Renewable Energy 154 (2020) 542e552

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Parametric study of a thermoelectric module used for both power

generation and cooling
Ding Luo a, b, Ruochen Wang a, *, Wei Yu b, Weiqi Zhou a
a
Automotive Engineering Research Institute, Jiangsu University, Zhenjiang, 212013, China
b
School of Automotive and Traffic Engineering, Jiangsu University, Zhenjiang, 212013, China

a r t i c l e i n f o a b s t r a c t

Article history: Thermoelectric module can be used as a generator to convert thermal energy into electricity, or as a
Received 14 January 2020 cooler to convert electricity into heat. In this work, a comprehensive model is proposed to predict the
Received in revised form performance of thermoelectric generator and cooler by setting different boundary conditions. Based on
4 March 2020
the proposed model, the influences of height, cross-sectional area, number of couples, ceramic plate, and
Accepted 9 March 2020
Available online 12 March 2020
heat loss on the generator and cooler are investigated. To balance the output performance and cooling
performance of thermoelectric modules simultaneously, a comprehensive study on the thermoelectric
module is conducted. The results indicate that (i) A relatively lower leg height enables the enhancement
Keywords:
Thermoelectric module
of output power, cooling power, and COP, despite a slight reduction in conversion efficiency; (ii) When
Thermoelectric generator increasing the total cross-sectional area of legs, thermoelectric generator should aim at adopting more
Thermoelectric cooler thermoelectric couples, whereas the cooler should apply a larger area for every single leg; (iii) For the
Numerical model optimization of ceramic plates, more attention should be paid in improving the thermal conductivity.
Parametric study Also, fillers are not recommended for thermoelectric module in general environment. The findings of this
work may guide the design and parametric optimization of thermoelectric module used for both power
generation and cooling.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction and high temperature materials, whose representative materials

are Bi2Te3, PbTe and SiGe, respectively, among which Bi2Te3 is the
The sandwich-shaped thermoelectric (TE) module is the most most popular one because its ZT value of around 1 can be reached at
commonly used thermoelectric device, which can be used as a TE room temperature. Bell [4] reported that the TE electronic com-
generator unit to convert thermal energy into electricity or as a TE ponents would become attractive if the ZT value can break through
cooler to turn electricity into heat. Also, the ring-shaped module [1] the threshold of 2. Disalvo [5] explored the current synthesis and
and the two-stage design [2,3] have been developed in some spe- characterization techniques for improving the conversion effi-
cific application fields, like engine exhaust waste heat recovery and ciency of TE materials and reported that an optimistic ZT > 4 would
refrigerators. Nevertheless, the low heat-to-electricity conversion be found in theory, which means TE devices have the potential to
efficiency and the high cost of TE materials have significantly become a promising alternative energy technology.
restricted their widespread commercialization. Generally, the TE Preliminary applications have been witnessed for power gen-
material performance is evaluated by the dimensionless constant of eration using TE modules in some specific fields, such as automo-
the figure of merit (ZT). Most of TE materials are still at a low- tive engine exhaust waste heat recovery [6,7], wearable TE devices
performance level at present, and their ZT values range from 1 to [8,9], and radioisotope TE generators in spacecraft [10,11]. Broader
1.6. According to the difference of optimal operating temperature, use of TE generators may be achieved in the future as long as the
TE materials can be classified into three main types: low, medium thermal-to-electricity energy conversion efficiency of TE materials
is close to that of other energy conversion technologies like
photovoltaic devices. Taking solar energy as heat sources, Kraemer
* Corresponding author. Jiangsu University, No.301, Xuefu Road, Jingkou District,
et al. [12] developed a kind of concentrating solar TE generator, and
Zhenjiang, 212013, Jiangsu, China. a peak efficiency of 7.4% was reached by their results. Photovoltaic
E-mail addresses: Ding_L@outlook.com (D. Luo), wrc@ujs.edu.cn (R. Wang), modules suffer from elevated temperature which results in the
downey_yu@hotmail.com (W. Yu), zwq@ujs.edu.cn (W. Zhou).

https://doi.org/10.1016/j.renene.2020.03.045
0960-1481/© 2020 Elsevier Ltd. All rights reserved.
 D. Luo et al. / Renewable Energy 154 (2020) 542e552 543

Nomenclature 4 electric potential, mV

h conversion efficiency

Symbols Subscripts
A area, mm2 a ambient
COP coefficient of performance c cooling or cold side
E electric field intensity vector, mV,mm
2 ce ceramic plates
h height (mm) or convective coefficient (W,m
2,K
1) co copper electrodes
I electric current, mA cond thermal conduction
J current density vector, mA,mm
2 f filler
N number of thermoelectric couples h hot side
P power, W in internal resistance or input
Q heat, W L load resistance
R resistance, K,W
1 or U leg thermoelectric legs
T temperature, K m material
U voltage, mV max maximum
n n-type thermoelectric legs
Greek symbols out output
l thermal conductivity, W,m
1,K
1 p p-type thermoelectric legs
a Seebeck coefficient, mV,K
1 solder solder material
s
1 electrical resistivity, 10
5U,m

