Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Solar Energy 206 (2020) 171–180

Contents lists available at ScienceDirect

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

Microeconomics of electrical energy storage in a fully renewable electricity T


system
David Timmonsa, , Khalil Elaheeb, Ming Lina

a
Department of Economics, University of Massachusetts Boston, USA
b
Mechanical and Production Engineering Department, The University of Mauritius, Reduit, Mauritius

ARTICLE INFO ABSTRACT

Keywords: The problem of climate change requires a transition to carbon–neutral energy. Given the variability of key
Decarbonization renewable energy sources such as solar photovoltaics (PV) and wind, fully renewable electricity systems require
Renewable energy economics some form of energy storage. But as described in this study, the economics of electrical energy storage are
Energy storage complex. Many reports on cost of electricity storage confuse the cost of charge/discharge power in kW with the
Pumped hydroelectric storage
cost of energy storage potential in kWh. These parameters vary greatly and create fundamentally different cost
Battery storage
structures for different storage technologies. A case study from the island-nation of Mauritius demonstrates that
simulation of a complete electricity system is needed to minimize cost of energy storage, and finds that reservoir-
type storage such as pumped hydroelectric (PHES) is less expensive than using batteries for the storage re-
quirements modeled. Large reductions in battery prices make them more competitive, but do not greatly reduce
total costs. PHES is a fully mature technology, and is more widely available than may be assumed. Policy makers
can thus be confident that commitments to deploy renewable energy do not depend on development of new
storage technology or reduction of battery prices.

This project was supported by a Fulbright Program grant sponsored by article about a utility shifting to natural gas states that “The utility
the Bureau of Educational and Cultural Affairs of the United States plans to get 7 percent of its power from solar by 2021, but says that
Department of State and administered by the Institute of International until storage technologies improve, gas will form the backbone of its
Education (IIE). energy mix…” (Plumer, 2019).
While better (or less expensive) batteries would be useful, and
battery technology is a critical component for decarbonization in sec-
1. Introduction
tors such as transportation, we show in this study that energy storage in
the electric utility sector may be dominated by reservoir storage options
Mitigating climate change requires providing most of society’s
such as pumped-hydroelectric storage (PHES). This is a mature tech-
electricity from carbon-neutral sources. Ambient sources such as solar
nology, used in electric grids since 1929 (Evans et al., 2012). Given its
and wind energy are renewable and carbon neutral in operation, but
cost structure (described in Section 2.3), PHES will likely remain
their availability varies over time. It is widely acknowledged that in-
competitive despite improvement of battery technology. And accom-
creasing energy storage is one approach to greater penetration of
modating variability of sources such as solar and wind does not depend
variable sources. In this paper we review microeconomic principles for
exclusively on energy storage; source complementarity, inclusion of
optimizing energy storage in a system with predominantly variable
dispatchable sources, demand-side management, and designed over-
sources, we report results from a case study in Mauritius incorporating
building also play critical roles in meeting demand while reducing the
these principles, and we discuss policy implications for renewable en-
energy storage requirement (Perez et al., 2019; Solomon et al., 2019).
ergy development.
There is a vast literature on electricity storage technologies, but only
A widely repeated misconception is that energy storage technology
a small portion of these studies focuses on the specific question of
is not sufficiently developed to support fully renewable electricity sys-
storage in a fully renewable electricity grid, which requires different
tems, and that broad deployment of wind and solar power depends on
energy storage services than current electric grids dominated by fossil
future battery development. For example, a recent New York Times
fuels. Reported energy storage prices sometimes vary widely, and it can


Corresponding author.
E-mail address: david.timmons@umb.edu (D. Timmons).

https://doi.org/10.1016/j.solener.2020.05.057
Received 8 September 2019; Received in revised form 29 April 2020; Accepted 19 May 2020
Available online 08 June 2020
0038-092X/ © 2020 International Solar Energy Society. Published by Elsevier Ltd. All rights reserved.
D. Timmons, et al. Solar Energy 206 (2020) 171–180

Nomenclature OPEX Operating expenditure


PHES Pumped hydroelectric storage
Abbreviations PV Photovoltaic

AC Alternating current Variables


CAPEX Capital expenditure
CAES Compressed air energy storage α Capital recovery factor
CF Capacity factor Cf Cost, fixed
DC Direct current Cv Cost, variable
H2 Hydrogen ht Hours in a time slice
kW Kilowatt (power measure) K Capital cost
kWh Kilowatt hour (energy measure) KE Capital cost, energy storage
LCOE Levelized cost of energy KP Capital cost, power conversion
LCOES Levelized cost of electricity system kWhD Discharged kWh per year
LCOS Levelized cost of storage kWhS System annual electricity demand
Li Lithium MWht MWh generated in a time slice
MW Megawatt (power measure) r Discount rate
MWh Megawatt hour (energy measure) T Useful life of a capital good

