Solar Energy 153 (2017) 628–654

Contents lists available at ScienceDirect

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

Review

A review on experience feedback and numerical modeling of packed-bed

thermal energy storage systems
Thibaut Esence a,⇑, Arnaud Bruch a, Sophie Molina a, Benoit Stutz b, Jean-François Fourmigué a
a
Université Grenoble Alpes (UGA), CEA-Liten-DTBH, Laboratoire de Stockage Thermique, F-38000 Grenoble, France
b
LOCIE, Univ. Savoie, Campus Scientifique, Savoie Technolac, F-73376 Le Bourget-du-Lac Cedex, France

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

Article history: Solar thermal energy is a clean, climate-friendly and inexhaustible energy resource. It is therefore
Received 28 April 2016 promising to cope with fossil fuel depletion and climate change. Thermal storage enables to make this
Received in revised form 6 December 2016 intermittent energy resource dispatchable, reliable on demand and more competitive. Nowadays, most
Accepted 10 March 2017
of the concentrated solar power plants equipped with integrated thermal storage systems use the two-
Available online 21 June 2017
tank molten salt technology. Despite its relative simplicity and efficiency, this technology is expensive
and requires huge amounts of nitrate salts. In the short to medium term, packed-bed thermal energy stor-
Keywords:
age with either liquid or gaseous heat transfer fluid is a promising alternative to reduce storage costs and
Thermal energy storage
Packed bed
hence improve the development of solar energy. To design reliable, efficient and cost-effective packed-
Thermocline bed storage systems, this technology, which involves many physical phenomena, has to be better under-
Concentrated solar power (CSP) stood.
This paper aims to sum up some key aspects about design, operation, and performances of packed-bed
storage systems. In the first part, most representative setups and their experience feedback are presented.
The controllability of packed-bed storage systems and the special influence of thermal stratification are
pointed out. In the second part, the various numerical models used to predict packed-bed storage perfor-
mances are reviewed. In the last part, some useful correlations enabling to quantify the main physical
phenomena involved in packed-bed operation and modeling are presented and compared. The correla-
tions investigated enable to calculate fluid/solid and fluid/wall heat transfer coefficients, effective ther-
mal conductivity and pressure drop in packed beds.
Ó 2017 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 630
2. Experimental studies and feedback. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
2.1. Experimental applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
2.2. Packed-bed storage behavior and operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
2.2.1. Operating principle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
2.2.2. Fluid flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
2.2.3. Stability of discharge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
2.2.4. Efficiency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
2.2.5. Cyclic operating conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
2.3. Tank design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 636
2.4. Heat transfer fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637
2.5. Solid filler material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637
3. Numerical models for packed-bed storage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638
3.1. Models with negligible thermal gradient inside solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638
3.1.1. The Schumann model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638

⇑ Corresponding author.
E-mail address: thibaut.esence@cea.fr (T. Esence).

http://dx.doi.org/10.1016/j.solener.2017.03.032
0038-092X/Ó 2017 Elsevier Ltd. All rights reserved.
 T. Esence et al. / Solar Energy 153 (2017) 628–654 629

Nomenclature

A viscous parameter in Ergun-type equations Greek symbols

As superficial area of solids (m2) a thermal diffusivity (m2 s
1), a ¼ k=ðq  cp Þ
ab external surface area of packed-bed per unit bed volume c ratio of volumetric heat capacity (qcp) to effective vol-
(m2 m
3) umetric heat capacity of the bed
as surface area of solid per unit bed volume (m2 m
3), see db dimensionless solid–fluid temperature difference
Eq. (4) e void fraction
aw surface area of tank wall per elementary unit wall vol- e emissivity
ume (m2 m
3) K dimensionless diffusion coefficient as defined by
B inertial parameter in Ergun-type equations Votyakov and Bonanos (2014), K ¼ 1 þ ðcs  cf  PeÞ2 =Bi0
Bi dimensionless Biot number, see Eq. (1) k thermal conductivity (W m
1 K
1)
Bi0 dimensionless Biot number as defined by Votyakov and l dynamic viscosity (Pa s)
Bonanos (2014), Bi0 ¼ h  as  L2 =keff m0 cinematic viscosity evaluated for surface conditions
cp specific heat capacity (J kg
1 K
1) (m2 s
1)
D diameter (m) pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi m1 cinematic viscosity evaluated for free stream conditions
E thermal effusivity (J K
1 m
2 s
1/2), E ¼ q  cp  k (m2 s
1)
f term of tortuosity, defined in Cheng and Hsu (1998) q density (kg m
3)
fv friction factor.h as defined by i Ergun (1952), r Stephan-Boltzmann constant,
f v ¼ DP  D2s  e3 L  lf  usup  ð1
eÞ2 r ¼ 5:67  10
8 W m
2 K
4
h convective heat transfer coefficient (W m
2 K
1) s dimensionless time as defined by Votyakov and
Hg dimensionless Hagen number, Hg ¼ qf  DP  D3s =ðL  l2f Þ Bonanos (2014), s ¼ t=ðL2 =aeff Þ
hv convective heat transfer coefficient per unit bed volume
(W m
3 K
1) Subscripts
L length of the packed bed in flow direction (m) 1 relating to ambient
m mass (kg) b relating to the whole packed bed
Nu dimensionless Nusselt number, Nu ¼ h  Ds =kf eff effective value
Pe dimensionless Péclet number as defined by Votyakov eq relating to/calculated with an equivalent diameter
and Bonanos (2014), Pe ¼ u  L=½k=ðq  cp Þeff eq, a relating to/calculated with the diameter of the sphere of
Pr dimensionless Prandtl number, Pr ¼ lf  cpf =kf equivalent specific area, see Eq. (37)
Q energy (J) eq, V relating to/calculated with the diameter of the sphere of
Q⁄ dimensionless
  energy, equivalent volume, see Eq. (38)
Q  ¼ Q = ðq  cp Þeff  Vb  ðTmax
Tmin Þ ext relating to the external surface of the walls
r radial coordinate (m) f relating to fluid phase
R radius (m) int relating to the internal surface of the walls
Re dimensionless Reynolds number, Re ¼ qf  usup  Ds =lf max maximal value
T temperature (K) min minimal value
T⁄ dimensionless temperature, T ¼ ðT
Tref Þ=ðTmax
Tref Þ mix relating to fluid mixing and braiding effect
t time (s) r relating to radial direction (perpendicular to flow direc-
t⁄ dimensionless time, t ¼ t=ðL=uÞ tion)
u interstitial fluid velocity (m s
1), u ¼ m=ð _ qf  e  p  R2b Þ ref relating to the reference condition
U overall heat transfer coefficient between the inside and s relating to filler solids
the outside of the tank (W m
2 K
1) tot total value
usup superficial fluid velocity (m s
1), usup ¼ e  u w relating to the walls of the storage tank
V volume (m3) z relating to axial direction (flow direction)
w velocity of the thermal front (m s
1), see Eq. (10)
x dimensionless time  of the  Schumann model, Superscripts
x ¼ hv  ðt
z=uÞ ð1
eÞ  qs  cps 0 with contribution of conduction in each phase (stagnant
y dimensionless length of the Schumann model, effective thermal conductivity)
y ¼ hv  z=ðe  qf  cpf  uÞ
C with contribution of conduction through contact sur-
z axial coordinate in flow direction (m) faces between solids
z⁄ dimensionless axial coordinate, z ¼ z=L R with contribution of radiative heat transfer

3.1.2. One-dimensional single-phase models: simplification of the Schumann model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640

3.1.3. Perturbation model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
3.1.4. One-dimensional three-phase models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
3.1.5. Two-dimensional models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
3.2. Models with non-negligible thermal gradient inside the solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
3.2.1. One-dimensional intraparticle conduction models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
3.2.2. Two-dimensional intraparticle conduction models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
3.3. Comparison between models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
3.3.1. Calculation time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
3.3.2. Parametric studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
4. Physical properties of porous media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
4.1. Fluid/solid heat transfer coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
630 T. Esence et al. / Solar Energy 153 (2017) 628–654

4.1.1. Determination methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643

4.1.2. Comparison of selected correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643
4.1.3. Effective heat transfer coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644
4.2. Fluid/wall heat transfer coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644
4.2.1. Comparison of some correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
4.3. Effective thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
4.3.1. Effective thermal conductivity contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646
4.3.2. Effective thermal conductivity with stagnant fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646
4.3.3. Contribution of fluid mixing and turbulences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 648
4.3.4. Integration of effective thermal conductivity in models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 648
4.4. Pressure losses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
4.4.1. The Ergun equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
4.4.2. Enhancement of the Ergun equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
4.4.3. Particular considerations for rock beds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 650
4.4.4. Comparison of some correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 650
5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 652

1. Introduction throughout the world gets integrated TES (NREL; Tian and Zhao,
2013) and this share is expected to grow in the next future since
Variable renewable energy sources, like wind and solar, are integrated TES is increasingly necessary to ensure the competitive-
expected to play a key role to cope with fossil fuel depletion and ness of CSP power generation (Philibert, 2014).
climate change. With interconnection and demand response, There are mainly three ways of storing heat: sensible heat stor-
energy storage is one of the most promising ways to enhance the age, latent heat storage, and chemical heat storage (Gil et al., 2009).
penetration of intermittent energy sources in power grids while Although promising in terms of performance, latent heat and
ensuring electricity security (Akinyele and Rayudu, 2014; chemical heat storage technologies are not commercially mature
Hohmeyer and Bohm, 2015; Pudjianto et al., 2014). Integration of yet. Latent heat storage for CSP is still at the prototype stage (Xu
energy storage system into a power plant enables the operator to et al., 2015) while chemical heat storage is still at the proof-of-
avoid frequent start up and shut down of electrical device, and to concept stage (Pardo et al., 2014). Currently, only sensible heat
increase load factor by dispatching production according to storage is widely used in CSP plants.
demand and current electricity price. This is therefore a way to The most widespread and state-of-the-art sensible TES technol-
reduce the levelized cost of energy (LCOE), thus making variable ogy is two-tank molten salt storage, an example of which is
renewables more competitive (Xu et al., 2012b). depicted in Fig. 1. It consists in using molten nitrate salts, which
At large scale, thermal energy storage (TES) is cheaper and more have high volumetric heat capacity and high operating tempera-
efficient than most electric storage systems (Philibert, 2011; ture, as storage medium. During load periods, cold molten salt is
Sabihuddin et al., 2015). Consequently, TES appears to be particu- heated in solar receptors and transferred from cold to hot tank.
larly suitable for concentrated solar power (CSP) plants which gen- The process is inverted to recover heat and run thermal processes.
erate heat before converting it into electricity. TES is therefore a This technology is well known, easy to handle and very efficient
key advantage for CSP compared to photovoltaic power plants given that hot and cold fluids are well separated (Gil et al.,
which directly generate electricity. About half of CSP plants 2009). However, this storage system is relatively expensive and

Fig. 1. Two-tank direct thermal storage system (Libby, 2010).

 T. Esence et al. / Solar Energy 153 (2017) 628–654 631

Fig. 2. Single tank direct thermocline system (Libby, 2010).

requires a huge amount of costly nitrate salt while world’s capacity 2. Experimental studies and feedback
production of natural nitrate is limited and won’t be able to supply
a soaring demand (García-Olivares et al., 2012). In this part, the main experimental setups are briefly presented
To face these difficulties, a possible alternative for sensible TES and feedback of the experimental studies is discussed in terms of
is single-tank packed-bed storage which consists in using solids system performances, operation and design. Finally, the main char-
as heat storage medium and a heat transfer fluid (HTF) in direct acteristics and some key aspects of filler materials and HTF are
contact with the solids to convey heat. As depicted in Fig. 2, con- presented.
trary to two-tank system, packed-bed storage requires a single
tank in which hot and cold zones are separated by a transition 2.1. Experimental applications
zone with a thermal gradient called ‘‘thermocline”. For a fixed
theoretical energy capacity, liquid/solid single-tank storage has Some of the most relevant packed-bed storage installations
almost the same size than one of the tanks of two-tank system with liquid or gaseous HTF are listed in Tables 1 and 2 respectively.
(Brosseau et al., 2005). Therefore, with typical fillers (i.e. rocks It should be noticed that some characteristics of the referenced
with molten salts or thermal oil), theoretical energy density of systems are not directly available in the original papers. Some of
packed-bed storage in terms of tank volume is about 50% higher them were easily deduced and calculated from given data, while
than the one of two-tank storage (Galione et al., 2015). Moreover, the others were estimated from partial data. The former are indi-
by using cheap solids like rocks and sand, it is possible to reduce cated with one asterisk (⁄) and the latter with two asterisks (⁄⁄).
the amount of expensive fluid by approximately 70% (Galione The first full-scale industrial packed bed for CSP storage was
et al., 2015). Theoretical energy density of most packed-bed implemented in 1982 in Solar One (Faas et al., 1986; McDonnell
storages calculated with specific heat of filler materials is about Douglas Astronautics Company, 1986). Solar One was a 10-MWe
180–250 MJ m
3 (i.e. 50–70 kWhth m
3) with a temperature dif- CSP tower plant using water as HTF and integrating a 182-
ference of 100 °C. MWhth storage of more than 3000 m3 composed of a mixture of
As a consequence, overall investment cost of packed-bed stor- rocks and sand with oil up to 290 °C as HTF. It was rapidly shut
age is up to 35% lower compared to two-tank storage (Brosseau down (in 1986) due to accidental admission of water in the tank
et al., 2005; Libby, 2010). Using packed-bed storage instead of which caused sudden pressure elevation and tank damage. As
two-tank storage thus enables to reduce LCOE of CSP plants preparatory work on Solar One, a similar pilot-scale 5.7-MWhth
(Cocco and Serra, 2015). storage was tested (Hallet and Gervais, 1977). From these two
Besides, given that the fluid is essentially used as heat transfer experiments only partial data and little feedback are available in
medium instead of as storage medium, the choice of the fluid is the literature.
more flexible. It is therefore possible to use a greater diversity of More detailed experimental results are available from the work
fluids, either liquid or gaseous. of Pacheco et al. (2002), who tested a 2.3-MWhth storage of about
However, while many two-tank storage systems have been 40 m3 comprised of rocks and sand with molten salt up to 390 °C as
implemented and tested, only a couple of packed-bed storage sys- HTF. Nowadays this reference paper is used by many authors to
tems have been experienced at large scale so far. In the first part, validate numerical models (Bayón and Rojas, 2013; Flueckiger
this literature review aims to present the main available feedback et al., 2014; Van Lew et al., 2011; Xu et al., 2012b). However,
about behavior, operation and performances of packed-bed storage exhaustive operative conditions used during tests were not pub-
systems. The second part presents and briefly compares numerical lished, and available temperature profiles suffer from significant
methods from the literature for thermal modeling of the system. and unexplained scattering (Fig. 5).
The third part deals with correlations for determination of the In addition to these two well-known experimental setups, some
main physical parameters involved in operation and modeling of other pilot-scale packed-bed storages have been implemented and
packed-bed storage. tested.
Table 1

632
Main characteristics of some liquid/solid packed-bed storage applications from the literature.

Publications Solid Fluid Db (m) L (m) Ds (mm) e Db/Ds T (°C) u Packed-bed

temperature
measurements

Hallet and Gervais (1977) River gravels (granite) and silica sand Caloria HT 43 (oil) 3.2 12.2 25/1.5 0.25 128 218–302 0.3–3.3 mm/sa Fluid
Axial/Radial
Faas et al. (1986), McDonnell Douglas Astronautics Company Rocks and sand Caloria HT 43 (oil) 18.3 13.7 25/– 0.22 732 200–290 – Fluid
(1986) Axial
Bhavsar and Balakrishnan (1990) Rocks HP Hytherm 500 (oil) 2.2 2.0 50 0.30 44 230–247 0.4 mm/sa Fluid/Solid
Axial/Radial
Pacheco et al. (2002) Quartzite rocks and silica sand Hitec XLÒ (Ca(NO3)2–NaNO3–KNO3, 42–15–43 wt%) 3.0 6.1 19/– 0.22 158 290–390 2.6 mm/sb Fluid
Axial/Radial
0.04–0.12 mm/sa
TM
Mawire et al. (2009), Mawire and McPherson (2009), Mawire Sandy stones CALFLO LT (oil) 0.29 0.3 5 0.42 58 20–240 Fluid
and Taole (2011) Axial
Mawire et al. (2010) Silica glass pebbles Shell thermia oil B (oil) 0.035 0.45 3 0.42 12 30–160 10–19 mm/sa Fluid
Axial
Yang et al. (2014) Ceramic spheres HitecÒ (KNO3–NaNO2–NaNO3, 53–40–7 wt%) 0.263 0.55 30 – 8.8 280–355 3 mm/sb Fluid
Axial
Bruch et al. (2014a, 2017) Silica gravels and silica sand TherminolÒ 66 (oil) 1.0 3.0 30/3 0.27 33 50–250 0.7–2.8 mm/sc Fluid/Solid
Axial/Radial
Rodat et al. (2015) Silica gravels and silica sand TherminolÒ 66 (oil) 2.5 6.0 30/3 0.27 83 100–250 0.5–1.7 mm/sc Fluid/Solid
Axial/Radial

T. Esence et al. / Solar Energy 153 (2017) 628–654

Hoffmann et al. (2016) Quartzite rocks Rapeseed oil 0.4 1.8 40 0.41 4.5 160–210 0.5 mm/sa Fluid
Axial/Radial
a
Recalculated value.
b
Estimated value from partial data.
c
There is a subscript mistake in Bruch et al. (2017): the velocity values correspond to interstitial velocities (and not to superficial velocities).

Table 2
Main characteristics of some gas/solid packed-bed storage applications from the literature.

