Solar Energy 174 (2018) 489–501

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

The error of neglecting natural convection in high temperature vertical T

shell-and-tube latent heat thermal energy storage systems
S. Saeed Mostafavi Tehrania,*, Gonzalo Diarceb, Robert A. Taylora
a
School of Mechanical and Manufacturing Engineering, University of New South Wales, Sydney, NSW 2052, Australia
b
ENEDI Research Group, Dpto. de Máquinas y Motores Térmicos, Escuela de Ingeniería de Bilbao, University of the Basque Country UPV/EHU, Rafael Moreno Pitxitxi 2,
Bilbao 48013, Spain

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

Keywords: There is little understanding of the relative importance of natural convection when designing latent heat thermal
High temperature energy storage (LHTES) systems based on geometric parameters and/or phase change material (PCM) properties.
Latent heat For high temperature shell-and-tube LHTES systems, this study aims: (i) to determine the error of ignoring
Phase change material (PCM) natural convection, and (ii) to quantify this error for different geometric parameters and PCM properties. In
Natural convection
particular, the study defines the circumstances under which natural convection is important and the error of
Error correlation
Shell-and-tube
choosing a ‘conduction-only modelling approach’. To do so, the performance of LHTES systems with nine
geometric aspect ratios and three commercial PCMs (of different melting points) were analyzed by means of a
validated CFD model.
The results showed that the error is a function of the process under analysis (melting or solidification) and the
ratio of stored/delivered energy divided by the maximum capacity of PCM (i.e. its effectiveness). Geometry also
plays a critical role in the relative importance of natural convection. The study demonstrates that a specific
system geometry (i.e. a dimensionless number defined based on the inner and outer radius as well as the length
R2 − ro2
of shell-and-tube geometry: S = 2ro L
) can be used to determine the relevance of natural convection. It was
found that regardless of PCM type, the error is of neglecting natural congestion is small if S < 0.005. For
r
S > 0.005, the error depends on the following non-dimensional groups: Lo , Ra, Ste, and Bi . As might be ex-
pected, the Rayleigh number was found to be the most influential group. Notably, a critical Rayleigh number
value (8 × 105) was found, below which the error of neglecting natural convection is < 1%. Finally, two cor-
relations were developed in order to quantify the error achieved – one for melting and another for solidification.

1. Introduction
Thermal energy storage (TES) is a key component in intermittent
energy conversion cycles like solar energy plants, where there is a
mismatch between supply and demand (Tehrani et al., 2013a, 2013b).
While sensible heat storage currently dominates the market for this type
of TES technology (Dincer and Rosen, 2002; Kuravi et al., 2013;
Seddegh et al., 2015b; Tehrani et al., 2017), latent heat thermal energy
storage (LHTES) systems have gained prominence in recent years as
they represent a promising alternative to traditional TES systems
(Cárdenas and León, 2013; Dutil et al., 2011). These systems use phase
change materials (PCMs), in a single or cascaded configuration (Tehrani
et al., 2018b), which store the latent heat of melting and release it upon
solidification. Compared to sensible heat storage, PCMs enable more
compact designs, which can result in lower storage media costs (Liu
et al., 2012; Zalba et al., 2003). Advanced high temperature systems are
currently under development to increase the efficiency of concentrated
solar power-tower (CSP-tower) plants, where the heat transfer fluid is
heated up to approximately 565 °C in order to produce electricity
(Mostafavi Tehrani et al., 2018; Gil et al., 2010; Jacob et al., 2016;
Kuravi et al., 2013; Liu et al., 2016; Tehrani and Taylor, 2016; Tehrani
et al., 2017).
Among the different configurations of LHTES systems, shell-and-
tube heat exchangers represent a promising and straightforward high
temperature PCM design (Seddegh et al., 2018). As a result, this con-
figuration is gaining interest (Agyenim et al., 2010; Nithyanandam and
Pitchumani, 2011; Tehrani et al., 2017). One of the most important
challenges facing this kind of system is the geometric design optimi-
zation (Li et al., 2017; Mahdavi et al., 2016; Nithyanandam and
Pitchumani, 2011, 2014; Tiari and Qiu, 2015) – task that is usually

*
Corresponding author.
E-mail addresses: s.mostafavi.tehrani@gmail.com (S.S. Mostafavi Tehrani), gonzalo.diarce@ehu.es (G. Diarce), robert.taylor@unsw.edu.au (R.A. Taylor).

https://doi.org/10.1016/j.solener.2018.09.048
Received 4 July 2017; Received in revised form 20 August 2018; Accepted 16 September 2018
Available online 21 September 2018
0038-092X/ © 2018 Elsevier Ltd. All rights reserved.
S.S. Mostafavi Tehrani et al. Solar Energy 174 (2018) 489–501

Nomenclatures Greek symbols

A area [m2] β volumetric expansion coefficient [1/K]

CP specific heat at constant pressure [J/kg K] ρ density [kg/m3]
d inner diameter of latent heat storage unit [m] μ dynamic viscosity [kg/m s]
f liquid fraction [–]
h enthalpy [J/kg] Subscripts
H , ΔH latent heat of fusion [J/kg]
k thermal conductivity [W/m K] l liquid
L length of tank [m] r radial direction
ṁ mass flow rate [kg/s] z axial direction
p pressure [Pa]
Q amount of stored/discharged energy [J] Abbreviations
Q̇ amount of stored/discharged energy [W]
R outer radius of a cylinder [m] ALF average liquid fraction
ro inner radius of pipe [m] HTF heat transfer fluid
t total time of simulation [s] LHTES latent heat thermal energy storage
T temperature [K] Nu Nusselt number
Tm1 lower melting point [K] PCM phase change material
Tm2 upper melting point [K] Ra Rayleigh number
→v velocity [m/s] Ste Stefan number
TES thermal energy storage

performed via modelling. However, the accurate simulation of two-
phase heat transfer problems is complex because of the moving
melting/solidification boundary (Fornarelli et al., 2016).
To date, several computational fluid dynamics (CFD) models have
been reported in the literature as for the modelling of phase change
processes (Riahi et al., 2017; Tehrani et al., 2016b). Conduction-only
models are much faster than full convection models, so they are often
used for design optimization. In fact, several recent studies only con-
sider the conduction mechanism (See Table 1). However, this approach
can lead to non-negligible errors depending on the conditions.
As demonstrated in a recent review on the topic (Dhaidan and
Khodadadi, 2015), upon melting, natural convection drives a
recirculation zone inside the liquid region. This phenomenon increases
the heat transfer within the liquid PCM and causes non-uniformities in
the solid-liquid interface and temperature distribution during melting.
Solidification, on the other hand, is mainly dominated by conduction,
although initially (at high liquid fractions) convection and conduction
transfer a similar amount of heat. Consequently, various researchers
have investigated the fundamentals of natural convection.
Fornarelli et al. (2016) compared the results obtained from con-
vective and pure conductive models of a high temperature shell-and-
tube LHTES system for a CSP application during the melting process.
The study confirmed that the convective motion increases the heat flux
to the PCM, effectively increasing the heat transfer rate (e.g. reducing

Table 1
Selected shell-and-tube LHTES system studies.
Ref. Tm (°C) Melting/Solidification Convection Considered? Method L (m) R/ ro L/d

