Solar Energy 164 (2018) 38–46

Contents lists available at ScienceDirect

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

Molten salt power towers operating at 600–650 °C: Salt selection and cost T
benefits
⁎
Craig S. Turchia, , Judith Vidalb, Matthew Bauerc
a
Thermal Systems R&D Group, National Renewable Energy Laboratory, Golden, CO 303-384-7565, USA
b
Thermal Systems R&D Group, National Renewable Energy Laboratory, Golden, CO, USA
c
ManTech International, on contract to U.S. Department of Energy, Solar Energy Technology Office, Washington, DC, USA

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

Keywords: This analysis examines the potential benefit of adopting the supercritical carbon dioxide (sCO2) Brayton cycle at
Molten salt 600–650 °C compared to the current state-of-the-art power tower operating a steam-Rankine cycle with solar salt
Concentrating solar power at approximately 574 °C. The analysis compares a molten-salt power tower configuration using direct storage of
Supercritical CO2 Brayton cycle solar salt (60:40 wt% sodium nitrate: potassium nitrate) or single-component nitrate salts at 600 °C or alternative
carbonate- or chloride-based salts at 650 °C.
The increase in power cycle efficiency offered by the sCO2 Brayton cycle is expected to reduce the size and
cost of the solar field required for a given thermal energy input. Power cycle capital cost is expected to decrease
compared to the superheated steam-Rankine cycle, based on projections from sCO2 cycle developers.
Maximizing the ΔT of the storage system is required for viable deployment of sensible-salt TES. In this regard,
the partial-cooling sCO2 cycle is noted as a better option than the recompression sCO2 cycle. In the current
analysis it is assumed that a ΔT = 180 K can be achieved with the partial-cooling cycle. Even with ΔT = 180 K,
the potential benefits of the sCO2 Brayton cycle are partially or completely offset by increased thermal storage
cost, albeit for reasons that differ for the different salts. An approximate 5% reduction in levelized cost of energy
(LCOE) is achieved with either solar salt at 600 °C or ternary magnesium chloride salt at 650 °C.
The potential of using pure sodium nitrate or potassium nitrate is considered because the cold tank tem-
perature for the sCO2 power cycle is estimated at 420 °C, which would allow use of a salt with a higher melting
point than solar salt. Sodium nitrate is the most cost effective, resulting in an overall LCOE reduction of 8.5%;
however, sodium nitrate is known to have lower thermal stability than potassium nitrate.
The strong influence of salt cost and hot-tank cost on overall economics led to the analysis of single-tank
thermocline options. The thermocline design significantly reduces salt inventory (by 50% or more) and in many
cases also reduces the tank size versus the hot salt tank of the 2-tank system. It is speculated that integration of
encapsulated phase-change material (PCM) in the thermocline could further increase the thermal-storage energy
density and reduce storage tank volume. The thermocline cases led to three scenarios with relative LCOE re-
ductions of approximately 10%; however, this must be tempered by possible operational inefficiencies of the
thermocline temperature profile.

1. Introduction salt temperature of about 570 °C. The SunShot goal requires an addi-
tional cost reduction of at least 50% from the current cost of this
The U.S. Department of Energy launched the SunShot Initiative in technology in the U.S. market (Mehos et al., 2016; IRENA, 2016). In
2011 with the goal of making solar electricity cost-competitive with addition to solar-field cost reductions, analysis suggests that the Sun-
conventionally generated electricity by 2020. The stated metric of this Shot goal requires development of new heat-transfer fluids (HTFs) and
initiative is a levelized cost of energy (LCOE) for utility-scale solar power systems operating at a temperature where the net power system
power of 0.06 USD/kWh (see, for example, Mehos et al., 2016). thermal-to-electric conversion efficiency will reach about 50%, for ex-
The state-of-the-art concentrating solar power (CSP) system is as- ample, near 700 °C. A molten-salt power tower is not the only possible
sumed to be a molten-salt power tower employing a 60:40 wt% blend of path for next-generation CSP; however, the operating flexibility, en-
sodium and potassium nitrate commonly known as “solar salt” at a hot- ergy-storage efficiency, and industry familiarity with this design makes

⁎
Corresponding author.
E-mail address: [email protected] (C.S. Turchi).

https://doi.org/10.1016/j.solener.2018.01.063
Received 3 November 2017; Accepted 18 January 2018
0038-092X/ © 2018 Published by Elsevier Ltd.
C.S. Turchi et al. Solar Energy 164 (2018) 38–46

Nomenclature ηrec,th receiver thermal efficiency

ηrec receiver annual average efficiency
Arec receiver area ρ salt density
Cp salt heat capacity CAPEX capital equipment expense
F view factor of receiver surface to ambient CSP concentrating solar power
Tmp salt melting or liquidus point DOE United States Department of Energy
Tmax salt maximum bulk operating temperature based on HTF heat transfer fluid
thermal stability LCOE levelized cost of energy
Trec receiver surface temperature LPPA levelized power purchase agreement
̇
Qinter power from the solar field intercepted by the receiver MSPT molten salt power tower
̇
Qrec thermal power delivered by the receiver to the heat NREL National Renewable Energy Laboratory
transfer fluid PCM phase change material
α receiver surface absorptance SAM System Advisor Model
∊ receiver surface emittance sCO2 supercritical carbon dioxide
ηrec receiver overall efficiency based on intercepted power TES thermal energy storage
from the solar field and power delivered to the heat USD U.S. dollars
transfer fluid
ηrec,opt receiver optical efficiency

