Journal of Energy Storage 41 (2021) 102961

Contents lists available at ScienceDirect

Journal of Energy Storage

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

Computational modeling of a thermal energy storage tank coupled to a

water-cooled household refrigerator
Felipe Lima Castro Gerhardt *, Ricardo de Vasconcelos Salvo, Ismael de Marchi Neto,
Rafael Sene de Lima, Rodrigo Corrêa da Silva
Universidade Tecnológica Federal do Paraná Campus Londrina, Avenida dos Pioneiros, 3131, CEP 86036-370, Londrina, PR, Brazil

A R T I C L E I N F O A B S T R A C T

Keywords: In this work, two-dimensional numerical simulations of a thermal energy storage tank coupled to a household
Thermal energy storage tank refrigerator through a shell and tube heat exchanger studies are performed. The geometry was developed in
2D unsteady CFD simulation SpaceClaim from ANSYS, whereas the unstructured quadrilateral mesh was developed in ANSYS ICEM© and the
Thermal stratification
simulations performed in ANSYS Fluent©. The boundary conditions used on the tank walls are the non-slip and
Energy reuse
adiabatic. Six simulations were performed: three of them with prescribed heat flux and three with prescribed
temperature, all values were associated with the experiment data. Due to the two-dimensional modeling, a heat
flux correction is proposed. Furthermore, the thermal coefficient expansion variation effect on the tank per­
formance is analyzed. The Boussinesq approximation is used to account for the density variation with the fluid
temperature. The sensible heat storage performance of the tank is assessed employing the Richardson Number,
Stratification Number, energy and exergy content. The numerical results are validated with experimental data.
Prescribed heat flux is the appropriate boundary condition to model the heat transfer inside the heat exchanger.
Also, the heat flux correction represents qualitatively and quantitatively the experiment. The thermal expansion
coefficient value influences significantly the water thermal stratification inside the TES tank.

(continued )
Nomenclature
A total exergy of stratified tank (J) u velocity vector (m /s) t time(s) exp experiment
Am total exergy of mixtured tank (J) Vol volume (m3) T temperature (K) cal calculated
C specific heat of the fluid (J /kg.K) v inlet velocity (m /s) Tm mixed temperature (K) top top of the tank
E total energy (J) Te equivalent exergy temperature (K) bot bottom of the tank
e specific energy (J /kg) Greek Symbols T0 reference temperature (K) cor corrected flux
g gravity (m /s2 ) β Termal expansion coefficient
(1 /K)
h height (m) ∇ vector operator
H total height (m) Δ difference operator 1. Introduction
k thermal conductivity (W /m.K) μ dynamic viscosity (Pa.s)
m fluid mass (kg) ρ Density (kg /m3 ) Usage of energy resources is an important theme in sustainable
mw mass of water between τ stress tensor vector (kg /m.s) development [1]. Studies show that almost 29% of the world electric
thermocouples (kg)
energy consumption is in the residential sector [2]. Refrigeration and
p pressure (Pa) ξ specific exergy (J /kg.K)
P perimeter (m)
air-conditioning systems represent about 17% of the electric energy
q linear heat flux (W /m) Subscripts and superscripts consumption, while 15.4% of this share is within the residential sector
Q volumetric flow rate (m3 /s) in inlet [3]. In Brazil, between the years of 2009 and 2018, the residential
r radius (m) out outlet electric energy consumption has increased about 35.4%. In 2018, the
Ri Richardson number acu accumulated residential sector represented nearly 25.4% of the country’s total elec­
Str Stratification number sim simulation
tric energy consumption (535.4 TWh) [4]. The Energy Research Office
(continued on next column) (EPE - Portuguese acronym) 2016 report has shown that water-heating

* Corresponding author.
E-mail address: flc.gerhardt@gmail.com (F.L.C. Gerhardt).

https://doi.org/10.1016/j.est.2021.102961
Received 12 December 2020; Received in revised form 12 June 2021; Accepted 16 July 2021
Available online 29 July 2021
2352-152X/© 2021 Elsevier Ltd. All rights reserved.
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