decrement of efficiency and life. To overcome this deficiency, re- model and thermal resistance model are two commonly used ap-
searchers [13,14] proposed a photovoltaic-thermoelectric (PV-TE) proaches, and the numerical one is more popular because it may
hybrid module, and the results show that this approach cannot only predict more realistic results [22,24]. When a temperature differ-
utilize the waste heat effectively but also improve the conversion ence exists on both ends of the TE elements, it will force the elec-
efficiency. Apart from power generation, TE modules can also be trons of n-type elements and the holes of p-type elements moving
used as coolers to chill electronic components or control temper- from hot side to cold side, so that the Seebeck voltage and the
ature. Naphon and Wiriyasart [15] studied the liquid cooling system parasitic heat of Peltier heat, Thompson heat and Joule heat are
in a mini-rectangular fin heat sink with and without TE cooler for generated. For the TE module used for power generation, Ebling
CPU and found that TE cooler can effectively reduce CPU temper- et al. [25] proposed a thermal-electric multi-physical finite element
ature. Considering the advantages of no moving parts, no noise, and model to compute the output power of TE modules and the results
no emissions, TE coolers have also been used in car-seat system [16] present a good agreement with experimental data. In consideration
and small space air conditioning system in buildings [17]. of the fluid flow, Luo et al. [26] reported a novel TE module struc-
Besides the exploration of high-performance TE materials, the ture to overcome the deficiency that the overall output current of
structure-based optimization approaches can also apparently TE module is limited by the smallest current among TE elements,
enhance the performance of TE modules. Typically, a TE module and its performance evaluation and structural optimization are
consists of a great number of p- and n-type thermoelements con- performed via a fluid-thermal-electric multi-physical model. For
nected electrically in serial and thermally in parallel, copper elec- the TE module used for cooling, the basic physical conservation
trodes to connect thermoelements, and two ceramic plates to equations are the same as those used for power generation, but the
clamp thermoelements and electrodes. A single TE couple, as a boundary conditions are different. Chen et al. [23], Huang et al. [27]
basic unit of TE modules, is often chosen as the research object to and HoSung Lee [28] have studied the cooling power and the co-
analyze and optimize the performance of TE devices. When heat is efficient of performance (COP) of TE coolers via numerical models
transferred from the hot side to the cold side of TE couples, a large and their research results may guide the design and theoretical
temperature drop occurs along TE materials, causing the inefficient modeling of TE coolers.
utilization of thermal energy if only one material is adopted, and Table 1 summarizes some commercial TE modules from pub-
the segmented TE structure with two or more TE materials is the lished documents. It is clear that TE module parameters are
most effective approach to address this problem [18,19]. Moreover, different among different types, and there are no practical guide-
the height and cross-sectional area of thermoelements have an lines on how to guide the design of TE modules. As mentioned
impressive effect on the performance of TE devices [20,21]. Meng above, structural parameters have a significant effect on the
et al. [22] have studied the output power and conversion efficiency behavior of TE modules. However, some parameters may be posi-
of a single TE couple of power generation via a numerical model tive for the thermoelectric generator, but negative for the cooler. To
and found that the TE generator geometry significantly affects the balance the output performance and cooling performance of TE
performance. Chen et al. [23] compared three different modules of module simultaneously, a comprehensive study on the thermo-
TE cooler consisting of 8, 20 and 40 pairs of TE couples and electric module was conducted in this work. To achieve this goal, a
observed that the cooling power grows drastically when the numerical model was proposed to predict the performance of TE
number of pairs increased from 8 to 40. Nevertheless, there are no generator and cooler by setting different boundary conditions.
practical guidelines on how to design and optimize the perfor- Taking the height and cross-sectional area of TE elements, number
mance of TE modules, and the effect of structural parameters on the of TE couples, height and thermal conductivity of ceramic plates,
TE modules used for power generation and cooling may be and fillers as the research objectives, some suggestions were pre-
paradoxical. sented to ensure the output performance and cooling performance
To figure out the output performance of TE modules, numerical of TE module at the same time. With the consideration of both
544 D. Luo et al. / Renewable Energy 154 (2020) 542e552

Table 1
Some commercial thermoelectric modules from published documents.

Module type Company Module dimensions (mm3) Number of TE couples Height of TE legs (mm) Area of TE legs (mm2) Fillers Ref.

TMH400302055 Wise Life Tech. 60  60  3 199 0.64 22 no [29]

TEG1-287 NA 55  55  4 287 1.5 1.5  1.5 no [30]
TEC1-1270 Thermonamic Co. 40  40  4.2 127 NA NA no [31]
GM250-127-28-10 Eur. Thermodynamics Ltd. 62  62  4 127 1 2.8  5.6 no [32]
CP10-127-05 Laird Tech. 30  30  3.2 127 1.253 11 no [28]

