Journal of Energy Storage 56 (2022) 105875

Contents lists available at ScienceDirect

Journal of Energy Storage

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

Research Papers

Optimizing phase change composite thermal energy storage using the

thermal Ragone framework
Nelson James *, Allison Mahvi , Jason Woods
National Renewable Energy Laboratory, 15013 Denver West Pkwy, Golden, CO 80401, USA

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

Keywords: Optimal design of thermal energy storage devices can allow systems to approach specific design objectives while
Thermal energy storage maintaining desired levels of performance. The thermal Ragone framework has been applied to the design of
Phase change composites thermal energy storage heat exchangers, specifically identifying relationships between their power and energy
Levelized cost of storage
capabilities. In this paper, a finite-difference model is used to optimize thermal storage heat exchanger designs
Design optimization
for three objectives given a discharge power constraint. The three objectives are maximizing energy density,
minimizing energy-specific capital costs, and minimizing the levelized cost of storage. This study focuses on the
design of planar thermal energy storage heat exchangers with phase change materials and thermal conductivity
additives. Key design parameters identified included the conductivity additive volume fraction, spacing between
heat transfer fluid tubes, and the phase transition temperature. The optimal design is found to depend strongly on
the required thermal power, with higher powers requiring transition temperatures further from the use tem­
perature, more conductivity additives, and closer tube spacing. There is also a trade-off between these design
parameters, where changing one (e.g., closer tube spacing) can allow another parameter to be relaxed (e.g., less
conductivity additive). To aid in future device design, models of reduced complexity were developed and
evaluated for their ability to predict optimal designs. These simplified models can predict the performance of
thermal energy storage heat exchangers up to 5000 times faster than the finite-difference model. For the design
objectives of maximizing energy density, minimizing costs, and minimizing levelized cost of storage, the
simplified models predicted optimal designs with average absolute deviations below 2 % in relation to the
optimal designs chosen by the finite difference model optimization. These models can serve as tools to further
study thermal storage heat exchangers incorporating the realistic trade-off between the device's power and
energy.

1. Introduction researchers and designers have looked to the use of fins and additives to
increase the effective thermal conductivity of storage mediums [7–10].
Thermal processes consume significant amounts of energy in the In general, PCM composites (PCCs) will have increased thermal con­
industrial and building sectors [1]. Because of this, thermal energy ductivity in comparison to pure PCMs, but they will also have reduced
storage (TES) can play an important role in the transition to a carbon- energy density because the conductivity enhancing material displaces
free economy by shifting these thermal loads while providing services some of the PCM's volume. This leads to a design trade-off between
ranging from space conditioning [2,3] to grid-scale storage [4,5]. storage capacity and thermal conductivity for PCM composites, which
Additionally, TES can provide crucial thermal management to electronic impacts the energy and power density of the storage device.
equipment [6]. The operational demands of TES systems can vary The design of TES devices typically involves the tuning of material
widely for different applications. Thermal storage charge and discharge and device fabrication parameters to optimize a desired objective
processes can take place over days, hours, minutes, and seconds. function for specific use cases. Conventional design practice has been to
Different storage power requirements will require different optimal develop numerical or analytical models to optimize PCM TES devices
thermal conductivities to maximize device performance. To increase the using objective functions based on costs or some other energy related
power capabilities of TES devices using phase change materials (PCM), metric. Numerical optimization methods typically require iterating on

* Corresponding author.
E-mail address: Nelson.James@nrel.gov (N. James).

https://doi.org/10.1016/j.est.2022.105875
Received 16 June 2022; Received in revised form 25 September 2022; Accepted 12 October 2022
Available online 21 October 2022
2352-152X/© 2022 Elsevier Ltd. All rights reserved.
N. James et al. Journal of Energy Storage 56 (2022) 105875

complex discretized models to arrive at optimal solutions [11–13]. For indicators for PCM thermal storage devices. Specifically, numerical
example, Bianco et al. [14] performed multi-objective optimization on models are used to determine how the PCM transition temperature,
finned PCC heat sinks to reduce cost while maximizing operational time thermal conductivity enhancements, thermal power, and other design
for electronics cooling. They used detailed finite element models choices impact energy density. These results are then used to develop
incorporating fin spacing, conductivity additive porosity, and other simplified analytical models that can rapidly investigate the design
geometric aspects to capture the impacts of design choices on system space and produce trends over a range of operating conditions. These
performance. A separate study by Xu et al. [15] numerically investigated simplified models and trends will reduce the time to optimize a TES heat
the impact of porous inserts on the performance of shell and tube PCM exchanger and provide generalized guidance on designs based on the
storage devices for solar thermal power applications. They used a finite- application.
volume model of the porous PCC to maximize the melting performance
relative to the device costs. 2. Methods
Numerical design processes like these typically face two main
drawbacks. First, when using numerical models for TES design, signifi­ There are multiple other factors that determine the viability of TES
cant computational effort goes into determining an optimal solution for heat exchangers in practical applications. These include energy density,
a given operational point. This can lead to lengthy design times and slow energy-specific capital costs, charging energy consumption, and the
device development. Second, the solutions are focused on a single use levelized cost of storage (LCOS) [23], among others. In this work, stor­
case, and are typically only applicable to a specific set of operating age power draws are represented using a C-Rate, which represents the
parameters for the TES device. This means that the results developed power requirement needed for the total storage capacity to be dis­
cannot be easily generalized to aid in the design of other TES systems. 1
charged in C− Rate hours. Higher C-Rates correspond to higher power
Attempts have been made to address these challenges individually, but delivery. Another important criterion used in this TES optimization
no work has looked at the combination of developing generalizable TES study is the concept of the cutoff temperature. In practical applications,
design trends in conjunction with simplified models for rapid design the temperature supplied by a thermal storage device must stay within a
evaluation. predefined range to provide useful heating or cooling. For example, in a
Analytical models offer the advantages of lower computational times cooling application the storage output temperature will increase as the
and typically capture physical phenomena to allow for greater extrap­ device discharges. Eventually this temperature will cross the cutoff
olation than purely empirical tools, such as artificial neural networks temperature, and no longer be useful for the given application.
[16–19]. These models have been employed in device design optimi­
zation and can be useful when incorporating TES performance into
larger system models [20]. Like numerical models, previously developed 2.1. Design objectives
analytical models typically focus on analyzing TES in the context of a
single application. Though analytical models have the computational The goal of this analysis is to determine TES designs that satisfy three
speed to rapidly evaluate systems with different operational constraints, objective functions: maximizing the effective energy density, EDeff
the required assumptions often limit the analysis to specific TES archi­ (kWhth/m3), minimizing the energy-specific capital cost, CTES,kWh
tectures. No study incorporating a generalizable framework character­ ($/kWhth), and minimizing the LCOS ($/kWhe). Expressions are pre­
izing the impacts of operational temperatures and thermal power sented for these design objectives and are used to evaluate the perfor­
requirements on TES design have been published in the literature. mance of TES designs.
Attempts have been made to develop generalized frameworks for the The thermal Ragone framework shows the influence of operating
evaluation of TES materials and devices through the use of Ragone plots. conditions, geometry, and material properties on the effective energy
Ragone plots are a common way of exploring the operational capabil­ density, or the utilized capacity per unit volume. The effective energy
ities of energy storage devices. They depict the change in the usable density can be calculated with the nominal energy density (EDnominal)
energy density of a storage device under different power requirements, and the cutoff state of charge (SOCcutoff), as shown in Eq. (1). The
due to internal resistances that build up during discharge. Plots repre­ nominal energy density is determined by the physical amount of storage
senting Ragone relations have been widely used in the electrochemical material in the device, and the state of charge is defined at any point as
battery space to characterize storage devices in terms of power and the fraction of charge left in the device relative to the total capacity.
energy storage capabilities. More recently these methods have been When the device output reaches the cutoff temperature, the remaining
applied to TES devices. Yazawa et al. [21] used analytical expressions to state of charge (the cutoff state of charge) can be used to determine the
map thermal energy storage materials into a power energy space. They amount of charge utilized during the discharge cycle. This formulation
introduced a figure of merit, which when plotted against energy density, assumes system inefficiencies such as heat leakage to the ambient are
illustrated power and energy trade-off in material selection. Woods et al. negligible. Additionally, electrical energy consumption associated with
[22] described the trade-off between power and energy for TES devices balance of plant equipment such as pumps and fans were neglected
using rate capability and Ragone plots. They showed that not only does a when evaluating TES energy density.
power-energy trade-off exist for PCM thermal storage devices, but they ( )
EDeff = EDnominal • 1 − SOCcutoff (1)
also describe a framework to evaluate their power and energy perfor­
mance. The Ragone framework allows for a more comprehensive un­ The cutoff state of charge depends on the thermal power and re­
derstanding of how PCM based TES performance changes with sistances in the device, which are functions of the C-Rate, thermal
operational conditions. It presents a systematic method of classifying conductivity, and geometry. The nominal energy density depends on the
TES operational conditions and can allow for a more straightforward amount of active storage material and the device geometry. The thermal
assessment of how the device design, and its corresponding perfor­ storage device considered in this work is a planar TES heat exchanger,
mance, vary with operational conditions. While various researchers where heat is passed in and out of the device by a heat transfer fluid
have identified the existence of Ragone power-energy trade-offs in flowing through rectangular aluminum channels. PCM composite layers
thermal energy storage devices, none have analyzed how the combina­ are placed between adjacent fluid channels and consist of PCM mixed
tion of material, design, and operational choices influence the perfor­ with thermal conductivity enhancing additives. A diagram depicting the
mance and cost characteristics of thermal storage devices for optimal geometry of a unit cell of the planar TES heat exchanger modeled is
designs. shown in Fig. 1.
In this paper, the thermal Ragone framework is leveraged to inves­ The nominal energy density for a unit cell of this design can be
tigate the impact of design parameters on multiple performance determined using Eq. (2), which relates the storage capacity to the unit

