Applied Thermal Engineering 154 (2019) 87–101

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Research Paper

Numerical investigation of combustion optimization in a tangential firing T

boiler considering steam tube overheating
⁎
Cong Yua, Wei Xionga, Huan Maa, Jianxin Zhoua, Fengqi Sia, , Xiaoming Jiangb, Xuwen Fangc
a
Key Laboratory of Energy Thermal Conversion and Control of Ministry of Education, School of Energy and Environment, Southeast University, Nanjing, 210096, PR
China
b
Datang Nanjing Environmental Protection Technology Co., Ltd, Nanjing, 210096, PR China
c
Maanshan Dangtu Power Generation Co., Ltd, Maanshan 243102, PR China

H I GH L IG H T S

• AA comprehensive boiler model that coupled the furnace and steam sides was developed.
• Thedetailed mapping method was provided for the grid systems of different models.
• The effects of the SOFA tilt angle on boiler efficiency and NO emissions were studied.
x

• Combustion
effects of the SOFA tilt angle on the tube outer surface temperature were studied.
• optimization was conducted under the constraint of tube overheating.

A R T I C LE I N FO A B S T R A C T

Keywords: A comprehensive numerical study of the combustion optimization of a 660-MW tangentially fired boiler under
Tangential firing boiler the constraint of steam tube overheating is presented. To obtain the chemical products and the heat flux at the
Coupled simulation heating surfaces, a three-dimensional (3D) combustion model was established using computational fluid dy-
Combustion efficiency namics (CFD) software. In addition, a 1D + 1D thermal hydraulic model was developed using MATLAB to de-
NOx emission
scribe the heating process inside the tubes. Detailed mapping methods were also provided to connect the mesh
Steam tube overheating
systems in CFD and MATLAB. As a result, the tube wall temperature can be calculated by considering the coupled
heat transfer on the gas and steam sides. The simulation results reveal that under full load conditions, the
upward tilting of the separated over-fire air (SOFA) burners can reduce the formation of CO and NOx, which
helps achieve higher boiler efficiency. However, when the SOFA tilt angle increases from +20° to +30°, 0.16%
of the outer surfaces of the final re-heater would exceed the allowable temperature of the TP347H material;
additionally, the improvement in the boiler efficiency and decrease in NOx formation are much less. Thus,
considering the unit’s safe operation, the optimal SOFA tilt angle for combustion optimization should be +20°.

1. Introduction threatening the unit’s safe operation. According to the statistics, ap-
proximately 40% of forced outages in power stations have been caused
Tangentially fired systems are the most widely used technology in by boiler tube failures due to material overheating [6]. Recent studies
utility boilers. In this type of system, the four corners of the furnace have suggested alleviating this gas deviation by adjusting boiler settings
have separate burners, from which the fuel and air jet flow are injected such as the opposing tangential angle of the air jets [7] and the vertical
into the boiler, forming a concentric swirling fireball [1] that provides burner tilt angle [8]. However, variations in these settings can also
good flame stability and high combustion efficiency. However, this affect the boiler thermal efficiency and NOx formation by changing the
combustion method causes an unexpected residual swirl at the furnace environment in the furnace [9–13], such as the turbulence intensity, the
exit, which can induce gas velocity and temperature deviations in the oxygen concentration distribution, and the residence time of the coal
horizontal pass [2,3]. Several studies [4,5] have shown that this de- particles. Thus, there exists an inherent connection between the com-
viation might accelerate steam tube rupture by local overheating, bustion performance and the overheating characteristics of steam tubes,

⁎
Corresponding author at: School of Energy & Environment, Southeast University, No. 2 Sipailou Road, Nanjing 210096, PR China.
E-mail address: [email protected] (F. Si).

https://doi.org/10.1016/j.applthermaleng.2019.03.074
Received 8 November 2018; Received in revised form 26 February 2019; Accepted 17 March 2019
Available online 18 March 2019
1359-4311/ © 2019 Published by Elsevier Ltd.
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

Nomenclature y y-coordinate (m)

A cross-sectional area (m2) Greek symbols

d diameter (m)
ET deviation factor α heat transfer coefficient (W·m−2·K−1)
f friction factor β angle of heater subtube
g gravity acceleration (m·s−2) ε roughness parameter (mm)
h enthalpy (kJ·kg−1) λ thermal conductivity (W·m−1·K−1)
L tube pitch (m) θ inclination angle of the steam tubes
m mass flow rate (kg·s−1) ρ density (kg·m−3)
MT nonuniformity coefficient ψ swirling momentum intensity (N·m)
M number of MATLAB discrete nodes Δs spatial size of control volume (m)
N number of CFD cells
Nu Nusselt number Subscripts
p fluid pressure (MPa)
pf frictional pressure (MPa) aux index of auxiliary cells
Pr Prandtl number b bulk
q heat flux (W·m−2) equiv equivalent
qr radiative heat flux (W·m−2) g flue gas side
r radius (m) i index of MATLAB discrete nodes
R equivalent radius (m) in inner
Re Reynolds number j index of CFD cells
S area of the unit cell (m2) k index of sub-tubes
T temperature (K) out outer
U circumference (m) r radiative
w tangential velocity of flue gas (m·s−1) w wall
x x-coordinate (m)

and this relationship is more closely for the modern boilers equipped Most researchers treat the heat exchanger inside the boiler as either a
with air-staging and deep-air-staging combustion systems [14–16]. group of surfaces with a uniform temperature distribution
To date, few studies have conducted integrated investigations on the [1,3,4,8,9,10,22,30] or a porous region with a homogeneous heat
boiler combustion economy, NOx emissions, and outer surface tem- source [2,5,20]; however, this method prevents the CFD model from
perature of steam tubes. Thus, the existing literatures cannot answer the calculating the actual temperature distribution on the heating surfaces.
question of how the possibility of tube overheating changes when Several recent studies have focused on the coupled heat transfer of
conducting combustion optimization. This problem might be attribu- combustion and fluid heating. Park et al. [31] and Schuhbauer et al.
table to the lack of an effective method that can evaluate the tube [32] separately used the commercial software PROATES and APROS to
overheating condition. In power plants, the number of thermocouples establish a one-dimensional (1D) model of the steam-water side, which
for monitoring outer surface temperature of the tubes is limited, and the was coupled with a 3D CFD model on the furnace side. These models
measuring points are usually located at the outlets of the tubes to iso- can consider the heat absorptions of all the heaters and can be used for
late from the high-temperature and high dust environment inside the the analysis of whole power plants. However, since the model on the
boiler; there is no guarantee that the maximum temperature and dan- steam-water side in these simulations applied the lumped parameter
gerous points on the heating surfaces can be monitored. Therefore, a method, the tube wall temperature along its path can be obtained only
comprehensive model is desired that not only can describe the complex by the linear interpolation of the inlet and outlet parameters. Yang et al.
physical and chemical processes within a furnace, but also have the [29] simulated the combustion and fluid heating process of a CO2 boiler
ability to predict the positions and cases in which overheating is more by combining two distributed parameter models and successfully cal-
likely to occur. culated the tube wall temperature distribution; however, the mapping
In the past decade, computational fluid dynamics (CFD) simulations method of the two grid systems was not reported. In addition, the re-
have been extensively applied in predicting the aerodynamic fields, search focused on 1D wall heaters, whereas the final re-heater and
temperature distributions and combustion products of utility boilers. super-heater with higher steam parameters and more complex struc-
Although some problems still exist in terms of NOx formation modeling, tures were not discussed.
the widely used models, consisting of Zeldovich’s thermal NOx model The aim of this study is to provide detailed information on the
[17] and De Soete’s Fuel NOx model [18], are proven to have a relative coupling of the distributed parameter models of both the furnace and
satisfactory approximation by estimating and tuning some simulation water-steam sides of a boiler; a more practical motivation is to si-
parameters [19–23]. However, modeling three-dimensional (3D) multaneously analyze the targets of combustion optimization (i.e.,
steam-water processes presents a major problem. The 3D models on the boiler efficiency and NOx emissions) and the overheating problem of
steam-water side can consider the influence of fouling by describing the steam tubes. In the simulation, a 3D CFD model was established to
process of ash deposition and oxide growth [24–27]; however, this kind describe the complex physical and chemical processes on the flue gas
of models can be hardly applied to simulate the heat transfer in one or (furnace) side, while a 1D + 1D thermal hydraulic model was devel-
more heat exchangers. In CFD models, the meshes are too large to oped using MATLAB to consider the fluid heating process. Considering
generate the actual structure of the heat exchangers [28,29] because the shape characteristics of different heating surfaces, detailed mapping
the mesh scale of a single tube is inconsistent with that of the entire methods were proposed separately for structured and unstructured
boiler flue. Qi et al. [24] established a 3D model of only one serpentine meshes in CFD with discrete nodes in MATLAB. In addition, the related
flow composed of 12 U-tubes in a final super-heater, and the numerical timer function, user-defined memory (UDM), and user-defined func-
grids exceeded 31 million. Thus, certain simplifications must be made. tions were applied to ensure that the tube wall temperature can be