devices represents 21.9% of the electric energy consumption of a com­ approximation and energy equation.
mon Brazilian domicile [5]. Furthermore, a previous study [6] shows Concerning works that combined numerical and experimental ob­
that household refrigerator possession has increased from 87.9% to servations, Álvares et al. [11] studied experimentally and numerically,
94.5% of the country total domiciles between the years of 2005 to 2012. at Fluent® software, the performance of a TES tank connected to a solar
The same study points that the household refrigerator represents 18.3% panel. Three-dimensional simulations and a k − − ε turbulence model
of the domestic electric energy consumption. Worldwide, there are were used. The authors compared experimental and numerical data
approximately one household refrigerator for about six people, making validating the model. Savicki et al. [24] studied numerical and experi­
relevant practices of energy reuse of this device [7]. mentally the thermal and hydrodynamic behavior of a TES tank heated
Sustainable energy generation development and the demand for by electric resistance. The three-dimensional simulations were per­
increasing the efficiency of energy reuse systems aroused interest on formed at transient state. The authors studied the effect of a deflector
energy storage, making it a key aspect for sustainable development that prevented thermocline degradation on the tank. Shin et al. [25]
[8-9]. Thermal energy storage tanks (TES) are generally used in energy compared the numerical results of an experimental study that evaluated
generation systems whose offer and demand are incompatible, such as the geometry and operational factors effects on a TES tank stratification.
solar energy [8]. Besides that, TES systems are also applicable for do­ The authors used two variations of the k – ε turbulence model and a
mestic usage [10]. Many experimental and numerical studies evaluated “plug-type” flow model.
aspect ratio, mass flow rate, obstacles, and fluid inlet positioning on the Previous investigations studied the performance of a TES tank in
TES tanks performance [10-15]. These parameters have relevance on the different operational and geometric conditions. TES systems perfor­
fluid’s thermal stratification, which influences TES efficiency [16-17]. In mances are usually evaluated by effectiveness and efficiency. Thermal
comparison, TES efficiency of a fully stratified tank is 6% to 10% higher efficiency is a method to evaluate how efficiently the thermal energy is
than of a fully mixed tank [10, 18]. Besides the energetic reuse, studies stored or removed. The terms “charging” and “discharging” process are
suggest that the refrigerator’s coefficient of performance (COP) in­ related to the tank’s application, whereas “charging” process usually
creases when connected to TES systems, generating savings of 7% to refers to hot fluid entering the tank and “discharging” to the fluid being
15% on the compressor electric energy consumption, due to the higher removed [26]. In the evaluated studies, only three authors studied
natural convection coefficient of the system [15, 20]. experimentally the system as a whole [15, 20, 22].
TES studies can be performed through one of three approaches: Considering numerical simulation there is no study that considers a
experimental, numerical simulation, or a combination of both. Consid­ sensible heat TES tank coupled to a household refrigerator. Besides, TES
ering the experimental work, one can highlight the studies proposed by tank researchers tend to perform three-dimensional simulations [7, 10,
Marchi [14], published in paper format later by Marchi et al. [15], 11, 12, 24, 40, 43] despite the computational cost and simulation time,
where the performance of a TES tank coupled to a household refrigerator as two-dimensional approach are considered to oversimplify
through a shell and tube heat exchanger was evaluated. The authors three-dimensional mixing effects. Also, authors tend not to reveal the
compared the COP of the conventional refrigerator system and the thermal expansion coefficient, which is an important parameter when
modified refrigerator system. In posterior work, the authors determined using the Boussinesq approximation. The potential practical contribu­
energy and exergy contents in the TES tank [19]. Zuzarte and tions of TES tank usage are extensively studied and its applicability
Simões-Moreira [20] studied the economic impact of domestic usage of ranges from solar energy storage [1, 8, 9, 11, 16] to water heating with
TES tanks. The researchers coupled a tank to the condenser of a household refrigerators [15, 19, 22]. Both can potentially generate
household refrigerator and used it as a water heater. The study has considerable energy savings for water heating, which can cause financial
shown that a 25 L TES tank can heat water from 20◦ C to 40◦ C in 4 hours and environmental impact. Also, the refrigerator operated without
for six times a day, representing an economy of BRL R$ 30.60 and BRL R thermal load, and the experimental project considered not only its
$ 34.20 per month comparing with electric and gas heating devices, cooling capacity, but also the compressor capacity, enabling water flow
respectively. Also, Khalifa et al. [21] studied experimentally the effect of through thermosiphon principle. Thus, the current study investigates
aspect ratio on a TES tank at cooling mode. The authors concluded that numerically the performance of a TES tank connected to a household
the main aspect that leads to stratification degradation for uninsulated refrigerator, comparing it with the results of the previous experimental
tanks is the heat loss for the environment. Moreover, Abu-Mulaweh [22] study of Marchi et al. [15]. A two-dimensional approach is used to
studied experimentally the performance of a TES tank coupled to an evaluate whether the mixing effect presents reasonable results. Also, this
air-conditioning system though different heat exchanger. The authors study contains significant information about thermal expansion coeffi­
showed that the system could heat the tank to about 43◦ C, with a top to cient effect on water stratification. The experimental apparatus used in
bottom temperature difference of 10◦ C. the experimental study is described in Section 2. The mathematical
Regarding numerical simulations, Ievers and Lin [10] performed model and scheme of the geometry model are described in Section 3. At
transient three-dimensional numerical simulations of a TES tank in the same section, the boundary conditions are presented. In Section 4,
order to evaluate aspect ratio, inlet mass flow, and inlet and outlet po­ the results of both studies are compared, and the numerical model is
sition influence on the tank’s performance. The authors developed the validated through the quali-quantitative comparison of the TES tank
turbulence model, and it was concluded that stratification degradation is model simulated with the experiment, also the performance of the tank
directly proportional to mass flow and inlet and outlet distance from top is evaluated using three parameters: Richardson Number, Stratification
and bottom. Furthermore, the authors concluded that the higher the Number, energy, and exergy contents. Finally, conclusions are presented
aspect ratio, the better the tank’s performance. Yaici et al. [12] used the in Section 5.
commercial software COMSOL® to study numerically the aspect ratio
and operational parameters variation effects on TES tank performance 2. Experimental setup and procedures
on three-dimensional simulations. Conjugate heat transfer and laminar
flow were used to model the heat transfer and slow-moving flow, The aim of this work is to create and validate a model of a previous
respectively. The study was performed at transient state and it has experimental study, performed by Marchi et al. [15]. The main experi­
shown that the higher the aspect ratio and inlet and outlet temperature mental apparatus is a refrigerator which consists of a 263 L refrigeration
differences, the better the tank’s performance. In addition, Barzegar and cabinet and a 74 L freezing cabinet. The refrigerator electric energy
Dehghan [23] studied the performance of a TES tank on discharging consumption was 49 kWh/month, with its condenser replaced by a
mode with Grashof and Reynolds number on two-dimensional axisym­ counter-current shell and tube heat exchanger that aims to condensate
metric transient numerical simulations. The study was performed with a the superheated refrigerant gas using circulating water. The average
low Reynolds number k − ω turbulence model, Boussinesq power of the compressor is 123 W. The water was stored in a PVC tank of

2
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

Fig. 1. Schematic diagram of the experimental apparatus [15].

Fig. 2. Model’s geometry and mesh.

122 L, 0.3 m in diameter, 1.75 m in length, and 0.004 m thickness. The drained to the heat exchanger through pressure differences induced by
tank was insulated with a 0.02 m acrylic blanket of approximately the inlet fluid flow promoting circulation. The temperature was
W
0.11 mK thermal conductivity, which provided the system with measured by 31 T-type copper-constantan thermocouples, spaced 50
approximately 0.683 mKW
global heat exchange coefficient and 1.464 m.K mm among each other along a probe placed on the entire vertical
W
linear thermal resistance, along with the PVC insulation. The heated extension of the tank to obtain the temperature profiles. The thermo­
water was transported to the upmost part of the tank through the couples were connected to a data acquisition system where the tem­
thermosiphon principle, which is known to provide fluid flow in closed perature information was collected every 30 min and stored in a
systems through density differences in fluids induced by temperature computer. The experiment was performed throughout 168 h, i.e., 7 days.
[20, 22, 46]. The colder water in the lowermost part of the tank was The schematic diagram of the experimental apparatus used by the au­
thors is in Fig. 1.

