Applied Thermal Engineering 53 (2013) 291e298

Contents lists available at SciVerse ScienceDirect

Applied Thermal Engineering

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

Thermodynamic analysis of pumped thermal electricity storage

Alexander White a, *, Geoff Parks a, Christos N. Markides b
a
Cambridge University Engineering Department, Trumpington Street, Cambridge CB2 1PZ, United Kingdom
b
Department of Chemical Engineering and Chemical Processing, Imperial College, London SW7 2AZ, United Kingdom

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

Article history: The increasing use of renewable energy technologies for electricity generation, many of which have an
Received 12 December 2011 unpredictably intermittent nature, will inevitably lead to a greater need for electricity storage. Although
Accepted 19 March 2012 there are many existing and emerging storage technologies, most have limitations in terms of
Available online 24 March 2012
geographical constraints, high capital cost or low cycle life, and few are of sufficient scale (in terms of
both power and storage capacity) for integration at the transmission and distribution levels. This paper is
Keywords:
concerned with a relatively new concept which will be referred to here as Pumped Thermal Electricity
Electricity storage
Storage (PTES), and which may be able to make a significant contribution towards future storage needs.
Thermal energy storage
Irreversibility
During charge, PTES makes use of a high temperature ratio heat pump to convert electrical energy into
Heat transfer thermal energy which is stored as ‘sensible heat’ in two thermal reservoirs, one hot and one cold. When
required, the thermal energy is then converted back to electricity by effectively running the heat pump
backwards as a heat engine. The paper focuses on thermodynamic aspects of PTES, including energy and
power density, and the various sources of irreversibility and their impact on round-trip efficiency. It is
shown that, for given compression and expansion efficiencies, the cycle performance is controlled chiefly
by the ratio between the highest and lowest temperatures in each reservoir rather than by the cycle
pressure ratio. The sensitivity of round-trip efficiency to various loss parameters has been analysed and
indicates particular susceptibility to compression and expansion irreversibility.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction 1.1. Other storage technologies

In order to meet the UK’s target for the EU Renewable Energy The list of possible, alternative storage methods is extensive and
Directive, forecasts suggest that more than 30% of electricity may includes: flywheels, super capacitors, batteries and flow batteries,
need to come from renewable sources by 2020 [1]. Much of this is Compressed Air Energy Storage (CAES), Superconducting Magnetic
likely to be in the form of wind generation, which produces Energy Storage (SMES) and Thermal Energy Storage (TES) in its
supplies that are highly variable, uncontrollable and to some extent various forms. A review of many of these technologies is given by
unpredictable. Combatting such variability will be one of the major Chen et al. [3]. Some (e.g., flywheels and super capacitors) have
challenges for the high penetration of renewable technologies and very high efficiency, fast response and high power density, but are
is likely to require a combination of demand side management, only able to supply power for short durations. They are therefore
reserve capacity and, especially, increased energy storage. The scale most appropriate for power quality management applicationsee.g.,
of the problem is discussed by MacKay [2] who suggests that up to bridging short duration interruptions and providing voltage and
1200 GW h of storage may be required in the UK if a quarter of frequency support during rapid supply or demand swings. For
today’s electricity demand is to be met by wind energy. To put this energy management applicationsee.g., levelling daily demand
into perspective, the UK’s current inventory of storage facilities fluctuations and smoothing the output from intermittent renew-
(nearly all of which is pumped hydro) can store about 30 GW h, and able sourceseCAES is probably the leading competitor to Pumped
its scope for extension is severely limited by geographic and Hydro Storage (PHS), but it too suffers geographic limitations since
planning constraints. There is thus a pressing need to investigate large, robust caverns are required for the storage of air at pressures
new, smaller-footprint alternatives. up to 100 bar. Amongst the other candidate technologies, few are
currently able to provide both multi-megawatt scale capacity and
long-duration (i.e., hours) discharge; many are expensive (often
several times the cost of open-cycle standing reserve gas turbines
* Corresponding author. Tel.: þ44(0) 1223 765310; fax: þ 44(0) 1223 765311. in terms of £/kW installed capacity), and many make use of
E-mail address: [email protected] (A. White). hazardous, toxic or scarce materials.