(Wang et al., 2015) 26 Melt. Yes 2D Num. – 1.2–5 25–150

(Wang et al., 2013) 26 Melt./Sol. Yes 2D Num. 1 2 78
(Tao and He, 2011) 26 Melt. Yes 2D Num. 1 1.8 83
(Trp, 2005; Trp et al., 2006) 26 Sol. No Exp./2D Num. 1 3.7 30
(Lacroix, 1993) 26 Melt. Yes 2D Num. 1 1.6 78
(Adine and El Qarnia, 2009; El Qarnia, 2009) 26 Melt./Sol. Yes 2D Num. 1 1.6 78
(Longeon et al., 2013) 35 Melt./Sol. Yes Exp./CFD 0.4 2.9 26
(Kalhori and Ramadhyani, 1985; Kemink and Sparrow, 1981) 36 Melt. Yes 2D Num. 0.2 8.4 10
(Mendes and Brasil, 1987) 36 Melt. Yes Exp. 0.17 6 5
(Esen et al., 1998) 29–46 Melt. No 2D Num. 3.2 1.09–1.2 20–48
(Fang and Chen, 2007) 20–80 Melt. No 2D Num. 2 2 100
(Seddegh et al., 2016) 51 Melt./Sol. Yes CFD 1 3.9 12
(Seddegh et al., 2015a) 58 Melt./Sol. Yes CFD 1 4 30
(Xiao and Zhang, 2015a, 2015b) 60 Melt./Sol. Yes Exp./2D Num./CFD 0.75 – 10.7
(Akgün et al., 2007) 75 Melt./Sol. Yes Exp. 0.5 3.3 17
(Wang et al., 2016) 121 Melt./Sol. Yes Exp. 0.9 2.4 20
(Fan et al., 2014) 169 Melt./Sol. No 2D Num. 1.4 1.25 87
(Pointner et al., 2016) 215 Melt. Yes 2D Num. 1 – –
(Fornarelli et al., 2016) 230 Melt./Sol. Yes CFD 0.5 5 35
(Tehrani et al., 2016b) 300–500 Melt./Sol. No 2D Num. 1–5 1.3–3 10–100
Present study 300–500 Melt./Sol. Yes CFD 0.5–2 1.5–3 15–60
(Riahi et al., 2017) 306 Melt./Sol. Yes CFD 0.6 3.15 29
(Muhammad et al., 2015a) 306 Melt./Sol. Yes CFD 0.9 5.41 155
(Tao and Carey, 2016) 400–600 Melt. No 2D Num. 1 2 40
(Li et al., 2013) 400–700 Melt. No 2D Num. 1.2 2 120
(Pirasaci and Goswami, 2016) 550 Sol. No 2D Num. 10–150 1–3 10–150
(Bellecci and Conti, 1993a, b) 734 Melt. No 2D Num. – 2.6–5 12–150
(Tao et al., 2014; Tao et al., 2012) 766 Melt. No 2D Num. 1.5 2 60

490
S.S. Mostafavi Tehrani et al.
Fig. 1. A (vertical) single unit cylindrical shell-and-tube LHTES system.
the charging time). Longeon et al. (2013) investigated the effect of top
and bottom HTF injection with respect to natural convection in a shell-
and-tube LHTES system. For both injection options, the PCM always
melted first at the top of the exchanger throughout its whole thickness
and natural convection transported the heat along the inner tube of the
exchanger. The natural convection effect, however, was less significant
during solidification as compared to melting. Wang et al. (2016) per-
formed an experimental study followed by CFD simulations for a
medium temperature shell-and-tube LHTES system. They demonstrated
that as the melting process progresses, the molten PCM first occupies
the top region and then spreads from top to bottom. Moreover, natural
convection was more pronounced as the volume and temperature of the
molten PCM region increased. The authors also explained that during
solidification, the PCM starts to solidify from the bottom region at the
beginning, and then the solid-liquid interface uniformly progresses
from the inner tube to the outer shell. These results confirm the initial
dominance of natural convection at high liquid fractions (Wang et al.,
2016).
Despite the number of papers that have considered natural con-
vection, there are still several that have neglected it, as demonstrated in
Table 1. Trp et al. (2006) compared a 2D-numerical model with ex-
perimental results and concluded that heat conduction controls the heat
transfer inside the PCM region in both melting and solidification. Apart
from this study, which appropriately justified neglecting convection,
most studies provide no justification when invoking this assumption
(Esen and Ayhan, 1996; Esen et al., 1998; Fan et al., 2014; Fang and
Chen, 2007; Li et al., 2013; Pirasaci and Goswami, 2016; Tao et al.,
2012, 2014; Tao and Carey, 2016; Tehrani et al., 2016b). Alternatively,
some numerical studies do consider natural convection through the
effective thermal conductivity approach (Adine and El Qarnia, 2009; El
Qarnia, 2009; Pointner et al., 2016; Tao and He, 2011; Wang et al.,
2013, 2015). Again, in most cases, no justification was provided for this
approach. Neglecting convection simply for the sake of simplification –
while tempting – can be a major source of error.
From the literature, it can be concluded that there is a clear lack of
direction on when convection can (or cannot) be neglected and as to
how much of a role it plays in a generalized shell-and-tube LHTES
system. The studies dealing with natural convection have been mostly
limited to a specific geometry or PCM (See Table 1). However, the re-
lative importance of natural convection is clearly affected by geometry
and the PCM properties (Agyenim et al., 2009; Ismail and Melo, 1998;
Longeon et al., 2013; Ng et al., 1998; Patankar and Ftamadhyani, 1977;
Wang et al., 2012, 2016) and no study published to date has evaluated
491
Solar Energy 174 (2018) 489–501
these factors comprehensively. In addition, the actual error of ignoring
convection has not yet been quantified. The present study addresses the
aforementioned aspects by comparing conduction-only and convection
models in a high temperature shell-and-tube LHTES system during both
melting and solidification. A full range of geometric parameters
R L
(0.5 ≤ L ≤ 2, 1.5 ≤ r ≤ 3, 15 ≤ d ≤ 60 ) and three high temperature
o
commercial PCMs (http://www.pcmproducts.net/, 2018): H325, H425,
H525 were considered. This enables a geometric and property depen-
dent quantification of the error in the form of some simple correlation
equations. These equations will help researchers decide when to neglect
natural convection for any high temperature shell-and-tube LHTES
system and/or to estimate the error of making this assumption. Fol-
lowing this work, Tehrani et al. (2018a) developed a general natural
convection Nusselt correlation to be used via effective thermal con-
ductivity method in conduction numerical models for the rapid mod-
elling of shell-and-tube LHTES system, which can minimize the error of
neglecting natural convection.
2. The high temperature shell-and-tube LHTES system
2.1. System description, operational parameters and materials
The generic system analyzed in this study is based on a LHTES
system previously proposed by the authors in (Tehrani et al., 2016b). It
consists of concentric tube(s) whereby the HTF and PCM are segre-
gated, as shown in Fig. 1.
The HTF flows through the inner tube and exchanges heat with the
PCM in the surrounding region (i.e. between the inner and outer tube
shown in Fig. 1). During the charging process (melting), the hot HTF
flows from top to bottom. During the discharging (solidification) pro-
cess, the flow direction is bottom up. The top of the tank is denoted as
Z/L = 1 and the bottom of the tank is referred to as Z/L = 0. The op-
erational conditions of the shell-and-tube LHTES system were adopted
from the design point operational conditions of CSP-tower plants as
tabulated in Table 2 (Tehrani and Taylor, 2016; Tehrani et al., 2016a).
The flow rate was also adopted from a previous work of the authors,
and represents the typical discharging flow rate inside each tube of an
optimized shell-and-tube LHTES system in a CSP-tower plant (Tehrani
et al., 2016b).
In order to perform the study under realistic conditions, and to
cover the temperature gap required for CSP-tower plants (i.e.,
286–565 °C), three commercially available PCMs based on eutectic
mixtures of carbonate salts were selected: H325, H425 and H525
(http://www.pcmproducts.net/, 2018). The commercial Solar Salt, a
eutectic nitrate salt mixture, was selected as the HTF (Tehrani and
Taylor, 2016; Tehrani et al., 2016b). The thermophysical properties of
all materials used in this study are given in Table 3.
2.2. Case studies: geometric parameters employed
As demonstrated in a previous publication of the authors (Tehrani
et al., 2016b), the maximum potential of a shell-and-tube LHTES system
is only achieved if the geometric parameters are selected appropriately.
Table 2
Operating conditions, adopted from (Tehrani et al., 2016a, 2016b, 2017).
Name Unit Value
HTF inlet temperature in charging (top °C 565
injection)
HTF inlet temperature in discharging °C 286
(bottom injection)
Mass flow rate (charging/discharging) kg/s 0.0025
Flow regime on HTF side – Laminar (Reynolds < 2200)
Nusselt number at HTF side (HTF/PCM – 3.66
interface)
S.S. Mostafavi Tehrani et al. Solar Energy 174 (2018) 489–501