3
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

3. Mathematical model Table 1

Water proprieties for the numerical simulations
The geometry of the transient two-dimensional planar model, which (
kg
)
998.3
included the pipes, shell and tube heat exchanger and tank, was devel­ ρ 3
(m )
1 3.10− 4
oped in SpaceClaim from ANSYS® and is illustrated along with the β ∘
( C)
mesh, Fig. 2. The mesh was made of unstructured quadrilateral ele­ kg 1.003 .10− 3
μ
ments. A mesh independency test was performed with three meshes with (m.s )
j 4182
different number of elements, 72.432, 141.735 and 296.135, by calcu­ C
( kg.k
w )
lating the Grid Convergence Index (GCI) as Roache [27] work describes. k 0.6
The error between the simulation with the lowest mesh element number
m.k

and the Richardson extrapolation is 1.12%, while the GCI index is

0.33%. Considering the computer’s processor was an i7-6700K, along including the approximated average (60◦ C), were evaluated. Consid­
with 8192 Mb of RAM and 292 GB space disk, an intermediate ering prescribed flux, the tank energy stored throughout 24h of the
time-consuming mesh composed of 141.735 elements was selected to experiment (4.506 kJ or 1.2517 kWh), calculated by Marchi et al. [15]
perform the simulations without compromising precision (Figs 7, and 8). as shows equation (4), was used to estimate the heat flux, as the adia­
The finite-volume method with fully implicit second-order temporal batic boundary condition could compensate for the unaccounted heat
discretization is applied to the numerical solution on the governing loss for the environment of the simulation. As the tank was insulated in
equations. Coupling between pressure and velocity fields is performed the experiment, the walls of the model were configured with no-slip and
with Simple algorithm. The criteria for iteration convergence are that adiabatic boundary conditions. Marchi et al. [15] experiment conditions
the residuals of continuity and momentum equations are smaller than led to a low refrigerator COP of 0.35, which was caused by the lack of
10− 5 and smaller than 10− 6 for energy equation. thermal load and the continuous compressor operation. A first law of
The following assumptions are considered to model the heat transfer Thermodynamics analysis of the TES tank, given the energy stored in the
inside the TES tank with water as a working fluid: tank and the energy output of the refrigerator condenser, shows that the
heat loss for the environment is approximately 2.430 kJ (or 0.675 kWh).
• The working fluid is incompressible. Although not negligible, the heat loss for the environment is already
• The thermophysical proprieties of the fluid are constant except for accounted in the prescribed heat flux of the simulations.
the density variation with temperature, using Boussinesq approxi­ nd ∑
∑ 30
[ ( )]
mation, responsible for generating the thermal buoyancy. Eacu = mw .C. Tm, ti − Tm, ti+1 (4)
• The fluid is considered Newtonian. i=1 n=1

• Negligible viscous dissipation.

• Fluid motion is laminar and two-dimensional. where mw is the mass of water between thermocouples, C is the specific
heat at constant pressure, and Tm the temperature collected in the
The two-dimensional energy and Navier-Stokes equations are used to thermocouples.
determine the transient thermal stratified behavior. The governing The heat flux was calculated as the ratio of accumulated energy and
equations consider gravity effect and can be written as: time, and was converted to linear heat flux as shown in equation (5):
The continuity equation: Eacu
qcal = (5)
∇.u = 0 (1) Δt.P

where Eacu is the accumulated energy stored in the TES tank, Δtis the
where u is the fluid velocity.
time interval of 24h, and P the perimeter of the line representing the
The momentum equation:
tube in the shell and tube heat exchanger.
∂u The total volume of the experiment, considering the measures pre­
ρ. + (ρ.u . ∇).u = − ∇.p + ∇ . τ − ρ.β.(T − T0 ) .g (2)
∂t sented in the paper, was approximately 122 L (Volexp), whereas the
extruded model concept of 1 m in width used to convert superficial to
where ρ, β, τ, p, g and (T − T0 ) are the fluid density, thermal expansion linear flux was 590 L (Volsim ). The corrected heat flux is the calculated
coefficient, the stress tensor, the static pressure, gravity, and the dif­ heat flux amplified by the ratio between the volumes, as shows equation
ference between current and reference temperature, respectively. (6):
The energy equation: ( )
Volsim
∂T qcor = .qcal (6)
ρ . C. + ρ . C.u.∇T = ∇.(k.∇.T) (3) Volexp
∂t

where C and k are the fluid specific heat at constant pressure and the
conduction coefficient, respectively. Table 2
Boundary conditions and thermal expansion coefficient configured in the
3.1. Boundary conditions simulations
Walls
The boundary conditions selected were applied to ensure realistic Non-slip and adiabatic
Heat exchanger’s tube
results considering the closed system. The temperature data is collected
Simulation Sim 1 Sim 2 Sim 3 Sim 4 Sim 5 Sim 6
at the respective heights in that the thermocouples were placed in the Wall Temperature - - - - 76.85 60 40
experiment. The boundary condition applied to the lines that represent T (◦ C)
the tube in the heat exchanger was a subject of the study, as prescribed Wall heat Flux –q 52.15 241.93 241.93 - - -
(W/m2)
flux and temperature were both reasonable options. The superheated
Water Thermal 3.10− 4
3.10− 4
3.10− 5
3.10− 4
3.10− 4
3.10− 4
refrigerant gas fluid inside the condenser transfers heat through a phase Expansion
change, keeping the temperature of the tube approximately constant. As Coefficient -
( )
the authors of the experiment had registered the temperature on the β
1
inlet (76.85◦ C) and outlet (40◦ C) of the condenser, these temperatures, K