1359-4311/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2012.03.030
292 A. White et al. / Applied Thermal Engineering 53 (2013) 291e298

devices are used, the flow directions of C and E can be reversed

Nomenclature during discharge such that they become an expander and
compressor respectively.) The heat engine and heat pump operate
c specific heat capacity, J kg1 K1 on the JouleeBrayton and reverse JouleeBrayton cycles respec-
p pressure, Pa tively, the temperature-entropy (Tes) diagram for which is shown
S sensitivity parameters defined by Eqs. (6)e(8). on the right of the figure for the case of ideal (reversible)
T temperature, K processes.
w specific work transfer Although PTES is in the early stages of development, it is based
a compression/expansion heat transfer factor on well-established technologies and it is therefore possible to
(defined in x2.3) estimate its likely performance (round-trip efficiency, storage
b pressure ratio density and capital cost) with a reasonable level of confidence.
c round-trip efficiency (work output/work input, Initial estimates suggest that it may well be able to compete with
during one storage cycle) CAES and PHS, but without the associated geographic limitations.
ε reservoir heat loss factor (defined in x2.3) The objectives of the present paper are to provide the thermody-
h compression/expansion efficiency namic background to estimates of PTES system performance,
q ¼ T3/T1 hot and cold reservoir temperature ratio when focusing in particular upon how the various sources of loss impinge
discharged upon round-trip efficiency and thereby pointing towards strategies
r density, kg m3 for optimisation. In contrast to work published thus far [4e6], the
rE energy storage density, J m3 or kWh m3 approach in the current paper makes use of simple models to
rP power density, W/(m3 s1) determine sensitivity of the cycle efficiency to various loss
s ¼ T2/T1 compressor temperature ratio during charge parameters (some of which are subject to uncertainty) and to
provide an understanding of the key dimensionless parameters
subscripts controlling performance.
0 ambient
1e4 cycle condition points defined by Fig. 1
C, D charged, discharged 2. Simplified cycle calculations
c, e compression, expansion
g, s gas, storage media The potential success of PTES hinges upon obtaining a satis-
factory round-trip efficiency (i.e., electrical energy output/elec-
Other symbols are defined in the text close to where they are used. trical energy input) whilst simultaneously keeping capital costs as
low as possible. Ultimately, successful design will require
comprehensive system modelling, taking into account a wide
variety of economic, thermodynamic, mechanical, electrical and
other factors. Such modelling is underway, but in anticipation of
1.2. Description of the PTES system its completion we present here instead a simplified model which,
nonetheless, provides a basic understanding of how the main
The subject of the current paper, PTES, is a recent concept which operating conditions and loss parameters influence performance.
is currently being developed in the UK (see for example ref. [4]) and The focus is upon thermodynamic aspects of PTES since electrical
is also being considered in France [5]. We include a brief description and mechanical issues are common to several other storage
of a PTES system here since it is a not widely known technology. methods.
The general principle of operation is straightforward and is illus-
trated in Fig. 1. The main components are a compressor (C), an
expander (E) and two thermal stores: one hot (HS) and one cold 2.1. Energy and power density
(CS). The latter are insulated cylindrical vessels containing an
appropriate thermal storage medium: e.g., a packed bed of pebbles For a given technology, the capital cost per unit energy storage
or gravel, or a uniform matrix of ceramic. The thermal stores are capacity (in £/kWh) and per unit power capacity (in £/kW) will
arranged vertically, as shown, so as to prevent buoyancy-driven depend inversely on the energy storage density, rE, and power
instabilities of the thermal fronts. The energy is stored as density, rP, respectively. These are thus key performance parame-
‘sensible heat’, enabling efficient integration with the thermody- ters for any storage method. We start by considering a fully
namic cycle and avoiding the ‘pinch-point’ difficulties associated reversible PTES system, as shown in the Tes diagram of Fig. 1. The
with phase-change storage methods. C and E may be either processes involved are the same as for the standard JouleeBrayton
turbomachinery-based or reciprocating devices. They are cycle (isentropic compression and expansion and isobaric heat
mechanically coupled and linked to a motor-generator. The heat addition and rejection), but traversed in the opposite direction
exchangers (HX1 and HX2) serve mainly to reject heat from the during charge. The round-trip efficiency for the reversible cycle is
cycle (via cooling water), this being required to combat the effects unity by definition, irrespective of the cycle pressures and
of system irreversibilities. temperatures, but it is nonetheless useful to consider this case as it
The working fluid is a gas (Argon) which flows in the direction provides reasonable estimates for rE and rP.
indicated by the arrows during charge, but is reversed for Defining rE as the average energy stored in the reservoirs per
discharge. During charge, the system operates as a high temper- unit volume of storage medium, and rP as the power output per
ature ratio heat pump, using electrical energy to extract heat from unit volumetric flow rate of the working fluid, straightforward
CS and deliver heat to HS. This takes place by the progression of analysis gives:
hot and cold fronts in the stores, as indicated in the layout figure.
During discharge, the processes are reversed so that the device
1 1
operates as a heat engine: heat is returned from the hot to the cold rE ¼ rs cs fðT2  T3 Þ  ðT1  T4 Þg ¼ rs cs T1 ðs  1Þð1  q=sÞ (1)
store and electrical energy is retrieved. (Note that if reciprocating 2 2
 A. White et al. / Applied Thermal Engineering 53 (2013) 291e298 293

