Journal of Energy Storage 17 (2018) 140–152

Contents lists available at ScienceDirect

Journal of Energy Storage

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

Pilot-scale demonstration of advanced adiabatic compressed air energy

storage, Part 2: Tests with combined sensible/latent thermal-energy
storage
V. Becattinia , L. Geissbühlera , G. Zanganehb , A. Haselbachera,* , A. Steinfelda
a
Department of Mechanical and Process Engineering, ETH Zurich, 8092 Zurich, Switzerland
b
ALACAES SA, 6900 Lugano, Switzerland

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

Article history:
Received 29 November 2017 Experimental and numerical results from the world’s first pilot-scale advanced adiabatic compressed air
Received in revised form 6 February 2018 energy storage plant with combined sensible/latent thermal-energy storage are presented. The combined
Accepted 6 February 2018 thermal-energy storage was composed of sensible and latent units with maximum capacities of
Available online 23 March 2018 11.6 MWhth and 171.5 kWhth, respectively. The latent thermal-energy storage consisted of a steel tank
with 296 stainless-steel tubes encapsulating an Al–Cu–Si alloy as phase-change material. The combined
Keywords: thermal-energy storage was investigated using four charging/discharging cycles with durations of about
Advanced adiabatic compressed air energy 3 h each and air inflow temperatures of up to 566
C. The experimental results showed that the latent
storage
thermal-energy storage reduced the drop in the air outflow temperature during discharging. Minor leaks
Thermal-energy storage
of the phase-change material were traced to the welding seams in the encapsulation as well as to holes
Packed bed
Pilot plant required to insert resistance temperature detectors. Analysis of the leaked phase-change material
Simulation revealed degradation and/or phase separation, which were attributed to the initial off-eutectic
Phase-change material composition of and impurities in the phase-change material and resulted in a reduced heat of fusion.
Simulations predicted the performance of the combined thermal-energy storage with good overall
accuracy. Discrepancies were put down to changes in the thermophysical properties.
© 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction time, which is unfavorable for the turbine in an AA-CAES plant. To

Among large-scale energy-storage technologies, advanced
adiabatic compressed air energy storage (AA-CAES) has recently
attracted much interest because of projected high power outputs
(above 100 MW), high efficiencies (about 60–75%), and low capital
costs, see Luo et al. [1], Budt et al. [2], and Sciacovelli et al. [3].
Experimental and numerical results from the first pilot-scale AA-
CAES plant with high-temperature sensible thermal-energy
storage (TES) were reported in Part 1, see Geissbühler et al. [4].
The results demonstrated the technical feasibility of both the AA-
CAES technology and the packed-bed TES that used rocks as the
storage material and air as the heat-transfer fluid (HTF).
Sensible TES systems using rocks as storage material and air as
HTF are of interest because experimental and numerical inves-
tigations have found them to be promising in terms of efficiencies
and costs. One drawback of such TES systems is that during
discharging, the temperature of the outflowing air decreases with
* Corresponding author.
E-mail address: haselbac@ethz.ch (A. Haselbacher).
overcome this drawback, a relatively small amount of phase-
change material (PCM) can be placed on top of the packed bed of
rocks, resulting in a so-called combined sensible/latent TES.
Combined sensible/latent TES has been investigated numerically
and experimentally by several authors. Zanganeh et al. [5] and
Geissbühler et al. [6] performed numerical and experimental
analyses of a 42 kWhth laboratory-scale combined storage consisting
of a packed bed of rocks and an eutectic aluminum–silicon alloy as
PCM encapsulated in steel tubes. They demonstrated stabilization of
the air outflow temperature around the melting temperature of the
PCM. In addition, Geissbühler et al. [6] used simulations to show that
23 MWhth and 1000 MWhth combined storages have higher exergy
efficiencies and lower specific material costs compared to sensible
storages for a given maximum outflow temperature drop during
discharging. Galione et al. [7] compared numerically several
industrial-scale combined storage configurations for a 50 MWhel
concentrated solar power plant and found that a storage consisting of
multiple layers of sensible materials and PCMs diminishes the
degradation of the thermocline and has a higher utilization factor.
Zhao et al. [8] investigated numerically storage concepts with an
output of 250 MWhth and found that a storage consisting of sensible

https://doi.org/10.1016/j.est.2018.02.003
2352-152X/© 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152 141

3. Assess the thermal and mechanical stability of the latent TES

Nomenclature
tank, the encapsulation, and the PCM in response to the
temperature and pressure variations during the charging/
discharging cycles.
Abbreviations
AA-CAES advanced adiabatic compressed air energy storage
DSC differential scanning calorimeter/calorimetry
2. Plant description
HTF heat-transfer fluid
LFA laser flash analysis
The pilot-scale AA-CAES plant and its main components are
PCM phase-change material
described in detail in Part 1. Because the plant used to produce the
RTD resistance temperature detector
results reported in this paper differs from that described in Part 1
TES thermal-energy storage
only through the TES, this section is restricted to a description of
the combined sensible/latent TES.
Greek characters
a thermal diffusivity (mm2/s) 2.1. Combined sensible/latent TES
r density (kg/m3)
As shown in Figs. 1 and 2, the combined sensible/latent TES
Latin characters
actually consists of two separate storages: a sensible storage,
cp specific heat capacity at constant pressure (kJ/kg K)
identical to that described in Part 1, and a latent storage. The main
m_ mass flow rate (kg/s)
reason why the two storages are separate is that they were
h specific enthalpy (kJ/kg)
designed and constructed at different times. Additional reasons
k thermal conductivity (W/m K)
that justify using two separate TES units will be given below. It is
T temperature (
C)
important to note that the addition of the latent storage affects the
t time (h)
performance of the sensible storage because during charging, the
x axial coordinate (m)
compressed air flows first through the latent storage from top to
bottom and then through the sensible storage from top to bottom.
Subscripts
Therefore, the melting temperature of the PCM and thermal losses
bot bottom of the storage unit
from the pipe connecting the two storages affect the temperature
el electric
of the air flowing into the sensible storage. As a result, the
exp experimental results
maximum capacity of the sensible storage is 11.6 MWhth (calcu-
lat latent unit
lated for an inflow temperature of 529
C) rather than 12 MWhth
ref reference
given in Part 1 (calculated for an inflow temperature of 550
C). The
sens sensible unit
maximum capacity of the latent storage is 171.5 kWhth (calculated
sim simulation results
for an inflow temperature of 566
C).
th thermal
During the preparation of the experiments with the combined
top top of the storage unit
TES and accounting for the constraints imposed by the previously
c charging
constructed sensible TES, three options were considered: (1)
d discharging
placing the encapsulated PCM directly on top of the packed bed of
f fusion
rocks, (2) placing the encapsulated PCM inside the air distributor,
l liquid
which is indicated in Fig. 2, and (3) placing the encapsulated PCM
m melting
in a separate storage. Option (1) was adopted in the laboratory-
pc pre-charging
scale combined storage by Zanganeh et al. [5] and Geissbühler et al.
s solid
[6]. Due to the limited space available above the packed bed and
the resulting poor heat transfer, this option could not be pursued.
Similarly, option (2) was discarded because of the limited space
available inside the distributor and because it would have required

