Вы находитесь на странице: 1из 16

73

Modeling Issues in CFD Simulation


of Brown Coal Combustion in a
Utility Furnace
Z. F. Tian*,1, P. J. Witt1, M. P. Schwarz1 and W. Yang2
1CSIRO,

Mathematics, Informatics and Statistics,


Process Science and Engineering,
P.O. Box 312, Clayton South, VIC 3169, Australia
*Email: zhaofeng.tian@csiro.au
2CSIRO,

Received: 13 May 2010, Accepted: 1 June 2010


Abstract
This paper describes the mathematical formulation and modelling issues of a
computational fluid dynamics (CFD) model of a 375 MW utility furnace. This
tangentially-fired furnace is fuelled by high moisture content brown coal from coal
mines at Latrobe Valley in Victoria, Australia. The influences of different turbulence
models, particle dispersion, and radiation models on the CFD prediction are
investigated. Two turbulence models, standard k- model and Shear-Stress Transport
(SST) model, provide similar predictions that are in good agreement with the plant
data. The effect of particle dispersion on the prediction is found to be insignificant for
this high-volatile brown coal. The predicted wall incident radiation flux based on two
radiation models, namely, discrete transfer (DT) model and P-1 model are compared
against power plant measurements. The comparison reveals that the DT model provides
good prediction of the radiation profiles, while the P-1 model considerably underpredicts the wall incident radiation flux.

1. INTRODUCTION
Pulverised coal combustion within utility furnaces generally involves the simultaneous processes
of three-dimensional gas-particle fluid dynamics, turbulent mixing, heat transfer, and chemical
kinetics (Viskanta and Menguc, 1987). In computational fluid dynamics (CFD) approaches,
mathematical descriptions of each of these processes are called submodels, as they are developed
and updated in the same way that modules in electronic circuit boards can be replaced (Niksa,
1996). The following submodels are required to calculate pulverised coal combustion in utility
furnaces: turbulence model, particle phase model including a devolatilisation and a char
combustion model, gaseous phase combustion model and a heat transfer model that includes a
radiation model. In some cases additional chemistry submodels for prediction of emissions such as
soot and NOx are also needed. Figure 1 shows the key submodels and some of the different options
available for each submodel in the CFD code ANSYS/CFX 12.0 (2009).
Since the 1980s a great deal of research effort and resources has been allocated to develop and
improve the submodels which were applied to understand coal combustion in furnaces (Truelove,
1986; Truelove and Holcome, 1990; Tu et al., 1997; Xu et al., 2001; Vuthaluru and Vuthaluru,
2006; Bris et al., 2007 and many others). Nevertheless, uncertainties still remain, the key one being
the modeling of dominant flow pattern in utility furnaces where the complex turbulent flow and
burner jet interactions take place. Improper modeling of turbulence in the furnace gas phase will
result in inaccurate calculation of turbulent mixing, possibly leading to significant errors in the
prediction of overall combustion and heat transfer processes (Tian et al., 2009). Another issue is the
modeling of radiative heat transfer that is the dominant heat transfer mechanism in such furnaces.
Radiative heat transfer limitations have been recognized as a prime source of errors in temperature
predictions, with the degree of error varying in different systems (IIBD, 2002).

74

Modeling Issues in CFD Simulation of Brown Coal


Combustion in a Utility Furnace
Coal combustion solver

Turbulence
models

Particle phase
models

Gas combustion
models

Radiation
models

Emissions
models

k- model

Lagrangian
model

EDM
model

P-1model

Soot
model

FRC
model

DT model

NOx
model

SST model
k- model

Devolatilisation
model

RNG k-
model

Char
oxidation

Monte Carlo
model

Combined
model

Figure 1. The structure of the CFD coal combustion solver ANSYS/CFX 12.0.
RSM: Reynolds Stress model; EDM: Eddy dissipation model; FRC: Finite rate
chemistry model; Combined model: Combined EDM/FRC model;
DT: Discrete transfer model. SST: Shear-stress transport model.

A CFD model of unit No. 3 furnace at TRUenergy Yallourn power plant has been developed.
This 375 MW tangentially-fired furnace is fuelled by raw brown coal from Latrobe Valley coal
mines in Victoria, Australia. Coal feed rates, coal properties, coal particle size distribution, and air
flow rates obtained from plant data are used as input conditions in the CFD model. Validation of
the CFD model and a study on the effects of operational conditions on the brown coal combustion
in the furnace has been reported elsewhere (Tian et al., 2010a). This paper describes the model
formulation and addresses the turbulence modeling, turbulence effect on particles and radiation
modeling issues in applying CFD models to this utility furnace.
2. DESCRIPTION OF CFD MODEL
The commercial CFD code, ANSYS CFX 12.0 (2009) has been applied to predict pulverised coal
combustion in the furnace model. As discussed above a number of submodels are required to describe
the various complex processes occurring a furnace. Details of these models are provided below.
2.1. Gas Phase Flow Model
Gas phase flow in the furnace is considered to be a gas mixture consisting of the gaseous
components, O2, H2O, CO2, CO, N2, NO, and volatiles. These components are assumed to be mixed
at the molecular level, hence, share the same mean velocity, pressure, temperature and turbulence
fields. The following mass conservation equation (or mass fraction equation) is solved for each
component I, except for N2 which is a constraint, in the gas mixture under the assumption of a
steady state flow field:

gU gYI = I .eff YI + SI

(1)

where YI is the mass fraction of the component I, g is the density of the gas mixture defined as:
Nc

g = YI I

(2)

I =1

here I is the mass density of the component I. Nc is the number of components in the gas mixture.
Other properties of the gas mixture including the molecular viscosity of the gas mixture g, the
Journal of Computational Multiphase Flows

Z. F. Tian, P. J. Witt, M. P. Schwarz and W. Yang

75

specific heat Cp, and the thermal conductivity g are also calculated in the same manner as
Equation (2):
Nc

g = YI I

(3)

I =1

where in the property being considered.


The effective diffusion coefficient, I.eff , in Equation (1) is defined as:
I .eff = I +

t
Sct

(4)

where the molecular diffusion coefficient I is assumed to be equal to I DI, where DI is the kinematic
diffusivity of the component. Sct is the turbulent Schmidt number. t is turbulent viscosity.
The source term SI in Equation (1) is due to chemical reaction involving the component. Each
gas component except the constraint gas N2 has its own transport equation. The mass fraction of N2
is determined by the constraint equation:
Nc

Y
I =1

=1

(5)

Bulk properties of the gas mixture are characterized by single velocity, pressure, temperature
and turbulence fields. Mean steady state motion of the gas mixture is determined by solving the
Reynolds averaged conservation equation for momentum:

gU gU g = pg + g U g + U g

U U +S
g g g
M

(6)

Here, pg is gas mixtures pressure. The term SM is the sum of body forces.
The Reynolds averaged energy equation for the gas mixture is:

gU g hg = g Tg + t hg + SE
Prt

(7)

where hg is the gas enthalpy. The energy source term SE includes thermal energy from chemical
reactions and thermal radiation.
2.2. Turbulence Model Gas Phase Flow Model
The terms g U gU g in Equation 6 are the Reynolds stresses which are modeled based on the
Boussinesq hypothesis:

g U gU g = t U g + U g

2 k + U
t
g
3 ij g g

(8)

2.2.1. k- Turbulence Model


For the standard k- model the turbulent viscosity is computed from:

t = C g kg2 / g

(9)

where kg is the gas mixture turbulence kinetic energy and g is the gas mixture kinetic energy
dissipation rate.
Turbulence kinetic energy equation of standard k- model is:


gU g kg = g + t kg + Pk g g
k

Volume 2 Number 2 2010

(10)

76

Modeling Issues in CFD Simulation of Brown Coal


Combustion in a Utility Furnace
Here, the rate of production of turbulence kinetic energy Pk is modeled by:

(11)

g

gU g g = g + t g + (C 1 Pk C 2 g g )
k

kg

(12)

2
Pk = t U g U g + (U g )T U g t U g + g kg
3
Kinetic energy dissipation rate equation is:

These equations contain five constants C, k, , C1 and C2. These constants have been arrived
at by a comprehensive data fitting exercise over a wide range of turbulent flow conditions (Launder
and Spalding, 1974), C = 0.09, k = 1.0, = 1.3, C1 = 1.44, C2 = 1.92.
2.2.2. Shear-Stress Transport (SST) Turbulence Model
The SST model (Menter, 1994) consists of a transformation of the k- model to a k- formulation
to eliminate an apparent deficiency in the original k- model, viz, the overestimation of the
spreading rates in the free shear flow. The k and equations of SST model are:


gU g kg = g + t kg + Pk g kg g
k3

(13)

g

1
gU g g = g + t g + 3
Pk + (1 F1 )2 g
g kg 3 g g2 (14)
k

g
3
2 g

For the SST model the turbulent viscosity is given as:

t =

g k g 1
Max (1 , SF2 )

(15)

where S = SS and S is the strain rate.


