Development of a Molten-Salt

Thermocline Thermal Storage

James E. Pacheco System for Parabolic Trough
e-mail: [email protected]

Steven K. Showalter
Plants
Thermal storage improves the dispatchability and marketability of parabolic trough
William J. Kolb power plants allowing them to produce electricity on demand independent of solar col-
lection. One such thermal storage system, a thermocline, uses a single tank containing a
Sandia National Laboratories,1
fluid with a thermal gradient running vertically through the tank, where hotter fluid (lower
Solar Thermal Technology Department,
density) is at the top of the tank and colder fluid is at the base of the tank. The thermal
Albuquerque, NM 87185-0703
gradient separates the two temperature potentials. A low-cost filler material provides the
bulk of the thermal capacitance of the thermal storage, prevents convective mixing, and
reduces the amount of fluid required. In this paper, development of a thermocline system
that uses molten-nitrate salt as the heat transfer fluid is described and compared to a
two-tank molten salt system. Results of isothermal and thermal cycling tests on candidate
materials and salt safety tests are presented as well as results from a small pilot-scale (2.3
MWh) thermocline. 关DOI: 10.1115/1.1464123兴

Introduction low-pressure steam, and returned to the bottom of the tank. The
low-pressure steam was fed to the turbine to make electricity or
Thermal storage for parabolic-trough power plants offers the
potential to dispatch electricity without fossil fuel backup. The was used as an auxiliary steam supply. Since the thermal storage
value of the electricity is often higher during peak periods. A plant system used Caloria® , its operating temperature was between
that can meet the peak demand independent of the solar resource 218°C and 302°C. The tank was designed for a thermal capacity
has the potential to generate more revenue. of 182 MWht . It used 6170 tonnes (1 tonne⫽1000 kg) of rock
The Solar Electric Generating Station 共SEGS兲 I plant had a and sand and 906 m3 of Caloria® . Physically, the storage tank
two-tank thermal storage system that used Caloria® heat transfer stood 13.3-m tall and 18.2-m diameter 关1兴. Because of the limited
fluid. This fluid was pumped through the collector field. The stor- upper temperature capability of the thermal storage system, the
age tanks were conventional atmospheric tanks. Caloria® is a liq- Rankine-cycle conversion efficiency was only about 21% 关2兴. De-
uid at atmospheric pressures when its temperature is below about spite the limitations of Caloria® , the thermocline storage concept
315°C. The SEGS I thermal storage system operated at tempera- worked well. It was able to establish a thermal gradient and main-
tures between about 175°C and 295°C. The hot Caloria® oil was tain the gradient until it was used. Heat loss through the tank walls
pumped through a steam generator to make saturated steam. Fossil was acceptable. Because of its upper temperature limitations,
fuel was used to superheat the steam to generate electricity. Caloria® is not practical for the current technology troughs.
The Solar One central receiver pilot plant also used Caloria® , The current baseline design for a SEGS plant uses Therminol
but in a thermocline tank. A thermocline tank is one that uses a VP-1® heat transfer fluid in the collector field. Therminol VP-1®
single tank to store thermal energy. A thermal gradient separates has been used successfully for 10 years in the SEGS II to
the hot from the cold fluid. A low-cost filler material is used to IX plants located in the Mojave Desert of California. Therminol
displace the higher-cost liquid. The filler material as well as buoy- VP-1® has a low freezing point (12°C) and is stable up to 400°C
ant forces help to maintain the thermal gradient. When the system 关3兴. The higher temperature capability of Therminol VP-1® 共rela-
is charged, cold fluid is drawn from the bottom, heated as it passes tive to Caloria® 兲 allows these plants to use higher pressure, higher
through a heat exchanger 共heated with the receiver heat transfer temperature, and more efficient Rankine turbines. It is difficult to
fluid兲 and returns to the top of the tank. When the tank is dis-
use Therminol VP-1® as a thermal storage media because its va-
charged, hot fluid is drawn from the top, cooled as it passes
por pressure is too high 共⬎1 MPa at 400°C兲 to practically store it
through a heat exchanger 共to transfer heat for steam generation兲,
in any significant quantity at its upper temperature. Multiple thick-
and returns to the bottom of the tank. In the Solar One ther-
mocline storage system, steam from the receiver passed through a walled pressure-vessels would have to be used to store the hot oil,
which would be too costly to be practical. Dispatching power in
heat exchanger, which heated the Caloria® oil drawn from the
these plants is done with natural gas-fired boilers.
bottom of the tank. The heated Caloria® returned to the top of the
Nexant—a subsidiary of Bechtel—and Pilkington performed
tank where it transferred heat to the rock and sand 共the filler
analyses of potential near-term thermal storage options for para-
material兲. During a discharge cycle, hot Caloria® was pumped
bolic troughs 关4,5兴. For a near-term solution of thermal storage for
from the top of the tank, through a steam generator that made
parabolic troughs, Nexant wanted to find a storage system that
1
could work with the current state-of-technology trough. Several
Sandia ia a multiprogram laboratory operated by Sandia Corporation, a Lockheed
Martin Company, for the United States Department of Energy under Contract DE- indirect thermal storage options were considered where heat from
AC04-94AL85000. the collector field was transferred from the Therminol® oil to an-
Contributed by the Solar Energy Division of THE AMERICAN SOCIETY OF ME- other thermal storage media, such as molten salt, which could be
CHANICAL ENGINEERS for publication in the ASME JOURNAL OF SOLAR ENERGY
ENGINEERING. Manuscript received by the ASME Solar Energy Division, Nov. stored at atmospheric pressure.
2000; final revision, May 2001. Associate Editor: T. Mancini. Because of the favorable experience of the two-tank molten-salt

