The
Influence
of
Regional
Geophysical
Resource
Variability
on
the
Value
of
Single-
and
Multistorage
Technology
Portfolios
Anna
X. Li,
*
,
⊥
Edgar
Virgu
̈
ez,
*
,
⊥
Jacqueline
A. Dowling,
⊥
Alicia
Wongel,
Dominic
Covelli,
Tyler
H. Ruggles,
Natasha
Reich,
Nathan
S. Lewis,
*
and
Ken
Caldeira
*
Cite
This:
https://doi.org/10.1021/acs.est.3c10188
Read
Online
ACCESS
Metrics
& More
Article
Recommendations
*
sı
Supporting
Information
ABSTRACT:
A stylized
macro-scale
energy
model
of least-cost
electricity
systems
relying
only
on wind
and
solar
generation
was
used
to assess
the value
of different
storage
technologies,
individually
and
combined,
for the contiguous
U.S.
as well
as for four
geographically
diverse
U.S.
load-balancing
regions.
For
the contiguous
U.S.
system,
at current
costs,
when
only
one
storage
technology
was
deployed,
hydrogen
energy
storage
produced
the lowest
system
costs,
due
to its energy-capacity
costs
being
the
lowest
of all storage
technologies
modeled.
Additional
hypothetical
storage
technologies
were
more
cost-
competitive
than
hydrogen
(long-duration
storage)
only
at very
low
energy-
capacity
costs,
but
they
were
more
cost-competitive
than
Li-ion
batteries
(short-duration
storage)
at relatively
high
energy-
and
power-capacity
costs.
In
all load-balancing
regions
investigated,
the
least-cost
systems
that
included
long-duration
storage
had
sufficient
energy
and
power
capacity
to also
meet
short-duration
energy
and
power
storage
needs,
so that
the addition
of short-duration
storage
as a second
storage
technology
did not
markedly
reduce
total
system
costs.
Thus,
in electricity
systems
that
rely
on wind
and
solar
generation,
contingent
on social
and
geographic
constraints,
long-duration
storage
may
cost-effectively
provide
the services
that
would
otherwise
be provided
by shorter-
duration
storage
technologies.
KEYWORDS:
Least-cost
electricity
systems,
energy
storage
technologies,
wind
generation,
solar
generation,
decarbonized
electricity
systems
■
INTRODUCTION
Energy
storage
is an important
component
of reliable,
cost-
effective,
deeply
decarbonized
electricity
systems
that
rely
on
substantial
generation
from
variable
renewable
energy
resources,
such
as wind
and
solar
energy.
1
Energy
storage
technologies
differ
in their
siting
and
supply
chain
constraints,
sociopolitical
challenges,
round-trip
efficiency,
energy-capacity
cost,
power-capacity
cost,
and
storage
duration.
2,3
Conse-
quently,
many
modeled
least-cost,
deeply
decarbonized
electricity
systems
contain
multiple
storage
technologies.
3,4
Short-duration
energy
storage
technologies
have
relatively
low
power-capacity
costs
and
thus
are
cost-effective
for
frequent
(hourly)
charging
and
discharging
to smooth
sharp
peaks
in electricity
generation
or demand.
5,6
Currently,
lithium-ion
(Li-ion)
batteries
with
1 to 4 h durations
are the
most
widely
deployed
short-duration
storage
technology.
7,8
In contrast,
long-duration
(>100
h) storage
technologies
such
as pumped-storage
hydropower
(PSH),
compressed
air
energy
storage,
and
electrolytic
hydrogen
have
relatively
high
power-capacity
costs
and
relatively
low
energy-capacity
costs,
as compared
to other
commercialized
storage
technologies
on
the market.
9
Herein,
energy-capacity
costs
refer
to overnight
capital
costs
for energy
storage
in $/kW
h, and
power-capacity
costs
refer
to overnight
capital
costs
for power
capacity
in
$/kW,
for a given
storage
technology.
Due
to these
low energy-
capacity
costs,
long-duration
energy
storage
can
compensate
for sustained
weather-related
events
that
last days
or weeks
and
can cost-effectively
buffer
seasonal
or interannual
variability
in
renewable
resource
availability,
even
if depleted
relatively
infrequently
in any
year.
