You are on page 1of 13

Applied Thermal Engineering 173 (2020) 115203

Contents lists available at ScienceDirect

Applied Thermal Engineering


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

A computational fluid dynamics (CFD) approach of thermoelectric generator T


(TEG) for power generation

Wei-Hsin Chena,b,c,d, , Yi-Xian Lina, Yi-Bin Chioua, Yu-Li Line, Xiao-Dong Wangf
a
Department of Aeronautics and Astronautics, National Cheng Kung University, Tainan 701, Taiwan
b
Department of Chemical and Materials Engineering, College of Engineering, Tunghai University, Taichung 407, Taiwan
c
Department of Mechanical Engineering, National Chin-Yi University of Technology, Taichung 411, Taiwan
d
Research Center for Energy Technology and Strategy, National Cheng Kung University, Tainan 701, Taiwan
e
Energy and Environmental Laboratories, Industrial Technology Research Institute, Hsinchu 310, Taiwan
f
State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China Electric Power University, Beijing 102206, China

H I GH L IG H T S G R A P H I C A L A B S T R A C T

• An advanced simulation technology


integrating CFD and TEM is devel-
oped.
• Performance of leading TEM in dual
TEM system is close to that of the
single TEM.
• Dual TEM can produce an additional
43% power when compared with the
single TEM.
• Modified channel with thinner height
improves output power and efficiency
significantly.
• This study provides the realistic ap-
proach of TEG harvesting waste heat
from flue gas.

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

Keywords: To harvest waste heat from flue gas in the industry, this study develops an advanced simulation technology by
Thermoelectric generator (TEG) integrating computational fluid dynamics (CFD) and a thermoelectric module (TEM) where the TEM is modeled
Output power and efficiency as a heat sink to absorb waste heat from flue gases. The influences of Reynolds number, convection heat transfer
Source term coefficient at the cold surface, flue gas inlet temperature, dual TEM, and channel geometry on the performance
TEG and CFD integration
of the TEM system are evaluated. The results clearly provide a measure in increasing the performance of TEM
Reynolds number
Convection heat transfer coefficient
with rising the Reynolds number, flue gas inlet temperature, and convection heat transfer coefficient at the cold
surface. In the dual TEM system, the performance of the leading TEM is very close to that of the single TEM, and
the dual TEM can produce an additional 43% power when compared with the single TEM. However, this also
implies that the output power of the trailing TEM drops 57% when compared to the leading one, stemming for its
impact upon the downstream TEM. When the channel geometry is modified to raise the flue gas velocity at
Re = 1,000, the output power and efficiency increase by 53.5% and 25.2%, respectively.


Corresponding author at: .
E-mail addresses: weihsinchen@gmail.com, chenwh@mail.ncku.edu.tw (W.-H. Chen).

https://doi.org/10.1016/j.applthermaleng.2020.115203
Received 26 November 2019; Received in revised form 10 February 2020; Accepted 13 March 2020
Available online 14 March 2020
1359-4311/ © 2020 Elsevier Ltd. All rights reserved.
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

Nomenclature V velocity (m·s−1)


∀ volume (m3)
A surface area (mm2)
a coefficient Greek letters
b contribution of the source term
C specific heat capacity (J) η efficiency (%)
D hydraulic diameter (m) ρ density (kg·m−3)
Ė energy per unit volume (W·m−3) ρe electrical resistivity (Ω·m)

E electric field intensity vector (V·m−1) μ dynamic viscosity (Pa·s)
ġ heat generation per unit volume (W·m−3) ∅ general variable
I electric current (A)
⇀ Subscripts
J electric current density vector (A·m−2)
k thermal conductivity (W·m−1·K−1)
c cold side of TE element
P pressure (atm)
h hot side of TE element
Q heat flow rate (W)
n n-type TE element
R universal gas constant (=8.314 m3·Pa K−1·mol−1)
p p-type TE element
RTE electric resistance (Ω)
st storage
Re external load resistance (Ω)
TE thermoelectric element
q̇ heat flux (W·m−2)
⇀ ch channel
q heat flux vector (W·m−2)
nb neighboring cells
S Seebeck coefficient (V·K−1)
P computational cells
T temperature (K)