it a leading contender. However, evolving from 570 °C to 700 °C will efficiencies. Physical property data for the different salts were added to
necessitate a new HTF to be developed, owing to solar salt’s decom- the SAM model via the user-defined HTF feature.
position around 600 °C. Furthermore, an advanced power cycle more
amenable to CSP requirements than steam-based turbines must be
2.1. Salt selection and properties
employed to achieve the LCOE objective. Each technology shift will
have several consequences on the CSP system.
Current parabolic trough and power tower systems use solar salt to
Deploying a new CSP technology operating at approximately 700 °C
provide thermal energy storage. Physical properties of this salt are well
entails a level of risk that makes financing such a technology difficult.
documented (SQM, 2016). One limitation of solar salt is a thermal
Developing the necessary technologies in a step-wise approach—first
decomposition temperature in the range of 600 °C, which limits the
demonstrating system concepts and power technologies at 600 °C to
upper temperature of power tower systems employing solar salt as the
650 °C and later evolving to higher-efficiency systems at 700 °C—offers
HTF and thermal storage media. A number of alternative salts have
a lower-risk path. Furthermore, financing high-risk technologies that
been proposed and explored; for this analysis we focus on the salts
can approach one billion dollars is highly challenging, and progressing
listed in Tables 1 and 2.
toward SunShot in steps that CSP industry members can support and
In addition to showing specific heat capacity, density and viscosity,
implement is essential for the health of the industry and the commercial
Table 1 highlights the volumetric heat capacity, ρCp, relative to the
viability of newly developed technologies.
value for solar salt. Volumetric heat capacity is an important factor in
determining the volume of the storage tanks, given that tank size is
2. Approach inversely proportional to ρCpΔT. The only salts with a larger volumetric
heat capacity than solar salt are sodium nitrate and the ternary car-
This study examines the benefits of operating a molten-salt power bonate.
tower with an advanced power cycle at 600–650 °C—temperatures that In addition to solar salt, this analysis considers pure sodium nitrate
are low enough to use the same or similar alloys to that in current CSP and potassium nitrate. Despite its higher Tmp, pure sodium nitrate is
plants while allowing for increases in power-cycle efficiency. The pro- considered a possible salt for use with the sCO2 Brayton cycle because
posed power cycle is the supercritical carbon dioxide (sCO2) re- this power cycle optimizes to a higher cold-salt temperature (∼400 °C)
compression Brayton cycle that is the subject of international devel- than the steam-Rankine cycle (∼300 °C). While exhibiting cost and
opment activity. heat capacity benefits versus solar salt, sodium nitrate is predicted to
NREL’s System Advisor Model (SAM) is a simulation tool with have lower thermal stability versus potassium nitrate (Bauer et al.,
technology models for various solar and other renewable energy sys- 2013a, 2013b). Thus, the study also considers pure potassium nitrate as
tems. In this analysis, SAM version 2017.01.17 was used to model a an alternative with similar physical properties but greater thermal
molten-salt power tower (MSPT). SAM’s default MSPT model was stability than solar salt.
modified to simulate higher salt temperatures and power-cycle Beyond the nitrate salts we consider chloride and carbonate salts as

Table 1
Properties of solar salt and alternative salts. Tmp represents the melting point or approximate liquidus point for non-eutectic salts. Physical properties shown at approximately 600 °C
unless noted.

Salt Tmp (°C) Tmax (°C) Heat Cap. Cp (kJ/kg K) Density ρ (kg/L) Relative ρCp Visc. (cP) Refs.

Solar salt (baseline) 238 585 1.55 1.71 1.00 1.03 SQM (2016)
NaNO3 306 520 1.62a 1.82a 1.11 – Bauer et al. (2013a)
KNO3 334 600 1.40a 1.78a 0.94 – Cordaro et al. (2011)
KCl/MgCl2 426 > 800 1.03 1.94 0.75 1.88 Mohan et al. (2018), and Williams (2006)
MgCl2/NaCl/KCl 385 > 800 1.14 1.93b 0.83 – Mohan et al. (2018)
ZnCl2/NaCl/KCl 200 > 800 0.92 2.08 0.72 4.5 Li et al. (2016)
K2CO3/Na2CO3/Li2CO3 398 800 1.79 2.01 1.36 10.7 An (2016)

a
Approximately 450 °C.
b
Value taken from binary NaCl/MgCl2 salt in Mohan et al. (2018).