4
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

The initial conditions are static fluid with a temperature of 22.5◦ C, and second laws of Thermodynamics and is considered a key parameter
the same as the experiment. Gravity is -9,81 m/s2. The Boussinesq in sensible heat storage tanks [10, 19, 42-45]. This parameter permits a
approximation, equation (7), requires the thermal expansion coefficient quantitative evaluation of stratification and thermal losses in the TES
to be set. The values of the fluid proprieties correspond to the average tank due to environment temperature. The energy and exergy content
temperature (30◦ C) of the first 24h of the experiment, Table 1, as a were described by Rosen [45] as:
model simplification. ∫
E = e dm (13)
(ρ − ρ0 ).g ≈ − ρ0 .β. (T − T0 ).g (7) m

The thermal expansion coefficient greatly influences water stratifi­ ∫

cation behavior. As the simulations are a 2D representation of a 3D A= ξ dm (14)
m
phenomenon, a simulation with a different thermal expansion coeffi­
cient, equivalent to this propriety of water at 6◦ C, was included in the where, given a reference temperature T0 , the specific energy (e) and
study. Table 2 is a summary of the boundary conditions of the six sim­ exergy (ξ) can be expressed as a function of height (h), given by:
ulations performed in the current study.
e(h) = C.(T(h) − T0 ) (15)
3.2. Richardson number ( ) ( )
T T
ξ(h) = e(h) − T0 .ln = C.(T(h) − T0 ) − T0 .ln (16)
T0 T0
The Richardson Number commonly describes stratification on sen­
sible heat storage tanks [28-36]. It measures the ratio of buoyancy forces The fluid mass (m) inside the TES tank can be expressed as:
to mixing forces, and can be expressed by [37]: m
( ) dm = dh (17)
g.β.H Ttop − Tbot H
Ri = (8)
V2 where H is the total height of the tank.
With equations (13), (14) and (15), E and A can be rewritten as:
where g is gravitational acceleration, β is the thermal expansion coef­
ficient, His the tank height, (Ttop − Tbot ) is the temperature gradient ∫H
m
between top and bottom fluid, and v the fluid inlet velocity, calculated E= e(h) dh (18)
H
by: 0

Q
v= (9) ∫H
π.r2in m
A= ξ (h) dh (19)
H
where Q is the volumetric flow rate and rin is the inlet pipe radius. 0

A large Richardson Number means a stratified storage tank, while a With equation (15), equation (18) can be rewritten as:
small Richardson Number means a mixed storage tank.
E = m.C.(Tm − T0 ) (20)
3.3. Stratification Number
where C = CP and Tm is expressed by:

TES tank researchers commonly use the Stratification Number ∫H

1
[38-41]. Fernández-Seara et al. [38] proposed this parameter as it de­ Tm = T(h)dh (21)
H
scribes the evolution of thermal stratification inside the tank during both 0
static and dynamic periods. The authors expressed the Stratification
Physically, Tm represents the temperature of the fluid when it is fully
Number as the ratio of the mean of the transient temperature gradients
mixed. By the conservation of energy principle, it can be inferred that
to the maximum mean temperature gradient for the charging and dis­
the energy of a fully mixed tank is the same as a stratified tank, that is:
charging processes, given by equation (10):
( ) E = Em (22)
∂T
∂h With equation (16), equation (19) can be rewritten as:
Str = ( ) t
(10) ( )
Te
(23)
∂T
∂h A = E − m.C.T0 .ln
max T0

where T is the temperature, his the height (axial position), and t the where Te represents the equivalent temperature of a mixed TES tank
time. Equations (11) and (12) represent the numerator and denominator with the same exergy as the stratified tank:
terms, respectively: ⎡ H ⎤
∫
( ) 1
∂T 1 ∑J− 1
Ti+1 − Ti Te = exp ⎣ lnT(h)dh⎦ (24)
= (11) H
∂h t J − 1 i=1 Δh 0

( ) Since Te is dependent on the degree of stratification, in general Te ∕

=
∂T
=
Tmax − Tmin
(12) Tm and the limit of Te = Tm is reached when the TES tank is fully mixed.
∂h max (J − 1).Δh This can be seen by noting that the exergy of a fully mixed tank is:
( )
Tm
3.4. Energy and exergy content Am = Em − m.C.T0 .ln (25)
T0
The energy analysis of a TES tank is a thermodynamic analysis based The difference in the TES tank exergy between the fully mixed and
on the first law of Thermodynamics. Energy analysis is a widely used stratified cases, with equations (23) and (25), can be expressed as:
parameter even though it does not compute stratification details. The
exergy content is a thermodynamic analysis technique based on the first

5
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

Fig. 3. Simulation 1: (a) height versus temperature for simulation hours; (b) temperature versus time for height position.

Fig. 4. Simulation 2: (a) height versus temperature for simulation hours; (b) temperature versus time for height position

( )
Tm initiating with the first vertical line at 22.5◦ C. In Fig. 3, referring to Sim
A − Am = m.C.T0 .ln (26)
Te 1, there is a gradual increase in the temperature of the isothermal layers
with time. A rise in the temperature of the lowermost fluid layer occurs
4. Numerical results only at the seventh hour. As in the experiment, in Fig. 3a, it is possible to
identify all isothermal layers, as they do not overlap each other. The
For the sake of understanding, numerical results are organized into highest temperature gradient between the top and bottom fluid is at the
two groups according to the type of boundary condition adopted: item seventh hour, reaching 0.84◦ C. At the end of 24h, the temperature
4.1 prescribed heat flux and item 4.2 prescribed temperature. The gradient stabilizes to 0.77◦ C. The temperature data collected from Sim 2
qualitative results, items 4.1 and 4.2, are presented as thermocline in a are in Fig. 4. In Sim 2, the temperature of the isothermal layers rises
function of time. This method allows a qualitative comparison of the faster than in Sim 1. A rise in the temperature of the lowermost data
numerical and experimental stratification. Quantitative results, item collector is at the fourth hour of the simulation. The highest temperature
4.3, are calculated by the difference and standard deviation of temper­ gradient is at the eighth hour, reaching 1.748◦ C, stabilizing in 1.721◦ C
ature data obtained in the simulations and the experiment given the throughout the simulation. From the last ten hours of the simulation, the
respectively collected data height and time. rate of temperature rising of the uppermost data collector is 0.323◦ C/h.
The temperature data collected from Sim 3 are in Fig. 5. The thermal
expansion coefficient has an important influence in qualitative results
4.1. Qualitative Analysis - Heat flux boundary condition simulations since the difference in the fluid stratification behavior observed in Sim 2
and Sim 3 is relevant. The lowermost data collector shows a rise in
Sim 1 through Sim 3 stratification behavior shows qualitative simi­ temperature only at the tenth hour. Furthermore, the temperature
larities to the experiment, as shown in Fig. 3 through Fig. 5. The lines gradient between the top and bottom fluid is higher throughout the
alternating in shape and color represent each hour thermocline, simulation, reaching its peak of 5.78◦ C at the tenth hour. This gradient

