Large Scale, Long Duration

Energy Storage, and the Future

of Renewables Generation

01
White Paper

Large Scale, Long Duration

Energy Storage, and the Future
of Renewables Generation

Form Energy, a Massachusetts based startup, is developing and commercia-

lizing ultra-low cost (<$10/kWh), long duration (>24hr) energy storage systems
that can match existing energy generation infrastructure globally. These systems
can reshape the electric system, making renewables fully firm and dispatchable
year-round. Form Energy has comprehensively assessed the electrochemical
landscape and screened for fundamental cost, abundance, and suitability for long
duration applications. The result is two technology platforms under development,
an aqueous sulfur (AqS) system, and a proprietary longer duration system. Addi-
tionally, the company has developed sophisticated analytical tools to model and
identify highest value applications and ensure a tight product-market fit.

Foundation

Enel Foundation is non-profit organization focusing on the crucial role of cle-

an energy to ensure a sustainable future for all. By developing partnerships with
pre-eminent experts and institution across the globe, leveraging on the vast know-
ledge of its founders, Enel Foundation conducts research to explore the implica-
tions of global challenges in the energy domain and offers education programs to
the benefit of talents in the scientific, business and institutional realms.
Acknowledgements

Stopping anthropogenic climate change is a major achievement we, individually

and collectively, aspire to, and believe is within reach. One of many reasons for
hope: our ability to cooperate with others to contribute to something bigger than
ourselves, a uniquely human trait that makes us the awesome species we are.

The following people contributed in significant and diverse ways to this paper.
Thank you.

Drew Beyer 5
Gianluca Gigliucci 3
Chris Harvey 3
Mateo Jaramillo 1
Ryan Kane 5
Luigi Lanuzza 4
Petar Litchev 5
Mark McGrail 3
Carlo Papa 2
Daniel Pappo 5
Massimo Schiavetti 3
Ted Wiley 1

1
Form Energy
2
Enel Foundation
3
Enel Green Power
4
Enel X
5
Enel Trading North America
Table of Contents

1 | NOTE FROM THE AUTHORS1

2 | INTRODUCTION 2

2.1 Objectives  2
2.2 Scope  2
2.3 Study Overview  2

3 | RENEWABLE ENERGY PENETRATION AND BARRIERS 5

3.1 Renewable Energy Development in the USA 5

3.2 Obstacles to RES development 6

4 | COMMERCIAL RES, US EXAMPLE  10

4.1 C&I evolution 10

4.2 Current C&I PPA Structures and Limitations 10
4.3 Charting the Pathway for Future C&I Off-taker Agreements 13

5 | ENERGY STORAGE SYSTEMS  15

5.1 Short Duration Storage 15

5.2 Long Duration Storage 16

6 | QUANTITATIVE ANALYSIS  19

6.1 Overview  19
6.1.1 Objectives 19
6.1.2 Scope 19
6.1.3 Methodology 20
6.1.4 Data Sets 22
6.2 Results 23
6.2.1 Results without Storage  23
6.2.2 The Impact of Storage - Pumped Hydro Storage like Example  24
6.2.3 Sensitivity Analysis  26
6.3 Summary of Storage Impact  36

7 | CONCLUSIONS37
1 Note from the authors

Renewables widespread adoption across the globe is at the core of sustainable

energy transition just for all.
To effectively manage larger scale of variable renewable energy, power system
flexibility is the name of the game and indeed storage is and will be one of the
core enablers of decarbonized energy systems.
In the United States Corporate Power Purchase Agreements (PPAs) are a
major driving force for renewable power deployment. With deeper renewables
penetration, energy intermittency is causing an increase in risks borne by parties
to a PPA, and will require effective mitigation to enable the continuation of fast-
paced deployments.
In this work, we use FormWare™, a proprietary asset optimization software
developed by Form Energy, to explore the impact of increased volume and basis
risk on the distribution of returns for long-term contracted windfarms in the
Southwest Power Pool (SPP) footprint, under a simple contract for differences
and operating in a day-ahead / real-time market environment. Quantitative results
demonstrate the ability of storage to effectively manage risk and returns across a
variety of potential storage technologies, while currently available short-duration
storage technology shows limited impact. We extend the framework to evaluate
a range of future renewables scenario and associated risk levels, and offer a
methodology to quantitatively assess the risk and return benefits of storage for
financially settled, long-term contracted renewable assets.
This work has been carried out considering assets located in the US deregulated
markets, and thanks to the methodology and scientific approach can be extended
to other locations and markets. In addition, the underlying conditions and risk
factors (e.g. intrinsic technology features, variability of renewable sources, limited
long distance energy transmission capacity and congestion, supply-demand
mismatch etc.) are common across locations and market settings. For this reason,
the results of this study have, at least qualitatively, a clear implication for a wide
range of geographies.