material and PCMs at the top and bottom of a tank with an optimum
configuration is more cost-competitive than other thermocline TES
systems for the same design requirements and operating conditions.
In summary, the literature shows that combined sensible/latent
TES can be superior to sensible-only TES in terms of both efficiency
and cost. While the combined sensible/latent TES has been
investigated experimentally at the laboratory scale, to our
knowledge it has so far not been tested at the pilot scale.
Therefore, the overall goal of this paper is to investigate
experimentally and numerically the performance of a combined
sensible/latent TES in the pilot-scale AA-CAES plant introduced in
Part 1. The specific objectives of this paper are:

1. Demonstrate the feasibility of a combined sensible/latent TES in

the pilot-scale AA-CAES plant using several charging/discharg-
ing cycles.
2. Validate the quasi-one-dimensional heat-transfer model of
Geissbühler et al. [6] using the experimental data collected with
the combined TES.
Fig. 1. Picture of combined sensible/latent TES, with latent storage in foreground
and sensible storage in background. The insulated pipes between the heater and the
latent storage and between the latent and sensible storages can be seen.
142 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152
Fig. 2. Schematic drawing of the combined sensible/latent TES and locations of air
inflow/outflow temperature measurements.
strengthening to withstand the weight of the encapsulated PCM.
Therefore, option (3) was chosen and a separate latent storage was
designed and constructed.
The combined sensible/latent storage with separate sensible
and latent storages is judged to have several advantages for the
current pilot-scale and/or for possible future industrial-scale
applications. First, the cross-sectional area of the latent storage
can be chosen independently of that of the sensible storage to
give good heat-transfer rates and hence melting behavior of the
PCM. Second, a separate latent storage provides the possibility to
switch between experiments either with only the sensible TES or
with the combined TES by disconnecting and reconnecting the
pipe linking the two storages. Third, at least for the current pilot-
scale plant, a separate latent storage was easier to construct and
adapt than if the encapsulated PCM had been placed in the
sensible storage, which would have required the removal of its
steel cover and the pipe attached to the cover. Fourth, with a view
to the possible use of industrial-scale combined storages, a
separate latent storage allows the PCM to be inspected and/or
replaced more easily. Periodic replacement of the PCM might be
required due to corrosion reactions between the PCM and the
encapsulation. For metallic PCMs and encapsulations, these
reactions can lead to the growth of intermetallic layers and
result in a decrease of the performance, see Yan and Fan [9],
Yeremenko et al. [10], and Binder and Haussener [11], as well as
potential leaking of the PCM through the encapsulation. Finally,
again with a view to possible industrial-scale combined storages,
a separate latent storage means that the top of the packed bed is
not obstructed by the encapsulated PCM, simplifying the
replacement of rocks near the top of the packed bed. These
rocks might have to be replaced periodically because the higher
temperatures near the top of the packed bed might cause them to
degrade more quickly, see Allen et al. [12], Tiskatine et al. [13],
and Becattini et al. [14].
Separating the sensible and latent storages results in at least
two potential disadvantages. The first is that the pipe linking the
two storages must be well insulated to avoid thermal losses. The
second disadvantage is that the total surface area of separate
storages is likely to be larger than that of a single storage, which
will increase thermal losses unless the storages are adequately
insulated. The additional insulation will increase material costs.
2.2. Latent TES
For the latent TES, the eutectic metal alloy with the theoretical
composition 68.5Al26.5Cu5Si (wt%) was chosen as the PCM. This
choice was dictated by the following considerations. The melting
temperature of the PCM must be sufficiently below the
temperature of the air flowing into the latent storage during
charging to allow complete melting. Based on simulations of
laboratory-scale combined sensible/latent storages, Zanganeh
et al. [5] recommend that the melting temperature of the PCM
should be about 20
C below the air inflow temperature during
charging. Because the design temperature of the air inflow during
charging was 550
C, suitable melting temperatures were
considered to be between 520 and 530
C. Among PCMs with
melting temperatures in this range, we restricted our choice to
metal alloys because of their high thermal conductivity, small
overcooling during solidification, and small volume change
during melting, see Kenisarin [15]. Among suitable metal alloys
with high heats of fusion, see Kenisarin [15], we selected
68.5Al26.5Cu5Si because of unpublished laboratory-scale expe-
rience with this material. The thermophysical properties of this
PCM were reported by Gasanaliev and Gamataeva [16] as a
melting temperature of 525
C, a heat of fusion of 365 kJ/kg, and a
density of 2938 kg/m3. The elemental composition and the
temperature-dependent properties of the alloy used in the
experiments will be presented in Section 3.
Based on prior experience with a laboratory-scale combined
storage by Zanganeh et al. [5] and Geissbühler et al. [6], we elected
to encapsulate the PCM in AISI316 stainless-steel tubes. Because
the laboratory-scale combined storage had a circular cross-section,
the length of the tubes was not constant. This did not present any
problems because only 68 tubes were needed. For the pilot-scale
storage, however, tubes of variable length would have increased
the cost and complicated the assembly of the storage. As a result,
the PCM in the pilot-scale combined storage was encapsulated in
tubes of equal length and therefore the latent storage had a square
cross section. To determine the cross section and height of the
latent storage as well as the number and diameter of the tubes
encapsulating the PCM, we used the quasi-one-dimensional heat-
transfer model of Geissbühler et al. [6].
The storage consists of two nested shells made from AISI304
stainless steel panels. The inner shell consists of a central body
with a square cross-section of 0.75 m by 0.75 m and a height of
0.65 m as well as two truncated pyramids that connect the central
body to the inlet/outlet pipes. The outer shell fixes the insulation,
made of microporous panels of powdered silica (MICROBIFIRE
1000 from Bifire S.R.L.) with a thickness of 15 cm. The properties of
the insulation material are given in Part 1. The pipe connecting the
latent storage to the sensible storage is made of AISI304 stainless
steel and wrapped with several layers of microporous insulation,
ceramic wool, and rock wool to reduce heat losses.
The PCM was encapsulated in 296 tubes made of AISI316
stainless steel. The tubes had a length of 73 cm, an inner diameter
of 3.2 cm, and a thickness of 0.15 cm. The tubes were closed with
caps made of AISI316 stainless steel with a thickness of 0.15 cm.
The length of the tubes was chosen to be 2 cm shorter than the
dimensions of the inner shell to allow for thermal expansion. The
tubes were arranged into 16 layers, with the tubes in each layer
resting on and oriented at 90
with respect to the tubes in the
layer below. The tubes in layers with same orientation were
staggered by half a tube diameter, as shown in the inset in Fig. 3,
resulting in 8 layers each with 18 and 19 tubes. The gaps between
the outermost tubes in a layer and the inner shell were stuffed
with strips of insulation material (felt), as can be seen from Fig. 4.
These strips ensured that the tubes in a layer were touching at the
caps, prevented the tubes from rolling, and eliminated bypass
flows. The tubes in the bottommost layer were resting on a
perforated plate of S235 JR steel with a thickness of 1 cm and a
void fraction of 0.16.
The tubes were filled while the PCM was liquid. A suitable void
space was left while filling the tubes to allow expansion of the PCM
upon solidification. Prior to filling, eight of the tubes were
internally coated with a protective layer that was shown to
prevent the formation of intermetallic layers in small-scale
experiments, see Binder and Haussener [11]. The coated tubes
were included in the pilot-scale experiments to assess the
 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152
Fig. 3. Drawing of the latent TES with positions of resistance temperature detectors
(RTDs). The inset shows a detail view of the staggered tubes. Note: triangles and circles
indicate that an RTD is measuring the PCM and air temperatures, respectively.
Fig. 4. Picture of interior of latent storage, partially filled with the steel tubes that
encapsulate the PCM. Some of the RTDs that were used to measure the air and PCM
temperatures are visible.
effectiveness of the coating under operating conditions that are
representative of industrial-scale AA-CAES plants.
2.3. Sensors
The temperatures in the two TES units were measured by
resistance temperature detectors (RTDs). Seven RTDs were placed
on the centerline of the latent TES as shown in Fig. 3. As indicated
in the figure, two RTDs measured the temperature of the PCM.