2
N. James et al. Journal of Energy Storage 56 (2022) 105875

Fig. 1. Representation of planar TES heat exchanger unit cell geometry during a melting process. PCC storage media are alternated with tubes carrying heat transfer
fluid for storage charging and discharge.

cell volume including both the active and inactive material. Eq. (3) cold storage). The temperature difference between the charging fluid
represents the total nominal capacity (Capnominal) of the storage material inlet temperature and the transition temperature will depend on the TES
with density (ρPCM). The thickness of the composite (thPCC) and tube design. As the difference between this charging temperature and the
layers (thtube) as well as the porosity of the conductivity additive (φ) are cutoff temperature increases, so will the required energy for recharging
design parameters. Here porosity is defined as the ratio between the the storage. For a cooling application, Eq. (5) describes the ratio be­
volume of the pores and the total volume of the conductivity additive. A tween the equipment coefficient of performance (COP) during the
porosity of 50 % would denote that half of the additive consists of void charging process of the storage media and the COP avoided during the
space while 100 % porosity would denote that there is effectively no time when the TES is discharging. The charging COP (COPchg) is pro­
conductivity additive and only void space that would be occupied by portional to the Carnot COP defined by the low side temperature during
PCM. For a unit cell of the storage device, the PCM composite layer charging (TL, chg) and the high side temperature during charging (TH, chg).
thickness was defined as half the distance between the heat transfer The avoided COP (COPav) is proportional to the Carnot COP defined by
channels. The planar geometry of the device is defined by a specified the low side temperature during discharge period (TL, av), and the high
length (LHX) and width (WHX). The enthalpy of the PCM depends on the side temperature during discharge period (TH, av). TL, av can be consid­
properties of the PCM and the initial and cutoff temperatures for the ered equivalent to the cutoff temperature during the TES discharge
application. In this study, the ideal charged state is defined as the PCM process. During the discharging process, the TES system displaces
uniformly at the initial temperature with corresponding enthalpy (hPCM, cooling load that would have been provided by the equipment. As the
charged) and the discharged state as the PCM uniformly at the cutoff ratio of COPchg to COPav decreases, the relative energy consumption
temperature with corresponding enthalpy (hPCM, discharged). during the charging period to the energy consumption avoided during
the discharge period will increase.
Capnominal
EDnominal = (2) TL,chg
(2thPCC + thtube )LHX WHX
COPchg (TH,chg − TL,chg )
[( ) ] = TL,av (5)
Capnominal = hPCM,discharged − hPCM,charged ρPCM • 2thPCC LHX WHX φ (3) COPav
(TH,av − TL,av )
The effective energy density determined from the nominal capacity The upfront and operational costs can be combined using the LCOS.
and SOCcutoff can also be used to estimate the energy-specific capital cost Such metrics have been widely used to compare different energy tech­
(CTES,kWh), which represents the material cost of the TES heat exchanger nologies on a consistent basis. Odukomaiya et al. [24] introduced a
relative to the effective storage capacity of the device. Eq. (4) gives the novel formulation for the LCOS for TES in applications where the storage
energy-specific cost of the TES heat exchanger using the volumetric costs is charged with a heat pumping device. The formulation they developed
of the PCM, tubing, and additive materials. These formulations neglect allows for the fair comparison of TES cost-effectiveness in relation to
costs associated with insulating material. Volumetric costs of the PCM alternative storage technologies. The LCOS formulation described in Eq.
(cPCM), tubes (ctubes), and additive material (cadditive) were set to 1000; (6) incorporates aspects of the device energy-specific capital costs, depth
9840; and 2500 $/m3, respectively. Justification for these cost values of charge utilization (DT), annual storage utilization factor (uT), storage
can be found in the supplemental information, section 5. As cost can efficiency (ηSt), and COP characteristics using Carnot performance and
vary significantly with manufacturing methods, parametric analysis 2nd law efficiencies (ηII). For this analysis, storage utilization was
provided in the supplemental information illustrate the impacts of assumed to be constant and independent of C-Rate. Additional economic
relative material cost assumptions. Similar to energy density, aspects parameters include the discount rate (r) and electricity price (p). Further
associated with balance of plant equipment such as the cost of pumps details on the assumed parameter values used in the formulation can be
and vapor compression equipment were neglected. found in section 7 of the supplemental information. This work represents
1 the first attempt at TES design optimization based on this novel LCOS
CTES,kWh = (cPCM (2thPCC representation using a thermal Ragone framework.
EDeff (2thPCC + thtubes )
⎛ ⎞
• φ) + ctubes (thtube ) + cadditive (2thPCC • (1 − φ) ) ) (4)
CTES,kWh ⎜ 1 ⎟
Eq. (4) gives information about the upfront capital cost of a TES LCOS = + p⎜ ⎟
⎝η ⋅COPchg − 1⎠ (6)
∑
lT
− i
device but does not consider how the device design impacts the opera­ ηSt ⋅η DCOP
II
T •uT
av
• (1 + r) St COPav
i=1
tional costs. In many cases, mechanical refrigerating or heat pumping
equipment charges the TES device. When the thermal storage is not in Eqs. (1), (4), and (6) represent different objective functions that can
use, this equipment will need to provide a thermal load at the cutoff be used to optimize the design of a TES device. Each equation depends
temperature; however, while charging the TES, the set point tempera­ heavily on the nominal capacity of the device and the cutoff state of
ture will need to be below the transition temperature of the PCM (for charge. The nominal capacity can be directly calculated with knowledge