power generation and cooling, this method can be used to thor- Eq. (6). Also, the current continuity can be expressed as Eq. (7).
oughly evaluate the performance of TE modules. Section 2 provides
the detailed principles and verifications of the model. Section 3 2.2. Boundary conditions of TE module used for power generation
presents the effect of different parameters on the TE generator and cooling
and cooler. The results and findings of this research are discussed in
Section 4. When TE module is used for power generation, its boundary
conditions include the temperature difference boundary and
2. Model voltage boundary. A high temperature (Th ¼ 450K) and a low
temperature (Tc ¼ 310K) are imposed on the hot and cold side of
2.1. Fundamental principles of the model the TE module respectively, so that a temperature difference is
generated. To form a complete circuit, the TE module is set to be
TE module follows the same physical laws whether it is used for connected with a load resistance, where one side of the terminals of
power generation or cooling, such as energy conservation, Seebeck TE module and load resistance is assumed to be ground (U ¼ 0V),
effect, Peltier effect, Joule effect, and Fourier heat conduction. To and another side is set to be connected electrically each other as a
figure out the complex thermoelectric multi-physical field coupling voltage coupling boundary condition.
effect of TE devices, numerical model has become a powerful tool In the case of cooling, the boundary conditions of TE module
for evaluating their performance, and COMSOL Multiphysics consist of the current boundary and temperature boundary.
[25,33,34] and ANSYS [26,35] are two frequently used commercial Compared to that of power generation, the difference is that elec-
finite element software to compute the numerical model. The tricity is not generated by a temperature difference but an inputted
fundamental principles of the model include the conservation current. Here, one terminal of the TE module is set as a constant
equations both thermal field and electric field, as follows: current input (Iin ) whereas another terminal is set to U ¼ 0V. The
cooling performance of TE module is investigated at DT ¼ 0K


V , lp ðTÞVTp ¼
s
1
p ðTÞJ þ Vap ðTÞJTp
2
(1) where both the hot and cold side temperatures are set to 310K.
In addition, heat loss occurs on the contact surfaces between TE
module and ambient air, which is defined as a natural convection
V , ðln ðTÞVTn Þ ¼
s
1
n ðTÞJ þ Van ðTÞJTn
2
(2) heat transfer boundary.

V , ðlco VTÞ ¼
s
1 2 vT

co J (3)
l ¼ ha ðT
Ta Þ (8)
vn
V , ðlce VTÞ ¼ 0 (4) where ha is the natural convection heat transfer coefficient and
Ta ¼ 300K is the ambient temperature.
E ¼
V4 þ aVT (5) After the above settings of boundary conditions, the numerical
model was computed using ANSYS 18.0/Thermal-Electric in this
J ¼ sE (6) work. Also, parameters of output voltage (UL ) and absorbed heat
(Qh ) for TE generator, and input voltage (Uin ) and cooling power
(Qc ) for TE cooler were obtained by a simple area integration of the
VJ ¼ 0 (7)
voltage density and heat flux over a selected cross section. Thus, the
where subscripts p, n, co, and ce represent p-type semiconductors, output power (Pout ) and conversion efficiency (h) of TE generator
n-type semiconductors, copper electrodes, and ceramic plates, can be estimated by Eqs. (9) and (10) respectively, and the input
respectively. The temperature dependence of thermoelectric ma- power (Pin ) and coefficient of performance (COP) of TE cooler can be
terials was taken into consideration in this work, including thermal expressed as Eqs. (11) and (12) respectively.
conductivity lp;n ðTÞ, electric resistivity s
1
p;n ðTÞ, and Seebeck coeffi-
UL2
cient ap;n ðTÞ. Eq. (1) and Eq. (2) denote the energy conservations of Pout ¼ (9)
p-type and n-type semiconductors respectively; In Eqs. (1) and (2), RL
the first term denotes the Fourier heat conduction, the second term
denotes Joule heat, and the last term represents Peltier heat or Pout
h¼ (10)
Thomson heat. Eq. (3) and Eq. (4) are the energy conservations of Qh
copper electrodes and ceramic plates respectively. Compared with
thermoelectric materials, the term related to Seebeck coefficient is Pin ¼ Iin Uin (11)
absent in copper materials, and only the Fourier heat conduction is
included in ceramic materials. The electric field density vector can Qc
be regarded as the sum of electric potential difference and Seebeck COP ¼ (12)
Pin
voltage, as shown in Eq. (5). The current density vector equals
electric conductivity times electric field density vector, as shown in Compared with thermal resistance network models, the
 D. Luo et al. / Renewable Energy 154 (2020) 542e552 545