Journal of Solar Energy Engineering Copyright © 2002 by ASME MAY 2002, Vol. 124 Õ 153

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/19/2013 Terms of Use: http://asme.org/terms

 Fig. 1 Schematic of a two-tank thermal storage system inte-
grated into a trough plant

Fig. 2 Temperature gradient every 0.5 h during a charging

cycle of a 688 MWht thermocline tank 16 m tall and 34.3 m di-
thermal storage system used at the Solar Two demonstration plant ameter filled with quartzite and sand
关6兴, its low risk, and cost-effectiveness, Nexant selected this de-
sign as a baseline. In this design to charge the storage system, hot
Therminol® oil coming from the collector field would transfer ⳵T f 共 mC p 兲 f ⳵ T f
heat through an oil-to-salt heat exchanger, and then it would re- 共␳Cp兲 f␧ ⫽⫺ ⫹h v 共 T b ⫺T f 兲 (1)
turn to the collector field to be heated again. Salt pumped from the ⳵␶ A ⳵y
cold tank would be heated as it passed through the oil-to-salt heat ⳵Tb
exchanger and would flow into the hot tank. To discharge the 共 ␳ C p 兲 b 共 1⫺␧ 兲 ⫽h v 共 T f ⫺T b 兲 (2)
storage system, hot salt would be pumped from the hot tank and ⳵␶
would transfer its heat through the same oil-to-salt heat exchanger where f subscripted variables denotes fluid properties and b de-
and return to the cold tank. Therminol® oil would be heated by the notes bed properties. The model consisted of a one-dimensional,
heat exchanger, then pass through a steam generator to make finite difference representation of the above equations. The tank
steam. A schematic of the two-tank molten-salt system integrated was divided into equal horizontal slices, typically 0.2 m thick. An
into a trough plant is shown in Fig. 1. Nexant estimated the capital initial vertical temperature distribution was specified. The model
cost of a 688 MWht two-tank molten-salt system used with the then calculated the local heat transfer coefficients, then calculated
current state-of-technology troughs to be about $31/kWht includ- the local temperatures of the fluid and the bed using the partial
ing the nitrate salt inventory, the installed tanks, and the oil-to-salt differential equations. It used the modified Euler’s method to cor-
heat exchangers. rect the predicted fluid and bed temperatures. The model would
In a recent roadmap of trough technology 关7兴, the long-term repeat the above process at fixed time intervals, typically every 4
goal for the capital cost of a thermal storage system is $10/kWht . seconds.
In order to realize this cost reduction, innovative development is All the code was written in Visual Basic using Microsoft Excel
needed. and ran as a macro. The following parameters could be specified:
The goal of this project was to improve on the economics of an bed material, fluid, void fraction, tank height, tank diameter, filler
indirect thermal storage system by assessing the feasibility of us- material and diameter, surface area shape factor, time step, differ-
ing molten salt in a thermocline. We chose to study a thermocline ential height thickness, initial temperature profile, time dependent
system because it has the potential to reduce the cost of the ther- inlet fluid temperature, and flow rate.
mal storage system and it can dispatch thermal energy at nearly a Figure 2 shows a plot of the thermal gradient during a charging
constant temperature over most of its discharge cycle. Both labo- cycle for a 688 MWht thermocline tank. Figures 3 and 4 show the
ratory and engineering-scale experiments were conducted. To inlet and outlet temperatures of the nitrate salt 共going into or com-
come up with a feasible molten salt thermocline system, we broke ing out of the tank兲 during charging and discharging cycles. As
the development activity into three major areas: can be seen from Fig. 2, the gradient requires a certain amount of
vertical space. Because of the height required for the gradient,
1. Modeling–where we simulated the performance and esti- taller tanks with smaller diameters are favored over shorter tanks
mated the economics of a thermocline system, with larger diameters. The tallest tanks that can practically be
2. Filler Evaluation: Isothermal Testing–where we screened fabricated have a bed height of 16 m. Note that during the charg-
potential candidate filler materials in isothermal baths of
molten salt—and Thermal Cycling Testing—where we took
the most promising fillers and subjected them to repeated
thermal cycling, and
3. Thermocline Test–where we built a small 共2.3 MWh兲 ther-
mocline system to validate performance and evaluate prac-
tical operational issues of a thermocline system.
Also as part of this effort, we evaluated potential low melting-
point salts and evaluated salt-safety issues related to using an
organic heat-transfer fluid alongside oxidizing nitrate salts.