88
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

calculated iteratively in CFD and MATLAB. A grid independence test 2.2. Experimental conditions
and a series of on-site measurements were also conducted for numerical
validation. Furthermore, the present study discusses the selection of the To validate the simulation results and support the analysis of the
optimal separated over-fire air (SOFA) vertical tilt angle by considering boiler characteristics, a series of on-site experiments were performed on
combustion optimization and the limitations of the tube wall tem- the boiler. The basic experimental case (Case 1) was at the full load
perature. condition. Huainan (HN) bituminous coal was burned with a total mass
flow rate of 282 t/h and was equally provided to mills ABCDEF. The
2. Experiments primary air had a total mass flow rate of 630 t/h and a temperature of
351 K. The secondary air flow rate was 1670 t/h at 581 K, and the SOFA
2.1. Tangentially fired pulverized coal boiler took up 35.8% of the air. The vertical tilt angles of the primary and
secondary air nozzles in the main burner zone were −5° and −7.5°,
The tangentially fired pulverized coal boiler considered in this study respectively, while the SOFA vertical tilt angle was −12°. On the water-
is an SG-2090/25.4-M975 type with supercritical pressure, one middle steam side, the mass flow rates of SH and RH were 1955 t/h and 1632 t/
reheater and steam power circulation in a 660-MW unit. The super- h, respectively. The inlet temperature and pressure of the water wall
heater (SH) has a capacity of 2090 t/h at a design pressure of 25.4 MPa heaters, the division SH, and the final RH were 596 K/26.9 MPa, 687 K/
and a design temperature of 571 °C, while the design temperature of the 25.9 MPa, and 733 K/4.4 MPa, respectively.
re-heater (RH) can reach 569 °C. As shown in Fig. 1(a), the division SH Case 2, Case 3, and Case 4 set the SOFA tilt angle to +30°, 0°, and
and platen SH are placed at the upper furnace, while the final RH and +20°, respectively, and the other boiler settings were identical to those
the final SH, with higher steam parameters, are separately arranged at in Case 1. For each case, great effort was made to assure minimum
the inlet and outlet of the horizontal crossover pass. The furnace walls variations in the boiler operating conditions and coal properties
from the hopper bottom to a height of approximately 41.4 m are spiral [14–16]. During the experiments, the dry-bulb and wet-bulb ambient
water walls, and the other walls above the furnace are vertical water temperatures were 278 K and 277 K, respectively.
walls. The economizer and primary RH can be found in the rear pass.
As shown in Fig. 1(b), a low-NOx concentric firing system (LNCFS) 2.3. In situ measurements
made of six groups of primary air nozzles (A–F), six groups of secondary
air nozzles, two groups of CCOFA nozzles and five groups of SOFA Fig. 2 shows the measurement point arrangement in the boiler, and
nozzles was installed in the furnace. The secondary air burners consist the main tested parameters can be divided into three parts.
of CFS (AI/AII–FI/FII) and auxiliary (AA–EF) air nozzles. As shown in The parameters of the combusted coal were obtained. The raw coal
Fig. 1(c), the CFS air nozzles are biased 25° relative to the jet direction was sampled before entering the mills, and the parameters of the
of the primary air, and the opposing tangential angle of the SOFA jets is proximate and ultimate analysis were tested, as listed in Table 1. The
fixed at −12°. Notably, the yaw angles in the horizontal direction pulverized-coal finenesses R90 and R200 were measured using different
cannot be adjusted in actual operation, whereas the SOFA vertical tilt sizes of sieves, and the values are 33.6% and 3.3%, respectively.
angle can be changed from −30° to +30°. According to the boiler in- To validate the simulation results of Case 1, the gas temperature was
structions, approximately 30% of the combustion air is supplied to the measured at the final RH exit using two K-type thermocouples, both
furnace through the SOFA nozzles. with a measurement uncertainty of 2.47 K. The concentrations of O2,

SOFA-V
SOFA-IV
SOFA-III
Final SH SOFA-II
SOFA-I
Primary SH
CCOFA-II
CCOFA-I
FII
Final RH F
Division SH FI
Platen SH EF
Economizer EII CFS air SOFA
E
EI Fuel and primary air
DE Left
Furnace DII
D
DI 51°
CD
CII 12°
C 12°
Front wall

Rear wall

CI
BC
BII
B
BI 12°
12°
AB
AII 51°
A
AI
AA Right
(a) (b) (c)
Fig. 1. Schematics of the (a) heat exchangers, (b) burners, and (c) yaw angles.

89
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

Tube wall temperature

Formed NO, CO, and O 2

Raw coal sample Gas
temperature
SCR
Pulverized-coal sample

Burner Burner
APH

Mills

ESP

Slag sample
Fly ash sample
Fig. 2. Measurement point arrangement in the boiler.

Table 1 final SH were tested by K-type thermocouples.

Properties of the pulverized coal. Finally, to analyze the boiler efficiency of Case 1 to Case 4, the slag
Proximate analysis (ar1, %) Ultimate analysis (ar1, %) Qnet,ar
and fly ash were sampled from the bottom of the furnace and the
(MJ/ hopper of the electrostatic precipitator (ESP), respectively, to determine
w(M) w(V) w(A) w(FC) w(C) w(H) w(O) w(N) w(S) kg) the content of combustible matter in them.

7.48 24.34 23.73 44.45 56.4 3.18 8.01 0.64 0.55 22.447
3. Methods and modeling
1
As received basis.
3.1. Flue gas side modeling
CO, and NO were tested by a Testo 350 gas analyzer, and the measured
value is the average of eight points evenly spaced at the entrances of the 3.1.1. Calculation domain and mesh system
two SCR reactors. The accuracy of the O2 sensor is ± 0.8% of the full- The calculation domain was established from the ash hoppers to the
scale value (fsv), while the CO and NO sensors can be accurate to ± 5% entrances of two selective catalyst reduction (SCR) reactors, as shown
of the measured value (mv). The heat absorptions of the division SH, in Fig. 3(a). In the simulation, the division SH, platen SH, final RH, and
the platen SH, the final RH, and the final SH were calculated according final SH are high-temperature heat exchangers that are treated as
to the pressure, temperature and mass flow rate at the inlet and outlet groups of surfaces. The primary RH and economizer are regarded as
of each heat exchanger, which were obtained from the plant on-line porous media. In Fig. 3(b), which shows the mesh system, it can be seen
measuring points and distributed control system (DCS). The tube outlet that most of the calculation domain, such as the furnace, the rear pass,
wall temperatures at the outermost tube coilers of the final RH and the and the regions containing the division SH, the platen SH and the final

Platen SH Final RH Final SH Y [m] Structured grids Unstructured grids

65
60
55
Division SH Primary RH
50

Economizer 45
Nose 40
35
SOFA 30
25
20

Rear wall 15
Right wall 10
5
X [m]
0
-15 -10 -5 0 5 10 15 20 25 30 35
(a) (b)

Fig. 3. (a) Geometric model and (b) mesh system of the calculation domain.