1. Introduction Waste heat recovery and output power of TEGs under various
configurations [11] and operating conditions [12] can be analyzed
Nowadays, many industrial processes and power plants generate a through numerical simulation. TEGs are appropriate to be utilized in
huge amount of waste heat from burning fossil fuels or utilizing other heat exchangers, and the use of TEG simulation is also important in heat
fuels. In the waste heat, low-temperature waste heat accounts for the exchangers. Numerical simulation can not only reduce the experimental
largest proportion (> 60%) [1]. Although the low-temperature waste runs and time but also provide a powerful tool in prediction and design.
heat can be used for preheating and drying, most of the heat, in fact, is It is even possible to optimize system parameter settings to get the best
directly discharged into the environment. On account of a large amount performance of TEGs [13]. Many studies simulated the performance of
of the low-temperature waste heat in the industry, if the waste heat TEGs applied to various heat sources. Zhao et al. [14] studied the waste
sources in the industry can be recovered effectively, it will be an ef- heat recovery from the wet and dry flue gases using TEG, and their
fective countermeasure to abate industrial energy consumption if the results showed that the maximum output power generated from the wet
waste heat can be efficiently recovered, thereby mitigating the impact flue gas was 5.1 times greater than that generated from the dry flue gas
on global warming. The thermoelectric generator (TEG) is suitable to at 120 °C. Fernández-Yañez et al. [15] presented an approach to study
harvest the waste heat [2]; it is especially a potential device for re- the performances of TEG for light-duty diesel engines. Their results
covering waste heat from low-temperature areas [3]. Through the showed a good agreement with the experiment data and found that flow
Seebeck effect, TEG can convert thermal energy into electrical energy diffusers could enhance the heat transfer. Lan et al. [16] designed the
from the waste heat that is not used or recycle [4]. It does not only dynamic model of a TEG system for vehicle waste heat recovery by
dissipate heat but also supplies additional power to increase the overall combining counter-flow heat exchangers and the TEG, and the model
energy efficiency of a system. showed a result of producing average 170–224 W electric power. Bai
TEG has the advantages of no moving parts, low maintenance, et al. [17] numerically investigated automotive TEG with metal foam
quietness, small size, wide adaptability, high reliability, and so on attached to inner wall and the out surface of heat exchanger, and in-
[5,6]. Owing to these advantages, TEG has been widely used in various dicated that the TEG with metal foam had better sound reduction and
industries. The applications of TEG are mainly divided into five cate- could generate a higher power output of 323.42 W. Muralidhar et al.
gories: (1) electricity generation in extreme environments; (2) waste [18] evaluated the benefits of using TEG in hybrid electric vehicles, and
heat recovery; (3) decentralized domestic power and combined heat pointed out that fuel saving of skutterudite and SiGe-based systems
and power generation systems; (4) microelectronics; and (5) solar TEG were 7.2% and 6.5%, respectively. Zhao et al. [19] proposed an exhaust
[7]. Many studies have focused on the use of TEG for waste heat re- TEG system with fluid media. Through the simulation results, the
covery, and have confirmed that TEG has a good performance in re- generation density and generation efficiency of the media fluid TEG
covering heat. Fernández-Yáñez et al. [8] compared TEGs in a spark- system reached 1,254 W/m2 and 4.51%, respectively, and the genera-
ignition engine and in a compression-ignition engine, and showed that tion density and generation efficiency of the media fluid TEGs were
the spark-ignition engine had greater potential than the compression 3.39 times and 1.43 times higher than traditional TEGs. Hsu et al. [20]
ignition engine for energy recovery. Sornek et al. [9] examined three numerically constructed a system to recover waste heat from the ex-
types of TEGs; two TEGs were designed on a flat hot surface (cooling by haust pipe of an automobile composed of 24 TEGs and slopping block.
air and water) and one was designed on a flue channel (cooling by The results demonstrated that the design of the slopping block made
water). Their results suggested that the water cooling system could thermal distribution more uniform than that without the slopping
increase the output power of the TEG, and the air-cooled system could block.
increase the combustion efficiency if a hot air was used as the com- Hoang et al. [21,22] found that adding a fishbone-shaped fin on the
bustion inlet fluid. Aravind et al. [10] proposed a high-intensity dual hot side of a TEM can effectively improve the waste heat recovery ef-
microcombustor based TEGs, which could achieve a maximum output ficiency. They changed the mass flow rates of hot inlet gas and cold
power and efficiency of 4.52 W and 4.66% when using the maximum inlet water, and the results were compared with a TEM without the fin
power point tracking. at different endothermic mass flows. The maximum output power and

2
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

efficiency under different hot inlet gas mass flow rates were 17.54 W
and 3.68%, respectively. Rodrigues et al. [23] proposed a system to
recover the energy released from the composting process. This system
produced a maximum power density of 175 mW/m2 at a temperature
gradient of 20 °C. Houshfar et al. [24] studied a waste-to-energy (WTE)
plant, replacing a condenser in it with a TEG, and optimized it using a
multi-criteria optimization (MCO) method. The plant’s effective effi-
ciency was 17.22% and the total cost rate was 184.2 $ / h. Ji et al. [25]
developed a numerical model that was constructed using in-
compressible Navier-Stokes equations to simulate a latent heat thermal
storage (LHTS) system for recovering low-temperature waste heat.
Amaral et al. [26] increased the heat transfer near a thermoelectric
module by changing the geometry of the flow tube because the con-
version efficiency of the module was not high in waste heat recovery.
They found that enhanced thermoelectric power generation was
achieved by inserting alternating flow obstruction plates in a channel.
Despite a number of numerical studied implemented in the litera-
ture, in most of the studies mentioned above the entire flow field in the
system was first calculated based on the specified boundary conditions.
After obtaining the hot and cold surface temperatures of the TEG, the
formula of the TEG was then used to obtain the output power. The
literature review of the different calculations is tabulated in Table 1.
From these studies, it can be seen that there are different ways to cal-
culate the output power of TEG. However, to date, no research has been
carried out to integrate TEG with computational fluid dynamics (CFD).
As a result, these studies lack the actual correlation between heat Fig. 1. Schematics of (a) physical configuration system with a 4 cm × 4 cm
TEM and a flue gas channel and (b) grid system.
transferred from a hot fluid and the TEG performance, that is, the effect
of heat transfer from hot fluid on TEG is absent. For this reason, this
study aims to develop a numerical model by integrating CFD and TEG. accordingly, the hydraulic diameter of the channel is 0.04 m. For the
From this model, the heat transfer from the hot flue gas to a thermo- flus gas, its composition consists of nitrogen (73.28 vol%), oxygen
electric module (TEM) is calculated, and the output power of the TEM is (17.00 vol%), water (7.28 vol%), and carbon dioxide (2.44 vol%) [27].
predicted simultaneously. A channel system to recover waste heat from The properties of the flue gas such as viscosity, thermal conductivity,
flue gas for the TEM is designed. A TEM is placed on the wall of the density, and specific heat are given in Table 2. he geometry of the TEM
channel and is regarded as a heat sink to absorb heat. The detailed heat is also cuboid and its size is 40 mm × 40 mm × 3.7 mm which is a
transport phenomena, TEG performance, and their links will be ex- typical size of the commercial TEM [28]. The TEM has 127 pairs of p-
plored in detail. As a result, the present study is able to provide a more type and n-type semiconductors. The material of the TEM is Bi2Te3, and
practical approach to TEM predictions which is conducive to waste heat its properties are assumed to be constant and given in Table 2 [28]. The
recovery in the industry. hot surface of the TEM is connected with the channel so that heat is
transferred from the channel into the TEM. The TEM will absorb the
2. Methodology heat to generate electricity. The cold surface of the TEM is cooled by
various fluids, depending on the adopted convection heat transfer
2.1. Physical configuration coefficient.