Modeling
To understand the parameters that affect the performance and
economics of a thermal storage system, a model was developed.
The thermal behavior of the tank is simulated by the Schumann Fig. 3 Salt temperature going into „upper line… and coming out
equations, which describe the heat transfer between the fluid and a „lower line… of a 688 MWht thermocline tank during a charge
packed bed 关Eqs. 共1兲 and 共2兲兴 cycle „same tank as in Fig. 2…

154 Õ Vol. 124, MAY 2002 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/19/2013 Terms of Use: http://asme.org/terms

 The witherite and barite samples reacted with the Ca(NO3 ) 2
during their initial 10 hour exposure to Hitec XL® . Bauxite, ill-
menite, and KS limestone samples crumbled after exposure to the
nitrate salt. Anhydrite experienced significant weight loss. Carbo-
rundum and taconite gained weight, most likely due to their po-
rous structure and the difficulty in removing trace amounts of salt.
The most successful candidates were quartzite, taconite, marble,
NM limestone, apatite, corrundum, scheelite, and cassiterite. Be-
cause taconite, marble, NM limestone, and quartzite are available
in bulk quantities for a reasonable price, we focused our attention
on these materials to conduct thermal cycling tests, which are
discussed later in this paper.
Fig. 4 Salt temperature coming out of „upper curve… and going 2 Nitrate Salt Mixtures. In addition to screening filler ma-
into „lower curve… a 688 MWht thermocline tank during a dis- terials, we evaluated mixtures of nitrate salts. Studies had shown
charge cycle „same tank as in Fig. 2… that the freezing point and upper temperature limit of nitrate salts
are functions of the amount of each constituent 关9兴. Mixtures of
sodium nitrate and potassium nitrate, as well as mixtures of so-
ing cycle, the temperature of the cold salt entering the heat ex- dium nitrate, potassium nitrate, and calcium nitrate were evalu-
changer is constant until the gradient reaches the bottom of the ated.
tank. At that point the temperature coming out of the bottom of In the Solar Two project, solar salt was used because of its
the tank starts to rise. Similarly during a discharge cycle, the hot stability at temperatures up to 600°C, its low cost, its compatibil-
salt coming off the top of the tank is nearly constant until the ity with common piping and valve materials, and its favorable
gradient enters the top. At that point the hot salt temperature starts performance in tests conducted at Sandia National Laboratories.
to decay. Hitec XL® is shipped premixed dissolved in water. To remove
One way to quantify the practical storage capacity of a thermal the water, the solution must be heated to drive off the water. After
storage system is to compare it to the amount of energy the stor- the water is driven off, it has a freezing point of about 120°C and
age media could hold if the entire inventory were at its upper is stable to temperatures around 500°C.
temperature and discharged completely to its lower temperature.
This has been named the percent theoretical capacity. In a two- 3 Safety Issues Related to Using Nitrate Salt and Thermi-
tank molten salt system, the percent theoretical capacity is typi- nol® . To store energy for dispatch, heat from the oil exiting the
cally about 85% because each tank has a heel of salt 共about 1 m collector field would be transferred to the salt through an oil-to-
deep兲 which cannot be used for its storage capacity. In a ther- salt heat exchanger. Because the nitrate salt is classified as an
mocline storage system, the percent theoretical capacity is a func- oxidizer and the Therminol® oil is a fuel, there was concern that
tion of the tank height and is typically about 69% because of the having these two fluids in close proximity posed an unacceptable
space required by the thermal gradient. hazard in the event that a component failed, leaking oil into the
