinate in CFD simulations. The dual-feed
swirl burner (also shown in Fig. 1), which resembles in design

where the torque is directly measured; Ro and Ri represent the outer

_ is the
and inner radius of the secondary air funnel, respectively; m
mass ow rate of the secondary air.
2.2. CFD modelling
When pulverized coal and biomass particles are injected into
the burner ow reactor, they will travel through gas and interact
with gas in the reactor until they reach the outlet. The different
particle streams can be tracked by solving the equation of motion
for particles. Along their trajectories, the fuel particles will experience different conversion processes (i.e., heating up and drying,
devolatilization and char oxidation), creating sources for reaction

Fig. 1. A scaled view of the reactor interior (left) and the burner cross-section (right).

4171

C. Yin et al. / Bioresource Technology 101 (2010) 41694178

Table 1
The wheat straw and coal co-red and the co-ring conditions.
Moisture [%, AR]

Volatiles [%, D]

FC [%, D]

Proximate analysis data (the bases: AR = as received; D = dry; DAF = dry ash-free; % on mass)
Coal
2.1
40.6
51.5
Straw
7.7
79.5
15.6

Ultimate analysis data

Coal
Straw

HHV [kJ/kg, D]

7.89
4.91

30,731
18,493

C [wt%, dry]

H [wt%, dry]

O [wt%, dry]

N [wt%, dry]

S [kJ/kg, dry]

74.8
47.3

5.08
5.68

10.1
41.6

1.53
0.54

0.58
<0.01

Min. dp [lm]

Max. dp [lm]

dp [lm]

n []

qp [kg/m3]

110.4
451.0

4.40
2.31

1400
600

The measured RosinRammler particle size distribution, and fuel particle density
Coal
25
200
Straw
50
1000
Pre-exponential factor, A [s1]

Activation energy, E [J/kmol]

Kinetic parameters, A and E, used in single kinetic rate devolatilization model

Coal
3.12  105
Straw
1.56  1010
Central fuel (straw) [kg/h]

Ash [%, D]

Annular fuel (coal) [kg/h]

Operational parameters of straw/coal co-ring

15.1
7.5

7.4  107
1.38  108

Central air [kg/h]

Annular air [kg/h]

Secondary air [kg/h]

Secondary air swirl number

12

160

1.0

in gas phase. Modelling of the particle conversion processes is to

update the mass (or composition), temperature, density and size
of the particles along their trajectories, by solving particle heat balance equation and particle conversion rate models. All the simulations are performed by using Fluent v.6.3.26 (Fluent, 2006), in
which fundamental equations are clearly documented. As a result,
only the user dened sub-models or reactions are introduced in
some details below.
2.2.1. Particle motion
Both the coal and biomass particles are assumed to be spherical
in this study. Non-spherical properties of the fuel particles are not
available during the experiments. Moreover, non-spherical effects
are negligible for particles of a few hundred microns in size in
the rst a few meters downstream of a burner quarl (Yin et al.,
2004), though these effects make a big difference in which particles
are carried up though a boiler. As a result, only drag and gravity are
retained in the equation of motion for particles to track the particles in all the simulations.
2.2.2. Particle conversion
The different conversion processes may occur sequentially or
simultaneously, as sketched in Fig. 2a, depending on fuel particle
properties and reactor conditions. By default in Fluent, the pro-

cesses are assumed to occur sequentially and solved step-by-step

by using discrete phase model (DPM) laws to update the mass,
temperature, density and size of the particles along their trajectories (Fluent, 2006). The kinetic parameters in Table 1, A and E, used
in the single kinetic rate devolatilization model for the straw particles, are from the work on the same fuel (Zhou et al., 2005).
To look into the effect of intra-particle heat and mass transfer
on conversion of large biomass particles, a 1D particle conversion
model is developed in this study, which resembles in principle
the works in literature, e.g., Johansson et al. (2007) and Lu
et al. (2008b). The large biomass particle is discretized into a
number of shells (i.e., control volumes, or cells), e.g., in the
way shown in Fig. 2b. For each of the control volumes, the mass,
energy and species balance equations for gas and solid phases
are solved, with necessary process rate equations and empirical
correlations used for the closure of the balance equations. Here,
only one issue is highlighted in order to assure the stability of
such a particle model. The similar problem also exists in modelling the conversion of a packed bed of fuels, e.g., the fuel bed on
the grate in a grate-red boiler (Yin et al., 2008). Here, the solidphase energy equation is used as an example to illustrate this issue. The discretized solid-phase energy equation on the outermost control volume, using fully implicit scheme, can be
expressed as,

Fig. 2. Particle conversion stages and discretization of a particle.