The blending function F1 is based on the distance to the nearest surface and on the flow
variables:

kg
500 g 4 g kg

F1 = tanh ( 14 , 1 = min max


,
,
2
+ 2

0.09 g y g y g ,2 D y

1 1
D = max 2 g
kh h ,10 10

,2
g

(16)

The blending function F2 is given as:

F2 = tanh ( 22 ,

2 kg
500 g
2 = max
,

2
0.09 g y g y g

(17)

2.3. Thermal Radiation Model


In a combustor, radiative heat transfer from the flame and combustion products to the surrounding
walls and coal particles can be predicted if the radiative properties and temperature distributions in
the medium and on the walls are available (Viskanta and Menguc, 1987). Two models for thermal
radiation, discrete transfer model (DT) and P-1 model, are tested in this paper.
Journal of Computational Multiphase Flows

Z. F. Tian, P. J. Witt, M. P. Schwarz and W. Yang

77

The radiative transfer equation (RTE) for an absorbing, emitting and scattering medium at
position r in the direction s is:
r r
dI (r , s )
K
r r
= ( K a + K s ) I ( r , s ) + K a I b (, T ) + s
4
ds

r r
r r
dI ( r , s ) ( s s ) d

(18)

Here, Iv is the spectral radiation intensity which depends on position r and direction s.
r , s, s
are position vector, direction vector and scattering direction vector, respectively. Ka is the spectral
absorption coefficient and Ks is the spectral scattering coefficient. Ib is blackbody emission
intensity. is the Stefan-Boltzman constant (5.762 108 W m2 K 4). is the solid angle. T is the
local temperature.

2.3.1. Discrete Transfer Model


In the DT model, scattering is assumed to be isotropic and the system is assumed to be reasonably
homogeneous, this model is then used to solve for the intensity Iv along rays leaving the boundaries
and then accounts for its concentration as source terms to the heat transport equation,
I v (r, s ) = I v 0 exp(( K av + K sv )s ) + I bv (1 exp( K a s )) + K sv I v

(19)

where Iv0 is the radiation intensity leaving the boundary. The radiation intensity is then integrated
over solid angle at discrete points to get the spectral incident radiation, and the radiative heat flux.
Based on the homogeneity assumption the solution is extended to the entire domain. Non-linearities
in the system due to scattering, diffuse reflection, or temperature dependency of radiation quantities
is overcome by iteration (Lockwood and Shah, 1981).
2.3.2. The P-1 Model
The P-1 model is the simplest case of P-Ns model that are based on the expansion of Iv in an
orthogonal series of spherical harmonics (Sazhin et al., 1996). It assumes that the radiation intensity
is isotropic or direction independent at a given location in space. The thermal radiation transfer
problem in the P-1 model reduces to the solution of the following equation:

1

Gv = K av ( Ebv Gv )

3( K av K sv ) AK sv

(20)

Here Gv is spectral incident radiation


Gv = I v d s

(21)

and A is the linear anisotropy coefficient. The thermal heat transfer is linked to the energy equation
by the term qrv (Sazhin et al., 1996):
qrv =

1
Gv
3( K av K sv ) AK sv

(22)

2.4. Gas Combustion Model


The coal combustion process is modeled by the following reactions:
Coal Volatiles + H 2 O(vapour) + C ( char )solid

(23)

Volatiles(HC) + O 2 CO + H 2 O

(24)

1
C ( char )solid + O 2 CO
2

(25)

1
CO + O 2 = CO 2
2

(26)

Volume 2 Number 2 2010

78

Modeling Issues in CFD Simulation of Brown Coal


Combustion in a Utility Furnace
The gas phase reactions (Equation 24 and 26) are calculated by a single step combustion model
using an eddy dissipation model, which is based on the concept that the chemical reaction is fast
relative to the transport processes in the flow. Equations (23) and (25) are heterogeneous reactions
and are discussed later in the particle model section. When the volatile gases or CO and oxygen
mix at the molecular level, they instantaneously react and form products. The source term of
component I in Equation (1), SI, is determined from the combined contributions of the reaction rates
by:
Nr

SI = WI nIJ RJ

(27)

J =1

Where WI is the molecular weight of component I. nIJ is the difference of the stoichiometric
coefficients for reactants and products, Nr is the number of reactions. The reaction rate of reaction
J, RJ, is assumed to be directly related to the time required to mix reactants at the molecular level,
i.e. a mixing time defined by the turbulent kinetic energy of gas mixture, kg, and dissipation rate, g:
RJ = A

g
X
min I
kg
vI

(28)

Where [XI] is the molar concentration of the reactant component I. vI is the stoichiometric
coefficient for I. A is a constant with a value of 4.
2.5. Particle Phase Transport Model Description
Temperature, composition and velocity of coal particles along their trajectories are predicted using
a Lagrangian particle tracking model. The equation of particle motion is:
dx p
dt

mp

du p
dt

= up

(29)

= FD + Fg

(30)