For a TEM, its hot surface absorbs the heat, while its cold surface 2.2. Formulation
remains a lower temperature by cooling. The temperature difference
between the hot and cold surfaces triggers the Seebeck effect, which 2.2.1. Governing equations and boundary conditions in the channel
converts the heat into electricity directly. In the present study, the In this study, physical phenomena are assumed in the steady state.
physical configuration for computation is shown in Fig. 1 in which a hot In the channel, the Reynolds number is in the range of 1–1000.
flue gas flow channel and a TEM are included. The geometry of the Therefore, the flow field is assumed to be laminar and incompressible.
channel is cuboid and its size is 30 mm × 60 mm × 100 mm, and Meanwhile, heat radiation and body force are ignored, and the thermal

Table 1
Literature review of TEG studies using different approaches.
Material Size of TEM Method of calculating output power Ref

Bi-Te 55 mm × 55 mm × 5 mm The electrical current was predicted. If the new current value differed from the [14]
predicted value, the predicted value was corrected and recalculated until the two
were equal.
– 62 mm × 62 mm × 4 mm Calculate the output power using the formula after obtaining the temperature [16]
difference.
– 50 mm × 50 mm × 5 mm Calculate the output power using the formula after obtaining the temperature [17]
difference.
SiGe-based Skutterudite- 60 mm × 60 mm × 7.1 mm60 mm × 60 mm × 6.1 mm Calculate the output power using the formula after obtaining the temperature [18]
based difference.
The output power of the entire TEG was calculated as the [19]
n
followingP = ∑i = 1 (qhi − qli )
BiTe based 40 mm × 40 mm × 6.4 mm Maximum output power of TEG was fitted as a second-ordered function of ΔT [20]

3
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

Table 2 the output power per unit volume of the TEM. The output power of the
Properties of the adopted TEG material and flue gas. TEM can be derived as follows.
Parameter Unit Value With the assumption of the constant properties of the TEM material
and ignoring the Thomson effect, the current induced by the tempera-
Parameters of TEG ture difference across TEM can be expressed as [31]
Seebeck coefficient V·K−1 Sp = −Sn = 226.8 × 10−6
Thermal conductivity W·m−1·K−1 kp = kn = 1.52 S ΔT¯
Resistivity Ω·m ρp = ρn = 1.447 × 10−5
I=
RTE + R e (10)
Parameters of flue gas
Density kg·m3 0.4469
where S, RTE, and Re are the Seebeck coefficient (S = Sp − Sn ), the
Specific heat at constant pressure J·kg−1·K−1 1149.6273 electrical resistance of the TEM, and the external load resistance, re-
Thermal conductivity W·m−1·K−1 0.0548 spectively. The output power of the TEM is given by [32]
Viscosity kg·m−1·s−1 3.56 × 10−5
2
S ΔT¯ ⎞
I 2R e = ⎛
⎜ Re

⎝ RTE + Re ⎠ (11)
resistance of copper chip and contact thermal resistance are neglected
[29,30]. The steady governing equations include continuity, mo- For maximizing the output power of the TEM, the assumption of
mentum, and energy equations as follows. impedance matching [33] is adopted, that is
Continuity equation RTE = R e (12)

∇∙ (ρV ) = 0 (1) This implies that internal resistance and external resistance are
equivalent. Under this condition, the maximum output power can be
Momentum equation acquired from Eq. (11) and expressed as [34]
⇀ ⇀ ⇀ ⇀
ρV ∙∇V = −∇P + ∇∙ [μ (∇V + (∇V )T )] (2) S 2ΔT¯ 2
Maximumpower =
4R2 (13)
Energy equation
Accordingly, the sink term ġ is given by

ρCp V ∇T = ∇∙ (k∇T ) (3) S2ΔT¯ 2
4R2
For the boundary conditions, the velocity and temperature at the ġ = −
∀ (14)
inlet are given, whereas the mass and energy conservation are followed
at the outlet. The pressure of the outlet is fixed at 1 atm. The no-slip where ∀ designates the volume of TEM. The sink term is negative be-
boundary condition is adopted at the channel walls. With respect to the cause the TEM absorbs heat and converts it into electricity. Substituting
temperature, the channel walls are assumed to be adiabatic, except for Eq. (14) to Eq. (9) along with the constant material properties yields
the interface contacting the TEM where the heat transfer between the S2ΔT¯ 2
∂ 2T ∂ 2T ∂ 2T ⎡ 2 ⎤
wall and TEM is equal, that is, + + − ⎢ 4R ⎥=0
∂x 2 ∂y 2 ∂z 2 ⎢ ∀k ⎥
dTch dT ⎣ ⎦ (15)
− kch = −kTE TE
dx dx (4)
To make the physical problem more trackable, the mean tempera-
where k is the thermal conductivity. tures of the hot surface and cold surface of the TEM are employed for
predicting output power, while the lateral surfaces of the TEM are as-
2.2.2. Thermoelectric module (TEM) sumed to be adiabatic. Therefore, Eq. (15) is simplified to one dimen-
The TEM is considered as a heat sink to the flue gas in the channel sional (x-dimensional) heat equation as
and is treated as a control volume, as shown in Fig. 2. According to the
d 2T ġ
first law of thermodynamics, a heat equation is expressed can be elu- + =0
dx 2 k (16)
cidated as
Furthermore, the temperature distribution in the TEM can be de-
∑ Qiṅ − ∑ Wouṫ + Eġ = Esṫ (5) termined by integrating Eq. (16) twice to obtain
where Eġ and Esṫ are the energy production rate per unit volume and ġ 2
T (x ) = − x + c1 x + c2
energy storage rate, respectively, and they are given by 2k (17)

Eġ = gdxdydz
̇ (6)

∂T¯
Esṫ = ρC dxdydz
∂t (7)
where ρ and C are the material density and specific heat capacity, re-
spectively. In the control volume system, assuming a sink term is in-
volved in this system and the heat equation can be further expressed as
(qẋ − qx +̇ dx ) dydz + (qẏ − qy +̇ dy ) dxdz + (qż − qz +̇dz ) dxdy + Eġ = Esṫ
(8)
where q̇ is heat flux. Substituting Eqs. (6) and (7) and Fourier’s law
(= −k∇T ) into Eq. (8) and the whole equation is divide by dxdydz yields

∂ ⎛ ∂T ⎞ ∂ ⎛ ∂T ⎞ ∂ ⎛ ∂T ⎞
k + k ⎜+ k⎟ + ġ = 0
∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠ ∂z ⎝ ∂z ⎠ (9)
where k and ġ are the thermal conductivity and sink terms, respec-
tively. From the theory of self-consistency [12], the term ġ stands for Fig. 2. A shematic of the control volume system.