Publications Solid Fluid Db (m) L (m) Ds (mm) e Db/Ds T (°C) u Packed-bed

temperature
measurements

Hollands et al. (1984), Shewen et al. (1978) River gravels Air 1.8  1.8 0.38–1.82 18.1 0.42 99 20–67 15–30 cm/sb Fluid/Solid
Axial/radial
Coutier and Farber (1982) Rocks Air 0.57 0.84 18–28 – 20–32 – – Fluid/Solid
Axial/Radial
Shitzer and Levy (1983) Crushed quarry rocks Air 1.0 2.45 18–45 0.34 22.2–55.6 30–75 20–46 cm/sb Fluid/Solid
Axial
Beasley and Clark (1984) Soda lime glass spheres Air 0.375 0.62 12.6 0.364 30 25–70 40–300 cm/sb Fluid/Solid
Axial/Radial
Meier et al. (1991) Porcelain spheres Air 0.15 1.20 20 0.40 7.5 25–550 90 cm/s Fluid
Axial
Zanganeh et al. (2012) Sedimentary rocks Air 2.5–4 2.9 20–30 0.342 83–200 20–500 3.0 cm/sb Fluid
Axial
Klein et al. (2013) Ceramic balls Flue gas/air 0.40 0.62 19 0.39 21 25–900 90–140 cm/sb Fluid/Solid
Axial/Radial
Kuravi et al. (2013) Bricks Air 0.508  0.203 1.07 50.8  203  178 0.20 – 20–530 510–590 cm/sb Fluid/Solid
Axial/Radial
Okello et al. (2014) Crushed rocks Air 0.40 0.40 14.6 0.3 27 20–350 12 cm/sb Fluid
0.40 0.90 8 22 cm/sb Axial
Anderson et al. (2014) Alumina spheres Air 0.572 3.05 6 0.40 20 25–120 485–975 cm/sa Fluid
Axial
Zanganeh et al. (2014) Rocks Air 5–6 4.0 – – – 250–550 – Fluid
Axial/Radial
Cascetta et al. (2015) Sintered alumina beads Air 0.58 1.80 7–9 0.39 64.5–83.0 38–239 90–225 cm/sb Fluid
Axial/Radial
a
Recalculated value.
b
Estimated value from partial data.
 T. Esence et al. / Solar Energy 153 (2017) 628–654 633

Zanganeh et al. (2012), built a 6.5-MWhth rock-bed storage run- One’s TES (more than 3000 m3) lost only 2.5% of its energy content
ning up to 500 °C with air as HTF. This truncated conical bed was in 20 h (Faas et al., 1986) with similar average temperature.
buried to tackle thermomechanical stresses on the walls. The Besides, absolute and relative dimensions of the tank and the
authors used the experimental results to validate a numerical solid particles of the bed influence flow distribution and velocity
model and to design the full-scale storage of the 3.9-MWth para- profiles, thereby affecting heat exchange and thermal stratification
bolic trough CSP plant of Aït-Baha (Morocco) which started pro- in the storage. In small tanks, the relative influence of the walls
duction in 2014 (Airlight Energy; Zanganeh et al., 2014). Since over velocity profiles is higher than in large tanks due to edge
Solar One, this is probably the first commercial CSP plant with effects. Flow channeling near the walls is also influenced by the
packed-bed storage. It consists of a pebble bed with air as HTF tank-to-particle diameter ratio.
(Zanganeh et al., 2014) similar to the one presented and designed Proportion effects may also affect laboratory-scale experiments.
in Zanganeh et al. (2012). As far as the authors know, no opera- While the height occupied by the thermocline, which affects stor-
tional data have been published so far. age behavior and efficiency, is relatively thin in industrial-scale or
Kuravi et al. (2013), implemented a structured packed bed of pilot-scale systems, many laboratory-scale results show thermo-
bricks of about 0.1 m3 with air up to 530 °C as HTF. They confirmed cline which occupies the whole height of the storage during charge
the viability of structured packed beds in terms of fluid distribution or discharge (Anderson et al., 2014; Bhavsar and Balakrishnan,
and thermal stratification and validated a numerical model with 1990; Mawire and Taole, 2011; Meier et al., 1991; Shitzer and
the experimental data. Levy, 1983; Yang et al., 2014).
The experimental studies carried out by CEA (Bruch et al., In these conditions, experimental results from laboratory-scale
2014a,b, 2017; Rodat et al., 2015) investigated the behavior of systems should be considered with precaution because their
packed beds of 2.4 and 30 m3 comprised of rocks and sand with behavior and their performances may significantly differ from
thermal oil up to 300 °C as HTF. The experimental results show industrial-scale.
smooth temperature profiles, with low experimental scattering,
and exhibit a very repeatable and robust stabilized state through
2.2. Packed-bed storage behavior and operation
various flow and temperature conditions fixed by the operator.
The studies highlight in particular the influence of various cycling
2.2.1. Operating principle
conditions on packed-bed performances.
Packed-bed charging and discharging consist in shifting hot and
This behavior through charge and discharge cycles was also
cold regions in the tank by circulating HTF through the solid stor-
observed by Cascetta et al. (2015), on a 0.5-m3 packed bed of alu-
age medium. For exergy and process efficiency purposes, the stor-
mina beads with air up to 240 °C. The authors investigated the
age should be kept well stratified, i.e. with hot and cold regions
influences of aspect ratio (i.e. height-to-diameter ratio of the bed),
well separated by a thin thermocline with steep thermal gradient
air flow rate, temperature level and inertia of the walls (highlighted
(Haller et al., 2009; Qin et al., 2012). The better the stratification,
by radial temperature profiles) on the storage performances.
the larger the amounts of hot and cold fluid (Fig. 4), and hence
Except the above mentioned installations, most of the experi-
the higher the storage exergy capacity and the longer the operation
mental data available in the literature come from laboratory-
of heat production process and thermal process at nominal tem-
scale setups. These experiments enable to investigate some phe-
peratures. In CSP plants, the heat production process and the ther-
nomena but are often non-representative of the overall behavior
mal process are respectively solar receiver and power block.
of large size installations (Fig. 3).
Given that hot and cold fluids coexist in the storage, buoyancy
For example, heat losses are proportionally higher in small
effects may occur. In order to preserve thermal stratification, espe-
tanks due to higher surface-to-volume ratio. Despite thermal insu-
cially during standby periods, this phenomenon is exploited
lation, several small size installations showed significant thermal
through the layout of the storage: hot fluid is injected and
losses which affected experimental results (Hoffmann et al.,
extracted by/from the top and cold fluid by/from the bottom of
2016; Klein et al., 2013; Kuravi et al., 2013; Okello et al., 2014;
the storage (Close, 1965; Zanganeh et al., 2012).
Shewen et al., 1978). As an example, the experimental system
During charging or discharging process, the thermocline moves
studied by Okello et al. (2014) (about 0.05 m3), lost about 25% of
respectively downwards or upwards inside the storage and slightly
its energy content during a 18-h standby period, while the Solar
spreads due to non-infinite heat transfer between solid and HTF,
thermal diffusion and fluid mixing. Upstream from the thermocline,
a new zone at relatively uniform temperature is generated. This
typical behavior has been experimentally observed or theoretically
predicted by numerous authors (Anderson et al., 2014; Beasley and
Clark, 1984; Bhavsar and Balakrishnan, 1990; Bruch et al., 2014a;
Cascetta et al., 2015; Coutier and Farber, 1982; Hallet and
Gervais, 1977; Klein et al., 2013; Kuravi et al., 2013; McDonnell
Douglas Astronautics Company, 1986; Meier et al., 1991; Okello
et al., 2014; Pacheco et al., 2002; Shewen et al., 1978; Shitzer and
Levy, 1983; Van Lew et al., 2011; Yang et al., 2014; Zanganeh
et al., 2012). Some representative results from installations with
different sizes, filler materials and temperature levels are depicted
through Figs. 5–7.

2.2.2. Fluid flow

Plug flow is the ideal way of circulating HTF in order to prevent
mixing and to keep a well stratified storage (Hollands et al., 1984).
That’s why a uniform distribution which prevents jet effects is
essential to preserve stratification and improve storage efficiency.
Fig. 3. Dimensions of referenced packed-bed storage systems. Reducing flow heterogeneities and natural convection, which cause
634 T. Esence et al. / Solar Energy 153 (2017) 628–654

Fig. 4. Differing degrees of stratification within a storage tank with the same amount of stored heat (a) left, highly stratified, (b) right, moderately stratified (Haller et al.,
2009).

Fig. 5. Measured temperature profiles during discharge of a 4.3-m3 packed bed of

rocks and sand with molten salts as HTF (Pacheco et al., 2002).

mixing during fluid circulation, is also one of the roles of the solid
filler material (Gil et al., 2009; Yang et al., 2014).
Furthermore, to some extent, reducing fluid velocity lowers the
spreading of the thermocline by reducing fluid mixing, Biot num-
ber of the solids (i.e. the influence of the thermal gradient inside Fig. 7. Measured and simulated centerline solid temperatures during discharge of a
solids and hence thermal inertia of the solids) and by coming close 0.08-m3 packed bed of ceramic balls with air as HTF (Klein et al., 2013).

to ideal process with thermal equilibrium between fluid and solids.

This influence of fluid velocity was numerically shown on both liq- Bonanos, 2015; Yang and Garimella, 2010b) and experimentally
uid/solid and gas/solid packed-bed storage systems by several observed on gas/solid systems by Anderson et al. (2014) and
authors (Durisch et al., 1986; Motte et al., 2015; Votyakov and Cascetta et al. (2015). However, given that reducing fluid velocity

Fig. 6. Dimensionless fluid temperatures during discharge of a 2.4-m3 packed bed of rocks and sand with thermal oil as HTF (Bruch et al., 2014b).
 T. Esence et al. / Solar Energy 153 (2017) 628–654 635

raises the residence time, this statement remains true as long as During typical operation, the thermocline has to ‘‘stay” inside
time dependent phenomena which lead to destratification of the the storage to keep outlet temperature as constant as possible.
storage, like heat losses or heat conduction through the bed, Otherwise, heat production and thermal processes would be
remain second order phenomena (Yang and Garimella, 2010a; respectively fed with increasing or decreasing fluid temperature
Votyakov and Bonanos, 2015). This phenomena was illustrated (with related efficiency drop).
by Bruch et al. (2017), with a liquid/solid packed-bed storage sys- On the other hand, if thermocline were not extracted at all from
tem: to some extent, increasing fluid velocity slightly increases the storage, the tank volume corresponding to the whole thermo-
both energy efficiency and utilization rate of charge/discharge cline thickness wouldn’t be exploited, which would lead to a signif-
cycles by lowering residence time and hence heat losses of the icant reduction of the actual capacity of the storage. Furthermore,
fluid. under cyclic operational conditions, non-extraction of at least a
part of the thermocline leads to unlimited increase in thermocline
thickness due to diffusion and mixing. In this case, the thermocline
2.2.3. Stability of discharge
would end up occupying the whole height of the tank after several
As explained by Pacheco et al. (2002), the outlet temperature of
charge/discharge cycles and the storage would lose its storage
packed-bed storage is constant as long as the thermocline remains
capacity (Bayón et al., 2013).
in the tank. This was experimentally illustrated in charge by
As a result, a trade-off should be found between maximization
Pacheco et al. (2002) (Fig. 8), and in discharge by Bruch et al.
of heat production and thermal process efficiencies on the one
(2014b) (Fig. 9). Fig. 8 shows the inlet and outlet fluid tempera-
hand, and optimization of the storage capacity over long-term
tures of the packed-bed storage during a charging process. While
operation on the other hand.
the inlet temperature remains approximately constant (except
during a transient stage at the very beginning), the outlet temper-
2.2.4. Efficiency
ature is first constant then gradually increases as the thermocline
In terms of efficiency, packed-bed storage systems may be
approaches the exit of the storage at the end of the charging pro-
assessed through an exergy efficiency parameter which accounts
cess. The same behavior is illustrated in Fig. 9 which shows inlet
for the temperature level at which energy is stored and recovered
and outlet dimensionless temperatures and recovered energy dur-
(Anderson et al., 2014), but in most cases, a simple energy effi-
ing the discharging process at constant flow rate presented in
ciency is used: the energy recovered above an arbitrary fixed tem-
Fig. 6. As soon as the thermal gradient reaches the top of the stor-
perature level is compared to the energy previously stored. The
age (t⁄ ffi 1.45), the outlet temperature and the slope of the energy
chosen temperature level is usually defined with a dimensionless
curve (which is proportional to discharge power due to constant
cut-off temperature which takes a zero value when it corresponds
flow rate) decrease.
to the temperature of discharge (the storage is totally discharged
until the outlet fluid temperature reaches the temperature of dis-
charge) and a value of one when it corresponds to the temperature
of charge (the storage is discharged only as long as the outlet fluid
temperature is equal to the temperature of charge). During the pre-
liminary experiments performed before the building of the Solar
One CSP plant, Hallet and Gervais (1977), reached an energy effi-
ciency of 87% on an oil-rock packed-bed storage of about 100 m3.
In operating conditions, the storage of Solar One (more than
3000 m3) had an overall energy efficiency of 92% (Faas et al.,
1986). Under cyclic operation with various operating conditions,
Bruch et al. (2017), obtained energy efficiencies between 88% and
92% on an oil-rock packed-bed storage of 2.4 m3. Values of the
same order of magnitude had been obtained numerically by sev-
eral authors. For molten salts-rock packed-bed storages between
about 2000 and 2500 m3 with a cut-off dimensionless temperature
of 0.95, Libby (2010), calculated energy efficiencies of 89%. Van
Lew et al. (2011), performed calculations on a 580-m3 oil-rock
Fig. 8. Measured temperatures of salt pumped from the bottom and returned to the
top of the thermocline tank during a charging cycle (Pacheco et al., 2002). packed-bed storage and showed typical energy efficiencies
between 90% and 95% depending on aspect ratio, solid size and
bed void fraction. Various configurations of oil-rock packed-bed
storage were simulated by Yang and Garimella (2010b), and
showed energy efficiencies between 75% and 95%. The calculations
performed by Xu et al. (2012b), on a 200-m3 molten salts-rock
packed-bed storage resulted in an energy efficiency of 90% with a
cut-off dimensionless temperature of 0.8. Zavattoni et al. (2014),
showed that the energy efficiency of an industrial-scale air-rock
packed-bed storage is about 92% (even when pumping costs are
accounted for). Therefore, there is a wide agreement in the litera-
ture about the good energy efficiency of packed-bed thermal
energy storage systems.

2.2.5. Cyclic operating conditions

Long term behavior of packed-bed storage has been poorly
investigated so far. This was mainly numerically studied by some
Fig. 9. Dimensionless inlet and outlet tank temperatures and dimensionless authors. As mentioned above, Bayón et al. (2013), have shown that
extracted energy during the discharging process (Bruch et al., 2014b). the thermocline region should be partly extracted from the tank at
636 T. Esence et al. / Solar Energy 153 (2017) 628–654

the end of charges and discharges so that the storage doesn’t lose
its storage capacity over cycles by unlimited expansion of the ther-
mocline. Zanganeh et al. (2012) and Zavattoni et al. (2015), per-
formed respectively first-law and second-law numerical analyses
of a high-temperature air/rock bed under cyclic operation. After a
transition phase depending on initial thermal conditions of the
storage, they show a stabilized state with constant performances
over cycles. Mertens et al. (2014), numerically showed that
steady-state packed-bed storage operation can withstand variation
of charging and discharging processes by drawing upon the heat
stored during initial preheating cycles. Therefore, operation of
packed-bed storages under cyclic operation is to some extent
flexible.
Experimental study with numerous successive charge/dis-
charge cycles has been carried out by Bruch et al. (2014a), on a
packed bed of rocks and sand with thermal oil as HTF. As foreseen
by Zanganeh et al. (2012), they show a repeatable and robust sta-
bilized periodic state under cyclic operation when constant inlet
fluid temperature and thermocline extraction strategy are
imposed. The stabilized state, which defines the long-term perfor- Fig. 11. End-cycle dimensionless temperature profiles in the ‘‘long cycle” config-
uration, i.e. with large extraction of the thermocline region at the end of each charge
mances of the storage, is controllable through operational condi-
and discharge (Bruch et al., 2014a).
tions, i.e. the way the thermocline is extracted at the end of
cycles, which is represented by DT⁄1 and DT⁄2 in Figs. 10 and 11.
The more the thermocline is extracted at the end of cycles, the files. According to the authors, this little influence is due to varia-
greater the effective storage capacity (represented by the greyed tion of residence time and hence heat losses. Similarly, excepted
area) and the faster the stabilization in terms of both time and through the slight influence of heat losses, the temperature level
number of cycles. This behavior was also observed by Cascetta does not influence the stabilized state as long as physical proper-
et al. (2015), on an air/alumina packed-bed storage system and ties of the fluid remain of the same order of magnitude. In addition,
confirms the prediction of Bayón et al. (2013), as the stabilized the stabilized state is also unaffected by punctual perturbations,
thermal gradient is steeper when the thermocline region is more like partial load, introduced during or after stabilization: as soon
extracted at the end of cycles. However, as mentioned above, larger as operational conditions are restored, the storage continues its
extraction of the thermocline leads to colder and hotter outlet fluid stabilization or returns to the former stabilized state.
temperature at the end of discharging and charging processes, To conclude, operation strategy of packed-bed storage is a key
respectively. aspect which determines above all its performances and its inte-
Furthermore, Bruch et al. (2017), have shown with a liquid/solid gration into the desired process. As a consequence, the operator
system that the packed-bed storage behavior is mostly governed has to define a custom-made operation strategy for thermocline
by the thermocline extraction strategy either in charge, in dis- extraction depending on process requirements (Biencinto et al.,
charge or both. Under cyclic operating conditions, they showed 2014).
that the stabilized state is only little influenced by the fluid flow
rate and the temperature level. On the range considered by the
authors, the flow rate which has been increased up to a factor 2.3. Tank design
3.7 has only a slight influence on the stabilized temperature pro-
To tackle heat losses almost all packed-bed storage applications
are insulated. In the case of molten salts, heat-trace cables may be
used to make up for thermal losses and prevent salt freezing during
standby periods (Pacheco et al., 2002). Standby tests carried out by
several authors (Faas et al., 1986; Kuravi et al., 2013; McDonnell
Douglas Astronautics Company, 1986; Okello et al., 2014;
Pacheco et al., 2002; Shewen et al., 1978) show that heat losses
occur mainly at the top of the storage because of natural convec-
tion and higher temperature difference with ambient. Given that
this part of the storage contains the most useful part of the stored
heat, it should be insulated with particular attention.
Packed beds generally have a height and a diameter of the same
order of magnitude (Fig. 3), which increases compactness (i.e.
volume-to-surface ratio) of the storage and hence reduces thermal
losses (Brosseau et al., 2005; Zanganeh et al., 2012). At fixed fluid
flow, the cross section and the height of the tank also affect ther-
mal losses through the residence time of fluid.
Most packed-bed storages are cylindrical. This shape is rela-
tively free from mechanical problems (Melanson and Dixon,
1985), minimizes lateral surface area for a given cross section sur-
face and may improve flow uniformity by avoiding corner effects.
Fig. 10. End-cycle dimensionless temperature profiles in the ‘‘intermediate cycle”
Some authors have used rectangular cross section to implement
configuration, i.e. with little extraction of the thermocline region at the end of each structured packed bed (Kuravi et al., 2013) or to use low-cost
charge and discharge (Bruch et al., 2014a). materials as walls like wooden boards (Hollands et al., 1984;
 T. Esence et al. / Solar Energy 153 (2017) 628–654 637