90
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

RH, was meshed with structured hexahedral cells. Unstructured tetra- dp dp m2 d ⎛ 1 ⎞

=− f − 2 ⎟ − ρg sin θ
hedral cells are used for the region of the final SH due to the complex
⎜
dsk dsk A dsk ⎝ ρ ⎠ (2)
shape of the heating surface. In the furnace, mesh refinements were
conducted in the neighborhood of the burners where the combustion dh 1 dp 1 dpf U
actively takes place. = + + qequiv
dsk ρ dsk ρ dsk m (3)

3.1.2. Pulverized coal combustion model The 1D calculation domain of the fluid is divided into M discrete
A commercial CFD code, ANSYS FLUENT 14.5, was used to simulate nodes. For Eqs. (1)–(3), the backward differencing scheme was adopted
the pulverized coal combustion, considering several closely coupled to discretize the spatial derivative terms of m, p, and h, and the frac-
subprocesses, including turbulent flow, coal particle transport and tional pressure drop was calculated using the Darcy–Weisbach formula
combustion, gas phase combustion, and multiple forms of heat transfer. [29,37]; then, the governing equations can be written as Eqs. (4)–(6).
In this study, the time-averaged conservation equations for mass, mo- mi = mi − 1 = const;i=2, ⋯, M (4)
mentum, enthalpy and species were solved to predict the velocity,
temperature and species concentration within the boiler. The turbulent
flow was calculated by the realizable k-ε model due to the strongly pi = pi − 1 −
m2
A2 ( 1
ρi
−
1
ρi − 1 )− fi | m | m
2din A2 ρi
× Δsk + Δyk ρi g sin θ
swirling flow inside the furnace. Coal particle transport was formulated i = 2, ⋯, M (5)
in the Lagrangian framework, and the stochastic particle trajectory
(SPT) model was introduced to consider the effect of instantaneous hi = hi − 1 +
1
(p − pi − 1 ) +
fi | m | m
×
Δsk U
+ Δsk m qequiv,i
ρi i 2din A2 ρi
turbulent fluctuations of gas on the particle trajectories. Coal devola- ρi

tilization was modeled by the two-competing-rates model proposed by i = 2, ⋯, M (6)

Kobayashi [33], and a non-premixed combustion model, in which the
The friction factor fi can be calculated using the Prandtl–von
chemical reaction rate is determined by a probability density function
Kármán formula [24]:
(PDF) table, was adopted to calculate the turbulent combustion of the
volatile gases [34]. The rate of char oxidation is dependent on the ex- 1
fi =
din 2
ternal diffusion rate of oxygen to the char surface and the intrinsic
reaction rate, which was computed according to the model of Field
(
4 log 3.7 ε ) (7)
[35]. The P-1 model, which considers the radiation scattering effect and where ε is a roughness parameter expressed in units of height, which
is suitable for combustion equipment with a large optical thickness and is 0.06 mm and 0.008 mm for the ferrite and austenite tubes, respec-
complex geometry, was selected for radiation modeling. The absorption tively.
coefficients of the gas phase were calculated by the weighted-sum-of- For Eqs. (4)–(6), considering that the tube inlet parameters are
gray-gases model (WSGGM) [36]. given as boundary conditions, there are 4 M – 3 unknown variables and
3 M – 3 equations. Since this study focuses on the full load operation
3.1.3. NOx formation model condition in a supercritical boiler, superheated steam was used as the
After the combustion calculation, the formation of NOx was mod- flow media in the high-temperature heat exchangers. Thus, the single-
eled as a postprocessing procedure. Due to its miniscule proportion in phase flow model expressed by Eqs. (1)–(7) is appropriate, and the fluid
the coal-fired boiler, prompt NOx was ignored in the simulation, and density at each node can be directly calculated according to the pres-
only thermal NOx and fuel NOx were predicted. Fig. 4 shows the NOx sure and enthalpy at the same node using the equations of IAPWS-IF97
formation reactions considered in this work. The thermal NOx, gener- [24]. Therefore, the number of unknown variables becomes 3 M – 3,
ated from the direct oxidation of N2, was simulated by the extended and the fluid parameters along the tube can be solved.
Zeldovich mechanism [17], as shown in reaction (1). The concentra- The convective heat transfer coefficient βi between the inner wall
tions of [O], [H], and [OH] in this reaction were calculated using the and steam-water fluid can be calculated by the Dittus model [38], as
partial equilibrium approach. Fuel NOx was calculated by the De Soete shown in Eq. (8). Then, the fluid temperature in the bulk phase Tb,i and
model [18], which assumes that the fuel-bound nitrogen was both in the tube inner and outer wall temperatures Tin,i and Tout,i can be com-
the volatile matter and the char. Many studies [20,22] have shown that puted by the energy equations inside the wall body, which are ex-
volatile N is mainly released as the intermediate HCN and that the rest pressed by Eqs. (9) and (10).
evolves as NH3 for bituminous coal; the proportion of HCN was 0.9 in
this work. These intermediate species competitively oxidize to NO or Nui λ 0.023Rei0.85 Pri0.4 λ
αi = =
are reduced to N2, as shown in reactions (3)–(8). As recommended by d in d in (8)
many researchers [1,8], approximately 60% of the char-bound N was
qequiv,i
directly transferred into NO, as NO can also be reduced by char parti- Tin,i = Tb,i +
cles through a heterogeneous reaction, as shown in reactions (2) and αi (9)
(9). In addition, NOx can be reacted with hydrocarbons through a re-
burning process [20], as shown in reaction (10).

3.2. Steam-water side modeling

3.2.1. Thermal hydraulic model in a single tube

In the present study, the circular tube was simplified to a 1D
structure, and the nonuniform heat flux distribution on the tube outer
surface of the same cross-section was neglected. To acquire the spatial
distributions of the fluid flow rate, pressure, and enthalpy in a single
tube, the 1D governing equations of mass, momentum and energy
conservation at steady state [29,37] can be written as follows:
dm
=0
dsk (1) Fig. 4. Illustration of the relative reactions in the NOx formation process.

91
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

qequiv,i d in d out,i rin L

Tout,i = Tin,i + ln
2λ d in,i (10)
rout

3.2.2. Wall temperature calculation for heat exchangers ǻsk Tb,i Tin,i Tout,i
This study mainly focuses on the overheating problem of high-
temperature heat exchangers, while the upstream water wall heaters in
the furnace were also considered in MATLAB modeling by the methods qj+qj,shadow
of Yang et al. [29] and Zima [37]. The high-temperature heat ex-
changers consist of the division SH, the platen SH, the final RH, and the
final SH, in which superheated steam is the flow media. As shown in
Fig. 5, the U-tubes in these heaters can be divided into several subtubes,
such as subtube k = 1 and subtube k = 2. Thus, the steam heating
process in a single U-tube can be regarded as a 1D + 1D problem, which Fig. 6. Illustration of the equivalent heat flux in MATLAB equations.
is described as follows:
x i, k = x i − 1, k + Δsk × sin β (11) node i, which is expressed as follows:

yi, k = yi − 1, k − Δsk × cos β 2πrin Δsk × qequiv,i = −L × Δsk × (qj + qj,shadow ) (13)
(12)

where xi,k and xi-1,k are the x coordinates of the discrete nodes i and i−1 L × (qj + qj,shadow )
qequiv,i = −
in subtube k, respectively; yi,k and yi−1,k are the y coordinates of the 2πrin (14)
discrete nodes i and i−1 in subtube k, respectively; and β is the angle
where L is the tube pitch; rin is the tube inner radius; j and j, shadow are
measured counterclockwise from the negative y-axis to the direction of
the CFD cells to which MATLAB discrete node i belongs on both sides of
fluid flow (e.g. β of subtube k = 1 is 0°, and β of subtube k = 2 is 90°, as
the heating surface; and qj and qj,shadow are the heat fluxes in CFD cells j
shown in Fig. 5).
and j, shadow, respectively.
Thus far, the outer wall temperature of all subtubes in the high-
qj is the total heat flux from the wall to the adjacent flue gas cell and
temperature heating surfaces can be calculated by solving Eqs. (4)–(12).
can be computed with the local convective and radiative heat fluxes:
However, there is still a key unknown parameter—the equivalent heat
flux qequiv,i. In the MATLAB modeling, qequiv,i is the heat flux that the qj = αg,j × (Tw,j − Tg,j1) + qr,j (15)
fluid in the bulk phase absorbs at discrete node i, as shown in Fig. 6.
According to the law of energy conservation [29,37], qequiv,i can also be Tw,j = ∑ Tout,i/Mj
calculated by the heat that the flue gas transfers to the outer wall at i∈j (16)

q aux,j

k =1 q j,i

qj T out,i

T out,i+1
y
x

k =2
(a ) (b )
Fig. 5. Mapping method between the nodes in MATLAB and meshes in CFD.