Fig. 1. Layout of a PTES system and Tes diagram of the ideal cycle shown during charge.

rP ¼ rg1 cpg fðT2  T1 Þ  ðT3  T4 Þg 2.2. Approximate susceptibility to irreversibility

¼ ðs  1Þð1  q=sÞ  gp1 =ðg  1Þ (2) System irreversibilities tend to reduce expansion work outputs
where rs and rg1 are the storage medium and gas density respec- and increase compressor work inputs during both charge and
tively (at state 1 for the latter), cs and cpg are the corresponding discharge. An approximate estimate of how operating conditions
specific heat capacities (isobaric for the gas), s ¼ T2/T1 ¼ T3/T4 is the influence the round-trip efficiency for the real (irreversible) system
compressor and expander temperature ratio, and q ¼ T3/T1 is the can thus be obtained by scaling compression work by 1/h and
ratio between the hot and cold reservoir temperatures when dis- expansion work by h, where h may be interpreted here as an
charged. The following may be deduced from these expressions: average isentropic efficiency. The round-trip efficiency is then
given by:
i Both the energy and power density are monotonically
increasing functions of the temperature ratio s, which hðT2  T1 Þ  ðT3  T4 Þ=h Rh2  1
c ¼ ¼ (3)
depends on the pressure ratio and the isentropic index (g) of ðT2  T1 Þ=h  hðT3  T4 Þ R  h2
the gas. where R ¼ T1/T4 ¼ T2/T3 ¼ s/q, and T1 to T4 represent the tempera-
ii. For a given pressure ratio (or s), rE and rP are both increased tures for the reversible rather than the real cycle. Eq. (3) shows that,
by reducing T3 or increasing T1. This is also shown by the for a given value of h, the round-trip efficiency is a continuously
dashed line in the Tes diagram of Fig. 1 which encloses increasing function of the ratio R. This remains approximately true
a larger area. even for a more realistic treatment of system losses (see ref. [6]) and
iii. The power density may also be increased by raising the is merely a reflection of the fact that R is the ratio between the
overall system pressure. The factor g/(g  1) also has compression and expansion work (during charge); losses therefore
a significant influence and takes the value 5/2 for monatomic have a smaller impact on the net work as R is increased. The ratio R
and 7/2 for diatomic gases. may be increased by increasing the pressure ratio, decreasing the hot
reservoir discharged temperature, T3, or increasing the cold reser-
Table 1 shows estimates of the energy and power densities for voir discharged temperature, T1, all of which are consistent with
PTES compared with pumped hydro and compressed air storage. In improving the energy and power densities. High pressure ratios,
all cases the estimates are based on ideal or reversible systems. however, lead to high storage costs since HS needs to be pressurised.
Note that CAES is not a pure storage technology because it operates It is for this reason that Argon is proposed as the working fluid,
as part of a gas turbine cycle so the power density of a low-spec gas rather than air, since the same value of s can be achieved at a lower
turbine is given to provide an approximate comparison. Compari- pressure ratio due to Argon’s higher isentropic index.
sons between technologies should be treated with caution, but it is Fig. 2 shows the approximate round-trip efficiency based on Eq.
nonetheless reasonable to conclude that PTES has very good energy (3) and the temperature dependent factor of rE (i.e., (s1)(1 q/s)/q,
density and a power density that is not too much below that of which is equivalent to Eq. (1) if T3 is fixed, as discussed further in
a low-spec gas turbine. section 2.3.3). Each quantity is plotted versus s and the energy
density factor has been multiplied by c to give a better represen-
Table 1 tation of the recoverable energy. Two sets of curves are shown, one
Energy and power densities for a few storage technologies.
for a turbomachinery-based system, for which h2 ¼ 0.80, and one
PTESa PHSb CAESc GTd for a system based on reciprocating devices, for which the assumed
Storage medium: Gravel Water Air e value of h2 is 0.95. In each case, curves are plotted for q ¼ 1 (bold)
Working fluid: Argon Water Air air and q ¼ 0.4 (faint). (These two values of q bracket the likely practical
Energy density, rE (kWh/m3) 50 1.4 10 e
range and intermediate values provide curves that are intermedi-
Power density, rP (kW/m3s1) 240 5000 e 330
ately located  i.e., the dependence on q is monotonic.) The graphs
a
With T1 ¼ T3 ¼ 300 K, p1 ¼ 1 bar, s ¼ 2.58 (solid line in Fig. 1 and the approximate clearly show the benefits of reducing q, especially for the
conditions for the system in ref. [4]).
b
With altitude drop Dh ¼ 500 m; both rP and rE are given by rH2O gDh.
turbomachinery-based system. Such benefits would not be real-
c
With storage at 100 bar; see ref. [6] for the energy density expression for CAES. ised, however, if the maximum and minimum temperatures within
d
With s ¼ 2.05 (pressure ratio w12) and Tmax ¼ 900  C. the cycle were to be constrained, as discussed below.
294 A. White et al. / Applied Thermal Engineering 53 (2013) 291e298