salt or vise versa. Earlier studies were conducted with Hitec® 共a
Screening of Potential Filler Materials and Nitrate mixture of sodium nitrate, potassium nitrate, and sodium nitrite兲
Salts. and organic liquids. Motor gasoline, cracked gasoline, gas oil, and
To develop a list of potential filler materials, we consulted with crude oil were separately mixed with sulfur and released below
a geologist and a nitrate salt expert who recommended several the surface of an open container of Hitec® at 593°C. In each case,
commonly mined minerals and crushed rock. We believed that the hydrocarbon burned. The principal reactions were between the
relatively pure materials could be found and delivered at a low vaporized hydrocarbons and the air above the salt bath and not
cost. The ideal filler material would be inexpensive and widely with the salt itself. Experimental data indicate that Hitec® is not
available, have a high heat capacitance, a low void fraction 共to explosive. Draw salt or any other mixtures of sodium nitrate and
reduce the amount of liquid required兲, be compatible with the potassium nitrate were expected to act similarly to Hitec® 关10兴.
nitrate salts, and be non-hazardous. Seventeen minerals were se- To verify that Hitec XL® behaved similarly to Hitec® , Thermi-
lected for testing. These included anhydrite, barite, bauxite, car- nol® was introduced into a beaker of Hitec XL® at approximately
borundum, cassiterite, corrundum, fluorapatite, hydroxyapatite, 400°C three different ways: on the surface, below the surface, and
illmenite, limestone 共from Kansas - KS兲, limestone 共from New trapped below the surface in a vial. In each case, the Therminol®
Mexico - NM兲, magnasite, marble, quartzite, scheelite, taconite, rapidly volatized when it came in contact with the hot nitrate salt.
and witherite. The oil did not burn under or on the surface of the salt. We
1 Compatibility of Common Minerals with Nitrate Salts repeated these tests with Caloria® oil. The first two tests behaved
A study had been performed on the compatibility of granite rock similarly to the Therminol® tests. However, in the third test after
and iron ore pellets 共taconite兲 in nitrate salts. It showed that taco- covering the beaker with insulation, trapped Caloria® oil vapor
nite should maintain its integrity for years at 550°C. Granite rock, caught fire and burned for approximately two minutes. The pres-
however, underwent significant erosion/dissolution, which could ence of a fuel, oxygen, and a heat source resulted in the small fire.
in time clog the system 关8兴. The fire was not a result of a chemical reaction with the nitrate
To evaluate our 17 candidate minerals, samples were weighed, salts that were used. In fact, any heat source would have yielded
photographed, and immersed in known masses of Hitec XL® ni- the same result. As a follow up test, we added an ignition source
trate salt—a commercially available salt containing calcium ni- 共spark plug兲 to simulate an external source of ignition such as a
trate, potassium nitrate and sodium nitrate 共circa. 42 wt% motor in a power plant to verify a fire would occur under the right
Ca(NO3 ) 2 /15% NaNO3 /43% KNO3 兲—then removed at specific conditions. We were able to ignite the Therminol® vapors above
intervals: 10, 100, and 1000 hours of exposure. A few candidates the salt after energizing the spark plug.
共limestone, marble, quartzite, and taconite兲 were also tested in The safety concerns in a plant that uses Therminol® and nitrate
solar salt 共a mixture of 60 wt% NaNO3 and 40 wt% KNO3 兲 and salt in a heat exchanger are not anymore dangerous than using
removed after 400 hours. The samples were washed, weighed, and Therminol® around other high-temperature heat sources. Acciden-
photographed again following removal. Samples of the salts were tally mixing the two components should not create combustion,
also taken for analysis of contaminants. but combining hot Therminol® vapors with oxygen from the air is