4172

C. Yin et al. / Bioresource Technology 101 (2010) 41694178

q C p DV
Dt

k Aw
T P  T W
Dr
hAe T g  T P Ae erh4R  T 4P Q_ Internal
|{z} |{z}

T P  T P 

convection

radiation

where q, Cp, k and e are all physical properties, representing the

density, specic heat, conductivity and emissivity of the solid in
the control volume, respectively. DV, Aw, Ae, Dr denote the volume,
area of the left control surface, area of the right control surface of
the control volume, and the distance between the two neighbouring
cell centres, respectively, as indicated in Fig. 2b. TP, TW, Tg and hR
represents the temperature at the centre of the cell under consideration, temperature at the neighbouring cell centre, local gas temperature, and local radiation temperature, respectively. Dt is the
time step; h is the convective heat transfer coefcient; and r is StefanBolzmann constant. The last term, Q_ Internal , represents the internal heat sources in the cell, due to, e.g., evaporation and char
oxidation. The superscript, *, is used to denote the value at the previous time step. Eq. (3) needs to be re-arranged into the standard
form, aPTP = aWTW + aETE + S. Care must be taken in all non-linear
sources, e.g., the radiation heat source in Eq. (3),

B
C
Q_ R  As er@h4R  T 4P A
|{z}

tile (CV) and straw-volatile (SV): CV  CH2.6O0.3246N0.056S0.0093

and SV  CH2.134O0.985N0.0146. The compositions and formation
enthalpies are determined from the proximate and ultimate analysis data of coal and straw, respectively. Then, the global mechanism that resembles the WD model is employed for the
oxidation of the two volatiles. In this mechanism, there are eight
species and four reactions, as listed in Table 2. The turbulencechemistry interaction is modelled by nite rate/eddy-dissipation
(Magnussen and Hjertager, 1976).
The JL 4-step global mechanisms are proposed for the combustion of alkane hydrocarbons up to butane (Jones and Lindstedt,
1988). In this mechanism, the two volatiles are considered as a
mixture of real species, as shown in Eqs. (7) and (8), respectively.
The split ratios of different species are determined by enforcing
the balance of elements and formation enthalpy. The inclusion of
NH3 and HCN is intended for the subsequent NOx simulations,
which are not covered in the paper. Then, the JL-mechanisms are
used for the combustion of the hydrocarbons (i.e., CH4 and C3H8)
in the mixture. All the reactions in this mechanism, together with
their kinetic rate data, are also listed in Table 2. The turbulencechemistry interaction is modelled by the eddy-dissipation concept
(EDC) (Magnussen, 1981).

CH2:5998 O0:3256 N0:0566 S0:0093  0:214 C3 H8 0:41 H2

|{z}
Coal Volatile

B

0:307 CO 0:051 HCN

The non-linear part, B  h4R  T 4P , needs to be handled in other way,

rather than the often used B  h4R  T P 4 which is inherently
responsible for any solution difculty. In this study, it is treated by

dB
B B
dT



T P 

T P

h4R

T P 4

4T P 3 T P

T P

CH2:1325 O0:986 N0:0145  0:172 CH4 0:0575 C3 H8 0:472 H2

|{z}
Straw Volatile

0:322 CO 0:0015 HCN 0:013 NH3

Substituting Eq. (5) into the radiation heat source, Eq. (3) can be
nally re-arranged into the standard form, in which the coefcients, aP, aW, aE and S, are calculated as,

8
>
a kADrw
aE 0
>
< W
qC DV
aP Dpt kADrw hAe 4Ae erT P 3
>
>
qC DV
:
S Dpt T p hAs T g As erh4R 3T p 4 Q_ Internal