Table 3 for the interface between the HTF and PCM continuum zones.
Thermophysical properties of the HTF and PCMs. • Adiabatic walls were used for the remaining external surfaces of the
HTF and PCMs properties HTF (Bauer PCMs (http:// PCM sub-domain.
Units et al., 2012) www.pcmproducts.net/, 2018)
The continuity equation for incompressible flows was employed
Name Solar Salt H325 H425 H525 within the domain [21]:
Dynamic viscosity kg/m s 0.00326 0.0035 0.0035 0.0035
∇→
v =0 (1)
Melting point °C 325 425 525
Density kg/m3 1820 2110 2100 2350 The enthalpy-porosity method was used to model the melting and
Specific heat J/kg K 1553 1505 1535 1560 solidification processes of the PCM (Muhammad et al., 2015a, 2015b;
capacity
Thermal W/m K 0.52 0.549 0.565 0.565
Voller et al., 1987). The term H is calculated by the sum of sensible and
conductivity latent enthalpies, as shown in Eq. (3). Eqs. (1)–(4) are combined to
Latent heat of J/kg – 80,000 220,000 155,000 obtain Eq. (5).
fusion
∂ (ρH )
Lower melting °C – 323 423 523 + ∇ ·(ρ→
v H ) = ∇ . (k∇T )
point (Solidus) ∂t (2)
Upper melting °C – 328 428 528
point H = h + f ΔH (3)
(Liquidus)
Thermal expansion 1/K – 0.0005 0.0005 0.0005 T
coefficient
h = href + ∫T ref
CP dT
(4)

f=0 if T < Tm1

Based on the findings of that study and on the values covered by the T − Tm1
if Tm1 ⩽ T ⩽ Tm2
f= Tm2 − Tm1
reviewed literature (Table 2), the following ranges of geometric para-
R L f=1 if T > Tm2
meters were herein employed: 0.5 ≤ L ≤ 2, 1.5 ≤ r ≤ 3, 15 ≤ d ≤ 60 . (5)
o
VHTF
Besides, the minimum adopted value of porosity (ε = is typically ) ∂ (ρh) ∂ρf ΔH
+ ∇ . (ρ→ −∇ . (ρ→
Vtotal
0.2 (Flueckiger et al., 2014; Showalter and Kolb, 2002), so that was v h) = ∇. (k∇T)− v f ΔH )
∂t ∂t (6)
adopted herein. Each combination of the geometric parameters corre-
sponds to a certain Case ID in Table 4. The outcomes of this study will The general form of the momentum equation employed is shown in
be therefore extracted from 54 simulations (nine case studies for each of Eq. (7) [21]. Where τ̄¯ represents the viscous stress tensor, g is the ac-
the 3 PCMs for both melting and solidification). celeration due to gravity (with a value of −9.81 m/s2, according to the
coordinates of Fig. 1), and S′ is a source term defined below in Eq. (8):
3. Modelling approach ∂ (ρ→v)
+ ∇ (ρ→→
v v ) = −∇p + ∇ (¯τ¯) + ρg + S′
∂t (7)
The influence of natural convection was elucidated by comparing
the results of a pure conduction model with an approach that includes (1−f )2
S '= vAmush
both conduction and convection within the PCM (referred, respectively, f3 + ∊ (8)
to as conduction and convection models hereafter). As such, for each 5
case study (Table 4), two simulations were run, one for each approach. Default values of 0.001 and 10 were adopted for the terms ∊and
The details of each of the models employed are outlined below. Amush , respectively (Fluent, 2011). The viscous stress tensor (τ̄¯ ) present
in Eq. (7) was re-arranged to be adapted to incompressible fluids, be-
3.1. Mathematical domain, governing equations and boundary conditions coming μ∇2 v . Consequently, the final version of the momentum equa-
tion employed is shown Eq. (9), where β represents the volumetric
A commercial CFD software package, ANSYS Fluent 15.0, was em- expansion coefficient of the PCM.
ployed to simulate the LHTES system under both modelling approaches, ∂→v
using the Leonardi cluster at the University of New South Wales ρref ⎛⎜ +→
v ∇→
v ⎞⎟ = −∇p + ∇ (μ∇2 v ) + ρref g−ρref β (T −Tref ) g
⎝ ∂t ⎠
(UNSW) (Fluent, 2011). For the two cases, transient simulations were
carried out by solving the continuity and the Navier–Stokes equations (1−f )2
+ 3 vAmush
for the PCM and the HTF. An overview of the modelled domain with the f +∊ (9)
boundary conditions applied is shown in Fig. 2. According to the The main difference between the convection and the conduction
symmetric nature of the shell and tube geometry, a 2D axisymmetric approaches comes from including the buoyancy effects within the PCM.
CFD approach was adopted (Tao and He, 2015). Two continuum fluid
zones were defined: the PCM and the HTF. The HTF and the PCM (in its Table 4
liquid state) always stayed within the laminar regime, based on the Case ID and geometric specifications.
maximum Rayleigh number achieved for the geometries and conditions
Case ID L a (m) R/ ro L/d L/ R−ro R2 − ro2 ε
involved – i.e. lower than 109 for all cases, which is within the laminar S=
2ro L
regime (Ghajar, 2011). As illustrated in Fig. 2, the boundary conditions
for the melting process were defined as follows (note that for the soli- 1 2 3 60 60 0.033 0.11
dification process, the inlet/outlet boundary conditions and the flow 2 2 2 60 120 0.013 0.24
3 2 1.5 60 240 0.005 0.44
direction were reversed):
4 1 3 30 30 0.067 0.11
5 1 2 30 60 0.025 0.24
• The inlet mass flow rate and temperature were specified at the top 6
7
1
0.5
1.5
3
30
15
120
15
0.010
0.133
0.44
0.11
section of the inner tube, and the pressure outlet was specified at the
8 0.5 2 15 30 0.050 0.24
bottom section of the inner tube.
•
9 0.5 1.5 15 60 0.021 0.44
The geometry is axisymmetric around the centreline.
• The wall was thermally coupled (termed as a shadow wall in Fluent) a
For all cases: Re < 2200, ro = 0.0166 m , NP = 1.

492
S.S. Mostafavi Tehrani et al. Solar Energy 174 (2018) 489–501

Fig. 2. CFD domain and boundary conditions during melting.

For the convection approach, the well-known Boussinesq approxima-
tion was employed to compute natural convection within the PCM (Al-
Abidi et al., 2013; Brent et al., 1988; Seddegh et al., 2015a). Con-
versely, for the pure conduction approach, these buoyancy effects were
not included in the model.
associated with the medium mesh (90,000 cells) had less than ∼1%
difference with the finest mesh, confirming the reliability of mesh se-
lected (i.e., with all PCMs). Note that the melting time is sensitive to
geometry (Case ID) and the thermophysical properties of the PCM
employed.

3.2. Numerical schemes, convergence criteria, and other considerations

3.4. Model validation
A pressure-based double-precision solver was selected to solve the
To the best of the authors’ knowledge, no experimental data are
set of equations with a PISO pressure–velocity coupling scheme.
available in the literature for the operational temperatures involved
Second-order upwind discretization schemes were applied on the mo-
(See Table 1). Therefore, a system operating at lower temperatures was
mentum and energy equations, while a PRESTO scheme was employed
used for validation purposes. Simulated results were compared against
for the pressure equation because second order schemes are in-
the experimental work performed by Longeon et al. (2013).
compatible with the enthalpy-porosity approach in ANSYS Fluent
In that work, the LHTES system consisted of two concentric cylin-
(Seddegh et al., 2015a, 2016).
ders, where the outer cylinder had a diameter of 44 mm and the inner
The numerical convergence of the model was checked based on the
diameter was 15 mm. The length of the whole system was 400 mm. The
scaled numerical residuals of all computed variables. A value of 10−6
HTF selected was water, which flowed through the inner tube. The PCM
was chosen for the continuity residual. This value assured that the mass
consisted of technical grade paraffin (RT35 by Rubitherm (Longeon
imbalance was less than 0.2% of the net flux through the domain. The
et al., 2013)) and it was placed within the outer tube. Its phase change
velocity residuals were set to 10−6 and the energy residual value was
temperature ranges from 30 °C to 36 °C and the mass employed was
10−9. Temperature and velocity magnitudes were monitored each
0.480 kg. According to the Reynolds number obtained from the flow
iteration at several points of the mesh to check convergence as well.
rate of the HTF inside the tube, 0.01 m/s, the flow regime was laminar
An adaptive time stepping method was used with a minimum time
(Ghajar, 2011). The charging of the system was simulated by injecting
step size of 10−4 s and a maximum of 1 s. This configuration minimized
the HTF at 53 °C (by the top) to a system that had initial PCM and HTF
the error attained, especially at the beginning of the process, when
temperatures of 22 °C.
larger time steps would not be able to capture the high heat flux values
The comparison was based on the experimental temperatures re-
due to the large initial temperature gradients inside the system. A first-
gistered in (Longeon et al., 2013) for thermocouples located in different
order implicit transient formulation scheme was employed because
radial positions – points a, b and c, placed 3, 6 and 9 mm away from the
second-order schemes are incompatible with adaptive time stepping.
inner tube, respectively. Their axial position was Z/L = 0.5 (the middle
of the tube). The comparison between the experiment and the simula-
3.3. Meshes tion is shown in Fig. 4, which has suitable agreement. The model was
able to successfully predict the time when the melting front reached the
Different mapped grid meshes with orthogonal quadrilateral cells axial position of the thermocouples. That moment can be identified in
were employed depending on the geometry under evaluation. In the Fig. 4 as the time where the PCM temperatures increase abruptly (at
radial direction, every mesh had 40 cells in the PCM zone and 20 cells ∼1 h of charging time). Thus, it was concluded that our CFD model can
in the HTF zone. The divisions in the axial direction varied according to be employed confidently for the purposes of this study.
the length of each geometry: 1500, 750 and 375 divisions for those
lengths with 2, 1 and 0.5 m, respectively. According to Table 4, Case ID
1 is the largest unit that has been considered in this study, while the
Case ID 9 is the smallest module. Thus, the number of elements in the
meshes employed were 22,500 cells (Case IDs: 7–9), 45,000 (Case IDs:
4–6) and 90,000 (Case IDs: 1–3). Each mesh was refined close to the
boundary walls where larger temperature and velocity gradients are
expected.
The numerical uncertainty of the grid was evaluated by the Grid
Convergence Method (GCI) (Diarce et al., 2014) for all 3 mesh sizes. As
an illustration, a detailed view of the mesh employed for the largest
geometry is shown in Fig. 3 (90,000 cells). For the GCI analysis, three
more meshes were built for Case ID 1, having 126,000, 90,000 and
65,000 elements. The melting time was chosen as the key variable to
evaluate the error. The results showed that the numerical error (GCI) Fig. 3. Mesh employed for the outlet of the CFD domain (Case ID 1).

