The Role of Energy Technical Report

NREL/TP-6A2-47187
Storage with Renewable January 2010

Electricity Generation

Paul Denholm, Erik Ela, Brendan Kirby,

and Michael Milligan
The Role of Energy Storage Technical Report
NREL/TP-6A2-47187
with Renewable Electricity January 2010

Generation

Paul Denholm, Erik Ela, Brendan Kirby,

and Michael Milligan

Prepared under Task No. WER8.5005

National Renewable Energy Laboratory

1617 Cole Boulevard, Golden, Colorado 80401-3393
303-275-3000 www.nrel.gov
NREL is a national laboratory of the U.S. Department of Energy
Office of Energy Efficiency and Renewable Energy
Operated by the Alliance for Sustainable Energy, LLC
Contract No. DE-AC36-08-GO28308
 NOTICE

This report was prepared as an account of work sponsored by an agency of the United States government.
Neither the United States government nor any agency thereof, nor any of their employees, makes any
warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or
usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not
infringe privately owned rights. Reference herein to any specific commercial product, process, or service by
trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement,
recommendation, or favoring by the United States government or any agency thereof. The views and
opinions of authors expressed herein do not necessarily state or reflect those of the United States
government or any agency thereof.

Available electronically at http://www.osti.gov/bridge

Available for a processing fee to U.S. Department of Energy

and its contractors, in paper, from:
U.S. Department of Energy
Office of Scientific and Technical Information
P.O. Box 62
Oak Ridge, TN 37831-0062
phone: 865.576.8401
fax: 865.576.5728
email: mailto:reports@adonis.osti.gov

Available for sale to the public, in paper, from:

U.S. Department of Commerce
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
phone: 800.553.6847
fax: 703.605.6900
email: orders@ntis.fedworld.gov
online ordering: http://www.ntis.gov/ordering.htm

Printed on paper containing at least 50% wastepaper, including 20% postconsumer waste
Acknowledgments
The authors wish to thank the following people for their review and other contributions:
Doug Arent, Easan Drury, Vahan Gevorgian, Michelle Kubik, Jim Leyshon, Sean Ong,
and Brian Parsons of the National Renewable Energy Laboratory (NREL); Jonah Levine
of the University of Colorado; Mark OMalley of the Electricity Research Center,
University College Dublin; and Samir Succar of the Natural Resources Defense Council.

iii
List of Acronyms
AC alternate current
AGC automatic generation control
AS ancillary services
CAES compressed-air energy storage
CCGT combined-cycle gas turbine
CCR capital charge rate
CSP concentrating solar power
DC direct current
ERCOT Electric Reliability Council of Texas
EVs electric vehicles
GW gigawatt
ISO independent system operators
LaaR Load Acting as a Resource (program)
MW megawatt
NERC North American Electric Reliability Corporation
PHS pumped hydro storage
PV photovoltaics
RTO regional transmission organization
RE renewable energy
SMES superconducting magnetic energy storage
T&D transmission and distribution
V2G vehicle to grid
VG variable generation

iv
Table of Contents
1 Introduction ..................................................................................................................................1
2 Operation of the Electric Grid ....................................................................................................2
3 Electricity Storage in the Existing Grid ..................................................................................... 6
3.1 Development of Energy Storage in Regulated Markets ....................................6
3.2 The Economics of Energy Storage in Restructured Markets .............................8
3.3 Other Applications of Energy Storage .............................................................12
3.4 Summary of Energy Storage Applications in the Current Grid .......................13
4 Impacts of Renewables on the Grid and the Role of Enabling Technologies ..................... 17
4.1 Costs of Wind and Solar Integration from Previous Studies ...........................19
4.2 Limiting Factors for Integration of Wind and Solar Energy............................23
5 Storage and Flexibility Options for Renewables-Driven Grid Applications ........................ 34
5.1 The Flexibility Supply Curve...........................................................................34
5.2 Deployment and Operational Balancing of Renewable Energy Individual
Plant Storage vs. Power System Storage ...............................................................36
5.3 Energy Storage Technologies and Applications ..............................................38
5.4 Electric Vehicles and the Role of Vehicle to Grid ...........................................44
6 Conclusions ............................................................................................................................... 46
Bibliography ..................................................................................................................................47
References ....................................................................................................................................48

v
List of Tables
Table 3.1. Historical Values of Energy Storage in Restructured Electricity Markets .......10
Table 3.2. Traditional Major Grid Applications of Energy Storage ..................................14
Table 4.1. Summary of Recent Wind Integration Cost Studies .........................................20
Table 5.1. Dedicated Renewable Applications of Energy Storage and Their Whole-Grid
Counterpart ........................................................................................................................37
Table 5.2. Three Classes of Energy Storage .....................................................................38

List of Figures
Figure 2.1. Hourly loads from ERCOT 2005 ......................................................................2
Figure 2.2. System load following and regulation ...............................................................3
Figure 2.3. Optimal and reserve constrained dispatch .........................................................5
Figure 3.1. Historical fuel costs for intermediate load power plants ...................................7
Figure 3.2. Relationship between the annual benefit of storage and capital cost using
different capital charge rates ..............................................................................................11
Figure 4.1. Impact of net load from increased use of renewable energy ...........................18
Figure 4.2. Dispatch with low VG penetration ..................................................................24
Figure 4.3. Dispatch with higher VG penetration ..............................................................25
Figure 4.4. Relationship between price and load in PJM in 2002 .....................................27
Figure 4.5. Contribution from inflexible generation ..........................................................28
Figure 4.6. Effect of decreasing minimum load point on increased use of RE .................29
Figure 4.7. Average curtailment rate as a function of VG penetration for different
flexibilities in ERCOT .......................................................................................................30
Figure 4.8. Marginal curtailment rate as a function of VG penetration for different system
flexibilities in ERCOT .......................................................................................................30
Figure 4.9. Relative cost of VG as a function of VG penetration for different system
flexibilities in ERCOT .......................................................................................................31
Figure 4.10. Option for increasing the use of VG by decreasing curtailment ...................32
Figure 4.11. Reduction of curtailment resulting from addition of energy storage ............33
Figure 5.1. Flexibility supply curve ...................................................................................34
Figure 5.2. Energy storage applications and technologies .................................................39

vi
1 Introduction
Renewable energy sources, such as wind and solar, have vast potential to reduce
dependence on fossil fuels and greenhouse gas emissions in the electric sector. Climate
change concerns, state initiatives including renewable portfolio standards, and consumer
efforts are resulting in increased deployments of both technologies. Both solar
photovoltaics (PV) and wind energy have variable and uncertain (sometimes referred to
as intermittent) 1 output, which are unlike the dispatchable sources used for the majority
of electricity generation in the United States. The variability of these sources has led to
concerns regarding the reliability of an electric grid that derives a large fraction of its
energy from these sources as well as the cost of reliably integrating large amounts of
variable generation into the electric grid. Because the wind doesnt always blow and the
sun doesnt always shine at any given location, there has been an increased call for the
deployment of energy storage as an essential component of future energy systems that
use large amounts of variable renewable resources. However, this often-characterized
need for energy storage to enable renewable integration is actually an economic
question. The answer requires comparing the options to maintain the required system
reliability, which include a number of technologies and changes in operational practices.
The amount of storage or any other enabling technology used will depend on the costs
and benefits of each technology relative to the other available options.

To determine the potential role of storage in the grid of the future, it is important to
examine the technical and economic impacts of variable renewable energy sources. It is
also important to examine the economics of a variety of potentially competing
technologies including demand response, transmission, flexible generation, and improved
operational practices. In addition, while there are clear benefits of using energy storage to
enable greater penetration of wind and solar, it is important to consider the potential role
of energy storage in relation to the needs of the electric power system as a whole.

In this report, we explore the role of energy storage in the electricity grid, focusing on the
effects of large-scale deployment of variable renewable sources (primarily wind and solar
energy). We begin by discussing the existing grid and the current role that energy storage
has in meeting the constantly varying demand for electricity, as well as the need for
operating reserves to achieve reliable service. The impact of variable renewables on the
grid is then discussed, including how these energy sources will require a variety of
enabling techniques and technologies to reach their full potential. Finally, we evaluate the
potential role of several forms of enabling technologies, including energy storage.

1
The use of the term intermittent has been questioned by the wind energy community as being
technically inaccurate. Intermittent implies a short-term on-off cycle while the output of wind
experiences maximum variations more typically on the order of 10% per hour. Solar PV is perhaps
somewhat more intermittent because it follows a daily on-off cycle. The description variable or
variable and uncertain has been proposed as a more technically accurate description of the output of a
wind power plant (Smith and Parsons 2007).

1
2 Operation of the Electric Grid
The operation of electric power systems involves a complex process of forecasting the
demand for electricity, and scheduling and operating a large number of power plants to
meet that varying demand. The instantaneous supply of electricity must always meet the
constantly changing demand, as indicated in Figure 2.1. It shows the electricity demand
patterns for three weeks for the Electric Reliability Council of Texas (ERCOT) grid
during 2005. 2 The seasonal and daily patterns are driven by factors such as the need for
heating, cooling, lighting, etc. While the demand patterns in Figure 2.1 are for a specific
region of the United States, many of the general trends shown in the demand patterns are
common throughout the country. To meet this demand, utilities build and operate a
variety of power plant types. Baseload plants are used to meet the large constant demand
for electricity. In the United States, these are often nuclear and coal-fired plants, and
utilities try to run these plants at full output as much as possible. While these plants
(especially coal) can vary output, their high capital costs, and low variable costs (largely
fuel), encourage continuous operation. Furthermore, technical constraints (especially in
nuclear plants) restrict rapid change in output needed to follow load. Variation in load is
typically met with load-following or cycling plants. These units are typically
hydroelectric generators or plants fueled with natural gas or oil. These load-following
units are further categorized as intermediate load plants, which are used to meet most of
the day-to-day variable demand; and peaking units, which meet the peak demand and
often run less than a few hundred hours per year.

1.0

60000 0.9

Load (Fraction of Annual Peak)

0.8
50000
0.7
Load (MW)

40000 0.6

0.5
30000
0.4

20000 0.3

Summer Maximum Winter Spring Minimum 0.2

10000
0.1

0 0.0
0 24 48 72 96 120 144 168
Hour
Figure 2.1. Hourly loads from ERCOT 2005

2
Most of Texas (about 85% of the population) is within the ERCOT grid, which is largely independent of
the two larger U.S. grids.

2
In addition to meeting the predictable daily, weekly, and seasonal variation in demand,
utilities must keep additional plants available to meet unforeseen increases in demand,
losses of conventional plants and transmission lines, and other contingencies. This class
of responsive reserves is often referred to as operating reserves and includes meeting
frequency regulation (the ability to respond to small, random fluctuations around normal
load), load-forecasting errors (the ability to respond to a greater or less than predicted
change in demand), and contingencies (the ability to respond to a major contingency such
as an unscheduled power plant or transmission line outage) (NERC 2008). 3 Both
frequency regulation and contingency reserves are among a larger class of services often
referred to as ancillary services, which require units that can rapidly change output.
Figure 2.2 illustrates the need for rapidly responding frequency regulation (red) in
addition to the longer term ramping requirements (blue). In this utility system, the
morning load increases smoothly by about 400 megawatts (MW) in two hours. During
this period, however, there are rapid short-term ramps of +/- 50 (MW) within a few
minutes.

Figure 2.2. System load following and regulation. Regulation (red) is the fast fluctuating
component of total load (green) while load following (blue) is the slower trend
(Kirby 2004)
Because of the rapid response needed by both regulation and contingency reserves, a
large fraction of these reserves are provided by plants that are online and spinning (as a
result, operating reserves met by spinning units are sometimes referred to as spinning
reserves.) 4 Spinning reserves are provided by a mix of partially loaded power plants or
responsive loads. The need for reserves increases the costs and decreases the efficiency of

3
Operating reserves are primarily capacity services (the ability to provide energy on demand) as opposed to
actual energy services.
4
The nomenclature around various ancillary services (especially spinning reserves) varies significantly.
While the NERC glossary indicates that spinning reserve applies to both contingency and frequency
regulation, the term spinning reserve often is used to refer to only contingency reserves. For additional
discussion of nomenclature around contingency and spinning reserves, see Rebours and Kirschen 2005.

3
an electric power system compared to a system that is perfectly predictable and does not
experience unforeseen contingencies. These costs result from several factors. First, the
need for fast-responding units results in uneconomic dispatch because plants providing
spinning reserve must be operated at part load, they potentially displace more economic
units. 5 (Flexible load-following units are often either less efficient or burn more
expensive fuel than baseload coal or nuclear units.) Second, partial loading can reduce
the efficiency of individual power plants. Finally, the reserve requirements increase the
number of plants that are online at any time, which increases the capital and O&M costs.

Figure 2.3 provides a simplified illustration of the change in dispatch (and possible cost
impacts) needed to provide operating reserves. The figure on the left shows an ideal
dispatch of a small electric power system. Two baseload units provide most of the
energy, while an intermediate load and two peaking units provide load following. In the
ideal dispatch, it is possible that the intermediate load unit cannot rapidly increase
output to provide operating reserves. Furthermore, during the transition periods when the
load-following units are nearing their full output but before additional units are turned
on there may be insufficient capacity left in the load-following units to provide
necessary operating capacity for regulation or contingencies. A dispatch that provides the
necessary reserves is provided on the right. In this case, lower-cost units reduce output to
accommodate the more flexible units providing reserves. This increases the overall cost
of operating the entire system.

5
This opportunity cost associated with uneconomic dispatch is the dominant source of reserve costs
(Kirby 2004).

4
 3000

2500
Insufficient Peak 2
2000 Reserve
Load (MW)

Peak 1
Capacity
1500 Int
Base2
1000 Higher Cost Unit Started
Insufficient Base 1 Early To Provide Reserve
500 Flexibility to Capacity
Provide Reserves
0 3000
0 4 8 12 16 20 24
Hour 2500

Ideal Dispatch 2000

Load (MW)
1500

1000

Lower Cost Unit Backed Off 500

To Accommodate Higher
Cost Flexible Generator 0
0 4 8 12 16 20 24
Reserve Constrained Hour
Dispatch
Figure 2.3. Optimal and reserve constrained dispatch
The need for operating reserves and the large variation in demand restricts the
contribution from low-cost baseload units and increases the need for units that can vary
output to provide both load-following and ancillary services. As a consequence, utility
operators have long pursued energy storage as one potential method of better utilizing
baseload plants and providing an alternative to lower efficiency thermal generators for
meeting variations in demand.