39
C.S. Turchi et al. Solar Energy 164 (2018) 38–46

Table 2 2.3. Supercritical CO2 power cycle

Composition and estimated market price of salts. Estimated costs taken from (Mehos
et al., 2017).
The sCO2 Brayton cycle has been identified as a potential replace-
Salt Composition by Estimated Estimated salt ment to the venerable superheated steam-Rankine cycle that is used in
weight fraction cost ($/kg) media cost with almost all CSP plants (Mehos et al., 2017). sCO2 Brayton cycles can
ΔT = 200 K have many variations, but attention has focused on the recompression
($/kWhth) cycle as the best combination of thermal efficiency and simplicity, and
Solar salt (NaNO3:KNO3) 0.60:0.40 800 10 this configuration has been selected for large-scale demonstration under
NaNO3 1 680 8 the U.S. DOE’s Supercritical Transformational Electric Power (STEP)
KNO3 1 1000 13 program. A slight variation referred to as the partial-cooling re-
KCl:MgCl2 0.625:0.375 350 6 compression cycle has been proposed as well-suited to CSP plants be-
MgCl2:NaCl:KCl 0.550:0.245:0.205 220 3
cause of the ability to increase the temperature differential across sto-
ZnCl2:KCl:NaCl 0.606:0.313:0.081 800 18
K2CO3:Na2CO3:Li2CO3 0.345:0.334:0.321 2500 28 rage with relatively little impact to thermal efficiency (Neises and
Turchi, 2014). sCO2 Brayton cycles have been projected to have higher
efficiency and lower capital cost than comparable superheated steam
options that exhibit greater thermal stability and would allow for op- cycles at turbine inlet temperatures greater than about 550 °C (Padilla
eration at temperatures in the range of 700 °C (Mehos et al., 2017). et al., 2015).
Three different chloride salt combinations are considered: a binary Thermal efficiency of the sCO2 recompression cycle is based on
potassium/magnesium chloride blend, a ternary magnesium/sodium/ input from power system developer Echogen Power Systems (Akron,
potassium blend, and a ternary zinc/potassium/sodium blend. The OH). Echogen is developing sCO2 turbomachinery and system designs
binary MgCl2 salt has been studied fairly extensively for application as a for multiple applications. In this study Echogen provided simulation
high-temperature HTF (Williams, 2006), although data with respect to results for configurations using a recompression cycle operating with
its heat capacity as a function of temperature are lacking. The ternary 590 °C or 640 °C turbine inlet temperature and design-point ambient
MgCl2 salt promises a lower melting point, higher heat capacity, and temperature of 35 °C (compressor inlet temperature of 40 °C). The
lower cost that make it an attractive alternative to the binary blend corresponding gross efficiency values were 0.437 and 0.468 and the
(Mohan et al., 2018; Williams, 2006; Mehos et al., 2017). Lastly, the design-point cooling fan parasitic load was estimated at 2.1% and 2.0%
listed ZnCl2 blend is one of several ZnCl2-based salts that have been of gross power, respectively. Pressure losses across each heat exchanger
examined for CSP application (Li et al., 2016; Jonemann, 2013). ZnCl2 were estimated at approximately 1% per unit. The sCO2 power cycle
salts emphasize low melting point over cost and heat capacity con- was simulated with off-design tables developed by NREL for the re-
siderations compared to the MgCl2-bearing salts. compression cycle and imported into SAM via the user-defined power
The ternary carbonate is a well-known eutectic salt blend (An, cycle option. While not part of the 2017.01.17 release, this code will be
2016), as alkali carbonate salts are used in molten carbonate fuel cells included in future releases of SAM.
at temperatures around 650 °C. In this application they are exposed to The capital cost of the power cycle is based on estimates made by
oxygen, H2O, and CO2 as part of operation, and they are inherently less Echogen Power Systems and Abengoa Solar as part of a prior study
corrosive than chloride salts under such conditions. Although corrosion exploring the design of a 10-MWe sCO2 turbine (Turchi, 2014). Under
is still an issue of concern, the greatest issue with the eutectic carbonate that project the team developed cost and performance estimates for a
salt blend listed in Table 2 is the cost of lithium carbonate. If Li2CO3 nominal 100 MWe partial-cooling cycle operating at 600 °C and 700 °C
remains at or near its current market price, it is unlikely that the ternary with a hypothetical molten salt that had properties comparable to solar
carbonate will be an economic option despite its attractive ρCp value. salt. These estimates are in overall agreement with more recent esti-
mates made by Carlson et al. (2017), although the distribution of cost
2.2. Receiver efficiency between components in the power block differ.
In the present analysis thermal efficiency estimates are based on a
One of the key metrics for the analysis is the calculated or assumed recompression cycle simulation; whereas, the turbine temperature dif-
annual average receiver efficiency, ηrec . Thermal power delivered by the ferential (180 K) and power cycle capital cost are based on the partial
receiver to the HTF is given by the product of the overall receiver ef- cooling cycle. Thus, we assume that the two cycle designs have similar
ficiency and the incoming energy intercepted by the receiver. The gross efficiency, which is supported by prior analysis (Neises and
overall receiver efficiency is often broken into the product of its optical Turchi, 2014).
and thermal efficiencies (Hirsch, 2017), thus:
̇ = ηrec Qinter
Qrec ̇
2.4. Relative LCOE comparison
ηrec = ηrec,opt ηrec,th
The primary metric of this analysis is a relative comparison of
Within SAM, ηrec,opt is given by the coating absorptance, α, on the system LCOE versus that for the baseline MSPT technology. The base-
Tower and Receiver inputs page, which is assumed to be a constant value line is assumed to be represented by the default MSPT case in SAM
during the simulation. The receiver thermal efficiency, ηrec,th , varies 2017.01.17. Input variables for that case can be found by downloading
with each timestep in the SAM simulation depending on the receiver
SAM and viewing the default MSPT case. Several of the key design and
conditions; however, the annual average receiver efficiency can be
performance parameters are shown in Table 3. The financial assump-
calculated as:
tions that exist in the SAM default case are retained, with the exception
ηrec = αηrec,th that the federal investment tax credit is set to zero for all cases. Relative
LCOE is selected to normalize the influence of financial assumptions on
Σ[Rec. incident thermal power ]−Σ[Rec . conv. + emissive losses]
=α the analysis. The nomenclature within SAM uses levelized power pur-
Σ[Rec. incident thermal power ]
chase agreement (LPPA) price as the metric that accounts for market
where the terms in brackets are SAM output variables, and the Σ in- factors such as time of delivery pricing. However, in the current ana-
dicates use of the annual sum of this variable from SAM’s Statistics lysis a flat price structure is used, and relative LCOE and relative LPPA
page. (Note: SAM neglects heat loss due to conduction through the provide the same results.
tower components.) In this analysis we report ηrec for the various cases.