Shewen et al., 1978). A truncated conical tank (with the largest at low HTF velocity while keeping a good heat transfer coefficient
cross section at the top) was studied by Zanganeh et al. (2012). This (HTC), which improves thermal stratification. Moreover, liquids
shape aimed to reduce thermomechanical stresses on the walls have a high viscosity compared to gasses, which leads to low Rey-
and heat losses by increasing the volume-to-surface ratio of the nolds number. As flow dispersion is low at small Reynolds num-
upper (i.e. the hottest) part of the tank. bers (Yang and Garimella, 2010b), stratification is generally
The aspect ratio of the packed-bed influences stratification and better in liquid/solid systems, thereby improving efficiency of the
hence efficiency of the storage (Haller et al., 2009) which may be storage. Besides, due to poor thermal properties of gasses, gas/solid
defined as the amount of energy recovered above a certain temper- systems require to operate at high flow rate, otherwise charging
ature divided by the amount of energy initially loaded. The ther- and discharging would be unacceptably long. However, this may
mocline occupies a certain height in the storage tank, thereby lead to non-negligible pressure losses and hence energy consump-
reducing its useful capacity. Since thermocline thickness increases tion due to high pumping costs of compressible fluids (Kuravi et al.,
more slowly than the tank height (at fixed solid filler size), a large 2013).
height-to-diameter ratio improves stratification and efficiency On the other hand, air is free, non-toxic and non-flammable,
(Cascetta et al., 2015; Mertens et al., 2014; Pacheco et al., 2002; which may significantly reduce installation costs and safety con-
Yang and Garimella, 2010b). A uniformly packed bed with large cerns. Moreover, air is chemically stable and can operate at very
height-to-diameter ratio also reduces flow heterogeneities since high temperature, thus increasing energy density of the storage
it acts as a flow straightener which evened out nonuniformities and efficiency of the electric conversion. The chemical compatibil-
introduced upstream (Szekely and Poveromo, 1975). ity between the fluid and the solids is usually less problematic with
However, a larger height-to-diameter ratio causes higher pres- air than with oils or molten salts.
sure losses which may affect the whole process efficiency
(Mertens et al., 2014). Furthermore, at industrial scale, the maxi-
2.5. Solid filler material
mum height of packed-bed tanks is practically limited to 16 m
for mechanical reasons (Libby, 2010; Pacheco et al., 2002). Thermal
Physical properties of some representative sensible heat storage
ratcheting has also to be considered in the dimensioning of the
solids are presented in Table 4. These values are estimated since
tank. Thermal ratcheting consists in plastic deformation of the tank
they depend on the quality and the origin of materials.
as a result of cyclic differential thermal expansion of the walls and
Packed bed can be made of either structured or non-structured
the solid filler with reorganization of the solids, mainly in non-
solid filler. Non-structured filler enables to use low cost solids like
structured packed beds (Flueckiger et al., 2013). In the long term,
pebbles. If a single size of spheroidal solids is used, void fraction of
this phenomenon can lead to the collapse of the tank (Dogangun
the bed is typically around 0.3–0.4 (Nellis and Klein, 2009). A void
et al., 2009). So far, thermal ratcheting of packed-bed storages
fraction around 0.25 can be achieved by using two sizes of particle,
has not been extensively investigated and would necessitate more
e.g. by mixing rocks and sand (Bruch et al., 2014a; Faas et al., 1986;
research effort. However, numerical studies show that a low
Hallet and Gervais, 1977; Pacheco et al., 2002). The decrease in
height-to-diameter ratio is likely to reduce wall stresses caused
void fraction enables to decrease the cost of liquid/solid systems
by thermal ratcheting (Charlas et al., 2010). By the way, the TES
since liquid HTF are often more expensive than the solids chosen
of Solar One, which is so far the only very large packed-bed storage
for thermal storage applications. In gas/solid systems, reducing
which has been built, had a height-to-diameter ratio less than one.
the void fraction enables to improve the storage density of the sys-
To sum up, the design of packed-bed storage tanks is a trade-off
tem since the volumetric heat capacity of solids is much higher
between mechanical, flow distribution, pressure drop, thermal loss
than the one of gases. However this method increases interstitial
and stratification issues. Generally speaking, a cylindrical tank
velocity of the fluid and pressure loss which may be restrictive
with a height-to-diameter ratio larger than one is often preferable
when using gaseous HTF. Manufactured materials like ceramic,
but each system should be specially adapted for the whole process
glass or alumina, usually of spherical shape, may be used in order
depending on technical requirements and economic constraints.
to run at very high temperature or to prevent fluid/solid chemical
interactions. Small solid size is preferable to improve stratification
2.4. Heat transfer fluid since it increases the total fluid/solid heat exchange area and
reduces the Biot number of solids (Van Lew et al., 2011; Yang
In most cases, liquid HTF consists either of thermal oil or molten and Garimella, 2010b). For stratification purpose, at fixed convec-
nitrate salts, while gaseous HTF consists of air (sometimes of flue tive heat transfer coefficient, the Biot number of the particles has
gas for high temperature tests). Physical properties at average to be as low as possible so that heat transfer is only governed by
operating temperature of some typical HTF are presented in convection (Adeyanju and Manohar, 2009), resulting in a sharper
Table 3. In particular, the volumetric heat capacity (qcp) enables thermal front and better stratification. This influence was illus-
to assess the heat storage density of materials, while the thermal trated by both numerical models (Durisch et al., 1986; Mertens
effusivity E enables to assess their ability to exchange heat. et al., 2014; Yang and Garimella, 2010b) and experimental studies
Liquid HTF have good heat capacity and thermal conductivity (Anderson et al., 2014). As long as fluidization is avoided, reducing
compared to gasses. This enables liquid/solid systems to operate particle size also improves flow uniformity by increasing pressure

Table 3
Physical properties of some usual HTF at average operating temperature.

Fluid Tmin/Tmax (°C) q (kg m
3) cp (J kg
1 K
1) k (W m
1 K
1) l (Pa s) qcp (kWh m
3 K
1) E (J K
1 m
2 s
1/2)

4
Caloria HT 43 0/315 695 2700 0.16 6.8  10 0.52 547
Therminol 66 0/345 845 2380 0.10 5.7  10
4 0.56 451
Jarytherm DBT 0/350 870 2350 0.11 4.7  10
4 0.57 469
Solar salt 220/600 1835 1510 0.52 1.8  10
3 0.77 1200
Hitec 142/535 1790 1560 0.33 1.8  10
3 0.78 960
Hitec XL 120/500 1990 1445 0.52 6.3  10
3 0.80 1224
Water 0/100 990 4180 0.64 5.8  10
4 1.15 1627
Air –/– 0.5 1075 0.05 3.4  10
5 1.5  10
4 5.3
638 T. Esence et al. / Solar Energy 153 (2017) 628–654

Table 4
Order of magnitude of physical properties of some sensible heat storage solids.

Solid q (kg m
3) cp (J kg
1 K
1) k (W m
1 K
1) qcp (kWh m
3 K
1) E (J K
1 m
2 s
1/2)
Quartzite 2600 850 5.5 0.61 3486
Basalt 2900 900 2.0 0.73 2285
Concrete 2200 850 1.5 0.52 1675
Bricks 3200 800 1.6 0.71 2024
Ceramic 3550 900 1.3 0.89 2038
Alumina 4000 900 11 1.0 6293

 
loss (Hollands et al., 1984) and by preventing flow channeling near @Tf @Tf
e  q f  c pf  þu ¼ h  as  ðTs
Tf Þ ð2Þ
the walls. In cylindrical beds, this latter phenomenon can be @t @z
avoided by respecting a minimum tank-to-particle diameter ratio
of 30–40 (Meier et al., 1991; Rose and Rizk, 1949). In the case of @Ts
rectangular cross section, the ratio between the smaller side of ð1
eÞ  qs  cps  ¼ h  as  ðTf
Ts Þ ð3Þ
@t
the tank and the solid diameter has to be greater than 50 to avoid
wall channeling (Hollands et al., 1984). In previous equations, as is the surface area of solid per unit bed
Structured filler material, like bricks or plates, may be used to volume, expressed in m2 m3. This shape factor can be determined
shape the bed. Although this is more expensive, this enables to as follows:
optimize the geometry of the bed in terms of heat exchange, con-
As;tot 6  ð1
eÞ
duction resistance of the solid and pressure losses. Low void frac- as ¼ ¼ ð4Þ
tion can be reached (e.g. 0.2 for Kuravi et al. (2013)) while Vb Deq;a
keeping acceptable pressure loss. Furthermore, structured filler The Schumann model is valid under some conditions presented
material enables to tackle with the near-wall channeling issue in Table 5. An analytical solution of this model was proposed by
and to prevent reorganization of solids over thermal cycles, which Schumann and presented in the form of graphs (Figs. 12 and 13).
is very likely to solve the great mechanical concern of thermal This analytical solution is valid when the volumetric heat capacity
ratcheting. For all these reasons, structured filler is particularly of the fluid is negligible compared to the one of the solids (i.e. for
suitable with gaseous HTF operated at high temperature. gas/solid systems), the solids are initially at a zero uniform temper-
ature and the material properties, the inlet fluid temperature and
3. Numerical models for packed-bed storage the fluid flow rate are assumed constant.
Shitzer and Levy (1983), simplified the writing of this analytical
Numerous numerical models for packed-bed storage have been solution and extended its validity to cases in which inlet fluid tem-
presented in the literature as summarized by Ismail and Stuginsky perature is time dependent.
(1999). All of them are derived from energy equations applied to Enhancement of the Schumann model can be performed by tak-
the components of the storage system and assume no mass ing into account additional phenomena.
exchange and no heat production inside the storage. The heat Thermal losses to the outside may be a significant phenomenon
exchange between the fluid and the solid is assumed proportional and should be taken into account mainly in poorly insulated sys-
to the average difference in temperature (Newton’s law). The phys- tems, for storage tanks with large surface-to-volume ratio, for pro-
ical properties of the materials may be considered constant and cess operation at high temperature or to simulate losses in energy
uniform or not depending on the studied system and the simplify- content during standby operation. Based on comparisons between
ing assumptions. The additional assumptions and deriving simpli- numerical and experimental results for laboratory-, prototype- and
fications (presented in Table 5) lead to various kinds of numerical commercial-scale packed beds, Hoffmann et al. (2016), showed
model. First, the models may be grouped into two main categories that heat losses have to be accounted for to simulate efficiently
depending on whether significant thermal gradient inside the laboratory-scale systems, while it is not always necessary for
solids is considered or not. Then, differentiation can be made prototype- and commercial-scale systems. Given that contact area
according to the number of dimensions (1D or 2D) and to the num- between the solids and the walls of the storage is usually small
ber of phases (fluid, solid, walls) accounted for. compared to the one between the fluid and the walls, this term
is often neglected in the equation of solid. However, with gaseous
3.1. Models with negligible thermal gradient inside solids HTF, heat transfer between the solids and the walls may become
non-negligible due to radiation (Klein et al., 2013), particularly at
This approach is used when the conduction resistance in the high temperature, i.e. more than 200–300 °C (Balakrishnan and
solids of the packed bed is negligible compared to the convection Pei, 1979b; Krupiczka, 1967; Kunii and Smith, 1960; Schröder
resistance (Saez and McCoy, 1982). This assumption is assessed et al., 2006). With liquid HTF, radiation can be neglected due to
thanks to the dimensionless Biot number (Bergman et al., 2011): the opacity of the fluid (Kunii and Smith, 1960).
Thermal losses can be expressed through an overall heat trans-
h  Vs =As
Bi ¼ 6 0:1 ð1Þ fer coefficient (U) which accounts for internal and external convec-
ks
tive thermal resistances, and conduction resistance of the walls. In
order to be expressed in a volumetric form, this coefficient has to
3.1.1. The Schumann model be coupled with the shape factor of the packed bed (ab) which cor-
The Schumann’s model is a reference for many authors. This responds to the heat exchange area between the bed and the out-
one-dimensional two-phase model of heat transfer in porous side divided by the bed volume (Ismail and Stuginsky, 1999).
media was originally developed by Schumann (1929). It consists A diffusion term including effective thermal conductivity which
in two energy equations, one for the fluid (2) and the other for enables to apply the Fourier’s law in the packed bed (Weidenfeld
the solid (3), linked by a convective heat exchange term. et al., 2004) may also be added to account for thermal destratifica-
 T. Esence et al. / Solar Energy 153 (2017) 628–654 639

(13), (14), (15) (16), (17), (18) conduction (19), (20), conduction (22), (23),
2D intraparticle

Ts/Tinlet
(24)

X
1D three-phase 2D three-phase 1D intraparticle

x
(21)

X
Fig. 12. Analytical solution of the Schumann model for the solid temperature
depending on the dimensionless time x and the dimensionless length y (Schumann,
1929).
X

Tf/Tinlet
X

X
X

In comparison with the

reference temperature
Perturbation model
(11), (12)

x
X

X
X

X
(otherwise only time and space

Fig. 13. Analytical solution of the Schumann model for the fluid temperature
depending on the dimensionless time x and the dimensionless length y (Schumann,
1D single-phase (liquid) (9)

1929).
If w calculated from (10)

tion. It should be noticed that the convection term in the original

Schumann model already leads to spreading of the thermocline
derivatives)

during circulation of the fluid due to non-infinite HTC (Hollands

et al., 1984). Diffusion is usually a second order phenomenon dur-
ing circulation of the fluid but becomes essential to simulate
X

X
X

degradation of the thermocline during standby periods (Gunn

(gaseous) (8)

and Vortmeyer, 1982; Handley and Heggs, 1969). Diffusion and

1D single-

effective thermal conductivity will be further discussed in

phase

Section 4.3.
X

X
X

The heat capacity of tank’s walls may also have a significant

influence, especially for laboratory-scale or pilot-scale systems
Schumann
Schumann Enhanced

(Bruch et al., 2014a; Hoffmann et al., 2016). The walls’ contribution

(6), (7)

can be accounted for through a modified solid density (q0s ) which

includes the walls’ heat capacity (Bruch et al., 2014a). This modi-
X

X
X

fied density may be expressed by (5).

(2), (3)

mw  cpw
q0s ¼ qs þ ð5Þ
Assumptions and simplifications for each model.

ð1
eÞ  cps  Vb
X

X
X

concentric temperature profile inside

the fluid compared to the one of the
Negligible thermal gradient inside the

Negligible volumetric heat capacity of

If heat losses (only through fluid phase), solid and fluid effective
Fluid and solid temperatures close to
derivatives in space close to each
Fluid and solid temperature second

conductivities, and walls’ heat capacity are added to the Schumann

Adiabatic storage with no thermal

Negligible diffusion compared to

1D system with fluid circulating

Uniform HTC around solids and

model, a more general two-phase one-dimensional model can be

Homogeneous porous medium

obtained and expressed by (6) and (7) with the modified or added
according to plug flow

influence of the walls

terms in bold characters.

Assumptions/conditions

   
@Tf @Tf › ›Tf
solids (Bi < 0.1)

e  qf  cpf  þu ¼  keff;f  þ h  as  ðTs
Tf Þ

@t @z ›z ›z
convection

each other

þ U  ab  ðT1
Tf Þ ð6Þ
1

solids
other

solid

 
cs cf Pe

@Ts › ›Ts
Table 5

Bi0

ð1
eÞ  q0s  cps  ¼  keff;s  þ h  as  ðTf
Ts Þ ð7Þ

@t ›z ›z
640 T. Esence et al. / Solar Energy 153 (2017) 628–654

3.1.2. One-dimensional single-phase models: simplification of the 3.1.3. Perturbation model

Schumann model The perturbation model was originally developed by Kuznetsov
Single-phase models, directly derived from the Schumann (1995), and further applied to packed-bed analysis by Votyakov
model, were mainly introduced by the work of Vortmeyer and and Bonanos (2014). This model comes from a Schumann model
Schaefer (1974), and consist in modeling the packed bed with a in which conduction in fluid and solid phases were added. The
single equation. The calculated temperatures are those of the solid underlying principle of the perturbation model is that, although
phase provided that required assumptions are perfectly respected different, temperatures of fluid and solid phases are considered
(Vortmeyer, 1989). close to each other compared to the reference temperature used
Two kinds of single-phase models involving different assump- to make the equations dimensionless. The solid phase temperature
tions and simplifications depending on whether the fluid is liquid can therefore be considered as a perturbation in the temperature
or gaseous may be considered. With gaseous HTF, it is reasonable field of the fluid phase. As a result, the fluid phase temperature
to assume that the internal energy of the fluid is negligible com- can be calculated thanks to the single dimensionless Eq. (11). Then,
pared to the one of the solids (qfcpf  qscps). Therefore, a temperature of the solid phase can be calculated with (12) from
Schumann-type model similar to (2) and (3) with additional diffu- the temperature of the fluid and characteristics of the packed
sion terms (with uniform effective thermal conductivities) but bed. To be valid, this model has to meet some conditions which
without the internal energy term of the fluid may be considered. mainly limit its application to liquid HTF (cf. Table 5).
As demonstrated by Vortmeyer and Schaefer (1974), based on this
@Tf @T @ 2 T
model a single-phase one-dimensional model (8) may be obtained þ cf  Pe  f ¼ K  2f ð11Þ
by only assuming that the second derivatives in space of the fluid @s @z @z
and the solid temperatures are equal (@ 2Ts/@ 2z = @ 2Tf/@ 2z), which
cs  cf  Pe @Tf
can be reasonably assumed in most packed bed storage systems Ts ¼ Tf þ dh ¼ Tf þ 0  ð12Þ
(Vortmeyer and Schaefer, 1974): Bi @z
The perturbation model is more precise than single phase mod-
@Ts @Ts
ð1
eÞ  qs  cps  þ e  qf  cpf  u  els and has an analytical solution in some conditions (constant
@t @z
" # inlet temperature and flow rate). A semi-analytical solution of this
ðe  qf  cpf  uÞ2 @Ts model was also developed for cases in which inlet fluid tempera-
¼ keff;s þ keff;f þ  ð8Þ
h  as @z ture and flow rate are time dependent (Mabrouk et al., 2015).