493
S.S. Mostafavi Tehrani et al. Solar Energy 174 (2018) 489–501

melting are depicted in Figs. 7 and 8, respectively. One significant

feature of Fig. 7 is that the solid/liquid interface has a ‘top-hat’ shape in
the presence of natural convection, whereas it has a parabola shape for
the conduction modelling approach. In the presence of natural con-
vection, the PCM melts mostly from top to the bottom. However, with
the conduction model, the PCM melts mostly in the axial direction,
from the hot inner surface (i.e., PCM/HTF interface) to the outer cold
surface. The conduction model does not show perfect axial melting
because of the non-uniform heat transfer rates caused by the initial
temperature gradient of the HTF as it is replaced by the hot fluid in-
troduced at the pipe inlet. It should be noted that these results are
consistent with the literature for HTF top injection in melting (Longeon
et al., 2013; Seddegh et al., 2015a).
Fig. 4. Validation of the CFD model against experimental data provided in
(Longeon et al., 2013). When comparing Figs. 7 and 8 it can be seen that the influence of
natural convection in the H325 system is more significant than the
H525 system. This is more easily noticed in the less radially shaped
4. Results and discussion: conduction vs. convection solid/liquid interface of H525 compared to the H325 system in the
convection model. This can be explained by the fact that the H525
All cases from Table 4 were simulated under the boundary condi- system has a very high melting point, allowing conduction to dominate
tions outlined in Section 2.1 for the three PCMs of Table 3. since melting takes much longer. Indeed, the melting of the system with
H325 is characterized by high heat fluxes since the PCM melts earlier.
4.1. Temperature profiles in melting/solidification According to Fig. 9, the heat transfer rate to the PCM has an initial peak
that decays over time (with much of the initial increase corresponding
This section investigates the behavior of the temperature profiles to the replacement of the initial cold HTF in the pipe). When the PCMs
under the two (conduction and convection) modelling approaches. Case start melting (after ∼1 and ∼5 h for the H325 and H525 systems, re-
IDs 1 and 3 as well as PCMs H525 and H325 were selected as re- spectively), the heat transfer rates to the PCM (i.e., the driving force of
presentative examples. Fig. 5 shows their timewise temperature dis- natural convection) is much higher for the system with H325 compared
tributions upon melting, while Fig. 6 depicts the evolution of the to that of H525. Thus, natural convection circulation is much more
temperatures during solidification. The following conclusions can be significant for low melting point PCMs. As such, for relatively high
drawn: melting point PCMs (i.e., when the melting point is close to the
The slope of the temperature profiles in the liquid phase is larger
compared to the solid phase (i.e., convection compared to conduction),
indicating that natural convection enhances the heat transfer rate in the
liquid phase. Thus, the melting process shows more discrepancy be-
tween the temperature profiles of conduction and convection modelling
approaches than the solidification process.
The influence of natural convection is a function of the geometric
parameters. The figures suggest that natural convection effect in Case
ID 1 is more significant than Case ID 3 (i.e., the deviation of the con-
vection profiles from the conduction profiles is lower in Case ID 3). This
is because of the larger non-dimensional radius of Case ID 1 (R/ ro=3, 3),
compared to Case ID 3 (R/ ro=1.5).
The influence of natural convection is not always identical at the top
and the bottom of the tank. In Fig. 5a, for instance, the difference be-
tween the conduction and the convection models is more significant at
the top of the tank, where the PCM melts much faster. Therefore, the
influence of natural convection cannot be evaluated at just one specific
location inside the PCM tank. Similarly, the influence of natural con-
vection must be monitored over the whole process, as the deviation
between the conduction and convection model changes from the be-
ginning to the end of the process.
The results also suggest that for a given geometry (e.g., Case ID 1),
the thermo-physical properties of the PCM significantly affect the pre-
sence of the natural convection. For example, the discrepancies be-
tween conduction and convection models are much higher for H325 as
compared to that of H525 (due to their 200 °C difference in melting
point), in both melting and solidification. The reasons for this are dis-
cussed in the following section.

4.2. Solid/liquid interface and heat transfer rates on melting

To visually demonstrate the differences between the conduction and

convection modelling approaches and to realize how the presence of
natural convection can be influenced by PCM selection, the behavior of
H325 and H525 were compared in detail for Case ID 1. The timewise Fig. 5. Melting temperature histories of the PCM at different axial locations for:
contours of the average liquid fraction (ALF) for H325 and H525 upon (a) Case ID 1-H525, (b) Case ID 3-H525, (c) Case ID 1-H325.

494
S.S. Mostafavi Tehrani et al.
Fig. 6. Solidification temperature histories of the PCM at different axial loca-
tions for: (a) Case ID 1-H525, (b) Case ID 3-H525, (c) Case ID 1-H325.
maximum operating temperature of the system), there is little deviation
between the conduction and convection models.
While the temperature evolutions and solid/liquid interfaces dis-
cussed so far give a qualitative representation of the ‘error’ in ne-
glecting natural convection, they do not provide a quantitative mea-
sure. In the next subsection, a method to quantify the error between
choosing conduction versus convection models is proposed.
Fig. 7. Solid/liquid interface during melting of H325 (Case ID 1) for conduction (left) and convection (right).
495
Solar Energy 174 (2018) 489–501
4.3. Effectiveness and error
In order to quantify the error between the conduction and convec-
tion models, a performance criterion is required. The authors suggest
that the storage effectiveness is the best performance metric for LHTES
systems as it reflects the real utilization of the PCM and covers the
whole range of sensible and latent phases of PCM. Effectiveness is de-
fined by the ratio of stored/delivered energy compared to the maximum
potential of the PCM, as shown in Eq. (10). Note that the transient value
of stored/delivered heat by PCM (QPCM ̇ ) was calculated from the CFD
simulations (i.e., the heat flux values of Fig. 9), while the maximum
potential of PCM (QPCM , max ) was case study dependent as described by
Eq. (11) and Table 5.
t
∫0 Q̇PCM (t ) dt
ε (t ) =
QPCM , max (10)
where
QPCM , max = mPCM CPCM , S (Tm1−Tinitial ) + mPCM ΔH
+ mPCM CPCM , L (Tinlet −Tm2) (11)
As can be seen in Table 5, Case ID 1 has the highest QPCM , max (due to
having the largest PCM volume), while Case ID 9 has the lowest
QPCM , max (due to having the minimum PCM volume). The amount of
PCM will clearly affect the potential for natural convection to play an
important role, but the volume/capacity of PCM is not the correct cri-
terion to assess the influence of natural convection. As will be shown in
this section, Case ID 7 most dramatically shows the presence of natural
convection.
The heat transfer rate and instantaneous effectiveness of H325 (Case
ID 1) are depicted in Fig. 10 for melting and in Fig. 11 for solidification.
In both conduction and convection models, the heat transfer rate is
maximum at the very beginning of the process because of the maximum
temperature difference between the inner and outer wall of the annulus.
After this point, the heat transfer rate is decreasing over time because
the temperature difference between the inner and outer wall of annulus
decreases until they reach equilibrium, which is the end of charging/
discharging process.
Another feature of Figs. 10 and 11 is that the heat transfer rate in
the convection model is not always higher than that of conduction
model for both melting and solidification. This occurs because con-
vection allows the system to reach equilibrium faster. Thus, the effec-
tiveness for the convection model is always higher than that of the
conduction model at any given time. The time difference between the
two modelling approaches is a function of the effectiveness ratio, so the
error between two approaches can defined as follows:
S.S. Mostafavi Tehrani et al. Solar Energy 174 (2018) 489–501

Fig. 8. Solid/liquid interface during melting of H525 (Case ID 1) for conduction (left) and convection (right).

to reach full capacity (∼100% effectiveness) as often as possible.