5
3 Electricity Storage in the Existing Grid
The challenges associated with meeting the variation in demand while providing reliable
services has motivated historical development of energy storage. While a number of
pumped hydro storage (PHS) plants 6 were built in the United States before 1970,
significant interest, research, and funding for new storage technologies began in the early
1970s, associated with dramatic increases in oil prices. This period also saw the largest
deployment of PHS based on its competitive economics compared to alternative sources
of intermediate load and peaking energy.

3.1 Development of Energy Storage in Regulated Markets

Deployment of energy storage is dependent on the economic merits of storage
technologies compared to the more conventional alternatives used to follow load. Before
the advent of low-cost, efficient gas turbines now typically used to follow load and
provide reserves, utilities often relied on oil- and gas-fired steam turbines (and
hydroelectric dams where available). In the 1970s, dramatic price increases in oil and
natural gas occurred, along with concerns about security of supply. This led to the
Powerplant and Industrial Fuel Use Act, restricting use of oil and gas in new power plants
(EIA 2009b). Utilities expected to bring online many new coal and nuclear plants to meet
baseload demand, but were left with limited options to provide load-following and
peaking services. 7 This led utilities to actively evaluate pumped hydro (along with other
storage technologies) as alternatives to fossil-fueled intermediate load and peaking units.
The economic analysis and justification of new energy storage facilities during this
period was based on a direct comparison of the energy and capacity provided by energy
storage to an equivalently sized fossil plant, (choosing the lower net-cost option) which
largely ignored any additional operational benefits energy storage can provide. 8
Figure 3.1 provides a simple framework of comparing these technologies over time. In
the figure, the variable (fuel-related) costs are shown for a storage device and fossil-
fueled alternatives. In this figure, the storage technology is assumed to be fueled with off-
peak coal and has an effective round-trip efficiency of 75%. 9

6
PHS stores energy by pumping water from a lower reservoir to an upper reservoir and releasing that
stored water through a conventional hydroelectric generator. Additional information about PHS is provided
in Section 5.
7
Concerns about the availability of oil and other peaking fuels in this period was so great that an
international conference (including the U.S. National Academy of Sciences) on the subject in 1979
described energy storage as a vital element in mankinds quest for survival and progress (Silverman
1980).
8
See, for example, EPRI 1976. Here, the proposed method for comparing energy storage to conventional
alternatives is based solely on the value of energy and firm capacity value without any actual quantification
of operational benefits.
9
This would be a typical assumption for a pumped hydro plant built during the 1970s and 1980s (EPRI
1976).

6
 14.0
Gas Steam
Oil Steam
12.0 Gas CCGT
Storage (Coal Charging)

10.0
Fuel Cost (2008 Cents/kWh)

8.0

6.0

4.0

2.0

0.0
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
Year

Figure 3.1. Historical fuel costs for intermediate load power plants 10

Figure 3.1 shows that the variable cost of providing energy from a storage device was
much lower than alternatives available in the mid-1970s and early 1980s. While
Figure 3.1 provides the fuel costs, the total economics of storage must also consider the
fixed costs. During the mid- to late 1970s, gas-fired combined cycle plants were not
significantly less expensive than pumped hydro, with cost estimates of $110-$280/kW for
a 10-hour PHS device and $175-275/kW for a combined-cycle generator (EPRI 1976).
As a result, pumped hydro appeared more economic than alternative generation sources
during this time period, even without considering additional operation benefits. It was
expected that oil and gas prices would remain high, and that off-peak energy would be
widely available (and even less expensive) due to anticipated large-scale deployment of
nuclear power plants.

During the mid- to late 1970s, much of the nations 20 GW of pumped hydro storage was
initiated (ASCE 1993), along with significant research and development in a variety of
other storage technologies including several battery types, capacitors, flywheels, and
superconducting magnetic storage (DOE 1977). Growth projections for energy storage
during this period included significant increases of several types (Boyd et al. 1983).
However, most PHS development, along with interest in and deployment of other
10
This figure is intended to represent general trends as opposed to absolute costs. In this figure, fuel prices
and generation characteristics are derived from various data sets from the Energy Information
Administration.

7
emerging storage technologies, ended in the 1980s after dramatic reduction in the price of
natural gas, increased efficiency and reduced costs of flexible combined-cycle and
simple-cycle natural gas turbines, and repeal of the Fuel Use Act in 1987. While
estimates from the 1970s place combined-cycle gas turbine (CCGT) units and PHS at
similar costs, by the early 2000s, PHS was estimated to be about twice the cost of a
CCGT. As a result, even with the increased cost of gas, this dramatic increase in PHS
cost (along with the many other factors discussed previously) limited the economic
competitiveness of PHS vs. gas-fired generators. 11 Furthermore, while coal prices
continued to drop, the limited nuclear build-out eliminated a source of low-cost off-peak
electricity. Finally, the simplistic treatment of the economic benefits of energy storage
technologies was also a limiting factor. One of the main benefits of energy storage is its
ability to provide multiple services, including load leveling (and associated benefits such
as a reduction in cycling-induced maintenance) along with regulation and contingency
reserves and firm capacity. 12 However, it has always been somewhat difficult to quantify
these various value streams without fairly sophisticated modeling and simulation
methods, (especially before the advent of energy and ancillary service markets, which
will be discussed in the next section). Because the economic analysis is difficult, and
benefits of storage are often uncertain, utilities tend to rely on more traditional generation
assets, especially in regulated utilities where risk is minimized and new technologies are
adopted relatively slowly. 13

Combined, these factors have restricted deployment of utility-scale energy storage in the
United States. Besides PHS, deployment has been limited to a single 110 MW
compressed-air energy storage (CAES) facility, and a variety of small projects. 14 A more
comprehensive discussion of energy storage technologies and their status is provided in
Section 5.

3.2 The Economics of Energy Storage in Restructured Markets

Despite the lack of significant new construction, interest in energy storage never
completely disappeared during the period of low-cost peaking fuels. Research and
development has continued, along with an increasing number of proposed projects.

11
PHS also takes longer to build (increasing the risk for investors), requires additional permits, and is
typically located farther from load centers, which requires more transmission than gas-fired generators.
PHS may also face greater environmental opposition (Strauss 1991).
12
This problem has been noted many times. For example, traditionally, when electric utilities evaluate
generating additions to their facilities, the evaluation process considers the contribution of each alternative
to both capacity and energy requirements. However, the evaluation process often neglects or inaccurately
measure potential costs and benefits not directly related to capacity and energy. Operating considerations
that reflect the ability (or inability) of a generation resource to respond to the electric systems dynamic
operating needs usually fall into this category. (Jabbour and Wells 1992).
13
Private investors tend to favor lower capital cost investments with faster construction times (i.e.,
combustion turbines and combined-cycle plants), even if they have higher operating costs, because this
reduces perceived economic risk.
14
To place these values in perspective, between 1993 and 2008, more than 320 GW of conventional
capacity was constructed in the United States. With the exception of the completion of previously started
PHS facilities and a few demonstration projects, no significant storage capacity was added. The total U.S.
utility storage capacity of about 20 GW in 2008 is less than 2% of the total installed generation capacity
(EIA 2009a).

8
Recent renewed interest in energy storage has been motivated by at least five factors:
advances in storage technologies, an increase in fossil fuel prices, the development of
deregulated energy markets including markets for high-value ancillary services,
challenges to siting new transmission and distribution facilities, and the perceived need
and opportunities for storage with variable renewable generators.

Emergence of wholesale electricity markets along with increased volatility in natural gas
prices has created new opportunities and interest in energy storage. As shown in Figure
3.1, rising natural gas prices in the early 2000s increased the cost-competitiveness of
energy storage. However, perhaps the single greatest motivation for proposals to build
new energy storage is the creation of markets for both energy and ancillary services
including regulation, contingency reserves, and capacity. As of 2009, wholesale energy
markets exist in parts of more than 30 states and cover about two-thirds of the U.S.
population (IRC 2009). The markets provide real, transparent data for both utilities and
independent power producers to consider the opportunities for energy storage. Market
data allows evaluation of both the economic yield and optimum location of energy
storage devices for arbitrage the ability to purchase low-cost off-peak energy and re-sell
this energy during on-peak periods. Furthermore, the benefits of providing operating
reserves and other ancillary services from energy storage can now be evaluated.
Previously, the value of these services was largely hidden in utilities cost of service,
and the cost of providing operating reserves, for example, was rarely calculated. The
high value of these services is now recognized, and the advantages of energy storage in
providing these services is evident, especially because these services generally require
fast response and limited actual energy delivery, two qualities that are well-suited to
many energy storage devices.

Historical market data can be used to evaluate the potential profitability of energy storage
devices that provide various services. Table 3.1 provides the results of several studies of
U.S. electricity markets.

9
 Table 3.1. Historical Values of Energy Storage in Restructured Electricity Markets
Market Location Years Annual Assumptions
Evaluated Evaluated Value
($/kW)
Energy PJMa 2002- $60-$115 12 hour, 80% efficient device. Range of
Arbitrage 2007 efficiencies and sizes evaluated 15
NYISOb 2001- $87-$240 10 hour, 83% efficient device. Range of
2005 (NYC) efficiencies and sizes evaluated.
$29-$84
(rest)
USAc 1997- $37-$45 80% efficient device, Covers NE, No Cal,
2001 PJM
CAd 2003 $49 10 hour, 90% efficient device.
Regulation NYISOb 2001- $163-248
2005
USAe 2003- $236-$429 PJM, NYISO, ERCOT, ISONE
2006
Contingency USAe 2004- $66-$149 PJM, NYISO, ERCOT, ISONE
Reserves 2005
a
Sioshansi et al. 2009
b
Walawalkar et al. 2007
c
Figueiredo et al. 2006
d
Eyer et al. 2004
e
Denholm and Letendre 2007

The values in Table 3.1 can be translated into a maximum capital cost for the applicable
storage technology (equal to the maximum cost of a storage device that can be supported
by the revenues available). Figure 3.2 provides a generic conversion between annual
costs and total capital costs. This conversion is performed by dividing the annual
revenues by the capital charge rate, which produces a total capital cost. 16 The capital
charge rate (also referred to as a fixed-charge rate or capital recovery factor) refers to the
fraction of the total capital cost that is paid each year to finance the plant. It should be
noted that this cost does not include any operation and maintenance costs. 17

15
This study analyzed devices up to 40 hours, and found rapidly diminishing returns for devices with
storage capacity greater than about 10 hours. The majority of arbitrage benefits are within a day, as
opposed to over larger time periods. In addition, while short-term price variation is highly predictable,
long-term variations are less predictable, which reduces the certainty of long-term arbitrage opportunities.
16
This is the inverse of the process of calculating an annualized cost from a capital cost by multiplying by
the capital charge rate.
17
These values also assume that frequency regulation is an energy- and cost-neutral service.

10
 4000
Capital
Charge Rate
3500
Required Storage Capital Cost ($/kW)

9.8%
3000 12.0%
13.9%
2500

2000

1500 Arbitrage Only

Regulation Only
1000

500
Contingency Only
0
0 50 100 150 200 250 300 350
Annual Benefit of Storage ($/kW)

Figure 3.2. Relationship between the annual benefit of storage and capital cost using
different capital charge rates 18
Figure 3.2 shows the range of values and corresponding capital costs for three types of
operation: energy arbitrage, contingency reserves, and frequency regulation. In general,
energy arbitrage provides the least value. Outside of New York City, the maximum
annual value in Table 3.1 for arbitrage was $115/kW (for a 20-hour device), which would
translate into a capital cost of $827-1,170/kW this is below current estimates for most
energy storage technologies. While there is significant uncertainty in costs, most energy
storage assessments indicate that few commercially available bulk energy storage
technologies are deployable for less than $1,000/kW. 19 This reveals the same challenging
economics as comparing a storage device to a conventional generator as discussed in
Section 3-1. The value of energy arbitrage alone does not appear to justify the
deployment of energy storage at current technology costs and electricity prices.

The value of energy storage increases when taking advantage of other individual sources
of revenue or even combined services. A device with sufficient energy capacity for
energy arbitrage would likely be able to receive capacity payments in locations where

18
Capital charge rate (CCR) of 9.8% from EPRI 2003, CCR of 12% from Butler et al. 2003, CCR of 13.9%
from Eyer et al. 2004.
19
CAES is one possible exception, but requires analysis of both the electricity price and natural gas prices.
See Section 5 and the references in the Bibliography for additional discussion of storage costs.

11
capacity markets now exist; recent data in the PJM market indicates an additional
potential value of $40-90/kW-year. 20

Alternatively, contingency reserves offer a higher value 21 than energy arbitrage and also
require less energy capacity. Obtaining the values for energy arbitrage in Table 3.1
generally requires a device with at least 10 hours of storage capacity; contingency
reserves can require as little as 30 minutes, depending on the market and market
reliability rules (PJM 2009b). The challenge for a device providing contingency reserves
is that the device must be able to respond rapidly, typically in a few minutes or less.
Frequency regulation is even more demanding, requiring continuous changes in output,
frequent cycling, and fast response. It is also the highest-value opportunity for an energy
storage device, and has been the focus of many potential energy storage applications,
especially given its fairly small energy requirements. 22

3.3 Other Applications of Energy Storage

In addition to energy arbitrage and operating reserves, there are several other services that
energy storage has provided or could provide in the current grid. Several of these
applications are discussed below.

Transmission and Distribution

In addition to generation, storage can act as an alternative or supplement to new
transmission and distribution (T&D). Distribution systems must be sized for peak
demand; as demand grows, new systems (both lines and substations) must be installed,
often only to meet the peak demand for a few hours per year. New distribution lines may
be difficult or expensive to build, and can be avoided or deferred by deploying distributed
storage located near the load (Nourai 2007). (Energy can be stored during off-peak
periods when the distribution system is lightly loaded, and discharged during peak
periods when the system may otherwise be overloaded.) Energy storage can also reduce
the high line-loss rates that occur during peak demand (Nourai et al. 2008).

20
PJMs capacity market (Reliability Pricing Model) data derived from http://www.pjm.com/markets-and-
operations/rpm.aspx.
21
More recent data shows a greater range of value for contingency reserves. For example, in 2008, the
average value for spinning reserves in NYISO was $10.1 in the east and $6.2 in the west, corresponding to
as little as $54/kW-yr. In the same year, the average price in ERCOT was $27.1, corresponding to
$237/kW-yr. In 2009, these values fell substantially to $36/kW-yr in NYISO west, and $87/kW-yr in
ERCOT.
22
Frequency regulation theoretically is a net zero energy service over relatively short time scales, meaning
the energy capacity of the device can be much smaller than those providing operating reserves and energy
arbitrage. Several markets in the United States have changed or have proposed to change their treatment of
regulation to accommodate energy-limited storage technologies. Furthermore, it has been suggested that
fast-responding storage devices could receive a greater value per unit of capacity actually bid, because they
could actually reduce the amount of reserves needed. For example, faster responsive resources can help to
reduce California ISOs regulation procurement by up to 40% (on average) and California ISO may
consider creating better market opportunities and incentives for fast responsive resources. (Makarov et al.
2008).