Journal of Solar Energy Engineering MAY 2002, Vol. 124 Õ 155

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/19/2013 Terms of Use: http://asme.org/terms

 produced, but not so much to be concerned with clogging the
system. Some of these may have been present in the initial mate-
rial. Taconite is acceptable for use in molten salt.
The quartzite rock held up remarkably well after 553 cycles.
The edges of individual rocks appeared sharp and unaffected.
Some pieces of quartzite went from a bluish to slightly reddish
color. The change in color was likely due to oxidation of iron that
is present in the quartzite. White pieces of quartzite did not
change color or appear affected at all. The silica filter sand be-
haved similarly to the quartzite. Some grains appear slightly red-
dish in color after exposure, but overall held up well.
Because of the successful experience of the quartzite in the
isothermal tests and quartzite and silica filter sand in the thermal
cycling tests, we chose these materials for the small thermocline
test. Also, quartzite and silica filter sand are widely available,
more so than taconite—the second choice.

Thermocline Test
Fig. 5 Schematic and photograph of thermal cycling system To evaluate the molten-salt thermocline concept on a larger
to test selected candidate materials scale, a small, 2.3 MWh, thermocline system was designed and
built. The goal of this test was to verify the operation and perfor-
mance of a molten-salt thermocline system on a larger-than-
laboratory scale. A mixture of quartzite and silica filter sand was
dangerous. The design of the thermal storage system should ad- chosen as the filler media because of the positive results from the
dress the unlikely event of having hot oil vapors released into the previous tests. Since the economics of a thermocline system are a
ullage space of a nitrate salt tank where oxygen could be present. strong function of the void fraction, tests were done to determine
the highest packing density of quartzite and silica filter sand. We
were able to achieve void fractions of 0.25 in half drum quantities
Thermal Cycling Tests by using a 2:1 ratio of quartzite to sand. The final void fraction
The purpose of these tests was to evaluate how well the mate- measurement of the rock/sand mixture in the tank based on actual
rials selected in the isothermal tests held up to thermal cycling quantities of salt used was 0.22.
conditions typical of a thermocline system. We focused on taco- 1 Description of the Thermocline Tank. A carbon steel
nite, NM limestone, marble, and quartzite for the thermal cycling tank, 6.1 m tall by 3.0 m diameter, was fabricated to American
experiments. We added filter sand to the test matrix because this Petroleum Institute 共API兲 Code 650 Appendix J–Shop Assembled
material is similar to quartzite and drastically reduces the void Storage Tank. A thermocouple tree was placed in the middle of the
fraction in a thermocline tank. tank with thermocouples every 15 cm, vertically. In addition, ther-
A small 10 cm diameter chamber 30 cm tall containing the filler mocouples were placed on three radial arms each at 0.36 m, 3.05
material was placed between two vessels, each 15 cm in diameter m, and 4.88 m from the bottom. The tank had a radial flow dis-
by about 2 m tall as shown in Fig. 5. One vessel contained hot salt tribution manifold at 10 cm from the bottom, which was con-
at 400°C and the other cold salt at 290°C. Hitec XL® salt was nected to a standpipe, housing the cold pump. The hot pump was
used in these tests. To cycle the temperature of the filler material, supported from below to pump salt from the top of the tank.
the vessel containing hot salt was pressurized with air to about 27 Seven electric mineral-insulated heat-trace cables, each rated at
kPa, forcing the hot salt through the rock bed, which heated it up 4.8 kWe, were wrapped on the exterior surface of the tank to
and increased the salt level in the cold vessel. After about a one provide heat input during the initial heating process and make up
half-hour soak, the pressure was bled off the hot vessel, which for heat loss. The tank was insulated with 23 cm of fiberglass
allowed cold salt to flow back through the filler material, cooling insulation on the sides and with 20 cm of calcium silicate ridged
it off. The cold salt was allowed to soak for approximately 21 hour block insulation on the top of the tank.
as well. This process continued 24 hours a day. Typically 15 to 19 The tank was filled with a mixture of quartzite and silica filter
cycles were actually completed per day. After a specified number sand. First, a 3-cm layer of sand was laid on the bottom of the
of cycles 共typically 350兲 the experiment was stopped and the tank, followed by a 20-cm layer of quartzite rock. A fine mesh
samples were inspected. stainless-steel screen was laid down to prevent the sand from
The NM limestone, a sedimentary rock consisting of hydrated working down to the manifold. Next, 5-cm layers of a 2:1 mixture
calcium carbonate, should be insoluble to nitrate salt. It fell apart of quartzite rock and sand was added and compacted in place with
after 365 cycles, becoming soft and looking like mud. This mate- pneumatic tampers until the filler level reached 5.2 m. In total, we
rial fared the worst and is unacceptable as a filler material. used 49.9 tonnes of quartzite and 21.9 tonnes of silica filter sand.
Marble is a metamorphic rock, which is calcium carbonate that All manifolds had a stainless steel screen installed on them to
has been exposed to tremendous heat and pressures. The marble prevent filler material from entering the piping system.
held together after cycling, but softened. The individual grains in The tank and inventory of rock and sand were initially heated
the samples appeared to grow in size and prominence during the by energizing the exterior heat trace and forcing hot air from a
test. At these temperatures, it is possible that the grains were re- small propane heater using a venturi into the lower manifold at the
crystallizing into larger crystals. Another explanation is that the bottom of the tank. Once the entire volume was above about
surface was being preferentially eroded by an unknown process 260°C, the salt melting process began. Because of time con-
during the test, making the surface features higher in contrast. The straints related to receiving technical grade calcium nitrate, we
marble also became discolored, most likely due to oxidation of chose to use a nearly eutectic mixture of sodium nitrate and po-
entrapped iron. This material fared poorly and is unacceptable as a tassium nitrate without any calcium nitrate. Equal numbers of
filler material. 23-kg bags of sodium nitrate and potassium nitrate were loaded
The taconite pellets held together fairly well. Because the pel- into a propane-fired salt melter, placed on the top of the tank. The
lets are porous, salt wicked into the pellets, but this did not seem salt was melted and heated to about 290°C, and then a valve was
to affect them. There were some fines 共small particles of material兲 opened to dump the salt into the tank. Once enough salt was