With liquid HTF, the heat capacity of the fluid cannot be 3.1.4. One-dimensional three-phase models
neglected. Vortmeyer and Schaefer (1974), therefore considered a In order to account more finely for the influence of the walls, a
Schumann-type model similar to (2) and (3) with additional diffu- dedicated energy equation may be added (Beasley and Clark, 1984;
sion terms (with uniform effective thermal conductivities). Based Hoffmann et al., 2016). This equation takes into account heat
on this model and by assuming that the fluid and the solid phases capacity and axial conduction in the tank’s walls. If this equation
have the same time and space derivatives (@Ts/@t = @Tf/@t and @Ts/@ is added to the general two-phase one-dimensional model pre-
z = @Tf/@z), the authors developed the following one-dimensional sented above ((6) and (7)), the new one-dimensional three-phase
model: model is expressed by (13)–(15). In (13) and (15), internal and
@Ts @Ts external wall heat transfer coefficients are associated with shape
½ð1
eÞ  qs  cps þ e  qf  cpf   þ e  qf  cpf  u  factors calculated respectively with internal and external surface
@t @z
"  2 # area of the walls (aw,int and aw,ext).
ð1
eÞ  qs  cps  w @Ts    
¼ keff;s þ keff;f þ  @Tf @Tf @ @Tf
h  as @z e  q f  c pf  þu ¼  keff;f  þ h  as  ðTs
Tf Þ
@t @z @z @z
ð1
eÞ  qs  cps  w  keff;s @ 3 Ts
þ  3 ð9Þ þ hw;int  ab  ðTw
Tf Þ ð13Þ
h  as @z
 
In Eq. (9), the last term is small and could be neglected since it is @Ts @ @Ts
ð1
eÞ  qs  cps  ¼  keff;s  þ h  as  ðTf
Ts Þ ð14Þ
a third derivative. The term w is the velocity of the thermal front in @t @z @z
m/s. As a consequence of the condition on the derivatives, w is
 
assumed the same for the fluid-phase and the solid-phase. Unless ›Tw @ ›Tw
qw  cpw  ¼  kw  þ hw;int  aw;int  ðTf
Tw Þ
measuring the thermal front velocity, it is possible to calculate it ›t @z ›z
from an overall energy balance:
þ hw;ext  aw;ext  ðT1
Tw Þ ð15Þ
e  qf  cpf
w¼ u ð10Þ
ð1
eÞ  qs  cps þ e  qf  cpf
3.1.5. Two-dimensional models
This method assumes thermal equilibrium between the fluid Consideration of non-negligible edge effects like flow hetero-
and the solid. If (10) is employed to calculate w, the model is there- geneities or radial thermal gradient due to significant lateral heat
fore restricted to packed beds in which the temperature of the fluid losses leads to the development of two-dimensional models.
and the solid are close to each other. Vortmeyer and Schaefer Parameters like temperature, fluid velocity or bed void fraction
(1974), pointed out that this model cannot be applied to steady- can vary according to radial direction. Two-dimensional models
state processes due to the fact that the part of the dispersion term also enable to consider anisotropic effective conductivity, particu-
due to liquid-solid heat transfer is related to the thermal front larly in structured packed beds.
velocity w. Given that heat exchange between the packed bed and the walls
It should be noticed that thermal destratification due to non- or outside occurs only at boundaries, this phenomenon is taken
infinite HTC is included in the diffusion term of single phase- into account through boundary conditions (usually Neumann
models. As a consequence, the additional simplifications of single boundary conditions). Although it is not always necessary, the
phase-models enable to reduce computational effort but don’t equation for the walls also may be two-dimensional (Klein et al.,
enable to dispense with determination of the HTC. 2013). A two-dimensional model similar to the one-dimensional
 T. Esence et al. / Solar Energy 153 (2017) 628–654 641

model presented in (13)–(15) and written with cylindrical coordi- around solids. These assumptions are expressed with Neumann
nates could be expressed by (16)–(18). In these equations, the boundary conditions.
radial component of the fluid velocity is assumed negligible. If thermal losses, axial conduction through the fluid phase and
    the influence of the walls are added to the original model of
@Tf @Tf @ @Tf
e  qf  cpf  þu ¼  keff;f;z  Handley and Heggs (1969), the resulting one-dimensional three-
@t @z @z @z phase model can be expressed by (19)–(21). In the solid equation
 
1 › ›Tf of this model, the fluid-solid heat exchange is accounted for thanks
þ  r  keff;f;r 
r ›r ›r to a Neumann boundary condition. Thermal diffusion through the
þ h  as  ðTs
Tf Þ ð16Þ solid phase of the packed bed is accounted for through the fluid-
phase equation by assuming that second space derivatives of the
    fluid and the solid temperature profiles are close to each other.
@Ts @ @Ts 1 › ›Ts
ð1
eÞ  qs  cps  ¼  keff;s;z  þ  r  keff;s;r  Therefore, the total effective thermal conductivity of the bed (keff)
@t @z @z r ›r ›r
is used in the fluid-phase equation (Wakao et al., 1979).
þ h  as  ðTf
Ts Þ ð17Þ
   
@Tf @Tf @ @Tf
    e  qf  cpf  þu ¼  keff  þ h  as  ðTs jrs ¼Rs
Tf Þ
@Tw @ @Tw 1 › ›Tw @t @z @z @z
qw  cpw  ¼  kw;z  þ  r  kw;r  ð18Þ
@t @z @z r ›r ›r þ hw;int  ab  ðTw
Tf Þ ð19Þ
More complete two-dimensional models may be implemented  
›Ts 1 › ›Ts
thanks to CFD commercial codes like Ansys Fluent (Chang et al., qs  cps  ¼ 2 r2s  ks  ð20Þ
2015; Yang and Garimella, 2010b). Thanks to these codes, continu- ›t rs ›rs ›rs
ity and momentum equations can be solved and some phenomena  
like compression work due to volume expansion/shrinkage, vis- @Tw @ @Tw
qw  cpw  ¼  kw  þ hw;int  aw;int  ðTf
Tw Þ
cous effects and kinetic energy changes can be accounted for, even @t @z @z
if negligible (Chang et al., 2015). þ hw;ext  aw;ext  ðT1
Tw Þ ð21Þ

3.2. Models with non-negligible thermal gradient inside the solids

3.2.2. Two-dimensional intraparticle conduction models
The condition (1) which enables to consider negligible thermal As far as the authors know, there is no two-dimensional
resistance inside solids compared to convective resistance may be intraparticle conduction model in the literature. However, if the
unfulfilled when the solids are too large, the solid conductivity is two-dimensional approach were applied to the intraparticle
too low or the HTC is too high. In this case, the temperature of conduction model described in (19)–(21), the resulting model
solids cannot be reasonably assumed uniform. The literature shows would be expressed as follows:
that two approaches may be used to face this issue: either calculat-    
@Tf @Tf @ @Tf
ing the temperature field inside the solids by using a devoted equa- e  qf  cpf  þu ¼  keff;z 
tion or using an effective HTC which accounts for conduction @t @z @z @z
 
resistance inside the solids without modifying the structure of 1 › ›Tf
þ  r  keff;r 
the models presented above. The former is described below, while r ›r ›r

the latter will be treated in Section 4.1.3. þ h  as  Ts jrs ¼Rs
Tf ð22Þ
 
3.2.1. One-dimensional intraparticle conduction models @Ts 1 @ @Ts
In this kind of model, which accounts for thermal gradient qs  cps  ¼ 2 r2s  ks  ð23Þ
@t rs @rs @rs
inside the solids and was introduced by Handley and Heggs
(1969), the solids are assumed to be spherical (with the same vol-    
@Tw @ @Tw kw › ›Tw
ume than real particles) and the packed bed is modeled as a super- qw  cpw  ¼  kw  þ  r ð24Þ
@t @z @z r ›r ›r
position of spheres (Fig. 14). In each elementary layer, the model
enables to compute the temperature of the fluid phase and the
temperature profile inside a representative solid. The overall 3.3. Comparison between models
behavior of the packed bed is considered to be one-dimensional.
Compared to Schumann-type models, two additional assumptions In the literature, a few comparisons of numerical models for
are necessary to implement this model: the thermal behavior of packed-bed storage have been carried out. The most complete
solids is concentric (central symmetry) and the HTC is uniform one was probably performed by Ismail and Stuginsky (1999). The
authors have compared various kinds of numerical model in terms
layer of calculation time and results.

n=1 3.3.1. Calculation time

n=2 Ismail and Stuginsky (1999), compared several numerical mod-
Rs rs els in terms of calculation time with the same number of time
n=3
increments and grid points, and the same initial and final condi-
tions for a packed bed comprised of stones with water as HTF.
Unless explicitly mentioned, every model takes into account heat
n=N losses and conduction in fluid and solid phases. The result obtained
by the authors is presented in Fig. 15.
Although this study is difficultly exploitable since operational
Fig. 14. Principle of the intraparticle conduction model. Reproduced from Meier conditions (temperature, flow rate, solid size, material properties,
et al. (1991). Biot number, etc.) which govern the scope of validity of each model
642 T. Esence et al. / Solar Energy 153 (2017) 628–654

In the two-phase model, the increase in solid size leads to a

Model CPU time slight decrease in the overall heat exchange between fluid and
solids (Fig. 16b). This is mainly due to the decrease in heat
1D 1-phase 1
exchange surface area between fluid and solids.
1D 2-phase without
4
In the intraparticle conduction model, the heat exchange reduc-
conduction tion caused by the increase in solid size is even more pronounced
1D 2-phase 7 because, in addition to the decrease in overall heat exchange sur-
face area, this model also accounts for the increasing conduction
2D 1-phase 20 resistance inside solids (Fig. 16c).
In this example, while the singe-phase model gives good results
1D intraparticle conduction 81 with small particle diameter, it is not relevant to assess the influ-
ence of the size of solids, because some involved phenomena are
2D 2-phase 180 neglected. However, this benchmark test should be considered
with precaution since the models should be compared within their
Fig. 15. Comparison of calculation times with various numerical models (basis 1 for scope of validity and flow rate and Biot number which would
one-dimensional single-phase model). Reproduced from Ismail and Stuginsky enable to check the validity of each model are not given.
(1999).
To sum up, various numerical models have been developed in
are not given, it enables to estimate the additional calculation time the literature. They come from the simplification of energy equa-
due to each phenomenon accounted for: tions which are applied to various elements of the storage system.
The numerical model should be carefully chosen according to the
– calculation time for two-phase models is 7–9 times longer than studied system, the assumptions which can be reasonably done,
for one-phase models; the phenomena investigated, the desired accuracy and the avail-
– calculation time for two-dimensional models is 20–25 times able computing power. Even if the results are representative of
longer than for one-dimensional models; the experimental reality, it should be kept in mind that many mod-
– accounting for thermal diffusion multiplies calculation time by els neglect some phenomena and are thereby limited to a particu-
1.7; lar scope of validity. In addition, experimental data from the
– accounting for thermal gradient inside solids multiplies calcula- literature on packed-bed storages are available only in a narrow
tion time by 11. range of operating conditions. Thus the scope of significant influ-
ence of each phenomenon and the scope of validity of each
assumption are not well known. Numerical models should there-
3.3.2. Parametric studies
fore be considered with precaution when they are used for extrap-
By definition, if the requirements for the applicability of a
olation or parametric studies.
numerical model are well fulfilled, the results obtained are close
to the experimental reality and similar to the results which would
be obtained with a less simplified and more general model. How- 4. Physical properties of porous media
ever, even if a simplified model gives good results, the estimation
through a simplified model of the influence of some phenomena As depicted in Fig. 17, many physical phenomena govern the
could be misleading. packed-bed storage behavior. To implement numerical models
As an example, Ismail and Stuginsky (1999), compared the and design setups, several parameters have to be determined:
influence of the size of solids in a packed bed comprised of stones fluid/solid and fluid/wall heat transfer coefficients, effective ther-
and water using three different models: a single-phase model mal conductivity and pressure loss of the bed. This section aims
which neglects convective thermal resistance and thermal gradient to present some models and correlations for these parameters.
inside the solids, a two-phase model which neglects thermal gradi-
ent inside the solids, and an intraparticle conduction model which 4.1. Fluid/solid heat transfer coefficient
accounts for the two above mentioned phenomena.
Due to the inherent assumptions of the single-phase model, this The fluid/solid HTC is the key parameter which governs the
model doesn’t exhibit any influence of the size of solids over the charge and discharge power and the thermal stratification of the
thermal behavior of the packed-bed storage (Fig. 16a). storage. This parameter may be expressed either in volume form

Fig. 16. Variation of the energy stored in terms of particle diameter for different bed models: (a) 1D 1-phase, (b) 2D 2-phase and (c) intraparticle conduction models; solid
and dotted lines refer to 25 mm and 50 mm diameter solids respectively (Ismail and Stuginsky, 1999).
 T. Esence et al. / Solar Energy 153 (2017) 628–654 643

Fig. 17. Diagram of some physical phenomena involved in the packed-bed storage behavior.

(hv in W m
3 K
1) or in classical surface form (h in W m
2 K
1). The third determination method is based on the Chilton-
These two forms are related by the shape factor as defined in Eq. Colburn analogy which links convective heat transfer and mass
(4) (Chandra and Willits, 1981): transfer phenomena through characterization of the boundary
layer. Several authors used this method (Acetis and Thodos,
hv As;tot
¼ as ¼ ð25Þ 1960; Achenbach, 1995; Beek, 1962; Gupta and Thodos, 1962;
h Vb Khan et al., 1991; Satterfield et al., 1954), but some of them
(Acetis and Thodos, 1960; Satterfield et al., 1954) obtained a heat
4.1.1. Determination methods transfer coefficient superior to the corresponding mass transfer
There are several ways to determine the HTC and the corre- coefficient, which is in contradiction with the Chilton-Colburn
sponding correlations. As pointed out by Balakrishnan and Pei analogy. According to Balakrishnan and Pei (1979a), this is due to
(1979a), the method used determines the heat transfer modes interaction of additional heat transfer modes and methodology
included in the HTC. When the methodology doesn’t enable to iso- issues. These biases were corrected by Gupta and Thodos (1962),
late fluid/solid convection from the other heat transfer modes (e.g. who obtained very similar heat and mass transfer coefficients.
conduction and radiation between solids), the determined HTC The developed correlation is therefore likely to be general.
corresponds to the total heat transfer rate. Given that total heat In addition to the three main methods mentioned above, other
transfer is influenced by physical and transport properties of the interesting approaches were developed in the literature. Galloway
bed materials, these correlations, although reliable, are only appli- and Sage (1970), used a boundary layer model based on the behav-
cable to the particular bed materials for which they were ior of heat and mass transfer from single particle and determined
developed. empirically some constants from local heat transfer measurements
The first determination method consists in using temperature in packed beds. Whitaker (1972) and Gunn (1978), used statistical
measurements and energy balance between the fluid and the method on numerous data from the literature to correlate the
solids. This method, which in principle correlates the total heat influence of several parameters. Ranz (1952) and Pfeffer (1964),
transfer, was introduced by Furnas (1930), and used by several estimated the heat transfer coefficient in a packed bed of spheres
authors (Achenbach, 1995; Inaba and Fukuda, 1984; Schröder from the heat transfer coefficient around an isolated sphere.
et al., 2006; Singh et al., 2006). Martin (2005), used the theory of Lévêque (1928), which correlates
The second determination method consists in comparing exper- the frictional pressure drop with heat or mass transfer, to build a
imental results with the ones of a numerical model. The HTC is fit- general correlation for heat transfer coefficient in packed beds
ted so that experimental and numerical results correspond. With comprised of particles of various shapes.
this method, the heat transfer modes included in the HTC depend
on the phenomena accounted for by the model: the transfer modes 4.1.2. Comparison of selected correlations
which are not considered separately in the model are actually Regarding to their determination method, some correlations of
included in the HTC. This method, introduced by Furnas (1932), the literature were selected and detailed in Table 6. They are
is often based on the Schumann model (Alanís et al., 1977; expressed in terms of dimensionless Nusselt number with the
Chandra and Willits, 1981; Handley and Heggs, 1968; Löf and nominal diameter of the solid particles (Nu) or with the diameter
Hawley, 1948; Morgan, 1980; Sagara and Nakahara, 1991). More of the equivalent sphere (Nueq) as characteristic length. Whatever
complete models were used by Coutier and Farber (1982) and the experimental conditions in which these correlations were
Beasley and Clark (1984). An intraparticle conduction model, i.e. developed and validated, it should be noticed that the column
one of the most complete models, was used by Wakao et al. ‘‘scope of validity” in Table 6 only indicates what have been explic-
(1979), to correlate numerous data from the literature. itly mentioned by the authors. The correlations are compared in
644 T. Esence et al. / Solar Energy 153 (2017) 628–654

Table 6
Correlations for fluid/solid HTC in packed beds.