These RTDs were inserted into the solid PCM by drilling holes
through the encapsulations. The remaining five RTDs measured air
temperatures. Due to the limited number of data-acquisition
channels, measuring temperatures in the latent TES meant using
fewer RTDs in the sensible TES compared to the experiments
described in Part 1. (See Fig. 5 of Part 1 for the locations of the RTDs
in the sensible TES.) The RTDs in the sensible TES that were
disconnected for the tests with the combined TES were chosen
such that the axial resolution of the thermocline was not affected.
The air mass flow rate was measured as described in Part 1.
143
Table 1
Actual composition of PCM in wt.% as obtained from supplier. In addition to the
elements shown, minor impurities of Cr, Ni, Ca, Na, and Sr were also detected.
Al Cu Si Mg Ti Fe Mn Zn
67.100 25.861 6.508 0.265 0.083 0.049 0.035 0.019
3. PCM composition and properties
Table 1 shows that the actual composition of the PCM obtained
from the supplier (Fonderia EMI S.R.L.) differed from the eutectic
composition. As a result, the thermophysical properties of the
supplied PCM were expected to differ from those reported by
Gasanaliev and Gamataeva [16]. Because the thermophysical
properties of the PCM can have a significant impact on the
performance of the latent storage, the thermophysical properties
of the supplied PCM were measured and are reported in the
following.
3.1. Density
The density was determined at room temperature by linear
measurements using three disks with average diameter and
thickness of 12.51 mm and 2.92 mm, respectively, and found to be
3246 kg/m3 with an estimated relative error of 0.2%. The measured
density is thus about 10% higher than that reported by Gasanaliev
and Gamataeva [16]. The difference is ascribed to the different
compositions.
3.2. Specific heat capacity, heat of fusion, and melting temperature
range
The specific heat capacity was determined by TOPEM1, a
temperature-modulated differential scanning calorimetry (DSC)
technique introduced by Mettler Toledo, see Schawe et al. [17]. The
measurements were performed on a PCM disk with a mass of
56.5 mg in the temperature ranges 100–500
C and 545–700
C
using a heating rate of 1
C/min. Sapphire was used as the reference
material. The heat of fusion and melting temperature range were
determined by classical DSC in the temperature range 500–550
C
using a sample of 2.3 mg with a heating rate of 5
C/min. Both
TOPEM1 and classical DSC were performed using 30 mL alumina
crucibles in Ar flows of 50 ml/min with a Mettler Toledo DSC3+,
previously calibrated using Zn, In, and Al.
Fig. 5(a) shows the measured specific heat capacity for the solid
and liquid phases. The specific heat capacity of the solid PCM
increases with temperature, in agreement with the behavior of the
solid metals forming the alloy, i.e., Al, Cu, and Si, see Chase [18]. At
about 350
C, the specific heat capacity exhibits an unexpected
change in slope. This can be attributed to a second-order phase
transition, e.g., a solid–solid phase transition (for which the Gibbs
free energy and its first derivatives are continuous but its second
derivatives are discontinuous), which would not occur in an
eutectic alloy. After the melting phase, the specific heat capacity
decreases with temperature. The decrease is in agreement with the
study of Bergman and Komarek [19], who report that the specific
heat capacity of liquid alloys generally decreases with increasing
temperature.
The average specific heat capacity of the solid PCM in the
temperature range 100–500
C is cp;s ¼ 1:004 kJ/kg. For the average
specific heat capacity of the liquid PCM, only the temperature
range 545–566
C is considered to reflect the maximum tempera-
ture reached in the experiments. The average specific heat capacity
of the liquid PCM is then cp;l = 1.103 kJ/kg.
Fig. 5(b) shows the measured heat flows during melting and
solidification. We first discuss the heat-flow curve obtained during
144 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152
Fig. 5. Measured (a) specific heat capacity and (b) heat flow during melting and
solidification of the PCM.
the endothermic melting process, which when integrated results
in a heat of fusion of 404 kJ/kg. The relatively broad heat-flow curve
as well as the presence of two peaks and the plateau at
temperatures of about 535–540
C are ascribed to the material
impurities and to the non-uniform structure resulting from the off-
eutectic composition, see Table 1.
The first peak at a temperature of about 510
C is attributed to a
solid–solid phase transition. The second peak corresponds to the
melting of the eutectic phase and reaches a maximum at about
527
C. The plateau at temperatures of about 535–540
C is
hypothesized to be caused by the superposition of the high-
temperature tail of the second peak and a third peak at about
540
C. The obscured third peak might be caused by the melting of
carbides that originate from the impurities. The maximum heat
flow associated with the hypothesized third peak is thought to be
small because the carbide amounts are expected to be small. A
detailed analysis of the secondary phases present in the PCM,
which requires that the binary and ternary phase diagrams of Al,
Cu, and Si be analyzed, was beyond the scope of this study.
Because of the broad heat-flow curve during melting, we infer
that the melting process does not occur at a distinct temperature
but rather within a temperature interval, and hence we refer to a
melting temperature range. In the absence of a sharp and narrow
peak of the heat-flow curve (as expected for an eutectic alloy), the
temperature of the peak only indicates that the sample is
completely molten but does not correspond to the melting
temperature, see Boettinger et al. [20]. Consequently, the melting
temperature range was extracted from Fig. 5(b) using two
considerations. First, for alloys the melting process is considered
to begin at the temperature at which the heat-flow signal departs
from the baseline rather than at the on-set temperature, see
Boettinger et al. [20]. Second, the temperature at which the heat-
flow signal returns to the baseline is not significant since it is
affected by experimental conditions (such as the heating rate and
sample mass) and is therefore merely indicative of thermal lags
within the instrument and the sample, see, e.g., He et al. [21]. Based
on these considerations, we extract the melting temperature range
of the PCM to be 509–527
C, corresponding to the range between
the temperature at which the heat-flow signal departs from the
baseline and the peak temperature.
We now move on to discuss the heat-flow curve obtained
during the exothermic solidification process. The two small peaks
at temperatures of about 518
C and 530
C are attributed to the
solidification of carbides that originate from the impurities and to
the formation of new secondary phases upon solidification
because of phase separation, see Mehling and Cabeza [22, pp.
65–66]. The large peak at about 511
C, which is caused by the
solidification of the eutectic phase, indicates a marked hysteric
behavior of the PCM with a temperature difference of about 16
C
between the melting and solidification peaks. This behavior could
be caused by a genuine subcooling effect or by an apparent
hysteresis introduced by the sample size and the heating and
cooling rates used in the experiments, see Mehling and Cabeza [22,
p. 66] and Kheradmand et al. [23]. Because of the uncertainty
associated with the experimentally determined solidification
temperature range, in the simulations the experimentally deter-
mined melting temperature range is used for both melting and
solidification.
Compared to the values reported by Gasanaliev and Gamataeva
[16], we measured a relatively large melting temperature range
with a mid-value of 518
C rather than a distinct melting
temperature of 525
C and a heat of fusion that is about 10% larger.
3.3. Thermal diffusivity and thermal conductivity
The thermal diffusivity was determined using laser-flash
analysis (LFA) with a NETZSCH LFA 457 MicroFlash. Three samples
were used with an average mass of 1163.2 mg and an average
diameter and thickness of 12.51 mm and 2.92 mm, respectively.
Measurements were taken between 50 and 450
C in 50
C
increments. The maximum temperature was chosen to be below
the melting range to avoid softening or melting of the PCM. Three
measurements were performed per sample at each temperature.
For all measurements, the Cowan model with pulse correction
from the NETZSCH software was used.
The thermal conductivity of the sample was calculated from
kðTÞ ¼ rcp;s ðTÞaðTÞ; ð1Þ
where r is the measured density at room temperature
(assuming negligible expansion during the measurements), cp,
s(T) is the temperature-dependent specific heat capacity measured
by DSC, and a(T) is the temperature-dependent thermal diffusivity
measured using LFA.
Fig. 6 shows the average measured thermal diffusivity and
calculated average thermal conductivity as a function of tempera-
ture. The averages at each temperature were computed from the
three measurements per sample for three samples. The vertical
bars indicate the standard deviation of the average values of the
three samples. From Fig. 6(a), the average thermal diffusivity is
seen to have a maximum at about 300
C, which is about 11% larger
 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152 145