40
C.S. Turchi et al. Solar Energy 164 (2018) 38–46

Table 3 overall temperature experience annual average efficiencies greater than

Primary parameters in the default MSPT model in SAM 2017.01.10 that 0.90. The results of this second analysis are compared to the initial
serves as the baseline case for this study.
results in Table 5.
Turbine gross output (MWe) 115
Hot-tank/cold-tank temp. (°C) 574/290
3.2. Pure nitrate salt as HTF
Thermal storage (h) 10
Solar multiple 2.4
Power cycle gross efficiency 0.412 The cold tank temperature in the sCO2 cycle cases is approximately
Cooling method Air cooled 420 °C, which implies that the melting point temperatures of NaNO3
Location (USA) Daggett, CA (308 °C) or KNO3 (334 °C) would not prohibit their use as the HTF/TES
salt. In terms of cost and physical properties, NaNO3 would be favored;
however, KNO3 is known to have greater thermal stability (Bauer et al.,
3. Results
2013a, 2013b; Cordaro et al., 2011). Each nitrate salt was examined to
determine the impact to relative LCOE. The performance simulation
The initial runs were performed starting with the default MSPT
uses values for KNO3 as the user-defined HTF for KNO3-sCO2-600 and
model in SAM 2017.01.17. This created the baseline for all subsequent
for solar salt in the NaNO3-sCO2-600 case, due to lack of available pure-
cases. Initially, four alternative cases were created by importing the
salt data for all properties. Solar salt properties should be a conservative
alternative salt physical properties, adjusting the hot and cold salt
surrogate for NaNO3. The primary difference in the case inputs is TES
temperatures, and applying the sCO2 Brayton cycle via a user-defined
cost, which ranges from 27 to 35 to 41 $/kWh for NaNO3, solar salt, and
power cycle. The results of these initial cases are summarized in
KNO3 respectively, due to the relative cost and Cp of the different salts.
Table 4.
The use of pure NaNO3 has a cost benefit, but it is unclear if this salt
could be used at the proposed temperature without significant de-
3.1. Influence of receiver efficiency gradation. While potentially more stable, the cost benefit of using KNO3
is rather small at the assumed cost of 1000 USD/tonne. Further study of
The data in Table 4 show that receiver efficiency decreases with the ability to stabilize NaNO3 (or solar salt) for operation at 600 °C
receiver temperature as one would expect due to the dependence of could enable this relatively low-risk alternative.
emissive and convective losses on receiver temperature. In terms of
factors that can be affected by the receiver design, optical and thermal
3.3. Interpretation of initial results
losses from the receiver are described by:
4
f (∊ ,F ,Trec ) + f (Arec ,Trec ) Fig. 1 highlights how assumptions within the alternative cases in-
ηrec = α− fluence certain key metrics. Solar Field CAPEX reflects the impact of
̇
Qinter
power cycle efficiency, because a more efficient power cycle will re-
This falloff in receiver efficiency with temperature offsets the ben- quire a smaller solar field. Thermal storage CAPEX is influenced by
efits of power cycle efficiency increase with temperature. The SunShot specific salt cost, salt volumetric heat capacity, hot-tank/cold-tank
Initiative established a target for annual average receiver efficiency of temperature differential (a function of power cycle), and tank material-
90%, acknowledging that operation at higher temperature must include of-construction. The high cost of the carbonate salt is apparent in the
advances in receiver design to mitigate the potential for greater optical 116% increase in thermal storage CAPEX for that case. On the other
and thermal losses. With this in mind, the analysis was repeated as- hand, the low volumetric heat capacity of the chloride salt is re-
suming an annual average receiver efficiency of 90% is achieved for sponsible for its 43% increase in thermal storage cost due to the in-
molten-salt receivers operating at an HTF outlet temperature of 650 °C. crease in tank volume and cost. Power cycle CAPEX changes with cycle
This value is obtained within SAM by setting the absorptance, α = 0.97, design and temperature. Total Installed Cost and LCOE show the overall
and the emittance, ∊ = 0.40. The absorptance value was selected based system impact. In these analyses, the thermal storage CAPEX cost re-
on the highest values reported in some selective surface research (Kim duces or overwhelms the advantages brought by the sCO2 power cycle.
et al., 2016). The emittance value was selected to force the lowest ef- The annual average receiver efficiency was raised for all cases, in-
ficiency receiver from the SAM-default simulations (Carb-sCO2-650) cluding the Baseline MSPT, thus the LCOE benefit is a moving target
equal to 0.90. While this emittance may not be realistic for the current- because the Baseline MSPT also experiences a significant boost in re-
design molten salt receivers, we assume that other design changes, such ceiver efficiency due this assumed technology advance.
as higher flux limits, smaller receivers, and/or cavity-type designs are Despite operation at the same nominal temperature, the influence of
implemented to achieve this “effective emittance” value. These emit- receiver efficiency on the Carb-sCO2-650 case is more pronounced than
tance and absorptance values were then applied to all the SAM cases in the MgCl2-sCO2-650 case. This results from the difference in heat ca-
the “ηrec = 0.90 ” scenarios. The functional dependence of thermal loss pacity of the two salts. The greater ρCp of the carbonate salt leads to a
with temperature is retained such that receivers running at lower slower modeled velocity within the receiver. The slower velocity

Table 4
Evaluation of alternative molten-salt/power-cycle systems that offer the potential for higher operating temperature.