92
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

where βg,j is the local heat transfer coefficient on the flue gas side, inside the pipelines was solved by MATLAB code. For the simulation of
which is computed by local flow field conditions such as the turbulence the whole boiler at steady state, the amount of heat that flue gas re-
level, temperature, and velocity profile [39]; Tw,j is the wall tempera- leases to the tube outer surface is equal to the quantity of heat that the
ture of CFD cell j, which is the average value of the tube outer wall steam absorbs from tube outer wall, and each point on the heating
temperatures for all MATLAB discrete nodes in this cell; Tg,j1 is the flue surfaces should satisfy this conservation relationship. Therefore, the
gas temperature in cell j1 neighboring j; qr,j is the radiative heat flux, strategy that calculates the outer surface temperature of the heat ex-
which is calculated by the P-1 model; and Mj is the number of MATLAB changers can be described as follows.
discrete nodes that belong to CFD cell j. The distribution of the tube outer surface temperature was first
assumed as the initial boundary condition in Fluent. After the flue gas
3.2.3. Mapping method between CFD and MATLAB meshes model was solved and achieved a convergent solution, Fluent delivers
In the CFD model, the serpentines, namely, heating surfaces com- the distribution of the heat flux on each heating surface to MATLAB,
posed of a group of tubes, of the division SH, the platen SH, and the and this distribution is treated as the boundary condition of the thermal
final RH were meshed with quadrilateral cells. The nodes of the CFD hydraulics model. Thus, the outer wall temperature of each tube can be
cells have equal intervals along the X- and Y-directions, as shown in calculated by MATLAB again and in turn serves as the boundary con-
Fig. 5(a). Thus, by calculating the numbers of intervals in the X- and Y- dition in Fluent. When the calculated distribution of the tube outer wall
direction between the MATLAB node and the top-left point of each temperature has almost no change compared with last time, the heat
serpentine, the CFD cells that a certain MATLAB node belongs to can be transfer of all points on the heating surfaces is considered to have
directly determined. reached the balance state, and the coupled calculation is completed.
However, the serpentines of the final SH have a complex structure Based on this strategy, Fig. 7 shows the detailed procedures for
and are meshed with triangular cells. To search the corresponding CFD Fluent and MATLAB during the coupled calculation, and the *.txt files
cells, there will be a high computational resource cost to compare the were chosen as the interface for the communication of these two soft-
spatial distances of each MATLAB node and all CFD cells. This com- ware programs. The main simulation process in Fluent can be described
parison will significantly decrease the computing speed of the MATLAB as follows. (1) According to the assumed wall temperature, calculate
model and affect the coupled simulation. Therefore, to promote the the pulverized coal combustion model in Section 3.1.2 until the gas-
efficiency of recognizing CFD cells, a set of auxiliary mesh systems, solid flow, temperature, and species concentration fields achieve a
which has a larger rectangular domain with equidistant quadrilateral convergent solution. (2) Run the user-defined function DEFI-
cells, as shown in Fig. 5(b), was established in MATLAB. Instead of NE_ADJUST, which was developed to write a heatflux.txt file every 120
directly mapping with the original CFD cells, it is simpler for MATLAB steps. This file serves as the boundary condition of the steam-water side
nodes to obtain the heat flux in the auxiliary cell qaux,j, which is the models in MATLAB. (3) After 25 steps of this operation, read the
area-weighted average of the inside triangular CFD cells qj,i. For one walltemp.txt file calculated by MATLAB. (4) Apply UDM to split the
serpentine meshed with M MATLAB nodes and N triangular CFD cells, original string in the walltemp.txt file into different cell arrays, which
this method can reduce the number of loops for linking the corre- are assigned to the corresponding serpentines as new temperature
sponding meshes from M × N to M + N. boundary conditions of Fluent using the DEFINE_PROFILE macro. (5)
Continue to solve the model on the flue gas side, and go back to step (2)
3.3. Coupled procedures of CFD and MATLAB until the maximum relative error of the wall temperature can be con-
trolled within 10−3. (6) Calculate the NOx formation model in Section
As mentioned above, the software Fluent was applied to establish 3.1.3.
the model of the boiler flue gas side, and the thermal hydraulics model The following are the calculation steps in MATLAB: (1) develop a

DEFINE_ADJUST DEFINE_PROFILE Models on flue gas side

If Step=120 If Step1=25 Split string and assign arrays using UDM

Yes Yes
export import
Fluent
heatflux.txt walltemp.txt End

If heatflux.txt exist export

Yes No Yes
If T=10min import If max{RE(Tw,j)}<10-3

Timer function Mapping function for MATLAB nodes and CFD cells

Thermal hydraulics model Tube wall model

MATLAB
Fig. 7. Coupled procedures for CFD and MATLAB.

93
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

timer function that examines whether Fluent has exported the heat- Table 2
flux.txt file every 10 min; (2) if the file exists, assign arrays for the CFD Comparisons between the predicted results and the measured results.
cells of each serpentine; (3) discretize the tubes in each serpentine by Parameter Unit Measurement point Plant data Simulation
Eqs. (11) and (12), and link the MATLAB nodes with the corresponding
CFD cells using the method proposed in Section 3.2.3; (4) calculate the Gas temperature K Final RH exit 962 1027
O2 (molar fraction) % SCR entrance 2.92 2.89
equivalent heat flux qequiv,i for each MATLAB node by Eq. (13); (5) solve
CO (molar fraction) ppm SCR entrance 575 593
Eqs. (4)–(10) to obtain the outer wall temperature of the tube along the NO mg/Nm3 SCR entrance 349 338
path; and (6) calculate the wall temperature in the CFD cells by Eq. Heat absorption MW Division SH tube bank 140.3 146.1
(16), and write the walltemp.txt file. Platen SH tube bank 105.6 109.7
Final RH tube bank 109.5 107.3
Final SH tube bank 109.3 113.4
3.4. Modeling cases and boundary conditions

The simulation cases are identical to the experimental cases, i.e., along line L1 (Z = 0 m, X = −9.072 to 9.072 m) on the cross-section
Case 1 to Case 4, which set the SOFA vertical tilt angle to −12°, +30°, that is 0.16 m above the B layer pulverized coal burner. It is clear that
0°, and +20°, respectively. The inlet boundary condition at each burner unlike the simulation results of the grid system with 2.0 million cells,
was set based on the operational data presented in Section 2.2. The the flow and temperature fields calculated by the grid systems with 2.8
boiler wall was regarded as a no-slip and no-quality penetration and 4.2 million cells are almost identical. Therefore, to ensure the ac-
boundary condition with an internal emissivity of 0.7 [8]. During the curacy of the numerical simulation while considering the computing
simulation, the wall temperatures of the water wall heaters, the division cost, the medium mesh with 2.8 million cells is selected in this work, as
SH, the platen SH, the final RH, and the final SH were initially set as it can provide sufficient computational efficiency.
650 K, 723 K, 773 K, 823 K, and 843 K, respectively, and then the actual After determining the mesh numbers, the coupled simulation was
temperature distributions can be obtained through the coupled simu- performed. Table 2 compares the predicted and measured parameters
lation. The coal properties was provided in Table 1. According to the on the flue gas and steam-water sides of Case 1. Apart from the gas
measured pulverized-coal fitness values of R90 (33.6%) and R200 temperature, the relative errors (REs) of the other seven parameters are
(3.3%), the mean diameter of the coal particles can be estimated as generally less than 5%, showing good agreement between the simula-
61 μm, and the distribution of the particle diameter was assumed to tion and the on-site test results. The RE of the gas temperature is 6.8%,
follow the Rosin-Rammler ruler with a spread parameter of 1.15 [12]. which can be attributed to the poor temperature distribution at the exit
In the CFD modeling, the SIMPLE algorithm was applied to consider the of the final RH. There might be a discrepancy between the average of
coupling of the velocity and pressure fields. the two on-site testing points and the area-weighted average tempera-
ture of the simulation. Therefore, an RE within 7% is acceptable, and
4. Results and discussion the coupled model can be applied in the following analysis.