Where FD is drag force and Fg is gravity force. The drag force FD is calculated from:
FD = m p

u g.inst u p

(31)

When turbulent dispersion of particles by gas phase turbulence is included in the CFD
model the trajectory equations for representative particles are integrated using the
instantaneous fluid velocity along the particle path during the integration process. With
u g.inst = U g + u where u is the fluctuating turbulent velocity calculated based on Gosman and
Ioannides (1981). When particle turbulent dispersion is not included the mean gas velocity is
used, u g.inst = U g . As a stochastic method is utilized 7,100 sample coal particles are injected
through each in-service firing group, making a total of 42,600 sample coal particles tracked in
the furnace.
The particle relaxation time r for the drag force is defined as:

r =

p d p2
18 g fD

(32)

Journal of Computational Multiphase Flows

Z. F. Tian, P. J. Witt, M. P. Schwarz and W. Yang

79

with the drag coefficient, fD, determined by the Schiller_Naumann (1933) expression for a sphere,
given as:
1 + 0.15Re 0.687
,
p
fD =
0.01833 Re p ,

CD =

Re p 1000
Re p > 1000

24
fD
Re p

(33)

(34)

Particle temperature, Tp, is obtained by solving an energy equation for the particle:
( mi Cpi )

dT p
dt

= QC + QM + QR

(35)

Where mi and Cpi are the mass and specific heat of component i in the particle. The terms QC , QM
and QR are energy transfers to the particle due to convection from the gas phase, combustion and
chemical reactions and radiation. Details of the various terms are given in ANSYS/CFX (2009).
2.6. Particle Phase Combustion Model
The chemical reaction in Equation (23) is devolatilisation of the coal particle and the SFOR (single
first order reaction) model is used to calculate the devolatilisation rate. The rate of devolatilisation
is assumed to be linearly dependent on the amount of volatiles remaining in the particle:
dm p
dt

= kv max[ 0, m p (1 fv,0 )m p,0 ]

(36)

where mp is the instantaneous particle mass, mp,0 is the initial particle mass. fv,0 is the mass fraction
of volatiles initially present in the particle and kv is the kinetic rate defined as:
T
kv = Av exp v
Tp

(37)

where Tp is the temperature of coal particle. The pre-exponential factor Av and activation
temperature Tv are, in this study, taken from a study on brown coal by Duong (1987).
The global reaction model is used to calculate the coal char oxidation described by the chemical
reaction Equation (25).
The equation used to calculate coal char oxidation is as follows:
dm p,C
dt

= d p2 g RTg ( XO 2 / M O 2 ) RT

(38)

where mp,C is the mass of char in the particle, dm p,C / dt is the rate of char oxidation to CO, dp is
particle diameter, Xo2 oxygen molar fraction, Mo2 molecular weight of oxygen. RT is overall reaction
rate calculated by:
1
RT = ( Rdiff
+ RC1 )1

Volume 2 Number 2 2010

(39)

80

Modeling Issues in CFD Simulation of Brown Coal


Combustion in a Utility Furnace
Rdiff is the diffusion rate coefficient given by:

Rdiff

Dref T p + Tg PA
=
d p 2Tref pg

(40)

where Tg is gas phase temperature. pg is local gas pressure and PA is atmospheric pressure. Dref is
the dynamic diffusivity and Tref is the reference temperature normally 293K.
The chemical rate coefficient Rc in Equation (39) is given as:
T
RC = AC exp C
TP

(41)