Matt Arnold 3
Sander Cohan 3
Aly Eltayeb 1
Marco Ferrara 1
Ben Jenkins 1
Giuseppe Montesano 2
Claudio Pregagnoli 2
Carlo Napoli 2

1
Form Energy, 2 Enel Foundation, 3 Enel Green Power

1
2 Introduction

2.1 | Objectives
The objectives of this paper are twofold:

1_ Evaluate the economic rationale for pairing utility scale renewable energy
with Long Duration Energy Storage (LODES), by analyzing the conditions
that would allow LODES to increase and/or stabilize the market revenues
of a renewable energy facility (specifically, a large utility scale wind farm);
2_ Understand how LODES can provide a key technology to add value and
bridge the gap between renewables intermittency and predictable,
dispatchable renewables, thus overcoming one of the most substantial
barriers to 100% adoption of renewable power.

2.2 | Scope
Using real-world examples describing current trends in utility scale wind power
generation farms, notably the trend towards private-party power purchase
agreements to support climate sustainability goals, the study will assess how
LODES can provide a key technology to add value and bridge the gap between
renewables intermittency and predictable, dispatchable renewables.

2.3 | Study Overview

Electrical power systems are in the first phases of a profound transformation as
the cost competitiveness of renewables puts irreversible pressure on natural
gas, coal and nuclear generation, globally. For example, McKinsey anticipates
that renewables will reach unsubsidized cost competitiveness with coal and gas
in the majority of geographies in the 2025 -2030 time-frame1. In another study,
the Carbon Tracker Initiative finds that 42% of global operating coal fleet is
unprofitable in 2018 and 72% will be by 2040, independent of additional climate
or air pollution policy2. Between 2019 and 2040, renewable energy will be the
fastest growing source of energy across the world at an average of 7.1% p.a. (BP
Energy Outlook, 2019).

1
McKinsey, Energy Insight Global Energy Perspective. January 2019
2
The Carbon Tracker, Powering Down Coal. November 2018

2
 2 | Introduction

In the US, low cost renewables are putting the $112 billion of gas-fired power
plants currently proposed or under construction -- along with $32 billion of
proposed gas pipelines to serve these power plants -- at risk of becoming
stranded assets. While it is true that extremely low cost of fuel is driving a short-
term increase in natural gas-powered generation in parts of the country where
coal and nuclear plants have retired, the combination of renewable energy cost
declines, environmental concerns, and regulatory and legislative pressure have
begun to place the long-term financial viability of natural gas assets in question.
According to a recent study by the Rocky Mountain Institute3, “across a wide range
of case studies, regionally specific clean energy portfolios already outcompete
proposed gas-fired generators, and/or threaten to erode their revenue within the
next 10 years. This has significant implications for investors in gas projects (both
utilities and independent power producers) as well as regulators responsible for
approving investment in vertically integrated territories.”

So what is to stop this powerful march towards a 100% clean, renewable future?

In the same study, McKinsey concludes that natural gas generation will continue
to play a critical role in the grid of the future acting as a stable baseload and
dispatchable capacity provider in renewable-heavy systems, thereby hindering
the objective of a deeply decarbonized grid. In another study specifically focused
on the European grid, Eurelectric estimates that ~400GW of dispatchable gas
reserves will be required in high decarbonization scenarios (>80%) to provide
system flexibility for days with low renewable generation. This gas capacity is in
addition to 100GW – 200GW of commercial battery technologies4.