Publications Correlations Scope of validity Determination methods

Ranz (1952) Nu ¼ 2 þ 1:8  Re 0:5
 Pr 1=3 Packed bed of uniform Correlation between HTC in packed
spheres beds of spheres (measured by
Gamson et al. (1943) and HTC
around a single sphere (correlation
of Frössling (1938))
Beek (1962) Nu ¼ 3:22  ðRe  PrÞ1=3 þ 0:117  Re0:8  Pr0:4 Re > 40 Correlation from mass transfer
experiments of Thoenes and
Kramers (1958)
Gupta and Nu ¼ RePr
1=3
 0:0108 þ Re0:58 0:929 Packed bed of uniform Correlation of numerous heat and
e
0:483
Thodos spheres mass transfer experimental data
(1962) Re > 20 from the literature
Pfeffer h i1=3 ‘‘High” Péclet number Theoretical model treating packed
1
ð1
eÞ5=3
Nu ¼ 1:26   Re  Pr
(1964) 2
3ð1
eÞ1=3 þ3ð1
eÞ5=3
2ð1
eÞ2 and ‘‘low” Reynolds beds as an assembly of
number (without independent particles associated
indication but with their own fluid layer
experimentally
validated with Re < 74)
Galloway 0:16 0:5 1=3 – Boundary layer model based on the
Nu ¼ 2 þ 0:5483  mm10 þ ½0:1212  at  ðat
0:04595Þ þ 0:001656  Re0:5  Pr1=6  Re Pr0:5
and Sage ðe
eb Þ established behavior of heat and
(1970) with at the apparent level of turbulence and eb the void fraction of stagnant boundary mass transfer from single cylinders
regions; these two parameters can be determined graphically in the original paper and are and spheres in turbulent fluid
both approximately equal to 0.2 for randomly packed beds with e = 0.4 streams. Constants determined
empirically from local heat transfer
measurements in packed beds of
spheres
Whitaker h i h i Non-structured packed Statistical analysis of numerous
1=3
Reeq;a 0:5 Reeq;a 2=3
Nueq;a ¼ ð1
eÞPre  0:5  1
e þ 0:2  1
e
(1972) beds and non-cubic experimental data from the
particle shape literature
h i
Gupta et al. 1=3
Nueq ¼ Pre  2:876 þ 0:3023  Re0:65 Reeq > 10 Statistical analysis of experimental
eq
(1974) data from the literature
Gunn 0:2 1=3
Nu ¼ ð7
10  e þ 5  e2 Þ  ð1 þ 0:7  Re  Pr Þ þ ð1:33
2:4  e þ 1:2  e2 Þ  Re  Pr0:7 1=3 Packed bed of uniform Statistical analysis of experimental
(1978) spheres data from the literature
Re < 105
0.35 < e < 1.0
Wakao et al. Nu ¼ 2 þ 1:1  Re0:6  Pr1=3 Only for models in Analysis of experimental data from
(1979) which the effective the literature treated and corrected
thermal conductivity of with an intraparticle conduction
the bed is specifically model
accounted for
Martin h i1=3 – Application of the theory of
Nu ¼ 0:4038  43  xf  Hg  Pr  e 2=3
(2005) ð1
eÞ Lévêque (1928), which correlates
with the frictional pressure drop with
0:4 0:78
Hg ¼ e  Re  ½150  ð1
eÞ þ 1:75  Re  ð1
3eÞ
e
heat or mass transfer
and xf the frictional fraction of total pressure drop depending on particles’ shape (0.45 for
spheres, 0.34 for saddles, 0.28 for rings, 0.25 for cylinders, 0.20 for cubes)

Fig. 18 on a case study with a packed bed of uniform spheres and enables to calculate the effective average temperature of solids.
thermal oil. For readability issues, only the most representative The effective HTC is calculated as:
of them are plotted. All the correlations of Table 6 which were
not plotted give results approximately situated in the envelop 1 1 Ds
¼ þ ð26Þ
formed by the plotted correlations. Fig. 18 shows that Nusselt heff h ks  k
number increases with Reynolds number. As a consequence, other The value of the coefficient k in (26) depends on the shape of the
things equal, convective heat transfer coefficient increases with solids and is 10, 8 and 6 for spheres, cylinders and slabs respec-
fluid velocity. However this increase is less than proportional. Sim- tively, with Ds the sphere or cylinder diameter or slab thickness
ilar evolutions of the Nusselt number with Reynolds number are (Dixon and Cresswell, 1979). Compared to analytical exact solu-
observed when case studies with molten salts or air at higher tem- tion, the effective HTC method gives very good results (Xu et al.,
perature are considered. Given that the correlation of Wakao et al. 2012a).
(1979), was developed on a large range of experimental data from
different setups, is easy to implement, widely used in the literature
4.2. Fluid/wall heat transfer coefficient
and in the average of the other correlations, it seems to be a rele-
vant choice to calculate the HTC in packed beds, at least in a first
In order to determine the influence of the walls on the packed
approach.
bed storage (thermal losses, heat capacity and axial conduction
through the walls), the HTC between the walls and the bed has
4.1.3. Effective heat transfer coefficient to be estimated. As shown by Yagi and Kunii (1960), several heat
As mentioned in Section 3.2, when the thermal gradient in the transfer modes which interact with each other govern the heat
solids cannot be reasonably neglected, it may be convenient to transfer near wall surfaces in packed beds: heat transfer through
use an effective HTC which accounts for conduction resistance of the thin fluid film near contact points, radiative heat transfer from
the solid. This method introduced by the work of Stuke (1948), solid surface to solid surface, thermal conduction in void spaces,
 T. Esence et al. / Solar Energy 153 (2017) 628–654 645

1000 1000 ®
®
Fluid: Therminol 66 Fluid: Therminol 66
Tf = 200 °C Tf = 200 °C
Ds = 0.01 m Db/Ds = 100
ρs = 2500 kg/m
3
ε = 0.4
ε = 0.4
100 100

Nuw
Nu

10 10
Ranz, 1952
Pfeffer, 1964 Yagi and Wakao, 1959
Galloway and Sage, 1970 Beek, 1962 (spheres)
Whitaker, 1972 Beek, 1962 (cylinders)
Gunn, 1978 Kunii and Suzuki, 1968
Wakao et al., 1979 Dixon et al., 1984
1 1
1 10 100 1000 10000 1 10 100 1000 10000
Re Re
Fig. 18. Influence of the Reynolds number on a given set of fluid/solid HTC Fig. 19. Influence of the Reynolds number on a given set of correlations of fluid/wall
correlations. HTC in packed beds.

radiative heat transfer from void to void (i.e. between solids not at higher temperature and lower tank-to-particle diameter ratio
directly in front of each other), heat transfer caused by the lateral are considered. Although Dixon et al. (1984), established their cor-
mixing of the flowing fluid in void spaces, heat transfer through the relation for low tank-to-particle diameter ratios, this correlation
boundary film. gives an average of several correlations from the literature over a
In the same way as for fluid/solid HTC, temperature measure- wide range of conditions (even at large tank-to-particle diameter
ments (Chu and Storrow, 1952; Colburn, 1931; Leva, 1947, 1950; ratios). This correlation could therefore be used to estimate the
Leva et al., 1948; Leva and Grummer, 1948; Yagi and Kunii, fluid/wall HTC in a first approach.
1960), Chilton-Colburn analogy (Beek, 1962; Dixon et al., 1984;
Kunii and Suzuki, 1968; Li and Finlayson, 1977; Yagi and Wakao, 4.3. Effective thermal conductivity
1959) or more theoretical approaches (Tsotsas and Schlünder,
1990) may be used to determine the HTC between the packed When the fluid is circulating through the packed bed, thermal
bed and the walls. diffusion is usually a second order phenomenon compared to con-
vective heat transfers. However, in some cases, especially when
4.2.1. Comparison of some correlations radiative heat transfer is high (in gas/solid packed beds at high
Some correlations from the literature were selected and pre- temperature), thermal diffusion may become non-negligible. Fur-
sented in Table 7. In all of them, the characteristic length for calcu- thermore, during standby periods, thermal diffusion is a first-
lation of the Nusselt number is the diameter of solids. As order phenomenon (with heat losses) which leads to destratifica-
previously mentioned, the column ‘‘scope of validity” only indi- tion of the storage. In these cases, thermal diffusion in the packed
cates what have been explicitly mentioned by the authors. The cor- bed, expressed as an effective thermal conductivity, should be esti-
relations are compared in a case study in Fig. 19. In spite of mated and taken into account.
discrepancies between the correlations, all of them show increas- The effective thermal conductivity of packed beds has been
ing wall Nusselt number with Reynolds number. The result of the investigated with various approaches by many authors and quite
comparison is similar when case studies with molten salt or air exhaustive reviews can be found in the literature (Kaviany, 1995;

Table 7
Correlations for fluid/wall HTC in packed beds.

Publications Correlations Scope of validity Determination methods

Yagi and Wakao 1=3
Nuw ¼ 0:6  Pr  Re 0:5 1 < Re < 40 Correlation from mass transfer experiments
(1959)
Yagi and Wakao 1=3
Nuw ¼ 0:2  Pr  Re 0:8 40 < Re < 2000 Correlation from mass transfer experiments
(1959)
Beek (1962) Nuw ¼ 0:203  Pr1=3  Re1=3 þ 0:220  Pr0:4  Re0:8 Packed bed of sphere-like Adaptation to the near wall region of correlations for fluid/solid HTC
solids
Re > 40
Beek (1962) Nuw ¼ 2:58  Pr1=3  Re1=3 þ 0:094  Pr0:4  Re0:8 Packed bed of cylinder-like Adaptation to the near wall region of correlations for fluid/solid HTC
solids
Re > 40
Kunii and Suzuki Nuw ¼ 0:062  Pr
1=3
 Re0:75 Packed bed of spheres Correlation from mass transfer experiments
ew
(1968) Re > 100
with e2w = 0.5 for packed beds of spheres
(Dixon, 1968)
Dixon et al. (1984) 1:5 3.0 < Db/Ds < 12 Correlation from mass transfer experiments
Nuw ¼ 1
1:5  DDbs  Pr1=3  Re0:59

Tsotsas and 0:5 Gaseous fluids and ‘‘high” Calculation of the thermal boundary layer considered equivalent to
Nuw ¼ 1:75  1
e
Schlünder (1990) e3  Re Péclet number the dynamic boundary layer
646 T. Esence et al. / Solar Energy 153 (2017) 628–654

Tsotsas and Martin, 1987; Van Antwerpen et al., 2010; Wakao and are accounted for. Breitbach and Barthels (1980) and Sih and
Kaguei, 1982). This part aims to point out some key aspects and to Barlow (1995), calculated the radiative contribution thanks to the
present some correlations for the estimation of thermal diffusion unit cell approach developed by Zehner and Schlünder (1970).
in packed beds. Integration of the effective thermal conductivity The contribution of conduction through contact surfaces kCeff is
into numerical models is also treated. negligible when ks/kf < 103 (Hsu et al., 1994), which is the case
for most packed-bed storage. However some correlations enable
4.3.1. Effective thermal conductivity contributions to estimate this contribution if needed. Bauer and Shlünder
The effective thermal conductivity (keff) of packed beds can be (1978), completed the ZS model by adding both radiative and con-
considered as the sum of three contributions (Bauer, 1990): the tact surfaces contributions. The resulting Zehner, Bauer and
first one accounts for the effective thermal conductivity of the Schlünder (ZBS) model, based on an empirical contact area fraction
packed bed with stagnant fluid (keff0,C,R), the second one accounts parameter, therefore enables to determine keff0,C,R. Hsu et al. (1994),
for the contribution of fluid mixing and turbulences (braiding also proposed a modification of the ZS model by including contact
effect) (kmix) which convey heat through the packed bed and simi- surfaces through an empirical deformation factor. Kaviany (1995),
larly the last one accounts for the motion of solids within the bed. developed a correlation to estimate the additive contribution of
The last contribution won’t be treated in this paper since the solids conduction through contact surfaces. While the radius of contact
are considered motionless in packed-bed storage systems. surfaces is calculated according to Hertzian elastic deformation,
The effective conductivity with stagnant fluid embodies four the number of contact surfaces is calculated by considering ran-
main heat transfer modes (Van Antwerpen et al., 2010): conduc- domly packed beds to be a combination of various packing
tion in solids, conduction in stagnant fluid, conduction through arrangements (simple cubic, body centered cubic or face centered
contact surfaces of adjacent solids and radiation between solids. cubic) depending on the void fraction. Hsu et al. (1995), developed
The contributions of conduction in each phase, of contact conduc- three different unit-cell models in which solids are considered to
tion and of radiation (respectively keff0, kCeff and kReff) may be added be square cylinders, circular cylinders or cubes with connecting
to determine the overall stagnant effective conductivity, keff0,C,R plate or connecting columns to account for conduction through
(IAEA, 2000): contact surfaces. The authors found that the cube model agrees
best with experimental data of a packed-sphere bed. Some of the

keff ¼ k0;C;R
eff þ kmix ¼ k0eff þ kCeff þ kReff þ kmix ð27Þ above mentioned correlations for contact conduction contribution
are relatively complex to implement and require many input
parameters which can be difficult to know and may introduce
4.3.2. Effective thermal conductivity with stagnant fluid additional uncertainties. That’s why only the model of Hsu et al.
Various correlations for effective conductivity of isotropic (1994), and the cubic model of Hsu et al. (1995), are developed
packed beds of spheroidal solids can be found in the literature. here. The other models give similar results and further information
The simplest ones neglect radiative contribution and contact about them can be found in the review of Van Antwerpen et al.
conduction (only keff0 is therefore considered) and enable to assess (2010), if needed.
the value or the bounds of the effective conductivity from the void All the above mentioned studies were developed for the core
fraction and the conductivity of the fluid and the solid. The effec- region of a packed bed and can be used in one-dimensional models
tive medium theory derived from the work of Maxwell (1873), for packed beds with negligible near-wall effects. For two-
enables to calculate these bound values when ks/kf 1. Less dimensional models, a correlation which accounts for radial varia-
restrictive bound values can be estimated from the correlations tion of the effective thermal conductivity due to near-wall effects
of Deissler and Boegli (1958), who considered the fluid and the should be used. Van Antwerpen et al. (2009), developed such a cor-
solids to be in parallel (upper bound) or in series (lower bound). relation which enables to account for radiative contribution and
A very simple correlation from Nield (1991), enables to roughly conduction through contact surfaces if needed.
assess the effective thermal conductivity of packed beds thanks The correlations developed in the above mentioned studies are
to a void fraction-weighted geometric average of the conductivity detailed in Table 8. The correlations which neglect radiative contri-
of the fluid and the solid. butions (i.e. which enable to determine keff0 or keff0,C) are compared
More complete models have been implemented to calculate the in Fig. 20. The correlations of Kunii and Smith (1960) (with
effective conductivity with stagnant fluid, keff0. A semi-empirical neglected radiative contribution) and Krupiczka (1967), are not
equation derived from numerical calculation of heat transfer plotted for readability purposes but they both give results close
through a bundle of cylinders was developed by Krupiczka to the ZS model. Experimental results collected in the literature
(1967), for the effective conductivity of a packed bed of spheres. by Kaviany (1995), are also reproduced. Fig. 20 shows that all the
Zehner and Schlünder (1970) (ZS), calculated the effective conduc- selected correlations and most of the experimental data are within
tivity of a unit cell comprised of fluid and solid and used results the bound values defined by Maxwell (1873) and Deissler and
from mass transfer experiments to correlate the influence of the Boegli (1958). At ks/kf ratio lower than 103, experimental data
void fraction. This so-called ZS model is a reference and is com- and all the selected correlations show similar trend. At ks/kf ratio
monly used in the literature. Hadley (1986), used the Maxwell larger than 103, the models which do not account for contact con-
upper bound and introduced empirically the influence of the void duction like the ones of Zehner and Schlünder (1970) and Gonzo
fraction from experiments on packed beds of disks or powder. Sim- (2002), show increasing discrepancy with experimental results.
ilarly, Gonzo (2002), empirically included the influence of both On the contrary, the correlations which account for contact con-
void fraction and material conductivities in Maxwell equations. duction like the ones of Hsu et al. (1994), and Hsu et al. (1995),
Radiative heat transfer is nonexistent when the HTF is liquid fit quite well with experimental data. It is interesting to note that
but may become non-negligible and should be taken into account the correlation of Hadley (1986), manages to reproduce experi-
with gasses, particularly at high temperature with large solids mental results at high ks/kf ratio while the contact conduction is
(Kunii and Smith, 1960). Several authors included radiative heat a priori not taken into account (since it derives from Maxwell
transfer contribution kReff in the effective conductivity of packed equations). This may be due to the very empirical approach which
bed. In the correlation of Kunii and Smith (1960), radiative heat captured the contact conduction contribution. As a consequence,
transfer between two neighboring solids and between two neigh- the relatively simple and well-known ZS model can be used for
boring voids (i.e. between solids not directly in front of each other) ks/kf ratio lower than 103, while correlations which account for
Table 8
Correlations of stagnant effective thermal conductivity in packed beds.

Publications Correlations Phenomena accounted for Determination methods/remarks

Maxwell (1873) k0eff 2eþkks ð3
2eÞ Conduction in both phases Effective medium theory
Lower bound: kf ¼ f
3
eþekks
f
2
ks
2ð1
eÞ þð1þ2eÞkks
k0eff kf
Upper bound: kf ¼ ð2þeÞkks þ1
e
f

Deissler and Boegli Parallel arrangement: k0eff ¼ e  kf þ ð1
eÞ  ks Conduction in both phases Fluid and solid phases in parallel or in series
(1958) h i
eÞ
1
Series arrangement: k0eff ¼ kef þ ð1
ks
Krupiczka (1967) 0:280
0:757logðeÞ
0:057log ks Conduction in both phases Semi-empirical equation derived from numerical
k0eff kf
kf ¼ ks
kf calculation of heat transfer through a bundle of cylinders
and a wide range of experimental data
The equation approximates 76% of 165 experimental
data of the literature with an error of less than ±30%
" #
Zehner and Schlünder pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi kf
B Conduction in both phases Theoretical unit cell model completed with results from
k0eff e   1
ks
(1970) kf ¼1
1
e þ 2 k1
2  ln ks
Bkf 2


Bþ1 B
1
k mass transfer experiments to correlate the influence of
f 1
ks B k
1
ksf B 1
ksf B
void fraction. The model is assumed valid for 0.2 < e < 0.6

T. Esence et al. / Solar Energy 153 (2017) 628–654

10=9
with the deformation parameter of solids, B ¼ Cð1
eeÞ and 10
3 < ks/kf < 104
k0 pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
If kksf  B ¼ 1 then kefff ¼ 1
1
e þ 1
e  1þ2B
3B
3
2
2
The shape parameter C is 1.25, 1.4 and 2.5 for spheres,
3ðB
1Þ
crushed solids and Raschig rings respectively. For
spheres, Hsu et al. (1994), proposed C = 1.364 with 1.055
instead of 10/9 for the exponent
2 3
Breitbach and Barthels pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi Conduction in solids and Adaptation of the ZS unit cell model by considering that
(1980) kReff ¼ 4ð1
1
eÞ  e þ 21


1
e  Bþ1 
B 1þ
1
1
5  4  r  T3  Ds radiation between solids base surfaces of the unit cell are black body surfaces
s
ð 2
Þ ks
s
1 4rT3 D
s B is the same as in the original ZS model
Hadley (1986) ks
2
Conduction in both phases Empirical enhancement of the Maxwell upper bound
ef 0 þkks ð1
ef 0 Þ 2 ð1
eÞþð1þ2eÞkks
k0eff kf
¼ ð1
a0 Þ  1
e f
þ a0  f
with the void fraction influence (experiments on packed
kf ð1
f 0 Þþkks  eð1
f 0 Þ ð2þeÞkks þ1
e
f f beds of disks or powder)
logða0 Þ ¼
4:898  e if 0 e 00827
logða0 Þ ¼
0:405
3:154  ðe
0:0827Þ if 0.0827 e 0.298
logða0 Þ ¼
1:084
6:778  ðe
0:298Þ if 0.298 e 0.580
f 0 ¼ 0:8 þ 0:1  e
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi
Hsu et al. (1994) 0;C
keff Conduction in both phases Enhancement of the ZS unit cell model by addition of
kf ¼ ð1
1
e Þ þ kf 
ks
1
e 1
1
2 þ kf 2 1
ekf
ð1þaBÞ 1
ks Bþ 1
ks aB and conduction through contact surfaces between solids which are taken into
"  #
k
1
ksf ð1þaÞB ðB
1Þ
contact surfaces between account through the empirical deformed factor a
   2  ln 1þaB
k
Bþ1þ2aB2
 k
 k

solids
k
1
ksf Bþ 1
ksf aB
k ð1þaÞBksf 2ð1þaBÞ 1
ksf Bþ 1
ksf aB ð1þaBÞ

with a ¼ 0:002 and B so that:

h h i i2
1
e
B2
 ðB2
4  B þ 3Þ þ 2  ð1 þ aÞ  ð1 þ a  BÞ  ln ð1þ aÞB 2
1þaB þ a  ðB
1Þ  ðB
2B
1Þ ¼0
ð1
BÞ ð1þaBÞ2
6