4
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

The values of c1 and c2 can be acquired from the thickness of TEM Re = 1, 10, 100, and 1000 are considered. The Reynolds number is
and the temperatures of hot and cold side surfaces. From Eq. (17), the defined as:
heat fluxes at the hot and cold surfaces are calculated by:
ρVD
Re =
dT μ (23)
qḢ = −kTE H
dx (18)
where ρ , V , D , and μ are the density and velocity of the flue gas, hy-
dT draulic diameter, and dynamic viscosity, respectively. The convection
qL̇ = −kTE L
dx (19) heat transfer coefficients under natural air convection, forced air con-
Then, the output power is (qḢ − qL̇ ) × ATEM , and the efficiency of the vection, oil forced convection, and water forced convection are in-
TEM, which is an energy conversion index from heat to electricity, cluded, and their values of approximated by 10, 100, 1000, and
pertains to energy conversion from heat to electricity, is written as 3000 W·m−2·K−1, respectively. As for the inlet temperature of the flue
gas, three values of 373.15 K, 573.15 K, and 773.15 K are considered. In
P q ̇ − qL̇ addition, a single 4 cm × 4 cm TEM and two TEM in tandem are tested,
η (%) = × 100 = H × 100
QH qḢ (20) and the effect of the size-reduced channel on the TEM is examined in
this study.
where P is output power, QH is the heat flow rate at the hot side surface,
and q̇ is the heat flux.
3. Results and discussion
2.3. Numerical method and validation
3.1. Influence of inlet Reynolds number
The integrated system was constructed using the commercial soft-
The isothermal contours at the channel surface contacting the
ware ANSYS 19.1. For the purpose of solving the aforementioned
thermoelectric module (TEM) under four different Reynolds numbers
governing equations, the SIMPLEC algorithm along with the second-
(i.e., Re = 1, 10, 100, and 10,00) are shown in Fig. 4a where the inlet
order upwind scheme was adopted. The residual sum of the computa-
flue gas temperature is 773.15 K and the convection heat transfer
tion is defined as:
coefficient at the cold surface is 3000 W·m−2·K−1. At the low Reynolds
Residualsum = ∑ |anb ∅nb + b − aP ∅P | number of 1 (Re = 1), the characteristic time of a flow is high and the
domain (21) role of heat sink played by the TEM [38] is significant. Therefore, the
where a , ∅, and b are coefficient, general variable, and the contribution entire temperature distribution is low, except for that close to the inlet.
of the constant part of the source term, respectively [35]. During the A higher Reynolds number brings more thermal energy into the
calculation, the numerical convergence was presumed when the re- channel. As a result, the high-temperature (red) zone extends with in-
sidual sums of continuity, momentum, and energy equations were all creasing the Reynolds number. In view of partial heat in the flue gas is
smaller than 10−6, and the iteration was terminated. absorbed in the TEM, the temperature at the contacting surface is
To ascertain the numerical system satisfying the grid independence lower.
requirement, four different mesh numbers of 688, 1292, 2676, and The profiles of the hot surface and cold surface mean temperatures
4206 were tested. The results of the test suggested that the perfor- and their difference (ΔT̄ ) across the (TEM) at the four Reynolds num-
mances of the TEM at the grid systems of 2676 and 4206 meshes were bers are shown in Fig. 4b. For Re = 1, the mean temperatures of the hot
almost equivalent, so the former was utilized for numerical simulations. side and cold side surfaces are close to each other, so the value of ΔT̄ is
Meanwhile, to provide rigorous numerical validation, a number of small (1.55 K). This clearly suggests that the waste heat transferred
comparisons were performed. The output power based on the devel- from the flue gas to the TEM is insufficient. ΔT̄ increases significantly
oped numerical method was compared with the result of the calculation when the Reynolds number rises. The value of ΔT̄ is raised to 22.20 K at
of a pair of TEG [31]. Specifically, for the flue gas temperature of Re = 1,000. This is ascribed to two factors: (1) more thermal energy is
773.15 K with Re = 1000 and the convection heat transfer coefficient transported from the flue gas and (2) a higher Reynolds number leads to
of 3000 W·m−2·K−1 at the cold surface, the attained mean temperatures a thinner thermal boundary layer [39], thereby intensifying the tem-
of the hot and cold surfaces were 332.01 K and 309.81 K, respectively. perature gradient and heat transfer rate. Nevertheless, it should be
The calculated output power of the TEM from the developed method in
this study was 0.13629 W (P1). Meanwhile, substituting the afore-
mentioned mean temperatures into Eq. (11) followed by the multi-
plication of 127 pairs of TEG, the calculated output power was
0.13589 W (P2). The relative error (= (P1 − P2)/ P1 × 100 ) was 0.29%.
Moreover, in the subsequent section (Section 3.1), it will be found that
the predicted result (0.13 W) at certain conditions is in line with that
(0.11 W) of past study [36]. Another numerical validation was also
carried out by comparing our prediction to the experimental results of
Chen et al. [37] in which the TEG’s cold side temperature was fixed at
303 K, and three hot side temperatures were selected. The comparison
shown in Fig. 3 suggests that the present predictions were in good
agreement with the experimental data [37], thereby validating the
developed numerical method.

2.4. Operating conditions in CFD

In this study, three parameters of the Reynolds number and tem-


perature of the flue gas and the convection heat transfer coefficient at
the cold surface on the performance of the TEM are taken into account.
For the Reynolds number, four different orders of magnitudes, namely, Fig. 3. Numerical validation between simulations and experimental data.

5
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

Fig. 4. (a) Isothermal contours at the channel wall contacting TEM and (b) distributions of temperatures of the hot and cold surfaces of TEM and their difference.

illustrated that the mean temperature at the hot surface is by far lower
than that of the flue gas at the inlet (773.15 K). This implies, in turn,
that the heat transfer between the flue gas and the TEM is not efficient.
The channel for the flue gas has a hydraulic diameter of 0.04 m. The
corresponding flow velocities at Re = 1, 10, 100, and 1000 are
0.001943, 0.01943, 0.1943, and 1.943 m/s, respectively. These low
velocities are responsible for the insufficient heat transfer, as observed.
The distributions of the heat transfer rate at the hot surface (Qh), as
well as the output power and efficiency of the TEM are displayed in
Fig. 5. Fourier’s law of heat conduction reveals that the mean tem-
perature difference (ΔT̄ ) across the TEM is the driving force of the heat
transfer. This is the reason why Qh increases considerably with in-
creasing the Reynolds number, having the same trend as ΔT̄ (Fig. 4b).
For waste heat harvest and energy consumption using a higher Rey-
nolds number, Borcuch et al. [40] studied the use of fins in thermo-
Fig. 5. Distributions of heat transfer rate (Qh) at the hot surface of TEM, output
electric systems to help recover waste heat, and analyzed thermo-
power and efficiency. electric output power and power requirements for the thermoelectric
systems. The energy required for the power supply of a thermoelectric
system increased when the mass flow rate increased, but this energy

6
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

Fig. 6. Distributions of (a) mean outlet temperature and (b) inlet and outlet
enthalpies and enthalpy absorption percentage.