However, in real operation the system may frequently undergo partial
charging and discharging (ε ≪ 100%) (Tehrani et al., 2017). Since the
objective in this study is to quantify the error of neglecting convection,
a value of 98% effectiveness was considered, as it provides a realistic
maximum error. According to Table 6, at effectiveness values lower
than 98%, the error decreases for the melting process, while it increases
for the solidification process. Note that the maximum error in melting is
always higher than that of solidification so that the priority was given
to melting. In the next subsection, general error correlations will be
presented to predict the error at 98% effectiveness.

Fig. 9. Heat flux to the PCM on melting for the convection modelling approach; 4.4. Error correlations at 98% effectiveness
H325 vs. H525.
So far, it has been demonstrated that the relative error between
Table 5 conduction and convection models is sensitive to: (1) geometry, (2)
Maximum stored/delivered energy. PCM, and (3) effectiveness ratio. Given the fact that the 98% effec-
Parameter QPCM , max [MJ ]
tiveness represents the maximum error that can occur in the simula-
tions (during melting), this section aims to incorporate these para-
ID case study PCM 1: H 325 PCM 2: H 425 PCM 3: H 525 meters to develop error correlations for researchers who wish to: (a)
decide whether to consider convection or not and/or (b) determine the
1 14.8 19 19.3
maximum error by neglecting convection.
2 5.5 7.1 7.2
3 2.3 2.9 3
4 7.3 9.5 9.7 4.4.1. Error as a function of geometry
5 2.7 3.5 3.6
6 1.1 1.4 1.5
Table 7 summarizes the error values presented in the different Case
7 3.6 4.7 4.8 IDs during melting and solidification. The maximum error observed was
8 1.3 1.7 1.8 100%, which corresponded to the melting of H325 in Case ID 7.
9 0.5 0.7 0.75 A regression analysis of the error values as a function of various
R−r L r
shell-and-tube LHTES geometric parameters (i.e., L o , R − r , Lo ) was
o

tconvection−tconduction
Error (%) = × 100, where t = time to reach #%
tconvection
of effectiveness
In Table 6, the error between the conduction and the convection
modelling approaches is summarized for various values of effectiveness.
It is evident that for the melting process, the maximum error occurs at
higher values of the effectiveness, while the maximum error in dis-
charging occurs at lower values of effectiveness. This result can be
explained by considering that during melting natural convection effect
increases as time progresses (i.e., the error accumulates from when
PCM first melts until the end of process), while during solidification it
has more influence at the beginning, when all the PCM is in the liquid
state.
Without selecting an appropriate effectiveness value, it is not pos-
sible to determine the actual error between conduction and convection
modelling approaches. Commonly, LHTES systems should be designed
performed. The results, which will be herein omitted for the sake of
R2 − r 2
brevity, indicated that the specific geometry, S = 2r L o , might be a way
o
to lump together the effect of all geometric parameters (see Fig. 12a and
(12)
b), since as this parameter increases, the error increases. If the specific
geometry is small enough (S < 0.005), natural convection can be safely
neglected (e.g. with an error of < 5%), for the PCM types and boundary
conditions tested in this study.
In Fig. 12, a simple linear correlation is presented that is suitable for
each specified PCM only and is a function of geometry (i.e., the un-
certainty of correlations is within 30% as shown by the error bars on the
graphs). It can also be observed that the error between the conduction
and the convection modelling approaches is lower when a PCM with a
higher relative melting point (i.e., close to the maximum operating
temperature) is implemented.
4.4.2. Error as a function of geometry and PCM properties
Despite the monotonic effect of geometric parameters on error

496
S.S. Mostafavi Tehrani et al. Solar Energy 174 (2018) 489–501

Fig. 10. Comparison of conduction and convection model simulations during melting of H325 in terms of (a) heat transfer rate and (b) effectiveness (ID case 1).

Fig. 11. Comparison of conduction and convection model simulations during solidification of H325 in terms of (a) heat transfer rate and (b) effectiveness (Case ID 1).

Table 6
Time needed to attain specific effectiveness values for melting and solidification (Case ID 1, PCM H325).
Melting Solidification

Effectiveness (%) Required time of conduction Required time of convection Error (%) Required time of conduction Required time of convection Error (%)
model (h) model (h) model (h) model (h)

98 13.11a 8.71 51 16.18 15.18 7

95 10.45 7.19 45 14.27 12.62 13
85 7.08 5.25 35 9.95 8.11 23
65 4.29 3.58 20 5.16 4.08 26
40 2.3 2.10 11 2.5 2.08 18

a
This is a very large melting time for most practical applications, but the aim in this study is to cover a wide range.

(Fig. 12), the value of error is different for the different PCMs. To in- hHTF Lc π (R2−ro2 ) L R2−ro2
Bi = , where Lc = =
vestigate the influence of PCM properties as well as geometric para- kPCM 2πro L 2ro (14)
meters on the error, with the results of Table 7, several sets of non-
dimensional groups that normally plays a role in the melting/solidifi- kHTF Nu R2−ro2
Bi = , where Nu = 3.66 in laminar flow
cation problems were analyzed through a multi-variant regression kPCM 4ro2 (15)
analysis in Excel as described by Table 8. The non-dimensional numbers
hHTF Lc, Nu
were defined as follows: Nu = , where Lc, Nu = 2ro
kHTF (16)
gβ (Tr = ro−Tm) (R−ro)3
Ra = Gr × Pr = , where Tr = ro = Thot = 565 ∘C CP, PCM (Thot −Tm)
υα Ste = , where Thot = 565 ∘C
ΔH (17)
(13)
It is worth noting that the Ra number was calculated when the PCMs
are in their liquid phase – i.e., at the end of the melting process and the

497
S.S. Mostafavi Tehrani et al. Solar Energy 174 (2018) 489–501

Table 7
Error at 98% effectiveness for various Case IDs and PCMs.
Case ID Melting Solidification

Error (%)

H325 H425 H525 H325 H425 H525

1 51 22 7 7 7 1
2 23 9 2 10 7 1
3 8 4 0 4 1 0
4 73 36 13 22 8 2
5 35 16 4 12 5 1
6 13 5 0 5 2 0
7 100 53 21 28 10 3
8 48 24 7 16 6 1
9 19 9 1 7 3 1

Fig. 13. Accuracy of global error correlations versus the error calculated by
simulations: (a) melting, (b) solidification.

Table 9
Range of Stefan and ro/ L numbers used in this study.
ID case study ro/ L H325 H 425 H525

1, 2, 3 0.008 Ste = 4.47 Ste = 0.97 Ste = 0.4

4, 5, 6 0.01
7, 8, 9 0.03

Table 10
Range of Biot and Rayleigh numbers used in this study.
H325 H 425 H525

Fig. 12. Error at 98% effectiveness as a function of geometry for each particular ID case R/ ro Bi Ra Bi Ra Bi Ra
study
PCM: (a) melting, (b) solidification.
1, 4, 7 3 6.93 1.51 6.91 8.94 6.73 3.16 E + 07
beginning of the solidification process (when the error is maximum). E + 08 E + 07
2, 5, 8 2 2.59 1.88 2.58 1.12 2.52 3.95 E + 06
Based on the natural convection literature for the vertical annular space
E + 07 E + 07
(Yunus A. Cengel, 2015; Prasad and Kulacki, 1984; El-Shaarawi and Al- 3, 6, 9 1.5 1.08 2.35 1.07 1.40 1.05 4.94 E + 05
Nimr, 1990; Al-Arabi et al., 1987; Abouali and Falahatpisheh, 2009; E + 06 E + 06
Afrand et al., 2015; Sankar et al., 2011; Keyhani et al., 1983), the
characteristic length should be the distance between the cold and hot

Table 8
Results of multi-variant regression analysis on different non-dimensional groups for the melting process.
Group Non-dimensional numbers Multiple Ra R Squaredb Adjusted R Squaredc Averaged

1 Ra, ro/ L, Bi, Ste 0.97 0.89 0.87 0.91

2 Ra, ro/ L 0.84 0.71 0.68 0.74
3 Ra, ro/ L, Ste 0.92 0.85 0.83 0.86
4 Ra, S 0.80 0.64 0.60 0.68
5 Ra, S, Ste 0.91 0.83 0.80 0.85
6 Ra, R−ro/ L 0.80 0.65 0.62 0.69
7 Ra, R−ro/ L, Ste 0.91 0.83 0.81 0.85
8 Ra, ro/ L, kHTF / kHTF 0.92 0.86 0.84 0.87
9 Ra, ro/ L, Ste , R2−ro2/4ro2 0.94 0.88 0.86 0.89
10 Ra, ro/ L, R/ ro 0.94 0.89 0.87 0.90
11 Ra, ro/ L, R/ ro, Ste 0.94 0.89 0.87 0.90

a
Square root of R-squared.
b
R-squared is a statistical measure of how close the data are to the fitted regression line.
c
The adjusted R-squared is a modified version of R-squared for the number of predictors in a model.
d
Average of Multiple R, R squared, and Adjusted R Squared.