10
−
13
Another
group
of demonstrated
storage
technologies
can
potentially
provide
mid-duration
storage,
i.e.,
storage
for
durations
of days
to weeks.
This
group
is characterized
by
intermediate
energy-
and
power-capacity
costs
(Figure
1, Table
S1).
For example,
deployed
redox-flow
batteries
have
durations
up to 10 h and
can theoretically
be designed
to provide
storage
for
even
longer
durations.
14
Thermal
energy
storage
can
reportedly
provide
storage
durations
from
8 to 192
h (8 days),
and
commercial
iron
−
air
batteries
are
projected
to provide
durations
from
100
to 150
h at a combined
energy-
and
power-
Received:
December
4, 2023
Revised:
June
19, 2024
Accepted:
June
21,
2024
Article
pubs.acs.org/est
© XXXX
The Authors.
Published
by
American
Chemical
Society
A
https://doi.org/10.1021/acs.est.3c10188
Environ.
Sci. Technol.
XXXX,
XXX, XXX
−
XXX
This article is licensed under CC-BY 4.0
Downloaded via CALIFORNIA INST OF TECHNOLOGY on July 16, 2024 at 19:50:48 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
capacity
cost
of <$20/kW
h.
15
−
17
Gravity-based
energy
storage
has the potential
to store
energy
for >12
h.
18
When
long-duration
storage
is used
in addition
to short-
duration
storage,
total
system
costs
are reduced
for wind-
and
solar-based
electricity
systems
that
meet
hourly
averaged
demand
in full
for over
a year
of resource
variability.
10
−
13
However,
the value
and
role
of deploying
two
or more
storage
technologies
are controversial.
For
a United
Kingdom
(U.K.)
electricity
system
modeled
with
mainly
wind
and
solar
generation
along
with
existing
nuclear
resources,
in con-
junction
with
demand
flexibility,
almost
all optimal
storage
portfolios
in least-cost
reliable
systems
used
only
Li-ion
batteries
and
electrolytic
hydrogen,
with
compressed
air energy
storage
deployed
only
in scenarios
with
electricity
oversupply
within
a specific
range.
19
However,
when
projected
2050
costs
were
assumed
for seven
independent
United
States
(U.S.)
electricity
system
load-balancing
regions
with
100%
renewable
or carbon-free
resources
(wind,
solar,
nuclear,
hydro,
biomass,
and
geothermal),
the
optimal
storage
portfolio
contained
4
types
of storage
technologies
with
mutually
different
durations.
20
When
2050
costs
were
assumed
for
storage
technologies
in three
different
U.S.
load-balancing
regions
that
rely
primarily
on wind
and
solar
generation,
with
constrained
amounts
of natural
gas generation,
only
low-cost
Li-ion
and
redox-flow
batteries
were
used
for storage,
obviating
a need
for
longer-duration
storage
technologies
including
electrolytic
hydrogen,
thermal
energy
storage,
or metal
−
air
batteries.
20
In scenarios
in which
Li-ion
batteries,
redox-flow
batteries,
and
a single
long-duration
storage
technology
(thermal,
metal
−
air,
or hydrogen)
were
available,
the
optimal
storage
portfolio
partially
substituted
deployment
of Li-ion
batteries
with
redox-
flow
batteries
and
the
long-duration
storage
technology.
21
Here,
we aim
to identify
generalizable
findings
for least-cost
energy
storage
portfolios,
based
on the
fundamental
geo-
physical
variability
of the resources
available
in different
load-
balancing
regions
over
the time
scales
required
to meet
hourly
averaged
demand
in full over
a year.
The
value
of a storage
technology
was
measured
by the
technology's
impact
on total
costs
of a least-cost
electricity
system
based
solely
on wind
and/or
solar
generation
that
met
hourly
averaged
demand
in full for an entire
year.
To assess
the
value
of different
storage
technology
portfolios
in a simple,
transparent
fashion,
we used
a stylized
macro-scale
energy
model
10,22
−
24
to obtain
asset
capacities
and
dispatch
schedules
in a least-cost
stylized
electricity
system
that
relies
only
on
wind
and
solar
generation,
assuming
no previously
existing
grid
technologies.