The pre-exponential factor Ac and activation temperature Tc of the char oxidation model are
497 kg m2 s1 and 8540 K respectively and based on Wall et al. (1976).
2.7. Numerical Solution
Solution of the above equations is achieved using ANSYS/CFX 12.0 with the details of the
numerical procedure given in ANSYS/CFX (2009) and more details of numerical solution can be
found in Tian et al. (2010a).
3. FURNACE MODEL
The TRUenergy Yallourn unit No. 3 is 77.5 m high and has a 15.9 15.9 m2 square cross-section.
The unit generates 319 kg s1 of steam, at 16.8 MPa and 541C when operating at full load
operation. Geometry of the CFD model for the furnace is shown in Figure 2. This furnace is
equipped with eight firing groups. Each firing group is comprised of two wall-mounted slot main
burners (upper and lower), two vapour burners (upper and lower), mill and duct system, and a gas
off-take that extracts hot furnace gas to heat and dry the raw brown coal in the mill. The raw brown
coal is put in the duct system without any pre-drying and ground by the mills to give a fine particle
size. The coal is dried, by hot furnace gas recycled from the off-takes, in the mill system before
being fed into the furnace through the vapour and main burners.
At full load operation six firing groups are normally required supplying a total of 156 kg s 1 of
raw brown coal particles. These coal particles and any steam or volatiles released during drying
enter the furnace through vapour and main burners. Hence in normal full load operation two firing
groups are out-of-service with no coal flow. About 20 kg s 1 of air flows through each of the outof-service firing groups protecting the burners from the large amount of radiant heat generated by
combustion in the furnace. In plant operation selection of the two out-of-service firing groups is
normally determined by maintenance requirements. Total secondary air flow under the full load
operation is about 370 kg s 1, which is supplied through the secondary air nozzles. These secondary
air nozzles are located in the main burners just above and below the primary nozzles through which
the gas and pulverised coal enters the furnace. In this study, the coal flow rate, air flow rate and
furnace gas flow rate are assumed to be evenly distributed to the six in-service firing groups.
Detailed geometry, such as tubes in the convective passes, has not been included in the current
CFD model, since the main focus of the current study is on coal combustion and heat transfer in the
radiant pass of the furnace. However, the regions where convective tube banks are located has source
terms added to the momentum and energy equations. The source terms account for the tubes by
damping the streamwise velocity components and generating a pressure drop as a function of gas
phase velocity. Heat absorption in the convective tube banks is also included via a source term as
given in Table 1. Properties of the high moisture content raw brown coal are also given in Table 1.
A level of confidence in the current CFD model has been established and reported by Tian et al.
(2010a) where a mesh independence test was performed and validation of model predictions against
the furnace gas exit temperature (FGET), concentration of flue gas components, total boiler heat
supply, and the wall incident heat fluxes as measured in the power plant. The effect of several operating
conditions at full load such as different out-of-service firing groups and different combustion air
distributions on the coal flame have been investigated and reported by Tian et al. (2010a).
Journal of Computational Multiphase Flows

Z. F. Tian, P. J. Witt, M. P. Schwarz and W. Yang

81

Convective
parts

Off-takes

Level 8

Vapour
burners

Level 7

Main
burners

Plane 1
Level 4

Firing group 5

Firing group 4

Normalised
distance 0
Normalised
distance 0

Line 1
Line 2
Normalised
distance 1

Figure 2. CFD geometry of the furnace, locations of ports, plane 1, line 1 (red line)
and line 2 (orange line).
In the current paper the performances of two turbulence models, k- model and SST model, and
two radiation models, DT model and P-1 model, are assessed for two cases where firing groups
4&6 are out-of-service and firing groups 7&8 are out-of-service.
4. EFFECT OF TURBULENCE MODELS
In a preliminary study (Tian et al., 2009), six two-equation RANS models were used to simulate a
non-swirling coal flame in a pilot-scale furnace. For that case predictions of the standard k- model,
RNG k- model, BSL and SST models were found to be generally in good agreement with the
experimental data. Predictions using the SST and BSL models were almost identical, while results
of the standard k- model and RNG k- model were also very similar. To understand the effect
turbulence models have of the fundamental gas-particle flow the standard k- model and SST
model were compared by Tian et al. (2010b) for isothermal gas-particle flow in three inclined
Volume 2 Number 2 2010

82

Modeling Issues in CFD Simulation of Brown Coal


Combustion in a Utility Furnace
Table 1. Raw brown coal properties and heat sink values for convective tube banks

Proximate analysis
Fixed carbon
Volatile matter
Ash
Ultimate analysis
C
H
N
S
O
Moisture (as received, w%)
Gross dry specific energy (MJ kg1)
Heat Exchanger
Superheater 1
Superheater 2
Superheater 3
Superheater 4
Reheater 1a
Reheater 1b
Reheater 2
Economizer

(db w%)
47.2
51.3
1.5
(daf w%)
66.44
4.5
0.57
0.19
28.3
66
26.2
Heat sink value (MW)
76.1
77.6
145.2
80.2
25.6
40.7
114.2
77.8