be difficult to make meaningful comparisons between prices reported Consider a simplified example of a renewable electricity system
from different sources (Jülch, 2016). We suggest three main reasons for with all energy obtained from solar photovoltaics (PV), an identical
these discrepancies. First, some storage technologies and prices have solar insolation pattern on each day, a PV capacity factor1 (CF) of 0.25,
changed quickly; in such cases, only the most recent studies are re- a constant 1 MW of demand, and the use of energy storage. Daily en-
levant. Second, prices for technologies such as PHES vary with scale ergy consumption is 24 MWh.2
and local conditions. But a third and more fundamental problem is that As illustrated in Fig. 1, to supply energy needed, the solar power
studies report storage costs using different metrics, including capital rating must be 4 MW.3 Solar intensity is sufficient to meet the 1 MW
cost based on charge–discharge power (e.g. $/kW capacity), capital cost demand directly for perhaps 8 h per day, so demand will be met from
based on energy storage potential ($/kWh), and levelized cost of sto- storage for approximately 16 h. The maximum discharge rate needed is
rage (LCOS; $/kWh per unit of energy actually stored). And each of 1 MW (to meet demand at night), but the more important constraint is
these metrics can be defined differently, leading to different conclusions the storage charging requirement, which is 3 MW.4 In this case the
about costs. In Section 2 we propose specific definitions for storage required storage is approximately 16 MWh energy and 3 MW power, or
power and energy costs, which are consistent with needs for optimizing 5.3 h of storage.
storage in a fully renewable electricity system. This is a simplified example that includes only one energy source
In Section 3, a case-study model of a fully renewable electricity (PV), constant weather conditions, constant demand, and no energy
system on the island-nation of Mauritius demonstrates that at current curtailment. In actual renewable electricity systems, none of these
prices, the cost-minimizing solution relies exclusively on PHES for the conditions hold, and computer simulation is the only practical way to
energy storage requirements modeled, and that this energy storage determine the power and energy requirements for storage.
accounts for 7.0% of total renewable electricity production cost. We test Furthermore, the objective is not simply to meet demand, but rather to
reductions of lithium (Li) battery prices, and find that batteries enter meet demand at the minimum cost, given various energy production
the optimal solution with large price reductions, but that this does not and storage options.
greatly reduce cost compared to the base case using PHES.
Several policy implications arise from this study, as we discuss in 1.2. Review of relevant literature
Section 4. Most importantly, policy makers should be confident that
energy storage technology is already available to support a transition to There are many studies of energy storage technologies and costs,
variable renewable energy sources. cataloging energy storage options including pumped hydroelectric
Given the problem of climate change, there is an urgent need to storage (PHES), compressed air energy storage (CAES), flywheels, ca-
replace legacy fossil-fuel systems with renewable energy sources. Fully pacitors, conventional batteries, flow batteries, fuel storage with fuel
renewable electricity systems require some form of energy storage. But cells, and thermal storage (Beaudin et al., 2010; Dunn et al., 2011;
the economics of electricity storage are often misunderstood. This study Evans et al., 2012; Schoenung and Hassenzahl, 2003). Many studies
applies economic principles to the problem of grid electricity storage to also discuss storage applications such as maintaining power quality,
assist in planning cost-effective energy storage for the renewable energy shifting loads in time, and bulk power management, with technologies
transition. appraised for suitability in such applications (Díaz-González et al.,
2012; EPRI, 2010; Sundararagavan and Baker, 2012). While con-
1.1. The energy storage problem ceptually useful, a limitation of the application-driven approach to
technology selection is that the applications are not mutually exclusive,
In engineering terms, power is a flow measured in watts, while i.e. a storage technology needed for bulk power management may also
energy is an accumulation of flows over time, measured in watt hours. be used for load shifting and to improve power quality. For a fully
This distinction is crucial for energy storage economics, because po-
tential storage devices must be described in terms of both power and 1
Capacity factor is the portion of maximum energy production actually ob-
energy parameters. Lazard (2018), for example, describes a commer- tained over period of time, which must be less than or equal to 1.
cial/industrial-scale battery as having a 1 MW power rating (can be 2
1 MW x 24 h = 24 MWh.
charged or discharged at a rate of 1 MW), and an energy rating of 2 3
4 MW x 24 h x 0.25 CF = 24 MWh.
MWh. The ratio of MWh to MW equals hours of storage (in this ex- 4
(4 MW maximum PV power) – (1 MW demand) = 3 MW, assuming no
ample, 2), which varies considerably depending on the application. curtailment.

172
D. Timmons, et al. Solar Energy 206 (2020) 171–180

Fig. 1. Schematic supply, demand, and storage for fully solar electricity generation.

renewable electricity system, the question is which storage technology energy is available, this may substitute for both energy production and
or combination of technologies should be used to minimize the total storage. One cannot determine how much energy storage is optimal
cost of the system. without knowing the availability and cost of biomass (and similar al-
From a cost-minimization perspective, energy storage must be op- ternatives).
timized for both power and energy: as in the example in Section 1.1, we At the minimum total cost, the equimarginal principle applies. In its
need to know both MW and MWh requirements. Costs for power and simplest from, this says that the marginal costs of all energy production
energy must thus be reported separately. Several studies adhere to this and storage technologies should be equal at the minimum total cost.
practice (Jülch, 2016; Lazard, 2018; Schoenung and Hassenzahl, 2003; However, the application of the equimarginal principle is more complex
Sundararagavan and Baker, 2012; Zakeri and Syri, 2015), providing the in a system with variable sources (Timmons, 2018). Given source
necessary information for economic optimization. variability, marginal costs for energy alternatives are equal only under
A few studies look specifically at the economics of energy storage in critical design conditions, which are the times when it is most difficult
electricity systems that are fully renewable, or nearly so. Solomon et al. to supply energy. Optimal system design is determined by difficult
(2017) advocate a system-level approach, noting that storage require- conditions—times without much wind or sun—rather than by average
ments depend on factors including portion of variable generation conditions.
sources, resource complementarity in time, possibilities for generation Many renewable energy sources are capital intensive and can thus
curtailment, and criteria selected for grid reliability and reserve re- be treated as fixed costs. Most of the expense is for initial construction,
quirements. In reviewing studies of energy storage in grids dominated with no fuel and little operating expense. The same may be true for
by variable sources, the authors find that for up to 90% use of variable energy storage. Costs may be similar regardless of how much a specific
sources, the optimum stored electricity potential is perhaps one day of energy generating or storage asset is used.
electricity consumption (Solomon et al., 2017). Most storage technologies exhibit economies of scale, i.e. cost per
The approach of Child et al. (2018) in modeling the European unit falls as scale increases. This may suggest using a small number of
electricity grid is similar to ours: an optimization model is used to large electricity storage facilities, though costs of transmission and
choose among different electricity generation and storage technologies needs for local grid stabilization and resilience must also be considered.
to arrive at a cost-minimizing portfolio for which supply always meets This combination of microeconomic factors drives particular con-
or exceed demand. Like our study, Child et al. show that a fully re- figurations for renewable energy and storage systems, for example, it
newable electric grid is feasible with a combination of dispatchable explains why energy storage skews toward large reservoir sources with
sources, variable sources, and nonbiomass energy storage equal to ap- relatively low cost for energy storage (as compared to battery storage).
proximately 1.5 days of average consumption. However, results for the First, there are some circumstances in which using stored energy from
composition of storage technologies differ markedly from ours due to large reservoirs has the lowest marginal cost. Since reservoir-type sto-
different cost assumptions, as we discuss in Section 3. rage has near-zero operating costs, if storage is constructed for these
difficult circumstances, it costs virtually nothing to use at other times,
and it is easily substituted for other energy and storage sources.
2. Methods and data
Reservoir-type sources may thus be the least costly option for energy
storage (also depending on environmental and other costs).
2.1. Microeconomics of energy storage