proposed model can obtain the distribution characteristics of module to eliminate the air gap. Two K-type thermocouples
physical parameters, such as temperature distribution, voltage (WRNT, Huarun, China) were attached on the surfaces of the heater
distribution, and so on. In addition, this model can evaluate both and the heat sink to gauge the temperature of heat source (Th ) and
output performance and cooling performance of TE modules by cooling source (Tc ), respectively. An electronic load (IT8500þ,
setting different boundary conditions, which can be used to ITECH, China) was connected with the TE module to measure the
conduct a comprehensive analysis of the behavior of TE modules. output parameters of voltage and output power. Here, the heat
source temperature was set at 450K, and the corresponding cooling
2.3. Model validation source temperature was 320.2K.
Under the boundary conditions of Th ¼ 450K, Tc ¼ 320:2K, and
2.3.1. Grid independence verification ha ¼ 0W,m
2 ,K
1 , the comparison of output power between
The mesh sensitivity analysis is performed for the present nu- numerical results and experimental data was shown in Fig. 4. The
merical model to check its grid independence, taking output power maximum error is about 1.47%, which can be explained by the
as the examination object. Here, a TE module (TEG-127020, P&N measurement error of devices and the neglect of thermal grease in
technology, China) was chosen for the performance analysis. The numerical simulations. The small experimental error shows a great
actual TE module is shown in Fig. 1. There are 128 pairs of TE validity of this developed model, and the model can be used for the
couples in total. The size of hot side ceramic plate is theoretical analysis of TE modules.
40  44  0.8 mm3 whereas the cold side one is 40  40  0.8 mm3,
and both of them are made of alumina ceramics. The electrode 3. Results and discussion
material is copper, and TE material is bismuth telluride based.
Detailed geometric features and material properties of the TE 3.1. Numerical results
module are listed in Table 2.
In this section, four grid systems of grid (i), grid (ii), grid (iii), and The TE module mentioned above, as geometric features and
grid (iv) are selected for simulations with the mesh size of 0.2 mm, material properties are listed in Table 2, is regarded as the base
0.4 mm, 0.6 mm, and 0.8 mm respectively and the mesh number of structure. Fig. 5 shows the numerical results of TE module at
429328, 61717, 19520, and 8346 respectively. The calculation time RL ¼ 4U and ha ¼ 0W,m
2 ,K
1 , when it is used for power gen-
for grid (i) is about 3 h, rather it is only 20 min for grid (ii). Fig. 2 eration. The boundary conditions of TE generator are shown in
shows the output power as a function of load resistance for four Fig. 5(a), where the hot side temperature is fixed at 450K and that of
grid schemes at Th ¼ 450K, Tc ¼ 310K, and ha ¼ 0W, m
2 , K
1 . cold side is fixed at 310K in present work. From Fig. 5(b) to (d) are
The maximum error between grid (ii) and (iv) is about 3.9%, it the temperature, voltage, and current density distributions of TE
reduced to 0.6% for grid (ii) and (i). Thus, grid (ii) with a mesh size of generator, respectively. As can be seen in the pictures, the tem-
0.4 mm was chosen for simulations as a tradeoff between calcula- perature drop mainly occurs in the TE legs due to their lowest
tion time and model accuracy. thermal conductively compared to other components. The electric
potential increment is in line with the series of TE legs, and the
2.3.2. Experimental validation potential difference of TE module from cathode to anode is equal to
A TE module testing system, which equipped a sandwiched the voltage on both sides of load resistance. Through a simple
structure, was designed to measure the performance of TE mod- calculation, the output power is 3.13W with the output voltage of
ules, as shown in Fig. 3. A heater (X200, Xinhaomai, China) was 3.54V, when RL ¼ 4U. It seems that the output voltage of TE
used to supply steady heat energy with a maximum power of generator increases with the number of TE legs, resulting in higher
800W, and its temperature can be adjusted in the range of output power. Besides, the current is determined by an area inte-
0e450  C. An aluminum heat sink with pipelines inside is attached gration of current density over a selected cross section, which is
on the cold side of TE module to dissipate heat. The coolant (here is 0.88A in Fig. 5(d).
water) is powered by a liquid cooling system (CW-3000, Ruitong, In the case of cooling, the numerical results of TE module are
China). Thermal grease was adopted on the hot and cold side of TE quite different from that of power generation. Fig. 6(a) shows the

Fig. 1. The actual TE module.

546 D. Luo et al. / Renewable Energy 154 (2020) 542e552

Table 2
Detailed geometric features and material properties of the TEM.

p-type legs n-type legs Copper electrodes Ceramic plates

Seebeck coefficient (mV,K
1)
7 4

ap ðTÞ ¼
1:80268  10 T þ 3:23632  10 T
4 3 an ðTÞ ¼
ap ðTÞ e e

0:21537T 2 þ 62:97444T
6616:56781
Thermal conductivity (W,m
1,K
1) lp ðTÞ ¼
3:05948  10
9 T 4 þ 4:56781  10
6 T 3 ln ðTÞ ¼ lp ðTÞ 165.64 22

2:51621  10
3 T 2 þ 0:61074T
53:98632
Electrical resistivity (10
5oh,m) s
1
p ðTÞ ¼
3:08802  10

9 4
T þ 4:56531  10
6 T 3 s
1
n ðTÞ ¼ sp ðTÞ

1 1.75  10
3 e

3 2

2:58541  10 T þ 0:65579T
60:58804
Dimensions (L  W  H mm3) 1.4  1.4  1 1.4  1.4  1 3.8  1.4  0.3 40  44 (and 40)  0.8

Fig. 4. Comparison of output power between numerical results and experimental data.

Fig. 2. Mesh sensitivity analysis.

Fig. 6(a)) move from the bottom side to the top side, so that the heat
is absorbed on the bottom side of TE module whereas the heat is
released on the top side. Consequently, the cooling power (Qc ) of TE
boundary conditions and the grid system of TE cooler, where the
module can be predicted by an area integration of heat flux over the
temperature both hot and cold side of TE module is set to 310K.
bottom ceramic plate. Also, it can be expressed by:
Fig. 6(b) shows the temperature distributions of TE cooler at Iin ¼
4:5A and ha ¼ 0W,m
2 ,K
1 . As can be seen, the temperature of TE
module varies from 308.77K to 324.2K, which is caused by Peltier 12
heat, Thomson heat, and Joule heat. When the current flows Qc ¼ Qpeltier þ Qcond
Iin Rin (13)
2
through TE legs, both the holes of p-type legs (colored in red in
Fig. 6(a)) and the electrons of n-type legs (colored in blue in where Qpeltier and Qcond are the Peltier heat on the bottom side of TE

Fig. 3. TE module testing system.