rectangular jets in crossflow. Geometry and conditions were scaled based on typical flow
conditions experienced in the Victorian brown coal furnace burners. The gas phase predictions were
validated against laser measurements of the gas phase velocity profiles along the centre plane of
the primary jet and secondary jet. Gas and particle flows predicted by both models were found to
be in reasonable agreement with the detailed experimental data, although the SST model showed a
slightly better agreement with the measurements than the standard k- model.
Whilst this previous work has indicated the effect of turbulence models on single coal
combustion and isothermal gas-particle flow in and near a burner jet, their effect on the complex
rotating flow field in a full scale furnace with combustion has not been assessed. To assess the
effect of turbulence models on the full furnace, results with firing groups 7&8 out-of-service
have been obtained with both the SST and k- turbulence models. Shown in Figure 3a is
predictions of the gas phase velocity component w along line 1, while Figure 3c is predictions
along line 2. The locations of line 1 and line 2 are given in Figure 2. The predicted velocity
profiles along line 1 based on the two turbulence models are quite close, while in Figure 3c
results from the standard k- model show a slightly higher jet decay than SST model in the burner
exit area. Further investigation found that the bulk flow fields predicted by the two turbulence
models in the furnace are quite similar. For example Figure 4 compares the predicted velocity
vectors on plane 1 for the case of firing groups 7&8 out-of-service based on two turbulence
models and the difference is small.
Comparisons between the predicted gas temperature for the two turbulence models along line 1
and line 2 are given in Figure 3b and Figure 3d, respectively. The predicted gas temperature
profiles on plane 1 are compared in Figure 5. It is evident that the two models give similar results
in the temperature field in the furnace. FGET predicted by k- turbulence model is 1074 C that is
very close to that of SST prediction, 1083 C.
A comparison of predicted wall incident heat flux profiles for the case of firing groups 7&8
out-of-service with measurements (Barlow, 2005) is shown in Figure 6. At the time of measuring,
the operating conditions of the unit were the same as those of simulation; the unit was operating
Journal of Computational Multiphase Flows

Z. F. Tian, P. J. Witt, M. P. Schwarz and W. Yang


(b)

1
0.8

Normalised distance along line 1

Normalised distance along line 1

(a)

83

Standard k-e model

0.6

SST model

0.4
0.2
0
18

12

0
6
6
Gas velocity w (m/s)

12

1
0.8
0.6

Standard k-e model

0.4

SST model

0.2
0
600

18

800

1000
1200
1400
Gas temperature (C)

1600

(d)
1
0.8

Normalised distance along line 2

Normalised distance along line 2

(c)

Standard k-e model

0.6

SST model

0.4
0.2
0
0

10

20
Velocity (m/s)

30

1
0.8
0.6
0.4

Standard k-e model

0.2

SST model

40

0
300

600
900
1200
Gas temperature (C)

1500

Figure 3. Comparison of different turbulence models for the case of firing groups 7&8
out-of-service: (a) gas velocity w along line 1, (b) gas temperature along line 1, (c) gas
velocity w along line 2, (d) gas temperature along line 2.
(a)

(b)
Burner 4

Burner 5
East wall

Burner 3

Burner 6

North
wall

South
wall

Burner 2

Burner 1

Burner 7
West wall

Burner 8

Figure 4. Predicted velocity vectors on plane 1 for the case of firing groups 7&8
out-of-service based on (a) standard k- model, (b) SST model.

at full load and burners 7&8 were out-of-service. Both measurement and prediction of wall heat
fluxes were taken through ports in the furnace walls. The heat flux for each level was the
average of the ports in all four walls for that level. Locations of the ports for the CFD calculation
are displayed in Figure 1. The difference between the predicted heat fluxes based on two
turbulence models is small.
Volume 2 Number 2 2010

84

Modeling Issues in CFD Simulation of Brown Coal


Combustion in a Utility Furnace

00

07

15

13

14
11

1
92

72

6
53

34

15

(b)

(a)

(C)

Figure 5. Predicted gas temperature on plane 1 for the case of firing groups 7&8
out-of-service based on (a) standard k- model, (b) SST model.

Boiler levels

Measurement

k-e model
5
SST model
4

Firing groups 7&8 out-of-service


3
0

200
100
Incident heat flux (kW/m2)

300

Figure 6. Comparison of calculated and measured wall incident heat flux profiles, firing
groups 7&8 out-of-service (Measurements from Barlow, 2005).

Journal of Computational Multiphase Flows

Z. F. Tian, P. J. Witt, M. P. Schwarz and W. Yang

85

5. EFFECT OF TURBULENCE DISPERSION


Turbulent dispersion of particles is handled by the stochastic method described both above and in
Gosman and Ioannides (1981). Truelove (1986) studied the influence of particle dispersion on the
CFD prediction of coal combustion in a swirling coal flame. It was found that for the swirling
flame, the effect of particle dispersion became increasingly more significant the lower the volatile
content of the coal. The effect of the particle dispersion on the prediction of the combustion of
Latrobe Valley brown coal in the furnace is investigated herein and is found to be small. For
example, the predicted wall incident heat fluxes for case of firing groups 4&6 out-of-service
with/without particle dispersion are illustrated in Figure 7. The difference between the predicted
heat fluxes with/without particle dispersion force is negligible. The predicted particle trajectories
from firing group 5 for the case of firing groups 4&6 out-of-service with particle dispersion are
shown in Figure 8a, while the trajectories without particle dispersion are shown in Figure 8b.
Many particles from the lower main burner (colored blue) circulate at the center of the furnace
following the center vortex that is formed by the gas and particle jets. Particles from the vapor
burners flow out of the furnace and into the convective pass without interacting with the centre
vortex. The trend of non-dispersed particle trajectories is very similar to that of dispersed particle
trajectories, though they are not coincident. Some dispersed particles from the lower main burner
travel into the ash hopper further than the non-dispersion case. This may explain the slightly
smaller heat flux at level 4 in Figure 7 with particle dispersion than that for the case without
particle dispersion