Several common microeconomic concepts apply to the problem of 2.2. Cost definitions
minimizing cost in an electricity system with variable renewable
sources. Since energy storage devices are available at vastly different scales,
Kilowatt hours of electricity are identical, so electricity storage a way is needed to normalize devices for cost comparison. For example,
devices are potential substitutes for each other. For example, multi-day EPRI (2010) provides system costs in $/kWh (or CAPEX per kWh):
storage can also provide storage needed at minute and hour time scales.
K
An electric grid operator should make cost-reducing energy pro- (1)
kWh
duction or storage substitutions whenever possible, but such substitu-
tions must be considered at a system level. For example, if biomass where K is total capital cost, and kWh is the energy potentially stored in

173
D. Timmons, et al. Solar Energy 206 (2020) 171–180

the device. Similarly, EPRI (2010) also provides storage costs in $/kW where r is a discount rate and T is the expected life of the energy storage
(or CAPEX per kW): device in years.
LCOS is useful for comparing storage technologies, but only if these
K
(2) technologies are used in identical applications. LCOS suffers from the
kW
same conflation of power and energy costs described above, and the
where kW is the power charge–discharge rating of a device. A problem denominator kWhD creates an additional complexity. Annual kWh dis-
with these metrics is that they conflate the costs of energy storage with charged depends on storage cycling, which may vary greatly between
charge–discharge costs: equations 1 and 2 both include all capital costs, applications, and optimal kWhD cannot be determined without a system
and there is no fixed ratio of kWh/kW (no standard number of hours) model. Reported LCOS statistics often include the cost of energy to
for an energy storage device. charge the storage as a variable cost, which also varies greatly. For
Energy and power attributes and costs may be usefully dis- these reasons, LCOS itself is not an appropriate statistic for optimizing
aggregated. For example in a battery system, the power conversion energy storage (though some LCOS reports, e.g. Lazard (2018), include
system determines the power rating, and the storage module (the bat- KE and KP statistics which are useful optimization inputs).
tery) determines the energy storage potential. Using a given power LCOS is similar to the more-common metric of levelized cost of
conversion system, one could buy a larger or smaller battery, depending energy (LCOE). In this study, our objective is to minimize LCOES, or
on the quantity of energy storage required (within limits). Similarly, for energy cost for the electricity system as a whole. This can be defined as:
a given PHES turbine-generator, a smaller or larger energy storage re- I
servoir could be constructed. i=1
((KEi + KPi ) i + Cfi + Cvi )
LCOES =
Power and energy costs must be considered separately to optimize kWhS (7)
an energy system. We thus use the energy-cost measure:
where i is an index of energy production or storage technologies, kWhS
KE is annual system kWh consumed, and the other terms are as defined
kWh (3) above.
where KE includes only capital cost associated with an energy storage
module (as well as proportional balance of plant costs), and similarly, 2.3. Cost of energy storage technologies
the power-cost measure:
The ultimate cost of energy storage is the sum of power and energy
KP costs plus operating cost, divided by the quantity of energy stored (Eq.
kW (4) (5)). But the more relevant question is how energy storage can be used
where KP includes only the capital cost for the power conversion to minimize the total cost of energy in a system (Eq. (7)). To determine
system. Distinguishing capital costs in this way allows the costs to be this, we must consider power, energy, and operating costs separately, as
considered separately for optimization, where both power and energy well as storage efficiency, which we describe here for characteristic
expenditures are optimized. storage types. Results in Section 3 show each technology’s role in the
Levelized cost of storage (LCOS) is another commonly used metric context of minimizing the cost of a renewable electricity system in
(Jülch, 2016). For example, the widely-cited energy storage cost report Mauritius.
from Lazard (2018) focuses on LCOS. In a system where annual energy
use is constant (or nearly so), LCOS can be defined as: 2.3.1. Battery-type storage
Power costs for battery storage relate to the power conversion
(KE + KP ) + Cf Cv
LCOS = + system (Sundararagavan and Baker, 2012), which consists of an in-
kWhD kWh (5) verter to transform DC battery power into AC grid power, and the re-
where KE and KP are capital costs for energy and power respectively, Cf verse. These are electronic components that have fallen greatly in price
is fixed cost, kWhD is energy discharged over the course of a year, and over recent years. Of the energy storage alternatives, we see in Table 1
Cv is variable operating cost. A capital recovery factor α is used to that battery-type storage has the lowest power costs for charging and
annualize the capital costs: discharging, i.e. it is relatively inexpensive to move energy in and out of
batteries. A battery’s electronic power system can also respond very
r (1 + r )T quickly to generation surpluses or deficits in the electric grid, so a
=
(1 + r )T 1 (6) battery is useful for applications that require quick response but little

Table 1
Representative costs for different types of electricity storage.
Capital Cost, Power (KP) Capital Cost, Energy (KE) Fixed operating cost, power Fixed operating cost, energy Life, years Efficiency (%)
USD/kW USD/kWh USD/kW USD/kWh (T)

Battery Type Storage


Lia 105 316.00 2.30 5.17 20 88
Leadb 206 370.00 9.61 12.95 10 72
Reservoir Type Storage
PHESc 2,743 11.62 57.14 na 40 81
CAESd 866 16.67 6.00 na 30 70
Fuel Type Storage
e
H2, geological 1,263 0.49 27.00 na 15 43
H2, steel tanke 1,263 19.00 27.00 na 15 43

Table notes:
a
Lazard (2018), 10 MW (EPC omitted).
b
Lazard (2018), 1 MW (EPC omitted).
c
Mauritius (Timmons et al. 2019).
d
Calculated from EPRI (2010).
e
Steward et al. (2009).