498
S.S. Mostafavi Tehrani et al.
surfaces for the enclosure, so L = R−ro was selected here. It should be
noted that if tube length of L was used in Rayleigh number definition
instead of the annular space gap of R−ro , the range of Rayleigh number
could vary from 105 < Ra < 108 to 1011 < Ra < 1013 for the geome-
tries under evaluation, which means the flow inside PCM should have
been treated as turbulent instead of laminar. However, using tube
length in Rayleigh number calculations is a common practice for fluid
inside a tube/cylinder (Lakeh et al., 2015; Evans et al., 1968) - that
should not apply to annular spaces. As for the temperature in Rayleigh
number definition, the temperature of inner surface depends on the
axial location of PCM cell. However, at the end of the melting process,
the inner wall temperature of the annulus has already reached the HTF
inlet temperature. Thus, Tr = ro = 838 K was assumed in Rayleigh number
calculations.
According to Table 8, the best set of non-dimensional groups for
error correlations is Group 1. The same groups were adopted for the
analysis of solidification results. The general error correlations attained
for that Group 1, including the coefficients and exponents achieved by
regression, are shown in Eqs. (18) and (19):
ro 0.92
Errormelting (%) = 2.5 × 10−11Ste−0.17Bi−3.4Ra2.1 ()L
[range of this study
: 1 < Bi < 7, 105 < Ra < 108, 0.4 < Ste < 4.5, 0.008 < ro/ L < 0.03]
(18)
r 0.47
Errorsolidification (%) = 1.2 × 10−4Ste 0.33Bi−1.36Ra0.84 ⎛ o ⎞
⎝L⎠ (19)
Fig. 13 illustrates the accuracy of the proposed correlations for
melting and solidification, respectively. The error lines in both figures
were set at 30%. Note that the error bars (lines) on Figs. 12 and 13 refer
to the difference percentage of the error calculated by the correlation
and the one obtained from CFD simulations. This indicates that the
maximum uncertainty of the proposed correlations is within ∼30%.
Note that the error correlations tend to overestimate the error at large
error values (e.g. > 100%), but at this point it is clear that convection
cannot be neglected.
Tables 9 and 10 summarize the range of non-dimensional groups
that were covered in this study.
According to Tables 7–10, the following conclusions can be drawn:
• As tube length increase (e.g. r /L becomes small), the error de-
o
creases.
• Both Rayleigh and Biot number are sensitive to the values of R andr .
Table 10 implies that the non-dimensional radius can be represented
by them. Thus, as R/ ro decreases, the Rayleigh and Biot also decrease
– along with the error.
• As Stefan number decreases, the error reduces. For the same case
IDs, the Rayleigh and Biot numbers are less for H525 as compared to
H325, which explains why the natural convection is less significant
when deploying H525.
• The error values can be best estimated by employing the general
error correlations (Eqs. (18) and (19)). However, as all of these
parameters are linked with each other, the Rayleigh number (e.g. as
defined by Eq. (13)) can be selected as a representative of all in-
fluential parameters in the error value. For Ra < 8 × 105 (considered
here as a critical Rayleigh number), natural convection can be ne-
glected (since the error in melting/solidification is < 1%). Note that
this value was obtained through the linear interpolation of the error
values for H525 in regards to the Rayleigh numbers indicated on
Table 10.
5. Conclusions
This study defines the conditions under which natural convection is
significant for high temperature shell-and-tube LHTES systems, and
Solar Energy 174 (2018) 489–501
determines the error between conduction-only and full convection
modelling approaches. From the results obtained, we can report the
following conclusions:
(1) To quantify the discrepancy between conduction and convection
modelling approaches, the error was defined as the difference be-
tween conduction and convection models in terms of the time re-
quired to reach 98% charging/discharging ratio. The errors were
clearly higher during the melting process compared to solidification
– i.e. 100% versus 30% maximum error values, for the cases stu-
died.
(2) The error was higher when tube length (L ) decreases and when the
non-dimensional radius R increases. The proposed specific geo-
rO
R2 − r 2
metry parameter, S = 2r L o , was shown to provide a good indicator
o
for the overall effect of geometry on the error. Using this parameter,
it was found that for S < 0.005 the natural convection could be
neglected for the boundary conditions and PCM properties range
used in this study.
(3) The results of a multi-variant regression analysis revealed that the
global error correlations depends on the following non-dimensional
r
groups: Lo , Ra, Ste , Bi . The error correlations were established for
melting and solidification and can be consulted in Eqs. (18) and
(19).
(4) While the error correlations are appropriate to quantify the error, a
critical Rayleigh number, Racritical = 8 × 105 , was obtained below
which the error is < 1% and the natural convection is negligible. It
is also worth noting that low Ra numbers are associated with the
high melting point PCMs (i.e. low Stefan numbers).
The outcomes of the present study also will help researchers to
identify cases where natural convection is negligible so that they can
use (much) faster conductivity models instead, to conduct coarse opti-
mizations for early design and feasibility analyses.
Acknowledgments
This work was supported by a grant from the Transfield Foundation
Research Grant – Round 2. Gonzalo Diarce would like to acknowledge
the Spanish's Ministry of Economy and Competitiveness (MINECO) for
financial support through the ekimProVe research project
(ENE2015–71083–R) as well as the Erasmus Mundus Panther exchange
program. The authors also would like to acknowledge the Faculty of
o
Engineering at the University of New South Wales (UNSW) for pro-
viding the Leonardi cluster without which this work would not have
been possible.
References
Abouali, O., Falahatpisheh, A., 2009. Numerical investigation of natural convection of
Al2O3 nanofluid in vertical annuli. Heat Mass Transf. 46 (1), 15.
Adine, H.A., El Qarnia, H., 2009. Numerical analysis of the thermal behaviour of a shell-
and-tube heat storage unit using phase change materials. Appl. Math. Model. 33 (4),
2132–2144.
Afrand, M., Farahat, S., Nezhad, A.H., Sheikhzadeh, G.A., Sarhaddi, F., Wongwises, S.,
2015. Multi-objective optimization of natural convection in a cylindrical annulus
mold under magnetic field using particle swarm algorithm. Int. Commun. Heat Mass
Transfer 60, 13–20.
Agyenim, F., Eames, P., Smyth, M., 2009. A comparison of heat transfer enhancement in a
medium temperature thermal energy storage heat exchanger using fins. Sol. Energy
83 (9), 1509–1520.
Agyenim, F., Hewitt, N., Eames, P., Smyth, M., 2010. A review of materials, heat transfer
and phase change problem formulation for latent heat thermal energy storage sys-
tems (LHTESS). Renew. Sustain. Energy Rev. 14 (2), 615–628.
Akgün, M., Aydın, O., Kaygusuz, K., 2007. Experimental study on melting/solidification
characteristics of a paraffin as PCM. Energy Convers. Manage. 48 (2), 669–678.
Al-Abidi, A.A., Mat, S., Sopian, K., Sulaiman, M.Y., Mohammad, A.T., 2013. Numerical
study of PCM solidification in a triplex tube heat exchanger with internal and ex-
ternal fins. Int. J. Heat Mass Transf. 61, 684–695.
Al-Arabi, M., El-Shaarawi, M., Khamis, M., 1987. Natural convection in uniformly heated
vertical annuli. Int. J. Heat Mass Transf. 30 (7), 1381–1389.
499
S.S. Mostafavi Tehrani et al.
Bauer, T., Breidenbach, N., Pfleger, N., Laing, D., Eckand, M., 2012. Overview of molten