0:0056 NH3 0:0093 SO2

This discretization secures the absolute stability of the custom

1D particle conversion model, which is independent on the volume
of the cells and the size of the time step. In this model, the different
conversion processes may occur simultaneously. This model is
implemented into Fluent via user dened function (UDF).
2.2.3. Gas phase combustion
During the different conversion processes of fuel particles along
their trajectories, some species are released into gas phase, e.g.,
volatiles and CO, which create sources for gas phase combustion.
For typical coal combustion, a great portion of the fuel conversion
occurs in gas phase. The volatiles often carry about 50% of the energy of coals. In addition, the homogeneous combustion of the volatiles plays a very important role in ignition, local stoichiometries,
ame stability and pollutant emissions. For biomass combustion,
volatile combustion is even more important since volatiles contain
the dominating part of the energy present in the biomass particles.
Therefore, gas-phase reaction mechanism is expected to play an
important role in CFD modelling of co-ring. In this study, the
Westbrook and Dryer two-step mechanism (WD) and the Jones
and Lindstedt four-step mechanism (JL) are used and compared.
The WD 2-step mechanism is derived for the oxidation of
hydrocarbon fuels in ames, and consists of two reactions with
CO as the intermediate species, in which the last step, oxidation
of CO to CO2, is reversible (Westbrook and Dryer, 1981). In this
study, two articial species are used to represent the coal-vola-

0:332 CO2

The rate coefcients of the reactions, kT  A  T b  exp REu T , can be

calculated using the kinetic rate data, A, b and E, as given in Table 2.
It has to be mentioned that the kinetic rate data of the reaction (3r)
in the JL 4-step scheme, H2 O ! H2 0:5 O2 , is derived by Andersen
et al. (2009). Since H2 0:5 O2 $ H2 O is a global reaction and the
forward reaction orders do not follow the stoichiometry, the reverse
rate constant can not be simply determined from the equilibrium
constant and the forward rate constant. The new kinetic rate data
involve a negative H2 concentration dependence, i.e., 0.75, which
may cause solution difculties. An examination of the reaction rate
coefcient shows that this reverse reaction is completely negligible
compared to the forward reaction at relatively low temperatures
(<2000 K). So, H2 O ! H2 0:5 O2 is removed from the JL 4-step
mechanism in this study.
2.2.4. Char combustion
The combustion of biomass char is quite complicated since it is
affected not only by the composition of the biomass fuel but also
by the shape and size of the fuel particles. Biomass char is believed
to be more reactive than coal char. Different measures in modelling
of biomass char combustion can be found in literature, for instance,
using diffusion-limited surface reaction rate model modied by aspect ratio-dependent enhancement factors (in the range of 11.6
for the aspect ratio from 1 to 11) due to the non-sphericity of biomass particles (e.g., Gera et al. 2002; Yin et al. 2004), and using
Smiths intrinsic model modied by a constant enhancement factor
of 4 in order to represent the high burning rate of the biomass char
particles (e.g., Ma et al. 2007, 2009). In this study, due to the lack of
non-spherical properties of the straw particles and the low char
content in the straw, the diffusion-limited surface reaction rate
model is used in all the simulations, without considering any
enhancement factor.

4173

C. Yin et al. / Bioresource Technology 101 (2010) 41694178

Table 2
Kinetic rate data of the reactions involved (Units in m, s, kmol, J, K).
No.

Reactions

Rate orders

WD 2-step global combustion mechanism for hydrocarbon fuels (Westbrook and Dryer, 1981)
2.119  1011
1
CV 0:997 O2 ! CO 1:3 H2 O 0:009 SO2 0:028 N2

2.027  108

CV0:2 O2 1:3

SV 0:541 O2 ! CO 1:067 H2 O 0:0073 N2

2.119  1011

2.027  108

SV0:2 O2 1:3

CO 0:5 O2 ! CO2

2.239  1012

1.702  108

3r

CO2 ! CO 0:5 O2

5.0  108

1.702  108

COO2 0:25 H2 O0:5

CO2 

JL 4-step global combustion mechanism for alkane hydrocarbons up to butane (Jones and Lindstedt, 1988)
1
CH4 0:5O2 ! CO 2H2
4.4  1011
0