4.1. Grid independence test and validation

4.2. Effect of SOFA tilt angle on combustion products
To eliminate the numerical error and to improve the accuracy of the
results, a grid independence test was conducted based on Case 1. In the The upward tilting of the SOFA burners has a significant effect on
present work, three grid systems with 2.0 million, 2.8 million, and 4.2 the combustion products, especially CO, which, as an economic para-
million cells were established. In the two grid systems of 2.8 million and meter, represents the chemical incomplete combustion loss, and the
4.2 million cells, the meshes were refined in the combustion chamber, pollutant NO. Fig. 9 shows the distributions of the turbulence intensity,
the platen zone, and the horizontal crossover pass, due to the large gas temperature, and main species concentrations on the cross-section
temperature gradients in these regions. Fig. 8(a) shows the gas tem- z = 0 m. Fig. 9(a) and Fig. 9(f) display the contours of the CO molar
perature along the furnace height, and Fig. 8(b) shows the gas velocity fraction for Case 1 with a SOFA tilt angle of −12° and for Case 2 with a
component Vx along line L2 (X = 0 m, Z = −9.408 to 9.408 m) and Vz SOFA tilt angle of +30°. According to the studies in reference [40], the

Vx [m/s]
50 Furnace Nose -30 -20 -10 0 10 20 30
-10
-30
40 2.0 million
-20 -5
4.2 million
30 2.8 million -10
Vz [m/s]
Y [m]

4.2 million
Z [m]

0 L1 0
20 2.8 million
10
2.0 million
5
10 20
L2

30
0 10
1000 1100 1200 1300 1400 1500 1600 1700 -10 -5 0 5 10
Temperature [K] X [m]
(a) (b)
Fig. 8. Modelling results of different CFD grid systems, (a) gas temperature along the furnace height, (b) gas velocity component Vx along line L2 (X = 0 m,
Z = −9.408 to 9.408 m) and Vz along line L1 (Z = 0 m, X = −9.072 to 9.072 m) on the cross section that is 0.16 m above the B layer pulverized coal burner.

94
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

65
60
55
50
45
40
Y [m]

35
30
25
20
15
10
5
0
(a) Case 1, CO (b) Case 1, Turbul. (c) Case 1, O2 (d) Case 1, Temp. (e) Case 1, NO
65
60
55
50
45
40
Y [m]

35
30
25
20
15
10
5
0
(f) Case 2, CO (g) Case 2, Turbul. (h) Case 2, O2 (i) Case 2, Temp. (j) Case 2, NO
Fig. 9. Distributions of the CO molar fraction [–], turbulence intensity [%], O2 molar fraction [–], gas temperature [K], and NO molar fraction [–] on the cross-section
z = 0 m.

P7
Temp. [K] Temp. [K]
1450 1450
1200 1200
950 950
P6 700 700
P5
Vx [m/s] Vx [m/s]
P4
25 25
P3
15 15
5 5
P2
-5 -5
-15 -15
P1
-25 -25
(a) Case 1 (b) Case 2
Fig. 10. Distributions of gas velocity and temperature on different planes.

95
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

time scale of the homogeneous reactions for CO generation and oxi- distance below the SOFA burners. The swirling momentum intensity on
dation is approximately three orders of magnitude shorter than that of P2 can be enhanced by the jet flow from both the E layer PA nozzles and
heterogeneous char oxidation. Thus, the analysis of the CO formation the SOFA nozzles. Compared with that of Case 1, the SOFA tilt angle is
should be focused on the homogeneous combustion process, in which +30° in Case 2. The swirling momentum intensity on planes 1 to 3 is
the reaction rate is controlled by the diffusion kinetics and affected by obviously weakened. In addition, due to the SOFA being introduced at a
the O2 concentration and turbulent mixing. As shown in Fig. 9(b) and higher place, the swirling momentum intensity is strengthened on cross-
(g), when the SOFA tilt angle increases, the zone with a high turbulence sections P4 and P5, both of which are above the SOFA burners. This
intensity moves upward from the region in the vicinity of the burners to effect illustrates that there is a shorter length in the furnace to decrease
the region near the furnace arch. This change illustrates that a smaller the swirling flow generated by the SOFA. Thus, compared to Case 1, the
SOFA tilt angle is advantageous for O2 mixing in the neighborhood of swirling momentum intensity in front of the platen zone (see P6) in
the burners. Thus, the generated CO in the main combustion zone of Case 2 is increased by approximately 11%.
Case 1 is less than that of Case 2. However, Case 2 induces a high Cross-section P7 in Fig. 10 is the entrance profile of the final RH at a
turbulence intensity near the furnace arch, which would enhance the distance of 0.1 m. It can be seen that the gas temperature on the left side
mixing of SOFA, and the flue gas comes from the main combustion is higher than that on the right side and that the difference decreases as
zone. Thus, the generated CO of Case 2 is more likely to burn out. the height increases. This phenomenon, which concurs with several
Overall, the CO generated from coal combustion decreases from published studies [3,5], can be attributed to the higher velocity and net
593 ppm to 63 ppm as the SOFA tilt angle increases from −12° flow rate on the left part of the platen zone, which can reduce the gas
to + 30°. residence time so that less heat is released from the flue gas to the
Fig. 9(e) and (j) are the distributions of the NO molar fraction of division SH and platen SH in this region. To perform a further quanti-
Case 1 and Case 2, respectively. Several studies [8,20] have noted that tative analysis of the gas temperature deviations for Case 1 and Case 2,
the formation of thermal NO and fuel NO is directly related to the gas the deviation factor ET and the nonuniformity coefficient MT [1] are
temperature and O2 concentration. Fig. 9(d) and (i) show that the high- employed, as shown in Eqs. (18) and (19). The results show that the ET
temperature region of the burnout zone moves upward with the in- values in Case 1 and Case 2 are 1.09 and 1.12, respectively, and the MT
creasing SOFA tilt angle because part of the air is brought into the values calculated for Case 1 and Case 2 are 1.33 and 1.48, respectively.
furnace at a higher place. Compared with Case 1, Case 2 shows a re- This result indicates that the upward tilting of the SOFA nozzles can
latively lean-oxygen and low-temperature atmosphere in the main enhance the residual swirling at the furnace exit, which increases the
combustion zone, which can reduce NO formation. Although Case 2 gas temperature nonuniformity and the deviation at the entrance of the
presents better air mixing around the furnace nose, the downstream gas final RH.
enters the platen zone, and its temperature is so low that it cannot
Tave,left
significantly contribute to NO formation. Therefore, compared with ET =
Tave,right (18)
that of Case 1, the NO concentration of Case 2 decreases from 253 ppm
to 234 ppm. Overall, the upward tilting of the SOFA burners can reduce
1 N
NO emissions and chemical incomplete combustion losses. Tave + 3 N−1
∑i = 1 (Ti − Tave )2
MT =
Tave (19)
4.3. Effect of SOFA tilt angle on flue gas deviation
where Tave,left and Tave,right are the gas area-weighted average tem-
In addition to the combustion behavior, the SOFA tilt angle also perature in the left and right parts on the P7 plane, respectively; Tave is
affects the aerodynamic field and swirling flow in the furnace, which the gas area-weighted average temperature on the whole P7 plane; N is
are associated with the flue gas deviation in the horizontal pass. As the number of cells on the P7 plane; and Ti is the gas temperature in the
shown in Fig. 10, the horizontal planes P1 and P2 are separately located ith cell on the P7 plane.
at the middle of the A and E layer PA burners, both of which are in the
main combustion zone. P3 and P4 are separately the inlet and outlet 4.4. Effect of SOFA tilt angle on tube wall temperature
cross-sections of the SOFA burners, while P5 and P6 are in the front of
the furnace arch and the platen zone, respectively. It is clear that the air Due to the high-temperature and high-dust environment inside the
from the burners is injected into the furnace, generating a positive boiler, the thermocouples for monitoring the outer surface temperature
velocity component Vx at the right side and a negative component Vx at of the tubes are usually isolated from the flue gas. A limited number of
the left side on these horizontal planes, which results in a counter- measuring points are located at the outlets of the tubes. Thus, a coupled
clockwise fire ball formed in the furnace center. It can be seen that the
velocity deviation of the left and right sides varies significantly along
the furnace height and that there still exists a residual swirling flow at
the furnace exit (see P6). To quantitatively describe the strength of the
swirling flow within the furnace, the parameter of the swirling mo-
mentum intensity ψ is calculated for planes P1 to P6 using the following
mathematical correlation [3]:
R
ψ = 2π ∫0 ρg w 2r 2dr = ∬S ρg w2rdS (17)

where ρg is the flue gas density; w is the tangential velocity of the flue
gas; r is the rotating radius; R is the equivalent radius; and S is the area
of the unit cell.
Fig. 11 shows the swirling momentum intensity along the furnace
height. It can be observed from Case 1 that the swirling momentum
intensity gradually decays from the main combustion zone to the fur-
nace exit, due to the spontaneous recovery effect of the flow field. The
maximum swirling intensity appears on cross-section P2 because the
SOFA tilt angles are −12° in Case 1 and because the P2 plane is a short Fig. 11. Swirling momentum intensity along the furnace height.