salt storage systems and material development for solar thermal power plants. In:
Proceedings of the 2012 National Solar Conference for (SOLAR 2012), Denver.
Bellecci, C., Conti, M., 1993a. Phase change thermal storage: transient behaviour analysis
of a solar receiver/storage module using the enthalpy method. Int. J. Heat Mass
Transf. 36 (8), 2157–2163.
Bellecci, C., Conti, M., 1993b. Transient behaviour analysis of a latent heat thermal
storage module. Int. J. Heat Mass Transf. 36 (15), 3851–3857.
Brent, A., Voller, V., Reid, K.t.J., 1988. Enthalpy-porosity technique for modeling con-
vection-diffusion phase change: application to the melting of a pure metal. Numeric.
Heat Transf., Part A Appl. 13 (3), 297–318.
Cárdenas, B., León, N., 2013. High temperature latent heat thermal energy storage: phase
change materials, design considerations and performance enhancement techniques.
Renew. Sustain. Energy Rev. 27, 724–737.
Dhaidan, N.S., Khodadadi, J., 2015. Melting and convection of phase change materials in
different shape containers: a review. Renew. Sustain. Energy Rev. 43, 449–477.
Diarce, G., Campos-Celador, Á., Martin, K., Urresti, A., García-Romero, A., Sala, J., 2014.
A comparative study of the CFD modeling of a ventilated active façade including
phase change materials. Appl. Energy 126, 307–317.
Dincer, I., Rosen, M., 2002. Thermal Energy Storage: Systems and Applications. John
Wiley & Sons.
Dutil, Y., Rousse, D.R., Salah, N.B., Lassue, S., Zalewski, L., 2011. A review on phase-
change materials: mathematical modeling and simulations. Renew. Sustain. Energy
Rev. 15 (1), 112–130.
El-Shaarawi, M., Al-Nimr, M., 1990. Fully developed laminar natural convection in open-
ended vertical concentric annuli. Int. J. Heat Mass Transf. 33 (9), 1873–1884.
El Qarnia, H., 2009. Numerical analysis of a coupled solar collector latent heat storage
unit using various phase change materials for heating the water. Energy Convers.
Manage. 50 (2), 247–254.
Esen, M., Ayhan, T., 1996. Development of a model compatible with solar assisted cy-
lindrical energy storage tank and variation of stored energy with time for different
phase change materials. Energy Convers. Manage. 37 (12), 1775–1785.
Esen, M., Durmuş, A., Durmuş, A., 1998. Geometric design of solar-aided latent heat store
depending on various parameters and phase change materials. Sol. Energy 62 (1),
19–28.
Evans, L., Reid, R., Drake, E., 1968. Transient natural convection in a vertical cylinder.
AIChE J. 14 (2), 251–259.
Fan, Z., Ferreira, C.I., Mosaffa, A., 2014. Numerical modelling of high temperature latent
heat thermal storage for solar application combining with double-effect H2O/LiBr
absorption refrigeration system. Sol. Energy 110, 398–409.
Fang, M., Chen, G., 2007. Effects of different multiple PCMs on the performance of a
latent thermal energy storage system. Appl. Therm. Eng. 27 (5), 994–1000.
Flueckiger, S.M., Iverson, B.D., Garimella, S.V., Pacheco, J.E., 2014. System-level simu-
lation of a solar power tower plant with thermocline thermal energy storage. Appl.
Energy 113, 86–96.
Fluent, A., 2011. ANSYS Fluent User's Guide. Ansys Inc, p. 2427.
Fornarelli, F., Camporeale, S., Fortunato, B., Torresi, M., Oresta, P., Magliocchetti, L.,
Miliozzi, A., Santo, G., 2016. CFD analysis of melting process in a shell-and-tube
latent heat storage for concentrated solar power plants. Appl. Energy 164, 711–722.
Ghajar, A.C.Y.a.A.J., 2011. Heat and Mass Transfer: Fundamentals & Applications, fourth
ed. McGraw-Hill, New York.
Gil, A., Medrano, M., Martorell, I., Lazaro, A., Dolado, P., Zalba, B., Cabeza, L.F., 2010.
State of the art on high temperature thermal energy storage for power generation.
Part 1—Concepts, materials and modellization. Renew. Sustain. Energy Rev. 14 (1),
31–55.
http://www.pcmproducts.net/, 2018. <http://www.pcmproducts.net/>, LAST
ACCESSED ON 15/04/2018. (Accessed 15/04/2018 2018).
Ismail, K., Melo, C., 1998. Convection-based model for a PCM vertical storage unit. Int. J.
Energy Res. 22 (14), 1249–1265.
Jacob, R., Belusko, M., Fernández, A.I., Cabeza, L.F., Saman, W., Bruno, F., 2016.
Embodied energy and cost of high temperature thermal energy storage systems for
use with concentrated solar power plants. Appl. Energy 180, 586–597.
Kalhori, B., Ramadhyani, S., 1985. Studies on heat transfer from a vertical cylinder, with
or without fins, embedded in a solid phase change medium. J. Heat Transfer 107 (1),
44–51.
Kemink, R., Sparrow, E., 1981. Heat transfer coefficients for melting about a vertical
cylinder with or without subcooling and for open or closed containment. Int. J. Heat
Mass Transf. 24 (10), 1699–1710.
Keyhani, M., Kulacki, F., Christensen, R., 1983. Free convection in a vertical annulus with
constant heat flux on the inner wall. J. Heat Transfer 105 (3), 454–459.
Kuravi, S., Trahan, J., Goswami, D.Y., Rahman, M.M., Stefanakos, E.K., 2013. Thermal
energy storage technologies and systems for concentrating solar power plants. Prog.
Energy Combust. Sci. 39 (4), 285–319.
Lacroix, M., 1993. Numerical simulation of a shell-and-tube latent heat thermal energy
storage unit. Sol. Energy 50 (4), 357–367.
Lakeh, R.B., Lavine, A.S., Kavehpour, H.P., Wirz, R.E., 2015. Study of turbulent natural
convection in vertical storage tubes for supercritical thermal energy storage.
Numeric. Heat Transf., Part A: Appl. 67 (2), 119–139.
Li, Q., Tehrani, S.S.M., Taylor, R.A., 2017. Techno-economic analysis of a concentrating
solar collector with built-in shell and tube latent heat thermal energy storage. Energy.
Li, Y., He, Y., Song, H., Xu, C., Wang, W., 2013. Numerical analysis and parameters op-
timization of shell-and-tube heat storage unit using three phase change materials.
Renew. Energy 59, 92–99.
Liu, M., Saman, W., Bruno, F., 2012. Review on storage materials and thermal perfor-
mance enhancement techniques for high temperature phase change thermal storage
systems. Renew. Sustain. Energy Rev. 16 (4), 2118–2132.
500
Solar Energy 174 (2018) 489–501
Liu, M., Tay, N.S., Bell, S., Belusko, M., Jacob, R., Will, G., Saman, W., Bruno, F., 2016.
Review on concentrating solar power plants and new developments in high tem-
perature thermal energy storage technologies. Renew. Sustain. Energy Rev. 53,
1411–1432.
Longeon, M., Soupart, A., Fourmigué, J.-F., Bruch, A., Marty, P., 2013. Experimental and
numerical study of annular PCM storage in the presence of natural convection. Appl.
Energy 112, 175–184.
Mahdavi, M., Qiu, S., Tiari, S., 2016. Improvement of a novel heat pipe network designed
for latent heat thermal energy storage systems. Appl. Therm. Eng. 108, 878–892.
Mendes, P.S., Brasil, A.P., 1987. Heat transfer during melting around an isothermal
vertical cylinder. J. Heat Transfer 109 (4), 961–964.
Muhammad, M., Badr, O., Yeung, H., 2015a. CFD modeling of the charging and dis-
charging of a shell-and-tube latent heat storage system for high-temperature appli-
cations. Numeric. Heat Transf., Part A: Appl. 68 (8), 813–826.
Muhammad, M., Badr, O., Yeung, H., 2015b. Validation of a CFD melting and solidifi-
cation model for phase change in vertical cylinders. Numeric. Heat Transf., Part A:
Appl. 68 (5), 501–511.
Mostafavi Tehrani, S. Saeed, Shoraka, Yashar, Nithyanandam, Karthik, Taylor, Robert,
2018. Transient modelling of shell- and -tube and packed bed PCM storage systems
integrated with a CSP plant: a techno-economic comparison with sensible alter-
natives. Appl. Energy (in press).
Ng, K., Gong, Z., Mujumdar, A., 1998. Heat transfer in free convection-dominated melting
of a phase change material in a horizontal annulus. Int. Commun. Heat Mass Transfer
25 (5), 631–640.
Nithyanandam, K., Pitchumani, R., 2011. Analysis and optimization of a latent thermal