Table 2
Measured and calculated thermophysical properties of PCM.

Quantity Value Units

DTm 509–527

C
Dhfus 404 kJ/kg
cp;s 1.004 kJ/kg K
cp;l 1.103 kJ/kg K
k 151 W/mK
r 3246 kg/m3

coefficient between the air and the tubes encapsulating the PCM.
The heat-transfer coefficient for the perforated plate is calculated
using the correlation of Tomi c et al. [26] with a pitch-to-diameter
ratio of 2.2. At the lateral walls, natural convection and conduction
losses through the insulation are considered while the thermal
inertia of the metal sheets is neglected. At the top and bottom of the
tank, adiabatic conditions are assumed. Grid- and time-refinement
studies were performed to ensure that the results were numerically
accurate.
The temperature-dependent thermophysical properties of the
insulation materials and the rocks are given in Part 1. The
temperature-independent thermophysical properties for the PCM
and the encapsulation material are those reported in Table 2 and by
Geissbühler et al. [6], respectively.

5. Experimental conditions

Since the behavior of the cavern was already investigated

Fig. 6. (a) Measured thermal diffusivity and (b) calculated thermal conductivity of
the PCM. Symbols indicate the average values of three measurements per sample
for three samples (nine measurements per temperature) and the vertical bars
indicate the corresponding standard deviations. The dashed lines indicate the
polynomial regression of the measured values.
than its value at 50
C. The decrease in the thermal diffusivity at
temperatures higher than 350
C might be linked to the second-
order phase transition to which we attributed the change in slope
of the specific heat capacity in Fig. 5(a).
Fig. 6(b) shows that the thermal conductivity at 450
C is about
35% larger than its value at 50
C. The increase of the conductivity
with temperature is consistent with the behavior of aluminum
alloys, see Drezet et al. [24]. The average thermal conductivity in
the temperature range 50–450
C is k ¼ 151 W=mK.
The measured and calculated thermophysical properties are
summarized in Table 2.
4. Numerical model
through the ambient-temperature tests presented in Part 1, only
high-temperature tests were performed with the combined
sensible/latent storage. Furthermore, because the behavior of
the cavern was already discussed extensively in Part 1, it will not be
discussed further here.
The tests with the combined sensible/latent storage consisted
of four charging/discharging cycles. Before the first charging phase,
a so-called pre-charging was performed to approach steady cycling
conditions more quickly, as described in Part 1. The durations of the
pre-charging, charging, and discharging phases are listed in
Table 3. During the pre-charging phase, the inflow temperatures
to the latent and sensible storages, indicated in Fig. 2 by Tair,top,lat
and Tair,top,sens, respectively, and the mass flow rate varied as shown
in Fig. 7. In this and subsequent figures, light blue, white, and gray
backgrounds indicate pre-charging, charging, and discharging
phases, respectively.
Fig. 8 shows the air inflow and outflow temperatures of the
latent and sensible storages during the charging/discharging
cycles. During the charging phases of each cycle, the heater was
used at full power. However, because of the thermal inertia of the
heater and the pipes and because of the aforementioned thermal
losses from the pipe connecting the heater and the latent storage,
Tair,top,lat and therefore also Tair,top,sens are seen not to have reached
their maximum values immediately during the second, third, and
fourth charging phases. In the first charging phase, Tair,top,lat and
Tair,top,sens attained their maximum values because of the preceding
pre-charging phase.