Fig. 7. Distributions of (a) output power and (b) efficiency versus convection
demand merely accounted for 3–13% of the power generation,. This
heat transfer coefficients at the cold surface of TEM.
suggested that the effect of the pump energy on the overall energy in
the present study might be small and ignored. Basically, the profiles of
output power and efficiency also have the same trend as the mean percentage is 84.8% (Fig. 6b), which is responsible for the drastic drop
temperature difference. For the case of Re = 1,000, the mean tem- in the flue gas mean temperature from 773.15 K at the inlet to 367.22 K
perature difference is 22.1 K and the output power of 0.13 W. Though at the outlet (Fig. 6a). However, in view of less energy input by the flue
the conditions mentioned above are somewhat different from those of gas at Re = 1, the mean temperature difference (1.5 K) and the effi-
the past experimental result [36] where the output power was 0.11 W ciency (0.06%) are low. Despite pronounced increases in the efficiency
at the temperature difference of 26.5 K. This qualitative consistency of the TEM with increasing the Reynolds number, the enthalpy ab-
also validate the developed model. From a practical point of view, the sorption percentage drops rapidly from 84.8% at Re = 1 to 3.34% at
low-temperature difference and efficiency at Re = 1,000 can be im- Re = 1,000. This implies, in turn, that the utilization of energy in the
proved by installing fins in the channel at the hot surface of TEM [41] flue gas for thermoelectric generation declines in a significant way.
or installing cooling fins at the cold surface [35] enhance the heat
transfer. Alternatively, increasing flow velocity or adopting turbulent
3.2. Influence of convection heat transfer coefficient at the cold surface
flow can also intensify the convection heat transfer coefficient, thereby
enlarging the temperature difference [42,43] and output power [44].
The influence of the convection heat transfer coefficient at the cold
To proceed farther into the recognition of the heat transfer phe-
surface on the output power and efficiency of the TEM at the four
nomena, the profile of mean outlet temperature as well as the profiles of
Reynolds numbers are examined in Fig. 7. Natural and forced convec-
the inlet and outlet enthalpies and enthalpy absorption percentage are
tion of air as well as forced convection of oil and water at the cold
displayed in Fig. 6. A TEM will follow self-consistency [12], that is, the
surface are taken into account where their typical values are 10, 100,
difference in heat flow at the hot surface and the cold surface is
1,000, and 3,000 W·m−2·K−1, respectively [45], and thus adopted in
equivalent to the output power. For Re = 1, the enthalpy absorption
the present study. An increase in the convection heat transfer

7
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

Fig. 8. Three-dimensional isothermal contours at the channel walls with the convection heat transfer coefficients of (a) 10, (b) 100, (c) 1000, and (d)
3000 W·m−2·K−1 at Re = 1000.

0.86 W, respectively. Consequently, enhancing the convection heat


transfer coefficient of the cold surface is an effective way to promote the
performance of a TEM.
It is noted that the output power and efficiency are lifted obviously
when the convection heat transfer coefficient rises from 10 to
100 W·m−2·K−1. When examining the isothermal contours of the
channel surface contacting the TEM, Fig. 8a and 8b clearly depict that
increasing the coefficient from 10 to 100 W·m−2·K−1 significantly
lowers the hot surface temperature of the TEM, implying that more heat
is transferred from the flue gas through the TEM. This is the reason why
the output power and efficiency are improved to a great extent (Fig. 7).
Once the coefficient further goes up, its impact on the power and effi-
ciency withers because the two physical quantities merely increase a
bit. This implies, in turn, that the difference in output power and effi-
ciency using the forced convection of oil and the forced convection of
water is small. This arises from the fact that the forced convection of oil
enables the maintenance of the cold surface temperature at around
room temperature (300 K) which is close to the result of the forced
convection of water. The isothermal contours in Fig. 8c and 8d suggest
Fig. 9. Distribution of outlet temperature under different convection heat that the contours are very close to each other. This is also the reason
transfer coefficients at Re = 1000. responsible for the small variation in output power and efficiency
shown in Fig. 7. Accordingly, an increase in the convection heat
transfer coefficient within 100 W·m−2·K−1 is more efficient in enhan-
coefficient raises the output power and efficiency inasmuch as it en-
cing the TEM’s performance.
hances the cooling performance at the cold surface [46]. The forced
The profile of the mean temperature of the flue gas at the exit of the
convection of water has the best cooling performance. For the case of
channel with respect to the convection heat transfer coefficient at
Re = 1,000, the efficiency of the TEM with the forced convection of
Re = 1,000 is shown in Fig. 9. The mean temperature decreases with
water is 2.14 times higher than that with the natural convection of air.
increasing the coefficient, ascribed to more heat delivered at the cold
Gou et al. [47] also studied the effects of natural and forced convection
surface. However, on account of the low fluid velocity in the channel
on thermoelectric performance. They reported that the temperature
and thereby low heat transfer rate, the mean temperature is merely
gradients across a TEG under the natural and forced convection were
lowered from 766 K to 757 K when the convection heat transfer coef-
4 K and 15 K, respectively, while the output powers were 0.015 W and
ficient increases from 1 to 3,000 W·m−2K−1.

8
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

Fig. 10. Profiles of (a) output power and (b) efficiency of TEM at different
Fig. 12. Profiles of (a) output power and (b) efficiency of single TEM, in-
Reynolds numbers.
dividual TEMs and dual TEM.