174
D. Timmons, et al. Solar Energy 206 (2020) 171–180

total energy. PHES also requires water, but unlike a hydropower facility, the
While battery-type power costs are low, energy costs are high same water may be used repeatedly; water is moved between elevations
(Table 1), i.e. batteries are an expensive way to hold energy for an rather than consumed, and replacement water is needed only because of
extended period (though costs have been declining quickly). Lazard evaporation and leakage. Lin (2019) models PHES on the relatively dry
(2018) provides cost estimates for lithium (Li) batteries, lead and ad- island of Rodrigues, and includes the cost of a small seawater desali-
vanced lead batteries, vanadium flow batteries, and zinc-bromide flow nation plant to provide replacement water for the PHES, increasing
batteries, all technologies which are sufficiently commercialized that annual storage cost by 6.4% over the same system without desalination.
cost data are available. Costs are similar to those for the Li and lead In other settings, a small stream or groundwater supply might provide a
batteries shown in Table 1 (though Lazard does not provide dis- similar quantity of replacement water. Some projects have considered
aggregated KE and KP costs for all battery technologies). Note that en- using wastewater from municipal treatment plants for PHES (Yang and
ergy costs are currently on the order of 20 times greater than for re- Jackson, 2011).
servoir storage types. PHES has typically been developed at large scales of
Storage efficiency is somewhat higher than reservoir storage, and 1000–4000 MW (Evans et al., 2012), but can be found at mini, micro,
can exceed 90%, i.e. less than 10% of energy may be lost in the charge- and pico-scales, including plants that supply only a few households
store-discharge cycle. (Rehman et al., 2015). Smaller-scale plants will likely have greater costs
than large-scale plants. PHES has been used to stabilize small isolated
2.3.2. Reservoir-type storage electric grids such as island grids. El Hierro in the Canary Islands of
Reservoir-type storage represents a fundamentally different ap- Spain has a wind farm coupled with a PHES of 6 MW charge capacity
proach from chemical storage of energy in batteries. For example with and 11.3 MW discharge capacity, which supplies most of the island’s
PHES, energy is stored as water in a high-elevation reservoir. When electricity (Godina et al., 2015).
there is energy demand, water is released to a lower-elevation reservoir, A region with hydroelectric resources might have sufficient energy
giving up potential energy in conventional hydroelectric turbines that storage behind hydroelectric dams without needing PHES (Child et al.,
generate electricity. When surplus electricity is available, energy is used 2018). In other areas, there may be potential to convert existing hy-
to pump water from the lower reservoir back to the upper. droelectric facilities to PHES (Adams, 2018). Rather just supplying
The cost structure of reservoir-type storage also differs from bat- electricity on demand, hydroelectric turbines converted to PHES could
teries, as can be seen in in Table 1. PHES power conversion is accom- also pump water back into reservoirs, storing energy for later use.
plished with a mechanical pump/turbine (typically a single reversible Compressed air energy storage (CAES) is another promising re-
unit; Yang and Jackson, 2011) coupled with a motor/generator rather servoir-type technology. As shown in Table 1, CAES power and energy
than with electronic components. These mechanical components are costs may both be less than PHES, making it a strictly dominant tech-
responsible for much greater power costs (KP) for reservoir storage than nology (if both costs were lower, this technology would always be
for batteries, on the order of 20 times greater. used). But CAES technology is not extensively deployed, with only two
On the other hand, reservoirs at elevation can hold large quantities utility-scale examples in the world (Zakeri and Syri, 2015), so cost es-
of energy at KE costs only a fraction of those for batteries, as seen in timates are unreliable. And both of the existing examples use natural
Table 1. The potential energy in stored water depends both on the gas to reheat compressed air as it expands, an approach that is not
volume of water and on the head (vertical difference) between re- strictly renewable. Adiabatic CAES does not use natural gas, but costs
servoirs; a reservoir contains more potential energy with greater head. are even less certain. For the current model, we thus include only PHES.
If reservoir capital cost is constant, KE increases as head falls. Lin (2019)
shows that KE increases sharply as head falls below 200 m, though as
shown in Table 1, KE could be increased considerably and still be less 2.3.3. Fuel-type storage
than for batteries. A final relevant storage strategy is using electricity to produce
Many authors remark that potential PHES sites are not widely dis- chemical fuels such as hydrogen, methane, or methanol (Burkhardt and
tributed (Beaudin et al., 2010; Esteban et al., 2012; Schoenung and Busch, 2013; Gahleitner, 2013). Chemical fuels may be stored in-
Hassenzahl, 2003), while several authors disagree (Díaz-González et al., definitely at low cost, yielding a low KE value.
2012; Rehman et al., 2015; Yang and Jackson, 2011). Yang and Jackson Hydrogen storage is perhaps the best known fuel-storage technology
(2011) remark that PHES use has been limited more by environmental (Steward et al., 2009). Hydrogen in water can be separated from oxygen
issues and uncertainty about market conditions than by availability of through electrolysis, stored in tanks or geological formations, and re-
suitable sites. combined with oxygen to release energy in a fuel cell or combustion
Unlike battery storage, PHES costs vary by site. Available head is engine. With underground storage in geologic formations, the high
one factor, and new reservoirs may be needed in upper, lower, or both energy density of hydrogen or methane results in very low storage cost
locations. Construction costs vary depending on conditions such per unit of energy (Table 1). But the combination of electrolysis and
geology, permeability of reservoir soils, land cover, land cost, and fuel cells currently has low energy conversion efficiency, with an ex-
permitting cost. pected loss of approximately 60% of energy in power conversion for
PHES costs in Table 1 costs reflect assumptions for our Mauritius storage (Table 1).
case study. We assume 335 m head, and given the existence of large
high-elevation storage reservoirs there, assume construction of only Table 2
lower reservoirs at the cost of a recent reservoir construction project in Capital cost comparisons for PHES power and energy.
the country (Havumaki, 2018). We use a power cost from a consultant Source Capital cost, power Capital cost, energy
report on potential hydropower cost in that country (Kantor (KP) USD/kW (KE) USD/kWh
Management Consultants, 2008; adjusted for inflation). In Table 2 we
This study 2,743 11.62
compare the resulting power and energy costs to other published fig- Sundararagavan and Baker 2,000 12.00
ures (some of which report ranges derived from other studies). (2012)
The main condition needed for PHES is a place for two ponds with a Díaz-González et al. (2012) 1,100 16.50
vertical difference of perhaps 200 m over a short horizontal span. This Rehman, Al-Hadhrami, and Alam 800 10.50
(2015)
condition is widely found around the world. And with sufficient
Zakeri and Syri (2015) 581 68.00
transmission capacity, storage facilities may be located far from elec- Jülch (2016) 833 27.75
tricity consumers.