6
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

Fig. 5. Simulation 3: (a) height versus temperature for simulation hours; (b) temperature versus time for height position

Fig. 6. Simulation 4: (a) height versus temperature for simulation hours; (b) temperature versus time for height position

does not appear to stabilize during the simulation. The rate of temper­ sixth hour of the simulation. In Fig. 10 are the temperature data
ature rising of the uppermost data collector is 0.299◦ C/h. In Fig. 3 and collected from Sim 6. The temperature gradient between the top and
Fig. 4, it can be seen that a higher thermal expansion coefficient causes bottom fluid is the highest in the first hour of the simulation, reaching
lower water layers to heat faster. 6.77◦ C, and it stabilizes at 2.94◦ C in the final hours. The top fluid rea­
ches 99.60% of the final temperature in the sixth hour of the simulation.
4.2. Qualitative Analysis - Temperature boundary condition simulations In Fig. 6 through Fig. 10, it can be seen that stratification behavior
differs qualitatively from the one observed in Fig. 9. Prescribed tem­
In Fig. 6 are the temperature data collected from Sim 4. The tem­ perature boundary condition, along with the absence of convection heat
perature gradient between the top and bottom fluid reaches its peak of loss, does not seem to represent faithfully the experimental system, in
18.15◦ C in the first hour of the simulation. In the final hours of the which the data is presented in Fig. 9. In addition, the thermal expansion
simulation, the temperature gradient tends to 3.70◦ C. The adiabatic coefficient should vary with temperature. For instance, the thermal
boundary condition configured on the wall makes the fluid temperature expansion coefficient of water at 75◦ C is 6.11 .10− 4 1/K, approximately
reaches the prescribed temperature rapidly, as the top fluid reaches 2 times greater than the one calculated by the average temperature of
99.04% of its final temperature at the fifth hour of the simulation. the first 24h of the experiment. This unaccounted variation influences
Convection inducted heat loss has a significant influence on fluid directly on the fluid buoyancy, as seen in Boussinesq approximation,
stratification behavior. The temperature data collected from Sim 5 are in equation (7), which leads to a lower stratification.
Fig. 9. Similarly to Sim 4, the temperature gradient between the fluid at
the top and bottom is the highest in the first hour of the simulation, 4.3. Quantitative Analysis
reaching 12.43◦ C. The temperature gradient behavior tends to stabilize
at 4.05◦ C. The top fluid reaches 99.47% of its final temperature at the The quantitative analysis was performed through the calculation of

7
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

Fig. 7. Simulation 5: (a) height versus temperature for simulation hours; (b) temperature versus time for height position

Fig. 8. Simulation 6: (a) height versus temperature for simulation hours; (b) temperature versus time for height position

temperature differences between simulation and experiment data, Fig 5, it can be seen that the temperatures reached in the simulations
measuring temporal and height disparities. Table 3 presents the tem­ underestimated the heat expansion coefficient until the 20th hour for
perature difference average and standard deviation for the simulations. Sim 2 and 16th hour for Sim 3. Given that, the bottom of the tank
In Fig. 10 are the temperature differences average and standard devia­ experienced an increase in temperature earlier than the experiment
tion accounting all temperature data for each simulation. Negative while the fluid from the top of the tank did not retain as much tem­
values mean the average temperature data in the simulated tank is lower perature as the experiment. This process can be explained by the higher
than in the experiment. Closer to zero averages means that the simula­ diffusion that occurs between the water layers because of the higher
tion results and the experimental data are in good agreement. fluid movement occasioned by the higher thermal expansion coefficient.
The data from Sim 2 have closer to zero average despite Sim 3 pre­ The data from Sim 3 shows that fluid from the top of the tank reaches
sents lower standard deviation. Although Sim 3 thermal expansion co­ temperature closer to the experiment, while fluid from the bottom of the
efficient is physically inconsistent, its qualitative and quantitative tank shows greater temperature differences with the experiment. This
results are more similar to those of the experiment, given the thermo­ behavior is because the thermal expansion coefficient at 6◦ C
cline behavior, temperature differences average, and standard devia­ (3.10− 5 1/K) causes less fluid movement, accumulating hotter fluid at
tion. This behavior occurs because the thermal expansion coefficient the top and colder fluid at the bottom because of the lower diffusion
changes approximately linearly with temperature, i.e., at 22.5◦ C to 34◦ C process. Most TES tank researches do not specify the thermal expansion
the thermal expansion coefficient of water assumes values of 2.31.10− 4 1 coefficient used in simulations.
/K to 3.36.10− 4 1/K. As the simulations were performed with a constant Fig. 10 and Fig. 11 present detailed information about the influence
thermal expansion coefficient at the average temperature of the exper­ of height and time on temperature differences. Fig. 10 shows that data
iment (30◦ C), the fluid inside the tank experienced more movement on collected from the top of the simulated tank does not reach precisely the
temperatures below 30◦ C and less movement above 30◦ C. In Fig. 3 to temperature of the uppermost thermocouple of the experiment. In Sim 2,

8
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

Fig. 9. Marchi et al. [15]: (a) height versus temperature for simulation hours; (b) temperature versus time for height position