The
stylized
electricity
system
relied
solely
on
wind
and/or
solar
generation,
with
no transmission
con-
straints,
no reserve
margins,
and
no firm
dispatchable
fossil
generation
such
as natural
gas,
to transparently
reveal
the
fundamental
geophysical
dynamical
relationships
between
energy
storage
and
wind
and/or
solar
resource
variability
over
a variety
of geographically
distributed
regions
in the U.S.
Hourly
averaged
resource
availability
data
and
concurrent
hourly
demand
data
were
obtained
for one
year
from
a weather
reanalysis
data
set for the contiguous
U.S.
(CONUS)
as well
as
for four
independent
system
operator
(ISO)
regions
within
CONUS
that
were
characterized
by very
different
qualities
and
quantities
of wind
and
solar
resources
(Table
S2).
The
modeling
was
subject
to the
strict
constraint
that
100%
of
hourly
averaged
demand
was
met
for
every
hour
in the
simulated
year.
Each
region
was
represented
by a single
node,
which
reduces
generation
variability
and
thus
decreases
the
value
of storage
technologies
compared
to more
realistic
representations
of the grid.
The
modeled
energy
storage
technologies
were
divided
qualitatively
into
three
categories:
short-,
mid-,
and
long-
duration
storage.
Li-ion
batteries
were
used
to represent
a
short-duration
storage
technology,
whereas
electrolytic
hydro-
gen
represented
a long-duration
storage
technology.
The
electrolyzers
used
electricity
to produce
hydrogen,
which
was
stored
in underground
salt caverns
and
subsequently
utilized
in
fuel
cells.
Various
technologies
represented
potential
mid-
duration
storage
systems:
redox-flow
batteries
(RFB),
com-
pressed
air energy
storage
(CAES),
pumped-storage
hydro-
power
(PSH),
thermal
energy
storage,
gravity
energy
storage,
and
metal
−
air
battery
storage.
In the
modeled
systems,
the
energy
and
power
capacities
of these
mid-duration
storage
technologies
were
independently
sizable,
potentially
allowing
them
to be optimized
to also
provide
short-
or long-term
storage.
Figure
S1 shows
the electricity
sources,
storage,
and
sinks
(electricity
demand
or curtailed
power)
in the
model
architecture.
The
modeled
electricity
systems
contained
portfolios
of 1
−
3
storage
technologies
that
comprised
various
combinations
of the
defined
short-,
mid-,
and
long-duration
storage
technologies.
The
robustness
and
generality
of the
findings
were
evaluated
by parameterizing
the
energy-
and
power-capacity
costs
of a hypothetical
storage
technology
(
Storage
X
) across
wide
ranges
for these
geographically
diverse
U.S.
load-balancing
regions.
■
METHODS
Wind
and
Solar
Generation
Data.
The
regions
considered
in this
analysis
were
the
contiguous
U.S.
(CONUS)
and
four
subnational
independent
system
operator
(ISO)
geographic
regions
(CAISO,
ERCOT,
ISO-NE,
and
MISO).
Hourly
capacity
factors
for solar
and
wind
data
for
Figure
1.
Energy-capacity
costs
and
power-capacity
costs
of energy
storage
technologies.
Ranges
of total
installed
energy-
and
power-
capacity
costs
of different
storage
technologies.
Numerical
values
and
sources
are
provided
in Table
S1.
*
Energy-capacity
and
power-
capacity
costs
were
combined
to obtain
the total
cost
of Li-ion
battery
and
metal
−
air
battery
storage.
Environmental
Science
& Technology
pubs.acs.org/est
Article
https://doi.org/10.1021/acs.est.3c10188
Environ.
Sci. Technol.
XXXX,
XXX, XXX
−
XXX
B
each
region
during
2018
were
generated
using
reanalysis
data
with
a grid-cell
resolution
of 0.5
°
latitude
by 0.625
°
longitude
from
the Modern-Era
Retrospective
analysis
for Research
and
Applications,
Version
2 (MERRA-2).
25
Solar
capacities
of
utility-scale
photovoltaics
were
calculated
for
a single-axis
tracking
system
with
0
−
45
°
of tilt.
Wind
capacity
factors
for
geographic
regions
with
the top
25%
generation
potential
of
land-based
wind
turbines
were
calculated
assuming
a General
Electric
1.6
−
100
turbine
with
a 1.6
MW
nameplate
capacity.