 D. Luo et al. / Renewable Energy 154 (2020) 542e552 547

Fig. 5. Numerical results of TE generator at RL ¼ 4U and ha ¼ 0W,m
2 ,K
1 . (a) Schematic diagram of the boundary conditions and the grid system of TE generator. (b) Temperature
distributions. (c) Voltage distributions. (d) Current density distributions.

Fig. 6. Numerical results of TE cooler at Iin ¼ 4:5A and ha ¼ 0W,m
2 ,K
1 . (a) Schematic diagram of the boundary conditions and the grid system of TE cooler. (b) Temperature
distributions. (c) Voltage distributions. (d) Current density distributions.

legs and the heat conduction of TE module respectively. cooling. Based on this, the parametric study and optimization of TE
The voltage distributions and current density distributions of TE module are conducted in the following sections.
cooler are shown in Fig. 6(c) and (d), respectively. The input voltage
is about equal to Iin Rin . The current density of copper electrodes is 3.2. Effect of leg height
the highest by reason of their significantly higher electrical con-
ductivity. According to the numerical results, the performance of TE In the sections of 3.2e3.5, when studying the effect of one
module can be evaluated accurately both for power generation and variable on the performance of TE module, other parameters are
548 D. Luo et al. / Renewable Energy 154 (2020) 542e552

consistent with the base structure. Besides, all simulations of TE release. Under the same current input, the lower hleg is, the greater
generator are performed at Th ¼ 450K and Tc ¼ 310K, and that of Qc would become. Further, the effective current range of
TE cooler is performed at DT ¼ 0K. The heat losses between TE hleg ¼ 0:6mm is apparently larger than that of other three situa-
module and ambient air are omitted in this paper, that is ha ¼ 0W, tions. When hleg decreases, the internal resistance (Rin ) will
m
2 ,K
1 , except the section of 3.5. Fig. 7 shows the effect of leg decrease at the same time, causing the drop of Joule heat, so that
height (hleg ) on the performance of TE module. Obviously, output both Qc and current working range are increased. As shown in
power (Pout ) is increased along with the decrease of hleg , and the Fig. 7(d), the COP is constantly decreased with the growth of cur-
lower the height, the sharper the curves, as shown in Fig. 7(a). The rent, especially for a small input current. When the current is
reason for this is that the internal resistance of TE module is further increased, TE cooler will lose the function of cooling and
significantly decreased with the drop of hleg , so that a larger Pout is work at the invalid current operating ranges, causing the ineffective
generated. Also, it seems that the decrement of Pout becomes elevation of COP. The reason for this can be explained by Fig. 7(c).
modest if hleg reaches a relatively high value. Fig. 7(b) shows the Moreover, the COP of TE cooler is inversely proportional to hleg ,
conversion efficiency (h) of TE generator versus load resistance thus, the leg height of TE module should be as small as possible to
under four leg heights. Unlike that of output power, the maximum h achieve a better cooling performance when it is used as a cooler. It
remains a considerably low change for different leg heights, and it may be concluded that TE module with a relatively low leg height
is slightly increased with the augment of hleg because the more can supply a higher output power, even if a tiny reduction of con-
absorbed heat is transferred into electricity. version efficiency for TE generator, and higher cooling power and
For TE cooler, its cooling power (Qc ) and coefficient of perfor- COP for TE cooler.
mance (COP) versus current are shown in Fig. 7(c) and (d),
respectively. It is clear that Qc is parabolic with the current, which
3.3. Effect of leg cross-sectional area and number of TE couples
means the input current should be at the optimal one to reach the
maximum Qc . The dotted lines in Fig. 7(c) indicate the invalid
In addition to the height of TE leg, its cross section also has a
operating ranges of TE cooler. According to Eq. (13), if the current is
significant impact on the performance of TE module. There are two
too large, Qc would become negative, and the working state of the
approaches to increase the total cross-sectional area of TE legs
bottom side of TE cooler would change from heat absorption to heat
(Atotal ), one is to increase the area of every single TE leg (Aleg ), and

Fig. 7. The effect of leg height on the performance of TE module. (a) and (b) Performance of TE generator. (a) Output power versus load resistance under four leg heights. (b)
Conversion efficiency versus load resistance under four leg heights. (c) and (d) Performance of TE cooler. (c) Cooling power versus current under four leg heights. (d) Coefficient of
performance (COP) versus current under four leg heights.
 D. Luo et al. / Renewable Energy 154 (2020) 542e552 549