96
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

Final SH Final SH

Heat flux [W/m 2] Temp. [K]

-10000 905
-70000 865
-130000 825
Final RH -190000 Final RH 785
Division SH Division SH
Platen SH -250000 Platen SH 745

(a) (b)
Fig. 12. Heat flux and wall temperature of the high-temperature heat exchangers from Case 1.

simulation is performed to predict the tube wall temperature inside the According to the boiler manufacturer, the final RH is assembled by
boiler and evaluate the positions and cases in which overheating is 33 serpentines, each of which is made of 20 U-tubes. By contrast, the
more likely to occur. final SH has 82 serpentines, and each serpentine consists of two U-tube
Fig. 12 shows the heat flux and tube outer surface temperature for panels equipped with 13 tubes. Fig. 13 displays the tube outer surface
the division SH, the platen SH, the final SH, and the final RH in Case 1. temperature of six representative serpentines, which are the 5th, 10th,
division SH and platen SH are typical semi-radiant heating surfaces 15th, 20th, 25th, and 30th serpentines of the final RH and the 4th, 18th,
located at the upper furnace. The around-gas temperature is relatively 32th, 46th, 60th, and 74th serpentines of the final SH from the right
high, and the heat exchangers can receive radiation from the furnace side to the left side. The final RH and the final SH are designed with
flame. The maximum heat flux and wall temperature in both heaters cross-parallel-flow and cross-counterflow configurations, respectively.
appear at the bottom of the serpentines. However, the final RH and final Unlike the semi-radiant heating surfaces, the high-temperature zones of
SH are convection heating surfaces. It is clear that the heat flux cal- the final RH and the final SH are mainly at the tubes in the outer coilers
culated by Eq. (15) is much less than that of the division SH and the and the outlet parts of the tubes. Additionally, the final RH is arranged
platen SH; however, a higher tube outer surface temperature can be directly above the furnace nose, and the bottom of its serpentines on the
found in the final SH and the final RH. This result illustrates that the furnace side also appears in a local high-temperature zone. This phe-
tube wall temperature is more dependent on the steam side parameters nomenon is similar to the temperature distribution in the division SH
than on the gas side parameters. In addition, the region of the final SH and the platen SH and can be attributed to the penetrating radiation of
and the final RH is subject to a stronger flue gas deviation, which leads the furnace flame.
to an intensive nonuniformity for the tube wall temperature. Therefore, Fig. 14 displays the tube outlet wall temperatures at the outermost
the following work analyzes these two heating surfaces. tube coilers, which are at the P1 and P2 points of the final RH and final

Left Left
#1 to #20
Right Right
Temp. [K] Temp. [K]
P1
905 905
865 865
825 825
R1 R3 785 785
745 745

R2
(a) Case 1, Final RH (b) Case 2, Final RH
#13 to #1
Temp. [K] Temp. [K]
P2 875 875
835 835
S4 855 855
795 795
S3 S1
815 815
775 775
S7 S5
S6 S2 (c) Case 1, Final SH (d) Case 2, Final SH

Fig. 13. Distributions of tube outer wall temperature for the final RH and the final SH.

97
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

915 Left Right 885 Left Right

905
Case 1, calc. 875 Case 1, calc.

Temperature at P2 [K]
Temperature at P1 [K]
895 Case 2, calc. Case 2, calc.
Case 1, test Case 1, test
885 865

875 855
865
845
855

845 835
-10 -5 0 5 10 -10 -5 0 5 10
Z [m] Z [m]
(a) Final RH (b) Final SH
Fig. 14. Tube outer wall temperature at P1 and P2 along the width direction.

SH, respectively, as shown in Fig. 13. The average relative errors (AREs) temperature of approximately 870 K.
of the simulation and measurement results are 0.72% for the final RH
and 0.42% for the final SH, respectively, and both results show that the 4.5. Effect of SOFA tilt angle on steam tube overheating
tube outer surface temperature is higher on the left side than on the
right side. The qualitative and quantitative results of the coupled si- The main steel materials of the final RH and final SH include
mulation show good agreement with the in situ measurements. Table 3 TP347H, T91, and T23. References [24,41] noted that the allowable
lists the simulated average, maximum, and minimum temperatures and metal temperatures of TP347H, T91, and T23 are 923 K, 866 K, and
their deviations at the P1 and P2 points. Compared with those in Case 1, 853 K, respectively. From a purely creep strength point of view,
the average tube wall temperatures in Case 2 have increased by 12 K TP347H and T91 steels could be applied up to approximately 948 K and
and 5 K for the final RH and the final SH, respectively. In addition, the 893 K, respectively. Although the tube wall temperature exceeding the
differences in the maximum tube wall temperature (i.e., 18 K for the allowable metal temperature cannot directly result in instantaneous
final RH and 14 K for the final SH) are much larger than those of the tube overheating failure, it would reduce the service life of the tubes.
minimum tube wall temperature (i.e., 3 K for the final RH and the final Thus, the allowable metal temperatures can be regarded as the
SH). These results indicate an increase of the tube wall temperature and thresholds. As shown in Table 4, when the SOFA angle tilts upward
its deviation between the maximum and minimum value in Case 2. from −12° to +30°, the percentages of tube outer surfaces in the final
Compared with the final RH, the downstream final SH shows a smaller RH exceeding the thresholds of 866 K and 853 K rise by 38.5% and
deviation of the tube wall temperature because the temperature dif- 53.2%, respectively, while the percentage of that exceeding the
ference in the adjacent serpentines can gradually decrease through the threshold of 923 K becomes 0.16%. Therefore, the tube overheating risk
heat transfer of the flue gas between the serpentines. increases. One reason is that the flame center of the burnout zone
As shown in Fig. 13, the maximum outer surface temperatures for moves upward with the increasing SOFA tilt angle, which increases the
the final RH and the final SH are not exactly at the tube outlet points, overall gas temperature in the horizontal pass. Fig. 16 shows the
where most thermocouples are located. Thus, predicting the position maximum outer wall temperatures of all tubes in the final RH and the
with the maximum wall temperature is necessary for the analysis of the final SH. It can be seen that even the maximum tube wall temperature
tube overheating. on the right side increases in Case 2. The other main cause is that a
To investigate the tube wall temperature along its path, the outer- greater SOFA tilt angle also increases the gas temperature deviation in
most tube coilers (i.e., #1 tube) on the 30th serpentine of the final RH the horizontal pass. It is clear in Fig. 16 that the discrepancy in the
and the 74th serpentine of the final SH are selected. As shown in maximum tube wall temperature between the left and right parts in-
Fig. 13, the tube coilers in the final RH and final SH are divided into 3 creases in Case 2 compared to Case 1, which further enhances the
parts (i.e., R1 to R3) and 7 parts (i.e., S1 to S7), respectively. Fig. 15(a) overheating risk for the tubes in the left part.
shows the results of Case 1. The equivalent heat flux increases along the Another conclusion that can be drawn from Table 4 is that the
length of R1 and reaches its maximum value at the bottom elbow to- overheating risk of the final SH is lower than that of the final RH. This
wards the furnace side. After the steam enters the R2 part, the difference is because the final SH is arranged downstream with a lower
equivalent heat flux decreases mainly because the flue gas around R2 is gas temperature and a smaller gas deviation.
in parallel flow over the tube, and so the heat flux in this part is smaller
than that in the tube parts with cross flow [24]. The R3 part is down-
stream of the flue pass with a relatively low heat flux. Although the 4.6. Combustion optimization under overheating constraint
steam temperature increases from R1 to R3 by accumulatively ab-
sorbing heat, the tube outer wall temperature is dependent on both the Fig. 17 shows the simulated and experimental results at off-design
heat flux and steam temperature. Overall, the maximum value in the
final RH appears at the end of the R1 part with a temperature of ap- Table 3
Simulated results at the P1 and the P2 points.
proximately 900 K.
As shown in Fig. 15(b), the S2, S4, and S6 parts of the final SH are in Item Tube wall temperature at the P1 Tube wall temperature at the P2
the local valley heat flux areas due to the effect of parallel flow in these point of the final RH (K) point of the final SH (K)

segments. In addition, because of the cross-counter configuration of the

Case 1 Case 2 Difference Case 1 Case 2 Difference
final SH, the heat flux in outlet part S7 is slightly higher than that in
inlet part S1. The steam temperature also shows an increasing trend Average 868 880 12 855 860 5
along the tube’s path. Therefore, the maximum outer wall temperature Maximum 889 907 18 865 879 14
Minimum 856 858 3 845 848 3
of the final SH is located in the region close to the tube outlet, with a