Fig. 2. Approximate round-trip efficiency and energy density for q ¼ 1 (bold) and q ¼ 0.4 (faint).

Despite being based on the same thermodynamic cycle, it is (a) a polytropic efficiency to account for thermal dissipation
clear from the figures that trends are quite different to those of a gas and mixing, and (b) a heat loss (or gain) factor, a, defined as
turbine. For example, gas turbines have optimal pressure ratios for the ratio between net heat transfer and work transfer. Valve
both efficiency and power density, partly as a consequence of the losses are included in the pressure loss factor described
turbine inlet temperature being limited by material constraints. In below.
practice, there are likely to be limits on T2 and T4 in a PTES system iii Pressure losses. These occur through piping, valves (as
for similar reasons, and if these two temperatures are fixed, then it described above), diffusers (especially for turbomachinery)
may be shown from Eq. (1) that the maximum energy density and through the hot and cold reservoirs. In terms of impact on
occurs for s2 ¼ (T2/T4) and q ¼ 1, provided losses are small. However, the round-trip efficiency, it is the fractional pressure loss,
optimisation cannot be properly undertaken without considering f ¼ Dp/p, in each device that is most relevant since this is
a range of extra-thermodynamic factors, particularly those associ- proportional to the entropy increase and hence to the lost
ated with capital cost and materials issues. work. Individual values of f are expected to be small and so
can be summed to give an overall pressure loss factor F.
2.3. Detailed loss analysis iv Thermal reservoir losses. Aside from pressure losses, there are
two other loss mechanisms within the thermal stores. Firstly,
The aim of this section is to describe the various sources of loss heat exchange with the exterior means that the available
and provide preliminary estimates of their impact on round-trip energy stored within each reservoir is depleted with time.
efficiency. Several of the loss parameters (for example, the This is quantified by specifying an energy loss factor, ε, defined
compression and expansion efficiencies for reciprocating devices) below. Secondly, the process of heat transfer between the gas
are difficult to specify with great certainty at this stage. The and storage medium requires a finite temperature difference
approach adopted here is therefore to determine the sensitivity of and thus incurs a net entropy increase. The associated losses
the round-trip efficiency to these loss parameters. The chief sources are quite complex and depend on the mode of operation of
of loss are as follows: the storage system (e.g., regular periodic cycling or long-term
storage). Further discussion is given in section 2.3.4 below.
i Electrical and mechanical losses. These lie outside the scope of
the present paper which focuses instead upon thermody- Inclusion of the various losses modifies the Tes diagrams as
namic factors. However, it is worth noting that electrical/ shown in Fig. 3 (the condition points are as indicated in the layout
mechanical energy conversion can be achieved very effi- of Fig. 1). Entropy increases during compression and expansion and
ciently; he  97% is not unreasonable for large devices. various pressure losses mean that the charge and discharge cycles
Mechanical losses are more uncertain and are likely to be are no longer coincident and, as expected, the work output falls
greater for reciprocating devices than for turbomachinery. In below the work input during charge. Heat must therfore be rejected
total, electrical and mechanical losses might reduce the from the cycle. If the discharge pressure ratio is the same as that for
round-trip efficiency by 10% or more. charging (Fig. 3a) then the compressor delivery temperature, T30 ,
ii Compression and expansion losses. For turbomachinery these lies above T3 and so heat rejected via HX2 (see Fig. 1) such that HS
stem chiefly from viscous effects and are quantified here by can be restored to its initial, discharged state. Likewise, heat must
specifying a polytropic (infinitesimal stage) efficiency. In the also be rejected between states 10 and 1 via HX1 in order to return
case of reciprocating devices, compression and expansion CS to its initial state. Alternatively, as shown in Fig. 3b, a lower
occur relatively slowly such that there may be significant heat discharge pressure ratio can be used such that T30 ¼ T3. In this case
transfer with the cylinder walls. It is important to distinguish all the heat is rejected between states 10 and 1 in HX1. In fact, the
between net heat exchange with the surroundings and the optimal pressure ratio lies somewhere between these two cases,
periodic exchange to and from the walls which, although but the variation in efficiency is usually small unless compression
globally adiabatic, is an irreversible process and incurs an and expansion losses are very large. Note also that using a lower
entropy rise. Other losses result from pressure drops through discharge pressure ratio and then bypassing HX1 enables q ¼ T3/T1
valves and from mixing of fresh charge with residual gas. to be reduced between successive cycles in order to obtain the
Reciprocating devices are thus modelled here by specifying benefits described in sections 2.1 and 2.2.
 A. White et al. / Applied Thermal Engineering 53 (2013) 291e298 295