energy storage system with embedded heat pipes. Int. J. Heat Mass Transf. 54 (21),
4596–4610.
Nithyanandam, K., Pitchumani, R., 2014. Cost and performance analysis of concentrating
solar power systems with integrated latent thermal energy storage. Energy 64,
793–810.
Patankar, S., Ftamadhyani, S., 1977. Analwsis of melting in the presence of natural
confection in the melt region.
Pirasaci, T., Goswami, D.Y., 2016. Influence of design on performance of a latent heat
storage system for a direct steam generation power plant. Appl. Energy 162,
644–652.
Pointner, H., De Gracia, A., Vogel, J., Tay, N., Liu, M., Johnson, M., Cabeza, L.F., 2016.
Computational efficiency in numerical modeling of high temperature latent heat
storage: comparison of selected software tools based on experimental data. Appl.
Energy 161, 337–348.
Prasad, V., Kulacki, F., 1984. Natural convection in a vertical porous annulus. Int. J. Heat
Mass Transf. 27 (2), 207–219.
Riahi, S., Saman, W.Y., Bruno, F., Belusko, M., Tay, N., 2017. Impact of periodic flow
reversal of heat transfer fluid on the melting and solidification processes in a latent
heat shell and tube storage system. Appl. Energy 191, 276–286.
Sankar, M., Park, Y., Lopez, J., Do, Y., 2011. Numerical study of natural convection in a
vertical porous annulus with discrete heating. Int. J. Heat Mass Transf. 54 (7–8),
1493–1505.
Seddegh, S., Tehrani, S.S.M., Wang, X., Cao, F., Taylor, R.A., 2018. Comparison of heat
transfer between cylindrical and conical vertical shell-and-tube latent heat thermal
energy storage systems. Appl. Therm. Eng. 130, 1349–1362.
Seddegh, S., Wang, X., Henderson, A.D., 2015a. Numerical investigation of heat transfer
mechanism in a vertical shell and tube latent heat energy storage system. Appl.
Therm. Eng. 87, 698–706.
Seddegh, S., Wang, X., Henderson, A.D., 2016. A comparative study of thermal behaviour
of a horizontal and vertical shell-and-tube energy storage using phase change ma-
terials. Appl. Therm. Eng. 93, 348–358.
Seddegh, S., Wang, X., Henderson, A.D., Xing, Z., 2015b. Solar domestic hot water sys-
tems using latent heat energy storage medium: a review. Renew. Sustain. Energy Rev.
49, 517–533.
Showalter, S.K., Kolb, W.J., 2002. Development of a molten-salt thermocline thermal
storage system for parabolic trough plants.
Tao, Y.-B., Li, M.-J., He, Y.-L., Tao, W.-Q., 2014. Effects of parameters on performance of
high temperature molten salt latent heat storage unit. Appl. Therm. Eng. 72 (1),
48–55.
Tao, Y., Carey, V., 2016. Effects of PCM thermophysical properties on thermal storage
performance of a shell-and-tube latent heat storage unit. Appl. Energy 179, 203–210.
Tao, Y., He, Y., 2011. Numerical study on thermal energy storage performance of phase
change material under non-steady-state inlet boundary. Appl. Energy 88 (11),
4172–4179.
Tao, Y., He, Y., 2015. Effects of natural convection on latent heat storage performance of
salt in a horizontal concentric tube. Appl. Energy 143, 38–46.
Tao, Y., He, Y., Qu, Z., 2012. Numerical study on performance of molten salt phase
change thermal energy storage system with enhanced tubes. Sol. Energy 86 (5),
1155–1163.
Tehrani, S.S.M., Saffar-Avval, M., Kalhori, S.B., Mansoori, Z., Sharif, M., 2013a. Hourly
energy analysis and feasibility study of employing a thermocline TES system for an
integrated CHP and DH network. Energy Convers. Manage. 68, 281–292.
Tehrani, S.S.M., Saffar-Avval, M., Mansoori, Z., Kalhori, S.B., Abbassi, A., Dabir, B.,
Sharif, M., 2013b. Development of a CHP/DH system for the new town of Parand: an
opportunity to mitigate global warming in Middle East. Appl. Therm. Eng. 59 (1),
298–308.
Tehrani, S.S.M., Shoraka, Y., Diarce, G., Taylor, R.A., 2018a. An improved, generalized
effective thermal conductivity method for rapid design of high temperature shell-and-
tube latent heat thermal energy storage systems. Renew. Energy.
Tehrani, S.S.M., Shoraka, Y., Nithyanandam, K., Taylor, R.A., 2018b. Cyclic performance
of cascaded and multi-layered solid-PCM shell-and-tube thermal energy storage
S.S. Mostafavi Tehrani et al.
systems: a case study of the 19.9 MW e Gemasolar CSP plant. Appl. Energy 228,
240–253.
Tehrani, S.S.M., Taylor, R.A., 2016. Off-design simulation and performance of molten salt
cavity receivers in solar tower plants under realistic operational modes and control
strategies. Appl. Energy 179, 698–715.
Tehrani, S.S.M., Taylor, R.A., Ghazani, A.S., Saberi, P., 2016a. Part load behavior of
molten salt cavity receiver solar tower plants under storage mode operational mode.
In: ASME 2016 10th International Conference on Energy Sustainability collocated
with the ASME 2016 Power Conference and the ASME 2016 14th International
Conference on Fuel Cell Science, Engineering and Technology. American Society of
Mechanical Engineers, pp. V001T004A027–V001T004A027.
Tehrani, S.S.M., Taylor, R.A., Nithyanandam, K., Ghazani, A.S., 2017. Annual compara-
tive performance and cost analysis of high temperature, sensible thermal energy
storage systems integrated with a concentrated solar power plant. Sol. Energy 153,
153–172.
Tehrani, S.S.M., Taylor, R.A., Saberi, P., Diarce, G., 2016b. Design and feasibility of high
temperature shell and tube latent heat thermal energy storage system for solar
thermal power plants. Renew. Energy 96, 120–136.
Tiari, S., Qiu, S., 2015. Three-dimensional simulation of high temperature latent heat
thermal energy storage system assisted by finned heat pipes. Energy Convers.
Manage. 105, 260–271.
Trp, A., 2005. An experimental and numerical investigation of heat transfer during
technical grade paraffin melting and solidification in a shell-and-tube latent thermal
energy storage unit. Sol. Energy 79 (6), 648–660.
Trp, A., Lenic, K., Frankovic, B., 2006. Analysis of the influence of operating conditions
and geometric parameters on heat transfer in water-paraffin shell-and-tube latent
501
Solar Energy 174 (2018) 489–501
thermal energy storage unit. Appl. Therm. Eng. 26 (16), 1830–1839.
Voller, V., Cross, M., Markatos, N., 1987. An enthalpy method for convection/diffusion
phase change. Int. J. Numer. Meth. Eng. 24 (1), 271–284.
Wang, S., Faghri, A., Bergman, T.L., 2012. Melting in cylindrical enclosures: numerical
modeling and heat transfer correlations. Numeric. Heat Transf., Part A: Appl. 61 (11),
837–859.
Wang, W.-W., Wang, L.-B., He, Y.-L., 2015. The energy efficiency ratio of heat storage in
one shell-and-one tube phase change thermal energy storage unit. Appl. Energy 138,
169–182.
Wang, W.-W., Zhang, K., Wang, L.-B., He, Y.-L., 2013. Numerical study of the heat
charging and discharging characteristics of a shell-and-tube phase change heat sto-
rage unit. Appl. Therm. Eng. 58 (1), 542–553.
Wang, Y., Wang, L., Xie, N., Lin, X., Chen, H., 2016. Experimental study on the melting
and solidification behavior of erythritol in a vertical shell-and-tube latent heat
thermal storage unit. Int. J. Heat Mass Transf. 99, 770–781.
Xiao, X., Zhang, P., 2015a. Numerical and experimental study of heat transfer char-
acteristics of a shell-tube latent heat storage system: Part I-charging process. Energy
79, 337–350.
Xiao, X., Zhang, P., 2015b. Numerical and experimental study of heat transfer char-
acteristics of a shell-tube latent heat storage system: Part II–discharging process.
Energy 80, 177–189.
Cengel, Yunus A., 2015. Heat and Mass Transfer: Fundamentals & Applications, fifth ed.
McGraw Hill Education, New York.
Zalba, B., Marı́n, J.M., Cabeza, L.F., Mehling, H., 2003. Review on thermal energy storage
with phase change: materials, heat transfer analysis and applications. Appl. Therm.
Eng. 23 (3), 251–283.