1.26  108

2
3

CH4 H2 O ! CO 3H2
H2 0:5O2 ! H2 O

3.0  108
6.8  1015

0
1

1.26  108
1.67  108

3r

H2 O ! H2 0:5O2

1.255  1017

0.877

4.096  108

4
10

CO H2 O $ CO2 H2
C3 H8 1:5O2 ! 3CO 4H2

2.75  109
4.0  1011

0
0

8.4  107
1.26  108

20

C3 H8 3H2 O ! 3CO 7H2

3.0  108

1.26  108

2.2.5. Numerical solution

In total, about 20 computational cases (steady and axisymmetric swirl) are performed. Three different meshes are tested for the
axisymmetric swirl simulations in order to secure grid-independent solutions: coarse, ne and ultra-ne, which have 6446,
62,600 and 192,640 quadrilateral cells, respectively. Two different
turbulence models, i.e., the standard ke model and the realizable
ke model, are compared: the latter is expected to outperform the
former in predicting swirling ows. The motion of the particles is
predicted by integrating the drag and gravity forces on the particles. The conversion of the large straw particles is solved by two
methods: one is the default DPM laws in Fluent and the other is
the user dened 1D particle conversion model. Two global reaction
mechanisms for gas phase combustion are compared: the WD 2step scheme combined with the nite rate/eddy-dissipation model
for turbulence-chemistry interaction vs. the JL 4-step mechanism
combined with the EDC. For radiation heat transfer, the Discrete
Ordinates (DO) model is used in all the computational cases. The
boundary conditions are as dened in Table 1.

CH4 0:5 O2 1:25

CH4 H2 O
H2 0:25 O2 1:5
H2 0:75 O2  H2 O
COH2 O
C3 H8 0:5 O2 1:25
C3 H8 H2 O

the entire reactor, even at the outer radius or far downstream of

the burner quarl, the ultra-ne mesh does not make any remarkable difference with the ne mesh in the CFD results. As a result,
the ne mesh, which has 62,600 quadrilateral cells in total, is sufcient to obtain a grid-independent CFD solution.

3.2. Effect of DPM laws

Fig. 4 shows the evolution of temperatures (including specic
heat-weighted mean temperature, temperatures at particle outer
surface and particle centre) and conversions of a straw particle
whose diameter is dp = 451 lm, as a function of its axial location

3. Results and discussion

3.1. Effect of meshes
The effect of meshes can be demonstrated by the calculated
proles of, e.g., species, temperature or velocity. Fig. 3 show such
an example. In the core zone where the grids are ne, to different
extents, in all the three meshes, the CFD results based on the three
meshes are quite close to each other, as shown in Fig. 3a. In the
outer radius, the coarse mesh will produce very different CFD results with the ne and ultra-ne meshes, as shown in Fig. 3b. In

Fig. 4. Temperature and conversion history of a straw particle as it travels in the

reactor.

Fig. 3. Predicted O2 proles based on the three different meshes.

4174

C. Yin et al. / Bioresource Technology 101 (2010) 41694178

when it travels in the burner reactor. This particle is discretized

into 15 spherical shells in radial direction. For each of the shells,
the temperature, conversion rates and compositions are updated.
The vertical axis on the right in Fig. 4 is the ratio of the mass of
jth species remaining in the fuel particle to the initial particle mass,
msp,j/mp0, where j = 1, 2, 3 and 4 correspond to moisture, volatiles,
char and ash, respectively. The gure indicates that before remarkable evaporation starts, the temperature difference inside the particle is very low: DT T p surf  T p centre < 5 K between the surface
and centre. When evaporation starts, this temperature difference
increases quickly as the evaporation front moves towards the particle centre. The maximum difference can be up to 150 K when the
evaporation front reaches the centre of the particle. After evaporation is completely nished, the big temperature difference between the surface and centre vanishes, and the difference turns
to be negligible again: |DT| < 5 K. Compared with the evaporation
rate, the char oxidation rate is very low. As a result, the temperature difference inside the particle is negligible even when the char
oxidation starts. It can also be seen from the gure that the devolatilization rate is much faster than the evaporation rate and char
oxidation rate. Here it has to be mentioned that the devolatilization does start simultaneously with the evaporation process,
although the plot of the volatile (i.e., curve 2 in Fig. 4) looks at before the evaporation is completely nished. This is just because the
particle temperature is still relatively low before the moisture is
completely driven off, <700 K inside all the shells throughout the
particle. As a result, the rate of the simultaneous devolatilization
is negligible. After the particle temperature rises above 700 K, a rapid devolatilization is observed.
In short, Fig. 4 indicates that a remarkable devolatilization is observed only after the moisture in the particle is completely driven
off and a remarkable char oxidation occurs only after all the volatiles are released. One can conclude that the user dened 1D particle model will not make a big difference with the default DPM
laws in the conversion of the straw particles red in this reactor,
which is also conrmed by the relevant CFD results. So for biomass
particles whose size is a few hundred microns in average, the intraparticle heat and mass transfer is a secondary issue at most in their
conversion and the default DPM laws can still be used with good
reliability.