26
−
28
Table
S2 presents
calculated
average
wind
and
solar
capacity
factors
for CONUS,
CAISO,
ERCOT,
ISO-NE,
and
MISO.
Electricity
Demand
Data.
Electricity
demand
data
for the
CONUS
and
ISO
regions
were
obtained
from
hourly
data
for
2018
from
the
U.S.
Energy
Information
Administration
(EIA).
29
The
EIA
data
was
cleaned,
and
missing
values
were
replaced
using
the multiple
imputation
by chained
equations
(MICE)
method.
30
Cost
and
Technological
Assumptions.
A complete
description
of the
model
formulation
is included
in the
Supporting
Information.
Base
case
costs
for solar
and
wind
generation
were
taken
from
the
National
Renewable
Energy
Laboratory
Annual
Technology
Baseline
(NREL
ATB)report
(Table
S3).
31
Tables
S3 and
S4 present
the base
case
costs,
efficiencies,
and
other
characteristics
for storage
technologies
used
in the model.
Parameters
for Li-ion
batteries,
hydrogen
storage,
RFB,
CAES,
PSH,
and
thermal
energy
storage
were
taken
from
a 2021
NREL
analysis
of long-duration
energy
storage
technologies.
10
Gravity
energy
storage
parameters
were
taken
from
the Pacific
Northwest
National
Laboratory’s
2020
Grid
energy
Storage
Technology
Cost
and
Performance
Assessment,
with
energy-
and
power-capacity
costs
separated
by linear
regression,
using
cost
estimates
for 1000
MW
storage
systems
at various
durations.
14
The
total
overnight
cost
for
metal
−
air
batteries
was
taken
from
press
releases
by Form
Energy,
and
the
O&M
costs
and
round-trip
efficiency
were
taken
from
the
2022
MIT
Future
of Energy
Storage
Report.
17,32
Li-ion
batteries
and
metal
−
air
batteries
were
each
modeled
using
one
total
cost
because
the energy
and
power
components
of these
batteries
are
nonseparable.
Li-ion
batteries
were
modeled
with
a duration
of 4 h, due
to technological
constraints.
8
Metal
−
air
batteries
were
assumed
to be iron
−
air batteries
with
a duration
of 100
h, matching
the duration
claimed
by Form
Energy
projects.
17
RFB,
PSH,
thermal
energy
storage,
and
gravity
energy
storage
were
modeled
with
separate
energy-
and
power-
capacity
components.
Charging
and
discharging
these
technologies
depend
on the same
physical
asset,
so only
one
power-capacity
cost
was
used
for each
system.
RFB
costs
were
based
on a vanadium-based
redox-flow
battery.
PSH
was
assumed
to be a closed-loop
pumped
hydroelectric
storage
system
using
upper
and
lower
water
reservoirs.
Thermal
energy
storage
was
modeled
after
a pumped-thermal
energy
storage
system,
utilizing
molten-salt
technology
for
heat
storage.
Gravity
energy
storage
was
assumed
to be a system
using
cranes
to lift heavy
bricks.
Figure
2.
System
costs
for combinations
of short-,
mid-,
and
long-duration
storage
for the contiguous
U.S.
Cost
contributions
of technologies
in
wind
and
solar
generation-based
systems
with
one,
two,
and
three
storage
technologies.
Tables
S7
−
S10
support
this figure.
System
costs
when:
(A)
no storage
technologies
were
deployed
and
the least-cost
100%
reliable
system
relied
only
on wind
and
solar
generation.
(B)
Only
one
storage
technology
was
available:
Li-ion
batteries,
redox-flow
batteries
(RFB),
pumped-storage
hydropower
(PSH),
gravity
energy
storage,
thermal
energy
storage,
compressed
air energy
storage
(CAES),
metal
−
air
battery
storage,
or hydrogen
energy
storage.
(C)
Two
storage
technologies
were
available:
Li-ion
batteries
with
the second
storage
technology
consisting
of either
a mid-duration
storage
technology
or hydrogen
energy
storage.
(D)
Two
storage
technologies
were
available:
hydrogen
energy
storage
with
the second
storage
technology
consisting
of a mid-duration
storage
technology.