Baseline MSPT SS-Steam-600 SS-sCO2-600 Carb-sCO2-650 MgCl2-sCO2-650

Salt Solar salt Solar salt Solar salt Na2CO3/K2CO3/Li2CO3 MgCl2/NaCl/KCl

Hot-tank temp. 574 °C 600 °C 600 °C 650 °C 650 °C
Cold-tank temp. 290 °C 316 °C 420 °C 470 °C 470 °C
Power cycle Steam Steam sCO2 Brayton sCO2 Brayton sCO2 Brayton
Power cycle gross efficiency 0.412 0.422 0.437 0.468 0.468
Annual HTF pumping parasitic load (% of gross) 2.5% 2.4% 3.1% 2.2% 3.9%
Annual avg. receiver efficiency 0.882 0.874 0.863 0.828 0.853
Annual net generation (MWh) 553,000 552,000 530,000 498,000 501,000
Overnight installed cost ($/kW) 6760 6720 6130 6490 5900
Relative LCOE – −0.4% −4.2% +6.8% −2.1%

41
C.S. Turchi et al. Solar Energy 164 (2018) 38–46

Table 5
Evaluation of alternative MSPT systems that offer the potential for higher operating temperature with a hypothetical coating that sets receiver efficiency to the SunShot target of 90% for
the Carb-sCO2-650 case.

Baseline MSPT SS-Steam 600 SS-sCO2 600 Carb-sCO2 650 MgCl2-sCO2 650 KNO3-sCO2 600 NaNO3-sCO2 600

Pyromark coating (SAM default)

Rec. efficiency 0.882 0.874 0.863 0.828 0.853 – –
Relative LCOE from Table 4 – −0.4% −4.2% +6.8% −2.1% – –

SunShot coating, ηrec = 90% at 650 °C for carbonate salt HTF

Rec. efficiency 0.929 0.924 0.919 0.900 0.913 0.919 0.919
Relative LCOEa – −0.8% −5.1% +1.9% −4.5% −1.9% −8.5%

a
Assumes Baseline MSPT also benefits from the SunShot coating.

reduces the heat transfer coefficient, which leads to a higher external strength than stainless steel. The cost ratio for Inconel 625 (IN625) to
surface temperature on the receiver. Hence, although the nominal HTF 347 stainless steel (347SS) is taken as 3.25:1. The greater tensile
temperatures are equal, the model predicts that the carbonate receiver strength of IN625 reduces the multiplier from this value to 2.44. The
loses more thermal energy due to its higher skin temperature. In real general results are summarized in Table 6 and Fig. 2.
terms, salt properties will affect receiver efficiency even at the same The combination of lower storage ΔT and higher specific cost (for
bulk salt temperature and it may be more difficult to realize the higher carbonates) or lower ρCp (for chlorides) leads to increased TES cost
receiver efficiency with the carbonate salt. versus the baseline MSPT and offsets the potential advantages of
Clearly the benefit of raising the power cycle temperature is con- switching to the sCO2 Brayton cycle. Given the effort that has been
tingent on other changes within the system. With respect to receiver dedicated to salt exploration, it is unlikely that new salt formulations
efficiency, one may need to redesign the receiver to reduce exposed will be forthcoming, so this remains an impediment to adoption of the
surface area via use of cavity designs or attainment of higher flux limits. new power cycle or advancement to higher temperature. An alternative
Improving the selective surface coating is another path to greater re- path for integration of the sCO2 Brayton cycle would be use of phase-
ceiver efficiency, albeit one that becomes more challenging as tem- change material (PCM) for thermal storage. Use of PCM storage relaxes
perature increases. The effect of internal heat transfer coefficient is also the incentive for a large ΔT across storage, which may allow for greater
noted as influencing the receiver external skin temperature for a given flexibility in power cycle optimization. Alternatives to conventional 2-
nominal HTF temperature. HTFs with greater thermal conductivity and tank TES are discussed next.
receiver designs with internal fins or other heat transfer enhancements
are beneficial in this regard.
3.5. Thermoclines to reduce salt inventory and tank costs