98
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

Temp. [K] Heat flux [W/m 2] Temp. [K] Heat flux [W/m 2]
950 120000 870 120000
qequiv,i qequiv,i
tout,i 100000 850 tout,i S2 S4 S6 100000
900 tb,i tb,i
80000 80000
850 830
60000 60000
800 810
40000 40000
750 20000 790 20000
R1 R2 R3 S1 S3 S5 S7
700 0 770 0
0 5 10 15 20 25 30 35 0 9 18 27 36 45
Length along the tube path [m] Length along the tube path [m]
(a) Final RH (b) Final SH
Fig. 15. Steam temperature tb,i, tube outer wall temperature tout,i, and equivalent heat flux qequiv,i along the tube’s path in Case 1.

Table 4
Percentages of tube outer surfaces exceeding different temperature thresholds.
Percentages (%) Exceeding 853 K Exceeding 866 K Exceeding 923 K

Final RH Final SH Final RH Final SH Final RH Final SH

Case 1 15.33 5.54 7.77 0.79 0 0

Case 2 24.93 6.94 16.59 2.2 0.16 0

Maximum temp. [K] Maxmium temp. [K]

960 960

920 920

880 880

840 840
33 33
25 25
17 17
9 9
1 1
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Tube Number Tube Number
(a) Case 1, Final RH (b) Case 2, Final RH
Maximum temp. [K] Maxmium temp. [K]
960 960

920 920

880 880

840 840
82 82
55 55

28 28
1 1
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Tube Number Tube Number
(c) Case 1, Final SH (d) Case 2, Final SH
Fig. 16. Maximum outer wall temperature for all tubes in the final RH and the final SH.

99
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

R923 NO [mg/Nm] CO [ppm] Unburned C [%] Boiler efficicency [%]

1.0 340 NO 750 14 93.4
Fly ash
CO Slag
12
0.8 RT 600 Efficiency
330 93.2
10
0.6 450 8
320 93.0
0.4 300 6

310 4 92.8
0.2 150
2

0.0 300 0 0 92.6

-15 -5 5 15 25 35 -15 -5 5 15 25 35
SOFA vertical tilt angle [°] SOFA vertical tilt angle [°]
(a) (b)
Fig. 17. Results at off-design conditions: (a) Calculated results of NO, CO, and tube overheating ratio R923; (b) Test results of fly ash unburned carbon, slag unburned
carbon and boiler efficiency.

conditions. As the SOFA vertical tilt angle increases from −12° allowable temperature of the material TP347H; concurrently, the
to + 30°, the generated NO and CO decrease by 25 ppm and 530 ppm, variations in the boiler efficiency and NOx decrease to a much lesser
respectively. That is, the pollutants generated from combustion and the degree. Considering the safe operation of the unit, the optimal
chemical incomplete combustion loss can be reduced. Additionally, SOFA tilt angle for the combustion optimization should be +20°.
according to the on-site experiments, despite an increase of 0.4% in the (4) Overall, the present model is a powerful tool for integrated boiler
unburned carbon in the fly ash, the unburned carbon content in the slag combustion optimization that considers the risk of steam tube
decreases by 3%, and the boiler thermal efficiency rises from 92.78% to overheating. In addition, the model can be used to evaluate the
93.21%. Therefore, provided that only a higher combustion efficiency positions at which overheating is more likely to occur; this cap-
and lower NO formation are the optimal boiler targets, the SOFA ver- ability is essential for boiler manufacturers to determine the mon-
tical tilt angle should be set at the maximum value +30°. However, itoring locations and select a proper steel alloy.
when the SOFA tilt angle increases from +20° to +30°, 0.16% of the
final re-heater would exceed the allowable metal temperature of Acknowledgements
TP347H (923 K); concurrently, the variations in the boiler efficiency
and NOx are 0.03% and 3 ppm, respectively, which are decreases of a This research on the coupled simulation of coal combustion and
much lesser degree. Thus, considering the long-term safe operation of steam tube heat transfer was funded by the Natural Science Foundation
the unit, the optimal SOFA tilt angle for the combustion optimization of China (Grant No. 51306032) and the Scientific Research Foundation
should be +20°. of the Graduate School of Southeast University (Grant No. YBPY1701).
Model validation and related experiments were supported by the Key
5. Conclusion Research Program of China Datang Corporation (Grant No. DTIG-KY-
2015-09).
A combined pulverized coal combustion and steam heating model
was established for a 660-MW tangentially fired boiler using CFD References
software and MATLAB. Detailed mapping methods were separately
provided for both structured and unstructured meshes in CFD with [1] D.F. Tian, L.J. Zhong, P. Tan, L. Ma, Q.Y. Fang, C. Zhang, D.P. Zhang, G. Chen,
discrete nodes in MATLAB. As a result, the tube outer surface tem- Influence of vertical burner tilt angle on the gas temperature deviation in a 700 MW
low NOx tangentially fired pulverized-coal boiler, Fuel Process. Technol. 138
perature can be calculated iteratively by coupling the heat transfers of (2015) 616–628.
the gas and steam sides. The REs of the calculated combustion products [2] H.Y. Park, S.H. Baek, Y.J. Kim, T.H. Kim, D.S. Kang, D.W. Kim, Numerical and
and the heat absorptions are less than 5%, and the REs of the gas and experimental investigations on the gas temperature deviation in a large scale, ad-
vanced low NOx, tangentially fired pulverized coal boiler, Fuel 104 (2013)
the tube outer wall temperature are controlled within 7% and 0.8%, 641–646.
respectively. The results of the coupled simulation thus have great [3] Y.C. Liu, W.D. Fan, M.Z. Wu, Experimental and numerical studies on the gas ve-
agreement with the on-site measurements. Under full load conditions, locity deviation in a 600 MWe tangentially fired boiler, Appl. Therm. Eng. 110
(2017) 553–563.
the main findings of this study can be concluded as follows: [4] G.W. Tang, B. Wu, K. Johnson, A. Kirk, D. Fu, C.Q. Zhou, Numerical study of a
tangentially fired boiler for reducing steam tube overheating, Appl. Therm. Eng.
(1) The upward tilting of the SOFA burners can reduce the formation of 102 (2016) 261–271.
[5] H.Y. Park, S.H. Baek, H.H. Kim, Y.J. Kim, T.H. Kim, H.S. Lim, D.S. Kang, Reduction
CO and NOx due to the reinforced turbulence intensity in the vici-
of main steam temperature deviation in a tangentially coal-fired, two pass boiler,
nity of the furnace arch and the increased air-staging condition, Fuel 166 (2016) 509–516.
which helps improve the boiler efficiency. From a purely combus- [6] H. Xu, B. Deng, D.F. Jiang, Y.Z. Ni, N.Q. Zhang, The finite volume method for
tion point of view, the optimal SOFA tilt angle is the maximum evaluating the wall temperature profiles of the superheater and reheater tubes in
power plant, Appl. Therm. Eng. 112 (2017) 362–370.
value of +30°. [7] Y.G. Zhou, M.C. Zhang, T.M. Xu, S.E. Hui, Effect of opposing tangential primary air
(2) A larger SOFA vertical angle leads to a higher gas temperature and a jets on the flue gas velocity deviation for large-scale tangentially fired boilers,
greater deviation in the horizontal pass. This effect can be attrib- Energy Fuels 23 (2009) 5375–5382.
[8] P. Tan, D.F. Tian, Q.Y. Fang, L. Ma, C. Zhang, G. Chen, L.J. Zhong, H.G. Zhang,
uted to the higher flame center in the reduction zone and the Effects of burner tilt angle on the combustion and NOx emission characteristics of a
smaller length in the furnace to eliminate the swirling flow caused 700 MWe deep-air-staged tangentially pulverized-coal-fired boiler, Fuel 196 (2017)
by the SOFA. 314–324.
[9] S.N. Chen, B.S. He, D. He, Y. Cao, G.C. Ding, X. Liu, Y. Duan, Z.P. Cao, X. Zhang,
(3) When the SOFA vertical tilt angle increases from +20° to +30°, J.G. Song, X.Z. Li, Numerical investigations on different tangential arrangements of
0.16% of the outer surfaces of the final re-heater would exceed the