(E)
Three
storage
technologies
were
available:
Li-ion
batteries
and
hydrogen
energy
storage,
with
the
third
storage
technology
consisting
of a mid-duration
storage
technology.
Environmental
Science
& Technology
pubs.acs.org/est
Article
https://doi.org/10.1021/acs.est.3c10188
Environ.
Sci. Technol.
XXXX,
XXX, XXX
−
XXX
C
CAES
and
hydrogen
storage
were
modeled
with
separate
energy-
and
power-capacity
components,
but
charging
processes
were
assigned
different
power-capacity
costs
than
the
ones
assigned
to discharging.
An
adiabatic
CAES
(A-
CAES)
system
was
assumed,
with
air compressed
into
a salt
dome
cavern,
the heat
of compression
stored
in thermal
energy
storage,
and
power
generated
by reheating
air with
stored
thermal
energy.
For
hydrogen
storage,
proton-exchange
membrane
(PEM)
electrolyzers
were
assumed
to split
water;
hydrogen
was
assumed
to be stored
underground
in salt
caverns;
and
hydrogen
was
combined
with
O
2
(g)
in PEM
fuel
cells
to generate
power.
Hydrogen
storage
was
conservatively
described
using
the
leakage
rate
characteristic
of hydrogen
stored
in pipelines,
as opposed
to the lower
leakage
rate
that
is
likely
characteristic
of hydrogen
stored
in salt caverns.
Figure
S2 shows
the base
case
costs
assumed
for the short-,
mid-,
and
long-duration
storage
technologies
considered
in this
study
(Table
S4).
9,14,17,32
Li-ion
batteries
use
the
same
technological
component
for energy
and
power
capacities,
so
their
energy
and
power
characteristics
are
not
mutually
separable.
The
capital
costs
of such
batteries
therefore
depend
on whether
the batteries
are sized
to meet
power
demand
or
energy
demand
(Figure
S3).
Li-ion
batteries
were
modeled
with
a fixed
duration
of 4 h, being
sized
to meet
short-term
power
demands,
because
they
are not
competitive
with
other
storage
technologies
on energy-capacity
costs,
especially
if used
relatively
infrequently.
8
In accord
with
currently
proposed
iron
−
air
battery
projects,
metal
−
air
batteries
were
constrained
to a fixed
duration
of 100
h, and
thus
were
sized
primarily
to
meet
energy
demand
over
their
storage
duration.
17
The
Figure
3.
System
costs
for combinations
of short-,
mid-,
and
long-duration
storage
for four
subnational
independent
system
operator
(ISO)
geographic
regions
(CAISO,
ERCOT,
ISO-NE,
and
MISO).
Tables
S7
−
S10
support
this
figure.
System
costs
when:
(A)
no storage
technologies
were
deployed,
(B)
only
one
storage
technology
was
available:
Li-ion
batteries,
redox-flow
batteries
(RFB),
pumped-storage
hydropower
(PSH),
gravity
energy
storage,
thermal
energy
storage,
compressed
air energy
storage
(CAES),
metal
−
air
battery
storage,
or hydrogen
energy
storage.
(C)
Two
storage
technologies
were
available:
Li-ion
batteries,
with
the second
storage
technology
being
a mid-duration
storage
technology
or hydrogen
energy
storage.
(D)
Two
storage
technologies
were
available:
hydrogen
energy
storage,
with
the second
storage
technology
being
a mid-duration
storage
technology.
(E)
Three
storage
technologies
were
available:
Li-ion
batteries
and
hydrogen
energy
storage,
with
the third
storage
technology
being
a mid-duration
storage
technology.
Environmental
Science
& Technology
pubs.acs.org/est
Article
https://doi.org/10.1021/acs.est.3c10188
Environ.
Sci. Technol.
XXXX,
XXX, XXX
−
XXX
D
durations
of the
other
storage
technologies
were
not
specifically
constrained
in the
modeling.
Costs
of a hypo-
thetical
Storage
X
technology
were
parameterized
over
the
entire
range
of energy-
and
power-capacity
costs
shown
in
Figure
S2, to address
the uncertainty
of storage
costs
on the
market
and
assess
the generalizability
of the findings
regarding
the
value
of different
storage
technology
portfolios
in these
stylized
electricity
systems.
■
RESULTS
CONUS
Storage
Portfolios.