3.4. Influence of thermal storage cost Single-tank thermocline systems have long been considered as a
pathway to lower-cost thermal storage for CSP plants, and the concept
Fig. 1 illustrates that an increasing thermal storage cost can offset was tested at pilot-scale by Sandia National Laboratories in the 1990s
much or all of the potential benefit of moving to higher temperature. (Pacheco et al., 2002). Thermocline systems use a single tank to hold a
The capital costs of the different thermal storage systems are calculated thermally stratified fluid, rather than having separate hot and cold
using the methodology outlined in Mehos et al. (2017). Starting from a tanks. The primary benefit comes from having only a single tank. Ad-
detailed cost analysis of the baseline MSPT (Abengoa Solar, 2010), the dition of a solid fill material such as loose rock or structured packing
amount of storage salt and the size of the salt tanks for the alternative brings the added ability to reduce the required inventory of liquid HTF.
cases are scaled based on thermal storage ΔT and the specific and vo- The ability to reduce salt inventory could bring significant cost benefits
lumetric heat capacities of the alternative salts. This scaling factor is to TES systems that rely on relatively expensive salts. Examining Fig. 2,
multiplied by an alloy multiplier that accounts for differences in the we postulate that the systems that would benefit most from a thermo-
alloy cost and alloy strength as a function of temperature. Alloy cost cline are the carbonate-salt, ZnCl2, and KNO3 cases. Elimination of the
estimates were obtained from discussion with Haynes International, a cold tank is a benefit in all cases, although the cold tank cost is small
provider of stainless steel and high-nickel alloys. Inconel 617 or 625 is fraction of the TES system cost. These benefits must be weighed against
selected for temperatures between 600 °C and 650 °C due to greater operational issues associated with non-constant temperature in the

Fig. 1. Relative change (versus baseline MSPT) of key metrics for the alternative cases.

42
C.S. Turchi et al. Solar Energy 164 (2018) 38–46

Table 6
TES assumptions and estimated cost inputs for SAM. The alloy multipliers account for differences in alloy cost and alloy mass (based on tensile strength) as material and temperature
change from the baseline conditions. Storage capacity is 2700 MWhth in all cases. CS = carbon steel, SS = stainless steel, IN = inconel.

Case Salt inventory (kg) Tank volume multiplier Hot tank alloy multiplier Cold tank alloy multiplier Hot/cold tank alloy TES system cost ($/kWhth)

Baseline 2.51 × 107 1 1 1 347SS/CS 24

SS-Steam-600 2.51 × 107 1 1.23 1 347SS/CS 25
SS-sCO2-600 3.86 × 107 1.539 1.23 1.11 347SS/CS 35
Carb-sCO2-650 3.30 × 107 1.144 2.44 2.30 IN625/347SS 59
MgCl2-sCO2-650 5.15 × 107 1.804 2.44 2.30 IN625/347SS 39
ZnCl2-sCO2-650 6.41 × 107 2.151 2.44 2.30 IN625/347SS 59
KNO3-sCO2-600 4.20 × 107 1.690 1.23 1.11 347SS/CS 41
NaNO3-sCO2-600 3.56 × 107 1.401 1.23 1.11 347SS/CS 27

Fig. 2. Breakdown of TES system cost for the baseline and alternative cases. Other category includes site work, electrical, and misc. structural steelwork.

exiting salt when the thermocline zone is discharged. beneficial reduction in cost due to elimination of the cold tank, reduced
In addition to replacing a large fraction of the liquid inventory, a salt inventory and, in most cases, a reduction in tank volume. Rock fill
rock fill will substitute the heat capacity of the rock for that of the salt. cost is taken from (Libby, 2010).
In the case of nitrate salts, the rock/mineral heat capacity is generally Lastly, the reduced TES system costs were introduced in the SAM
lower than the salt; however, for chloride-based salts, potential rock cases to determine the relative LCOE, see Fig. 5. Comparing Fig. 5 with
materials have greater heat capacity than the salt, which results in the Fig. 1 highlights the potential benefits of using a thermocline for re-
added benefit of a smaller tank, see Table 7. Unlike liquid salts, the heat ducing the TES system cost. This analysis oversized the thermocline by
capacity of rocks and minerals generally increase with temperature 20% to account for operational inefficiencies (Yang and Garimella,
(Waples and Waples, 2004), which is an important consideration in 2010). The greatest benefit versus the baseline MSPT is about a 10%
these systems. decrease in LCOE for each nitrate-salt case at 600 °C and the MgCl2-
As an initial estimate we assume that a thermocline will have an sCO2-650 case.
overall salt volume fraction (including solids-free space at the top and
bottom of the tank) of 35%. This is consistent with a packed-bed void
fraction of 25% (Libby, 2010). The cold tank is eliminated, and the hot
tank volume is adjusted by the difference in volumetric heat capacity of
the salt and the fill material, as well as a utilization factor of 80% to
account for unusable fraction of the thermocline zone. The volumetric Table 7
heat capacity of the rock will impact whether the tank size increases or Comparison of volumetric heat capacity for molten salts, three earth-abundant rocks
(Waples and Waples, 2004), and an encapsulated PCM.
decreases with the switch to a thermocline design. Fig. 3 shows how the
thermocline tank volume compares to the hot tank volume of a 2-tank Material Heat capacity Density Volumetric heat
system based on the ratio of volumetric heat capacities of the rock and (kJ/kg K)* (kg/m3)* capacity (kJ/m3 K)
the salt. A tank volume ratio of less than 1.0 indicates the thermocline
KNO3 (properties at 1.40 1780 2492
tank is smaller than the corresponding 2-tank hot tank, which is a de-
450 °C)
sired result. Most of the combinations show some benefit, with the ZnCl2/NaCl/KCl 0.92 2080 1907
notable exception of the ternary carbonate salt (Carb-Quartzite), K2CO3/Na2CO3/Li2CO3 1.79 2010 3597
which has a large volumetric heat capacity. (Carb)
EPRI’s 2010 thermocline study indicated a 20% increase in ther- Quartzite 1.15 2640a 3040
Basalt 1.35 2870a 3870
mocline volume, which is generated in Fig. 3 (EPRI) by assuming the Amphibolite 1.75 3010a 5250
heat capacity of quartzite is equal to its value at room temperature. The Encapsulated salt (NaCl/ 3.39b 1590 5390
other data here uses rock heat capacity at 600 °C as estimated by KCl, Tmp = 659 °C)
Waples and Waples, 2004.
* Measured or estimated at 600 °C unless noted.
The overall impact to TES system cost of the switch to a single-tank a
20°C.
thermocline is shown in Fig. 4. Each of the salt systems experiences a b
Effective Cp across 180 K.