175
D. Timmons, et al. Solar Energy 206 (2020) 171–180

Greater conversion efficiencies and/or lower conversion costs in the


future could make hydrogen or other forms of fuel storage viable for
renewable electricity systems. Renewable energy systems based on
variable supply will have times when energy production is curtailed due
to excess solar and/or wind energy (Perez et al., 2019). During these
times, the cost of electricity is effectively zero, which could make hy-
drogen attractive in spite of low conversion efficiencies. Power costs
may be partly mitigated by using a larger electrolyzer (inexpensive) to
charge storage and a smaller fuel cell (expensive) to discharge stored
energy. Combined thermal and electricity production from stored hy-
drogen (not included in our model) may also have potential to increase
efficiency and reduce costs in regions with significant thermal loads.

2.4. Storage economics in an energy system Fig. 3. Hypothetical annual storage exceedance.

Storage cost in an energy system depends on how the storage is


used. In Fig. 2 we illustrate how levelized cost of storage (LCOS) for area of 2040 km2 and approximately 1.3 million inhabitants. The island
is fully electrified, with annual consumption of 3132 GWh in 2018, 21%
PHES and Li batteries changes with the storage requirement in hours,
based on the cost parameters shown in Table 1 and an assumed op- of which was from renewable sources (Statistics Mauritius, 2019). Most
electricity is generated from coal and oil, and fossil fuels are also used
eration regime (but note that LCOS varies greatly with the number of
storage cycles). For durations of less than 6 h, Li batteries have lower in transportation, manufacturing, agriculture, and household cooking.
Mauritius is useful as a case-study site, being a small-scale re-
cost, but this cost rises quickly as more hours are required. By contrast,
PHES cost is nearly constant for any number of hours in the scenario presentation of the more complex problem of a global transition to
renewable energy. Mauritius has no domestic fossil fuels, but is well
illustrated, and has costs less than Li batteries for storage durations of
7 h or greater. situated for solar, wind, biomass, some hydropower energy, and ver-
tical relief suitable for PHES. From a renewable energy perspective,
Consider also a hypothetical annual storage exceedance function,
where there are a small number of hours with large power charge or anything (except geothermal) is possible, so the question is what
combination of renewable energy sources and energy storage minimizes
discharge requirements, and more hours with lower power needs, as
shown in Fig. 3. Since the vertical axis is power in MW and the hor- cost while meeting demand at all times.
Biomass in Mauritius comes chiefly from sugarcane bagasse, sugar
izontal is hours, areas in the figure can be interpreted as energy in
MWh. The energy for area A, with high power requirements for few being the principle agricultural crop. Bagasse already accounts for
about 14% of electricity production and the majority of renewable
hours, is least expensively provided by batteries, and similarly, the
energy for area C with low power for many hours is best supplied by energy production (Statistics Mauritius, 2019). The sugar industry also
produces bioethanol fuel on a small scale, which could be expanded.
reservoir storage. The important question from Fig. 3 is whether area B
should be supplied by battery or reservoir storage. Using batteries re- Approximately 23% of the island is forested, which could be sustainably
managed for additional biomass production. As with biomass produc-
quires increasing power from (p2-p1) to p2, while using reservoir sto-
rage would increase hours from (h2-h1) to h2. Which option is less tion anywhere, sustainable production is not assured, and must be
managed.
expensive depends on the shape of the storage exceedance function and
the cost parameters for battery and reservoir technologies. From Fig. 3, Hydropower currently accounts for about 4% of electricity genera-
tion (Statistics Mauritius, 2019). As at the world scale, hydropower is
it is clearly possible for one technology to provide almost all of the
energy storage services, for example, if reservoir storage is used to renewable, inexpensive, and dispatchable, but is limited in supply.
Year-round irrigation and drinking water is provided to the island by
supply both areas C and B.
large reservoirs at greater than 500 m elevation. We assume that these
may have potential for dual use as PHES reservoirs, though we have not
2.5. Mauritius optimization model
specifically evaluated reservoirs for this purpose. Engineering, agri-
cultural, social, human health, energy, and land-use criteria are all re-
2.5.1. Mauritius description
levant to the question of water reservoir use.
Mauritius is an island nation in the south Indian Ocean, with a land

Fig. 2. Levelized Cost of Storage* for Lithium Batteries (Li) and Pumped Hydroelectric Storage (PHES).