12
Black-Start
Black-start provides capacity and energy after a system failure. A black-start unit
provides energy to help other units restart and provide a reference frequency for
synchronization. Pumped hydro units have been used for this application. 23

Power Quality and Stability

Energy storage can be used to assist in a general class of services referred to as power
quality and stability. Power quality refers to voltage spikes, sags, momentary outages,
and harmonics. Storage devices are often used at customer load sites to buffer sensitive
equipment against power quality issues. Electric power systems can also experience
oscillations of frequency and voltage. Unless damped, these disturbances can limit the
ability of utilities to transmit power and affect the stability and reliability of the entire
system. System stability requires response times of less than a second, and can be met by
a variety of devices including fast-responding energy storage.

End-Use/Remote Applications
Other applications for energy storage are at the end use. Storage can provide firm power
for off-grid homes, but also can provide value when grid-tied through management of
time-of-use rates, or demand charges in large commercial and industrial buildings.
Energy storage also provides emergency and backup power for increased reliability. In
many cases, end-use applications have analogous applications in the grid as a whole (and
potentially compete with these applications). For example, using energy storage to time-
shift end use is functionally equivalent to energy arbitrage, and a flatter load on the
demand side reduces the potential need for load-leveling in central storage applications
(and vice-versa). To be economic, end-use applications require time-varying prices and
are extremely site-specific.

3.4 Summary of Energy Storage Applications in the Current Grid

Table 3.2 provides a summary of the various applications of energy storage commonly
discussed in the literature. Each of these applications provides a potential value to a
merchant storage operator in a restructured market or a source of cost reduction to the
system.

23
Large PHS units or other black-start generators must themselves be black-started and may use batteries
or small generators for this purpose. Many transmission substations also use batteries partly to maintain
reliability during power failures.

13
 Table 3.2. Traditional Major Grid Applications of Energy Storage
Application Description System Benefits when Provided by Timescale of Operation
Storage
Load Leveling/ Purchasing low-cost off-peak energy Increases utilization of baseload power Response in minutes to hours.
Arbitrage and selling it during periods of high plants and decrease use of peaking Discharge time of hours.
prices. plants. Can lower system fuel costs, and
potentially reduce emissions if peaking
units have low efficiency.
Firm Capacity Provide reliable capacity to meet Replace (or function as) peaking Must be able to discharge continuously
peak system demand. generators. for several hours or more.
Operating
Reserves

Regulation Fast responding increase or Reduces use of partially loaded thermal Unit must be able to respond in
decrease in generation (or load) to generators, potentially reducing both fuel seconds to minutes. Discharge time is
respond to random, unpredictable use and emissions. typically minutes. Service is
variations in demand. theoretically net zero energy over
extended time periods.

Contingency Fast response increase in generation Same as regulation. Unit must begin responding immediately
Spinning (or decrease load) to respond to a and be fully responsive within 10
24
Reserve contingency such as a generator minutes. Must be able to hold output for
failure. 30 minutes to 2 hours depending on the
25
market. Service is infrequently called.

Replacement/ Units brought on-line to replace Limited. Replacement reserve is typically Typical response time requirement of
Supplemental spinning units. a low-value service. 30-60 minutes depending on market
minutes. Discharge time may be several
hours.

24
Contingency reserves may be provided by both spinning and non-spinning units, depending on the market. The requirements for non-spinning reserves are the
same except the resource does not need to begin responding immediately. Full response is still within 10 minutes.
25
For example, in the PJM regional transmission organization (RTO) in 2008 (covering about 50 million people), synchronized reserves were called a total of 40
times with an average duration of 10 minutes. See http://www.pjm.com/markets-and-operations/ancillary-services.aspx

14
 Ramping/Load Follow longer term (hourly) changes Reduces use of partially loaded thermal Response time in minutes to hours.
Following in electricity demand. generators, potentially reducing both fuel Discharge time may be minutes to
use and emissions. Price is embedded hours.
in existing energy markets, but not
explicitly valued, so somewhat difficult to
capture.
T&D Reduce loading on T&D system Provides an alternative to expensive and Response in minutes to hours.
Replacement during peak times. potentially difficult to site transmission Discharge time of hours.
and Deferral and distribution lines and substations.
Distribution deferral is not captured in
existing markets.
Black-Start Units brought online to start system Limited. May replace conventional Response time requirement is several
after a system-wide failure (blackout). generators such as combustion turbines minutes to over an hour. Discharge time
or diesel generators. requirement may be several to many
hours. 26
End-Use
Applications

TOU Rates Functionally the same as arbitrage, Same as arbitrage. Same as arbitrage.
just at the customer site.

Demand Functionally the same as firm Same as firm capacity. Same as firm capacity.
Charge capacity, just at the customer site.
Reduction
Functionally the same as contingency Benefits are primarily to the customer. Instantaneous response. Discharge
Backup Power/ reserve, just at the customer site. time depends on level of reliability
UPS/Power needed by customer.
Quality

26
The black-start performance standard in PJM is 90 minutes to start, with an ability to run for 16 hours (PJM 2009a).

15
It should be noted that Table 3.2 does not include any dedicated renewables applications.
Historical motivations for energy storage deployment were based on the challenges of
meeting variations in demand using conventional thermal generators. Much of the current
attention for energy storage is based on its potential application with renewable energy
(primarily solar and wind). Energy storage is seen as a means to firm or shape the
output from variable renewable generators. However, the actual need for storage to
perform these roles has yet to be quantified. In addition, it is unclear whether the electric
power industry must create entirely new classes of energy and capacity services to deal
with the increased uncertainty and variability created by large-scale deployment of
variable renewables. Determining the role of energy storage with renewables first
requires examining the impacts of variable generators on the grid and how these impacts
may require the use of various enabling technologies.

16
4 Impacts of Renewables on the Grid and the Role of
Enabling Technologies
The introduction of variable renewables is now one of the primary drivers behind
renewed interest in energy storage. A common claim is that renewables such as wind and
solar are intermittent and unreliable, and require backup and firming to be useful in a
utility system energy produced by wind and solar should be smoothed or shifted to
times when the wind is not blowing or the sun is not shining using energy storage. These
statements are generally qualitative in nature and provide little insight into the actual role
of renewables in the grid, (including their costs and benefits) or the potential use of
energy storage or other enabling technologies.

To evaluate the actual role of energy storage in a grid with large amounts of variable
renewable generation, we must first return to our previous discussion of the variability of
electric demand, and how the conventional generators currently meet this demand. As
discussed in Section 2, tremendous variation in daily demand is met by the constant up-
and-down cycling of generators. In addition to this daily cycling, frequency regulation
and contingency reserves are provided by partly loaded generators and responsive load. 27
Most of these flexible generators are hydro units, combustion turbines, some
combined-cycle plants, and even large thermal generators, as well as the existing PHS.

Variable generation (VG) 28 will change how the existing power plant mix is operated,
because its output is unlike conventional dispatchable generators. It is easiest to
understand the impact of VG technologies on the grid by considering them as a source of
demand reduction with unique temporal characteristics. Instead of considering wind or
PV as a source of generation, they can be considered a reduction in load with
conventional generators meeting the residual load of normal demand minus the
electricity produced by renewable generators.

Figure 4.1 illustrates this framework for understanding the impacts of variable
renewables. In this figure, renewable generation is subtracted from the normal load,
showing the residual or net load that the utility would need to meet with conventional
sources. 29 The benefits to the utility include reduced fuel use (and associated emissions) 30
and a somewhat reduced need for overall system capacity (this is relatively small for
wind but can be significant for solar given its coincidence with load.) 31 There are also
four significant impacts that change how the system must be operated and affect costs.

27
In some locations such as Texas, demand response typically provides half of the contingency reserve
requirements. Other regions also use (or are evaluating) load to provide regulation.
28
From this point on, variable renewable generators will be referred to as variable generation (VG)
following NERC 2009.
29
This figure uses ERCOT load data from 2005 along with 15 GW of spatially diverse simulated wind data
from the same year. See Section 4.2 for more details about the data used.
30
A reduction in demand from VG or load will reduce the total marginal cost of generation, which includes
fuel, emissions costs, and variable O&M. This ignores any additional cost impacts of variability on the
remaining generation fleet, which is discussed in the next section.
31
There are a number of other benefits provided by renewable energy sources such as reduced volatility of
fuel prices. This work is not intended to be an analysis of the total benefits of renewables or VG.

17
First is the increased need for frequency regulation, because wind can increase the short-
term variability of the net load (not illustrated on the chart). 32 Second is the increase in
the ramping rate, or the speed at which load-following units must increase and decrease
output. The third impact is the uncertainty in the wind resource and resulting net load. 33
The final impact is the increase in overall ramping range the difference between the
daily minimum and maximum demand and the associated reduction in minimum load,
which can force baseload generators to reduce output; and in extreme cases, force the
units to cycle off during periods of high wind output. Together, the increased variability
of the net load requires a greater amount of flexibility and operating reserves in the
system, with more ramping capability to meet both the predicted and unpredicted
variability. The use of these variable and uncertain resources will require changes in the
operation of the remaining system, and this will incur additional costs, typically referred
to as integration costs.

Ramp Range (Increases in

50000 this two-week period from
Load Wind Net Load 19.3 GW/day to 26.2
45000 GW/day)
40000

35000

30000
MW

25000

20000

15000

10000

5000

0
1-Apr 8-Apr 15-Apr
Uncertainty in wind output
Variation in wind output increases net increases uncertainty in net
load ramp ate (Increases in this period load to be met with
from 4,052 MW/hour to 4,560 MW/hour) conventional generators

Figure 4.1. Impact of net load from increased use of renewable energy

32
As discussed later, the impact of short-term wind variability is often overstated, especially considering
the benefits of spatial diversity. The impact on minute-to-minute regulation requirements is mitigated by
aggregating large amounts of wind because individual wind plant variability is uncorrelated in the
regulation time frame. Furthermore, the limited ability to forecast wind and use of persistence forecasts
may be a major factor in increased short-term variability of the net load. Finally, newer wind turbines
meeting low-voltage ride-through standards can add short-duration stability, and have the capability to
provide frequency regulation to the grid.
33
This is actually the combination of the uncertainty in load and the uncertainty in wind. As VG
penetration increases, it begins to dominate the net load uncertainty.

18
4.1 Costs of Wind and Solar Integration from Previous Studies
Concerns about grid reliability and the cost impacts of wind have driven a large number
of wind integration studies. These studies use utility simulation tools 34 and statistical
analysis to model systems with and without wind and calculate the integration costs of
wind.

The basic methodology behind these studies is to compare a base case without wind to a
case with wind, evaluating technical impacts and costs. The studies calculate the
additional costs of adding operating reserves as well as the other system changes needed
to reliably address the increased uncertainty and variability associated with wind
generation.

Table 4.1 provides examples of several integration studies from various parts of the
United States. In these studies, integration costs are typically divided into three types,
based on the timescales important to reliable and economic power system operation.
These three types are the first three of the four impacts discussed previously:

Regulation the increased costs that result from providing short-term ramping (seconds
to minutes) resulting from wind deployment.

Load following the increased costs that result from providing the hourly ramping
requirements resulting from wind deployment.

Wind uncertainty the increased costs that result from having a suboptimal mix of units
online because of errors in the wind forecast. This is typically called unit commitment or
scheduling cost because it involves costs associated with committing (turning on) too few
or too many slow-starting, but lower operational-cost units than would have been
committed if the wind forecast been more accurate.

34
These tools have several names such as production cost or security-constrained unit commitment and
economic dispatch models.

19
 Table 4.1. Summary of Recent Wind Integration Cost Studies (DeCesaro et al. 2009)
Date Study Wind Regula- Load- Unit Other Tot Oper.
Capacity tion Cost Following Commit- ($/MWh) Cost Impact
Penetra- ($/MWh) Cost ment ($/MWh)
tion (%) ($/MWh) Cost
($/MWh)
2003 Xcel-UWIG 3.5 0 0.41 1.44 Na 1.85
2003 WE Energies 29 1.02 0.15 1.75 Na 2.92
2004 Xcel-MNDOC 15 0.23 na 4.37 Na 4.6
2005 PacifiCorp-2004 11 0 1.48 3.16 Na 4.64
Calif. (multi-
2006 year)a 4 0.45 trace trace Na 0.45
b
2006 Xcel-PSCo 15 0.2 na 3.32 1.45 4.97
c
2006 MN-MISO 36 na na na na 4.41
Puget Sound
2007 Energy 12 na na na na 6.94
Arizona Pub.
2007 Service 15 0.37 2.65 1.06 na 4.08
d
2007 Avista Utilities 30 1.43 4.4 3 na 8.84
2007 Idaho Power 20 na na na na 7.92
2007 PacifiCorp-2007 18 na 1.1 4 na 5.1
e
2008 Xcel-PSCo 20 na na na na 8.56
a
Regulation costs represent 3-year average.
b
The Xcel/PSCO study also examine the cost of gas supply scheduling. Wind increases the uncertainty of
gas requirements and may increase costs of gas supply contracts.
c
Highest over 3-year evaluation period. 30.7% capacity penetration corresponding to 25% energy
penetration
d
Unit commitment includes cost of wind forecast error.
e
This integration cost reflects a $10/MMBtu natural gas scenario. This cost is much higher than the
integration cost calculated for Xcel-PSCo in 2006, in large measure due to the higher natural gas price: had
the gas price from the 2006 study been used in the 2008 study, the integration cost would drop from
$8.56/MWh to $5.13/MWh.

20
The overall cost impact of accommodating wind variability in these studies is typically
less than $5/MWh (0.5 cents/kWh), adding less than 10% to the cost of wind energy. 35
The majority of these costs appear to be a result of wind forecasting errors and
uncertainty resulting in unit commitment errors where too little or too much capacity
is kept online. It is worth noting that wind forecasting is an area of active research, and
these errors are expected to decrease in time, which could potentially lead to a
corresponding decrease in unit commitment errors and associated costs.