43
C.S. Turchi et al. Solar Energy 164 (2018) 38–46

Fig. 3. Relative change in the tank volume and salt mass for a thermocline and 2-tank-system hot tank as a function of the rock and salt heat capacity ratio. SS = solar salt, others from
Table 7.

4. Conclusions are partially or completely offset by increased thermal storage cost,

albeit for reasons that differ for the different salts. The specific cost of
The current analysis examines the potential benefit of adopting the the carbonate salt, in particular, the lithium carbonate component,
sCO2 Brayton at 600–650 °C as an interim step towards the U.S. DOE makes this salt cost prohibitive. The ternary magnesium chloride salt
SunShot goals. The analysis assumes a molten-salt power tower con- has an attractive specific cost, but its low volumetric heat capacity,
figuration using direct storage of solar salt or single-component nitrate necessitates a large and costly increase in tank size.
salts at 600 °C or alternative carbonate- or chloride-based salts at Receiver efficiency is a key metric, and in the absence of design
650 °C. changes, the receiver efficiency decreases from the default case value as
The increase in power cycle efficiency offered by the sCO2 Brayton temperature increases in the alternative cases. Subsequently, the study
cycle is expected to reduce the size and cost of the solar field required assumes that receiver efficiency achieves the SunShot target of an
for a given thermal energy input. Power cycle CAPEX is expected to overall annual average efficiency of 0.90 for a 650 °C receiver. Under
decrease compared to the superheated steam-Rankine cycle, based on this condition, an approximate 5% reduction in LCOE is achieved with
projections from sCO2 cycle developers. Maximizing the ΔT of the either solar salt at 600 °C or MgCl2/NaCl/KCl at 650 °C. This relative
storage system is required for viable deployment of sensible-salt TES. In comparison assumes the baseline receiver efficiency is also increased by
this regard, the partial-cooling sCO2 cycle is noted as a better option the hypothetical new coating or receiver design from its current SAM
than the recompression Brayton cycle. In the current analysis it is as- value of 0.882–0.929 (at 574 °C). This study does not stipulate how to
sumed that a ΔT = 180 K can be achieved with the partial-cooling cycle increase receiver efficiency, but a combination of improved selective
with a negligible change in efficiency versus the recompression cycle. surface coating, higher flux limits, reduced receiver aperture area or
Even with ΔT = 180 K, the potential benefits of the sCO2 Brayton cycle reduced view factor to ambient are suggested.

Fig. 4. Breakdown of TES system cost for the baseline and alternative cases assuming the use of a single-tank thermocline with quartzite rock fill. Compare to 2-tank designs in Fig. 2.

44
C.S. Turchi et al. Solar Energy 164 (2018) 38–46

Fig. 5. Relative change (versus baseline MSPT) of key metrics for the alternative cases using a new 90% efficiency receiver and single-tank thermocline with quartzite fill.

The potential of using pure sodium nitrate or potassium nitrate is avoid thermal ratcheting while providing PCM containment.
considered because the cold tank temperature for the sCO2 power cycle
is estimated at 420 °C, which would allow use of a salt with a higher Acknowledgements
melting point. Sodium nitrate is the most cost effective, resulting in an
overall LCOE reduction of 8.5%; however, sodium nitrate is the less Work performed under U.S. Department of Energy Contract No. DE-
stable of the two nitrate salts. Estimates with pure potassium nitrate AC36-08GO28308.
result in an LCOE reduction of only 1.9% due to its higher cost and
lower heat capacity versus solar salt. References
The strong influence of salt cost and hot tank cost on overall eco-
nomics led to the analysis of single-tank thermocline options. The Abengoa Solar, Inc., 2010. Advanced Thermal Storage for Central Receivers with
thermocline design significantly reduces salt inventory (by 50% or Supercritical Coolants, DOE Contract Report under Grant DE-FG36-08GO18149.
An, X., 2016. Determination and evaluation of the thermophysical properties of an alkali
more) and in many cases also reduces the tank size versus the hot salt carbonate eutectic molten salt. Faraday Discuss. 190, 327.
tank of the 2-tank system, because in most cases the volumetric heat Bauer, T., Pfleger, N., Laing, D., Steinmann, W-D., Eck, M., Kaesche, S., 2013a. High-
capacity of the rock fill exceeds that of the salt, leading to smaller tank temperature molten salts for solar power application. In: Molten Salts Chemistry,
Elsevier.
volume. It is speculated that integration of encapsulated PCM in the Bauer, T., Pfleger, N., Breidenbach, N., Eck, M., Laing, D., Kaesche, S., 2013b. Material
thermocline could further increase the TES energy density and reduce aspects of Solar Salt for sensible heat storage. Appl. Energy 111, 1114–1119.
storage tank volume. The thermocline cases led to three scenarios with Carlson, M.D., Middleton, B.M., Ho, C.K., 2017. Techno-economic comparison of solar-
driven SCO2 Brayton cycles using component cost models baselined with vendor data
relative LCOE reductions of approximately 10%; however, this must be and estimates, Paper ES2017-3590. In: Proceedings of the ASME 2017 11th
tempered by possible operational inefficiencies of the thermocline International Conference on Energy Sustainability, ES2017, June 26–30, 2017,
temperature profile. Charlotte, North Carolina, USA.
Cordaro, J.G., Kruizenga, A.M., Altmaier, R., Sampson, M., Nissen, A., 2011.
Thermodynamic properties of molten nitrate salts. SolarPACES.
4.1. Recommendations Hirsch, T., 2017. Editor, “SolarPACES Guideline for Bankable STE Yield Assessment”,
SolarPACES.