another one is to increase the number of TE couples (N). Here, three increased, which can be explained by Fig. 7(c). Under the same
cases are designated as base, case1, and case2, respectively. Both Atotal , the structure with greater Aleg has a much wider current
case1 and case2 include two structures where structure (i) is that operating range than the structure with more TE couples, due to the
the N is consistent with that of base structure, whereas structure effect of internal resistance, despite the almost same maximum Qc
(ii) is that the Aleg , that is 1.4  1.4 mm2, is consistent with that of for these two structures. However, the COP is mainly affected by the
base structure. The Atotal for base, case1, and case2 are 501.76 mm2, area of every single TE leg (Aleg ), and the bigger Aleg is, the greater
282.24 mm2, and 784 mm2, respectively. Fig. 8(a) shows the output COP becomes, as shown in Fig. 8(d).
power (Pout ) versus load resistance under three cases. As can be In case 2 with the total cross-sectional area of 784 mm2, the
seen, Pout is apparently increased with the growth of Atotal . And maximum output power and maximum efficiency of TE module
under the same Atotal , the structure with more TE couples cannot using more TE couples are increased by 1% and 1.2% respectively,
only reach a marginally higher Pout, but also a larger high Pout and it has a larger high-performance resistance range. For TE cooler,
working range, compared with that of a higher Aleg. The reason for despite the almost same maximum cooling power, the COP of TE
this is that Pout is proportional with Atotal , and the maximum Pout of module using a greater cross-sectional area of legs is increased by
the structure with more TE couples enables a larger load resistance 118.7% when I ¼ 4.5A, and it enables a larger current operating
due to its greater internal resistance. Fig. 8(b) shows the conversion range. Combining the results of this section, it may be drawn that a
efficiency (h) versus load resistance under three cases. There is a bigger Atotal of TE module can enhance the performance of both TE
small decrease of maximum h along with the augment of Aleg under generator and TE cooler. However, under the same Atotal , the TE
the same amount of TE couples. However, it seems that the Atotal module should be designed with more TE couples to achieve a
has very little effect on the maximum conversion efficiency. greater output performance for TE generator, by contrast, it should
The effect of leg cross-sectional area and number of TE couples be designed with the larger area of every single TE leg to reach a
on the performance of TE cooler is shown in Fig. 8(c) and (d), where better cooling performance for TE cooler.
Fig. 8(c) illustrates the variation of cooling power (Qc ) with input
current, and Fig. 8(d) is the variation of coefficient of performance 3.4. Effect of ceramic plates
(COP) with input current, respectively. It can be seen that Qc is
proportional with Atotal because the more electricity is converted On the other hand, the ceramic plate has an effect on the per-
into heat. The TE cooler will release heat if the current is further formance of TE module as well, including its height (hce ) and

Fig. 8. The effect of leg cross-sectional area and number of TE couples on the performance of TE module. (a) and (b) Performance of TE generator. (a) Output power versus load
resistance under three cases. (b) Conversion efficiency versus load resistance under three cases. (c) and (d) Performance of TE cooler. (c) Cooling power versus current under three
cases. (d) Coefficient of performance (COP) versus current under three cases.
550 D. Luo et al. / Renewable Energy 154 (2020) 542e552

thermal conductivity (lce ). Here, the hce is set as 0.6 mm, 0.8 mm, and the heat is exchanged in the form of convective heat transfer. To
and 1 mm respectively when lce is fixed at 10 W, m
1 , K
1 , and lce overcome this deficiency, researchers [36] applied isolating mate-
is set as 10 W,m
1 ,K
1 , 20 W,m
1 ,K
1 , and 30 W, m
1 , K
1 rials with electrical insulation and low thermal conductivity into
respectively when hce is fixed at 0.6 mm. The simulations for these the air gaps to eliminate the convective heat transfer, like glass fi-
five situations are performed and the results are shown in Fig. 9. As bers. However, the TE module may work at the different ambient
we all know, the thermal resistance of ceramic plates is propor- convective coefficients (ha ), and the thermal conductivity (lf ) of
tional to hce but inversely proportional to lce , and the temperature different fillers is various. In this study, the convective coefficients
difference on both sides of TE legs is increased with the reduction of of 10 W,m
2 ,K
1 , 50 W,m
2 ,K
1 , and 100 W,m
2 ,K
1 are
thermal resistance of ceramic plates. Consequently, both the output selected for simulations when TE module is unfilled, and fillers with
power and conversion efficiency are increased with the decrease of lf ¼ 0.1 and 0.5 W,m
1 ,K
1 are chosen for simulations when TE
hce and the elevation of lce . In practical application, an appropriate module is filled. Fig. 10 shows the output performance of TE
height of ceramic plates is indispensable to withstand a sufficiently generator with RL ¼ 4U and the cooling performance of TE cooler
large mechanical stress, and thus, more attention should be paid in with Iin ¼ 4:5A. As can be seen, for TE generator, both the output
improving the thermal conductivity of ceramic plates. power and conversion efficiency are decreased with the increase of
Fig. 9(c) and (d) show the variation of cooling power (Qc ) and ha , and the thermal conductivity of fillers should be as low as
coefficient of performance (COP) with current for TE cooler, possible, at least less than 0.1 W,m
1 ,K
1, to achieve a better
respectively. To achieve higher Qc and COP, a relatively low hce performance than that of unfilled TE module. For TE cooler, the
should be designed and the lce should be as high as possible, which effect of ha on the cooling performance is opposite to that of ha on
presents the same trend with TE generator. However, it seems that the output performance of TE generator, and the higher the ha , the
the effect of hce and lce on the COP is obviously lower than that of greater the Qc and COP. Therefore, the fillers are not recommended
hce and lce on the Qc . The reason for this may be the COP remains a when the TE module is used for cooling.
relatively low value and its increase cannot be recognized clearly in
Fig. 9(d), especially for the high input current range.
4. Conclusions

3.5. Effect of heat losses In order to balance the output performance and cooling per-
formance of TE module simultaneously, a comprehensive study on
Generally, the heat loss occurs between TE legs and ambient air, the thermoelectric module is conducted in this work. Besides, a

Fig. 9. The effect of ceramic plates on the performance of TE module. (a) The variation of output power with load resistance for TE generator. (b) The variation of conversion
efficiency with load resistance for TE generator. (c) The variation of cooling power with current for TE cooler. (d) The variation of COP with current for TE cooler.
 D. Luo et al. / Renewable Energy 154 (2020) 542e552 551

conductivity of fillers is at an extremely low level, at least less

than 0.1 W,m
1 ,K
1.
(5) Furthermore, the numerical results present a great consis-
tency with experimental data, which means this numerical
model can be used to conduct the theoretical analysis of TE
modules with reasonable accuracy.
(6) In future works, the fabrication of TE modules with a greater
performance will be conducted according to the findings of
this research. The experimental test rig will be improved to
measure the cooling performance of TE modules, and then,
the further comparison of TE modules with different pa-
rameters will be conducted at the experimental level.