Boiler levels

Measurement
6

DT & dispersion
DT & non-dispersion

P1 & dispersion

Firing groups 4&6 out-of-service


3
0

100
200
Incident heat flux (kW/m2)

300

Figure 7. Comparison of calculated and measured wall incident heat flux profiles, firing
groups 4&6 out-of-service (Measurements Barlow, 2005).

Volume 2 Number 2 2010

86
(a)

Modeling Issues in CFD Simulation of Brown Coal


Combustion in a Utility Furnace
(b)

Figure 8. Predicted particle trajectories for the case of firing groups 7&8 out-of-service:
(a) with particle dispersion, (b) without particle dispersion.

6. SENSITIVITY TO RADIATION MODEL


Several categories of radiation models have been developed including zonal model, Monte Carlo
(MC) model, flux models (e.g. P-1 model) and hybrid model (e.g. DT model). Detailed review and
comparison of different models can be found in literature (e.g. Viskanta and Menguc, 1987;
Carvalho and Farias, 1998). Among these models, MC model, DT model and P-1 model are mostly
widely used for coal combustion in furnaces, e.g. Xu et al. (2001) made use of the DT model,
Vuthaluru and Vuthaluru (2006) used the P-1 model while Fan et al. (1999) employed the MC
method.
Monte Carlo models simulate the rays of radiation as photon trajectories and have been
recognised as the best method for modelling radiation (IIBD, 2002). However for coal-fired boilers,
the computational cost is extensive, making this model less attractive. An attempt was made as part
of this study to use an MC based radiation model for coal combustion in the furnace but it was
found to be less stable. Converged solutions could not been obtained for this problem and thus no
results for the MC method are presented.
The predicted wall incident heat flux for the case of firing groups 4&6 out-of-service calculated
using a DT and a P-1 radiation model are compared against power plant measurement in Figure 7.
Results from the P-1 model show that it considerably under predicts the incident heat flux up the
furnace. The reason for the under-prediction of P-1 model is not clear although this is under
investigation. As shown in Figure 9, the P-1 model predicts peak heat flux at the gas offtake for
firing group 4 (right hand side of east wall) which is out-of-service. While the DT model predicts
the peak wall incident heat flux at the location between the upper main burners and lower vapor
burners. This indicates that most of the combustion occurs at the burner level and is consistent with
the design concept and operation observations.
7. CONCLUSION
A CFD model of a 375 MW tangentially fired furnace that burns high-moisture brown coal has been
developed. A substantial amount of work has been undertaken to gain an understanding of the
sensitivity of CFD results to modelling parameters.

Journal of Computational Multiphase Flows

Z. F. Tian, P. J. Witt, M. P. Schwarz and W. Yang


(a)

87

(b)

East wall

50
85 0
00
12 0
75
17 00
00
21 00
25
0
25 0
50
29 00
75
34 00
00
38 00
25
42 00
50
00

South wall

42

West wall

North wall

(W m2)

Figure 9. Predicted wall incident heat flux: (a) DT model, (b) P-1 model.

Two turbulence models, the standard k- model and SST model, are used to model the gas phase
turbulence. Both turbulence models provide similar predictions that are in good agreement with the
plant data.
The effect of the particle dispersion on the CFD prediction is found to be insignificant. This is
consistent with the observation of Truelove (1986) who found the particle dispersion has little
effect on the high-volatile coal combustion although in a smaller and simpler combustion
environment. When turbulent dispersion is included it is found the computing effort in the
simulation in this furnace model is much greater than that without particle dispersion. As the results
show little or no change when turbulent dispersion is included is therefore found that when large
computational resources are required, such as sensitivity studies, the model can be run without
turbulent dispersion without compromising the solution.
The predicted wall incident radiation flux calculated using both the DT radiation model and
P-1 model were compared against power plant data. The comparison revealed that DT model
provides good prediction of the radiation profiles, while the P-1 model considerably under predicts
the wall incident radiation flux. However, this finding is inconsistent with that of Marakis et al.
(2000) who observed that the P-1 model gave results that were in satisfactory agreement with those
from a Monte Carlo model in cylindrical coal-fired furnaces. This may indicate that the radiation
models are not universal and are more likely to be case sensitive and that further investigation into
the P-1 models failure in this case is needed.
ACKNOWLEDGEMENTS
The financial support provided by the Victorian Government Department of Primary Industries
under the ETIS Program is gratefully acknowledged. The authors also thank Steve Pascoe and
Yorrick Nicolson of TRUenergy Yallourn power plant for providing plant data for the CFD
simulation and discussion about the predicted results.