176
D. Timmons, et al. Solar Energy 206 (2020) 171–180

2.5.2. OSeMOSYS model a site slated for development of a new wind farm with Suzlon 2.1 MW
For this study we use an OSeMOSYS model (Open-Source Energy turbines. Data observed at 60 m are adjusted for the Suzlon 90 m hub
Modeling System; Howells et al., 2011) to represent the Mauritius height. Time-slice capacity factors range from 0.0 to 1.0, with a mean
electricity system in the year 2040 and determine the optimum energy value of 0.29.
production and storage components for fully renewable generation. In We model hydroelectric water as an aggregate flow into a single
this section we present a brief description of the model and input data; a reservoir based on observed data from a representative gauging station
full model description is available in Timmons et al. (2019). on the Central Plateau (station J04; Ministry of Energy and Public
The model includes time slices for each day and night of the year, or Utilities, 2011). We assume that a 12% increase in net hydroelectricity
730 time slices (365 × 2). Time slices for days represent the hours is feasible (Elahee, 2013), with all new capacity in pico or micro-hydro
8:00–16:00, when insolation is potentially sufficient to meet demand. plants.
For each time slice where electricity generation exceeds demand, For pumped hydro storage, we assume use of existing upper re-
electricity can be stored, subject to available charging capacity and the servoirs on the Central Plateau and construction of new lower re-
available quantity of electricity storage. Similarly, electricity can be servoirs. Reservoirs may be charged by both river flow and storage
removed from storage if needed to meet demand in a subsequent time pumps.
slice. The time structure of the model precludes making conclusions We include a waste-to-electricity plant of 30 MW and force this
about balancing of supply and demand within time slices, though in- plant into the model solution, since waste disposal is the plant’s primary
cluded storage can be used within time slices (e.g. within a day). purpose.
The OSeMOSYS linear programming solver identifies the cost- Mauritius currently has several biomass electricity plants collocated
minimizing quantities of both generating and storage resources for with sugar processing facilities. During the six-month sugar harvest
which electricity supply equals or exceeds demand in each time period, season (approximately July-December), these plants burn primarily
subject to constraints on land, fuels, etc. We annualize all capital costs sugarcane bagasse (crushed cane residue), and currently burn coal
using a capital recovery factor (Eq. (6)) and an 8% discount rate, and during the six-month non-harvest season. To replace coal, the model
include both fixed and variable operating costs. Cost assumptions for includes off-season biomass sources including domestic biomass crops,
generating sources are shown in Table 3. domestic wood chips, and imported wood chips. Biomass assumptions
For variable sources such as solar and wind, generating performance and costs are provided in Timmons et al. (2019). The model also in-
enters the model as capacity factors (CF) for each time slice: cludes bioethanol produced from domestic sugar as a potential fuel for
electricity generation in engines or turbines.
MWht Daily electricity demand data for the 2015 data year are provided
CF =
ht (8) by the Central Electricity Board (CEB). To reflect possible demand in
the 2040 model year, we use demand growth assumptions from the
where MWht is electricity production estimated for 1 MW of capacity in
Mauritius Renewable Energy Agency (MARENA).
a particular time slice, and ht is hours in the time slice. Solar and wind
We use a base scenario to identify the optimum mixture of PHES, Li
capacity factors are calculated from observed ambient conditions in
batteries, and H2 storage at current costs, and to observe storage charge
each time slice.
and discharge patterns over the model year. We then decrease Li bat-
Given data availability, we combine data from different sources and
tery KE costs in 10% increments to determine the extent to which Li
years. For example, solar data are from 2017 to 2018, wind data are
battery technology may be preferred if prices fall in the future.
from 2010 to 2011, river-flow data are from 2009 to 2010, and demand
data are from 2015. We construct a composite year; solar, wind, hydro,
and demand conditions are all observed on the same calendar days, 3. Results and discussion
albeit in different years.
Solar data is from stations at Goodlands (north), La Mivoie (west), Overall results for fully renewable electricity generation in
and Souillac (south) (Solarmap, 2017). Daily capacity factors range Mauritius are shown in Table 4, given base cost assumptions. The
from 0.03 to 0.29, with mean capacity factors of 0.173, 0.169, 0.147, minimum-cost scenario includes more capacity for solar PV capacity
for the north, south, and west respectively. than for any other source, though solar wind generate almost identical
Wind data are measured at Plaine Sophie (Suzlon-Padgreen, 2012), quantities of energy. Biomass from various domestic and imported

Table 3
Cost parameters for electricity generation.
Capital cost, USD Fixed operating cost, USD/kW Mean annual capacity factor

g
Photovoltaic panels, ground-mount utility scale, per kW 1238 10.50 0.173
Wind turbines, per kW 2066 35.00 0.289
Hydroelectric and pumped storage plants, per kW 2743a 57.14a nah
Pumped storage reservoir, per potential kWh 12b na na
Lithium battery AC charge/discharge power, per kW 107c 2.30 nah
Lithium battery energy storage, per potential kWh 376c 5.17 na
Solid fuel plants, per kW 2850 50.00 nah
Bioethanol power plants, per kW 650d 10.00 nah
Municipal solid waste plant, per kW 9250e 50.00f 0.950

Costs from Lazard (2017) except as noted.


a. Kantor Management Consultants (2008), adjusted for inflation.
b. Havumaki (2018).
c. Lazard (2018).
d. Assumed same cost as diesel internal-combustion generator.
e. Carnegie Clean Energy (2017).
f. Assumed same as solid fuel plant.
g. Based on best site monitored at Goodlands (Solarmap, 2017).
h. Capacity factors for dispatchable sources determined within model.

177
D. Timmons, et al. Solar Energy 206 (2020) 171–180

Table 4 also fails to enter the optimum solution. Hydrogen stored in geologic
Results: Mauritius renewable electricity generating costs. formations has a much lower KE cost (Table 1), and our base-model
MW capacity GWh generation % GWh solution for Mauritius would include 41 MW of electrolysis power
(storage charge), 17 MW of fuel-cell power (storage discharge), and 3.6
Wind turbines 475 1112 27 GWh of storage potential, reducing total cost by 0.25% compared to the
Solar PV, utility scale 785 1120 27
base case with only PHES. Given the small change in total cost and the
Hydropower and PHES 114 130 3
Lithium batteries 0 na na
lack of data about suitability of Mauritian geology for hydrogen sto-
Hydrogen storage 0 na na rage, we exclude this result. But it is worth noting that hydrogen is on
Municipal solid waste plant 30 250 6 the cusp of economic viability, given low enough energy storage cost
Solid fuel plant 280 (KE), despite relatively high power cost (KP) and low conversion effi-
fuel: sugarcane bagasse 453 11
ciency.
fuel: cane tops and leaves 147 4
fuel: domestic wood chips 120 3 Results are summarized in Fig. 5. The horizontal axis labels show
fuel: biomass crop 75 2 both the Li price-reduction input (KE) and the corresponding result in
fuel: imported wood chips 626 15 system levelized cost of energy (LCOES). With a 50% reduction in Li
Liquid fuel plant 334
battery price, LCOES falls to only $0.129 from $0.130/kWh, less than a
fuel: bioethanol 61 2
Total 2018 4094 100
1% reduction in total cost. At 90% Li energy price reduction, LCOES
falls to $0.111/kWh, a 15% reduction in cost.
As indicated in Fig. 5, the composition of the generating portfolio

Fig. 4. Storage charge and discharge for Mauritius model year.