In theory, utilities could deploy lithium-ion or other commercially available battery

technologies in large enough quantities to ride through periods of wind lulls, cloud
formations or grid outages and offset the need for carbon emitting gas generation.
Lithium-ion batteries in particular offer high energy and power density, high cycling
efficiency, low self-discharge rate, fast response time, and low cost of maintenance
(Argyrou et al, 2018). Moreover, the cost of lithium-ion packs have come down
rapidly, from $1,160/kWh in 2010 to $176/kWh in 2018 (Goldie-Scot, Bloomberg
NEF 2019). Correspondingly, lithium-ion has seen a dramatic uptake over recent
years, dominating 95% of all new energy storage capacity in the US since 2013 and
seeing a 43% increase in installed capacity from 2017 to 2018 (IHS Markit, 2019).

So, why aren’t commercially available battery technologies good enough?

In reality, even at the lowest price forecasts for lithium-ion or other commercially
available battery technology, the modular nature of lithium-ion technology
drives deployment costs linearly higher, making durations greater than 10 hours
economically challenging. Lithium-ion and other “short-duration” energy storage

3
Rocky Mountain Institute, The Economics of Clean Energy Portfolios. 2018
4
Eurelectric, Decarbonization Pathways, page 62. May 2018

3
 2 | Introduction

technologies will have an important role in distributed residential, commercial

and industrial systems as well as in applications related to electric mobility;
however, there is a need and opportunity for long duration electrical storage
systems (LODES), which can be broadly defined as electrical storage systems
with durations greater than 10 hours.

In a recent review of 40 academic studies of decarbonization pathways, Jenkins

et al. observe that, “while [renewable] overgeneration arises during periods of
abundant supply, periods of scarce wind or solar production are the flip side
of the variability challenge. Prolonged periods of calm wind speeds lasting
days or weeks during winter months with low solar insolation are particularly
challenging for [variable renewable energy, VRE]-dominated systems. These
sustained lulls in available wind and solar output are too long to bridge with
shorter-duration batteries or flexible demand. Power systems with high VRE
shares consequently require sufficient capacity from reliable electricity sources
that can sustain output in any season and for long periods (weeks or longer)”5.
In acknowledgement of the need, the US Advanced Research Project Agency
- Energy (ARPA-E) has launched a federally funded grant program to develop
energy storage systems that provide power to the electric grid for durations
of 10 to approximately 100 hours with the scope of “opening significant new
opportunities to increase grid resilience and performance”6.

The rest of this paper will focus on a specific use case where the variable
nature of renewable generation combined with the limitations of transmission
infrastructure are already causing operational and financial risks and threatening
the march to high-renewable futures in the US. In particular, the paper will:

1_ Introduce the concept and general structure of commercial power purchase

agreements (PPAs), the fastest growing mechanism of contracting and financing
new renewable generation infrastructure in the US.
2_ Identify operational and financial risks arising from price volatility and
congestion in regions of the grid with excellent renewable resources but slow
transmission expansion.
3_ Lay the analytical foundation to investigate how LODES can reduce such
sources of risk.
4_ Perform extensive modeling and simulation based on real-world scenarios
to drive quantitative conclusions with regard to the value of LODES in said
applications.

In conclusion, the last section of the paper will discuss new opportunities for
hybrid renewable and LODES products well suited for future grids with majority
fractions of renewable generation.