3.3. Influence of inlet temperature

The influence of the inlet temperature of the flue gas on the output
power and efficiency of the TEM is evaluated in Fig. 10 where three
temperatures of 373.15, 573.15, and 773.15 K are considered, covering
a wider range of flue gas temperature for harvesting waste heat in the
industry. It can be seen that the output and efficiency of the TEM de-
pend strongly on the inlet temperature of the flue gas. The increasing
trend of the output power with the inlet temperature is not linear
(Fig. 10a). Specifically, the higher the inlet temperature, the more
pronounced the increase of the power. Yu et al. [48] simulated the inlet
temperature ranged from 100 °C to 200 °C in waste heat fluid for a
thermoelectric system, and found that the conversion efficiency in-
creased from 2.75% to 6.26%. Chen et al. [49] studied the effects of
thermal fluids at 110 to 150 °C for TEM power generation. Maximum
power changed from 0.74 to 1.75 W. This result is consistent with this
Fig. 11. Physical configuration and grid system consisting of two TEMs in studies. The behavior in Fig. 10a can also be explained by Eq. (4) which
tandem and a flue gas channel. reveals that the output power is linearly proportional to the tempera-
ture difference square. It follows that the inlet temperature has a pro-
nounced influence on the temperature difference. Alternatively, the

9
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

Fig. 13. Three-dimensional isothermal contours at the channel walls at Re = (a) 1, (b) 10, (c) 100, and (d) 1000.

individually considered in the two TEMs for the simulations. The output
power profiles of the two TEMs, their summation, and a single TEM at
the four Reynolds number are presented in Fig. 12a. For the dual TEM
system, the first (upstream) TEM possesses a higher power output than
the downstream one. This is attributed to the absorption of some heat in
the flue gas by the first TEM, giving rise to a lower temperature around
the second TEM, as shown in Fig. 13, whereby the second TEM has a
lower output power. It is noteworthy that the output power of the
leading TEM is close to that of the single TEM (Fig. 5). It follows that
the influence of the trailing TEM on the leading one is slight. The output
power values of the single and the dual TEMs at Re = 1,000 are 0.14 W
and 0.20 W, respectively, implying that around 43% of additional
power is produced if the second TEM is installed behind the first one.
However, it also reflects that the output power of the trailing TEM
drops 57% when compared to the leading one. This can also be ob-
Fig. 14. Physical configuration and grid system with modified flue gas channel. served in Fig. 13. Accordingly, the interaction of the two TEMs in
tandem can be figured out.
Similar to the output power, the efficiencies of the leading TEM and
efficiency of the TEM is linearly proportional to the inlet temperature in
the single TEM are close to each other, as shown in Fig. 12b. Seeing that
this study, as shown in Fig. 10b, reflecting that the increase in output
the efficiency is a ratio of output power to the heat transfer rate across
power and heat transfer rate at the hot surface (Qh) has the equivalent
the hot surface (Qh), it follows that the effect of the trailing TEM on the
rate. The results are also consistent with the study of Niu et al. [42]
heat absorption of the leading TEM is fairly slight. The mean efficiency
when they compared conversion efficiencies at hot fluid temperatures
of the dual TEM system is defined as
of 70 to 150 °C and found that the conversion efficiency was as the
linear function of the inlet temperature. P1 + P2
η¯ =
Qh1 + Qh2 (24)

3.4. Performance of two TEMs where the subscripts 1 and 2 designate the first and second TEMs, re-
spectively. The interaction between the two TEMs results in the mean
The attention of the study is shifted to the waste heat harvested by efficiency profile being between those of the first and the second TEMs.
two TEMs in tandem to figure out their performance. The investigated
configuration is displayed in Fig. 11 where two heat sinks are

10
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

Fig. 15. (a) Isothermal contours at the channel wall contacting TEM and (b) distributions of temperatures of the hot and cold surfaces of TEM and their difference.

3.5. Flue gas channel with a thinner height this reason, the higher hot surface means that the temperatures in
Fig. 15 are exhibited when compared to those in Fig. 4. Furthermore,
In view of the lower efficiency (Fig. 5) under the designed config- the modified system also has a higher mean temperature difference
uration in Fig. 1, a smaller channel with the half-height (1.5 cm) of the across the TEM. For example, the mean temperature differences at
original one (3.0 cm), as shown in Fig. 14, is adopted to examine the Re = 1,000 in Fig. 4b and 15a are 22.18 K and 27.51 K, respectively.
performance of the single TEM at the same Reynolds numbers. Based on The distributions of the heat transfer rate at the hot surface (Qh),
this modified channel, the hydraulic diameter of the channel is output power, and efficiency of the TEM are displayed in Fig. 16.
0.024 m, and the flue gas velocities corresponding to Re = 1, 10, 100, Fig. 16a shows that not only the output power but also the efficiency
and 1,000 are 0.003238, 0.03238, 0.3238, and 3.238 m·s−1, respec- are superior to the original system (Fig. 5). For instance, the output
tively, magnifying around 1.67-fold speed compared to the original power and efficiency of the TEM reach 0.21 W and 1.16% at
configuration. Compared to the results in Fig. 4a, as a whole, Fig. 15a Re = 1,000, accounting for 53.51% and 25.2% improvements, re-
gives higher isothermal contours. This can be explained by a higher gas spectively, when compared to the results in Fig. 5. It is known that the
velocity in this modified system, which attenuates the momentum and convection heat transfer coefficient increases with increasing the flow
thermal boundary layers, thereby intensifying the heat transfer rate. For velocity, that is, the capacity of heat transfer is increased. This is the

11
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

a factor of 1.43. The performance of the upstream TEM is close to that


of a single TEM. It is thus recognized that the interaction between the
two TEMs mainly imposes on the downstream TEM, yielding the pro-
nounced drop in the output power of the trailing TEM. To intensify the
performance of the TEM, the channel geometry is modified with a
thinner height, which yields a higher flue gas velocity. The resulting
output power and efficiency of the TEM are magnified by factors of 1.54
and 1.25, respectively. Overall, the developed method is capable of
providing a more realistic approach in predicting the performance of
TEM from harvesting waste heat from flue gas in the industry.

Declaration of Competing Interest

The authors declare that they have no conflicts of interest to this


work. We declare that we do not have any commercial or associative
interest that represents a conflict of interest in connection with the
work submitted.

Acknowledgments

The authors acknowledge financial support from the Bureau of


Energy, Ministry of Economic Affair, Taiwan, R.O.C., for this research.

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://


doi.org/10.1016/j.applthermaleng.2020.115203.