Fig. 3. Temperature-entropy diagrams for irreversible PTES cycles (note that the temperature scale is in  C).

2.3.1. General expression for round-trip efficiency vc 2ðg  1Þ=g

Ignoring reservoir losses for the time being, inspection of Fig. 3 SF ¼ ¼ (7)
vF ðs  1Þð1  q=sÞ
shows that the thermodynamic round-trip efficiency can be
expressed as:
vc 4lns
Sh ¼ ¼ (8)

0 


 vh ðs  1Þð1  q=sÞ
w0  w0c T  T10 1  a0e  T30  T40 1  a0c
c ¼ e ¼ 2 (4) where s is the temperature ratio under loss-free conditions, and for
wc  we ðT2  T1 Þ=ð1  ac Þ  ðT3  T4 Þ=ð1  ae Þ
Eq. (8) it has been assumed that all compressions and expansions
where the primes denote quantities during discharge. The 1/(1 e a) have the same polytropic efficiency, h. Sa, SF and Sh are effectively
factors reflect the occurrence of both heat and work transfer during sensitivity parameter, so Sa ¼ 2 implies the round-trip efficiency,
each of the compression and expansion processes (see ref. [7]). c, falls by two percent for each one percent heat leakage (as
Eq. (4) can be simplified by first noting that (in the absence of a fraction of work transfer) in the compressors and expanders.
heat losses from the reservoirs) T20 ¼ T2 and T40 ¼ T4. We will also Comparison with numerical cycle calculations (e.g. [6],) suggest
assume that T30 ¼ T3, corresponding to Fig. 3b above. Finally, we will a reasonable estimate of overall cycle loss can be obtained from
assume that ac ¼ ae ¼ a during charge, and ac0 ¼ ae0 ¼ ea during a linear combination of the loss contributions based on these
discharge. This reflects the fact that heat transfer will be to the sensitivity parameters, provided a, F and (1 e h) are each only a few
surroundings for hot components and from the surroundings for percent.
cold ones. With these simplifications, Eq. (4) becomes:
2.3.3. Impact of reservoir heat leakage on efficiency
  