Fig. 10. Temperature differences average and standard deviation in function of height for (a) Sim 2; (b) Sim 3

time. Both curves exhibit a local peak at the fifth, ninth, eleventh, and
Table 3 thirteenth hour, followed by a rise in the average. In Sim 3, the standard
Simulation, temperature differences average and standard deviation deviation reaches less than 1◦ C from the thirteenth hour. Lower values
Simulation Average (◦ C) Standard Deviation (◦ C) of thermal expansion coefficient accumulate energy on higher water
Sim 1 − 6.44 3.40 layers by decreasing the intensity of mixing between layers. Otherwise,
Sim 2 − 3.41 2.26 less energy reaches lower water layers.
Sim 3 − 3.44 1.65
Sim 4 41.15 9.52
4.4. Richardson Number
Sim 5 25.26 6.01
Sim 6 7.43 2.70
Table 4 presents the average for Richardson Number, inlet velocity,
and temperature gradient between the top and bottom fluid for the
simulations. Fig. 12 shows the values of the Richardson Number for the
the greatest average is -6.33◦ C and is located at the topmost data col­
simulations. Fig. 13 presents the results for inlet velocity and tempera­
lector, while the lowest average is -1.23◦ C and is located at the lower­
ture gradient between the top and bottom. The calculation used Sim 2
most data collector. In Sim 3, the greatest average is -4.90◦ C, at the third
data once experimental inlet velocity was not measured.
position from top to bottom data collector, while the lowest average is
The order of magnitude of Richardson Number for flux prescribed
-2.49◦ C, at the fifth from bottom to top data collector. In Fig. 11, Sim 2
simulations, seen in Fig. 12a, is 2 to 3 times smaller than temperature
greatest average is -4.81◦ C, located at the twelfth hour, while Sim 3
prescribed simulations, seen in Fig. 12b. As seen in Table 4 and Fig. 13b,
greatest average is -4.78◦ C, and it is located at the fourteenth hour. A
Sim 3 average temperature gradient is higher than other flux prescribed
pattern is observed in the temperature difference average behavior with
simulation. In addition, its average inlet velocity is smaller than all

9
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

Fig. 11. Temperature differences average and standard deviation in function of time for (a) Sim 2; (b) Sim 3.

that the temperature gradient between the top and bottom of the tank is
Table 4
significantly higher than the simulations. The results for prescribed
Average values of Richardson number, inlet velocity and temperature gradient
temperature simulations are in Fig. 12b. For prescribed temperature
for the simulations and experiment
simulations, Sim 4 and Sim 5 presents larger Richardson Numbers. Even
Simulation Richardson Inlet velocity (m/ Temperature though the temperature gradient between the top and bottom fluid is not
Number (103 ) s.10− 3 ) gradient (◦ C)
considerably different to establish a correlation between this parameter,
Sim 1 1.250 1.76 0.68
velocity inlet results show significant variation, seen in Fig. 13a. For
Sim 2 0.496 3.99 1.57
temperature prescribed simulations, inlet velocity peaks at initial hours,
Sim 3 1.650 1.28 4.42
as temperature gradient is considerably higher and fluid movement is
Sim 4 1540 3.66 5.24
dependent on a fluid density gradient. Richardson Number is inversely
Sim 5 1530 3.14 5.00
proportional to the square of the inlet velocity, making it sensible to the
Sim 6 1140 3.13 3.51
variation of this variable. In Fig. 6 and Fig. 9 the temperature gradient
Marchi et al. 2.16 3.99∗ 6.72
between the tank top and bottom fluid reaches constant values early at
[15]
the simulation, making the inlet velocity rapidly decrease. The physical
interpretation of the parameter leads to the conclusion that less fluid
simulations. Analyzing equations (7) and (8), one can infer that the movement is beneficial for maintaining the temperature of the
thermal expansion coefficient greatly influences the fluid buoyancy and isothermal layers.
makes it more relevant to inlet velocity and fluid stratification than its
direct correlation between Richardson Number. The main reason why
the Richardson Number of the experiment is higher in 0h-8h period is

Fig. 12. Richardson number for (a)flux boundary condition; (b) temperature boundary condition on the heat exchanger

10
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

Fig. 13. (a) Inlet velocity with time; (b) temperature gradient

Fig. 14. Stratification Number for (a) flux boundary condition; (b) temperature boundary condition on the heat exchanger

4.5. Stratification Number temperature prescribed simulations. The Stratification Number exhibits
a behavior pattern. For temperature prescribed simulations, the
The Stratification Number results are in Fig. 14. In Fig. 14a are the parameter oscillates in the first hours of the simulation and stabilizes
results for prescribed flux simulations, while in Fig. 14b is the result of afterward. These results agree with the thermocline profile shown in

Fig. 15. Energy content for stratified tank for (a) Sim 2, (b) Sim 3 and (c) experiment