References

[1] C. Forman, I.K. Muritala, R. Pardemann, B. Meyer, Estimating the global waste heat
potential, Renew. Sustain. Energy Rev. 57 (2016) 1568–1579.
[2] W.-H. Chen, C.-C. Wang, C.-I. Hung, Geometric effect on cooling power and per-
formance of an integrated thermoelectric generation-cooling system, Energy
Fig. 16. Distributions of (a) heat transfer rate (Qh) at the hot surface of TEM, Convers. Manage. 87 (2014) 566–575.
[3] S.B. Riffat, X. Ma, Thermoelectrics: a review of present and potential applications,
output power and efficiency and (b) inlet and outlet enthalpies and enthalpy Appl. Therm. Eng. 23 (2003) 913–935.
absorption percentage. [4] J. He, T.M. Tritt, Advances in thermoelectric materials research: Looking back and
moving forward, Science 357 (2017) eaak9997.
[5] R. Shen, X. Gou, H. Xu, K. Qiu, Dynamic performance analysis of a cascaded
reason why the output power and efficiency of the TEM are intensified. thermoelectric generator, Appl. Energy 203 (2017) 808–815.
As for the inlet and outlet enthalpies and enthalpy absorption percen- [6] W.-H. Chen, C.-C. Wang, C.-I. Hung, C.-C. Yang, R.-C. Juang, Modeling and simu-
lation for the design of thermal-concentrated solar thermoelectric generator, Energy
tage presented in Fig. 16b, the entire profiles resemble those in Fig. 6.
64 (2014) 287–297.
The best enthalpy absorption percentage occurs at Re = 1 with a [7] D. Champier, Thermoelectric generators: a review of applications, Energy Convers.
percentage of 95.69%. A comparison with the result (84.79%) in Fig. 6 Manage. 140 (2017) 167–181.
at the same Reynolds number suggests that the enthalpy absorption [8] P. Fernández-Yáñez, O. Armas, R. Kiwan, A.G. Stefanopoulou, A.L. Boehman, A
thermoelectric generator in exhaust systems of spark-ignition and compression-ig-
percentage is improved by 12.85% (=(95.69 − 84.79) × 100/84.79 ). nition engines. A comparison with an electric turbo-generator, Appl. Energy 229
Accordingly, it is summarized that the newly designed channel can (2018) 80–87.
efficiently intensify the performance of the TEM. [9] K. Sornek, M. Filipowicz, M. Żołądek, R. Kot, M. Mikrut, Comparative analysis of
selected thermoelectric generators operating with wood-fired stove, Energy 166
(2019) 1303–1313.
[10] B. Aravind, B. Khandelwal, S. Kumar, Experimental investigations on a new high
4. Conclusions intensity dual microcombustor based thermoelectric micropower generator, Appl.
Energy 228 (2018) 1173–1181.
An advanced numerical method by integrating computational fluid [11] J. Yu, H. Zhao, K. Xie, Analysis of optimum configuration of two-stage thermo-
electric modules, Cryogenics 47 (2007) 89–93.
dynamics (CFD) and thermoelectric generation (TEG) in association [12] W.-H. Chen, Y.-X. Lin, X.-D. Wang, Y.-L. Lin, A comprehensive analysis of the
with a heat sink model has been successfully developed in this study. performance of thermoelectric generators with constant and variable properties,
Through this developed method, the fluid dynamics in a channel and Appl. Energy 241 (2019) 11–24.
[13] W.-H. Chen, P.-H. Wu, Y.-L. Lin, Performance optimization of thermoelectric gen-
the power generation of a thermoelectric module (TEM) can be com-
erators designed by multi-objective genetic algorithm, Appl. Energy 209 (2018)
bined together and simultaneously predicted. The predictions suggest 211–223.
that, for the Reynolds number covering four orders of magnitude of [14] Y. Zhao, S. Wang, M. Ge, Y. Li, Z. Liang, Y. Yang, Performance analysis of a ther-
moelectric generator applied to wet flue gas waste heat recovery, Appl. Energy 228
1–1,000, the flue gas velocity, mean temperature difference across the
(2018) 2080–2089.
TEM, output power, and efficiency are in the ranges of [15] P. Fernández-Yañez, O. Armas, A. Capetillo, S. Martínez-Martínez, Thermal analysis
0.001943–1.943 m·s−1, 1.55–22.18 K, 0.000664–0.135998 W, and of a thermoelectric generator for light-duty diesel engines, Appl. Energy 226 (2018)
0.0652–0.9373%, respectively. These results clearly reveal the influ- 690–702.
[16] S. Lan, Z. Yang, R. Chen, R. Stobart, A dynamic model for thermoelectric generator
ence of Reynolds number or flue gas velocity on the TEM’s performance applied to vehicle waste heat recovery, Appl. Energy 210 (2018) 327–338.
is significant. However, the performance of the TEM is relatively low, as [17] W. Bai, X. Yuan, X. Liu, Numerical investigation on the performances of automotive
a consequence of low mean temperature difference. When the two thermoelectric generator employing metal foam, Appl. Therm. Eng. 124 (2017)
178–184.
TEMs in tandem are examined, it is found that the output power of the [18] N. Muralidhar, M. Himabindu, R.V. Ravikrishna, Modeling of a hybrid electric
dual TEM system at Re = 1,000 is higher than that of the single TEM by