The explanation for this relatively modest impact on costs is largely based on the already
significant variation in normal load. The large amount of flexible generation already
available to meet the variability in demand has the ability to respond to the greater
variability caused by the large-scale deployment of wind. Furthermore, these studies have
found significant benefits of spatial diversity just because the wind isnt blowing in one
location, it may be in another. The combination of multiple wind sites tends to smooth
out the aggregated wind generation in a system, which reduces the per-unit size of ramps
and mitigates the range of flexibility required. 36

Far less work has been performed on the operational impacts of large-scale solar
generation due largely to lower deployment rates compared to wind. 37 In addition, there
is insufficient solar data available to estimate impacts on frequency regulation and
ramping (Lew et al. 2009.) One study of PV on the Xcel Colorado utility system found
integration costs of between $3.51/MWh and $7.14/MWh for a scenario examining 800
MW of solar in a 6,922 MW peaking system, with gas prices ranging from $7.83 to
$11.83/MMBTU (EnerNex 2009). Additional studies are ongoing, but it will be some
time until knowledge of solars impact on the grid and associated costs are understood to
the degree of wind. 38

In reality, while these studies divide the costs into three main categories, the source of
actual integration costs is largely associated with the fuel costs needed to provide the
additional required reserves, along with some variable operations and maintenance costs.
To provide regulation and load following, the additional variability requires that utilities
run more flexible generators (such as gas-fired units instead of coal units, or simple-cycle
turbines instead of combined-cycle turbines) to ensure that additional ramping
requirements can be met. This was illustrated previously in Figure 2.3 where a

35
This actually oversimplifies the situation. Some research indicates that these costs are not entirely
integration costs but a modeling artifact of how the base case in these studies is actually simulated
(Milligan and Kirby 2009).
36
Combining geographically diverse wind and solar resources into a single portfolio tends to reduce
hourly and sub-hourly variations in real-time output. This would result in a more consistent level of output
over a longer time frame, which could reduce the cost of wind integration (Hurlbut 2009). For additional
analysis of spatial diversity, see also Palmintier et al. 2008.
37
In 2008, more than 8.5 GW of wind was installed in the United States, reaching a total capacity of about
25.3 GW by the end of the year (AWEA 2009). In the same year, 0.3 GW of solar PV was installed,
reaching a total capacity of about 1 GW (Sherwood 2009).
38
The impacts of CSP are largely unquantified as well. However, it is expected that the impact of CSP on
short time scales will be significantly less than PV, because CSP has significant thermal inertia in the
system that will minimize high-frequency ramping events. Furthermore, CSP has the potential advantage of
utilizing high-efficiency thermal storage discussed in Section 5.

21
nonoptimal mix of generators was needed to provide the ability to adjust output in
response to contingencies and other variations in demand. This combination of higher
fuel costs and lower-efficiency units results in increased cost of fuel per unit of electricity
generated as opposed to the no-wind cases. 39 Additional fuel costs occur from keeping
units at part-load, ready to respond to the variability, or from more frequent unit starts.
The costs associated with unit commitment or scheduling are also largely captured in
increased fuel costs. Most large thermal generators must be scheduled several hours (or
even days) in advance to be ready when needed. Ideally, utilities schedule and operate
only as many plants as needed to meet energy and reserve requirements at each moment
in time. If utilities over-schedule (turn on too many plants), they will have many plants
running at part-load, and have incurred higher than needed start-up costs. So, if they
under-predict the wind, they will commit and start up too many plants, which incurs
greater fuel (and other) costs than needed if the wind and corresponding net load had
been forecasted accurately. Conversely, if the wind forecast is too high and the net load is
higher than expected, insufficient thermal generation may be committed to cover the
unexpected shortfall in capacity. A worst-case scenario would be a partial blackout; but
the likely result is the use of high-cost quick-start units in real-time, purchasing
expensive energy from neighboring utilities (if available), or paying customers a
premium to curtail load all while lower cost units are sitting idle. An example of this is
the ERCOT event of Feb. 26, 2008 (Ela and Kirby 2008). On this date, a combination of
events including a greater than predicted demand for energy, a forced outage of a
conventional unit, the wind forecast not being given to the system operators, and a lower
than expected wind output resulted in too little capacity online to meet load. As a
result, the ERCOT system needed to deploy high-cost quick-start units, as well as pay
customers to curtail load through its load acting as a resource program. 40 All of this
occurred while lower-cost units were idle because the combination of events was
unanticipated. 41

39
It should be noted that these increased costs are associated with the residual part of the system that
provides the load not met by VG. This is often a source of confusion and is sometimes interpreted as an
increase in fuel use and emissions of the total system implying that the additional reserve requirements of
VG somehow actually increase fuel use and emissions of the system, or that VG has a net negative impact
on emissions. This is not the case, and integration studies have universally concluded that any increase in
fuel use associated with reserves for VG is much smaller than overall avoided fuel and emissions from
displaced conventional generation.
40
Customers with interruptible loads that can meet certain performance requirements may be qualified to
provide operating reserves under the Load Acting as a Resource (LaaR) program. In eligible ancillary
services (AS) markets, the value of the LaaR load reduction is equal to that of an increase in generation by
a generating plant. See http://www.ercot.com/services/programs/load/laar/
41
This issue has important implications for the use of storage or any other device used to mitigate
uncertainty. Energy storage, like any other generator, needs to be scheduled a storage device used for
load leveling may not be able to simultaneously provide hedging against under-forecasted wind, because it
may already be discharging.

22
While the bulk of the costs associated with wind integration are due to fuel use, the
increased cycling also increases wear and tear on generators, which imposes extra
maintenance costs. 42

Results from wind integration studies almost universally come to the conclusion that at
the penetrations studied to date (up to about 30% on an energy basis), the analyzed
systems do not need additional energy storage to accommodate winds variability and
maintain reliable service. 43 In the studied systems, no new generation technologies are
required; but there are some potentially significant operational changes needed to
maintain the present level of reliability (along with significant transmission additions
needed to exchange resources over larger areas). 44 However, they do not necessarily find
the cost optimal solution, which may include energy storage or some alternative mix of
generation to further reduce the cost of wind integration. Furthermore, these studies have
not evaluated much higher penetration levels of RE, where additional system constraints
may require additional enabling technologies such as energy storage.

4.2 Limiting Factors for Integration of Wind and Solar Energy

To date, integration studies in the United States have found that variable generation
sources can be incorporated into the grid by changing operational practices to address the
increased ramping requirements over various timescales. At higher penetrations (beyond
those already studied), the required ramp ranges will increase, which adds additional
costs and the need for fast-responding generation resources. However, there are
additional constraints on the system that will present additional challenges. These
constraints are based on the simple coincidence of renewable energy supply and demand
for electricity, combined with the operational limits on generators providing baseload
power and operating reserves. Of the four operational cost impacts listed in the beginning
of Section 4 (regulation, load following, scheduling, and ramping range), only the first
three present major quantifiable costs in U.S. studies as of the end of 2009. Yet, it is the
fourth constraint that may present an economic upper limit on variable renewable
penetration without the use of enabling technologies. 45

As discussed in Section 1, in current electric power systems, electricity is generated by

two general types of generators: baseload generators, which run at nearly constant output;
and load-following units (including both intermediate load and peaking plants), which
meet the variation in demand as well as provide operating reserves. At current
penetrations of wind and solar in the United States, and at the levels studied in most
integration studies, wind and solar generation primarily displaces flexible load-following
42
A number of utilities have expressed the opinion that these costs are not well-captured in previous wind
integration studies. This issue is discussed later in this report.
43
This also explains why no significant new storage has been developed in the United States or Europe
despite the 25 GW and 65 GW of wind development, respectively, as of the end of 2008 (EWEA 2009).
44
The more recent U.S. studies of very high penetration (the Western Wind and Solar Integration Study
and the Eastern Wind Integration Study) require power and energy exchanges over larger areas than
typically occur in the existing system (Milligan et al. 2009a).
45
One additional challenge in a high-VG grid is the potential decrease in mechanical inertia that helps
maintain system frequency. This concern is not well understood and could be mitigated by a variety of
technologies including improved controls on wind generators, or other sources of real or virtual inertia that
could include energy storage. See, for example, Doherty et al. (forthcoming).

23
generators. Figures 4.2 and 4.3 illustrate this issue by providing the impacts of increasing
amounts of wind generation in a simulated grid. 46 In Figure 4.2, wind provides 8.5 % of
the energy in this four-day period and displaces the output from a mix of gas-fired units,
which are already typically used to follow load. Because these generators are designed to
vary output, they can do this with modest cost penalties as analyzed in the wind
integration studies discussed previously.

Load Net Load Load Met by Inflexible Generation Wind

70000

60000

50000

40000
MW

30000

20000

10000

0
0 24 48 72 96
Hours

Figure 4.2. Dispatch with low VG penetration (wind providing 8.5% of load)

At higher penetration of RE, the ability of conventional generators to reduce output

becomes an increasing concern. In Figure 4.3, wind now provides 16% of the total
demand in this four-day period. Variable renewables begin to displace units that are
traditionally not cycled, and the ability of these thermal generators to reduce output may
become constrained. If the baseload generators cannot reduce output (and some other use
cannot be found for this excess generation), 47 then wind energy will need to be
curtailed this occurs in the overnight periods in the first two days of this scenario.

46
This simulation uses historical load data from ERCOT from 2005 and simulated wind data for the same
year provided by AWS Truewinds. While the wind and load data is from ERCOT, the mix of generators
(flexible and inflexible) is hypothetical and used only to illustrate the impact of VG. For additional
discussion of the wind data, see GE Energy 2008.
47
The alternative to curtailment is finding some alternative use for this energy through enabling techniques
and technologies, which may include energy storage. These options are discussed in Section 5.

24
 Wind Net Load Met by Inflexible Generation

70000 Curtailed Wind Net Load Met by Flexible Generation

Load
60000

50000

40000
MW

30000

20000

10000

0
0 24 48 72 96
Hours

Figure 4.3. Dispatch with higher VG penetration (wind providing 16% of load)
Utilities in the United States have expressed concern about their systems bottoming out
due to the minimum generation requirements during overnight hours, and being unable to
accommodate more variable generation during these periods. Minimum generation
constraints (and resulting wind curtailment) are already a real occurrence in the Danish
power system, which has a large installed base of wind generation (Ackermann et al.
2009). Due to its reliance on combined heat and power electricity plants for district
heating, the Danish system needs to keep many of its power plants running for heat.
Large demand for heat sometimes occurs during cold, windy evenings, when electricity
demand is low and wind generation is high. This combination sometimes results in an
oversupply of generation, which forces curtailment of wind energy production. The need
to curtail wind due to minimum load constraints has also been identified as an important
component of future power systems in the United States. 48 Modern wind turbines can
reliably curtail output, but this is largely undesirable because curtailment throws away
cost-free and emissions-free energy. 49

The actual minimum load is a function of several factors including the mix of
conventional generation, as well as the amount of reserves and the types of generators
providing those reserves. The ability to cycle conventional units is both a technical and
economic issue there are technical limits to how much power plants of all types can be

48
During over-generation periods, when dispatchable generation plants are already operating at their
minimum levels, the California ISO needs to have an ability to curtail wind generation on an as-needed
basis. (CAISO 2007)
49
When curtailing output, wind or other VG can supply operating reserves. In some cases, the value of
curtailed energy may actually exceed the value of energy, but the primary value of VG is displacing
conventional generation. While Figure 4.3 curtails renewable generation, it not clear which plant should
be curtailed. Ignoring operational constraints, from a strict economic sense, VG has a lower cost of energy
and lower emissions; therefore, it should be curtailed last.

25
turned down. Large coal plants are often restricted to operating in the range of 50-100%
of full capacity, but there is significant uncertainty about this limit. This is partly because
utilities often have limited experience with cycling large coal plants, and have expressed
concern about potentially excessive maintenance impacts 50 and even safety. It should be
noted that because cycling costs are not universally captured in operational models, they
may be ignored or underestimated in wind and solar integration studies. 51 The impact of
VG on power plant cycling is an active area of research, especially considering the
evolving grid and introduction of more responsive generation. 52

It is unclear what changes in operation practices or generator modification utilities will

need to make to cycle below their current minimum load points, which now typically
occur during the early morning in spring. In a number of markets, energy prices have
dropped below the actual variable (fuel) cost of producing electricity on a number of
occasions. This indicates that power plant operators are willing to sell energy at a loss to
avoid further reducing output. Figure 4.4 provides one example in the PJM market in
2002, where the price of electricity fell below the variable cost of generation (indicated
by the dotted line) 53 from coal-fired units for about 100 hours during periods near the
annual minimum. While not definitive, this indicates that under current operational
practices, utilities in some systems may be uncomfortable or unable to cycle much below
current minimum load levels. 54 In other locations, there may be a greater operating range
if plant operators become more comfortable with cycling individual units. As an example,
a recent wind integration study of the existing ERCOT system suggested the capability of
the system to cycle down to a net load of about 13 GW, compared to the recent annual
minimum loads of about 20 GW. However, this would require the coal fleet to cycle to
below 50% of rated capacity, and gas plants to perform over an even greater cycling
range (GE Energy 2008).

50
Cycling operations, that include on/off startup/shutdown operations, on-load cycling, and high
frequency MW changes for automatic generation control (AGC), can be very damaging to power
generation equipment. However, these costs can be very difficult to quantify, especially isolating the
additional costs associated with cycling above and beyond normal operations (Lefton et al. 2006).
51
Wind integration studies typically use proprietary software and data sets, and do not always state which
costs are and are not included. However, in the Western Wind and Solar Integration Study (the highest
penetration U.S. integration study as of 2009) the study states: Wear and tear costs due to increased or
harder cycling of units were not taken into account because these have not been adequately quantified.
(Lew et al. 2009).
52
A more detailed analysis of the relationship between wind penetration and plant cycling is provided by
Troy et al. (forthcoming).
53
During this year, the average price paid for coal in this region was about $31/ton. With a typical heat rate
of about 10,500 BTU, this translates into a variable cost of $12-14/MWh. This cost can be observed in
Figure 4.4 as the floor cluster of points at this level. Points below this level represent bids that are less
than the cost of generation, but not necessarily uneconomic if the alternative is excessive cycling-induced
maintenance or even a forced shutdown and very expensive restart of a coal generator.
54
Minimum load points would be less of a constraint if power plants could be quickly shut down and
started up at low costs. With the exception of certain peaking plants such as aeroderivative turbines and
fast-starting reciprocating engines, most plants have minimum up-and-down times, and require several
hours to restart (at considerable cost).