3
N. James et al. Journal of Energy Storage 56 (2022) 105875

of the material properties and the geometry, but the cutoff state of latent heat and density are parameters that will always be maximized to
charge is difficult to determine without a heat transfer model of the increase the storage capacity of TES heat exchangers. As such, they were
device. In this work, a detailed finite-difference model (FDM) and two excluded from the optimization study. Throughout this analysis, con­
simplified models are used to determine the cutoff state of charge for stant latent heat and density values were used to represent the PCM. The
planar TES heat exchangers with varying material and geometric initial temperature of the charged storage device was held constant at
properties. The design variables and models used to determine the cutoff 0 ◦ C.
state of charge are described in more detail below.
2.3. TES heat exchanger models

2.2. Design variables Several models were used to estimate the cutoff state of charge for
different TES device designs. These models have varying degrees of
Many design variables influence the effective energy density of a complexity and computation expense requirements. The most detailed
thermal storage device. To narrow the design space, a variance-based model used in this work is an FDM that discretizes both the fluid and
sensitivity analysis [25] was used to determine the material proper­ PCM composite material. Additionally, two simplified models were
ties, geometries, and operating conditions with the largest impact on developed to capture the power-energy trade-off characteristics without
effective energy density. The sensitivity analysis was performed using a the computational requirements of the FDM.
two-dimensional transient FDM developed and experimentally validated
by Woods et al. [22]. This model is publicly available from (https://gith 2.3.1. Finite-difference model
ub.com/NREL/ThermalRagone), and allows for the specification of PCM The FDM developed and experimentally validated by Woods et al.
thermophysical properties, incorporates the impacts of conductivity [22] was used for the optimization studies. Starting from a specified
enhancing additives, and uses a glycol-water as the heat transfer fluid. initial temperature, the model simulates the progression of phase change
Further details on the sensitivity analysis can be found in section 1 of the through the storage media when heat is added or removed from the
supplemental information. device and provides the fluid outlet temperature. During discharge, the
The sensitivity analysis showed that the PCM composite layer simulation continues until the outlet fluid temperature exceeds the
thickness, C-Rate, porosity of the conductivity additive, and driving cutoff temperature. While charging, the simulation runs until the TES
temperature difference (difference between the cutoff and phase tran­ device reaches 100 % state of charge. A rate-capability curve repre­
sition temperatures) have the largest impact on the effective energy senting a discharge process can be found in section 3 of the supplemental
density. These parameters were varied to optimize the TES design for the information. The model assumes that heat transfer through the phase
remainder of this analysis. To simulate a range of different applications, change composite is conduction dominated.
C-Rates between 1/6 and 3.0 and driving temperature differences from 3 Given the relatively large computational expense of the FDM, a
to 11 ◦ C were evaluated. The driving temperature difference was varied parametric assessment was used to approach the optimal design for each
by changing the transition temperature (material property) of a baseline performance objective (maximum energy density, minimum cost, and
PCM and setting the cutoff temperature to 12 ◦ C (system constraint) to minimum LCOS). The operating and design parameter values included
simulate a space cooling application. For real PCMs thermophysical in the parametric assessment are shown in Table 2. After using the FDM
properties like density, phase change enthalpy, and density are directly to generate results for each combination of parameters in Table 2, the
tied to the type of PCM material chosen. For this analysis a fictitious design characteristics of the optimal conditions for each design objective
PCM material with tunable transition temperature was analyzed in order were determined using the model outputs and Eqs. (1) through (6).
to isolate the effect of parameter changes such as driving temperature Energy density and energy-specific capital costs were directly
differences independently of other material property changes for a fixed calculated using Eqs. (1) through (4) and the state of charge at the end of
end-use application. The important geometric and material properties the discharge cycle (SOCcutoff). To determine LCOS, information on
were varied by evaluating devices with PCM composite layer thickness charging energy requirements for the TES device are needed. This was
between 1 and 20 cm and conductivity additive porosities between 80 % determined by simulating a 4-hour charge cycle (representative of
and 100 %. Two common conductivity enhancing additives were building and grid applications), starting at the cutoff state of charge for a
investigated. The properties of these additives are shown in Table 1. For given discharge C-rate and driving temperature difference. It was
expanded graphite, relative density values as well as latent heat ratios assumed that the temperature throughout the fluid and phase change
taken form literature were used to develop correlations for thermal material was uniform (fully relaxed) at the start of the charge cycle. Heat
conductivity and density as a function of porosity. For graphene plate­ transfer to the ambient was not considered in this analysis. The fluid
lets density values were taken from literature and linearly varied with inlet temperature was tracked throughout the process to determine an
porosity. Additionally, data points were also taken form literature to average inlet temperature. Using this temperature as a proxy for
develop correlations between graphene platelet thermal conductivity charging energy requirements (TL, chg), Eqs. (5) and (6) could be used to
and porosity. Expanded graphite was used as the default conductivity determine the LCOS for the TES heat exchanger. The high side
additive for this analysis.
Several design parameters were held constant through the analysis.
Table 2
Heat transfer fluid channel thicknesses of 3 mm were assumed to Variables used in TES heat exchanger model parametric assessment.
simulate microchannel devices with wall thicknesses of 0.5 mm. The
Parameter Values
planar length and width of the TES heat exchanger was held constant at
approximately 0.5 m and 0.25 m, respectively. The combination of PCM Initial temperature [◦ C] 0
Fluid temperature 4
difference [◦ C]
Table 1 Cutoff temperature [◦ C] 12
Properties of conductivity enhancing materials. PCM latent heat [kJ/kg] 220
PCM density [kg/m3] 784
Material Conductivity enhancer Conductivity enhancer Reference
C-Rate [1/h] 1/6, 1/4, 1/3, 1/2, 3/4, 1, 1.25, 1.5, 1.75, 2, 2.5, 3
conductivity [W/m-K] density [kg/m3]
Phase change temperature 1, 2, 3, 4.6, 6.5, 8, 9
Expanded 0.2 + 121.27 • (1 − φ) 1253.9 − 1246.5(φ) [26,27] [◦ C]
graphite Thickness [cm] 1, 1, 3, 4, 5, 6, 7, 8, 9, 10, 12.5, 15, 17.5, 20
Graphene 0.2 + 24.67 • (1 − φ) 2260 • (1 − φ) [28] Porosity 1, 0.99, 0.98, 0.97, 0.95, 0.9375, 0.925, 0.9125, 0.9,
platelet 0.875, 0.85, 0.8