sources provides the greatest quantity of energy. LCOES (system LCOE) changes significantly with the price of battery energy storage. With less
is estimated at $0.130/kWh, similar to the current cost of generating expensive batteries, the proportion of solar PV capacity grows mark-
electricity in Mauritius. See Timmons et al. (2019) for a complete dis- edly,5 while wind energy production declines and biomass generation
cussion of these results. capacity falls somewhat. Lower battery prices allow for substitution of
Total storage is equal to 0.5 days of average demand. As seen from more-variable PV for less-variable wind and biomass, without in-
the striped pattern in Fig. 4, storage is typically charged during the day creasing total cost.
and discharged at night. Solid black areas of Fig. 4 indicate consecutive In their study of energy storage optimization for a fully renewable
time slices with continuous charging (positive numbers) or discharging European grid, Child et al. (2018) reach different conclusions for sto-
(negative numbers). The longest continuous period of storage discharge rage composition of a 2050 scenario, with fuel in the form of synthetic
occurs at the end of July, with storage discharged for 3 consecutive time methane providing 82% of the nonbiomass energy storage, batteries
slices (2 nights and 1 day), due to the combination of below-average providing 16% of storage, and PHES providing just 2%. These results
solar insolation and wind during those days. can be traced to assumptions about cost reductions by 2050, with
With the occasional need for multi-day energy storage, PHES is the synthetic methane technology assumed to decrease 60% in price by
only energy storage technology to appear in the base-model cost- 2050, and Li batteries assumed to decrease 88%. While we agree that
minimizing solution, given its low energy (KE) cost in relation to battery future technology improvements and price reductions for energy sto-
storage. While PHES costs are sensitive to local conditions, our results rage are likely, the extent and timing of improvements is uncertain, and
arise more from the price structure of reservoir vs. battery storage predicting storage price paths is problematic for making current policy
(differences in power and energy costs) than from the particular price of decisions.
PHES that we assume for the Mauritius case study. We also test our
power cost (highest in Table 2) with the median reservoir cost reported
4. Conclusions and policy implications
by Zakeri and Syri (2015; highest in Table 2) and find that reservoir
storage is still used exclusively.
As suggested by the results of the scenarios for decreasing battery
While Li batteries have low power costs (KP), their current energy
prices, the optimum use of any particular technology depends on its
costs preclude their use when energy must be stored for more than a
own price and the prices of all other technologies in an energy system.
few hours. However, the model time-slice structure likely results in an
underestimate of the need for Li batteries to match supply and demand
within the day and night time slices (Timmons et al. 2019). 5
We have assumed sufficient land availability and constant marginal cost for
Fuel-type storage as represented by hydrogen stored in steel tanks all sources, which could affect these results.

178
D. Timmons, et al. Solar Energy 206 (2020) 171–180

Fig. 5. Mauritius optimum generation and storage portfolios at different price reductions for Li batteries.

For example, the optimum amount of PV energy depends on the price of creation. While PHES power equipment may be imported, the civil
PV, the price of Li battery storage, the price of wind energy, etc. This works associated with reservoir construction would typically create
implies that transitions to fully renewable electric grids should be stu- jobs in the areas where they are built.
died at a holistic system level. This study provides economic analysis of portfolios of electricity
Prices for storage batteries and fuel conversion technologies are sources and electricity storage that minimize the cost of supplying re-
widely expected to fall in the future, including prices for Li batteries, newable electricity, which is needed to mitigate the problem of climate
hydrogen fuel cells, and a host of competing technologies. But the level change. A fully renewable electric grid is shown to be feasible with
of price reductions and the timing of these are impossible to predict. We current electricity storage technology in the form of PHES. Policy ma-
make no prognostications about how likely a 50% reduction in Li bat- kers face many challenges around creating appropriate incentives for
tery costs might be, but a transition to a fully renewable electricity developing renewable energy and required storage infrastructure.
system does not depend on energy storage price reductions, at least While there remain many hurdles to deploying renewable energy sys-
where PHES has costs similar to those in Mauritius. In the case study, tems, a transition to renewable electricity does not require developing
even a 50% reduction in Li battery prices had virtually no impact on the new energy storage technologies.
total cost of renewable energy. Policy makers should be confident that a
transition to fully renewable electricity systems is feasible with current Declaration of Competing Interest
technology at current prices.
Pumped hydropower storage has been widely viewed as a tech- The authors declare that there is no conflict of interest.
nology with limited expansion potential, though this conclusion may
have been based more on the characteristics of legacy fossil-fuel sys- References
tems than on emerging renewable systems. A few studies report new
interest in developing PHES facilities, as well as some new approaches Adams, T.B., 2018. Feasibility of retrofitting existing hydropower infrastructure for use in
to this historic technology (Rehman et al., 2015; Yang and Jackson, renewable energy storage, Department of Civil and Environmental Engineering.
2011; Zakeri and Syri, 2015). It has been suggested that “…lack of Massachusetts Institute of Technology, Cambridge, MA, USA.
Beaudin, M., Zareipour, H., Schellenberglabe, A., Rosehart, W., 2010. Energy storage for
public awareness of the benefits of bulk electricity storage is a con- mitigating the variability of renewable electricity sources: an updated review. Energy
siderable barrier for PHES development” (Yang and Jackson, 2011, p. Sustain. Dev. 14 (4), 302–314.
842). The same authors observe that closed-loop and off-stream PHES Burkhardt, M., Busch, G., 2013. Methanation of hydrogen and carbon dioxide. Appl.
Energy 111, 74–79.
designs do not impact existing aquatic ecosystems, thus reducing en- Carnegie Clean Energy, 2017. High penetration renewable energy roadmap for the
vironmental problems and increasing probability of public acceptance Republic of Mauritius. Mauritius Research Council.
for PHES. The temporal patterns of variable wind and solar generation Child, M., Bogdanov, D., Breyer, C., 2018. The role of storage technologies for the tran-
sition to a 100% renewable energy system in Europe. Energy Procedia 155, 44–60.
as well as the cost structure of PHES suggest it as a strong candidate for Díaz-González, F., Sumper, A., Gomis-Bellmunt, O., Villafáfila-Robles, R., 2012. A review
fully renewable electricity systems. Policy makers can encourage re- of energy storage technologies for wind power applications. Renew. Sustain. Energy
search on potential PHES sites, ensure a stable market for electricity Rev. 16 (4), 2154–2171.
Dunn, B., Kamath, H., Tarascon, J.-M., 2011. Electrical energy storage for the grid: a
storage services, develop clear environmental standards for PHES, and
battery of choices. Science 334 (6058), 928–935.
provide public education on the emerging need for energy storage. Elahee, K., 2013. Potential of hydropower in mauritius: myth or reality? Energy Sources
Though not focuses in our study, other considerations are environ- Part A 35 (10), 921–925.
mental and local labor-market effects of battery production. EPRI, 2010. Electricity Energy Storage Technology Options. Electric Power Research
Institute, Palo Alto, CA, USA.
Environmental issues with Li batteries are widely reported (Peters et al., Esteban, M., Zhang, Q., Utama, A., 2012. Estimation of the energy storage requirement of
2017), and batteries have relatively short lives with recycling or other a future 100% renewable energy system in Japan. Energy Policy 47, 22–31.
disposal costs at end of life. These issues are particularly problematic Evans, A., Strezov, V., Evans, T.J., 2012. Assessment of utility energy storage options for
increased renewable energy penetration. Renew. Sustain. Energy Rev. 16 (6),
(and potentially costly) in an island setting such as Mauritius. In many 4141–4147.
regions, relying on battery storage also implies buying technology from Gahleitner, G., 2013. Hydrogen from renewable electricity: an international review of
outside of a region, potentially affecting trade balance and local job power-to-gas pilot plants for stationary applications. Int. J. Hydrogen Energy 38(5),