1a sc =s0e  1 The effect of heat leakage to or from the reservoirs (as opposed
c ¼ 1 (5)
1þa ðsc  1Þ  qð1  1=se Þ to losses due to irreversible heat transfer) can be assessed by
considering the total availability, B, stored in the HS and CS. If each
The various temperature ratios, s, are related to the corre- reservoir has a thermal capacity of Mscs, then this is given by:
sponding pressure ratios, b, by expressions of the form s ¼ bn,
where n ¼ (g e 1)(1 e ae)he/g for expanders and n ¼ (ge1)(1 e ac)/
hcg for compressors (see ref. [7] for derivation). Note that a 10% heat B=Ms cs ¼ ðT2  T3 Þ  T0 lnðT2 =T3 Þ þ ðT4  T1 Þ  T0 lnðT4 =T1 Þ (9)
loss from a compressor combined with a 90% polytropic efficiency T0 is the temperature of the environment. (Note that for
gives rise to an isentropic process, but the effect is quite different to a reversible PTES system, T4/T1 ¼ T3/T2 so the logarithmic terms
that of adiabatic reversible compression. Pressure losses result in cancel and an expression similar to Eq. (1) is recovered.) The effect
expanders seeing a lower pressure ratio than compressors such of heat loss is to reduce T2 and to increase T4. Rates of heat loss will
that, for example, be ¼ (1 e F) bc. be proportional to the difference between the reservoir and envi-
ronment temperatures, hence we write dT2 ¼ eε(T2eT0) and
2.3.2. Impact of pressure, compression and expansion losses on dT4 ¼ eε(T4eT0). The factor ε, which, for simplicity, is assumed the
efficiency same for the HS and CS, will depend on the level of insulation and
A useful indication of how the various losses impact upon the storage duration. If, for example, the HS is designed to be at
round-trip efficiency is given by the partial derivatives of c with ambient temperature when discharged (as in the system proposed
respect to each of the loss parameters in turn. The algebra is in ref. [4]) then ε ¼ 0.01 per day corresponds to a 1% loss in the
considerably simplified by evaluating these derivatives at loss-free stored internal energy per day. Differentiating Eq. (9) with respect
conditions (i.e., h ¼ 1, F ¼ a ¼ 0). Straightforward but lengthy to T2 and T4 and substituting the expressions for dT2 and dT4 gives:
manipulations then give:
( )
dB ðT2  T0 Þ2 =T2 þ ðT4  T0 Þ2 =T4
vc ¼ ε (10)
Sa ¼ ¼ 2 (6) B ðT2  T3 Þ þ ðT4  T1 Þ
va
296 A. White et al. / Applied Thermal Engineering 53 (2013) 291e298

Note that this is equivalent to the reduction in c for an otherwise thereby causing further reduction in the stored availability. Strat-
reversible PTES system. egies to overcome the latter problem are, however, under
Either T1 or T3 (or both) will normally be close to ambient investigation.
temperature for the purpose of heat rejection. If, as is the case for Cyclic operation. The shape of the temperature profiles for cyclic
both the systems proposed in refs. [4,5], T3 z T0, then Eq. (10) can operation (Fig. 4b) depends on the length of the chargeedischarge
be rearranged to give: period relative to the time taken for an ideal (abrupt) thermal front
  to pass through the reservoir; in the case shown this ratio is about
vc sq s1 50%, meaning that the energy stored per cycle is only half of the
Sε ¼ ¼  þq (11)
vε s1 sq maximum possible. (The three curves in the figure correspond to
5%, 50% and 95% of the charge period.) Longer period cycles allow
This equals 2 if q ¼ 1 (i.e., T3 ¼ T1, as in ref. [4]), reflecting the fact
more energy to be stored but at the expense of steeper fronts and
that heat losses reduce both the stored thermal energy and the
thus higher losses. It is worth pointing out that steady state, peri-
efficiency with which it can be converted back into useful work.
odic operation necessarily incurs an exit loss, as suggested by curve
(iii) in Fig. 4b which shows the situation near the end of the charge
2.3.4. Impact of reservoir thermal irreversibility
phase and indicates the temperature at the exit of the reservoir
Krane [8] has analysed the destruction of availability within
beginning to rise. The exit loss reflects the need for heat to be
a thermal reservoir and concluded that the majority of the entering
rejected between successive cycles in order to counter the effects of
available energy would be lost. However, his analysis was for
irreversible heat transfer.
a fluid-based storage system for which much of the loss is due to
Since reservoir thermal losses clearly depend on the charge-
mixing. Heat transfer in packed beds (as proposed here) has
discharge history, accurate modelling can only really be under-
received extensive attention in the literature (see, for example, ref.
taken by developing an overall system model that couples unsteady
[9]), but little of this has focused on availability loss, which is crucial
heat transfer calculations with thermodynamic cycle calculations,
to an electricity storage scheme. Losses due to the irreversible
and includes the time-varying characteristics of the electrical
nature of the heat transfer between the gas and storage medium
network to which the storage system is connected. It is nonetheless
are not straightforward and their magnitude depends on a variety
possible to estimate how reservoir losses will vary with operating
of factors, including the mode of operation of the reservoirs. For
temperatures for a given reservoir geometry and mode of opera-
example, if the reservoir is taken from a fully discharged to a fully
tion. The net entropy generation rate due to heat transfer between
charged state (the so-called single blow case), temperature gradi-
gas and solid is given by:
ents tend to be steep and consequently losses are high. On the other
hand, periodic cycling of the reservoirs, as might be used for Z
2
Tg  Ts
smoothing daily electricity demand fluctuations, yields gentler S_ ¼ h dA (12)
Tg Ts
thermal gradients and lower losses, but at the expense of reduced
storage capacity. Fig. 4, taken from ref. [10], shows examples of where h is a surface heat transfer coefficient, Tg and Ts are the local
these two cases. gas and solid temperatures, and the integration is carried out over
Single charge. The solid lines in the left-hand figure show the the entire solid-gas interfacial area, A. The loss in availability is
thermal front (gas temperature) at three time intervals for the given by integrating this entropy generation rate over the charge-
single charge case. The temperature profile becomes progressively discharge periods and multiplying by the environment tempera-
less steep as it moves through the reservoir and consequently the ture, T0. It was shown in ref. [10] that the result can be expressed in
temperature difference between gas and solid (dashed lines) the form:
gradually decreases, thereby reducing the entropy generation rate.