4
N. James et al. Journal of Energy Storage 56 (2022) 105875

temperature during the charging (TH, chg) and discharging (TH, av) pe­ simplicity, the fluid convective, wall conductivity, and contact re­
riods were assumed to be 21 ◦ C and 40 ◦ C, respectively, to simulate a sistances are lumped into a single area-specific surface resistance (Rsurf′′ )
location with a diurnal temperature swing. in Eq. (9). The derivation of eq. 9 is shown in section 2 of the supple­
The model predicted fluid pressure drop through the channels from 2 mental information.
to 100 Pa, which varied with C-rate and tube spacing. Using these √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( ) ̅
pressure drops, the mechanical pumping power was 4 to 7 orders of ′ ′
2
′
− kPCC Rsurf + ′
kPCC Rsurf + Crate4kϕΔh
PCC LMTD
c− d ρ
magnitude smaller than the thermal power delivered by the storage (9)
PCM
thPCM =
2
device. This allowed for this work to neglect the pumping work re­
quirements and focus solely on the thermal transfer optimization of the The effective energy density was then determined using Eqs. (1) and
TES devices. (2), with a cutoff state of charge of 0 %. Like the FDM, capital costs were
estimated with Eqs. (3) and (4) using this approximated energy density.
2.3.2. Parallel phase front approximation analytical model Estimating the charging energy and LCOS (Eq. (6)) require estimates of
To reduce the computational time for design optimization compared the fluid temperature during charging. A similar resistance-based
to the relatively slow FDM, simplified representations of the TES heat approach was used to find the charge temperature bounds, as shown
exchanger were developed. Previous efforts have attempted to develop in Eqs. (10) to (12), using the volume of the TES heat exchanger (VTES).
analytical models for phase change-based TES systems [29,30]. How­ The low side temperature during charging is then used with the ambient
ever, none were found that explicitly incorporated the operating con­ conditions in Eq. (5).
dition dependent power-energy trade-offs of the present Ragone ( )
CRate,charge EDnominal VTES ′ ′ ΔTf
framework. Simplifications of the TES heat exchanger were developed to Tf ,ininitial = Tt,min − Rsurf − (10)
3600 2
capture these influences while still minimizing computational re­
quirements. The Parallel Phase Front Approximation (PPFA) analytical ( )
CRate,charge EDnominal VTES ( ) ΔT
model removes some of the complexity of the FDM by assuming that the (11)
′′ f
Tf ,infinal = Tt,min − RPCM,cutoff + Rsurf −
3600 2
phase front remains parallel to the fluid channel during the discharge
process. This approach also assumes that for a melting process, the liquid Tf ,ininitial + Tf ,infinal
fraction is equivalent to the state of charge. TL,chg = (12)
2
The resistance between the fluid and the phase front will dictate the
driving temperature difference needed to reach the specified discharge More traditional optimization techniques could be used in conjunc­
rate. TES heat exchangers have several thermal resistances between the tion with the PPFA analytical model given its significantly lower
fluid and the PCM that do not change while the device discharges, computational requirements. The optimal porosity for each objective
including the fluid convective (Rf), wall conductive (Rwall), and channel- function was found using a quadratic approximation optimization
PCM contact resistances (Rc). Heat also must pass through the PCM method [31] as a function of the driving temperature difference and
liquid layer that builds up next to the fluid channel while the device discharge C-Rate requirements.
discharges. The value of this liquid PCM resistance (RPCM) when the fluid
outlet temperature reaches the cutoff can be calculated using Eq. (7), 2.3.3. Lumped mass analytical model
where the driving temperature difference is defined with the log-mean A lumped mass analytical model was also developed to estimate the
temperature difference between the phase front (at the maximum three objective functions. Starting with the formulations of Yazawa et al.
transition temperature, or Tt, max) and the fluid that operates with a [21], the time dependence of the lumped mass normalized temperature
temperature change (ΔTf). The discharge rate is defined using the C-rate (θ) can be defined using the cutoff and initial temperatures for the
and the total device capacity. operational temperature range.
⎡ ⎤ ( ( t) )
θ(t) = 1 − exp − (13)
1 ⎢ ⎥ τ
RPCM,cutoff = ⎢ ( ΔTf )⎥ (7)
CRate Capnominal ⎣ln Tcutoff +ΔTf − Tt,max ⎦ − Rf − Rwall − Rc
T(t) − Tinitial
Tcutoff − Tt,max
θ(t) = (14)
Tcutoff − Tinitial
During discharge, the liquid layer in PCM composites is conduction- A nominal time constant (τo) represents the product of the effective
dominated [22], and therefore the PCM resistance is related to the thermal resistance and capacitance of the TES material, as shown in Eq.
thickness of the liquid layer (thl = RPCMkPCMAHX). The liquid fraction (15). To capture the transient nature of the discharge process, the
when the fluid outlet temperature reaches the cutoff can be calculated resistance was defined as half of the conduction resistance in the PCC
by comparing this thickness to the total PCM thickness (thPCM). In this layer along the direction of the characteristic length. The surface resis­
analysis, it was assumed that the liquid fraction was equivalent to one tance of the heat transfer fluid was assumed to be negligible in com­
minus the state of charge, which allowed for directly calculating the parison to the conduction resistance. The effective capacitance of the
cutoff state of charge (Eq. (8)). The effective energy density can then be PCC was determined using a similar formulation to that found in Mallow
calculated using the nominal energy density and the utilization as et al. [26]. A scaling factor based on the ratio of the temperature dif­
determined by the state of charge at cutoff, as described by Eq. (1). ference between cutoff and initial temperatures and the temperature
( ) difference between the cutoff and phase change transition temperature
thl,cutoff
SOCcutoff ≈ max 0, 1 − (8) was appended to the nominal time constant to determine the effective
thPCM
time constant for the storage medium (τ) using Eq. (16). This accounts
Using this simplified model, the energy density is always maximized for the variations in driving temperature differences between a lumped
when the state of charge reaches 0 %, or when the liquid thickness (in mass model with sensible heat storage and the PCC under investigation.
the case of melting) at the cutoff state of charge equals the total PCM [ ] [( )
thickness. This relationship allows for the calculation of the optimal τo =
1 thPCC
cpPCC +
hmelt,PCC
ρ (LHX WHX
PCM thickness for a given porosity φ (Eq. (9)), which sets both the 2 kth • 2(LHX WHX ) Tcutoff − Tinitial pcc
]
composite conductivity using Table 1 and the PCM composite enthalpy
• 2thPCC ) (15)
change (Δhc− d, PCC = ϕΔhc− d where Δhc− d is the enthalpy difference
between the charged and discharged state of the pure PCM). For