179
D. Timmons, et al. Solar Energy 206 (2020) 171–180

2039–2061. program. Sandia National Laboratories.


Godina, R., Rodrigues, E., Matias, J., Catalão, J., 2015. Sustainable Energy System of El Solarmap, 2017. Solar map of Mauritius, Rodrigues, and Agalega. http://solarmap.uom.
Hierro Island, International Conference on Renewable Energies and Power Quality ac.mu/. (Accessed 7 October, 2018).
(ICREPQ’15). Renewable Energy and Power Quality Journal (RE&PQJ). pp. 46–51. Solomon, A., Bogdanov, D., Breyer, C., 2019. Curtailment-storage-penetration nexus in
Havumaki, B.M., 2018. Hydropower in the Decarbonized Mauritian Grid: A Prospective the energy transition. Appl. Energy 235, 1351–1368.
Study, Economics. University of Massachusetts Boston. Solomon, A., Child, M., Caldera, U., Breyer, C., 2017. How much energy storage is needed
Howells, M., Rogner, H., Strachan, N., Heaps, C., Huntington, H., Kypreos, S., Hughes, A., to incorporate very large intermittent renewables? Energy Procedia 135, 283–293.
Silveira, S., DeCarolis, J., Bazillian, M., 2011. OSeMOSYS: the open source energy Statistics Mauritius, 2019. Energy and Water Statistics - Year 2018. Ministry of Finance
modeling system: an introduction to its ethos, structure and development. Energy and Economic Development, Port Louis, Mauritius.
Policy 39 (10), 5850–5870. Steward, D., Saur, G., Penev, M., Ramsden, T., 2009. Lifecycle cost analysis of hydrogen
Jülch, V., 2016. Comparison of electricity storage options using levelized cost of storage versus other technologies for electrical energy storage. National Renewable Energy
(LCOS) method. Appl. Energy 183, 1594–1606. Lab.(NREL), Golden, CO (United States).
Kantor Management Consultants, 2008. Energy Policy for the Republic of Mauritius: Draft Sundararagavan, S., Baker, E., 2012. Evaluating energy storage technologies for wind
Final Energy Policy: TASK C: Electricity Sector. Ministry of Energy and Public power integration. Sol. Energy 86 (9), 2707–2717.
Utilities, Ebene, Mauritius. Suzlon-Padgreen, 2012. Environmental Impact Assessment (EIA) Reports: Proposed
Lazard, 2017. Lazard's levelized cost of energy analysis–version 11.0. construction and operation of a wind Farm comprising 14 wind turbines with a total
Lazard, 2018. Lazard's levelized cost of storage analysis–version 4.0. installed capacity of 29.4 mw by Consortium Suzlon-Padgreen Co Ltd, Annex 6.
Lin, M., 2019. A 100% Renewable Electricity System Model on the Island of Rodrigues, Ministry of Social Security, National Solidarity, and Environment and Sustainable
International Conference on Energy, Environment and Climate Change. Mauritius. Development (Environment and Sustainable Development Division).
Ministry of Energy and Public Utilities, 2011. Hydrology Data Book 2006-2010. Timmons, D., 2018. Optimal Renewable Energy Systems: Minimizing the Cost of
Perez, M., Perez, R., Rabago, K.R., Putnam, M., 2019. Overbuilding & curtailment: the Intermittent Sources and Energy Storage. In: Letcher, T.M., Fthenakis, V.M. (Eds.), A
cost-effective enablers of firm PV generation. Sol. Energy 180, 412–422. Comprehensive Guide to Solar Energy Systems. Elsevier, pp. 485–504.
Peters, J.F., Baumann, M., Zimmermann, B., Braun, J., Weil, M., 2017. The environmental Timmons, D., Dhunny, A.Z., Elahee, K., Havumaki, B., Howells, M., Khoodaruth, A.,
impact of Li-Ion batteries and the role of key parameters – a review. Renew. Sustain. Lema-Driscoll, A.K., Lollchund, M.R., Ramgolam, Y.K., Rughooputh, S., Surroop, D.,
Energy Rev. 67, 491–506. 2019. Cost minimization for fully renewable electricity systems: a mauritius case
Plumer, B., 2019. As Coal Fades in the U.S., Natural Gas Becomes the Climate study. Energy policy 133.
Battleground, New York Times. Yang, C.-J., Jackson, R.B., 2011. Opportunities and barriers to pumped-hydro energy
Rehman, S., Al-Hadhrami, L.M., Alam, M.M., 2015. Pumped hydro energy storage system: storage in the United States. Renew. Sustain. Energy Rev. 15 (1), 839–844.
a technological review. Renew. Sustain. Energy Rev. 44, 586–598. Zakeri, B., Syri, S., 2015. Electrical energy storage systems: a comparative life cycle cost
Schoenung, S.M., Hassenzahl, W.V., 2003. Long-vs. short-term energy storage technolo- analysis. Renew. Sustain. Energy Rev. 42, 569–596.
gies analysis: a life-cycle cost study: a study for the DOE energy storage systems

180

You might also like