Declaration of competing interest

We declare that we do not have any commercial or associative

interest that represents a conflict of interest in connection with the
work submitted.

Fig. 10. Radar map about the performance of TE module with or without fillers.
CRediT authorship contribution statement

numerical model was developed to work out the output power and Ding Luo: Conceptualization, Methodology, Writing - original
conversion efficiency of the TE generator, and the cooling power draft. Ruochen Wang: Supervision, Writing - review & editing. Wei
and COP of the TE cooler. Considering some parameters may be Yu: Data curation, Visualization. Weiqi Zhou: Investigation, Fund-
positive for the thermoelectric generator, but negative for the ing acquisition.
cooler, the parametric study of TE module were performed via
numerical simulations, taking the height and cross-sectional area of Acknowledgment
TE legs, number of TE couples, height and thermal conductivity of
ceramic plates, and fillers as the research objectives. Besides, the The authors are grateful for the financial support from the Na-
experimental validation of the numerical model was accomplished. tional Natural Science Foundation of China (51977100).
Based on the results above, the following conclusions were
reached: References

(1) Despite a slight reduction in conversion efficiency, the output [1] D.M. Rowe, Thermoelectrics Handbook: Macro to Nano, CRC press, 2005.
[2] L. Chen, J. Li, F. Sun, C. Wu, Performance optimization of a two-stage semi-
power of TE generator is apparently increased with the conductor thermoelectric-generator, Appl. Energy 82 (2005) 300e312.
decrease of TE leg height. For TE cooler, both the cooling [3] L. Hongkai, P. Yuzhuo, C. Jincan, Optimum design on the performance pa-
power and COP are inversely proportional to leg height. In rameters of a two-stage combined semiconductor thermoelectric heat pump,
Semicond. Sci. Technol. 19 (2004) 17.
practical application, a high enough output power is needed [4] L.E. Bell, Cooling, heating, generating power, and recovering waste heat with
to power the electrocircuit of the recovery system, thus, a thermoelectric systems, Science 321 (2008) 1457.
relatively low leg height should be designed for both TE [5] F.J. DiSalvo, Thermoelectric cooling and power generation, Science 285 (1999)
703.
generator and cooler.
[6] D. Luo, R. Wang, W. Yu, Z. Sun, X. Meng, Modelling and simulation study of a
(2) Both the output performance and cooling performance of TE converging thermoelectric generator for engine waste heat recovery, Appl.
module are enhanced with the increase of the total cross- Therm. Eng. 153 (2019) 837e847.
[7] Y. Zhao, S. Wang, M. Ge, Z. Liang, Y. Liang, Y. Li, Performance investigation of
sectional area of TE legs. However, under the same total
an intermediate fluid thermoelectric generator for automobile exhaust waste
cross-sectional area, the TE module should be designed with heat recovery, Appl. Energy 239 (2019) 425e433.
more TE couples to achieve a greater output performance for [8] Z. Wang, V. Leonov, P. Fiorini, C. Van Hoof, Realization of a wearable minia-
TE generator, by contrast, it should be designed with the turized thermoelectric generator for human body applications, Sens. Actua-
tors, A 156 (2009) 95e102.
larger area of every single TE leg to reach a better cooling [9] S.J. Kim, J.H. We, B.J. Cho, A wearable thermoelectric generator fabricated on a
performance for TE cooler. To be specific, the maximum glass fabric, Energy Environ. Sci. 7 (2014) 1959e1965.
output power and maximum efficiency of TE module using [10] R.C. O’Brien, R.M. Ambrosi, N.P. Bannister, S.D. Howe, H.V. Atkinson, Safe
radioisotope thermoelectric generators and heat sources for space applica-
more TE couples are increased by 1% and 1.2% respectively, tions, J. Nucl. Mater. 377 (2008) 506e521.
and the COP of TE module using a greater cross-sectional [11] A.D. LaLonde, Y. Pei, H. Wang, G. Jeffrey Snyder, Lead telluride alloy thermo-
area of legs is increased by 118.7% when I ¼ 4.5A. electrics, Mater. Today 14 (2011) 526e532.
[12] D. Kraemer, Q. Jie, K. McEnaney, F. Cao, W. Liu, L.A. Weinstein, et al.,
(3) The lower height and the higher thermal conductivity of Concentrating solar thermoelectric generators with a peak efficiency of 7.4%,
ceramic plates are, the better performance of TE module Nat. Energy 1 (2016) 16153.
becomes, both for TE generator and TE cooler. Further, more [13] Sark WGJHMv, Feasibility of photovoltaic e thermoelectric hybrid modules,
Appl. Energy 88 (2011) 2785e2790.
attention should be paid in improving the thermal conduc-
[14] W. He, J. Zhou, J. Hou, C. Chen, J. Ji, Theoretical and experimental investigation
tivity due to the need for a high enough height of ceramic on a thermoelectric cooling and heating system driven by solar, Appl. Energy
plates to withstand mechanical pressure. 107 (2013) 89e97.
[15] P. Naphon, S. Wiriyasart, Liquid cooling in the mini-rectangular fin heat sink
(4) For TE generator, both the output power and conversion ef-
with and without thermoelectric for CPU, Int. Commun. Heat Mass Tran. 36
ficiency are decreased with the increase of ambient (2009) 166e171.
convective coefficient, rather it is opposite for TE cooler. [16] H.-S. Choi, S. Yun, K-i Whang, Development of a temperature-controlled car-
Besides, the fillers are not recommended unless the thermal seat system utilizing thermoelectric device, Appl. Therm. Eng. 27 (2007)
2841e2849.
552 D. Luo et al. / Renewable Energy 154 (2020) 542e552