26
 250

Wholesale Price ($/MWh) 200

150

100

50

0
0 10000 20000 30000 40000 50000 60000
Load (MW)
35

30

Wholesale Price ($/MWh)

25

20

15

10

0
18000 20000 22000 24000 26000
Load (MW)

Figure 4.4. Relationship between price and load in PJM in 2002

Overall, the ability to accommodate a variable and uncertain net load has been described
as a systems flexibility. System flexibility varies by system and over time as new
technologies are developed and power plants are retired. In addition, VG deployment in
the United States is still relatively small, and utilities have yet to evaluate the true cycling
limits on conventional generators and their associated costs. Also, the additional reserve
requirements due to VG at high penetration are still uncertain. As a result, it is not
possible to precisely estimate the costs of VG integration or the amount of curtailment at
very high penetration; it is also difficult to define with certainty the value of energy
storage or other enabling technologies. It is clear, however, that substantial increase in the
penetration of wind energy without storage will require changes in grid operation to
reduce curtailment.

Figure 4.5 illustrates the fraction of a systems peak load derived from inflexible units as
a function of the minimum load point. This is also expressed more generally as the
systems flexibility factor, which is defined as the fraction below the annual peak to
which conventional generators can cycle (Denholm and Margolis 2007a and 2007b). In
the current grid, the annual minimum load point is typically 30%-40% of annual peak
load (or a flexibility factor of 60%-70%). If units cannot be cycled below this point, they
will provide 55%-70% of a systems energy; and even with storage, this level of
inflexibility leaves only 30%-40% of a systems energy for variable resources. One of the
major conclusions of wind integration studies looking at higher penetrations is that

27
minimum load points will need to be lowered substantially below their current annual
minimums. 55

Flexibility Factor
100% 90% 80% 70% 60% 50%
100%
% of Energy From Inflexible

Typical
90%
Minimum Load
80%
70%
Generators

60%
50%
40%
30%
20%
10%
0%
0% 10% 20% 30% 40% 50%
Minimum Load as a Fraction of Peak Load

Figure 4.5. Contribution from inflexible generation

Alternatively, it is possible to evaluate the relationship between system flexibility and
large-scale penetration of VG, both with and without energy storage or other enabling
technologies. Figure 4.6 shows an example of how flexibility affects the potential
curtailment of wind and solar energy when deployed without storage. These two charts
superimpose load data in ERCOT from 2005 with a spatially diverse set of simulated
wind and solar data from the same year. 56,57 The simulation places 19 GW of wind and
11 GW of solar (representing an 80%/20% mix of wind and solar on an energy basis) into
the 2005 ERCOT system, which had a peak demand of 60.3 GW. In the left chart, the
system is assumed to be unable to cycle below the 2005 minimum point of 21 GW,
resulting in substantial VG curtailment. In this simulation, wind and solar provide 20% of
the grids annual energy, and 21% of the total renewable energy production is curtailed.
The right graph shows the result of increasing flexibility, allowing for a minimum load
point of 13 GW. Curtailment has been reduced to less than 3%, and the same amount of
variable renewables now provides about 25% of the systems annual energy.

55
This is noted in previous wind integration studies such as the ERCOT study (GE 2008), and the more
recent Eastern and Western Interconnect studies. As the penetration of wind and solar increase, the impact
on base-load coal increases, becoming very challenging at the 30% penetration. (Milligan et al. 2009b).
56
Historically, many estimates of the limits of wind penetration have used data from a single or very small
set of wind power plants, and often a small balancing area. Without spatial diversity of resource and load,
this leads to both excessive ramp rates and excessive curtailment.
57
The data set is the same as from Figures 4.2 and 4.3, with solar data derived from the National Solar
Radiation Database. See Denholm and Margolis 2008 for additional details. The data was processed using
the REFlex Model, which compares hourly electricity demand with VG supply and calculates curtailments
as a function of system flexibility based on minimum load constraints. The model is described in more
detail in Denholm and Margolis 2007a and 2007b.

28
 Load Met by Renewable Energy Net Load Met by Flexible Generation
50000 Curtailed Renewable Energy Net Load Met by Inflexible Generation
45000 Renewable Energy
40000
35000
30000

Inflexible System -
MW

25000
20000 Minimum Load of
15000 21 GW
10000
5000
0 Decreased
0 24 48 72 96 curtailment
Hours

Net Load Met by Inflexible Generation Curtailed Renewable Energy

50000 Net Load Met by Flexible Generation Load Met by Renewable Energy
Renewable Energy
45000
40000
35000
30000
MW

25000
More Flexible 20000
System - 15000
Minimum Load 10000
of 13 GW 5000
0
0 24 48 72 96
Hours

Figure 4.6. Effect of decreasing minimum load point on increased use of RE

These results can be evaluated more generally by examining the amount of curtailment
that will result without energy storage as a function of system flexibility. Figure 4.7
shows example results, with three different minimum load/flexibility factors. The first
curve illustrates the curtailment that would result if the system could not cycle below its
2005 annual minimum load of 21 GW. The second curve uses a minimum load of 12
GW, which corresponds roughly to the assumptions used in the 2008 ERCOT wind study
(GE 2008). The third curve corresponds to a 6 GW minimum load, which largely
eliminates baseload units from the generation mix, replacing them with more flexible
units. 58

58
It should be noted that significant changes in the generation mix and corresponding changes in system
flexibility could result from CO2 emissions constraints. Increased cost of carbon could motivate greater use
of natural gas generation, and reduce use of coal, both of which would tend to increase system flexibility,
and allow greater economic use of VG. This, in turn, would decrease fuel use and the capacity factor of
thermal generators, which would also tend to increase the use of gas-fired generation and decrease the use
of coal as the economic optimal mix of generation. For additional discussion, see Lamont 2008.

29
 50% 21 GW Min Load
(65% FF)
45% 12 GW Min Load
(80% FF)
40% 6 GW Min Load
(90% FF)

Average VG Curtailment
35%
FF=Flexibility Factor
30%

25%

20%

15%

10%

5%

0%
0% 10% 20% 30% 40% 50% 60% 70% 80%
Fraction of Energy From RE

Figure 4.7. Average curtailment rate as a function of VG penetration for different

flexibilities in ERCOT
Figure 4.7 somewhat obscures the fact that at the margin, curtailment rates can be very
high. Figure 4.8 illustrates the marginal curtailment rate, or the curtailment rate of each
incremental unit of VG installed in the system. For example, in the 12 GW minimum
load curve, the average curtailment rate (Figure 4.7) when VG is providing 25% of the
systems electricity demand is less than 3%, which means that less than 3% of all the VG
at this penetration level is curtailed. However, the last unit of energy installed to get to
this 25% point has a curtailment rate of more than 10%, as illustrated in Figure 4.8.

50%

45%

40%
Marginal VG Curtailment

35%

30%
21 GW Min Load
25% (65% FF)

20% 12 GW Min Load

(80% FF)
15%
6 GW Min Load
10% (90% FF)

5% FF=Flexibility Factor

0%
0% 10% 20% 30% 40% 50% 60% 70% 80%
Fraction of Energy From RE

Figure 4.8. Marginal curtailment rate as a function of VG penetration for different system
flexibilities in ERCOT (marginal curtailment is defined as the percentage curtailed as a
function of the incremental penetration)

30
Curtailment increases the cost of the VG that is actually used, because curtailment
reduces the net capacity factor of the wind and solar generators. Figure 4.9 shows how
the effective cost of VG increases due to curtailment. Costs are illustrated in relative
terms a generator with no curtailment has a base cost of 1, which increases with a scale
factor equal to (1/1-curtailment rate).

2.0
21 GW Min Load
1.9 (65% FF)

1.8 12 GW Min Load

(80% FF)
Relative Cost of VG (Average)

1.7 6 GW Min Load

(90% FF)
1.6

1.5

1.4

1.3

1.2

1.1

1.0
0% 10% 20% 30% 40% 50% 60% 70% 80%
Fraction of Energy From Wind and Solar

2.0

1.9

1.8
Relative Cost of VG (Marginall)

1.7

1.6

1.5

1.4 21 GW Min Load

(65% FF)
1.3 12 GW Min Load
(80% FF)
1.2
6 GW Min Load
1.1 (90% FF)

1.0
0% 10% 20% 30% 40% 50% 60% 70% 80%
Fraction of Energy From Wind and Solar

Figure 4.9. Relative cost of VG average (top chart) and marginal (bottom chart) as a
function of VG penetration for different system flexibilities in ERCOT
This example can be compared to the results of the previous U.S. wind integration studies
that find very little curtailment up to about 20% on an energy basis, due to the existing
grid flexibility. Beyond this level (up to about 30%), curtailment increases without
significant operational changes (Corbus et al. 2009). In the United States, penetrations of
VG beyond 30% have yet to be extensively studied; however, the examples in this

31
section suggest curtailment rates will continue to rise substantially without a significant
increase in system flexibility or deployment of other enabling technologies such as
energy storage.

The results in Figure 4.9 apply to only a specific mix of VG and a single system so
they cannot be applied generally. However, they do illustrate the trends that may limit the
contribution of VG without enabling technologies. Ultimately, the cost of curtailment
must be compared to alternatives such as the cost of storage or other enabling
technologies, illustrated in Figure 4.10 and discussed in more detail in Section 5. 59

Figure 4.10. Option for increasing the use of VG by decreasing curtailment

Storage provides one solution to avoiding curtailment by absorbing otherwise unusable
generation and moving it to times of high net system load (where net load is defined as
normal load minus VG). The additional flexibility storage provides is also important.
Storage can provide operating reserves, which reduces the need for partially loaded
thermal generators that may restrict the contribution of VG. Finally, by providing firm
capacity and energy derived from VG sources, storage can effectively replace baseload
generation, which reduces the minimum loading limitations.

Figure 4.11 illustrates how the curtailment rate can be reduced by introducing energy
storage. 60 The base case (no storage) is identical to Figure 4.7 with a 12 GW minimum

59
The large number of options available for increasing the penetration of RE is one of the reasons why the
question of when storage becomes necessary is very difficult to answer. There is less of a technical limit
than an economic one that depends on a large number of factors such as the cost of storage compared to a
vast array of alternatives.
60
The data and methods used for this analysis are the same as before, using the REFlex model to place
otherwise curtailed energy into storage and using that stored energy at a later time (Denholm and Margolis
2007b).

32
load (or an 80% flexibility factor). In addition, a storage device with 5% of the systems
power capacity (or 3 GW in a 60 GW peaking system), 20 hours of energy capacity, and
a 75% round-trip efficiency is introduced. If this device is used only to absorb otherwise
unusable energy, the curtailment rate is reduced substantially. For example, when VG is
providing 50% of the systems energy, about 30% of the VG is curtailed without storage,
and about 25% with the 3 GW of storage. This includes the additional losses that occur in
the storage process. The devices ability to replace firm capacity and potentially reduce
the minimum load constraint further reduces curtailment. Adding the ability to reduce the
minimum load by 50% of the devices capacity (or about 1.5 GW) reduces curtailment
further in the 50% VG case, the curtailment rate now drops from 30% without storage
to 20% of the total VG. 61

50%
No Storage
45%

40% 5% Storage Capacity

Average RE Curtailment

35%
5% Storage Including
Minimum Load Reduction
30%

25%

20%

15%

10%

5%

0%
0% 10% 20% 30% 40% 50% 60% 70%
Fraction of Energy From Wind and Solar

Figure 4.11. Reduction of curtailment resulting from addition of energy storage

While storage provides one solution to the mismatch of VG supply and normal demand,
there are a variety of options for increasing the use of VG in the grid. Any evaluation of
energy storage should consider the many alternative technologies that can increase grid
flexibility and enable VG renewables.

61
Alternatively, for a given amount of allowed curtailment, the contribution of VG increases with the
addition of storage. In the case illustrated in Figure 4.11, a maximum VG curtailment of 10% allows VG to
provide about 35% of the total electricity demand. Adding energy storage (with its ability to reduce the
minimum load) increases the contribution of VG (for the same amount of allowable curtailment) to about
42% of total demand.

33
5 Storage and Flexibility Options for Renewables-Driven
Grid Applications
The previous section indicates that at high penetration of VG, fundamental changes to the
grid may be required to accommodate the increased variability of net load and the limited
coincidence of VG supply and normal electricity demand. A number of techniques and
technologies described as flexibility resources have been proposed to accommodate
the impacts of VG and ensure the generation mix matches the net-load requirement.

5.1 The Flexibility Supply Curve

Energy storage is one of many technologies proposed to increase grid flexibility and
enable greater use of VG. This set of technologies has been described in terms of a
flexibility supply curve that can provide responsive energy over various timescales. The
flexibility supply curve is conceptually similar to other resource supply curves where
(ideally) the lowest-cost resources are used until they are exhausted, then the next
(higher-cost) resource is deployed. The analysis in Section 4, for example, was restricted
to ERCOT, and did not consider the opportunity to exchange wind and solar energy with
surrounding areas by building interconnections with the other U.S. grids. It also did not
consider the ability to shift load by incentivizing customers to use less electricity when
VG output is low. Overall, utilities have many flexibility options for incorporating
greater amounts of VG into the grid, many of which may cost less than using energy
storage. Figure 5.1 provides a conceptual supply curve including some of these options.

Figure 5.1. Flexibility supply curve 62

62
Based on an original by Nickell 2008. While the figure hypothesizes an order of the supply curve, the
actual costs and availability of the individual components are conjectural.