ability with average differences of 4.6% and 3.5%, respectively,

whilst O2 and CO showed the least repeatability, with average differences of 20.3% and 15.1%, respectively (Damstedt, 2007). Taking
these into account, one can conclude from Fig. 5 that the CFD predictions in all the three cases show a generally good agreement
with the measured data, which also agree with expectations on
the proles of the species.
The prole of O2 correlates inversely with the product (CO2 and
H2O) and intermediate (CO) concentrations. Upon injection into
the burner quarl, the fuel particles travel in the reactor, dry rapidly
and devolatilize, producing an attached ame, which results in a
rapid decrease in O2 concentration and a rapid increase in the
product concentration, followed by a more gradual variation to
the efuent concentration. In the CFD simulations, regions of zero
O2 concentration in the core zone are captured. However, in the
experiments no region of zero O2 is found. The regions of high
CO concentration (up to 89%) exhibited by the experiments indicate that fuel-rich zones do exist in the core zone. The coexistence
of O2 and CO in the ame region can be understood because of the
turbulent and uctuating nature of the system. A xed location inside the ame alternates between fuel-rich eddies and oxidant-rich
eddies. Samples containing high CO and zero O2 combine with
samples containing high O2 and low CO, which gives rise to a
none-zero O2 concentration in the measurement even inside the
fuel-rich ame (Damstedt et al., 2007).
For the predicted CO2 and H2O concentrations, the most
remarkable difference between case 0 and case 2 seems to be that
case 2 predicts a bimodal prole of CO2 in the core zone. Such a bimodal behaviour may be related with the break in the ame
structure, which was visually observed during sampling. The break
was found to occur near x = 0.9 m, and was partly attributed to larger straw particles which show different combustion behaviour
compared to the rest of particles (Damstedt, 2007).
The realizable ke turbulence model is found to make a bigger
difference with the standard ke model in the core region than in
the near-wall region, as expected. For combusting swirling ows,
the delity of CFD modelling is mainly determined by the accuracy
of the turbulence model in the core region, rather than in the nearwall zone.
3.4. Contours of key parameters

3.3. Comparison between CFD results and experimental data

Table 3 only lists the differences between the three computational cases whose results are presented in details in this paper.
All the other settings in the three cases are precisely same.
Before comparing the CFD results with the measured data, one
has to be aware of the difculties of measuring gas species in the
down-red turbulent ames in this reactor. Ideally, the measured
data are expected to be perfectly symmetric about the geometric
centreline. In reality, none of the measured species exhibit perfect
symmetry. Compared to lifted ames, down-red ames tend to be
more unstable in regard to symmetry due to the buoyant forces of
the hot combustion products. Relatively small forces can produce a
deviation in the ow from one direction to another. Among the
major species measured, CO2 and H2O showed exceptional repeat-