11
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

Fig. 5, Fig. 6 and Fig. 9 but do not represent the experiment results. Declaration of Competing Interest
Otherwise, flux prescribed simulations results are similar to the exper­
iment, oscillating, and increasing simultaneously throughout the There are no conflicts to declare.
simulation.
References
4.6. Energy and exergy content
[1] Y. TIAN, C.Y. ZHAO, A review of solar collectors and thermal energy storage in
solar thermal applications, Applied Energy. Vol. 104 (2013) 538–553.
The energy and exergy content are calculated in Python IDLE by [2] INTERNATIONAL ENERGY AGENCY (IEA), Statistics Report. International Energy
numerical integration along with three models proposed by Rosen [45]. Outlook, OECD/IEA, Paris (France), 2019. Available :< https://www.eia.gov/
The models are the stepped temperature-distribution model, outlooks/ieo/pdf/ieo2019.pdf>. Last accessed in apr/2020.
[3] INTERNATIONAL INSTITUTE OF REFRIGERATION (IIR). The role of refrigeration
continuous-linear temperature-distribution model, and linear in the economy. 2016. Available: <http://www.iifiir.org>. Last accessed in may/
temperature-distribution model. The numerical results are compared 2019.
with the experimental results compiled by Marchi et al. [15]. The energy [4] EPE [Empresa de PesquisaEnergética] Brazilian Energy Balance (BEN) 2019: Base
Year 2018. Available in <https://ben.epe.gov.br >. Last accessed in apr/ 2020.
and exergy contents from Sim 2 and Sim 3 are compared to the exper­ [5] EPE [Empresa de Pesquisa Energética], Série: Projeções da Demanda de Energia
imental in Fig. 15, Fig. 16 and Fig. 17. For both simulation and exper­ Elétrica 2016-2020. Rio de Janeiro, Brasil: Ministério de Minas e Energia (MME) e
imental calculations, the reference temperature is constant at 22.5◦ C. Empresa de Pesquisa Energética (EPE) (2016).
[6] EPE [Empresa de Pesquisa Energética], Série: Estudos da Eficiência Energética. Rio
This approach leads to slightly different results from Del Rio Oliveira de Janeiro, Brasil: Ministério de Minas e Energia (MME) e Empresa de Pesquisa
et al. [19]. The exergy difference between the fully mixed and stratified Energética (EPE) (2014).
TES tanks are in the order of 10 J. From Fig. 15, Fig. 16 and Fig. 17, the [7] A.C. MARQUES, G.F. DAVIES, J.A. EVANS, G.G. MAIDMENT, I.D WOOD,
Theoretical modelling and experimental investigation of a thermal energy storage
simulations and experiment energy and exergy contents shows different
refrigerator, Energy. Vol. 55 (2013) 457–465.
behavior with time. Both simulation’s energy and exergy contents in­ [8] AMROUCHE S. OULD, D. REKIOUA, T. REKIOUA, S BACHA, Overview of energy
crease indefinitely, whereas the experiment tends to reach a peak by the storage in renewable energy systems, International journal of hydrogen energy. Vol
end of the 24h. This behavior can be attributed to the lack of convection 41 (2016) 20914–20927.
[9] J. XU, R.Z. WANG, Y LI, A review of available technologies for seasonal thermal
heat loss in the simulations as opposed to the experiment, where the energy storage, Solar Energy. Vol. 103 (2014) 610–638.
accumulated energy stored in the tank increases slower with time [10] S. IEVERS, W LIN, Numerical simulation of three-dimensional flow dynamics in a
because of this phenomenon. The energy analysis in Marchi et al. [15] hot water storage tank, Applied Energy. Vol. 86 (2009) 2604–2614.
[11] A. ÁLVARES, M. BAZ, O. CABEZA, J.L. FERRÍN, M.C. MUÑIZ, L.M VARELA,
experiment shows there was heat loss in the energy storage system, Experimental and numerical simulation of a storage tank connected to a flat-plate
which was addressed in the calculation of the prescribed flux boundary solar collector, Renewable Energy and Power Quality Journal. Vol 11 (2013)
condition. The non-linear behavior seen in Fig.15c, Fig. 16c and Fig. 17c 186–191.
[12] W. YAICI, M. GHORAB, E. ENTCHEV, S HAYDEN, Three-dimensional unsteady
can be explained by the increasing heat loss caused by higher temper­ CFD simulations of a thermal storage tank performance for optimum design,
ature gradients between the environment and the fluid. Applied Thermal Engineering. Vol 60 (2013) 152–163.
[13] N. ALTUNTOP, M. ARSLAN, V. OZCEYHAN, M KANOGLU, Effect of obstacles on
thermal stratification in hot water storage tanks, Applied Thermal Engineering.
5. Conclusion Vol. 25 (2005) 2285–2298.
[14] NETO MARCHI, I. Survey of domestic refrigerators coefficient of performance
The major purpose of this study is to analyze and compare the fluid associated a thermal storage unit. 2007. 132f. Thesis (MechanicalEngineering
Masters) – Faculdade de Engenharia, Universidade Estadual Paulista, Bauru, SP,
behavior of a CFD two-dimensional model of an experiment of a TES
Brazil, 2007.
tank coupled to a household refrigerator. The qualitative and quanti­ [15] MARCHI NETO I, A. PADILHA, V.L. SCALON, Refrigerator COP with thermal
tative results were considered satisfactory given the simulations two- storage, Applied Thermal Engineering. Vol. 29 (2009) 2358–2364.
dimensional nature, the thermal expansion coefficient constant [16] N.K. GHADDAR, Stratified storage tank influence on performance of solar water
heating system tested in Beirut, Renewable Energy. Vol 4 (1994) 911–925.
approach, the adiabatic walls assumption and the heat exchanger pre­ [17] N.K. GHADDAR, A.M AL-MARAFIE, Numerical simulation of stratification
scribed flux correction. This study shows that two-dimensional model­ behavior in thermal storage tanks, Applied Energy. Vol. 32 (1989) 225–239.
ling can be used in systems with similar conditions, providing [18] Y.M. HAN, R.Z. WANG, Y.J DAI, Thermal stratification within the water tank,
Renewable and Sustainable Energy Reviews. Vol. 13 (2009) 1014–1026.
considerable satisfactory results with low computational cost. Also, the [19] OLIVEIRA S. RIO, I. MARCHI, A. PADILHA, V.L SCALON, The energy and exergy
energy consumption of the refrigerator compressor slightly increased contents of stratified thermal energy storages, in: 19th International Congress of
with TES tank usage because of the refrigerator back-wall and heat Mechanical Engineering, Brasília, DF, Brazil, 2007. Nov/.
[20] L.A.C. ZUZARTE, J.R SIMÕES-MOREIRA, Recuperação do calor rejeitado em
exchanger proximity, resulting in heat entering the refrigerator thus condensadores de refrigeradores de pequeno porte para aquecimento de água,
increasing compressor consumption. Nevertheless, considering the en­ 2015, 10º Congresso sobre Geração Distribuída e Energia no Meio Rural.
ergy stored in sensible heat form, the COP of the complete system is [21] A.J.N. KHALIFA, A.T. MUSTAFA, F.A KHAMMAS, Experimental study of
temperature stratification in a thermal storage tank in the static mode for different
higher. The thermal expansion coefficient role on fluid stratification was aspect ratios, ARPN Journal of Engineering and Applied Sciences. Vol. 6 (2011)
considered significant, once the temperature data and profile along the 53–60.
vertical axis resulting from equally configured simulations were signif­ [22] H.I ABU-MULAWEH, Design and performance of a thermosiphon heat recovery
system, Applied Thermal Engineering. Vol 26 (2006) 471–477.
icantly different. Results showed that a smaller thermal expansion co­
[23] A. BARZEGAR, A.A DEHGHAN, Transient thermal behavior of a vertical solar
efficient leads to greater stratification. The smaller inlet velocity and storage tank with a mantle heat exchanger during no-flow operation, Journal of
fluid movement inside the TES tank can explain this stratification Applied Fluid Mechanics. Vol. 2 (2009) 55–69.
behavior difference, besides explaining differences in Richardson [24] D.L. SAVICKI, H.A. VIELMO, A KREZINGER, Three-dimensional analysis and
investigation of the thermal and hydrodynamic behaviors of cylindrical storage
Numbers for the simulations. The Richardson Number parameter tanks, Renewable Energy. Vol. 36 (2011) 1364–1373.
sensitivity to inlet velocity was enlightened. The Stratification Number [25] M. SHIN, H. KIM, D. JANG, S. LEE, Y. LEE, H YOON, Numerical and experimental
behavior with time showed that the prescribed temperature on the heat study on the design of a stratified thermal storage system, Applied Thermal
Engineering. Vol. 24 (2004) 17–27.
exchanger along with adiabatic boundary conditions in the walls results [26] G LI, Sensible heat thermal storage energy and exergy performance evaluations,
in unrealistic stratification behavior. The Stratification Number results Renewable and Sustainable Energy Reviews. Vol 53 (2016) 897–923.
of prescribed flux simulations were a better fit with the experiment re­ [27] P.J. ROACHE, Verification and Validation in Computational Science and
Engineering, Hermosa Publishers, Albuquerque, New Mexico, 1998.
sults. As in the experiment, the energy and exergy contents of Sim 2 and [28] B.K. SLIWINSKI, A.R. MECH, T.S SHIH, Stratification in thermal storage during
Sim 3 increased over time, although the data trend is significantly charging, in: Proceedings of the 6th International Heat Transfer Conference 4,
different. Numerically, the results were compatible with the experi­ Toronto, Canada, 1978, pp. 149–154.
[29] R.L. COLE, F.O BELLINGER, in: Thermally stratified tanks. Conference: ASHRAE
mental ones. The exergy analysis performed was useful to quantify the National Meeting, Toronto, Ontario, Canada, 1982.
stratification degradation of the simulations.