156 Õ Vol. 124, MAY 2002 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/19/2013 Terms of Use: http://asme.org/terms

 Fig. 7 Measured temperatures of salt pumped from the bottom
and returned to the top of the thermocline tank during a charg-
ing cycle

Fig. 6 Schematic of the 2.3 MWh thermocline flow loop off the top of the tank had a slight downward trend that became
more pronounced when the gradient reached the top of the tank.
1
Figure 8 shows the temperature gradient through the bed at 2 -hour
increments while discharging. During this discharge cycle, the
melted so the cold pump could be started, the propane heater was energy extracted was approximately 2.44 MWht , which is about
brought online and solid salt was added directly to the top of the equal to the amount of energy added to the salt during the charg-
tank to be melted by the hot salt returning from the heater. ing cycle, within experimental error. The final hot salt temperature
exiting the thermocline was 361°C.
2 Description of the Flow Loop. A schematic of the flow
Over a 41-hour period, average heat loss was measured to be
system is shown in Fig. 6. The system incorporated a propane
approximately 20 kW, which is much higher than the predicted
heater to simulate the heat input from the oil-to-salt heat ex-
value of 12 kW. It should be noted that there is a large uncertainly
changer. To charge the system, cold salt was drawn off the bottom
associated with heat loss measurement because it is based on the
of the tank through a manifold using a multistage turbine pump.
change in internal energy of the tank, which cannot be measured
The cold salt was heated with a propane heater and returned to the
very accurately with the instrumentation installed. Also, in this
top of the tank. Both the flow rate of the salt and the heat input
small tank there are heat sinks such as pump penetrations, which
from the propane heater could be controlled. In general, the flow
were not accounted for in the heat loss model. The penetrations in
rate was fixed, then the propane flow valve was adjusted to get the
a larger system will have less affect on the overall heat loss.
desired salt-outlet temperature. As the hot salt made its way
The shape of the thermocline profile is illustrated over this 41-
through the sand and rock, it heated the filler materials and estab-
hour period in Fig. 9. Since no heat was added during this test, the
lished a thermal gradient.
thermocline profile went from a fairly flat profile 共0.0 hours in
To discharge the system, hot salt was drawn off the top of the
Fig. 9兲 to a more tapered profile after 41 hours. Because this tank
tank through a manifold and pumped through a forced-air salt
had a high volume-to-surface-area ratio, edge effects are more
cooler to reject the heat from the salt. The cold salt exiting the
pronounced than would be the case in larger tanks. Despite this
cooler returned to the bottom of the tank. At the end of a test, both
fact, after nearly two days, the thermocline profile was still well
the heater and cooler were drained into a sump. The sump’s in-
pronounced and could yield useful energy extracted at a reason-
ventory could then be pumped back into the tank.
able temperature potential.
3 Test Plan. The main purposes of this test were to: 1兲
5 Economic Analysis of a Thermocline System. Since the
verify the heat capacity of a thermocline system, 2兲 evaluate the
goal of this development effort was to reduce the costs of an
size and shape of the thermal gradient, 3兲 evaluate the change in
indirect thermal storage system, we wanted to compare the costs
shape of the thermal gradient over time, and 4兲 evaluate the heat
loss of the thermocline.
The following tests were planned to gather the data necessary to
meet the objectives above: 1兲 initial charging test, 2兲 full dis-
charge test, 3兲 full charge test, 4兲 partial charging test, 5兲 partial
discharging test, and 6兲 heat loss test.
4 Test Results. Figure 7 shows the temperatures of the cold