Table 3
Three CFD cases with results presented in details.
Turbulence model

Chemistry model

Combustion model

Case 0

Realizable ke

WD 2-step

Case 1
Case 2

Standard ke
Realizable ke

JL 4-step
JL 4-step

Finite rate/eddydissipation
EDC
EDC

More details of the CFD results for case 0 and case 2 are presented in this section. Fig. 6 shows the ow structure, fuel particle
reaction rate, ame volume (i.e., the regions with low O2 and high
CO concentrations in this context) and ame temperature.
In both the cases, two main recirculation zones are predicted in
the top section of the reactor, just as expected for such a conned
swirl burner ow (Syred and Ber, 1974). One is the internal recirculation zone (IRZ), and the other is the external recirculation zone
which is located in the upper corner bounded by the swirling secondary air (SA) jet and the walls of the enclosure. It has to be mentioned that the bulk of the SA is not recirculated and passes
between the two recirculation zones in a semi-conical jet. The
IRZ recycles hot gas ow toward the burner along a portion of
the centreline, largely slowing down the incoming fuel particles.
The fuel particles can be quickly heated up and undergoes a series
of conversion processes (e.g., drying, devolatilization and char oxidation), producing an attached ame. The IRZ is an aerodynamic
ame holding mechanism. Therefore, inserting a probe into it will
disrupt the ow pattern and then the ame structure. The probe
impacts are minimal on the close side in the outer radius (e.g.,
y = 15 cm), and maximal on the corresponding far side (e.g.,
y = 60 cm) since the probe passes completely though the IRZ. That
is why the measured gas species tends to be more symmetric about
the centreline in the core zone than in the outer radius, as shown in

C. Yin et al. / Bioresource Technology 101 (2010) 41694178

4175

Fig. 5. The predicted O2, CO2, H2O and CO vs. measured data along different measuring lines for the co-ring ame. Co-ring conditions: central straw 15.1 kg/h; central air
9 kg/h; annular coal 7.5 kg/h; annular air 12 kg/h; swirling secondary air 160 kg/h (swirl number 1).

Fig. 5. For instance, the measured CO2 concentrations along the

pair-line, y = 35 cm and y = 40 cm, are quite close to each other;
whilst the values measured along another pair-line, y = 15 cm
and y = 60 cm show a great difference.
The contours of DPM mass sources give a good indication of
the fates (e.g., motion and conversion) of different fuel particles.
The central straw stream travels in a nearly straight line around

the centreline, penetrates though the fuel-rich IRZ and burns in

an axially expanded region. The coal stream released from the surrounding annulus is greatly affected by the swirling SA jet and the
majority of the coal stream is swirled into the oxygen-rich outer
radius. It is the different particle momentum response time which
2
is proportional to dp and the different fuel/air jet momentum that
cause the very different motion characteristics of the coal and

4176

C. Yin et al. / Bioresource Technology 101 (2010) 41694178

Fig. 5 (continued)

straw streams. The remarkable difference in the trajectories of coal

and straw particles is much better illustrated from our modelling
results of two pure fuel ames (Yin and Kr, 2009), i.e., pure coal
ame and pure straw ame, both of which have the same thermal
output (about 150 kW) and in both of which the solid fuels are fed
into the burner through the centre tube whilst the surrounding
annulus is free of ow. Most of the coal particles are swirled into

the oxygen-rich outer radius, which also increases the residence

time of coal particles (8.1 s in average in the 3 m high reactor).
Both of the factors favour a complete burnout of coal particles:
the overall burnout predicted is about 93% for the coal char. One
the contrary, a high un-burnout will be expected for the straw particles because of the oxygen-lean environment through which the
particle streams travel and the short residence time in the reactor.

C. Yin et al. / Bioresource Technology 101 (2010) 41694178

4177

Fig. 6. Comparison of CFD results of case 0 (left) vs. case 2 (right) for the co-ring ame. Co-ring conditions: central straw 15.1 kg/h; central air 9 kg/h; annular coal 7.5 kg/h;
annular air 12 kg/h; swirling secondary air 160 kg/h (swirl number 1).

The average residence time is 5.2 s and the overall burnout predicted is about 73% for the straw char, both of which are notably
lower than those of the coal streams.
Biomass containing ames were found to occupy a larger volume than coal ames due to a number of factors. For instance, biomass has a lower energy density per unit mass, requiring an
increase in the mass feeding rate to maintain the thermal input,
which also needs an increase in the primary air ow for transportation. These will increase the biomass/air jet momentum, which,
together with the large particle response time, will make it easier
for biomass particles to penetrate through the IRZ. The higher volatile yields of biomass fuels produce more off-gas, requiring more
O2 for the fast homogeneous reactions, causing the off-gas to proceed to lower reactor regions prior to mixing with oxidizer. This
co-ring ame extended axially beyond 125 cm, and expanded
outward radially for 1520 cm. In regard to the prediction of the
ame volume, case 2 perform much better than case 0. Gas-phase
reaction plays an important role in combustion of solid fuels, particularly biomass fuels. As a result, the JL 4-step mechanism combined with the EDC is expected to be an attractive modelling
approach for solid fuel combustion.
In summary, simply replacing part of the coal stream in existing
coal-red swirl-stabilized burners with biomass stream of equivalent thermal input may not achieve efcient and clean co-ring.
The increased momentum of biomass/air jet and the larger particle
response time will break the balance between the fuel/air jet and
the IRZ previously existing in the burner before co-ring is implemented. When co-ring biomass fuels whose mean particle size is
greater than a few hundred microns, biomass particles could be
better injected from the surrounding annulus or even together
with the swirling secondary air in some cases (e.g., in case the swirl
is induced by inclined or tangential inlet). In this way, large biomass particles may be swirled to the oxy-rich outer radius and
can have a longer residence time in the reactor, both of which help