34
Overall, there are two general types of flexibility required by variable sources and
offered by technologies in this curve. The first can be described as ramping flexibility, or
the ability to follow the variation in net load (in the second-to-minute timescale needed
for frequency regulation, or the minutes-to-hours timescale needed for load following and
forecast error.) This flexibility is the primary requirement at low penetration, as discussed
in Section 4.2. The second type of flexibility is energy flexibility, or the ability to
increase the coincidence of VG supply with demand for electricity services, which is
described in Section 4.3. A description of several of the sources of flexibility is provided
below.

a.) Supply and Reserve Sharing. This includes the sharing of renewable and
conventional supply, operating reserves, and net loads through markets or other
mechanisms that effectively increase the area over which supply and demand is
balanced. 63 Greater aggregation of loads and reserves has historically been one
of the least-cost methods of dealing with demand variability, especially because
it often requires operational changes and relatively little new physical
infrastructure. 64 It may also require transmission development to increase the
spatial diversity of VR resources.
b.) Flexible Generation. This includes deploying new, more flexible conventional
generators as well as increasing flexibility of existing generators. This can be
accomplished by modifying equipment and operational practices 65 to increase
the load-following, ramping rate, and ramping range of the grid. This also
includes introducing new generators that can be brought online quickly to
respond to forecast errors. 66 This may also require increased use of natural gas
storage to increase use of flexible gas turbines and decrease contractual
penalties for forecast errors in natural gas use (Zavadil 2006). Another source of
flexible generation is improved use of existing storage and hydro assets.
c.) Demand Flexibility. This includes introducing market or other mechanisms to
allow a greater fraction of the load to respond to price variations and provide
ancillary services. Responsive demand can provide flexibility over multiple
timescales by curtailing demand for short periods or shifting load over several
hours. Many of these technologies and processes have been described in terms

63
Larger markets and balancing areas that are a central feature of ISOs and RTOs can improve the
physical conditions needed to integrate large amounts of wind energy. ISOs and RTOs, with their day-
ahead and real-time markets, large geographies to aggregate diverse wind resources, large loads to
aggregate with wind, large generation pools that tap conventional generator flexibility offer the best
environments for wind generation to develop. (Milligan et al. 2009b).
64
This includes introducing sub-hourly markets that allow systems faster response to variability. Large
ISOs with 5-minute markets typically have substantially lower wind integration costs.
65
This could include scheduling generators over shorter time periods, and using sub-hourly wind forecasts
instead of persistence forecasts that may actually contribute to short-term scheduling errors and increase
regulation requirements.
66
Extensive changes will be required in the type of new generation built in the state: new units must have
greater operating flexibility to start up and shut down without long delays; they must be able to operate at
lower minimum loading levels; and they must have faster ramping capability and regulation capability
(CAISO 2007). Examples of more responsive generators include certain aeroderivative gas turbines and
reciprocating engines. For additional discussion of flexible generators, see Northwest Power and
Conservation Council 2009.

35
 of a smart grid and will require regulatory and policy changes in addition to
new technologies. In many locations in the United States, demand is
increasingly used as a source of grid services. 67
d.) VG Curtailment. Overbuilding VG may result in curtailment of low-value
springtime generation, but would allow for a greater overall VG contribution.
(This is functionally equivalent to cycling baseload generators in the spring.)
Furthermore, curtailed VG provides additional benefits, because it provides a
source of operating reserves and potentially allows for de-commitment of
thermal units that typically provide these services.
e.) New Loads. New controllable loads can be added to absorb otherwise unusable
VG. Examples include space and process heating, which currently use fossil
fuels. Another possibility is fuel production such as hydrogen via electrolysis or
shale oil heating. Electrification of transportation using electric vehicles or plug-
in hybrid vehicles is also a potential large-scale application. This may also
include electric vehicles providing regulation and contingency reserves with or
without the use of vehicle to grid (V2G).
f.) Electricity Storage. Electricity storage encompasses a large number of
technologies discussed in Section 5.2

The general classes of flexibility resources listed above represent dozens or even
hundreds of individual technologies, each with a potential contribution to increasing grid
flexibility. The cost and availability of many flexibility resources has yet to be quantified,
and there are a variety of regulatory barriers to completely deploying many flexibility
options such as demand response.

The cost of storage needs to be compared to the alternatives this includes the efficiency
losses in the storage process that may be avoided by using other enabling technologies.
There are, of course, limits to each option on the curve; and the benefits of spatial
diversity and demand response are limited because the VG supply cannot be expected to
exactly match the demand for electricity services on the multiple timescales. Because of
this, electricity storage is considered a potentially important step on the flexibility supply
curve when lower-cost options are saturated or otherwise unavailable.

5.2 Deployment and Operational Balancing of Renewable Energy

Individual Plant Storage vs. Power System Storage
The previous section suggests that incorporating increased levels of VG most efficiently
will require a variety of flexibility options. Likewise, the most efficient operation and
location of flexibility options including storage must be considered. While flexibility
resources can be considered renewable enabling technologies, their historical application
has been to benefit the grid as a whole. It is often suggested that energy storage be co-
located with, and operationally tied to, the output of individual VG facilities. However,
this is not how the current system balances the large variability in net demand.

67
As an example, ERCOT currently obtains half of its spinning reserve requirements from responsive load,
and the ERCOT event of February 2008 discussed previously is an example of an application of load as a
source of up ramping, used to meet demand until conventional generators could be started and
dispatched.

36
Furthermore, the services needed to address variability and uncertainty are generally the
same services that storage currently provides to the grid. Table 5.1 lists several
renewable-specific applications that have been proposed (EPRI 2004). As an example,
one potential renewable-specific application of storage is time-shifting of wind from
periods of low demand to periods of high demand. However, this application is
fundamentally the same as energy arbitrage, and the benefits of this application are
greatest when the energy storage operator can choose from all of the generators in a
system, and store energy when the cost is lowest, instead of storing only wind generation.

Table 5.1. Dedicated Renewable Applications of Energy Storage and Their Whole-Grid
Counterpart
RE Specific Application Whole Grid Application
Transmission Curtailment Transmission Deferral
Time Shifting Load Leveling/Arbitrage
Forecast Hedging Forecast Error
Frequency Support Frequency Regulation
Fluctuation Suppression Transient Stability

Despite the attractiveness of a smoothed and shifted output from each generator, this
approach results in a significant decrease in the efficiency of the entire system, and
eliminates the benefits of resource aggregation. For example, demand of any individual
electricity consumer can be very irregular, with rapid unpredictable ramps. Distributed
energy storage could be used to smooth these individual demands. However, this would
result in storage devices being charged in one location, while simultaneously discharged
in another, wasting both the cost of storage devices and the losses in the storage process.
The aggregated net demand of many individual consumers is considerably smoother, and
would require much less storage to flatten if desirable. Likewise, operational
integration between energy storage and any individual or groups of VG would be
nonoptimal and likely result in simultaneous charging and discharging. By aggregating
the entire net load of a system, including all loads and VG supply, storage or other
flexibility options can be deployed at the lowest cost and greatest efficiency. 68 This is
especially the case when spatial diversity substantially reduces variability over multiple
timescales.

There are some exceptions when there are benefits of operationally combining VG and
energy storage, typically through co-location and sharing of certain high-cost
components. The best example is integrating thermal storage into a concentrating solar
power (CSP) plant; another example is sharing power electronics in a distributed
PV/battery system.

There are several other applications where co-location of VG and storage may make
sense. Wind plants placed in areas of weak transmission can potentially introduce power
quality and stability issues, and storage can be a mitigating technology; however,
68
In effect, combining individual VG and storage is essentially the creation of very small balancing areas.
This is actually the opposite of how the grid is evolving, with the creation of larger balancing areas and the
use of reserve sharing agreements across utilities. Balancing individual VG would dramatically increase
reserve requirements and would incur much greater costs than at the system level.

37
improved power electronics in modern wind turbines may be a lower-cost alternative.
Finally, combining wind and energy storage has been proposed as an alternative (or
supplement) to developing new transmission capacity. Increased deployment of wind
energy will require substantial new transmission, and storage co-located with remote
wind resources can help decrease the need for new transmission. 69 This has been
proposed to relieve congestion in the ERCOT grid, for example, where the states best
wind resources are located largely in the sparsely populated western part of the state, and
transmission capacity is limited (Desai et al. 2003). Despite these potential applications,
the majority of storage deployed in the grid will likely be a shared resource, which will
benefit the entire system and not just a single generator or load (Smith et al. 2007). Just
as loads are balanced in aggregate, the net load in the future grid after all VG sources
are included will be balanced by a mix of conventional generation, plus flexibility
options that include energy storage.

5.3 Energy Storage Technologies and Applications

This section provides a brief overview of commercially available energy storage
technologies. It is not intended to be a comprehensive discussion, because there are a
large number of sources available that discuss the technical performance, current
applications, vendors, and costs in detail. A number of more comprehensive reviews of
energy storage technologies are provided in the Bibliography.

The choice of an energy storage device depends on its application in either the current
grid or in the renewables/VG-driven grid; these applications are largely determined by
the length of discharge. Energy storage applications are often divided into three
categories, based on the length of discharge. Table 5.2 indicates the three regimes of
energy storage applications commonly discussed.

Table 5.2. Three Classes of Energy Storage

Common Name Example Applications Discharge Time Required
Power Quality Transient Stability, Frequency Seconds to Minutes
Regulation
Bridging Power Contingency Reserves, Minutes to ~1 hour
Ramping
Energy Management Load Leveling, Firm Capacity, Hours
T&D Deferral

The first two categories of energy storage applications in Table 5.2 correspond to a range
of ramping and ancillary services, but do not typically require continuous discharge for
extended periods of time. In the case of renewables-driven applications, this could require
discharge times of up to about an hour to allow fast-start thermal generators to come
online in response to forecast errors. (Bridging power typically refers to the ability of a

69
Use of dedicated long-distance transmission for wind or solar will be limited by the relatively low
capacity factor of the resource. Storage could increase line-loading and help reduce curtailment due to
transmission constraints. For additional discussion, see Denholm and Sioshansi 2009.

38
storage device to bridge the gap from one energy source to another.) 70 The third
category (energy management) corresponds to energy flexibility, or the ability to shift
bulk energy over periods of several hours or more.

The references in the Bibliography provide a number of assessments and charts with
estimates of technical performance and costs. Figure 5.2 provides one example of the
range of technologies available for these three classes of services, and shows that many
technologies can provide services across various timescales.

Energy
Management

Bridging
Power

Power
Quality

Figure 5.2. Energy storage applications and technologies 71

70
This often refers to the time it takes after a power failure for an isolated system to switch from the grid to
a backup generator. However, this term may also be useful to describe the ability to address forecast errors
and bring up standby generators during times of unforeseen decreases in wind or conventional generation.
71
This chart represents technologies actually deployed or proposed as of November 2008. It does not
include a number of pre-commercial products or represent the total range of applications. For example,
most of the batteries listed could be scaled up in either energy or power capacity, while at least 1 CAES
plant of greater than 1,000 MW has been proposed. Alternatively, PSH plants of less than 50 MW have
been constructed (ESA 2009).

39
It should be noted that this chart does not include thermal energy storage, which would
cover a power range of a few kilowatts (kW) for thermal energy storage in buildings to
more than 100 MW in CSP plants, with a discharge time of minutes to several hours.

When considering the technical performance or costs of energy storage, there are a
number of caveats to consider. The first is the technical and commercial maturity of the
storage technology. As of 2009, only four energy storage technologies (sodium-sulfur
batteries, pumped hydro, CAES, and thermal storage) 72 have a total worldwide installed
capacity that exceeds 100 MW. 73 This doesnt mean that there isnt market potential for
any individual technology or storage, in general; but it makes it difficult to assess the
state of any individual technology given its limited deployment to date. This also leads to
some uncertainty in two of the primary performance indicators for energy storage
devices: efficiency and cost. Both of these values are often imprecisely reported, which
makes it difficult to perform an accurate assessment of the potential for individual
technologies or compare different technologies. Some major caveats when considering
electricity storage include:

Efficiency
1) The standard measure of an electricity storage devices efficiency in the grid is
the AC to AC round-trip efficiency, or AC kWhout/kWhin. 74 However, this is not
always the value reported, especially for devices that store DC energy such as
batteries and capacitors. In some cases, the DC-DC round-trip efficiency may be
reported, and additional losses in power conversion efficiencies must be
considered if the device is to provide applications in the grid.
2) Reported round-trip efficiencies may not include parasitic loads. These include
heating and cooling of batteries and power-conditioning equipment. These
parasitic loads can vary considerably depending on use, climate, and the length of
each storage cycle.
3) The round-trip efficiency of several technologies cannot be directly compared.
Thermal storage provides some, but not all of the services of a pure electricity
storage device; while compressed-air energy storage is a hybrid device that
requires both electricity and natural gas. These factors limit the value of a direct
comparison, as discussed in more detail later in this section.

Cost
1) As stated before, only a few storage technologies have been deployed at large
scale (greater than 100 MW). Estimated prices for emerging technologies may be
for a semi-custom product (and consequently very high) or projected costs based
on mass production (and perhaps overly optimistic). Even with more mature

72
This excludes small distributed applications including uninterruptible power supplies, off-grid homes,
and the substation batteries. These applications are dominated by lead-acid batteries.
73
100 MW is equivalent to a small power plant and negligible in terms of overall grid-scale capacity. As
of January 2008, the United States had 1,087,791 MW of installed capacity (EIA 2009a).
74
The U.S. electric grid uses alternative current (AC) while batteries, capacitors, and several other electric
storage technologies charge and discharge direct current (DC).

40
 technologies such as PHS and CAES, it has been some time since either has been
built in the United States, so the cost of the next plant is somewhat uncertain. 75
2) As with any generation technology, large variations in prices occur from year to
year due to commodity prices and the global economy. Therefore, cost estimates
of storage technologies from different years may reflect market conditions as
opposed to real differences.
3) Storage technologies offer different classes of services and are comprised of an
energy component and power component. The total cost of a storage device
includes both components, with the limits of the target application. As a result, a
direct comparison of a PHS device with a flywheel, for example, has limited
value.

5.3.1 Storage Technologies for Power Quality Applications

Power quality applications require rapid response often within less than a second and
include transient stability and frequency regulation. As with the other applications, the
timescales of discharge may vary; but this class of services typically requires discharge
times of up to about 10 minutes and nearly continuous cycling. Technologies for these
applications include flywheels, capacitors, and superconducting magnetic energy storage
(SMES).

Flywheels
Flywheels store energy in a rotating mass. Flywheels feature rapid response and high
efficiency, making them well-suited for frequency regulation. Several flywheel
installations have been planned or deployed to take advantage of high prices in frequency
regulation markets (Lazarewicz 2009).

Capacitors
Capacitors 76 store electricity in an electric charge. Capacitors have among the fastest
response time of any energy storage device, and are typically used in power quality
applications such as providing transient voltage stability. However, their low energy
capacity has restricted their use in longer time-duration applications. A major research
goal is to increase their energy density and increase their usefulness in the grid (and
potentially in vehicle applications.)

Superconducting Magnetic Energy Storage (SMES)

SMES stores energy in a magnetic field in a coil of superconducting material. SMES is
similar to capacitors in its ability to respond extremely fast, but it is limited by the total
energy capacity. This has also restricted SMES to power applications with extremely
short discharge times. Several demonstration projects have been deployed.

75
Furthermore, PHS and CAES depend on site-specific geologic conditions, which makes costs difficult to
generalize.
76
These devices have several names such as ultracapacitors and supercapacitors. There is some
uncertainty within the industry on the exact name for capacitors with massive storage capability. This is in
part due to the many names of products by different manufacturers, but also due to the relative newness of
the industry and recent advances. (EPRI 2003).