salt pumped from the bottom of the tank and of the hot salt re-
turning to the top of the tank after being heated by the propane
heater. The propane heater slightly overshot the set point during
start up. Also, towards the end of the charging cycle, the salt
temperature pumped from the bottom of the tank started to climb
as the gradient reached the bottom of the tank. During this charg-
ing cycle, the amount of energy added to the salt was approxi-
mately 2.6 MWht . This is slightly higher than the system rating of
2.3 MWht because the salt in the space above the rocks and sand
was at a temperature of about 330°C instead of 390°C as assumed
in the model. Fig. 8 Measured temperatures of the thermocline profiles dur-
During the discharge cycle, the temperature of the salt coming ing a discharge cycle

Journal of Solar Energy Engineering MAY 2002, Vol. 124 Õ 157

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/19/2013 Terms of Use: http://asme.org/terms

 Table 2 Cost of nitrate saltes, Hitec® XL, and Hitec® , not in-
cluding transporation

Fig. 9 Degradation of the thermocline profile over a period of

41 hours without supplementary heat

of a two-tank molten-salt system with a thermocline molten-salt

system. We did a survey of actual costs of delivered crushed rock $3.49/kg. In large quantities 共e.g., Megaton兲, sodium nitrate and
and nitrate salt components in early 2000. Table 1 summarizes the potassium nitrate can be purchased at reduced rates of approxi-
costs of several candidate crushed-rock, sand, and taconite. As can mately 10%.
be seen from the table, a significant amount of the cost is for Table 3 shows the cost breakdown of a 688 MWht two-tank
transportation, when the product is not locally produced. We also molten salt system and a 688 MWht thermocline thermal storage
found that sometimes the cost of crushed rock can be reduced system. The thermocline system assumes that the filler material is
3
significantly, if purchased directly from the quarry. a mixture of 4 inch crushed quartzite and silica filter sand (8
Table 2 shows the cost of nitrate salt components and mixtures, ⫻12) and has a void fraction of 0.24. The most costly component
not including transportation. The calcium nitrate, is the tetrahy- in a two-tank molten salt system is the nitrate salt inventory. The
drate form, Ca(NO3 ) 2 •4H2 O. As it is heated, the molecule se- thermocline system is about 66% of the cost of a two-tank molten
quentially dehydrates with increasing temperature. The water mol- salt system.
ecules comprise 30.5% of the mass of the tetrahydrate salt. The
cost of the calcium nitrate without the water is $2.06/kg.
The Hitec XL® we employed was shipped as a water solution. Conclusions
The inventory we received contained 59% water by mass, which A thermocline indirect storage system has been developed that
had to be evaporated. The cost of dewatered Hitec XL® is actually is about 2/3 the cost of a two-tank molten salt system for parabolic
trough power plants. Screening studies on candidate filler materi-
als showed that silica filter sand, quartzite rock, and taconite were
Table 1 Cost of crushed rock, sand, and taconite delivered to compatible with nitrate salts, both in isothermal and cyclic condi-
Albuquerque, NM tions. We chose quartzite and silica sand as the most practical
filler materials for a small, pilot-scale test because of their avail-
ability and cost. Results from the pilot-scale test confirmed the
thermal capacity of the thermocline was near the calculated value.
The height of the thermal gradient corresponded to that predicted
by the model. Heat losses were higher than predicted primarily
due to the heat losses through pump penetrations on the top of the
tank, though in a large thermal storage system (100s MWht) that
effect will have a smaller 共near negligible兲 impact. The results of
this work show that a molten-salt thermocline system is a feasible
option for thermal storage in a parabolic trough plant.