to improve the burnout of large biomass particles. Biomass fuels

often contain a large amount of volatiles. As a result, air supply
needs to be adjusted according to the trajectories of the majority
of biomass particle streams, in order to assure sufcient O2 is available in zones where the volatiles are released. When co-ring biomass fuels whose particle size is sufciently small (e.g., on the
same level of the co-red coal particles), the injection point of biomass fuels may not be very important and the pulverized biomass
fuels can even be pre-blended with coal before being fed into the
burner. However, care still needs to be taken in air supply because
of the often high content of volatiles in biomass fuels.

4. Conclusions
For pulverized biomass particles of a few hundred microns in
diameters, the intra-particle heat and mass transfer is a secondary
issue at most in their conversion. Both the JL 4-step global mechanism combined with the EDC and the WD 2-step scheme with the
nite rate/eddy-dissipation reasonably well predict the major species. The former outperforms the latter in regard to the ame volume prediction, and is expected to be a good modelling method for
volatiles combustion. When co-ring biomass in existing coal-red
burners, measures to increase the residence time of large biomass
particles in oxygen-rich zone will be helpful, which demand special care in feeding of biomass fuels and supply of air.

Acknowledgements
This work was nancially supported by Grant PSO 4806, Optimization of low NOx burner for co-ring. The authors would like
to thank a number of anonymous reviewers for their valuable comments that helped improve the quality of the paper.

4178

C. Yin et al. / Bioresource Technology 101 (2010) 41694178

References
Abbas, T., Costen, P., Kandamby, N.H., Lockwood, F.C., Ou, J.J., 1994. The inuence of
burner injection mode on pulverized coal and biomass cored ames.
Combustion and Flame 99, 617625.
Andersen, J., Rasmussen, C.L., Giselsson, T., Glarborg, P., 2009. Global combustion
mechanisms for use in CFD modeling under oxy-fuel conditions. Energy and
Fuels 23, 13791389.
Backreedy, R.I., Fletcher, L.M., Jones, J.M., Ma, L., Pourkashanian, M., Williams, A.,
2005. Co-ring pulverized coal and biomass: a modeling approach. Proceedings
of the Combustion Institute 30, 29552964.
Baxter, L., 2005. Biomass-coal co-combustion: opportunity for affordable renewable
energy. Fuel 84, 12951302.
Chao, C.Y.H., Kwong, P.C.W., Wang, J.H., Cheung, C.W., Kendall, G., 2008. Co-ring
coal with rice husk and bamboo and the impact on particulate matters and
associated polycyclic aromatic hydrocarbon emissions. Bioresource Technology
99, 8393.
Damstedt, B., 2007. Structure and Nitrogen Chemistry in Coal, Biomass and Coring
in Low-NOx Flames, Ph.D. Thesis, Brigham Young University.
Damstedt, B., Pedersen, J.M., Hansen, D., Knighton, T., Jones, J., Christensen, C.,
Baxter, L., Tree, D., 2007. Biomass coring impacts on ame structure and
emissions. Proceedings of the Combustion Institute 31, 28132820.
Fluent Inc., 2006. FLUENT 6.3.26 Users Guide.
Gera, D., Mathur, M.P., Freeman, M.C., Robinson, A., 2002. Effect of large aspect ratio
of biomass particles on carbon burnout in a utility boiler. Energy and Fuels 16,
15231532.
Hansson, J., Berndes, G., Johnsson, F., Kjrstad, J., 2009. Co-ring biomass with coal
for electricity generation an assessment of the potential in EU27. Energy
Policy 37, 14441455.
Hvid, S.L., 2006. PSO4806 Project Sub-report: Determination of BRF Secondary Air
Inlet Swirling Boundary Conditions. DONG Energy, Fredericia, Denmark.
Johansson, R., Thunman, H., Leckner, B., 2007. Inuence of intraparticle gradients in
modeling of xed bed combustion. Combustion and Flame 149, 4962.
Jones, W.P., Lindstedt, R.P., 1988. Global reaction schemes for hydrocarbon
combustion. Combustion and Flames 73, 233249.
Lu, G., Yan, Y., Cornwell, S., Whitehousem, M., Riley, G., 2008a. Impact of co-ring
coal and biomass on ame characteristics and stability. Fuel 87, 11331140.
Lu, H., Robert, W., Peirce, G., Ripa, B., Baxter, L.L., 2008b. Comprehensive study of
biomass particle combustion. Energy and Fuels 22, 28262839.
Laux, S., Tillman, D., Seltzer, A., 2002. Design issues for co-ring biomass in wallred low NOx burners. In: Proc. of 27th Int. Tech. Conf. on Coal Utilization and
Fuel Systems, Clearwater, FL.