5
N. James et al. Journal of Energy Storage 56 (2022) 105875

[ ]
T − Tinitial
τ = τo cutoff (16)
Tcutoff − Tt,avg

The time required to fully discharge the lumped mass storage device
(tlumped) can then be estimated by setting the normalized temperature to
99 % and solving for the time.
tlumped = − τ • ln(1 − 0.99) (17)

A value of 99 % was chosen to capture most of the discharge process

in a finite period. Knowing the time required for discharge as defined by
the lumped mass model and the time available for discharge as deter­
mined by the specified C-Rate, an expression similar to that for the PPFA
analytical model can be determined to represent the state of charge at
cutoff.
⎛ ⎞
3600
⎜ ⎟
SOCcutoff ≈ max⎜
⎝0, 1 −
CRate ⎟ (18)
tlumped ⎠

This was then used in Eq. (1) to calculate effective energy density Fig. 2. Variation in nominal energy density with TES heat exchanger conduc­
from nominal energy density. Eq. (3) and Eq. (4) then give the energy- tivity enhancer porosity and PCM composite layer thickness.
specific capital costs from this effective energy density.
The charging temperature was estimated as the average between the the total volume, lowering the nominal energy density. Once the PCC
inlet fluid temperature at the beginning and end of the charging process. layer's thickness gets sufficiently larger than the channel's thickness, the
The time required to charge was determined by the specified charging C- rate of change of nominal energy density with layer thickness begins to
rate. The initial fluid temperature was taken as the difference between decrease. As the porosity of the conductivity enhancing material de­
the average transition temperature (Tt, avg) and the fluid temperature creases, more phase change material is displaced, and the nominal en­
difference across the device. The final temperature was estimated by ergy density drops. Because the nominal energy density depends
setting the normalized temperature in Eq. (13) to be equal to 99 % and primarily on the material properties and geometry, these trends will
solving for the effective time constant. The nominal time constant hold regardless of the PCM used in planar TES devices.
scaling factor was modified to represent the driving temperature dif­
ference between the transition temperature and the unknown final
3.2. Maximizing effective energy density
temperature. To account for the fluid temperature glide, half of the fluid
temperature difference was subtracted from the resulting expression to
The effective energy density depends on the nominal energy density
estimate the fluid conditions at the inlet.
and the charge utilization (1 – SOCcutoff), as shown in Eq. (1). Fig. 3
⎛ ⎞
( ) depicts the charge utilization of a planar TES heat exchanger as a
τo (Tcutoff − Tinitial )
function of C-Rate, driving temperature difference, PCC layer thickness,
⎝ ΔTf ⎠
Tt,avg − ΔTf + Tt,avg − − 3600 − 2
and conductivity enhancer porosity, as calculated by the FDM.
CRate,charge

(19)
ln(1− 0.99)
TL,chg =
2 Expanded carbon was used as the conductivity enhancing material. The
This temperature was used in Eq. (5) to estimate TES heat exchanger x and y axis in each subplot correspond to the values of the axis in Fig. 2
design effects on charging energy requirements using the previous for thickness and porosity, respectively. Fig. 3 shows that the charge
ambient conditions. Eq. (6) was used to estimate the LCOS using the utilization increases with decreasing PCC layer thickness and decreasing
outputs of Eq. (5) plus the estimated energy-specific capital costs. porosity of the conductivity-enhancing material. Additionally, charge
In a similar manner to the PPFA analytical model, optimal configu­ utilization increases with larger driving temperature differences and
rations for the lumped mass model could be found using more conven­ lower C-Rates for any given porosity and thickness combination. These
tional optimization methods. Optimal configurations for each design trends can be readily explained through an examination of Fourier's
objective were found using combinations of nonlinear optimization Law. As thermal conductance increases (thinner composite layers or
routines [32] and exhaustive search approaches. more conductivity additives), greater amounts of heat can be transferred
through a material for a given temperature difference. Similarly,
3. Results and discussion increasing the driving temperature difference allows for higher rates of
heat transfer for a given thermal resistance.
The three modeling approaches developed were used as tools to Fig. 4 illustrates the combined effects of nominal enegy density and
investigate the design trade-off for the TES heat exchanger objective charge utilization for the TES heat exchanger described by the FDM.
functions of effective energy density, energy-specific capital costs, and Again, the axes of the subplots correspond to the axes of the plot in Fig. 2
LCOS. By incorporating the thermal Ragone framework, these models and expanded carbon was used as the conductivity enhancing material.
capture the impacts of operational, material, and geometric character­ At low C-Rates and high driving temperature differences, near-optimal
istics on the effective energy density of TES devices. energy densities can be achieved for a wide range of porosity and
thickness combinations. Because the power requirements are low and
the driving temperature differences are high, the TES device can tolerate
3.1. Nominal energy density higher thermal resistances before reaching the cutoff condition. As the
driving temperature difference is reduced or the C-Rate increases, the
The nominal energy density represents the upper limit for the number of combinations of porosity and thickness that can achieve large
effective energy density. It depends on the PCC layer thickness and effective energy densities shrinks. This is because designs with increased
porosity of the carbon conductivity enhancer (Fig. 2), which were thermal resistance (greater porosity and thickness) reach the cutoff
calculated using the parameters in Table 2. As the PCC layer gets condition sooner.
thinner, the heat transfer fluid channels take up more space relative to From Fig. 4 it is clear that for a given combination of C-Rate and

6
N. James et al. Journal of Energy Storage 56 (2022) 105875

Fig. 3. Variation in charge utilization (1-SOCcuotoff) with driving temperature difference and discharge C-Rate. Subplot axes represent increasing PCC thickness (x-
axis) and increasing conductivity enhancer porosity (y-axis).

Fig. 4. Variation in effective TES heat exchanger energy density with driving temperature difference and discharge C-Rate. Subplot axes represent increasing PCC
thickness (x-axis) and increasing conductivity enhancer porosity (y-axis).

driving temperature difference there exists a conductivity enhancer assumed PCM latent heat capacity and density of 220 J/g and 784 kg/
porosity and PCC layer thickness combination that results in the highest m3, respectively, used in the baseline FDM. Changes to these values will
energy density for the TES heat exchanger. The results shown are valid result in changes to the optimal thickness and porosity corresponding to
for the specified conductivity enhancer properties of expanded graphite, the maximum effective energy density.
the specified heat transfer fluid channel thickness (3 mm), and the To aid in the design of TES heat exchangers, the ability of the