Hsu et al. (1995) 0;C

keff c2a
c2c c2a   2ðcc ca
cc c2a Þ  Conduction in both phases Theoretical unit cell model in which solids are assumed
kf ¼ 1
c2a
2  cc  ca þ 2  cc  c2a þ c2c  c2a  kksf þ  k þ k
1
ca þca ksf 1
cc ca þcc ca ksf and conduction through cubic and contact surfaces are modeled with square
with cc ¼ 0:13 and ca so that: ð1
3:c2c Þ  c3a þ 3  c2c  c2a
ð1
eÞ ¼ 0 contact surfaces between connecting column, the relative dimensions of which (ca
solids and cc) are empirically determined
 
Sih and Barlow (1995) pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi Conduction in solids and Addition in the ZS model of radiation term from
kReff ¼ ½ð1
1
eÞ  e þ 1
e  4  2
: r  T3  Ds
s
0:264
2
radiation between solids Damköhler (1937), with the view factor calculated
according to Wakao and Kato (1969)
Gonzo (2002) k0eff
¼ 1þ2bð1
eÞþð2b
0:1bÞð1
eÞ2 þð1
eÞ3 0:05expð4:5bÞ
3
Conduction in both phases Empirical enhancement of the Maxwell equation with
kf 1
bð1
eÞ

kf
the void fraction and the conductivity influences. The
with b ¼ kks sþ2k f model is assumed valid for 0.15 < e < 0.85 and 10
3 < ks/
kf < 104. Compared to 35 experimental data of the
literature, the mean deviation is 10.6% (+55%/
23%)

647
648 T. Esence et al. / Solar Energy 153 (2017) 628–654

ε = 0.4
100

10
λeff 0,(C)/λf

Experimental data collected by Kaviany, 1995

Maxwell, 1873 (lower bound)
1
Maxwell, 1873 (upper bound)
Deissler and Boegli, 1958 (serial)
Deissler and Boegli, 1958 (parallel)
Zehner and Schlünder, 1970 (ZS)
Hadley, 1986
Gonzo, 2002
Hsu et al., 1994
Hsu et al., 1995, (cube)
0.1
0.1 1 10 100 1000 10000
λs/λ f

Fig. 20. Comparison of correlations of stagnant effective thermal conductivity of packed-sphere bed without radiative contribution (contact conduction is included or not
depending on the correlation).

contact conduction should be used for larger ks/kf ratio. In typical 4.5
rock beds, ks/kf ratio ranges from unity to several tens with liquids Reference case
and from several tens to several hundred with air. 4 λs = 100 W·m-1·K-1
s = 0.5
The relative contribution of radiative heat transfer in various Fluid: air
3.5 s = 1.0
conditions is illustrated in Fig. 21. Various cases in which each ε = 0.3
Ds = 1 mm
ε = 0.4
parameter is varied one by one (other things being equal) are plot- ε = 0.5 λs = 2 W·m-1·K-1
3
λeff0,R/λeff0

ted and compared to a reference case. The comparison is carried Ds = 0.1 mm s = 0.8

out with the correlation of Breitbach and Barthels (1980). The cor- Ds = 5 mm
2.5
Ds = 10 mm
relations of Sih and Barlow (1995) or Kunii and Smith (1960), show
similar trends (excepted that the latter is insensible to solid con- 2
ductivity). The figure shows that radiative contribution depends
1.5
mainly on solid diameter and temperature. The bed void fraction,
the solid conductivity and the solid emissivity also influence the 1
radiative contribution but their influence is smaller since they
are unlikely to vary a lot from one packed bed to the other. 0.5
0 200 400 600 800 1000 1200
T [°C]
4.3.3. Contribution of fluid mixing and turbulences
The contribution of fluid mixing kmix can be added to the stag- Fig. 21. Influence of various parameters on the relative contribution of radiative
nant effective conductivity keff0,C,R. This contribution is not the heat transfer according to the correlation of Breitbach and Barthels (1980).
same in axial and radial directions (respectively parallel and per-
pendicular to flow direction) and is given by Wakao and Kaguei in the near-wall region is not clear but it seems that Eq. (28) can be
(1982): applied over the whole cross section (Tsotsas and Schlünder,
1990).
kmix;z
¼ 0:5  Re  Pr ð28Þ According to Eqs. (28) and (29), effective conductivity of the
kf
mixing contribution can be as high as several dozens of times the
fluid conductivity in some packed-bed storage systems. However,
kmix;r this contribution exists only when the fluid is circulating, i.e. when
¼ 0:1  Re  Pr ð29Þ
kf thermal diffusion is usually a second-order phenomenon com-
In two-dimensional models, the inhibition of radial mixing due pared to the fluid/solid convective heat transfer. Furthermore, the
to the wall should be considered in the near wall-region (i.e. at less mixing contribution may be artificially incorporated in the fluid/-
than one particle radius from the wall). This inhibition is due to solid HTC when very empirical correlations for fluid/solid HTC
lower fluid velocity caused by the higher local void fraction, the are used (Balakrishnan and Pei, 1979a). That’s why the mixing con-
permanent barrier to fluid movement which discourages radial dis- tribution is neglected in many numerical models for packed-bed
placement of fluid parcels toward the wall and the viscous bound- storage simulation.
ary layer on the surface of the wall. As a consequence, radial
mixing in the near-wall region may be considered to be ten times 4.3.4. Integration of effective thermal conductivity in models
smaller than in the core region and a coefficient of 0.01 instead of Correlations of effective thermal conductivity enable to deter-
0.1 can be used in Eq. (29) (Baddour and Yoon, 1961). Axial mixing mine the conductivity of the whole packed bed. This is therefore
 T. Esence et al. / Solar Energy 153 (2017) 628–654 649

suitable for intraparticle conduction models (Eqs. (19) and (22)) The two above mentioned methods for integration of the effec-
which require only the whole effective conductivity keff. This is also tive thermal conductivity must lead to similar results since heat
appropriate for single-phase models (Eqs. (8) and (9)), in which the exchange between the fluid and the solids generally prevails over
second space derivatives of the fluid and the solid temperatures are heat exchange inside each phase (i.e. diffusion) as assumed by
considered equal, since the conductivity of the whole media can be Schumann (1929). However this point should be validated with
reasonably considered to be the sum of the effective conductivity dedicated simulations.
of the fluid phase keff,f and the effective conductivity of the solid
phase keff,s: 4.4. Pressure losses

keff ¼ keff;f þ keff;s ð30Þ

Fluid flow through a packed bed induces pressure losses. It
For two-phase models, two methods can be applied. Since sec- could be useful to calculate pressure losses to size the installation
ond space derivatives of the fluid and the solid temperatures can or to estimate pumping costs. This issue should be carefully
be assumed close to each other, the first method simply consists addressed when gas is used as HTF since pumping costs of com-
in including a single diffusion term, either in the fluid-phase or pressible fluids are much higher than for incompressible fluids
the solid-phase equation with the whole effective thermal conduc- (i.e. liquids) and may affect the overall energy efficiency of the
tivity keff of the bed (Gunn and Vortmeyer, 1982). This approach is storage system.
similar to what proposed Wakao et al. (1979), for intraparticle con-
duction models (cf. Section 3.2). 4.4.1. The Ergun equation
The second method consists in sharing out the effective thermal The Ergun equation (Ergun, 1952) is a reference for calculation
conductivity contributions between the fluid and the solid phases. of pressure losses through packed beds. This equation associates
If conduction through contact surfaces and radiative contribution the Kozeny-Carman equation which was developed by Carman
are accounted for, they should be included in the solid effective (1937), from the work of Kozeny (1927), for lowly to moderately
conductivity since these contributions depend on the solid thermal porous media, and the Burke-Plummer equation (Burke and
gradient. Conversely, the contribution of fluid mixing kmix has to be Plummer, 1928) which was developed for highly porous media.
included in the fluid effective conductivity. Single conduction in The Ergun equation is the sum of a viscous term, corresponding
each phase modeled by keff0 is more difficult to separate since it to the Kozeny-Carman equation, and an inertial term, correspond-
accounts for the combination of conduction in solid and fluid ing to the Burke-Plummer equation. They are respectively repre-
phases. As shown by Cheng and Hsu (1998), the effective conduc- sented by the left-hand term and the right-hand term in (36).
tivity due to single conduction in each phase can be expressed as: Each term corresponds to an asymptotic behavior of the Ergun
equation and their combination enables to treat intermediate
k0eff;f ¼ e  kf þ f  kf ð31Þ cases.

DP ð1
eÞ2 lf  usup ð1
eÞ qf  u2sup

k0eff;s ¼ ð1
eÞ  ks
f  ks ð32Þ ¼A  þB  ð36Þ
L e3
Ds2 e3 Ds
The coefficient f accounts for the tortuosity of the bed and
A and B are empirical constants determined by Ergun:
depends on many parameters. In Fig. 20, this coefficient represents
the gap between the actual values of thermal conductivity and the A ¼ 150 B ¼ 1:75
values of the correlation of Deissler and Boegli (1958), with parallel
arrangement (upper bound). This term can be used to share out the
whole effective conductivity between the two phases. Since Eq. 4.4.2. Enhancement of the Ergun equation
(30) can be reasonably assumed true even in two-phase models, For packed beds of spherical solids, the validity of the Ergun
then: equation has been confirmed by several authors (Lakota et al.,
2002; Nemec and Levec, 2005). However numerous authors dis-
ðk0eff;f þ k0eff;s Þ
e  kf
ð1
eÞ  ks puted the value or the constancy of parameters A and B, particu-
f¼ larly when the solids are not spherical. While several authors
kf
ks
fitted the value of the constants with experimental results
k0eff
e  kf
ð1
eÞ  ks (Handley and Heggs, 1968; Lakota et al., 2002; Yu et al., 2002),
ffi ð33Þ
kf
ks others give correlations to calculate A and B according to various
As a result, the effective conductivity of each phase may be parameters (Achenbach, 1995; Foumeny et al., 1996; Hicks,
expressed as follows: 1970; Macdonald et al., 1979; Nemec and Levec, 2005). Particu-
larly, the inertial parameter B seems to depend on the shape of
keff;f ffi ½e  kf þ f  kf  þ kmix ð34Þ the solids, since it accounts for pressure losses due to the tortuosity
of the flow (Handley and Heggs, 1968; Lakota et al., 2002). This
affirmation was strengthened by several experimental results
keff;s ffi ½ð1
eÞ  ks
f  ks  þ kReff þ kCeff ð35Þ
(Achenbach, 1995; Allen et al., 2015; Foumeny et al., 1996;
The coefficient f is determined thanks to Eq. (33) and correla- Macdonald et al., 1979). Du Plessis and Woudberg (2008), theoret-
tions of keff0 like the ZS model. The mixing contribution kmix may ically validated the empirical Ergun equation for packed beds of
be calculated with Eqs. (28) or (29) according to the direction con- uniform spherical granules with a unit cell model. This model is
sidered. The radiative contribution kReff may be determined thanks therefore applicable over the entire void fraction range. Some cor-
to the correlations of Sih and Barlow (1995) or Breitbach and relations were specifically developed for rock beds (Allen et al.,
Barthels (1980). The single contribution of contact conduction 2015; Chandra and Willits, 1981; Dunkle and Ellul, 1972;
may be determined thanks to the correlation from Kaviany Hollands and Sullivan, 1984) but they may be limited to the exper-
(1995), or by subtracting keff0 from keff0,C (determined with correla- imental conditions for which they were established.
tions like the ones of Bauer and Shlünder (1978), Hsu et al. (1994) In accordance with what proposed Ergun (1952), when the
or Hsu et al. (1995)). It should be noticed that radiative contribu- solids are not spherical and/or not uniform in size, most authors
tion kReff in Eq. (35) disappears if liquid HTF are considered. implement the Ergun equation and calculate the Reynolds number
650 T. Esence et al. / Solar Energy 153 (2017) 628–654

with the diameter of the sphere of equivalent specific area Deq,a as

expressed by Eq. (37) (Allen et al., 2015; Du Plessis and Woudberg,
2008; Foumeny et al., 1996; Hollands and Pott, 1971; Lakota et al.,
2002; Macdonald et al., 1979; Martin, 2005; Nemec and Levec,
2005; Rose and Rizk, 1949). However, for irregular solids like grav-
els, the external surface area is difficult to measure. That’s why

Correction factor compared to infinite bed

some authors use the diameter of the sphere of equivalent volume
Deq,V which is easier to determine and is expressed by Eq. (38)
(Allen et al., 2015; Chandra and Willits, 1981; Dunkle and Ellul,
1972; Singh et al., 2006). An alternative approach expressed by
Eq. (39) and based on masses of solids was proposed by Hollands
and Sullivan (1984). According to the authors, this representative
diameter D0eq;a is similar to the diameter of equivalent specific area
defined in (37) but altered from that one to simplify its determina-
tion. Ideally, the sums in Eqs. (37)–(39) should be taken over all
solids in the bed, but may be taken only over a representative num-
ber N (Hollands and Pott, 1971).
PN
Vs;i
Deq;a ¼ 6  PiN ð37Þ
i As;i

PN !1=3
6 Vs;i
Deq;V ¼  i
ð38Þ
p N

 1=3 PN Db/Ds
6 1 ms;i
D0eq;a ¼   PNi 2=3 ð39Þ
p qs i ms;i Fig. 22. Correction factor to account for near-wall effects on pressure losses
according to the tank-to-particle diameter ratio and the particle Reynolds number.
An enhanced Ergun-type equation was proposed by Bruch et al. Established for packed bed of spheres but valid for other particle shapes according
(2014a), for packed beds with two very different sizes of solids. For to the authors (Rose and Rizk, 1949).
example, when a packed bed is comprised of stones with void
spaces completely filled with sand, they assumed that both the
stones and the sand influence the interstitial velocity (usup/e) of
the storage utilization due to friction caused by thermal expansion
the fluid by reduction of the available cross section but only the
and contraction of the solids.
sand is directly responsible for pressure losses. As a consequence,
Allen et al. (2015), showed the influence of the direction in
the interstitial velocity is calculated with the global void fraction
which the rocks were poured into the test section relative to the
(eglobal) of the bed comprised of stones and sand, while the void
fluid flow direction. When rocks are not perfectly spherical, ellip-
fraction of the sand only (esand) is used in the Ergun equation. With-
soidal rocks tend to get into position parallel to the bottom of
out changing the original values of A and B, the resulting Eq. (40)
the tank. As a consequence, if the flow direction is the same as
predicted experimental pressure losses of a packed bed of 3-mm
the one in which the rocks were poured (co/counter-current), pres-
sand and 30-mm rocks with an error less than +5%/
15%.
sure losses may be as 80% higher as if a perpendicular flow direc-
   2 tion is adopted (cross-current). Moreover, the experimental data of
DP ð1
esand Þ2 lf usup ð1
esand Þ qf usup
¼A   þ B    the authors tend to show that rearrangement of the rocks due to
L e2sand D2sand eglobal esand Dsand eglobal
their expansion/shrinkage caused by thermal solicitations can irre-
ð40Þ versibly increase pressure losses of the bed (by about 15–20%).
Ergun-type equations don’t enable to account for near-wall
effects due to higher local void fraction close to the walls, and 4.4.4. Comparison of some correlations
which may affect overall pressure losses of the bed. In order to Some correlations of pressure drop in packed beds are detailed
account for near-wall effects, Rose and Rizk (1949), constructed a in Table 9. To simplify the writing, they are given in the form of
chart (Fig. 22) to determine the correction factor which should friction factor, fv, as defined by Ergun (1952). The column ‘‘scope
be applied to pressure losses of an infinite bed according to the of validity” only indicates what have been explicitly mentioned
tank-to-particle diameter ratio and the Reynolds number. The by the authors. Some of the selected correlations are plotted versus
influence of near-wall effects on pressure drops was also pointed the Reynolds number in Fig. 23. The figure shows that the Ergun
out by Foumeny et al. (1996), while Nemec and Levec (2005), con- equation and the correlation of Du Plessis and Woudberg (2008),
sidered it to be negligible when the tank-to-particle diameter ratio (which aims to generalize the Ergun equation), give very similar
is larger than 10, which was already suggested by Ergun (1952). results. The correlations of Achenbach (1995), Lakota et al.
(2002), for spheres and Yu et al. (2002), (which were not plotted
4.4.3. Particular considerations for rock beds for readability purposes) also give similar results. Almost all the
Hollands and Sullivan (1984), pointed out the influence of dust other correlations which were developed for non-spherical solids
on pressure losses in air/rock beds: with unwashed gravels, they result in higher friction factor than the Ergun equation. The corre-
measured pressure losses more than twice as high as pressure lation of Chandra and Willits (1981), is not plotted but gives results
losses with washed gravels. The presence of dust should therefore very similar to the correlation of Dunkle and Ellul (1972). Correla-
be taken into account in air/rock beds and could be responsible for tions given by Lakota et al. (2002), for porous cylinders and Raschig
some large scattering in experimental results of the literature. In rings give friction factor between the one for porous extrudates
addition, even with washed gravels, dust may be generated over and the Ergun equation.
 T. Esence et al. / Solar Energy 153 (2017) 628–654 651

100000
Combination of the Kozeny-Carman and the Burke-Plummer equations with empirical determination of the constants A
Ergun, 1952

Mean deviation between 5.3% and 10.5% for spheres, and 4.4% and 15.3% for the other shapes compared to experimental

On their experimental data, the coefficient of determination (R2) is 0.93 for the co/counter-current configuration and 0.96
Experimental data with air and washed river gravel and crushed granite (0.38 < e < 0.46 and 9.9 mm < Deq,V < 26.9 mm)
Du Plessis and Woudberg, 2008

Experimental data with air and spheres, cylinders, extrudates, and Raschig rings (0.375 < e < 0.580 and 3 mm < Deq,a
Dunkle and Ellul, 1972
Mcdonald et al., 1979 (smooth solids)

Modeling of the bed with representative unit cells composed of fluid volumes occupied by a cubic solid particle
Prediction of experimental results with an accuracy of ±50% for a wide variety of unconsolidated porous media
Experimental data with air and spheres, cylinders, sand, and fibers (0.123 < e < 0.919 and 18.3  10
3 mm < Deq,a
Mcdonald et al., 1979 (rough solids)

Crushed rocks and Experimental data with air and crushed rock in co/counter- or cross-current configuration (0.38 < e < 0.46 and
Lakota et al., 2002 (porous extrudates)
10000 Allen et al., 2015 (co/counter-c)
and B from 640 experiments with various-sized spheres, sand, pulverized coke, and CO2, N2, CH4 and H2

fv

1000
Determination methods/remarks

ε = 0.4
100
Mean deviation of 8% between the correlation and experimental data

1 10 100 1000 10000

Re

Fig. 23. Comparison of several correlations of pressure losses in packed beds.