12
F.L.C. Gerhardt et al. Journal of Energy Storage 41 (2021) 102961

[30] Y.H. ZURIGAT, P.R. LICHE, A.J. GHAJAR, Influence of the inlet geometry on [39] J. FERNÁNDEZ-SEARA, F. UHIA, J SIERES, Experimental analysis of a domestic
mixing in thermocline thermal energy storage, International Journal of Heat and electric hot water storage tank. Part II: Dynamic mode of operation, Applied
Mass Transfer. Vol. 34 (1991) 115–125. Thermal Engineering. Vol 27 (2007) 137–144, b.
[31] A.J. GHAJAR, Y.H ZURIGAT, Numerical study of the effect of inlet geometry on [40] O. ABDELHAK, H. MHIRI, P BOURNOT, CFD analysis of thermal stratification in
stratification in thermal energy storage, Numerical Heat Transfer. Vol 19 (1991) domestic hot water storage tank during dynamic mode, Building Simulation. Vol. 8
65–83. (2015) 421–429.
[32] J.V. BERKEL, C.C.M. RINDT, STEENHOVEN A.A VAN, A Modelling of two-layer [41] T. BOUHAL, S. FERTAHI, Y. AGROUAZ, RHAFIKI T. EL, T. KOUSKSOU, A JAMIL,
stratified stores, Solar Energy. Vol. 67 (1999) 65–78. Numerical modeling and optimization of thermal stratification in solar hot water
[33] C.K. YEE, F.C LAI, Effects of a porous manifold on thermal stratification in a liquid storage tanks for domestic applications: CFD study, Solar Energy. Vol. 157 (2017)
storage tank, Solar Energy. Vol. 71 (2001) 241–254. 441–455.
[34] U. JORDAN, S FURBO, Thermal stratification in small solar domestic storage tanks [42] M.A. ROSEN, R. TANG, I DINCER, Effect of stratification on energy and exergy
caused by draw-offs, Solar Energy. Vol. 78 (2005) 291–300. capacities in thermal storage systems, International Journal of Energy Research.
[35] E. HAHNE, Y CHEN, Numerical study of flow and heat transfer characteristics in Vol 28 (2004) 177–193.
hot-water stores, Solar Energy. Vol. 64 (1998) 9–18. [43] Z. WANG, H. ZHANG, B. DOU, H. HUANG, W. WU, Z WANG, Experimental and
[36] E.M. KLEINBACK, W.A. BECKMAN, S.A KLEIN, Performance study of one- numerical research of thermal stratification with a novel inlet in a dynamic hot
dimensional models for stratified thermal storage tanks, Solar Energy. Vol. 50 water storage tank, Renewable Energy. Vol 111 (2017) 353–371.
(1993) 155–166. [44] Z. YANG, H. CHEN, L. WANG, Y. SHENG, Y WANG, Comparative study of the
[37] A. CASTELL, M. MEDRANO, C. SOLÉ, L.F CABEZA, Dimensionless numbers used to influences of different water tank shapes on thermal energy storage capacity and
characterize stratification in water tanks for discharging at low flow rates, thermal stratification, Renewable Energy. Vol 85 (2016) 31–44.
Renewable Energy. Vol. 35 (2010) 2192–2199. [45] M.A ROSEN, The exergy of stratified thermal energy storages, Solar Energy. Vol 71
[38] J. FERNÁNDEZ-SEARA, F. UHIA, J SIERES, Experimental analysis of a domestic (2001) 173–185.
electric hot water storage tank. Part I: Static mode of operation, Applied Thermal [46] A HASAN, Thermosyphon solar water heaters: effect of storage tank volume and
Engineering. Vol 27 (2007) 129–136, a. configuration on efficiency, Energy Conversion and Management. Vol 38 (1997)
847–854.

13