41
5.3.2 Storage Technologies for Bridging Power
Bridging power applications include providing contingency reserves, load following, and
additional reserves for issues such as forecast uncertainty and unit commitment errors.
This set of applications generally requires rapid response (in seconds to minutes) and
discharge times in the range of up to about an hour. Far less cycling is required than for
power quality applications.

This application is generally associated with several battery technologies, which include
lead-acid, nickel-cadmium, nickel-metal hydride, and (more recently) lithium-ion. Due to
their rapid response, they can provide power quality services such as frequency
regulation; but the continuous cycling requirement can limit battery life. Several
demonstration projects have been built using these technologies to provide operating
reserves.

5.3.3 Storage Technologies for Energy Management

Energy management applications include moving power over longer timescales, and
generally require continuous discharge ratings of several hours or more. Technologies for
these applications include several battery types, pumped hydro, compressed air, and
thermal energy storage.

High-Energy Batteries
For many batteries, there is considerable overlap between energy management and the
shorter-term applications discussed previously. Furthermore, batteries can generally
provide rapid response, which means that batteries designed for energy management
can potentially provide services over all the applications and timescales discussed.

Several battery technologies have been demonstrated or deployed for energy management
applications. In addition to the chemistries discussed previously, the commercially
available batteries targeted to energy management include two general types: high-
temperature batteries and liquid electrolyte flow batteries

The most mature high-temperature battery as of 2009 is the sodium-sulfur battery, which
has worldwide installations that exceed 270 MW (Rastler 2008). Alternative high-
temperature chemistries have been proposed and are in various stages of development
and commercialization. One example is the sodium-nickel chloride (ZEBRA) battery.

The second class of high-energy batteries is the liquid electrolyte flow battery. This
battery uses a liquid electrolyte that flows across a membrane. The advantage of this
technology is that the power component and energy component can be sized
independently. As of 2009, there has been limited deployment of two types of flow
batteries vanadium redox and zinc-bromine. Other combinations such as polysulfide-
bromine have been pursed, and new chemistries are under development.

In the United States, a primary application of energy management batteries has been
T&D deferral; however, demonstration projects have been deployed for multiple
applications (Nourai 2007, EPRI 2003).

42
Pumped Hydro Storage (PHS)
Pumped hydro is the only energy storage technology deployed on a gigawatt scale in the
United States and worldwide. In the United States, about 20 GW is deployed at 39 sites,
and installations range in capacity from less than 50 MW to 2,100 MW. 77 Many of the
sites store 10 hours or more, making the technology useful for load leveling. PHS is also
used for ancillary services. PHS uses conventional pumps and turbines and requires a
significant amount of land and water for the upper and lower reservoirs. PHS plants can
achieve round-trip efficiencies that exceed 75% and may have capacities that exceed 20
hours of discharge capacity. Environmental regulations may limit large-scale above-
ground PHS development. However, given the high round-trip efficiencies, proven
technology, and low cost compared to most alternatives, conventional PHS is still being
pursued in a number of locations. 78 Alternative lower-impact configurations have been
studied, including using a natural or mined underground formation for the lower
reservoir, but this configuration has yet to be commercialized.

Compressed Air Energy Storage (CAES)

CAES technology is based on conventional gas turbine technology and uses the elastic
potential energy of compressed air. Energy is stored by compressing air in an airtight
underground storage cavern. To extract the stored energy, compressed air is drawn from
the storage vessel, heated, and then expanded through a high-pressure turbine that
captures some of the energy in the compressed air. The air is then mixed with fuel and
combusted, with the exhaust expanded through a low-pressure gas turbine. The turbines
are connected to an electrical generator.

CAES is considered a hybrid generation/storage system because it requires combustion in

the gas turbine. The performance of a CAES plant is based on its energy ratio (energy
in/energy out) and its fuel use (typically expressed as heat rate in BTU/kWh). CAES
performance is estimated at an energy ratio of 0.6-0.8 and a heat rate of 4,000-4,300
BTU/kWh (Succar and Williams 2008). Because CAES uses both electricity and natural
gas, a single-point definition of the round-trip efficiency of a CAES device does not
represent an economic figure of merit.

The primary disadvantages of CAES are the need for an underground cavern and its
reliance on fossil fuels. Alternative configurations for CAES have been proposed using
manufactured above-ground vessels, new turbine designs to reduce fossil fuel use, or
designs that re-use the heat of compression and avoid fuel use altogether.

Thermal Energy Storage

Thermal energy storage is sometimes ignored as an electricity storage technology
because it typically is not used to store and then discharge electricity directly. However,
in some applications, thermal storage can be functionally equivalent to electricity storage.
One example is storing thermal energy from the sun that is later converted into electricity
77
A complete list including capacity, location, date of initial operation, and ownership is available from
EIA Form EIA-860, Annual Electric Generator Report.
78
About 30 GW of new pumped hydro capacity has been proposed between 2006 and 2009. This represents
more than double the existing capacity, and certainly implies there are considerable opportunities for new
pumped hydro capacity (Adamson 2009).

43
in a conventional thermal generator. Another example is converting electricity into a
form of thermal energy that later substitutes for electricity use such as electric cooling or
heating.

The first example (storing thermal energy that is later converted into electricity) can be
used with many types of thermal generators but is most often associated with
concentrating solar power. In this application, thermal energy from the solar field is
stored in molten salt or another medium. This energy can be recovered later and used to
generate electricity, which turns this technology into a dispatchable source of energy.

Care must be used when discussing the efficiency of thermal energy storage. One of the
major issues with electricity storage is efficiency losses. Electricity is a high quality
source of energy, and transforming electricity into a stored medium and back incurs
considerable losses. Thermal energy is a much lower quality of energy, but can be stored
with much higher efficiency. In a CSP plant, thermal energy is stored before conversion
to electricity. As a result, the round-trip efficiency of CSP thermal storage may be close
to 100%, much higher than any electricity storage technology. However, CSP thermal
storage can only store thermal energy produced from the solar field, as opposed to other
storage technologies that can store electricity produced from any source.

Likewise, end-use energy storage can have extremely high round-trip efficiencies.
Demand for electric-power cooling can be shifted by storing cold energy in the form of
chilled water or ice during off-peak times and releasing that cold energy during times of
peak demand. This effectively stores electricity with high round-trip efficiency. 79

End-use hot storage can also be used in both space heating and water heating
applications. (Controllable water heating somewhat blurs the line between energy storage
and demand response.) As with other forms of thermal storage, the effective round-trip
efficiency of end-use hot storage is much higher than pure electricity storage devices
but is limited by daily and seasonal heating demands.

5.4 Electric Vehicles and the Role of Vehicle to Grid

Electric vehicles (EVs used here to represent both pure electric vehicles or plug-in
hybrid electric vehicles) are a potential source of flexibility for VG applications. The
charging of EVs can potentially be controlled, and provide a source of dispatchable
demand and demand response. Controlled charging can be timed to periods of greatest
VG output, while charging rates can be controlled to provide contingency reserves or
frequency regulation reserves. Vehicle to grid (V2G) (where EVs can partially discharge
stored energy to the grid) may provide additional value by acting as a distributed source
of storage. EVs could potentially provide all three grid services discussed previously.
Most proposals for both controlled charging and V2G focus on short-term response
services such as frequency regulation and contingency. Their ability to provide energy

79
Losses in the storage system are relatively small and occur through heat exchange from the stored cold
energy and the surrounding environment. These losses can be partially offset by the potential increase in
compressor efficiency when making ice or chilled water in the cooler evening compared to the daytime,
which achieves net efficiencies close to 100%.

44
services is more limited by both the storage capacity of the battery, as well as the high
cost of battery cycling. This could restrict their ability to provide time-shifting (energy
arbitrage) beyond their ability to perform controlled charging. 80 The role of V2G is an
active area of research, and because electrified vehicles in any form have yet to achieve
significant market penetration, it is difficult to assess their potential as a source of grid
flexibility. However, analysis has demonstrated potential system benefits of both
controlled charging and V2G (Denholm and Short 2006). The role of EVs as an enabling
technology requires additional analysis of their unique temporal characteristics of
availability, unknown battery costs and lifetimes, and the availability of smart charging
stations to maximize their usefulness while parked.

80
This conclusion depends on the anticipated cycle life and cost of EV batteries. See Sioshansi and
Denholm 2009 and Peterson et al. 2010 for a discussion of the impact of battery life and cycling on the
value of V2G. However, controlled charging (without V2G) is still a potentially significant source of
flexibility, with the ability to raise the minimum load and avoid curtailment.

45
6 Conclusions
The increasing role of variable renewable sources (such as wind and solar) in the grid has
prompted concerns about grid reliability and raised the question of how much these
resources can contribute before enabling technologies such as energy storage are needed.
Fundamentally, this question is overly simplistic. In reality, the question is an economic
issue: It involves the integration costs of variable generation and the amount of various
storage or other enabling technologies that are economically viable in a future with high
penetrations of VG. To date, integration studies of wind to about 20% on an energy basis
have found that the grid can accommodate a substantial increase in VG without the need
for energy storage, but it will require changes in operational practices, such as sharing of
generation resources and loads over larger areas. Beyond this level, the impacts and costs
are less clear, but 30% or more appears feasible with the introduction of low-cost
flexibility options such as greater use of demand response. However, these studies have
not necessarily focused on storage and generally do not attempt to determine the optimal
system (including the amount of storage) that provides the lowest cost of energy.

There are technical and economic limits to how much of a systems energy can be
provided by VG without enabling technologies based on at least two factors: coincidence
of VG supply and demand and the ability to reduce output from conventional generators.
At extremely high penetration of VG, these factors may cause excessive (and costly)
curtailment, which will require methods to increase the useful contribution of VG
However, the concern regarding how much VG can be used before storage is the most
economic option for further integration currently has no simple answer, primarily because
the availability and cost of grid flexibility options are not well understood and vary by
region.

It is clear that high penetration of variable generation increases the need for all flexibility
options including storage, and it also creates market opportunities for these technologies.
Historically, storage has been difficult to sell into the market, not only due to high costs,
but also because of the array of services it provides and the challenges it has in
quantifying the value of these services particularly the operational benefits such as
ancillary services. The challenge of simulating energy storage in the grid, estimating its
total value, and actually recovering those value streams continues to be a major barrier.
VG complicates this issue because variability adds additional analysis challenges. The
ability to simulate the cost impacts of VG and benefits of storage is still limited by the
methods and data sets available. It is understood that VG increases the need for flexible
generation and operating reserves, which can be met by energy storage. However, the
value of energy storage is best captured when selling to the entire grid, instead of any
single source. Evaluating the role of storage with VG sources requires continued
analysis, improved data, and new techniques to evaluate the operation of a more dynamic
and intelligent grid of the future.

46
Bibliography
The following are general overviews and reviews of energy storage technologies that
provide additional information.

American Physical Society (APS). (2007). Challenges of Electricity Storage

Technologies. Accessed December 2009 at http://www.aps.org/policy/reports/popa-
reports/upload/Energy-2007-Report-ElectricityStorageReport.pdf

Chi-Jen, Y.; Williams, E. (2009). Energy Storage for Low-Carbon Electricity. Duke
University Climate Change Policy Partnership.
http://www.nicholas.duke.edu/ccpp/ccpp_pdfs/energy.storage.pdf

Electric Power Research Institute/Department of Energy (EPRI/DOE). (2003).

Handbook of Energy Storage for Transmission and Distribution Applications. Palo Alto,
CA: 2003. 1001834.

Electricity Advisory Committee. (2008). Bottling Electricity: Storage as a Strategic Tool

for Managing Variability and Capacity Concerns in the Modern Grid. Accessed
December 2009 at http://www.oe.energy.gov/DocumentsandMedia/final-energy-
storage_12-16-08.pdf

Electricity Storage Association. (2009). Accessed December 2009 at

http://www.electricitystorage.org/site/home/

Hall, P.J.; Bain, E.J. (2008). Energy-storage technologies and electricity generation.
Energy Policy 36(2008)43524355.

Ibrahima, H.; Ilincaa, A.; Perron, J. (2008). Energy storage systemscharacteristics and
comparisons. Renewable and Sustainable Energy Reviews 12 (2008) 12211250.

Pew Center on Global Climate Change. (2009). Electric Energy Storage. Accessed
December 2009 at http://www.pewclimate.org/technology/factsheet/EnergyStorage

Rastler, D. (2008). New Demand for Energy Storage, September/October 2008.

Accessed December 2009 at
http://www.eei.org/magazine/EEI%20Electric%20Perspectives%20Article%20Listing/20
08-09-01-EnergyStorage.pdf.

47
References
Ackermann, T.; Ancell, G.; Borup, L.D.; Eriksen, P.B.; Ernst, B.; Groome, F.; Lange, M.;
Mohrlen, C.; Orths, A.G.; O'Sullivan, J.; de la Torre, M. (2009). Where the wind
blows. IEEE Power and Energy Magazine, Vol. 7, No. 6. (30 October 2009), pp. 65-75.

Adamson, D. M. (2009). Realizing New Pumped-Storage Potential Through Effective

Policies, Hydro Review, April 2009, pages 28-30.

American Society of Civil Engineers (ASCE). (1993). Task Committee on Pumped

Storage of the Hydropower Committee of the Energy Division of the American Society
of Civil Engineers. Compendium of Pumped Storage Plants in the United States.
American Society of Civil Engineers, New York.

American Wind Energy Association. (2009). Annual Wind Industry Report Year Ending
2008. Accessed December 2009 at http://www.awea.org/publications/reports/AWEA-
Annual-Wind-Report-2009.pdf

Boyd, D. W.; Buckley, O.E.; Clark, C.E. (1983). Assessment of Market Potential of
Compressed-Air Energy-Storage Systems. Journal of Energy 1983, 7, 549-556.

Butler, P.C.; Iannucci, J.; Eyer, J. (2003). Innovative Business Cases For Energy Storage
In a Restructured Electricity Marketplace, SAND2003-0362 February 2003.

California Independent System Operator (CAISO). (2007). Integration of Renewable

Resources. November 2007. California Independent System Operator.

Corbus, D.; Milligan, M.; Ela, E.; Schuerger, M.; Zavadil, B. (2009). Eastern Wind
Integration and Transmission Study -- Preliminary Findings: Preprint. 9 pp.; NREL
Report No. CP-550-46505.

DeCesaro, J.; Porter, K.; M. Milligan. (2009). Wind Energy and Power System
Operations: A Review of Wind Integration Studies to Date. The Electricity Journal,
Volume 22, Issue 10, Pages 34-43 (December 2009).