12
W.-H. Chen, et al. Applied Thermal Engineering 173 (2020) 115203

heavy duty vehicle to assess energy recovery using a thermoelectric generator, [34] Y. Wang, Y. Shi, D. Mei, Z. Chen, Wearable thermoelectric generator to harvest
Energy 148 (2018) 1046–1059. body heat for powering a miniaturized accelerometer, Appl. Energy 215 (2018)
[19] Y. Zhao, S. Wang, M. Ge, Z. Liang, Y. Liang, Y. Li, Performance analysis of auto- 690–698.
mobile exhaust thermoelectric generator system with media fluid, Energy Convers. [35] J.-Y. Jang, Y.-C. Tsai, C.-W. Wu, A study of 3-D numerical simulation and com-
Manage. 171 (2018) 427–437. parison with experimental results on turbulent flow of venting flue gas using
[20] C.-T. Hsu, G.-Y. Huang, H.-S. Chu, B. Yu, D.-J. Yao, Experiments and simulations on thermoelectric generator modules and plate fin heat sink, Energy 53 (2013)
low-temperature waste heat harvesting system by thermoelectric power generators, 270–281.
Appl. Energy 88 (2011) 1291–1297. [36] S. Wu, Y. Ding, C. Zhang, D. Xu, Improving the performance of a thermoelectric
[21] T.H. Hoang, A.T. Hoang, V.S. Vladimirovich, Power generation characteristics of a power system using a flat-plate heat pipe, Chin. J. Chem. Eng. 27 (2019) 44–53.
thermoelectric modules-based power generator assisted by fishbone-shaped fins: [37] M. Chen, L.A. Rosendahl, T. Condra, A three-dimensional numerical model of
Part I – effects of hot inlet gas parameters, Energy Sources Part A (2019) 1–12. thermoelectric generators in fluid power systems, Int. J. Heat Mass Transf. 54
[22] A.T. Hoang, X.P. Nguyen, A.T. Le, M.T. Pham, T.H. Hoang, A.R.M.S. Al-Tawaha, (2011) 345–355.
S. Yondri, Power generation characteristics of a thermoelectric modules-based [38] H. Su, F. Zhou, H. Qi, J. Li, Design for thermoelectric power generation using
power generator assisted by fishbone-shaped fins: Part II – Effects of cooling water subsurface coal fires, Energy 140 (2017) 929–940.
parameters, Energy Sour. Part A: Recovery Utilizat Environ. Effects (2019) 1–13. [39] D.P. Ghosh, R. Raj, D. Mohanty, S.K. Saha, 2 - Onset of Nucleate Boiling, Void
[23] C.R.S. Rodrigues, T. Machado, A.L. Pires, B. Chaves, F.S. Carpinteiro, A.M. Pereira, Fraction, and Liquid Film Thickness, in: S.K. Saha (Ed.), Microchannel Phase
Recovery of thermal energy released in the composting process and their conversion Change Transport Phenomena, Butterworth-Heinemann, 2016, pp. 5–90.
into electricity utilizing thermoelectric generators, Appl. Therm. Eng. 138 (2018) [40] M. Borcuch, M. Musiał, S. Gumuła, K. Sztekler, K. Wojciechowski, Analysis of the
319–324. fins geometry of a hot-side heat exchanger on the performance parameters of a
[24] E. Houshfar, Thermodynamic analysis and multi-criteria optimization of a waste-to- thermoelectric generation system, Appl. Therm. Eng. 127 (2017) 1355–1363.
energy plant integrated with thermoelectric generator, Energy Convers. Manage. [41] J.Y. Jang, C.Y. Tseng, Experimental and numerical analysis of built-in thermo-
205 (2020) 112207. electric generator modules with an elliptical pin-fin heat sink, Int. J. of Mech.
[25] C. Ji, Z. Qin, S. Dubey, F.H. Choo, F. Duan, Three-dimensional transient numerical Aerospace Industrial Mechatronic Manufact. Eng. 9 (2015) 702–708.
study on latent heat thermal storage for waste heat recovery from a low tempera- [42] X. Niu, J. Yu, S. Wang, Experimental study on low-temperature waste heat ther-
ture gas flow, Appl. Energy 205 (2017) 1–12. moelectric generator, J. Power Sour. 188 (2009) 621–626.
[26] C. Amaral, C. Brandão, É.V. Sempels, F.J. Lesage, Net thermoelectric generator [43] F.J. Lesage, N. Pagé-Potvin, Experimental analysis of peak power output of a
power output using inner channel geometries with alternating flow impeding pa- thermoelectric liquid-to-liquid generator under an increasing electrical load re-
nels, Appl. Therm. Eng. 65 (2014) 94–101. sistance, Energy Convers. Manage. 66 (2013) 98–105.
[27] M. Mofarahi, Y. Khojasteh, H. Khaledi, A. Farahnak, Design of CO2 absorption plant [44] Z. Niu, H. Diao, S. Yu, K. Jiao, Q. Du, G. Shu, Investigation and design optimization
for recovery of CO2 from flue gases of gas turbine, Energy 33 (2008) 1311–1319. of exhaust-based thermoelectric generator system for internal combustion engine,
[28] W.-H. Chen, S.-R. Huang, Y.-L. Lin, Performance analysis and optimum operation of Energy Convers. Manage. 85 (2014) 85–101.
a thermoelectric generator by Taguchi method, Appl. Energy 158 (2015) 44–54. [45] C. Lundgaard, O. Sigmund, Design of segmented thermoelectric Peltier coolers by
[29] Y. Wang, C. Dai, S. Wang, Theoretical analysis of a thermoelectric generator using topology optimization, Appl. Energy 239 (2019) 1003–1013.
exhaust gas of vehicles as heat source, Appl. Energy 112 (2013) 1171–1180. [46] S. Lv, W. He, Q. Jiang, Z. Hu, X. Liu, H. Chen, M. Liu, Study of different heat
[30] A. Eldesoukey, H. Hassan, 3D model of thermoelectric generator (TEG) case study: exchange technologies influence on the performance of thermoelectric generators,
Effect of flow regime on the TEG performance, Energy Convers. Manage. 180 Energy Convers. Manage. 156 (2018) 167–177.
(2019) 231–239. [47] X. Gou, H. Xiao, S. Yang, Modeling, experimental study and optimization on low-
[31] W.-H. Chen, S.-R. Huang, X.-D. Wang, P.-H. Wu, Y.-L. Lin, Performance of a ther- temperature waste heat thermoelectric generator system, Appl. Energy 87 (2010)
moelectric generator intensified by temperature oscillation, Energy 133 (2017) 3131–3136.
257–269. [48] J. Yu, H. Zhao, A numerical model for thermoelectric generator with the parallel-
[32] Y. Yan, J.A. Malen, Periodic heating amplifies the efficiency of thermoelectric en- plate heat exchanger, J. Power Sour. 172 (2007) 428–434.
ergy conversion, Energy Environ. Sci. 6 (2013) 1267–1273. [49] W.-H. Chen, C.-Y. Liao, C.-I. Hung, W.-L. Huang, Experimental study on thermo-
[33] W.-H. Chen, P.-H. Wu, X.-D. Wang, Y.-L. Lin, Power output and efficiency of a electric modules for power generation at various operating conditions, Energy 45
thermoelectric generator under temperature control, Energy Convers. Manage. 127 (2012) 874–881.
(2016) 404–415.

13

You might also like