7
N. James et al. Journal of Energy Storage 56 (2022) 105875

analytical models to determine optimal configurations was compared to As discussed earlier, the results above are applicable to just the
that of the FDM based parametric assessment (Fig. 5). First the optimal baseline design. The simplified models can be used to easily explore
thicknesses and porosities for each combination of C-Rate and driving other baseline designs. This includes using the less effective graphite
temperature difference were found using the analytical models. Then platelets, which have lower conductivities for a given porosity. This
these optimal thickness and porosity values were input into the FDM, approach requires thinner PCC layers and lower porosities to maximize
which then gave the FDM-predicted effective energy density. The the TES heat exchanger energy density. Cost or availability may also
analytical models predicted most optimal energy densities to within 5 % necessitate the use of heat transfer fluid channels with larger thick­
of the FDM optimal values, with the lumped mass model showing less nesses, such as the 6-mm case shown in Fig. 6. The optima for this case
error than the PPFA analytical model. This indicates that the assumption are thicker PCC layers and lower porosities. The thicker PCC layers help
of the phase front being parallel to the fluid channel was not as accurate counteract the displacement of the PCM by using thicker tubes, and the
as desired for these specific design conditions. Further explaination of lower porosites help maintain sufficient thermal conductance as the PCC
this behavior can be found in section 3 of the supplemental informaiton. layer thickness increases. Fig. 7 depicts the resulting changes in the
The discrete nature of the thicknesses and porosities from the para­ energy density distribution for these design changes. For moderate C-
metric FDM optimization are evident from Fig. 5. This highlights an Rates and driving temperature differences, energy density reductions in
advantage of the simplified models in that they are able to evaluate the range of 4 kWh/m3 result from using graphene platelets, and re­
combinations in a continuous manner with minimal computational ductions around 3 kWh/m3 result from using 6-mm tubes. Altering the
time. The PPFA model and the lumped mass analytical models can PCM material properties can also lead to changes in optimal configu­
simulate effective energy densities more than 5000 times faster than the rations for the TES heat exchanger. Additional analysis on the impacts of
FDM for specified TES heat exchanger operating conditions and design varying the intrinsic latent heat of the PCM on the optimal designs for
characteristics. This can speed up the evaluation of potential design maximizing energy density can be found in section 4 of the supplemental
options. Although the analytical models do not choose the exact same informaiton.
thickness and porosity values as the FDM optimization, the combina­
tions they choose approach the values of the optimal condition and as
3.3. Minimizing energy-specific capital costs
such can be used to understand design trends for TES heat exchangers.
The lumped mass model was used to develop the porosity-thickness
The energy-specific capital costs (Eqs. (3) and (4)) was estimated
combination in Fig. 6. As the driving temperature difference increases
using the FDM for a range of C-Rates and driving temperature differ­
and the discharge C-Rate decreases, the highest energy densities come
ences. Fig. 8 shows the inverse of the energy specific capital costs with C-
from the combination of thicker PCC layers and more porous conduc­
Rate, driving temperature difference, thickness, and porosity. Expanded
tivity enhancers. This is because high porosities displace less PCM while
carbon was used as the conductivity enhancing material. The inverse of
still maintaining sufficiently high utilization. At lower C-Rates, the
the costs was plotted to show distinctions more clearly between optimal
optimal porosities and thicknesses are less sensative to variations in
points. Larger thicknesses and lower porosities are generally favored in
driving temperature differences. Higher C-Rates require higher thermal
comparison to the optimal values for maximizing the effective energy
conductance, and therefore the optimal design uses thinner PCC layers
density.
with more conductivity enhancement. Even though this displaces more
Like the energy density analysis above, the analytical models were
PCM, this approach more fully utilizes the available capacity, allowing
also used to find thickness and porosity combinations that minimized
for higher effective energy densities.
the energy-specific capital costs ($/kWh) of the TES heat exchanger. The

Fig. 5. Parity plots comparing thickness and porosity values chosen by the approximate models and resulting FDM energy densities to the values determined by the
FDM parametric optimization. Error quantified using average absolute deviation (AAD).

8
N. James et al. Journal of Energy Storage 56 (2022) 105875

Fig. 6. Variation in optimal composite layer thickness and carbon porosity for specified operating conditions to maximize energy density as predicted by the lumped
mass model. Scenarios represent impacts of different material choices in the TES heat exchanger.

Fig. 7. Variation in TES heat exchanger volumetric energy density with changes in conductivity additive material and heat transfer fluid channel thickness as
predicted by the lumped mass model.

thickness and porosity selected by the analytical models were input into approximation. Parametric assessments highlighting the sensitivity of
the FDM to determine the FDM-predicted energy densities. Fig. 9 dis­ the calculated energy specific capital costs to the model input parame­
plays the accuracy of the approximate models in finding conductivity ters can be found in section 5 of the supplemental material.
enhancer porosity and thickness combinations that minimize the capital
costs. Expanded carbon was used as the conductivity enhancing mate­ 3.4. Minimizing levelized cost of storage
rial. For the design conditions investigated, the PPFA heat exchanger
model proved to be highly accurate in estimating the TES heat To estimate the LCOS, the energy-specific capital costs was combined
exchanger costs. This is likely because the parallel phase front with the costs for charging the TES, which depend on the fluid tem­
assumption holds for the conditions investigated. Although less accu­ perature required during charging. Comparisons between the charging
rate, the lumped mass approach also provided reasonable estimates of temperatures predicted by the two simplified models and the FDM are
the capital costs. The optimal solution for minimizing capital costs uti­ shown in section 6 of the supplemental material. In all cases, designs
lizes higher PCC layer thicknesses, which may have impacted the ac­ that maximize thermal conductance in the TES heat exchanger will lead
curacy of the averaged properties used in the lumped mass to lower energy consumption during the charging phase. Fig. 10 shows

9
N. James et al. Journal of Energy Storage 56 (2022) 105875

Fig. 8. Variation in the inverse of the energy specific captial costs with driving temperature difference and discharge C-Rate. Subplot axes represent increasing PCC
thickness (x-axis) and increasing conductivity enhancer porosity (y-axis).

Fig. 9. Parity plots comparing thickness and porosity values chosen by the approximate models and resulting FDM energy-specific capital costs to the values
determined by the FDM parametric optimization. Error quantified using average absolute deviation (AAD).

the inverse of the LCOS with C-Rate, driving temperature difference, combinations that minimize the LCOS is shown in Fig. 11. Expanded
thickness, and porosity, as estimated by the FDM. The inverse of LCOS carbon was used as the conductivity enhancing material. Both approx­
was plotted to show distinctions more clearly between optimal points. imate models show similar accuracy and select near-optimal points for
Expanded carbon was used as the conductivity enhancing material. minimizing LCOS. The lumped mass model better predicts the charging
The ability of the approximate models to find porosity and thickness temperature while the PPFA model better determines capital costs. The

10
N. James et al. Journal of Energy Storage 56 (2022) 105875

Fig. 10. Variation in the inverse of the LCOS with driving temperature difference and discharge C-Rate. Subplot axes represent increasing PCC thickness (x-axis) and
increasing conductivity enhancer porosity (y-axis).