Ma, L., Gharebaghi, M., Porter, R., Pourkashanian, M., Jones, J.M., Williams, A., 2009.
Modelling methods for co-red pulverized fuel furnaces. Fuel 88,
24482454.
Ma, L., Jones, J.M., Pourkashanian, M., Williams, A., 2007. Modelling the combustion
of pulverized biomass in an industrial combustion test furnace. Fuel 86, 1959
1965.
Magnussen, B.F., 1981. On the structure of turbulence and a generalized eddy
dissipation concept for chemical reaction in turbulent ow. 19th AIAA
Aerospace Meeting, AIAA, New York.
Magnussen, B.F., Hjertager, B.H., 1976. On mathematical modeling of turbulent
combustion with special emphasis on soot formation and combustion. In: 16th
Symposium (International) on Combustion, Combustion Institute, Pittsburgh,
PA, pp. 719729.
McKendry, P., 2002. Energy production from biomass (part 2): conversion
technologies. Bioresource Technology 83, 4754.
Molcan, P., Lu, G., Le Bris, T., Yan, Y., Taupin, B., Caillat, S., 2009. Characterisation of
biomass and coal co-ring on a 3 MWth combustion test facility using ame
imaging and gas/ash sampling techniques. Fuel 88, 23282334.
Pallars, J., Gil, A., Corts, C., Herce, C., 2009. Numerical study of co-ring coal and
cynara cardunculus in a 350 MWe utility boiler. Fuel Processing Technology 90,
12071213.
Robinson, A.L., Rhodes, J.S., Keith, D.W., 2003. Assessment of potential carbon
dioxide reductions due to biomass-coal coring in the United States.
Environmental Science and Technology 37, 50815089.
Sami, M., Annamalai, K., Wooldridge, M., 2001. Co-ring of coal and biomass fuel
blends. Progress in Energy and Combustion Science 27, 171214.
Spliethoff, H., Hein, K.R.G., 1998. Effect of co-combustion of biomass on emissions in
pulverized fuel furnaces. Fuel Processing Technology 54, 189205.
Syred, N., Ber, J.M., 1974. Combustion in swirling ows: a review. Combustion and
Flame 23, 143201.
Westbrook, C.K., Dryer, F.L., 1981. Simplied reaction-mechanisms for the oxidation
of hydrocarbon fuels in ames. Combustion Science and Technology 27,
3143.
Yin, C., Kr, S.K., 2009. PSO4806 Project Sub-report: Modelling and Design of Low
NOx Burners for Biomass/Coal Co-ring, Aalborg University.
Yin, C., Rosendahl, L., Kr, S.K., 2008. Grate-ring of biomass for heat and power
production. Progress in Energy and Combustion Science 34, 725754.
Yin, C., Rosendahl, L., Kr, S.K., Condra, T., 2004. Use of numerical modeling in
design for co-ring biomass in wall-red burners. Chemical Engineering Science
59, 32813292.
Zhou, H., Jensen, A.D., Glarborg, P., Jensen, P.A., Kavaliauskas, A., 2005. Numerical
modeling of straw combustion in a xed bed. Fuel 84, 389403.