100
C. Yu, et al. Applied Thermal Engineering 154 (2019) 87–101

burners for a 600 MW utility boiler, Energy 122 (2017) 287–300. 112 (2018) 929–938.
[10] S.D. Secco, O. Juan, M.L. Louisy, J.Y. Lucas, P. Pierre, L. Porcheron, Using a genetic [25] L. Sun, W. Yan, Prediction of wall temperature and oxide scale thickness of ferri-
algorithm and CFD to identify low NOx configurations in an industrial boiler, Fuel tic–martensitic steel superheater tubes, Appl. Therm. Eng. 134 (2018) 171–181.
158 (2015) 672–683. [26] J. Purbolaksono, A. Khinani, A.Z. Rashid, A.A. Ali, N.F. Nordin, Prediction of oxide
[11] X.H. Zhang, J. Zhou, S.Z. Sun, R. Sun, M. Qin, Numerical investigation of low NOx scale growth in superheater and reheater tubes, Corros. Sci. 51 (2009) 1022–1029.
combustion strategies in tangentially-fired coal boilers, Fuel 142 (2015) 215–221. [27] K. Waclawiak, S. Kalisz, A practical numerical approach for prediction of particu-
[12] L. Shi, Z.G. Fu, X.N. Duan, C.Y. Cheng, Y.Z. Shen, B.H. Liu, R.X. Wang, Influence of late fouling in PC boilers, Fuel 97 (2012) 38–48.
combustion system retrofit on NOx formation characteristics in a 300 MW tan- [28] M.B. Gandhi, R. Vuthaluru, H. Vuthaluru, D. French, K. Shah, CFD based prediction
gentially fired furnace, Appl. Therm. Eng. 98 (2016) 766–777. of erosion rate in large scale wall-fired boiler, Appl. Therm. Eng. 42 (2012) 90–100.
[13] P. Tan, J. Xia, C. Zhang, Q.Y. Fang, G. Chen, Modeling and reduction of NOx [29] Y. Yang, W.G. Bai, Y.M. Wang, Y.F. Zhang, H.Z. Li, M.Y. Yao, H.J. Wang, Coupled
emissions for a 700 MW coal-fired boiler with the advanced machine learning simulation of the combustion and fluid heating of a 300 MW supercritical CO2
method, Energy 94 (2016) 672–679. boiler, Appl. Therm. Eng. 113 (2017) 259–267.
[14] M. Costa, J.L.T. Azevedo, M.G. Carvalho, Combustion characteristics of a front-wall- [30] J.H. Yang, J.E.A. Kim, J. Hong, M. Kim, C. Ryu, Y.J. Kim, H.Y. Park, S.H. Baek,
fired pulverized-coal 300 MWe utility boiler, Combust. Sci. Technol. 129 (1997) Effects of detailed operating parameters on combustion in two 500-MWe coal-fired
277–293. boilers of an identical design, Fuel 144 (2015) 145–156.
[15] M. Costa, P. Sliva, J.L.T. Azevedo, Measurements of gas species, temperature, and [31] H.Y. Park, M. Faulkner, M.D. Turrell, P.J. Stopford, D.S. Kang, Coupled fluid dy-
char burnout in a low-no x pulverized-coal-fired utility boiler, Combust. Sci. namics and whole plant simulation of coal combustion in a tangentially-fired boiler,
Technol. 175 (2003) 271–289. 89 (2010) 2001–2010.
[16] M. Costa, J.L.T. Azevedo, Experimental characterization of an industrial pulverized [32] C. Schuhbauer, M. Angerer, H. Spliethoff, F. Kluger, H. Tschaffon, Coupled simu-
coal-fired furnace under deep staging conditions, Combust. Sci. Technol. 179 lation of a tangentially hard coal fired 700 °C boiler, Fuel 122 (2014) 149–163.
(2007) 1923–1935. [33] H. Kobayashi, J.B. Howward, A.E. Sarofilm, Coal devolatilization at high tem-
[17] S.C. Hill, L.D. Smoot, Modeling of nitrogen oxides formation and destruction in peratures, Symp. Int. Combust. 16 (1977) 411–425.
combustion systems, Prog. Energ. Combust. 37 (2010) 873–885. [34] J.M. Richardson, H.C. Howard, R.W. Smith, The relation between sampling-tube
[18] G.G. De Soete, Overall reaction rates of NO and N2 formation from fuel nitrogen, measurement and concentration fluctuation is a turbulent gas jet, Symp. Int.
Symp. Int. Combust. 15 (1975) 1093–1120. Combust. 4 (1953) 814–817.
[19] L.I. Díez, C. Cortés, J. Pallarés, Numerical investigation of NOx emissions from a [35] M. Field, Rate of combustion of size-graded fractions of char from low rank coal
tangentially-fired utility boiler under conventional and overfire air operation, Fuel between 1200 and 2000 K, Combust. Flame 13 (1969) 237–252.
87 (7) (2008) 1259–1269. [36] C.G. Yin, Refined weighted sum of grey gases model for air-fuel combustion and its
[20] C.R. Choi, C.N. Kim, Numerical investigation on the flow, combustion and NOx impact, Energy Fuels 27 (2013) 6287–6294.
emission characteristics in a 500 MWe tangentially fired pulverized-coal boiler, [37] W. Zima, M. Nowak-Ocłon, P. Ocłon, Simulation of fluid heating in combustion
Fuel 88 (2009) 1720–1731. chamber waterwalls of boilers for supercritical steam parameters, Energy 92 (2015)
[21] W.D. Fan, Y. Li, Q.H. Guo, G. Chen, Y. Wang, Coal-nitrogen release and NOx evo- 117–127.
lution in the oxidant-staged combustion of coal, Energy 125 (2017) 417–426. [38] F.W. Dittus, L.M.K. Boelter, Heat transfer in automobile radiators of the tubular
[22] Q.L. Zha, D.B. Li, C.A. Wang, D.F. Che, Numerical evaluation of heat transfer and type, Int. Commun. Heat Mass Transfer 12 (1985) 3–22.
NOx emissions under deep-air-staging conditions within a 600 MWe tangentially [39] Ansys Inc. Ansys fluent 14.5 User’s guide, Canonsburg: Ansys Inc: 2012.
fired pulverized-coal boiler, Appl. Therm. Eng. 116 (2017) 170–181. [40] Y.X. Wu, P.J. Smith, J.S. Zhang, J.N. Thornock, G.X. Yue, Effects of turbulent
[23] M. Chernetskiy, A. Dekterev, N. Chernetskaya, K. Hanjalić, Effects of reburning mixing and controlling mechanisms in an entrained flow coal gasifier, Energy Fuels
mechanically-activated micronized coal on reduction of NOx: Computational study 24 (2009) 1170–1175.
of a real-scale tangentially-fired boiler, Fuel 218 (2018) 215–229. [41] R. Viswanathan, Materials for ultra supercritical fossil power plants, EPRI Rport
[24] J. Qi, K.Y. Zhou, J.L. Huang, X.D. Si, Numerical simulation of the heat transfer of NO. TR-114750, 2000.
superheater tubes in power plants considering oxide scale, Int. J. Heat Mass Tran.

101