Table 3 Cost of two-tank and thermocline molten-salt thermal

storage systems each 688 MWh. Assumptions: Tank ⌬ T
Ä84°C, Mixed filler costÄ$72Õtonne, Salt costÄ$0.45Õkg, Tank
costÄ$155Õm3 , Thermocline practical capacityÄ69%.

158 Õ Vol. 124, MAY 2002 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/19/2013 Terms of Use: http://asme.org/terms

 Nomenclature Solar Thermal Central Receiver Pilot Plant: Thermal Storage Subsystem
Evaluation–Final Report, Sandia National Laboratories Report, SAND86-
A ⫽ cross sectional area of bed, m 2 8212.
关3兴 Solutia Technical Bulletin 7239115B, 1999, Therminol® VP-1, Vapor Phase/
Cp ⫽ specific heat of fluid ( f ) or bed (b), J/kg K Liquid Phase Heat Transfer Fluid.
D ⫽ particle diameter, m 关4兴 Nexant, LLC, 2000, Task 2 Report, Thermal Storage for Rankine Cycle Power
h ⫽ heat transfer coefficient between bed and fluid, Plants, NREL Task Order Authorization Number KAF-9-29765-09.
W/m2 K 关5兴 Pilkington, Solar International, 2000, Task 5: Annual Performance Simulation
hv ⫽ 6h(1⫺␧) ␣ /D⫽volumetric heat transfer coefficient, for ISCCS with Storage, NREL subcontractor AAR-9-29442-05 report.
W/m3 K 关6兴 Pacheco, J. E., and Gilbert, R., 1999, ‘‘Overview of Recent Results of the
Solar Two Test and Evaluations Program,’’ Renewable and Advanced Energy
T ⫽ temperature of fluid ( f ) or bed (b), C Systems for the 21st Century, Proc. of the 1999 ASME Int. Solar Energy Conf.,
y ⫽ vertical dimension, m Maui, HI.
␣ ⫽ surface area shape factor 关7兴 Price, H., and Kearney, D., 1999, Parabolic-Trough Technology Roadmap: A
␧ ⫽ void fraction of bed Pathway for Sustained Commercial Development and Deployment of
␳ ⫽ density of fluid ( f ) or bed (b), kg/m3 Parabolic-Trough Technology, NREL/TP-550-24748, NREL, Golden, CO.
␶ ⫽ time, s 关8兴 Burolla, V. P., and Bartel, J. J., 1979, The High Temperature Compatibility of
Nitrate Salts, Granite Rock and Pelletized Iron Ore, Sandia National Labora-
tories Report, SAND79-8634.
References 关9兴 Bradshaw, R. W., and Tyner, C. E. 1988, Chemical and Engineering Factors
关1兴 McDonnell Douglas Astronautics Company, 1986, 10 MWe Solar Thermal Affecting Solar Central Receiver Applications of Ternary Molten Salts, Sandia
Central Receiver Pilot Plant Mode 5 共Test 1150兲 and Mode 6 共Test 1160兲 Test National Laboratories Report, SAND88-8686.
Report, Sandia National Laboratories Report, SAND86-8175. 关10兴 Martin Marietta, 1980, Molten Salt Safety Study, Sandia National Laboratories
关2兴 Faas, S. E., Thorne, L. R., Fuchs, E. A., and Gilbertsen, N. D., 1986, 10 MWe Report, SAND80-8179.

Journal of Solar Energy Engineering MAY 2002, Vol. 124 Õ 159

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/ on 04/19/2013 Terms of Use: http://asme.org/terms