Volume 2 Number 2 2010

88

Modeling Issues in CFD Simulation of Brown Coal


Combustion in a Utility Furnace

REFERENCES
ANSYS/CFX 12.0, ANSYS Inc., the USA, 2009.
Barlow, M. HRL Report No. HLC/2005/051, HRL Limited, Australia, 2005.
Bris TL, Cadavid F, Caillat S, Pietrzyk S, Blondin J, Baudoin B. Coal combustion modelling of large power plant for NOx
abatement. Fuel 2007;86: 22132220.
Carvalho MG, Farias TL. Modelling of heat transfer in radiating and combusting systems. Transactions in IChemE, 1998;
76: 175184.
Duong TH. Modelling of Brown Coal Combustion in One-dimension, NERDDP Project 931, End of grant Report, Report
No ND/87/040, the State Electricity Commission of Victoria, Australia, 1987.
Fan JR, Sun P, Zha XD, Cen KF. Modeling of combustion process in 600 MW utility boiler using comprehensive models
and its experimental validation. Energy & Fuels 1999;13:10511057.
Gosman AD, Ioannides, E. Aspects of computer simulation of liquid-fuelled combustors. AIAA Papers 81-0323, 1981.
IIBD, Improving industrial burner design with computational fluid dynamics tools: Progress, Needs, and R&D Priorities.
Workshop Report, Department of Energys Office of Industrial Technologies (OIT) and the Sandia National
Laboratories (SNL), the USA, 2002.
Launder E, Spalding DB. The numerical computation of turbulent flows. Computer Methods in Applied Mechanics and
Engineering, 1974; 3: 269289.
Lockwood FC, Shah NG. A new radiation solution method for incorporation in general combustion prediction procedures.
In Eighteenth Symposium (International) on Combustion, The Combustion Institute, 1981; 14051414.
Marakis JG, Papapavlou C, Kakaras E. A parametric study of radiative heat transfer in pulverised coal furnace. International
Journal of Heat and Mass Transfer 2000;43:29612971.
Menter FR. Two-equation eddy viscosity turbulence models for engineering applications. AIAA Journal 1994;
32: 15981605.
Niska S. Coal combustion modelling. IEA Coal Research. 1996.
Sazhin SS, Szhina EM, Saravelou OF, Wild P. The P-1 model for thermal radiation transfer advantages and limitations. Fuel
1996;75: 289294.
Schiller L, Nauman A. Uber die drundlegende berechnung beider schwekraftaufbereitung. Ver Deutch Ing 1933;
44:318320.
Tian ZF, Witt PJ, Schwarz MP, Yang W. Comparison of two equation turbulence models in simulation of a non swirl coal
flame in a pilot scale furnace. Combustion Science and Technology 2009;181: 954983.
Tian ZF, Witt PJ, Schwarz MP, Yang W. Numerical modeling of Victorian brown coal combustion in a tangentially-fired
furnace, accepted by Energy & Fuels, 2010a.
Tian ZF, Witt PJ, Yang W, Schwarz MP. Numerical simulation and validation of gas-particle rectangular jets in crossflow,
submitted, 2010b.
Truelove JS, Holcome D. Measurement and modelling of coal flame stability in a pilot-scale furnace. In Twenty-Third
Symposium (International) on Combustion, The Combustion Institute, 1990, 963971.
Truelove JS. Prediction of the near burner flow and combustion in swirling pulverized coal flames. Proceedings of
Combustion Institute 1986; 21: 275284.
Tu JY, Fletcher AJ, Behnia M. Numerical modeling of three-dimensional fly-ash flow in power utility boilers. International
Journal for Numerical Methods in Fluids 1997;24:787807.
Viskanta R, Menguc MP. Radiation heat transfer in combustion systems. Progress in Energy and Combustion Science
1987;13: 97160.
Vuthaluru R, Vuthaluru HB. Modelling of a wall fired furnace for different operating conditions using FLUENT. Fuel
Processing Technology 2006;87: 633639.
Wall TF, Phelan WJ, Bartz S. Doc. F388/a/3, International Flame Research Foundation, The Netherlands, 1976.
Xu M, Azevedo JLT, Carvalho MG. Modelling of the combustion process and NOx emission in a utility boiler. Fuel
2001;79: 16111619.

Journal of Computational Multiphase Flows

Вам также может понравиться