Denholm, P.; Sioshansi, R. (2009). The Value of Compressed Air Energy Storage with
Wind in Transmission-Constrained Electric Power Systems, Energy Policy 37, 3149-
3158.

Denholm, P.; Margolis, R.M. (2008). Supply Curves for Rooftop Solar PV-Generated
Electricity for the United States, NREL/TP-670-44073.

Denholm, P.; Margolis, R.M. (2007a). Evaluating the Limits of Solar Photovoltaics
(PV) in Traditional Electric Power Systems Energy Policy 35, 2852-2861.

48
Denholm, P.; Margolis, R.M. (2007b). Evaluating the Limits of Solar Photovoltaics
(PV) in Electric Power Systems Utilizing Energy Storage and Other Enabling
Technologies. Energy Policy 35, 4424-4433.

Denholm, P.; Letendre, S.E. (2007). Grid Services from Plug-in Hybrid Electric
Vehicles: A Key to Economic Viability? Electrical Energy Storage Applications and
Technology Conference. September 25, 2007.

Denholm, P.; Short, W. (2006). An Evaluation of Utility System Impacts and Benefits of
Optimally Dispatched Plug-In Hybrid Electric Vehicles. NREL/TP-620-40293.

Department of Energy (DOE). (1977). DOE Interagency Coordination Meeting on

Energy Storage, 1977, CONF 7709116.

Desai, N.; Nelson, S.; Garza, S.; Pemberton, D.J.; Lewis, D.; Reid, W.; Lacasse, S.;
Spencer, R.; Manning, L.M.; Wilson, R. (2003). Study Of Electric Transmission In
Conjunction With Energy Storage Technology, Lower Colorado River Authority, Texas
State Energy Conservation Office, August 21, 2003.

Doherty, R.; Mullane, A.; Nolan, G.; Burke, D. J.; Bryson, A.; O'Malley, M.
(forthcoming) An Assessment of the Impact of Wind Generation on System Frequency
Control. IEEE Transactions on Power Systems, forthcoming.

Energy Information Administration (EIA). (2009a). Electric Power Annual, 2009 (data
for 2008). Accessed December 2009 at
http://www.eia.doe.gov/cneaf/electricity/epa/epa_sprdshts.html

EIA. (2009b). Repeal of the Powerplant and Industrial Fuel Use Act (1987) Accessed
December 2009 at
http://www.eia.doe.gov/oil_gas/natural_gas/analysis_publications/ngmajorleg/repeal.html

Ela, E.; Kirby, B. (2008). ERCOT Event on February 26, 2008: Lessons Learned.
National Renewable Energy Laboratory. NREL/TP-500-43373, July 2008.

EnerNex Corporation. (2009). Solar Integration Study for Public Service Company of
Colorado. February 2009 for Xcel Energy. Accessed December 2009 at
http://www.xcelenergy.com/SiteCollectionDocuments/docs/PSCo_SolarIntegration_0209
09.pdf

Electric Power Research Institute/Department of Energy (EPRI/DOE). (2004). Energy

Storage for Grid Connected Wind Generation Applications, EPRI-DOE Handbook
Supplement 1008703, December 2004.

EPRI. (2003). Handbook of Energy Storage for Transmission and Distribution

Applications. Palo Alto, CA: 2003. 1001834.

49
EPRI. (1976). Assessment of energy storage systems suitable for use by electric
utilities. EPRI-EM-264, July 1976.

Electricity Storage Association (ESA). (2009). Technology Comparisons. Accessed

December 2009 at http://www.electricitystorage.org/site/technologies/

European Wind Energy Association (EWEA). (2009). Global Installed Wind Power
Capacity. Accessed December 2009 at
http://www.ewea.org/fileadmin/ewea_documents/documents/press_releases/2009/GWEC
_Press_Release_-_tables_and_statistics_2008.pdf

Eyer, J.M.; Iannucci, J.J.; Corey, G.P. (2004). Energy Storage Benefits and Market
Analysis Handbook. A Study for the DOE Energy Storage Systems Program.
SAND2004-6177, December 2004.

Figueiredo, F.C.; Flynn, P.C.; Cabral, E.A. (2006). The economics of energy storage in
14 deregulated power markets. Energy Studies Review 14, 131152.

General Electric (GE) Energy. (2008). Analysis of Wind Generation Impact on ERCOT
Ancillary Services Requirements. Prepared for the Electric Reliability Council of Texas
(ERCOT), March 2008. Accessed December 2009 at http://www.uwig.org/AttchA-
ERCOT_A-S_Study_Exec_Sum.pdf

Hurlbut, D. J. (2009). Colorado's Prospects for Interstate Commerce in Renewable

Power. Prepared for the Colorado Governor's Energy Office, Renewable Energy
Development Infrastructure (REDI). NREL Report No. TP-6A2-47179.

Independent System Operators and Regional Transmission Organizations Council (IRC).

(2009). 2009 State of the Markets Report, prepared by the ISO/RTO Council. Accessed
December 2009 at http://www.isorto.org/atf/cf/%7B5B4E85C6-7EAC-40A0-8DC3-
003829518EBD%7D/2009%20IRC%20State%20of%20Markets%20Report.pdf

Jabbour, S.J.; Wells, W.M. (1992). Optimal Dispatching of Storage Plants with
Dynamics, in Proceedings: Second International Conference on Compressed Air Energy
Storage. December 1992. EPRI TR-101770.

Kirby, B. (2004). Frequency Regulation Basics and Trends, Oak Ridge National
Laboratory, December 2004, ORNL/TM 2004/291.

Lamont, A. D. (2008). Assessing the long-term system value of intermittent electric

generation technologies. Energy Economics 30, No. 3 (May): 1208-1231.

Lazarewicz, M. (2009). Commercial Flywheel Based Frequency Regulation Status,

Electricity Storage Association annual meeting 2009. Washington, D.C.,
May 21-22, 2009.

50
Lefton, S.A.; Besuner, P. (2006). The Cost of Cycling Coal Fired Power Plants, Coal
Power Magazine, Winter 2006.

Lew, D.; Milligan, M; Jordan, G.; Freeman, L.; Miller, N.; Clark, K.; and Piwko, R.
(2009). How do Wind and Solar Power Affect Grid Operations: The Western Wind and
Solar Integration Study. NREL/CP-550-46517.

Makarov, Y.V.; Ma, J.; Lu, S.; Nguyen, T.B. (2008). Assessing the Value of Regulation
Resources Based on Their Time Response Characteristics. June 2008, PNNL 17632.

Milligan, M.; Lew, D.; Corbus, D.; Piwko, R.; Miller, N.; Clark, K.; Jordan, G.;
Freeman, L.; Zavadil, B.; Schuerger, M. (2009a). Large-Scale Wind Integration Studies
in the United States: Preliminary Results. NREL/CP-550-46527, September 2009.

Milligan, M.; Kirby, B.; Gramlich, R.; Goggin, M. (2009b). The Impact of Electric
Industry Structure on High Wind Penetration Potential. NREL/TP-550-46273, July
2009.

Milligan, M.; Kirby, B. (2009). Calculating Wind Integration Costs: Separating Wind
Energy Value from Integration Cost Impacts. NREL TP-550-46275.

Nickell, B.M. (2008). Wind Dispatchability and Storage Interconnected Grid

Perspective. Presentation accessed December 2009 at
http://www.austinenergy.com/About%20Us/Newsroom/Reports/taskForce/tfWindDispatc
hability.pdf

North American Electric Reliability Corporation (NERC). (2009). Accommodating

High Levels of Variable Generation, April 2009. Accessed December 2009 at
http://www.nerc.com/docs/pc/ivgtf/IVGTF_Outline_Report_040708.pdf

NERC. (2008). NERC Glossary of Terms Used in Reliability Standards, February 12,
2008. Accessed December 2009 at www.nerc.com/files/Glossary_12Feb08.pdf

Northwest Power and Conservation Council. (2009). Draft Northwest Sixth Power
Plan, September 2009.
http://www.nwcouncil.org/energy/powerplan/6/DraftSixthPowerPlan_090309.pdf

Nourai, A. (2007). Installation of the First Distributed Energy Storage System (DESS) at
American Electric Power (AEP), June 2007, SAND2007-3580.

Nourai, A.; Kogan, V.I.; Schafer, C.M. (2008). Load Leveling Reduces T&D Line
Losses, IEEE Transactions on Power Delivery, October 2008, Volume: 23, Issue: 4:
2168-2173.

Palmintier, B.; Hansen, L.; Levine, J. (2008). Spatial and Temporal Interactions of Solar
and Wind Resources in the Next Generation Utility, Solar 2008, San Diego, CA.

51
Peterson, S.B.; Whitacre, J.F.; Apt, J. (2010). The Economics of Using PHEV Battery
Packs for Grid Storage. Journal of Power Sources, 2010. 195: 2377-2384.
PJM Interconnections (PJM). (2009a). PJM Manual 12: Balancing Operations,
Revision: 20, October 5, 2009. http://www.pjm.com/markets-and-operations/ancillary-
services/~/media/documents/manuals/m12.ashx

PJM. (2009b). PJM Manual 11: Scheduling Operations, Revision: 43. September 24,
2009. http://www.pjm.com/markets-and-operations/ancillary-
services/~/media/documents/manuals/m11.ashx

Rastler, D. (2008). New Demand for Energy Storage, Electric Perspectives,

September/October 2008. Accessed December 2009 at
http://www.eei.org/magazine/EEI%20Electric%20Perspectives%20Article%20Listing/20
08-09-01-EnergyStorage.pdf

Rebours Y.; Kirschen, D.S. (2005). A survey of definitions and specifications of reserve
services, University of Manchester, 2005.

Sherwood, L. (2009). U.S. Solar Market Trends 2008. Interstate Renewable Energy
Council, published July 2009. Accessed December 2009 at
http://www.irecusa.org/fileadmin/user_upload/NationalOutreachDocs/SolarTrendsReport
s/IREC_Solar_Market_Trends_Report_2008.pdf .

Short, W.; Denholm, P. (2006) A Preliminary Assessment of Plug-In Hybrid Electric

Vehicles on Wind Energy Markets, NREL/TP-620-39729.

Silverman, J. (ed). (1980). Energy Storage: a vital element in mankinds quest for
survival and progress. Transactions of the First International Assembly held at
Dubrovnik, Yugoslavia, 27 May-1 June, 1979. Pergamon Press.

Sioshansi, R.; Denholm, P.; Jenkin, T.; Weiss, J. (2009) Estimating the Value of
Electricity Storage in PJM: Arbitrage and Some Welfare Effects, Energy Economics. 31,
269-277.

Smith, J.C.; Parsons, B. (2007). What Does 20% Look Like? IEEE Power & Energy,
Nov/Dec 2007 22-33.

Smith, J.C.; Milligan, M. R.; DeMeo, E.A.; Parsons, B. (2007). Utility Wind Integration
and Operating Impact State of the Art, IEEE Transactions On Power Systems, Vol. 22,
No 3, August 2007.

Strauss, P. L. (1991). Pumped storage, the environment and mitigation. Waterpower

91, New York

52
Succar, S.; Williams, R.H. (2008). Compressed Air Energy Storage: Theory, Resources,
And Applications For Wind Power, published for the Princeton Environmental Institute,
April 2008.

Troy, N.; E. Denny, E.; M. OMalley, M. (forthcoming) Base-load Cycling on a System

with Significant Wind Penetration IEEE Transactions on Power Systems, forthcoming.

Walawalkar, R.; Apt, J.; Mancini, R. (2007). Economics of electric energy storage for
energy arbitrage and regulation in New York, Energy Policy 35 (2007) 25582568.

Zavadil, R. (2006). Wind Integration Study for Public Service Company of Colorado,
Enernex Corporation for Xcel Energy, May 2006.

53
 Form Approved
REPORT DOCUMENTATION PAGE OMB No. 0704-0188
The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources,
gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this
collection of information, including suggestions for reducing the burden, to Department of Defense, Executive Services and Communications Directorate (0704-0188). Respondents
should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a
currently valid OMB control number.
PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ORGANIZATION.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (From - To)
January 2010 Technical Report
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
The Role of Energy Storage with Renewable Electricity Generation DE-AC36-08-GO28308
5b. GRANT NUMBER

5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) 5d. PROJECT NUMBER

P. Denholm, E. Ela, B. Kirby, and M. Milligan NREL/TP-6A2-47187
5e. TASK NUMBER
WER8.5005
5f. WORK UNIT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION

National Renewable Energy Laboratory REPORT NUMBER
1617 Cole Blvd. NREL/TP-6A2-47187
Golden, CO 80401-3393

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR'S ACRONYM(S)

NREL

11. SPONSORING/MONITORING
AGENCY REPORT NUMBER

12. DISTRIBUTION AVAILABILITY STATEMENT

National Technical Information Service
U.S. Department of Commerce
5285 Port Royal Road
Springfield, VA 22161
13. SUPPLEMENTARY NOTES

14. ABSTRACT (Maximum 200 Words)

Renewable energy sources, such as wind and solar, have vast potential to reduce dependence on fossil fuels and
greenhouse gas emissions in the electric sector. Climate change concerns, state initiatives including renewable portfolio
standards, and consumer efforts are resulting in increased deployments of both technologies. Both solar photovoltaics
(PV) and wind energy have variable and uncertain (sometimes referred to as intermittent) output, which are unlike the
dispatchable sources used for the majority of electricity generation in the United States. The variability of these sources
has led to concerns regarding the reliability of an electric grid that derives a large fraction of its energy from these
sources as well as the cost of reliably integrating large amounts of variable generation into the electric grid. In this report,
we explore the role of energy storage in the electricity grid, focusing on the effects of large-scale deployment of variable
renewable sources (primarily wind and solar energy).
15. SUBJECT TERMS
NREL; energy analysis; energy storage; electric grid; renewable energy; wind energy; solar energy; variability; solar
photovoltaics; PV; variable generation; energy cost; renewable electricity generation; Paul Denhom; Erik Ela; Michael
Milligan; Brendan Kirby
16. SECURITY CLASSIFICATION OF: 17. LIMITATION 18. NUMBER 19a. NAME OF RESPONSIBLE PERSON
OF ABSTRACT OF PAGES
a. REPORT b. ABSTRACT c. THIS PAGE
Unclassified Unclassified Unclassified UL
19b. TELEPHONE NUMBER (Include area code)

Standard Form 298 (Rev. 8/98)

Prescribed by ANSI Std. Z39.18

F1147-E(10/2008)