The sloping front constitutes a loss of stored available energy and dB ¼ kT0  fnðTC ; TD Þ ¼ kT0 D2 = 1 þ D þ aD2 (13)
also prevents the reservoir from being fully charged without hot (or
cold) gas first issuing from the exit, thereby incurring an exit loss. where TC and TD are the reservoir charged and discharged
This means that the charged reservoir will have a non-uniform temperatures, D ¼ (TC e TD)/TD and k is a factor that depends on the
temperature distribution which will tend to equilibrate with time, geometry, flow conditions and mode of operation of the reservoir.

Fig. 4. Reservoir temperature profiles for different modes of operation. The solid curves are dimensionless gas temperatures defined as (Tg e TD)/(TC e TD). Distance is normalised by
l ¼ dp/{6 St (1 e ε)}, where St is the Stanton number, ε is the packed bed porosity and dp is the (equivalent) particle diameter. See [10] for details.
 A. White et al. / Applied Thermal Engineering 53 (2013) 291e298 297

Fig. 5. Variation of the loss sensitivity factors with isentropic temperature ratio for q ¼ 1 (bold) and q ¼ 0.4 (faint). Note that the x curves (right-hand figure) are with k ¼ 1/2.

The quadratic denominator of Eq. (13) stems from the inverse is improved by reducing the ratio q, in accord with the
dependence of the entropy generation rate on TsTg, as shown in Eq. approximate analysis based on work ratio given in section 2.2.
(12). The coefficient a varies from 0 for high frequency cycles This has been the strategy adopted for the turbomachinery-
(analytical solution) to 1/12 for single charge operation (numerical based design presented in ref. [5].
approximation, but with very small error). It is most likely that ii. Pressure losses seem to have a relatively small impact,
PTES will be used in the periodic cyclic mode and, in any case, provide the pressure ratio is not too low. (In addition, the
the effect of a is relatively small so it is set to zero in what follows. values of F as well as their impact will tend to be larger at
For simplicity, values of k will be assumed equal for the hot and lower pressure ratios, due to higher flow velocities for a given
cold reservoirs, although in practice the higher Reynolds mass flow rate.)
numbers would tend to give a larger k for the CS. Substituting the iii. Heat leakage to and from the compressors and expanders has
cycle temperatures and normalising by the stored availability a very different impact to that of the polytropic efficiency. A
gives the following expression for the overall (fractional) avail- compression heat loss during the charging phase, for
ability loss: example, will reduce the storage temperature (and hence
reduce the stored energy), but it will also reduce the work
dB ðT2  T3 Þ2 =T2 T3 þ ðT4  T1 Þ2 =T4 T1 input for the compression process. The net effect is a reduc-
x ¼ ¼ kT0
B ðT2  T3 Þ þ ðT4  T1 Þ tion in round-trip efficiency, but it is not as dramatic as the
  effect of a similar magnitude reduction in compression
T sq
¼ 2k 0 (14) efficiency.
T3 s  1
iv. The advantage of reducing q is not so clear once reservoir
This expression does not include the exit loss and the loss losses are also taken into account. Note that both the heat loss
associated with thermal equilibration during storage, but these sensitivity and thermodynamic loss factor (with k ¼ 1) take
losses are relatively small provided the reservoirs are operated on the constant value of 2 when q ¼ 1, so reducing q to 0.4
sensibly [10]. causes a significant increase in the thermodynamic compo-
nent of loss and only marginally improves the effect of heat
leakage at high pressure ratios.
2.4. Graphical representation of sensitivity factors