The numerical model of the combined sensible/latent TES in the

cavern is based on the model of Geissbühler et al. [6], modified as Table 3
Operating conditions of the tests. Dtpc, Dtc, and Dtd denote the durations of the pre-
described in Part 1. The boundary conditions for the simulations
charging, charging, and discharging phases, respectively.
presented here are the measured air mass flow rate and inflow/
outflow temperatures to/from the storages, as indicated in Fig. 2. Cycle Dtpc (h) Dtc (h) Dtd (h) _ air,top,lat/sens
m,T
The sensible TES is treated as described in Part 1, assuming a 15% 1 70:36 1:30 1:30 Figs. 7 and 8
bypass flow with respect to the measured air mass flow rate. 2 – 1:29 1:31 Fig. 8
3 – 1:31 1:27 Fig. 8
The latent TES is treated as described by Geissbühler et al. [6] using
4 – 1:32 1:29 Fig. 8
the correlation of Žukauskas [25] with Crow = 1.0 for the heat-transfer
146 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152
Fig. 7. Measured air mass flow rate and inflow temperatures to the sensible and
latent TES during the pre-charging phase.
Fig. 8. Measured air temperature at the top (inlet during charging) and bottom
(inlet during discharging) of the latent TES, measured air temperature at the top
(inlet during charging) and bottom (inlet during discharging) of the sensible TES,
and measured air mass flow rate during the four consecutive charging/discharging
cycles.
During the discharging phases, air flows from the cavern into
the sensible storage. Therefore, Tair,bot,sens is approximately equal to
the air temperature in the cavern, see Fig. 8, and decreases with
time due to the expansion of the air in the cavern. From the
sensible storage, air enters the latent storage with a temperature of
Tair,bot,lat that also decreases with time, see Fig. 8.
The air mass flow rates during the charging and discharging
phases, denoted by m _ c and m
_ d , respectively, are shown in Fig. 8.
During charging, the mass flow rate is dependent on the
compressor, while during discharging, it is controlled through
outlet valves.
6. Results and discussion
6.1. Cycling experiments
6.1.1. Pre-charging phase
Fig. 9 shows the comparison between the measured and
computed temperature profiles in the latent and sensible TES units
at several times during the pre-charging phase. In this and the
following figures, the melting range DTm of the PCM is indicated by
a red background. First, we find that good agreement between
experimental and numerical results for both TES units is achieved,
as for the sensible-only TES in Part 1, meaning that the heat-
transfer model can accurately predict the behavior of a pilot-scale
combined sensible/latent TES. Second, we note that the tempera-
ture evolution over time differs between the two TES units. A
thermocline develops in the sensible TES, but the latent TES shows
little or no thermal stratification.
The lack of stratification in the latent TES is already visible after
10 h of pre-charging, when the average measured temperature is
357
C and the difference between the highest and lowest
measured temperatures is only about 10
C. With the exception
of the time when the PCM melts, after 10 h this difference is never
larger than 22
C. We attribute the lack of stratification partly to the
slow heating rate during the pre-charging phase. For example,
Fig. 7 shows that between 9:27 and 10:00, the temperature of the
air entering the latent TES varies between 363
C and 350
C. The
average air inflow temperature is therefore close to the above-
mentioned average measured temperature of 357
C after 10 h. The
Fig. 9. Temperature profiles as a function of the vertical position for latent TES (top)
and sensible TES (bottom) at several times during the pre-charging phase. Note: The
shaded red area indicates the melting range of the PCM; the RTD measuring T11
failed at t = 60.2 h; the symbols in the bottom plot represent the air temperature
measured by the RTDs located inside the sensible TES, as shown in Fig. 5 of Part 1.
(For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.)
 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152 147