Fig. 11. Parity plots comparing thickness and porosity values chosen by the approximate models and resulting FDM LCOS to the values determined by the FDM
parametric optimization. Error quantified using average absolute deviation (AAD).

porosity values chosen by the approximate models are notably different investigated, the optimal LCOS values are relatively insensitive to the
than those chosen by the FDM, while the PCC layer thicknesses chosen porosity of the conductivity enhancer when the porosity is below the
more closely align with those of the FDM. This leads to the conclusion range of 95 %. Additional information on the sensitivity of the LCOS to
that for the model assumptions and TES heat exchanger device the chosen input parameters can be found in section 7 of the

11
N. James et al. Journal of Energy Storage 56 (2022) 105875

supplemental material. 4. Conclusion

3.5. Variations in optimal designs The presented modeling study investigated the impacts of design
choices on the performance of TES heat exchangers under a wide range
The analysis above demonstrates how optimal designs for TES heat of operating conditions using a thermal Ragone Framework. The
exchangers are dependent on the expected C-Rate and driving temper­ thickness of the PCC layers, the porosity of conductivity enhancing ad­
ature difference imposed by the end-use requirements and PCM transi­ ditives, the C-Rate at which the device is operated, and the driving
tion temperature. Comparisons of the designs chosen by the FDM for the temperature difference are among the most impactful design parameters
three design objectives are shown in Fig. 12. In each of the subplots the for PCM heat exchangers. Primary design objectives included maxi­
optimal porosity or thickness for a given design objective is plotted on mizing effective energy density, minimizing energy-specific capital
one axis and compared to the corresponding optimal value for another costs, and minimizing the LCOS. A novel expression for the evaluation of
design objective on the other axis. Expanded carbon was used as the LCOS for building based TES has been incorporated into detailed TES
conductivity enhancing material. A parity line is drawn through each heat exchanger design for the first time.
subplot to indicate the locations where the two design objectives would Optimal ranges for thickness and porosity were shown to vary
be optimized by the same value of porosity or thickness. Points above significantly depending on the PCM transition temperature relative to
the parity line indicate that for the design variable under consideration, the cutoff temperature, the desired C-Rate, and the target objective
the design objective on the Y axis requires higher values of the specified function. For all scenarios investigated, as the C-Rate decreases and the
thickness or porosity. Similarly points below the parity line indicate the driving temperature difference increases, the porosities and thicknesses
design objective on the X axis requires higher values of thickness or needed to optimize energy density, cost, or LCOS will both increase.
porosity to be optimized. Conversely, higher C-Rates and lower driving temperature differences
The optimal design for maximizing effective energy density uses thin necessitate thinner PCC layer thicknesses and lower porosity conduc­
PCC layers with high porosity conductivity enhancers. The optimal tivity additives for optimizing energy density, cost, or LCOS. These
design shifts toward thicker PCC layers and lower porosity conductivity design trends can serve as quick guides for future TES heat exchanger
enhancers for minimizing cost and LCOS. This is because the benefit-to- designers seeking to implement design changes in response to different
cost ratio of adding more graphite exceeds the benefit-to-cost ratio for operating criteria or device objectives. For the assumed material cost
adding more tubes for the cost assumptions used in this study. estimates, minimizing LCOS or energy-specific capital costs requires
Using the thermal Ragone framework to understand how opera­ thicker PCC layers with less conductivity enhancing material compared
tional, material, and geometric characteristics influence design out­ to those needed to maximize the effective energy density.
comes can allow for more intentional design of TES devices to meet user Simplified models that incorporate elements of the Ragone frame­
needs. For example, achieving a target energy specific capital cost or work were presented to capture the trends of more detailed finite-
LCOS can be done by varying a combination of discharge powers, difference numerical simulations. These models can help understand
driving temperature differences, tube spacing, and conductivity addi­ design trends and trade-offs for numerous design objectives with rela­
tives. The results and models presented in this study can help designers tively minimal computational expense (as much as 5000 times less than
more intelligently configure systems toward specific objectives. that required for FDMs). Even though this study was performed with a
fictitious PCM, real materials can be incorporated into the models
through the appropriate modification of thermophysical properties. The
PPFA analytical model predicted optimal design configurations for

Fig. 12. Comparison of optimal thickness and porosity combinations chosen based on TES heat exchanger design objective function.

12
N. James et al. Journal of Energy Storage 56 (2022) 105875

energy density, energy specific capital costs, and LCOS with average PCM, discharged PCM at fully charged state
absolute deviations of 1.6 %, 0.15 % and 0.24 % respectively. The s solid
lumped mass analytical models predicted optimal design configurations st storage
for energy densities, energy specific capital costs, and LCOS with sub subcooling
average absolute deviations of 0.47 %, 1.7 % and 0.28 % respectively. super superheating
These simplified models can be easily integrated into more compre­ surf surface
hensive system models incorporating cost and performance character­ tube heat transfer tube channels
istics of other balance of plant equipment. This will allow system T thermal energy storage
designers to evaluate plant configurations with more realistic repre­ th thermal
sentations of TES heat exchanger capabilities under varying operating t, avg average transition temperature
conditions. Future work can explore the use of these modeling tech­ t, max maximum transition temperature accounting for glide
niques in conjunction with multi-objective optimization methods to t, min minimum transition temperature accounting for glide
designing TES heat exchangers that optimally satisfy multiple design TES thermal energy storage
goals and user needs. TES, kWh thermal energy storage capacity

Nomenclature CRediT authorship contribution statement

A area [m2] Nelson James: Writing – original draft, Methodology, Software,

c costs [$] Visualization. Allison Mahvi: Methodology, Software, Writing – review
cp specific heat capacity [J/kg-K] & editing. Jason Woods: Conceptualization, Writing – review & editing,
COP coefficient of performance [− ] Supervision, Funding acquisition.
D depth of discharge [− ]
ED energy density [kWh/m3] Declaration of competing interest
η efficiency [− ]
h enthalpy [J/kg] Nelson James reports financial support was provided by Department
k thermal conductivity [W/m-K] of Energy EERE Building Technologies Office.
LHX heat exchanger length [m]
l lifetime [yr] Data availability
LCOS levelized cost of storage [$/kWhe]
M mass [kg] Data will be made available on request.
P electricity price [$/kWhe]
φ porosity [m] Acknowledgements
r discount rate [− ]
ρ density [kg/m3] This work was authored in part by the National Renewable Energy
R resistance [K/W] Laboratory (NREL), operated by Alliance for Sustainable Energy, LLC,
SOC state of charge [− ] for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-
T temperature [K] 08GO28308. Funding provided by U.S. DOE Building Technologies Of­
t time [s] fice. The views expressed in the article do not necessarily represent the
τ time constant [s] views of the DOE or the U.S. Government. The U.S. Government retains
th PCM composite layer thickness [m] and the publisher, by accepting the article for publication, acknowledges
thl liquid layer thickness [m] that the U.S. Government retains a nonexclusive, paid-up, irrevocable,
ΔT temperature difference [K] worldwide license to publish or reproduce the published form of this
u utilization frequency [1/yr] work, or allow others to do so, for U.S. Government purposes.
V volume of TES heat exchanger [m3] The authors would like to thank S. Mumme from the DOE Building
WHX heat exchanger width [m] Technologies Office for his support for thermal energy storage analysis.