Figure
2A shows
the
cost
contributions
of generation
assets
in a least-cost
system
that
relies
solely
on wind
and
solar
generation,
with
no storage
technologies
included.
In this
system,
total
system
costs
are
dominated
by costs
attributed
to wind
generation
capacity.
Figure
2B
−
E
shows
the cost
contributions
of generation
and
storage
assets
of least-cost
systems
optimized
de novo
in each
case.
Figure
2B
shows
scenarios
in which
one
storage
technology
(short-,
mid-,
or long-duration)
was
deployed.
Figure
2C,D
shows
scenarios
in which
mid-duration
storage
was
deployed
as well
as either
short-
or long-duration
storage,
respectively.
Figure
2E shows
scenarios
in which
short-,
mid-,
and
long-duration
storage
were
deployed.
Figure
S3 shows
analogous
results
for regional
ISOs.
Using
base-case
cost
assumptions,
the least-cost
system
that
used
only
short-duration
storage
(i.e.,
Li-ion
and/or
RFB)
had
the highest
total
system
costs
(Figure
2) of all scenarios
with
storage
technologies
evaluated,
representing
a
∼
55%
reduction
in total
CONUS
system
costs
as compared
to the least-cost
100%
reliable
system
that
had
only
wind
and
solar
generation
without
storage.
The
high
total
system
costs
resulted
primarily
from
the large
wind
generation
capacity
that
was
still
required
to meet
demand
in full,
given
the seasonal
and
weather-related
variability
of the wind
resource
over
CONUS.
33
System
costs
were
reduced
when
any
other
type
of storage
was
used,
either
instead
of or in combination
with
short-duration
storage
(Figure
2), to obtain
reliable,
least-cost
systems.
The
observed
cost
reductions
between
these
various
least-cost
systems
were
dominated
by a decrease
in the installed
wind
capacity.
At current
costs,
least-cost
CONUS
systems
that
used
hydrogen
energy
storage
alone
or in combination
with
other
storage
technologies
resulted
in the lowest
total
system
costs
and
constituted
a
∼
72%
reduction
in total
system
costs
as
compared
to the least-cost
100%
reliable
system
that
had
only
wind
and
solar
generation
without
storage
(Figure
2).
Regional
Storage
Portfolios.
In ISO
regions
with
high
wind
energy
potential
(e.g.,
MISO),
long-duration
energy
storage
resulted
in the
lowest
system
costs
and
thus
had
a
higher
value
than
short-duration
storage.
These
trends
were
also
observed
in regions
with
high
solar
resources
(e.g.,
CAISO),
although
the difference
in added
value
between
the
two
storage
types
was
less
pronounced
than
in regions
with
high
wind
resources
(Figure
3, Tables
S2 and
S7).
The
cost
reductions
in all regions
considered
were
associated
with
a
substantial
decrease
in the wind
generation
capacity,
as well
as
with
a comparatively
smaller
reduction
in the solar
generation
capacity.
Long-duration
storage
compensated
effectively
for
the
seasonal
variability
and
discharge
needs
associated
with
wind
and
solar
generation
in CONUS
and
regional
ISOs
and
Figure
4.
Energy
in storage
over
one
year
for combinations
of short-,
mid-,
and
long-duration
storage.
The
role
(optimized
discharge
time)
of mid-
duration
storage
technologies
(here
represented
by redox-flow
batteries,
RFB)
depended
on the availability
of short-
and
long-duration
storage.
Figures
S9
−
S38
show
analogous
results
for regional
indepedent
system
operators
and
other
mid-duration
energy
storage
technologies.
Energy
in
storage
over
one
year
when:
(A)
RFB
was
the only
storage
technology.
(B)
RFB
had
lower
power
costs
than
Li-ion
batteries
and
thus
acted
as
short-duration
storage.
(C)
RFB
had
lower
energy
costs
than
electrolytic
hydrogen
and
thus
acted
as long-duration
storage.
(D)
RFB
was
not
present
in the least-cost
system,
because
less
expensive
short-
and
long-duration
storage
technologies
were
available.
Environmental
Science
& Technology
pubs.acs.org/est
Article
https://doi.org/10.1021/acs.est.3c10188
Environ.
Sci. Technol.
XXXX,
XXX, XXX
−
XXX
E