lack of stratification is attributed only partly to the slow heating

rate because, as will be shown in Section 6.1.2, limited stratification
is observed also during the cycling phase in which the latent TES is
exposed to higher heating rates. As a result, the lack of
stratification is put down also to the heat transfer between the
air and the PCM being effective, which is supported by the
simulations that predict the differences between the air and PCM
temperatures never to exceed 5
C with the exception of the time
when the PCM melts.
At 27 h, the measured temperatures of layers 11 and 14 are
within the melting range, suggesting that the PCM located in the
bottom half of the tank is in the melting phase, whereas the
measured temperatures of layers 2, 5, and 8 are already higher than
the melting range. However, the model predicts the PCM in layers
2, 5, 8, 11, and 14 to be in the melting process.
After 70 h of pre-charging, the average measured temperature
in the latent TES is 561
C and the difference between the highest
and lowest measured temperatures is 7
C. Since the latent TES has
been charged for 44:49 h with an average air inflow temperature of
553
C, see Fig. 7, the storage can be considered to be fully charged.
At the end of pre-charging, the thermal energies stored in the
latent and sensible TES units, computed from the simulation
results, are 167.4 kWhth and 4.5 MWhth, respectively, correspond-
ing to 98% and 39% of their maximum storage capacities.
Fig. 9 shows that at the end of the pre-charging phase, the
temperature of the air flowing out of the latent TES is 557
C, while
that of the air flowing into the sensible TES reaches only 530
C. The
difference is ascribed (1) to thermal losses from the pipe
connecting the two storages, and (2) to air losses from the steel
cover of the sensible TES, which were discussed already in Part 1.
Since the behavior of the sensible TES was thoroughly studied in
Part 1, it will not be discussed further.
Fig. 10. Temperature evolution during the four charging/discharging cycles as a
6.1.2. Cycling phase function of time at four locations in the latent TES.
Fig. 10 shows the computed and measured PCM and air
temperatures as a function of time during the four charging/
discharging cycles for layers 2, 5, 8, and 14. As indicated in Fig. 3, for time is highly variable. On the one hand, in layer 14 the
layer 2 the measured temperature is that of the PCM, whereas for discrepancies between the simulated and measured air temper-
layers 5, 8, and 14 the measured temperature is that of the air. We atures over the four charging–discharging cycles are very small. On
make three observations from Fig. 10. the other hand, in layer 2 the maximum discrepancy between the
First, we observe that the measured PCM temperatures in layer measured and simulated PCM temperatures is about 43
C.
2 at the ends of the discharging phases decrease with successive Interestingly, the simulated air temperatures approximate the
cycles. At the end of the first discharging phase, the temperature of measured PCM temperatures in layer 2 more closely than the
the PCM has only just entered the melting range and therefore the simulated PCM temperatures. The discrepancies between the
PCM has only just begun solidifying. At the ends of the second and measurements and the simulations might be due to approxima-
third discharging phases, the temperature of the PCM is well tions in the model, uncertainties in the experimental conditions,
within the melting range and hence the PCM is partially solidified. and uncertainties in the thermophysical properties. The effect of
At the end of the fourth discharging phase, the PCM temperature is uncertainties in the thermophysical properties on the simulated
below the melting range, indicating that the PCM is fully solidified. temperatures will be quantified through a sensitivity analysis in
The trend of decreasing PCM temperatures at the ends of the Section 6.3.
discharging phases is visible not only in the measurements for The third and final observation concerns thermal stratification.
layer 2, but also in the simulations for layers 5, 8, and 14. For In Section 6.1.1, it was shown that during the pre-charging phase,
example, the simulations for layer 14 predict that the PCM there is little or no stratification in the latent TES. The lack of
temperature barely enters the melting range during the charging stratification is reflected in Fig. 10 through all layers being at the
phases of the third and fourth cycles. As a consequence of the same temperature at the beginning of the first discharging phase.
effective heat transfer between the PCM and the air, the measured At the end of the first discharging phase, however, the simulated
air temperatures at the ends of the discharging phases for layers 5, air temperatures in layer 14 are already about 43
C lower than
8, and 14 decrease with each successive cycle just like the those in layer 2. At the end of the fourth discharging phase, the
measured PCM temperatures. difference between the air temperatures in these two layers has
The second observation relates to the agreement between the grown to 83
C. The development of thermal stratification can be
measurements and simulations. The overall agreement is reason- seen more clearly in Fig. 11, which presents the temperature
ably good in the sense that general trends, such as the above- profiles as a function of the axial coordinate at the ends of the
mentioned decreases of the PCM and air temperatures at the ends charging and discharging phases for each cycle. The profiles at the
of the discharging phases, are predicted by the simulations with ends of the charging phases show that the portion of the latent TES
good accuracy. Conversely, the agreement between measured and with temperatures above the melting range decreases with each
simulated PCM and air temperatures in a specific layer at a specific cycle. Similarly, the profiles at the ends of the discharging phases
148 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152
Fig. 11. Temperature profiles in the latent TES as a function of the axial coordinate at the ends of the charging and discharging phases.
show that the portion of the latent TES with temperatures below
the melting range increases with each cycle. In other words, in each
cycle a smaller fraction of the PCM melts and solidifies and
therefore the latent TES behaves more like a sensible TES.
The results presented above indicate a decrease in the
performance of the latent TES with each cycle. The decrease in
performance is ascribed to four factors, in descending order of
importance: (1) the thermal losses from the pipe connecting the
latent and sensible TES units and the air leakages from the cover
of the sensible TES, (2) the experimentally measured mass flow
rates being smaller than the nominal mass flow rate used during
the design of the latent TES, (3) the experimentally measured
average mass flow rates during the discharging phases being
smaller than those during the charging phases, and (4) the non-
constant air inflow temperature during charging caused by the
start-up times of the heater. The effect of these factors was
assessed through simulations, the results of which are not shown
for brevity.
Despite the reduction in its performance, the latent TES can
effectively reduce the decrease in the air outflow temperature
during discharging. This may be deduced by comparing the
measured air inflow and outflow temperatures during discharging,
i.e., Tair,bot,lat and Tair,top,lat. During discharging, these temperatures
can be interpreted as the air outflow temperatures from a sensible-
only and from a combined-sensible/latent TES, respectively. Fig. 12
shows that Tair,bot,lat drops by about 99
C during the first
discharging phase, while the drop in Tair,top,lat is reduced to 35
C
thanks to the latent heat released by the PCM. Therefore, the
combined TES as used in an AA-CAES plant can provide smaller
variations in the temperature of the air flowing to the turbine,
which should be beneficial for the overall plant efficiency.
6.2. Post-experiment inspection and analysis
After the experimental campaign, we inspected the latent TES
to assess the thermal and mechanical stability of the tank and the
tubes encapsulating the PCM. The inner and outer shells and the
insulation were observed to have withstood the thermal cycling
well. After opening the inner shell, we inspected the tubes and
found solidified pieces of PCM such as those shown in Fig. 13. Based
 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152 149

pieces weighed between 2.8 mg and 3.7 mg and are denoted by A,

B, and C in the following.
Fig. 14 compares the measured heat flows during melting for
the PCM samples after thermal cycling (indicated by the solid lines)

Fig. 12. Measured air temperature at the inlet and outlet of the latent TES during the
first discharging phase.

Fig. 13. Pictures of leaked PCM after experiments.