[17] M. Gillott, L. Jiang, S. Riffat, An investigation of thermoelectric cooling devices thermoelectric coolers using simplified conjugate-gradient method, Energy
for small-scale space conditioning applications in buildings, Int. J. Energy Res. 59 (2013) 689e697.
34 (2010) 776e786. [28] H. Lee, The Thomson effect and the ideal equation on thermoelectric coolers,
[18] G.J. Snyder, Application of the compatibility factor to the design of segmented Energy 56 (2013) 61e69.
and cascaded thermoelectric generators, Appl. Phys. Lett. 84 (2004) [29] C.-T. Hsu, G.-Y. Huang, H.-S. Chu, B. Yu, D.-J. Yao, An effective Seebeck coef-
2436e2438. ficient obtained by experimental results of a thermoelectric generator mod-
[19] M.S. El-Genk, H.H. Saber, T. Caillat, Efficient segmented thermoelectric uni- ule, Appl. Energy 88 (2011) 5173e5179.
couples for space power applications, Energy Convers. Manag. 44 (2003) [30] X. Lu, X. Yu, Z. Qu, Q. Wang, T. Ma, Experimental investigation on thermo-
1755e1772. electric generator with non-uniform hot-side heat exchanger for waste heat
[20] D. Luo, R. Wang, W. Yu, W. Zhou, A novel optimization method for thermo- recovery, Energy Convers. Manag. 150 (2017) 403e414.
electric module used in waste heat recovery, Energy Convers. Manag. 209 [31] W. Li, J. Peng, W. Xiao, H. Wang, J. Zeng, J. Xie, et al., The temperature dis-
(2020) 112645. tribution and electrical performance of fluid heat exchanger-based thermo-
[21] Q. Ma, H. Fang, M. Zhang, Theoretical analysis and design optimization of electric generator, Appl. Therm. Eng. 118 (2017) 742e747.
thermoelectric generator, Appl. Therm. Eng. 127 (2017) 758e764. [32] S. Lan, Z. Yang, R. Chen, R. Stobart, A dynamic model for thermoelectric
[22] J.-H. Meng, X.-X. Zhang, X.-D. Wang, Characteristics analysis and parametric generator applied to vehicle waste heat recovery, Appl. Energy 210 (2018)
study of a thermoelectric generator by considering variable material prop- 327e338.
erties and heat losses, Int. J. Heat Mass Tran. 80 (2015) 227e235. [33] X. Hu, P. Jood, M. Ohta, M. Kunii, K. Nagase, H. Nishiate, et al., Power gener-
[23] W.-H. Chen, C.-Y. Liao, C.-I. Hung, A numerical study on the performance of ation from nanostructured PbTe-based thermoelectrics: comprehensive
miniature thermoelectric cooler affected by Thomson effect, Appl. Energy 89 development from materials to modules, Energy Environ. Sci. 9 (2016)
(2012) 464e473. 517e529.
[24] D. Luo, R. Wang, W. Yu, Comparison and parametric study of two theoretical [34] E. Sandoz-Rosado, R. Stevens, Robust finite element model for the design of
modeling approaches based on an air-to-water thermoelectric generator thermoelectric modules, J. Electron. Mater. 39 (2010) 1848e1855.
system, J. Power Sources 439 (2019) 227069. [35] W.-H. Chen, P.-H. Wu, Y.-L. Lin, Performance optimization of thermoelectric
[25] D. Ebling, M. Jaegle, M. Bartel, A. Jacquot, H. Bo €ttner, Multiphysics simulation generators designed by multi-objective genetic algorithm, Appl. Energy 209
of thermoelectric systems for comparison with experimental device perfor- (2018) 211e223.
mance, J. Electron. Mater. 38 (2009) 1456e1461. [36] Q. Zhang, J. Liao, Y. Tang, M. Gu, C. Ming, P. Qiu, et al., Realizing a thermo-
[26] D. Luo, R. Wang, W. Yu, W. Zhou, Performance evaluation of a novel ther- electric conversion efficiency of 12% in bismuth telluride/skutterudite
moelectric module with BiSbTeSe-based material, Appl. Energy 238 (2019) segmented modules through full-parameter optimization and energy-loss
1299e1311. minimized integration, Energy Environ. Sci. 10 (2017) 956e963.
[27] Y.-X. Huang, X.-D. Wang, C.-H. Cheng, D.T.-W. Lin, Geometry optimization of