• Develop and test selective coatings and receiver designs that mini- IRENA, 2016. The Power to Change: Solar and Wind Cost Reduction Potential to 2025.
Available at. <www.irena.org/publications>.
mize thermal losses for higher operating temperatures. Jonemann, M., 2013. Advanced Thermal Storage System with Novel Molten Salt, National
• Explore means to increase the thermal stability of solar salt or so- Renewable Energy Laboratory, Golden, CO, NREL/SR-5200-58595.
Kim, T.K., et al., 2016. Copper-alloyed spinel black oxides and tandem-structured solar
dium nitrate to allow operation at 600 °C. The greatest cost benefit
absorbing layers for high-temperature concentrating solar power systems. Sol. Energy
was found with use of pure sodium nitrate, assuming thermal de- 132, 257–266.
gradation can be controlled. Li, P., Molina, E., Wang, K., Xu, X., Dehghani, G., Kohli, A., Hao, Q., Kassaee, M.H., Jeter,
• Thermoclines remain a potential path to reduce TES costs while S.M., Teja, A.S., 2016. Thermal and transport properties of NaCl–KCl–ZnCl2 eutectic
salts for new generation high-temperature heat-transfer fluids. J. Sol. Energy Eng.
retaining the benefits of a direct-storage system. In addition to re- 138.
duced salt inventory, thermoclines with high volumetric-heat-ca- Libby, C., 2010. Solar Thermocline Storage Systems, Preliminary Design Study. Electric
pacity fill media will decrease required tank size. Power Research Institute, Palo Alto, California, USA Final Report 1019581.
Mehos, M. Turchi, C., Jorgenson, J., Denholm, P., Ho, C., Armijo, K., 2016. On the Path to
o Test rock fills with potentially higher volumetric heat capacity for
SunShot: Advancing Concentrating Solar Power Technology, Performance, and
compatibility with the candidate salts. Quartzite has been shown Dispatchability, National Renewable Energy Laboratory, NREL/TP-5500-65688.
to be compatible with nitrate salts, but other rocks have higher Mehos, M., Turchi, C., Vidal, J., Wagner, M., Ma, Z., Ho, C., Kolb, W., Andraka, C.,
Kruizenga, A., 2017. Concentrating Solar Power Gen3 Demonstration Roadmap.
volumetric heat capacity and data are needed with chloride salts.
National Renewable Energy Laboratory, Golden, Colorado, USA NREL/TP-5500-
o Simulate system operations with an sCO2 power cycle and ther- 67464.
mocline TES to determine the annual performance impacts of this Mohan, G., Venkataraman, M., Gomez-Vidal, J., Coventry, J., 2018. Assessment of a novel
system design. ternary eutectic chloride salt for next generation high-temperature sensible heat
storage, Energy Convers. Manage., submitted.
o Thermal ratcheting of thermocline tanks remains a concern for Neises, T., Turchi, C., 2014. A comparison of supercritical carbon dioxide power cycle
loose-fill designs. This phenomenon would have to be explored configurations with an emphasis on CSP applications. Energy Procedia 49,
before large-scale, loose-fill thermocline tanks could be im- 1187–1196.
Pacheco, J., Showalter, S.K., Kolb, W.J., 2002. Development of a molten-salt thermocline
plemented. thermal storage system for parabolic trough plants. J. Sol. Energy Eng. 124.
o Model the integration of encapsulated PCMs into a thermocline Padilla, R.V., Too, Y.C.S., Benito, R., Stein, W., 2015. Exergetic analysis of supercritical
design to provide higher energy-storage density and more stable CO2 Brayton cycles integrated with solar central receivers. Appl. Energy 148,
348–365.
temperature outlet profiles. Structured packing could be used to SQM, 2016. SQM’s Thermo-Solar Salts (Salt Factsheet), Antwerpen, Belgium: SQM

45
C.S. Turchi et al. Solar Energy 164 (2018) 38–46

Europe, N.V. < www.sqm.com > . Natural Resources Research, vol. 13, no. 2.
Turchi, C., 2014. 10 MW Supercritical CO2 Turbine Test, Final Report Under DE- Williams, 2006, Assessment of Candidate Molten Salt Coolants for the NGNP/NHI Heat-
EE0001589. National Renewable Energy Laboratory, Golden CO, USA. Transfer Loop, ORNL/TM-2006/69, Oak Ridge, TN.
Waples, D.W., Waples, J.S., 2004. A Review and Evaluation of Specific Heat Capacities of Yang, Z., Garimella, S.V., 2010. Thermal analysis of solar thermal energy storage in a
Rocks, Minerals, and Subsurface Fluids, Part 1: Minerals and Nonporous Rocks, molten-salt thermocline. Sol. Energy 84.

46