on the locations of these pieces, we concluded that the PCM had

leaked from either the welding seams or the two RTD insertion
points. On one tube, a small hole was found in a welding seam. The
PCM pieces were collected and found to correspond to about 0.2%
of the total PCM mass.
While a more detailed analysis of the PCM, the encapsulation,
and the growth of intermetallic layers is underway, the recovered
PCM pieces provided a valuable first opportunity to assess the
effect of thermal cycling on the PCM. To this end, the DSC analysis
Fig. 14. Measured heat flows during melting for PCM samples after thermal cycling
described in Section 3.2 was repeated for three PCM pieces that (solid lines) and comparison to measured heat flow during melting before thermal
differed in color and texture. The samples extracted from these cycling (dashed line).
150 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152
to the measured heat flow during melting before thermal cycling
(indicated by the dashed line). We make four observations.
First, it is noted that all heat-flow curves after cycling are
qualitatively similar and that they differ from the heat-flow curve
before cycling. This indicates that the phase composition of the
PCM changes upon cycling and that degradation and/or phase
segregation occur.
Second, compared to the heat-flow curve before cycling, the
heat-flow curves after cycling show a larger low-temperature peak
and a smaller eutectic peak. The low-temperature peak after
cycling could be caused by the same solid–solid phase transition as
before cycling (see Section 3.2) or phase transitions of Al–Fe
compounds. These compounds could have originated from the
dissolution of stainless steel from the encapsulation into the liquid
PCM. Corrosion processes would also explain the decrease of the
eutectic peak through the loss of the eutectic phase and the
formation of Al–Fe compounds.
Third, by integrating the heat-flow curves, the heats of fusion of
samples A, B, and C are found to be 374 kJ/kg, 386 kJ/kg, and 326 kJ/
kg respectively. Relative to the heat of fusion prior to cycling, this
corresponds to reductions of 7.4%, 4.5%, and 19.3%, respectively.
Finally, we observe that the heat-flow curves of samples A and B
are closer to each other than either of them is to the heat-flow curve of
sample C. This suggests that samples A and B have a very similar phase
composition and that it differs from that of sample C. Therefore, at
least two alloys appear to have formed during thermal cycling.
In addition to the DSC analysis, the density was determined as
for the PCM before cycling for two samples each extracted from the
same PCM pieces from which samples A and C were extracted. The
average density of the four samples was found to be only 0.5%
smaller than the density of the PCM before cycling. The standard
deviation of the densities of the four samples was 1.4% of the
average density.
The DSC results show that the PCM is not stable under thermal
cycling. The lack of stability is attributed to the off-eutectic
composition of and impurities in the PCM as well as to corrosion
phenomena between the liquid PCM and the stainless-steel
Fig. 15. Results of sensitivity study in terms of Tair,top,lat during the fourth discharging phase. “Ref.” indicates the values before thermal cycling.
encapsulation. We assume that other thermophysical properties,
such as the mid-temperature of the melting range, the melting
range, and the specific heat capacity, are also affected by thermal
cycling. Changes in the thermophysical properties may have
contributed to the decrease in the performance of the latent TES in
addition to the factors listed in Section 6.1.2.
6.3. Sensitivity study
To assess the degree to which changes in the thermophysical
properties of the PCM might explain the discrepancies between the
experimental and numerical results, a sensitivity study was
performed. In the sensitivity study, the simulations of the pre-
charging and cycling phases were repeated by varying one
thermophysical property at a time relative to its value before
thermal cycling, see Table 2. Variations in the following properties
were studied: heat of fusion, specific heat capacity, mid-tempera-
ture of the melting range, and melting range. For the specific heat
capacity, the variations were applied to both the solid- and liquid-
phase specific heat capacities. The effect of the PCM melting range
was investigated in two ways. First, by keeping the range constant
and varying the mid-temperature of the range between 512 and
524
C in increments of 3
C. Second, by varying the range by 25%
and 50% with respect to its value before cycling and keeping its
mid-temperature constant.
The results of the sensitivity study are presented in Fig. 15 in
terms of Tair,top,lat during the fourth discharging phase. It is
immediately apparent that variations in the heat of fusion have the
largest impacts and that variations in the specific heat capacity, the
mid-temperature of the melting range, and the melting range have
limited impacts.
It is interesting to observe that the simulations and experiments
agree well when the heat of fusion is decreased by 20% relative to
its value prior to thermal cycling. That such large decreases are
plausible is clear from Section 6.2, in which post-experiment
analysis of the PCM showed decreases in the heat of fusion of up to
19.3%. Because a decrease of 20% is larger than the mean of the
 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152
measured decreases of 10.4%, we assume that the decrease in the
heat of fusion is not the sole explanation for the discrepancies
between the experimental and numerical results.
Additional results, not presented for brevity, show that
changes of up to 20% of the heat-transfer coefficient and the
PCM thermal conductivity have small impacts, while changes of
up to 20% of the PCM density have significant impacts. In
particular, a decrease of 20% in the density relative to the value
measured before cycling led to significantly improved agreement
between the simulations and experiments. This result appears
irrelevant because we reported in Section 6.2 that our measure-
ments did not find significant changes in the density after thermal
cycling. However, in our numerical model the PCM density
appears only as a multiplier of the PCM volume, so relative
variations in the PCM density are equivalent to relative variations
in the PCM volume. Variations in the PCM volume could have
been caused by some tubes having been filled with less PCM and/
or by tubes having been shorter than specified. Nevertheless, it is
very unlikely that these two causes can account for a 20%
decrease in the PCM volume.
The sensitivity study shows that the thermophysical properties
of the PCM have a significant impact on reducing the discrepancies
between simulations and experiments. Variations in the heat of
fusion of the PCM that give improved agreement are plausible
given the measured changes reported in Section 6.2, but it seems
unlikely that they are the sole explanation for the discrepancies.
Instead, it seems likely that uncertainties in the total volume of the
PCM are an important factor also. Although other factors that are
not included in the model, such as radial effects and the thermal
inertia of the steel support structure, might have contributed to the
discrepancies, they are believed to be less important.
7. Summary, conclusions, and further work
A pilot-scale AA-CAES plant with a combined sensible/latent TES
was built in an unused tunnel. The combined sensible/latent TES
differs from the sensible-only TES presented in Part 1 through a
latent TES with maximum storage capacity of 171.5 kWhth. The latent
TES consists of a steel tank containing 296 stainless-steel tubes
encapsulating an Al–Cu–Si alloy. The combined sensible/latent TES
was investigated using four charging/discharging cycles with