In Fig. 23, it should be noticed that the correlations are com-

pared only for a given void fraction (e = 0.4). According to the Ergun
equation, pressure drop increases with the decrease in void frac-
tion. This trend is represented by all correlations with varying
degree except for the one of Dunkle and Ellul (1972), which is
for the cross-current configuration
Experimental data on air/rock bed

not related at all with void fraction, which is unlikely to be physi-

cally correct.
To sum up, pressure losses can be estimated thanks to the Ergun
9 mm < Deq,V < 49 mm)

equation in a first approach for spherical granules of uniform size

with ‘‘standard” void fractions (0.35 < e < 0.5). When the shape of
solids is very irregular the correlations of Macdonald et al.
(1979), which were tested over a wide range of shape and void
< 65.8 mm)

fraction may be used. Finally, when solids of two very different

< 6 mm)

sizes are used, the approach developed by Bruch et al. (2014a),

data

should be considered.
spherical granules
Scope of validity

buoyancy effects
Uniformly sized
1 < Reeq,V < 1000
Unconsolidated
1 < Reeq,V < 105

porous media

5. Conclusion
0.3 < e < 0.5
Rock beds

negligible

In the short to medium term, packed-bed energy storage is a

–

promising technology to enhance the development of solar ther-

Correlations of pressure drop in packed beds expressed in the form of friction factor.

Allen et al. (2015) Co/counter-current (cf. 4.4.3) with 9 mm < Deq,V < 49 mm:

mal energy. However, few studies have been carried out on setups
þ 1:9   e 2 2  1
eq;a
e
Re

Cross-current (cf. 4.4.3) with 9 mm < Deq,V < 30 mm:

large enough to be fully representative of commercial-scale sys-

h

tems in which edge effects are second order phenomena. The stud-
i

2 1
ð1
eÞ3
i

Porous extrudates: f v ¼ 150 þ 7:4  1
eq;a

Reeq;a

e
e0:6 þ e0:6  1
e

i
Reeq;a

ies performed so far showed that thermal stratification of the

e0:6 þ e0:6  1
e

Reeq;a

2
Re
1
e

i
i

i
i

Raschig rings: f v ¼ 150 þ 6  1
eq;a

Reeq;V

 1
eq;V

 1
eq;V
e
eq;V
h

h
1
e

e
eÞ  1
e

storage system is a key parameter which strongly influences stor-

Reeq;a
h

1
e

Re
h

Re

Re
Re
1:8

Porous cylinders: f v ¼ 150 þ 3 

h
h

4:0

i0:08 h

i0:14 h
h

age efficiency and operation. Thermal stratification of the storage

i

h
e 
f v ¼ 150 þ 1:75  1
eq;a
e

Spheres: f v ¼ 150 þ 1:75 

eÞ

0:4
Re

3
þ 21  ð1

depends on many parameters like the aspect ratio of the storage

For ‘‘smooth” solids: f v ¼ 180

f v ¼ 185  e 2 þ 1:7  ð1

For ‘‘rough” solids: f v ¼ 180
e
h
Correlations

eq;V

eq;V

tank, the fluid and solid physical properties, the solid filler geome-
1
e

1
e
2 2

f v ¼ 310 þ 6:85  Re

f v ¼ 300 þ 6:15  Re
ð1
eÞ3  1
ð1
eÞ3  1
ð1
eÞ3

try and the charging and discharging operating conditions. Exper-

h

h
ð1
eÞ2

imental and numerical studies showed that long-term operation of

ð1
eÞ
3

0:4
f v ¼ 1750  e

1 

packed-bed storage under cyclic conditions leads to a stabilized

e3

state which can be controlled by the operator and is able to with-

stand punctual perturbations. Packed-bed storages may work with
2 

either liquid or gaseous heat transfer fluid. Liquid heat transfer flu-
ids have good physical properties and enable to reach high effi-
f v ¼ 25:4 

ciency whereas air used as gaseous heat transfer fluids is free,

enables to operate at very high temperature and is free from chem-
ical issues.
Willits (1981)
Dunkle and Ellul

Macdonald et al.

Since packed-bed storage performances depend on operating

Publications

Du Plessis and
Woudberg
Ergun (1952)

conditions and many interacting physical phenomena, numerical

Chandra and

Lakota et al.
(1972)

(1979)

modeling is a good way to investigate and optimize packed-bed

(2002)

(2008)

storage design and operation. Several numerical models from the

Table 9

literature have been presented. All of them are based on energy

equations applied to the components of the system. Numerical
652 T. Esence et al. / Solar Energy 153 (2017) 628–654

models differ mainly as a result of the assumptions made and the Bruch, A., Fourmigué, J.F., Couturier, R., Molina, S., 2014b. Experimental and
numerical investigation of stability of packed bed thermal energy storage for
degree of simplification. Assumptions and simplifications should
CSP power plant, SolarPACES 2013. Energy Proc. 49, 743–751.
be a trade-off between the desired accuracy and the available com- Bruch, A., Molina, S., Esence, T., Fourmigué, J.F., Couturier, R., 2017. Experimental
puting power, and depend on the studied system. Numerical mod- investigation of cycling behaviour of pilot-scale thermal oil packed-bed thermal
els should be considered with precaution when they are used for storage system. Renew. Energy 103, 277–285.
Burke, S.P., Plummer, W.B., 1928. Gas flow through packed columns. J. Ind. Eng.
extrapolation and parametric studies since the scope of validity Chem. 20, 1197–1200.
of each simplifying assumption is not precisely known. Further- Carman, P.C., 1937. Fluid flow through granular beds. Trans. Instn. Chem. Engrs. 15,
more, it should be noticed that, while packed-bed storage systems 150–166.
Cascetta, M., Cau, G., Puddu, P., Serra, F., 2015. Experimental investigation of a
comprised of solids of two very different sizes (e.g. mixture of packed bed thermal energy storage system. J. Phys. Conf. Ser. 655.
rocks and sand) are commonly considered, there is no specific Chandra, P., Willits, D.H., 1981. Pressure drop and heat transfer characteristics of
numerical model to simulate its behavior and its performances. air-rockbed thermal storage systems. Sol. Energy 27, 547–553.
Chang, Z.S., Li, X., Xu, C., Chang, C., Wang, Z.F., 2015. The design and numerical study
Many correlations are available in the literature to model and of a 2 MWh molten salt thermocline tank. Energy Proc. 69, 779–789.
compute fluid/solid and fluid/wall heat transfer coefficients, ther- Charlas, B., Doremus, P., Gillia, O., Imbault, D., 2010. Numerical analysis of a metal
mal diffusion and pressure drop in packed beds. Some of them hydride powder bed under swelling or lateral compaction by the discrete
element method. Proc. World Powder Metall. Congr. Exhibition, vol. 1.
were presented, compared and advocated. However, these are only Cheng, P., Hsu, C.T., 1998. Heat conduction. In: Ingham, D.B., Pop, I. (Eds.), Transport
general recommendations and the correlations should be chosen Phenomena in Porous Media. Pergamon, pp. 57–76.
depending on each specific case. Chu, Y.C., Storrow, J.A., 1952. Heat transfer to air flowing through packed tubes.
Chem. Eng. Sci. 1, 230–237.
In addition to better understanding of the physical mechanisms
Close, D.J., 1965. Rock pile thermal storage for comfort air conditioning. Mech.
which govern packed-bed storage behavior and performances, Chem. Eng. Trans. MC-1, 11–22.
some mechanical issues like thermal ratcheting as well as chemical Cocco, D., Serra, F., 2015. Performance comparison of two-tank direct and
issues like fluid/solid long-term compatibility should be further thermocline thermal energy storage systems for 1MWe class concentrating
solar power plants. Energy 81, 526–536.
investigated to enable wide spread of this storage technology. Colburn, A.P., 1931. Heat transfer and pressure drop in empty, baffled, and packed
tubes. Ind. Eng. Chem. 23, 910–913.
Coutier, J.P., Farber, E.A., 1982. Two applications of a numerical approach of heat
References transfer process within rock beds. Sol. Energy 29, 451–462.
Damköhler, G., 1937. Technische Reaktionsgeschwindigkeit. In: Eucken, A., Jakob,
M. (Eds.), Der Chemie-Ingenieur. Akademische Verlagsgesellschaft, Leipzig, pp.
Acetis, J., Thodos, G., 1960. Mass and heat transfer in flow of gases through spherical
359–485.
packings. Ind. Eng. Chem. 52, 1003–1006.
Deissler, R.G., Boegli, J.S., 1958. An investigation of effective thermal conductivities
Achenbach, E., 1995. Heat and flow characteristics of packed beds. Exp. Therm. Fluid
of powders in various gases. ASME Trans. 80, 1417–1425.
Sci. 10, 17–27.
Dixon, A.G., 1968. Heat Transfer in Packed Beds of Low Tube/Particle Diameter
Adeyanju, A.A., Manohar, K., 2009. Theoretical and experimental investigation of
Ratio. Univ. Edinburgh, Scotland.
heat transfer in packed beds. Res. J. Appl. Sci. 4, 166–177.
Dixon, A.G., Cresswell, D.L., 1979. Theoretical prediction of effective heat transfer
Airlight Energy. Aït-Baha CSP Pilot Plant. <http://www.airlightenergy.com/ait-baha-
parameters in packed beds. AIChE J. 25, 663–676.
csp-pilot-plant/>, Apr. 2016.
Dixon, A.G., DiCostanzo, M.A., Soucy, B.A., 1984. Fluid-phase radial transport in
Akinyele, D.O., Rayudu, R.K., 2014. Review of energy storage technologies for
packed beds of low tube-to-particle diameter ratio. Int. J. Heat Mass Transf. 27,
sustainable power networks. Sustain. Energy Technol. Assess. 8, 74–91.
1701–1713.
Alanís, E., Saravia, L., Rovetta, L., 1977. Measurement of rock pile heat transfer
Dogangun, A., Karaca, Z., Durmus, A., Sezen, H., 2009. Cause of damage and failures
coefficients. Sol. Energy 19, 571–572.
in silo structures. J. Perform. Constr. Facil. 23, 65–71.
Allen, K.G., von Backström, T.W., Kröger, D.G., 2015. Rock bed pressure drop and
Du Plessis, J.P., Woudberg, S., 2008. Pore-scale derivation of the Ergun equation to
heat transfer: simple design correlations. Sol. Energy 115, 525–536.
enhance its adaptability and generalization. Chem. Eng. Sci. 63, 2576–2586.
Anderson, R., Shiri, S., Bindra, H., Morris, J.F., 2014. Experimental results and
Dunkle, R.V., Ellul, W.M.J., 1972. Randomly packed particulate bed regenerators and
modeling of energy storage and recovery in a packed bed of alumina particles.
evaporative coolers. Inst. Eng. Aust. Mech. Chem. Eng. Trans. MC8, 117–121.
Appl. Energy 119, 521–529.
Durisch, W., Frick, E., Kesselring, P., 1986. Heat Storage in solar power plants using
Baddour, R.F., Yoon, C.Y., 1961. Local radial effective conductivity and the wall effect
solid beds, high temperature technology and its applications. Solar Thermal
in packed beds. Chem. Engg. Prog. Symp. Ser. 57, 35–50.
Central Receiver Systems: Proceedings of the Third International Workshop, vol.
Balakrishnan, A.R., Pei, D.C.T., 1979a. Heat transfer in gas-solid packed bed systems.
2, pp. 879–896.
1. A critical review. Ind. Eng. Chem. Process Des. Dev. 18, 30–40.
Ergun, S., 1952. Fluid flow through packed columns. Chem. Eng. Prog. 48, 89–94.
Balakrishnan, A.R., Pei, D.C.T., 1979b. Heat transfer in gas-solid packed bed systems.
Faas, S.E., Thorne, L.R., Fuchs, E.A., Gilbertsen, N.D., 1986. 10 Mwe Solar Thermal
2. The conduction mode. Ind. Eng. Chem. Process Des. Dev. 18, 40–46.
Central Receiver Pilot Plant - Thermal storage subsystem evaluation - Final
Bauer, R., 1990. Stagnant packed beds. In: Hewitt, G.F. (Ed.), Hemisphere Handbook
report, SAND86-8212.
of Heat Exchanger Design. Hemisphere Pub. Corp., New York.
Flueckiger, S.M., Iverson, B.D., Garimella, S.V., Pacheco, J.E., 2014. System-level
Bauer, R., Shlünder, E., 1978. Effective radial thermal conductivity of packings in gas
simulation of a solar power tower plant with thermocline thermal energy
flow – 2. Thermal conductivity of the packing fraction without gas flow. Int.
storage. Appl. Energy 113, 86–96.
Chem. Eng. 18, 189–204.
Flueckiger, S.M., Yang, Z., Garimella, S.V., 2013. Review of molten-salt thermocline
Bayón, R., Rivas, E., Rojas, E., 2013. Study of thermocline tank performance in
tank modeling for solar thermal energy storage. Heat Transf. Eng. 34, 787–800.
dynamic processes and stand-by periods with an analytical function. Energy
Foumeny, E.A., Kulkarni, A., Roshani, S., Vatani, A., 1996. Elucidation of pressure
Proc. 49, 725–734.
drop in packed-bed systems. Appl. Therm. Eng. 16, 195–202.
Bayón, R., Rojas, E., 2013. Simulation of thermocline storage for solar thermal power
Frössling, N., 1938. Über die Verdunstung fallender Teilchen. Gerlands Beitr.
plants: from dimensionless results to prototypes and real-size tanks. Int. J. Heat
Geophys. 52, 170–175.
Mass Transf. 60, 713–721.
Furnas, C.C., 1930. Heat transfer from a gas stream to a bed of broken solids. Ind.
Beasley, D.E., Clark, J.A., 1984. Transient response of a packed bed for thermal
Eng. Chem. 22, 26–31.
energy storage. Int. J. Heat Mass Transf. 27, 1659–1669.
Furnas, C.C., 1932. Heat transfer from a gas stream to a bed of broken solids. U.S.
Beek, J., 1962. Design of Packed Catalytic Reactors 203–271.
Department of Commerce, Bulletin 361, 88 p.
Bergman, T.L., Lavine, A.S., Incropera, F.P., Dewitt, D.P., 2011. Fundamentals of Heat
Galione, P., Pérez-Segarra, C.D., Rodríguez, I., Torras, S., Rigola, J., 2015. Numerical
and Mass Transfer. John Wiley & Sons, p. 1048.
evaluation of multi-layered solid-PCM thermocline-like tanks as thermal
Bhavsar, V.C., Balakrishnan, A.R., 1990. Pebble bed-oil thermal energy storage for
energy storage systems for CSP applications. Energy Proc. 69, 832–841.
solar thermo-electric power systems. Int. J. Energy Res. 14, 233–240.
Galloway, T.R., Sage, B.H., 1970. A model of the mechanism of transport in packed,
Biencinto, M., Bayón, R., Rojas, E., González, L., 2014. Simulation and assessment of
distended, and fluidized beds. Chem. Eng. Sci. 25, 495–516.
operation strategies for solar thermal power plants with a thermocline storage
Gamson, B.W., Thodos, G., Hougen, O.A., 1943. Trans. AIChE 39, 1–35.
tank. Sol. Energy 103, 456–472.
García-Olivares, A., Ballabrera-Poy, J., García-Ladona, E., Turiel, A., 2012. A global
Breitbach, G., Barthels, H., 1980. The radiant heat transfer in the HTR core after
renewable mix with proven technologies and common materials. Energy Pol.
failure of the afterheat removal systems. Nucl. Technol. 49, 392–399.
41, 561–574.
Brosseau, D., Kelton, J.W., Ray, D., Edgar, M., Chisman, K., Emms, B., 2005. Testing of
Gil, A., Medrano, M., Martorell, I., Lázaro, A., Dolado, P., Zalba, B., Cabeza, L.F., 2009.
thermocline filler materials and molten-salt heat transfer fluids for thermal
State of the art on high temperature thermal energy storage for power
energy storage systems in parabolic trough power plants. J. Sol. Energy Eng.
generation. Part 1 – Concepts, materials and modellization. Renew. Sustain.
Trans. ASME 127, 109–116.
Energy Rev. 14, 31–55.
Bruch, A., Fourmigué, J.F., Couturier, R., 2014a. Experimental and numerical
Gonzo, E.E., 2002. Estimating correlations for the effective thermal conductivity of
investigation of a pilot-scale thermal oil packed bed thermal storage system
granular materials. Chem. Eng. J. 90, 299–302.
for CSP power plant. Sol. Energy 105, 116–125.
 T. Esence et al. / Solar Energy 153 (2017) 628–654 653