Subscripts Appendix A. Supplementary data

additive conductivity enhancing additive Supplementary data to this article can be found online at https://doi.
av avoided usage period org/10.1016/j.est.2022.105875.
chg charging
c contact References
e electric
eff effective property [1] International Energy Agency, Renewables 2019: Analysis and forecast to 2024,
2019. Paris.
f fluid [2] A. Mavrigiannaki, E. Ampatzi, Latent heat storage in building elements: a
H, av high side temperature during avoided usage period systematic review on properties and contextual performance factors, Renew. Sust.
H, chrg high side temperature during charging Energ. Rev. 60 (2016) 852–866.
[3] F. Souayfane, F. Fardoun, P. Biwole, Phase change materials (PCM) for cooling
II second law efficiency applications in buildings: a review, Energy Build. 129 (2016) 396–431.
in inlet stream [4] P. Sharan, C. Turchi, P. Kurup, Optimal design of phase change material storage for
l liquid steam production using annual simulation, Sol. Energy 185 (2019) 494–507.
[5] P. Albertus, J. Manser, S. Litzelman, Long-duration electricity storage applications,
L, chrg low side temperature during charging
economics, and technologies, Joule 4 (2020) 21–32.
L, av low side temperature during avoided usage period [6] M. Barako, S. Lingamneni, J. Katz, T. Liu, K. Goodson, J. Tice, Optimizing the
melt melting process design of composite phase change materials for high thermal power density,
PCC phase change composite J. Appl. Phys. 124 (2018), 145103.
[7] A. Abdulateef, S. Mat, J. Abdulateef, K. Sopian, A. Al-Abidi, Geometric and design
PCM phase change material parameters of fins employed for enhancing thermal energy storage systems: a
PCM, charged PCM at fully charged state review, Renew. Sust. Energ. Rev. 82 (2018) 1620–1635.

13
N. James et al. Journal of Energy Storage 56 (2022) 105875

[8] H. Ji, D. Sellan, M. Pettes, X. Kong, J. Ji, S. Li, R. Ruoff, Enhanced thermal [20] F. Rouault, D. Bruneau, P. Sebastian, J. Lopez, Multiobjective optimization of an
conductivity of phase change materials with ultrathin-graphite foams for thermal air-phase change material heat exchanger for free cooling in a single-family house,
energy storage, Energy Environ. Sci. 7 (2014) 1185–1192. J. Energy Storage 32 (2020), 101944.
[9] N. Kumar, J. Hirschey, T. LaClair, K. Gluesenkamp, S. Graham, Review of stability [21] K. Yazawa, P. Shamberger, T. Fisher, Ragone relations for thermal energy storage
and thermal conductivity enhancements for salt hydrates, J. Energy Storage 24 technologies, Front. Mech. Eng. 5 (2019). Paris.
(2019), 100794. [22] J. Woods, A. Mahvi, A. Goyal, E. Kozubal, A. Odukomaiya, R. Jackson, Rate
[10] N. Ibrahim, F. Al-Sulaiman, S. Rahman, B. Yilbas, A. Sahin, Heat transfer capability and ragone plots for phase change thermal energy storage, Nat. Energy 6
enhancement of phase change materials for thermal energy storage applications: a (2021) 295–302.
critical review, Renew. Sust. Energ. Rev. 74 (2017) 26–50. [23] O. Schmidt, S. Melchior, A. Hawkes, I. Saffell, Projecting the future levelized cost of
[11] J. Mahdi, S. Lohrasbi, D. Ganji, E. Nsofor, Simultaneous energy storage and electricity storage technologies, Joule 3 (2019) 80–100.
recovery in the triplex-tube heat exchanger with PCM, copper fins and Al2O3 [24] A. Odukomaiya, J. Woods, N. James, S. Kaur, K. Gluesenkamp, N. Kumar,
nanoparticles, Energy Convers. Manag. 180 (2019) 949–961. S. Mumme, R. Jackson, R. Prasher, Addressing energy storage needs at lower cost
[12] R. Srikanth, P. Nemani, C. Balaji, Multi-objective geometric optimization of a PCM via on-site thermal energy storage in buildings, Energy Environ. Sci. 14 (2021)
based matrix type composite heat sink, Appl. Energy 156 (2015), 703714. 5315–5329.
[13] S. Sridharan, R. Srikanth, C. Balaji, Multi objective geometric optimization of phase [25] J. Herman, W. Usher, SALib: an open-source python library for sensitivity, J. Open
change material based cylindrical heat sinks with internal stem and radial fins, Source Softw. 2 (2017) 97.
Therm. Sci. Eng. Prog. 5 (2018) 238–251. [26] A. Mallow, O. Abdelaziz, S. Graham, Thermal charging performance of enhanced
[14] N. Bianco, S. Busiello, M. Iasiello, G. Mauro, Finned heat sinks with phase change phase change material composites for thermal battery design, Int. J. Therm. Sci.
materials and metal foams: pareto optimization to address cost and operation time, 127 (2018) 19–28.
Appl. Therm. Eng. 197 (2021), 117436. [27] A. Mills, M. Farid, J. Selman, S. Al-Hallaj, Thermal conductivity enhancement of
[15] Y. Xu, M. Li, Z. Zheng, X. Xue, Melting performance enhancement of phase change phase change materials using a graphite matrix, Appl. Therm. Eng. 26 (2006)
material by a limited amount of metal foam: configurational optimization and 1652–1661.
economic assessment, Appl. Energy 212 (2018) 868–880. [28] A. Yu, P. Ramesh, M. Itkis, E. Bekyarova, R. Haddon, Graphite Nanoplatelet− Epoxy
[16] F. Aghbalou, F. Badia, J. Illa, Exergetic optimization of solar collector and thermal composite thermal Interface materials, J. Phys. Chem. C 111 (2007) 7565–7569.
energy storage system, Int. J. Heat Mass Transf. 49 (2006) 1255–1263. [29] P. Lamberg, K. Siren, Analytical model for melting in a semi-infinite PCM storage
[17] R. Raud, M. Cholette, S. Riahi, F. Bruno, W. Saman, G. Will, T. Steinberg, Design with an internal fin, Heat Mass Transf. 39 (2003) 167–176.
optimization method for tube and fin latent heat thermal energy storage systems, [30] A. Agrawal, A. Kumar, D. Rakshit, A novel analytical approach to study the
Energy 134 (2017) 585–594. charging characteristics of a shell and tube thermal energy storage system, Energy
[18] K. Venkateshwar, S. Tasnim, S. Mahmud, Nanoparticles to augment melting Storage 4 (2021) 129–146.
process in dynamicall adjusted thermal energy storage systems: an analytical [31] A. Ravindran, K. Ragsdell, G. Reklaitis, Powell's method of successive quadratic
investigation, J. Energy Storage 43 (2021), 103202. approximations, in: Engineering Optimization, John Wiley, New York, 1983.
[19] A. Hoe, M. Barako, A. Tamraparni, C. Zhang, A. Elway, J. Felts, P. Shamberger, [32] J. Lagarias, J. Reeds, M. Wright, P. Wright, Convergence properties of the Nelder-
Objective oriented phase change material composite heat sink design, Appl. Therm. Mead simplex method in low dimensions, SIAM J. Optim. 9 (1998) 112–147.
Eng. 209 (2022), 118235.

14