durations of about 3 h each and air charging temperatures of up
to 566
C.
The main conclusions are:
1. The latent TES reduced the decrease in the air outflow
temperature during discharging, confirming prior work on
combined sensible/latent TES at the laboratory scale and
demonstrating the potential of sensible/latent TES as an attractive
option for industrial-scale high-temperature storage.
2. The performance of the latent TES decreased with each cycle. The
decrease was traced primarily to thermal losses from the pipe
connecting the latent and sensible TES units, to air leakages from
the cover of the sensible TES, and to mass flow rates that were
smaller in the experiments than during the design of the latent TES.
3. Although minor leaks of the PCM were detected after the
experiments, the latent TES structure and the PCM encapsula-
tion exhibited good thermal and mechanical stability to
temperature and pressure variations. The leaks were traced to
the welding seams between the tubes and the caps that
encapsulated the PCM and to the holes drilled through the tubes
to insert RTDs into the PCM.
4. The PCM exhibited degradation and/or phase segregation upon
thermal cycling, resulting in a decrease of its heat of fusion. These
changes were attributed to the initial off-eutectic composition of
and impurities in the PCM as well as to corrosion phenomena
151
between the liquid PCM and the stainless-steel encapsulation at
high temperatures.
5. Simulations predicted the behavior of the combined sensible/
latent TES with good overall accuracy. The discrepancies were
attributed primarily to changes in the heat of fusion of the PCM
due to thermal cycling and uncertainties in the volume of the
PCM contained in the latent TES.
From the experience gathered with the pilot-scale combined
sensible/latent TES, it is clear that further work should focus on the
PCM and the encapsulation. For the PCM, it is imperative that
eutectic compositions with minimal impurities can be guaranteed
when ordering quantities commensurate with pilot-scale or
industrial-scale applications. For the encapsulation, further work
must address the large-scale production of leak-proof encapsula-
tions and continue the development of coatings that prevent
corrosion phenomena.
Acknowledgments
The authors gratefully acknowledge funding by the Swiss
National Science Foundation under the National Research Program
70 (grant no. 407040_153776), the Swiss Federal Office of Energy
(SI/501001-01), the Commission for Technology and Innovation
through the Swiss Competence Center for Energy Research on Heat
and Electricity Storage, and the European Union under the 7th
Framework Program (SFERA-II, grant no. 312643). The authors are
grateful to Luciano Serio for the engineering support during the
design, building, and operational phases of the combined TES, to
Mattia Fransioli and Vittorio Lo Vaglio for the support during the
design of the storage, to Elke Hempel and Urs Jörimann for
performing DSC and TOPEM1 analysis on the PCM and for helpful
discussions, and to Mihai Stoica for valuable inputs on metal
physics.
References
[1] X. Luo, J. Wang, M. Dooner, J. Clarke, Overview of current development in
electrical energy storage technologies and the application potential in power
system operation, Appl. Energy 137 (2015) 511–536.
[2] M. Budt, D. Wolf, R. Span, J. Yan, A review on compressed air energy storage:
basic principles, past milestones and recent developments, Appl. Energy 170
(2016) 250–268.
[3] A. Sciacovelli, Y. Li, H. Chen, Y. Wu, J. Wang, S. Garvey, Y. Ding, Dynamic
simulation of adiabatic compressed air energy storage (A-CAES) plant with
integrated thermal storage-link between components performance and plant
performance, Appl. Energy 185 (2017) 16–28.
[4] L. Geissbühler, V. Becattini, G. Zanganeh, S. Zavattoni, M. Barbato, A.
Haselbacher, A. Steinfeld, Pilot-scale demonstration of advanced adiabatic
compressed air energy storage, Part 1: Plant description and tests with
sensible thermal-energy storage, J. Energy Storage 17 (2018) 129–139.
[5] G. Zanganeh, R. Khanna, C. Walser, A. Pedretti, A. Haselbacher, A. Steinfeld,
Experimental and numerical investigation of combined sensible-latent heat
for thermal energy storage at 575
C and above, Sol. Energy 114 (2015) 77–90.
[6] L. Geissbühler, M. Kolman, G. Zanganeh, A. Haselbacher, A. Steinfeld, Analysis
of industrial-scale high-temperature combined sensible/latent thermal
energy storage, Appl. Therm. Eng. 101 (2016) 657–668.
[7] P. Galione, C.D. Pérez-Segarra, I. Rodríguez, S. Torras, J. Rigola, Multi-layered
solid-PCM thermocline thermal storage for CSP. Numerical evaluation of its
application in a 50 MWe plant, Sol. Energy 119 (2015) 134–150.
[8] B.C. Zhao, M.S. Cheng, C. Liu, Z.M. Dai, Thermal performance and cost analysis
of a multi-layered solid-PCM thermocline thermal energy storage for CSP
tower plants, Appl. Energy 178 (2016) 784–799.
[9] M. Yan, Z. Fan, Review durability of materials in molten aluminum alloys, J.
Mater. Sci. 36 (2) (2001) 285–295.
[10] V. Yeremenko, Y.V. Natanzon, V.I. Dybkov, The effect of dissolution on the
growth of the Fe2Al5 interlayer in the solid iron–liquid aluminium system, J.
Mater. Sci. 16 (7) (1981) 1748–1756.
[11] S.R. Binder, S. Haussener, Characterization and Growth Kinetics of
Intermetallic Phases in the Liquid Al–13Si/316L Stainless Steel Diffusion
Couple, (2017) in preparation.
[12] K.G. Allen, T.W. von Backström, D.G. Kröger, A.F.M. Kisters, Rock bed storage for
solar thermal power plants: rock characteristics, suitability, and availability,
Sol. Energy Mater. Sol. Cells 126 (2014) 170–183.
152 V. Becattini et al. / Journal of Energy Storage 17 (2018) 140–152
[13] R. Tiskatine, A. Eddemani, L. Gourdo, B. Abnay, A. Ihlal, A. Aharoune, L.
Bouirden, Experimental evaluation of thermo-mechanical performances of
candidate rocks for use in high temperature thermal storage, Appl. Energy 171
(2016) 243–255.
[14] V. Becattini, T. Motmans, A. Zappone, C. Madonna, A. Haselbacher, A.
Steinfeld, Experimental investigation of the thermal and mechanical stability
of rocks for high-temperature thermal-energy storage, Appl. Energy 203
(2017) 373–389.
[15] M.M. Kenisarin, High-temperature phase change materials for thermal energy
storage, Renew. Sustain. Energy Rev. 14 (3) (2010) 955–970.
[16] A.M. Gasanaliev, B.Y. Gamataeva, Heat-accumulating properties of melts, Russ.
Chem. Rev. 69 (2) (2000) 179–186.
[17] J. Schawe, T. Hütter, C. Heitz, I. Alig, D. Lellinger, Stochastic temperature
modulation: a new technique in temperature-modulated DSC, Thermochim.
Acta 446 (1) (2006) 147–155.
[18] M.J. Chase, NIST-JANAF thermochemical tables, J. Phys. Chem. Ref. Data.
Monogr. (9) (1998).
[19] C. Bergman, K. Komarek, Heat capacity of liquid alloys, Calphad 9 (1) (1985) 1–14.
[20] W.J. Boettinger, U. Kattner, K. Moon, J. Perepezko, DTA and Heat-Flux DSC
Measurements of Alloy Melting and Freezing, US Govern. Printing Office, 2006.
[21] B. He, V. Martin, F. Setterwall, Phase transition temperature ranges and storage
density of paraffin wax phase change materials, Energy 29 (11) (2004) 1785–1804.
[22] H. Mehling, L.F. Cabeza, Heat and Cold Storage with PCM, Springer, 2008.
[23] M. Kheradmand, M. Azenha, J.L. de Aguiar, K.J. Krakowiak, Thermal behavior of
cement based plastering mortar containing hybrid microencapsulated phase
change materials, Energy Build. 84 (2014) 526–536.
[24] J.-M. Drezet, M. Rappaz, G.-U. Grün, M. Gremaud, Determination of
thermophysical properties and boundary conditions of direct chill-cast
aluminum alloys using inverse methods, Metall. Mater. Trans. A 31 (6) (2000)
1627–1634.
[25] A. Žukauskas, Heat transfer from tubes in crossflow, Adv. Heat Transfer 8
(1972) 93–160.
[26] M.A. Tomi c, P.Z. Živkovi
c, M.V. Vukic, G.S. Ili
c, M.M. Stojiljkovi
c, Numerical
study of perforated plate convective heat transfer, Therm. Sci. 18 (3) (2014)
949–956.