Gunn, D.J., 1978. Transfer of heat or mass to particles in fixed and fluidised beds. Int. Mabrouk, M.T., Kheiri, A., Feidt, M., 2015. Using generalized integral transforms to
J. Heat Mass Transf. 21, 467–476. solve a perturbation model for a packed bed thermal energy storage tank. Int. J.
Gunn, D.J., Vortmeyer, D., 1982. Heat transfer in fixed bed chemical reactors. Heat Mass Transf. 84, 633–641.
Proceedings of the 7th International Heat Transfer Conference, vol. 6, pp. 13–18. Macdonald, I.F., El-Sayed, M.S., Mow, K., Dulllen, F.A.L., 1979. Flow through porous
Gupta, A.S., Thodos, G., 1962. Mass and heat transfer in the flow of fluids through media – the Ergun equation revisited. Ind. Eng. Chem. Fundam. 18, 199–208.
fixed and fluidized beds of spherical particles. AIChE J. 8, 608–610. Martin, H., 2005. The Lévêque-analogy or How to predict heat and mass transfer
Gupta, S.N., Chaube, R.B., Upadhyay, S.N., 1974. Fluid-particle heat transfer in fixed from fluid friction. 4th International Conference on Heat Transfer, Fluid
and fluidized beds. Chem. Eng. Sci. 29, 839–843. Mechanics and Thermodynamics, HEFAT2005.
Hadley, G.R., 1986. Thermal conductivity of packed metal powders. Int. J. Heat Mass Mawire, A., McPherson, M., 2009. Experimental and simulated temperature
Transf. 29, 909–920. distribution of an oil-pebble bed thermal energy storage system with a
Haller, M.Y., Cruickshank, C.A., Streicher, W., Harrison, S.J., Andersen, E., Furbo, S., variable heat source. Appl. Therm. Eng. 29, 1086–1095.
2009. Methods to determine stratification efficiency of thermal energy storage Mawire, A., McPherson, M., van den Heetkamp, R.R.J., Mlatho, S.J.P., 2009. Simulated
processes – review and theoretical comparison. Sol. Energy 83, 1847–1860. performance of storage materials for pebble bed thermal energy storage (TES)
Hallet, R.W. Jr., Gervais, R.L., 1977. Central receiver solar thermal power system – systems. Appl. Energy 86, 1246–1252.
Phase 1 - CDRL ITEM 2 - Pilot Plant Preliminary Design Report - Vol V - Thermal Mawire, A., McPherson, M., van den Heetkamp, R.R.J., Taole, S.H., 2010.
Storage Subsystem, SAN/1108-8/5. Experimental volumetric heat transfer characteristics between oil and glass
Handley, D., Heggs, P.J., 1968. Momentum and heat transfer mechanisms in regular pebbles in a small glass tube. Energy 35, 1256–1263.
shaped packings. Trans. Instn. Chem. Engrs. 46, T251–T264. Mawire, A., Taole, S.H., 2011. A comparison of experimental thermal stratification
Handley, D., Heggs, P.J., 1969. The effect of thermal conductivity of the packing parameters for an oil/pebble-bed thermal energy storage (TES) system during
material on transient heat transfer in a fixed bed. Int. J. Heat Mass Transf. 12, charging. Appl. Energy 88, 4766–4778.
549–570. Maxwell, J.C., 1873. A Treatise on Electricity and Magnetism. Clarendon Press,
Hicks, R.E., 1970. Pressure drop in packed beds of spheres. Ind. Eng. Chem. Fundam. Oxford. 425 p.
9, 500–502. McDonnell Douglas Astronautics Company, 1986. 10 MWe Solar Thermal Central
Hoffmann, J.F., Fasquelle, T., Goetz, V., Py, X., 2016. A thermocline thermal energy Receiver Pilot Plant Mode 5 (Test 1150) and Mode 6 (Test 1160) Test report,
storage system with filler materials for concentrated solar power plants: Sandia National Laboratories, SAND86-8175.
experimental data and numerical model sensitivity to different experimental Meier, A., Winkler, C., Wuillemin, D., 1991. Experiment for modelling high
tank scales. Appl. Therm. Eng. 100, 753–761. temperature rock bed storage. Sol. Energy Mater. 24, 255–264.
Hohmeyer, O.H., Bohm, S., 2015. Trends toward 100% renewable electricity supply Melanson, M.M., Dixon, A.G., 1985. Solid conduction in low dt/dp beds of spheres,
in Germany and Europe: a paradigm shift in energy policies. Wiley Interdiscip. pellets and rings. Int. J. Heat Mass Transf. 28, 383–394.
Rev. Energy Environ. 4, 74–97. Mertens, N., Alobaid, F., Frigge, L., Epple, B., 2014. Dynamic simulation of integrated
Hollands, K.G.T., Pott, P., 1971. Development and testing of evaporative cooler pads. rock-bed thermocline storage for concentrated solar power. Sol. Energy 110,
Mech. Chem. Eng. Trans., 66–72 830–842.
Hollands, K.G.T., Sullivan, H.F., 1984. Pressure drops across rock bed thermal storage Morgan, G.R., 1980. 10 Mwe Pilot Plant Thermal Storage Subsystem – Analysis
systems. Sol. Energy 33, 221–225. Report, RADL ITEM 5-1.
Hollands, K.G.T., Sullivan, H.F., Shewen, E.C., 1984. Flow uniformity in rock beds. Sol. Motte, F., Bugler-Lamb, S.L., Falcoz, Q., 2015. Thermocline storage filled with
Energy 32, 343–348. structured ceramics. Numerical consistency of the developed numerical model
Hsu, C.T., Cheng, P., Wong, K.W., 1994. Modified Zehner-Schlunder models for and first observations. High Temp. Mater. Process. 34, 353–365.
stagnant thermal conductivity of porous media. Int. J. Heat Mass Transf. 37, Nellis, G., Klein, S., 2009. Heat Transfer. Cambridge, 1107 p.
2751–2759. Nemec, D., Levec, J., 2005. Flow through packed bed reactors: 1. Single-phase flow.
Hsu, C.T., Cheng, P., Wong, K.W., 1995. A lumped-parameter model for stagnant Chem. Eng. Sci. 60, 6947–6957.
thermal conductivity of spatially periodic porous media. J. Heat Transf. Trans. Nield, D.A., 1991. Estimation of the stagnant thermal conductivity of saturated
ASME 117, 264–269. porous media. Int. J. Heat Mass Transf. 34, 1575–1576.
IAEA, 2000. Heat Transport and Afterheat Removal for Gas Cooled Reactors Under NREL. Concentrating Solar Power Projects. <http://www.nrel.gov/csp/solarpaces/>,
Accident Conditions, TECDOC-1163. Apr. 2016.
Inaba, H., Fukuda, T., 1984. Transient heat transfer behaviors in cylindrical porous Okello, D., Nydal, O.J., Banda, E.J.K., 2014. Experimental investigation of thermal de-
beds at relatively large Reynolds numbers. Wärme- und Stoffübertragung 18, stratification in rock bed TES systems for high temperature applications. Energy
109–116. Convers. Manage. 86, 125–131.
Ismail, K.A.R., Stuginsky, J., 1999. A parametric study on possible fixed bed models Pacheco, J.E., Showalter, S.K., Kolb, W.J., 2002. Development of a molten-salt
for pcm and sensible heat storage. Appl. Therm. Eng. 19, 757–788. thermocline thermal storage system for parabolic trough plants. J. Sol. Energy
Kaviany, M., 1995. Principles of Heat Transfer in Porous Media, second ed., 709 p. Eng. Trans. ASME 124, 153–159.
Khan, J.A., Beasley, D.E., Alatas, B., 1991. Evaporation from a packed bed of porous Pardo, P., Deydier, A., Anxionnaz-Minvielle, Z., Rougé, S., Cabassud, M., Cognet, P.,
particles into superheated vapor. Int. J. Heat Mass Transf. 34, 267–280. 2014. A review on high temperature thermochemical heat energy storage.
Klein, P., Roos, T.H., Sheer, T.J., 2013. Experimental investigation into a packed bed Renew. Sustain. Energy Rev. 32, 591–610.
thermal storage solution for solar gas turbine systems, SolarPACES 2013. Energy Pfeffer, R., 1964. Heat and mass transport in multiparticle systems. I&EC Fund. 3,
Proc. 49, 840–849. 380–383.
Kozeny, J., 1927. Uber Kapillare Leitung des Wasser im Boden. Sitzungsbericht der Philibert, C., 2011. Renewable Energy Technologies. Solar Energy Perspectives.
Akademie der Wissenschaften 136, 271–306. OECD, IEA, Paris, France. 228 p.
Krupiczka, R., 1967. Analysis of thermal conductivity in granular materials. Int. Philibert, C., 2014. IEA Technology Roadmap. Solar Thermal Electricity. OECD, IEA,
Chem. Eng. 7, 122–144. Paris, France. 46 p.
Kunii, D., Smith, J.M., 1960. Heat transfer characteristics of porous rocks. AIChE J. 6, Pudjianto, D., Aunedi, M., Djapic, P., Strbac, G., 2014. Whole-systems assessment of
71–78. the value of energy storage in low-carbon electricity systems. IEEE Trans. Smart
Kunii, D., Suzuki, M., 1968. Heat and mass transfer from wall surface to packed beds. Grid 5, 1098–1109.
Sympos. Heat Mass Transf. Qin, F.G.F., Yang, X., Ding, Z., Zuo, Y., Shao, Y., Jiang, R., Yang, X., 2012. Thermocline
Kuravi, S., Trahan, J., Goswami, Y., Jotshi, C., Stefanakos, E., Goel, N., 2013. stability criterions in single-tanks of molten salt thermal energy storage. Appl.
Investigation of a high-temperature packed-bed sensible heat thermal energy Energy 97, 816–821.
storage system with large-sized elements. J. Sol. Energy Eng. Trans. ASME 135. Ranz, W.E., 1952. Friction and transfer coefficients for single particles and packed
Kuznetsov, A.V., 1995. An analytical solution for heating a two-dimensional beds. Chem. Eng. Progr. 48, 247–253.
porous packed bed by a non-thermal equilibrium fluid flow. Appl. Sci. Res. 55, Rodat, S., Bruch, A., Dupassieux, N., Mourchid, N.E., 2015. Unique Fresnel
83–93. demonstrator including ORC and thermocline direct thermal storage:
Lakota, A., Levec, J., Carbonell, R.G., 2002. Hydrodynamics of trickling flow in packed operating experience. Energy Proc. 69, 1667–1675.
beds: relative permeability concept. AIChE J. 48, 731–738. Rose, H.E., Rizk, A.M.A., 1949. Further researches in fluid flow through beds of
Leva, M., 1947. Heat transfer to gases through packed tubes – general correlation for granular material. Proc. Inst. Mech. Eng. 160, 493–503.
smooth spherical particles. Ind. Eng. Chem. 39, 857–862. Sabihuddin, S., Kiprakis, A.E., Mueller, M., 2015. A numerical and graphical review of
Leva, M., 1950. Packed-tube heat transfer. Ind. Eng. Chem. 42, 2498–2501. energy storage technologies. Energies 8, 172–216.
Leva, M., Grummer, M., 1948. Heat transfer to gases through packed tubes – effect of Saez, A.E., McCoy, B.J., 1982. Dynamic response of a packed bed thermal storage
particle characteristics. Ind. Eng. Chem. 40, 415–419. system – a model for solar air heating. Sol. Energy 29, 201–206.
Leva, M., Weintraub, M., Grummer, M., Clark, E.L., 1948. Cooling of gases through Sagara, K., Nakahara, N., 1991. Thermal performance and pressure drop of rock beds
packed tubes. Ind. Eng. Chem. 40, 747–752. with large storage materials. Sol. Energy 47, 157–163.
Lévêque, A., 1928. Les Lois De La Transmission De Chaleur Par Convection. Satterfield, C.N., Resnick, H., Wentworth, R.L., 1954. Simultaneous heat and mass
Université de Paris. transfer in a diffusion-controlled chemical reaction. Chem. Eng. Progr. 50, 460–
Li, C.H., Finlayson, B.A., 1977. Heat transfer in packed beds: a re-evaluation. Chenz. 466.
Eng. Sci. 32, 1055–1066. Schröder, E., Class, A., Krebs, L., 2006. Measurements of heat transfer between
Libby, C., 2010. Solar thermocline storage systems: preliminary design study, EPRI, particles and gas in packed beds at low to medium Reynolds numbers. Exp.
2010.1019581, 188 p. Therm. Fluid Sci. 30, 545–558.
Löf, G.O.G., Hawley, R.W., 1948. Unsteady state heat transfer between air and loose Schumann, T.E.W., 1929. Heat transfer: a liquid flowing through a porous prism. J.
solids. Ind. Eng. Chem. 40, 1061–1070. Franklin Inst. 208, 405–416.
654 T. Esence et al. / Solar Energy 153 (2017) 628–654

Shewen, E.C., Sullivan, H.F., Hollands, K.G.T., Balakrishnan, A.R., 1978. A heat storage Wakao, N., Kato, K., 1969. Effective thermal conductivity of packed beds. J. Chem.
subsystem for solar energy – Final Report - Phase 2, STOR - 6. Eng. Jpn. 2, 24–33.
Shitzer, A., Levy, M., 1983. Transient behavior of a rock-bed thermal storage system Weidenfeld, G., Weiss, Y., Kalman, H., 2004. A theoretical model for effective
subjected to variable inlet air temperatures – analysis and experimentation. J. thermal conductivity (ETC) of particulate beds under compression. Granul.
Sol. Energy Eng. Trans. ASME 105, 200–206. Matter 6, 121–129.
Sih, S.S., Barlow, J.W., 1995. The prediction of the thermal conductivity of powders. Whitaker, S., 1972. Forced convection heat transfer correlations for flow in pipes,
6th SFF Symposium, pp. 397–401. past flat plates, single cylinders, single spheres, and for flow in packed beds and
Singh, R., Saini, R.P., Saini, J.S., 2006. Nusselt number and friction factor correlations tube bundles. AIChE J. 18, 361–371.
for packed bed solar energy storage system having large sized elements of Xu, B., Li, P., Chan, C., 2015. Application of phase change materials for thermal
different shapes. Sol. Energy 80, 760–771. energy storage in concentrated solar thermal power plants: a review to recent
Stuke, B., 1948. Berechnung des Wärmeaustausches in Regeneratoren mit developments. Appl. Energy 160, 286–307.
zylindrischen und kugelförmigen Füllmaterial. Angew. Chem. B20, 262. Xu, B., Li, P.W., Chan, C.L., 2012a. Extending the validity of lumped capacitance
Szekely, J., Poveromo, J.J., 1975. Flow maldistribution in packed beds: a comparison method for large Biot number in thermal storage application. Sol. Energy 86,
of measurements with predictions. AIChE J. 21, 769–775. 1709–1724.
Thoenes, D.J., Kramers, H., 1958. Mass transfer from spheres in various regular Xu, C., Wang, Z., He, Y., Li, X., Bai, F., 2012b. Sensitivity analysis of the numerical
packings to a flowing fluid. Chem. Eng. Sci. 8, 271–283. study on the thermal performance of a packed-bed molten salt thermocline
Tian, Y., Zhao, C.Y., 2013. A review of solar collectors and thermal energy storage in thermal storage system. Appl. Energy 92, 65–75.
solar thermal applications. Appl. Energy 104, 538–553. Yagi, S., Kunii, D., 1960. Studies on heat transfer near wall surfaces in packed beds.
Tsotsas, E., Martin, H., 1987. Thermal conductivity of packed beds: a review. Chem. AIChE J. 6, 97–104.
Eng. Proc. 22, 19–37. Yagi, S., Wakao, N., 1959. Heat and mass transfer from wall to fluid in packed beds.
Tsotsas, E., Schlünder, E.U., 1990. Heat transfer in packed beds with fluid flow: AIChE J. 5, 79–85.
remarks on the meaning and the calculation of a heat transfer coefficient at the Yang, X., Qin, F.G.F., Jiang, R., 2014. Experimental investigation of a molten salt
wall. Chem. Eng. Sci. 45, 819–837. thermocline storage tank. Int. J. Sustain. Energy.
Van Antwerpen, W., Du Toit, C.G., Rousseau, P.G., 2010. A review of correlations to Yang, Z., Garimella, S.V., 2010a. Molten-salt thermal energy storage in thermoclines
model the packing structure and effective thermal conductivity in packed beds under different environmental boundary conditions. Appl. Energy 87, 3322–
of mono-sized spherical particles. Nucl. Eng. Des. 240, 1803–1818. 3329.
Van Antwerpen, W., Rousseau, P.G., Du Toit, C.G., 2009. Accounting for porous Yang, Z., Garimella, S.V., 2010b. Thermal analysis of solar thermal energy storage in
structure in effective thermal conductivity calculations in a pebble bed reactor. a molten-salt thermocline. Sol. Energy 84, 974–985.
Int. Congr. Adv. Nuc. Power Plants, ICAPP 1, 258–269. Yu, J., Zhang, M., Fan, W., Zhou, Y., Zhao, G., 2002. Study on performance of the ball
Van Lew, J.T., Li, P., Chan, C.L., Karaki, W., Stephens, J., 2011. Analysis of heat storage packed-bed regenerator: experiments and simulation. Appl. Therm. Eng. 22,
and delivery of a thermocline tank having solid filler material. J. Sol. Energy Eng. 641–651.
Trans. ASME 133, 021003-1–021003-10. Zanganeh, G., Ambrosetti, G., Pedretti, A., Zavattoni, S., Barbato, M., Good, P.,
Vortmeyer, D., 1989. Packed bed thermal dispersion models and consistent sets of Haselbacher, A., Steinfeld, A., 2014. A 3 MWth parabolic trough CSP plant
coefficients. Chem. Eng. Proc. 26, 263–268. operating with air at up to 650 °C. Proc. Int. Renew. Sustain. Energy Conf., 108–113
Vortmeyer, D., Schaefer, R.J., 1974. Equivalence of one- and two-phase models for Zanganeh, G., Pedretti, A., Zavattoni, S., Barbato, M., Steinfeld, A., 2012. Packed-bed
heat transfer processes in packed beds: one dimensional theory. Chem. Eng. Sci. thermal storage for concentrated solar power – pilot-scale demonstration and
29, 485–491. industrial-scale design. Sol. Energy 86, 3084–3098.
Votyakov, E.V., Bonanos, A.M., 2014. A perturbation model for stratified thermal Zavattoni, S.A., Barbato, M.C., Pedretti, A., Zanganeh, G., 2015. Single-tank TES
energy storage tanks. Int. J. Heat Mass Transf. 75, 218–223. system – transient evaluation of thermal stratification according to the second-
Votyakov, E.V., Bonanos, A.M., 2015. Algebraic model for thermocline thermal law of thermodynamics. Energy Proc. 69, 1068–1077.
storage tank with filler material. Sol. Energy 122, 1154–1157. Zavattoni, S.A., Barbato, M.C., Pedretti, A., Zanganeh, G., Steinfeld, A., 2014. High
Wakao, N., Kaguei, S., 1982. Heat and Mass Transfer in Packed Beds. Gordon and temperature rock-bed TES system suitable for industrial-scale CSP plant – CFD
Breach Science Publishers, New York. 364 p. analysis under charge/discharge cyclic conditions. Energy Proc. 46, 124–133.
Wakao, N., Kaguei, S., Funazkri, T., 1979. Effect of fluid dispersion coefficients on Zehner, P., Schlünder, E.U., 1970. Wärmeleitfähigkeit von Schüttungen bei mäßigen
particle-to-fluid heat transfer coefficients in packed beds. Correlation of Nusselt Temperaturen. Chem. Ing. Tec. 42, 933–941.
numbers. Chem. Eng. Sci. 34, 325–336.