Fig. 5 shows the various loss sensitivity factors (Eqs. (6)e(8), 3. Conclusions
(11), (14)) plotted as functions of the isentropic temperature ratio
s ¼ T2/T1. Only one line is shown for the compressor/expander heat A new method of electricity storage (PTES) has been described
leakage in the left-hand figure since Sa is independent of q. For the and aspects of its thermodynamic performance investigated, with
reservoir losses it has been assumed that T3 ¼ T0, as discussed in particular focus on how various sources of loss affect the round-trip
section 2.3.3. Note also that the reservoir thermodynamic loss efficiency. The analysis presented has been very much simplified in
factor x is not strictly a sensitivity factor (i.e., it is not a partial order to show general trends which will help guide design. In
derivative of c), but the curves nonetheless show the relative particular, no account has been taken of how material properties
temperature dependence of these losses; an arbitrary value of (such as the thermal capacity of the storage medium) vary with
k ¼ 1/2 has been used for the results shown. temperature, and compressors and expanders have been modelled
The results presented in Fig. 5 demonstrate a number points by means of simple polytropic expressions.
which are of use in guiding research and design efforts. These may The round-trip efficiency and storage density both increase with
be summarised as follows: the compressor temperature ratio. High temperature ratios,
however, imply high pressure ratios which in turn imply high cost
i The round-trip efficiency is particularly susceptible to the for the hot reservoir. This is mitigated by the use of a monatomic
compression and expansion polytropic efficiency, especially gas such as Argon for the working fluid. Approximate analysis also
at low temperature (and hence pressure) ratios. The situation indicates that, for given compression and expansion efficiencies, it
298 A. White et al. / Applied Thermal Engineering 53 (2013) 291e298

is really the ratio between the highest and lowest temperatures in References
each of the reservoirs (R ¼ T2/T3 ¼ T1/T4) and not the compression
temperature ratio that determines the performance. Obtaining [1] The UK Renewable Energy Strategy [CM 7686] DECC Publications. (2009).
[2] D.J.C. MacKay, Fluctuations and storage, in: Sustainable Energy Without the
a satisfactory round-trip efficiency clearly requires highly efficient Hot Air, UIT Cambridge, 2009, pp. 186e202.
compression and expansion processes, and it is anticipated that [3] H. Chen, T.N. Cong, W. Yang, C. Tan, Y. Li, Y. Ding, Progress in electrical energy
this may be achieved by the use of reciprocating devices. For storage system: a critical review, Prog. Nat. Sci. 19 (2009) 291e312.
[4] J. MacNaghten and J. S. Howes, ‘Energy Storage’, Int. Patent No. WO 2009/
a turbomachinery-based PTES system, the effects of compression 044139 A2. (2009).
and expansion irreversibility can be mitigated by reducing the [5] T. Desrues, J. Ruer, P. Marty, J.F. Fourmigué, A thermal energy storage process
ratio between hot and cold store discharged temperatures, which for large scale electric applications, Appl. Therm. Eng. 30 (2010) 425e432.
[6] A. J. White, ‘Electricity storage using a thermal storage scheme’, 8th Interna-
also has the advantage of increasing the energy and power tional Conference of Computational Methods in Science and Engineering
densities. Ultimately, however, selecting the optimal operating (ICCMSE), Kos, Greece. (2010).
conditions will require reliable estimates of the various loss [7] A.J. White, Thermodynamic analysis of the reverse JouleeBrayton cycle heat
pump for domestic heating, Applied Energy 86 (11) (2009) 2443e2450.
parameters which depend on detailed design and, in some cases,
[8] R.J. Krane, A second law analysis of the optimum design and operation of
are currently subject to a degree of uncertainty. Areas in particular thermal energy storage systems, Int. J. Heat Mass Trans. 30 (1) (1987) 43e57.
need of further investigation include the reservoir thermal losses [9] A. Willmott, Dynamics of Regenerative Heat Transfer, Taylor & Francis, 2002.
[10] A.J. White, Loss analysis of thermal reservoirs for electricity storage schemes,
and the compression and expansion efficiency for reciprocating
Applied Energy 88 (2011) 4150e4159.
devices.