5
Jesse D. Jenkins, Max Luke, Samuel Thernstrom, Getting to Zero Carbon Emissions in the
Electric Power Sector. Volume 2, issue 12, P2498-2510, December 19, 2018.
6
ARPA-E DAYS

4
3 Renewable Energy
Penetration and Barriers

3.1 | Renewable Energy Development

in the USA
According to the Energy Information Administration (EIA), US solar capacity
is projected to grow by 17% in 2020. Wind will similarly grow by 14%.7 Within
the ISO territories evaluated for this study, namely the Southwest Power Pool
(SPP) and ERCOT, renewable energy development has accelerated rapidly.
For example, in 2008, wind energy made up just 3% of SPP’s annual energy
production. By 2018, wind’s share had increased to 23%. Additionally, SPP
has reliably met as much as 69% of its load with renewable resources and
64% with wind alone at a given point in time. 8 This momentum is expected
to continue: total wind capacity in SPP is expected to leap from 20.5 GW
(2018) to 40.5 GW by 2030. 9 Similarly, ERCOT has so far been able to meet
as much as 54% of its load with wind alone at a given point in time.10 More
importantly, as of August 2018, 86% of ERCOT’s pipeline of new energy
projects is either wind or solar, with solar projected to make substantial gains
through to 2030.11

This rapid growth will be driven by a number of factors; including favorable

economic conditions. According to Bloomberg New Energy Finance, by 2050
the cost of an average PV plant will fall by 71% and wind by 58%, the cost per
kilowatt hour of lithium ion batteries will fall to roughly $70, global share of coal
generation will shrink to 11%, and natural gas will only grow modestly.12
With growing concern over climate change, the American policy environment
is also increasingly supportive of rapid growth in renewables. For example, 29

7
Tyler Hodge, Short Term Energy Outlook. January 2019.
8
Derek Wingfield. As it turns five, Southwest Power Pool’s Integrated Marketplace is saving
billions and enabling big changes in energy dispatch. Southwest Power Pool Press Release
February 28, 2019.
9
Gary Cate, presentation. SPP’s Integrated Marketplace and Renewable Energy Evolution.
Southwest Power Pool. October 17, 2017.
10
Jeff St. John. Texas Grid Operator Reports Fuel Mix is now 30% Carbon Free. Greentech Media,
January 23, 2019.
11
John Weaver. Texas is going green: 86% of future capacity solar or wind, zero coal. PV Magazine
USA. August 23, 2018.
12
Seb Henbest et al. BNEF New Energy Outlook 2018. Bloomberg NEF 2018.

5
 3 | Renewable Energy Penetration and Barriers

of the 50 states have established Renewable Portfolio Standards (RPS) as of

2018, many of which have undergone multiple iterations to ratchet up renewable
generation requirements further.13 Perhaps the most radical political gesture
involves the February 2019 publication of the “Green New Deal” Resolution
by Democratic members of the US Congress that called for 100% renewable
generation by 2030.

Large American Commercial and Industrial (C&I) entities have also signaled
growing appetites for renewable energy, partly out of concern for sustainability,
having more than doubled the annual record of corporate off-take renewable
power purchase agreements in 2018 by capacity added.14 In ERCOT alone,
corporate PPAs jumped from 292 MW in 2017 to 1.661 GW in 2018.15

3.2 | Obstacles to RES development

Despite the widespread interest across the US to rapidly increase renewable
energy use, meeting that demand is not without its obstacles. Over recent
years, the cost of renewables has achieved parity or even better with more
conventional, baseload technologies. Renewables like wind and solar, however,
disrupt the conventional methods for planning the daily operation of the electric
grid because their output fluctuates depending on the availability of the resources
(solar irradiance and wind speed). This variability forces the grid operator to
adjust its day-ahead, hour-ahead, and real-time scheduling while also disallowing
the renewable power producer to sell at the most ideal times, hindering them
from maximizing revenue. Because wind and solar increase the magnitude of
sudden power generation shortfalls or excesses, the grid operator requires
more reserve power ready to respond at a moment’s notice to ensure the grid
remains balanced. This process starkly contrasts with the baseload power from
fossil-fuel generators such as coal and natural gas, which can reliably produce
set quantities of electricity consistently.

Variability manifests in several ways that, without ample availability of storage

assets, leads to lost economic value. Price volatility --between morning, midday,
and evening energy prices-- can be most commonly reflected through the classic
“duck curve”, in which solar production hits its peak output during midday, causing
demand from other energy sources to be at its lowest. Demand rises rapidly later in
the day as solar production falls, placing pressure on grid operators to quickly bring
online other generating sources to compensate. With more renewable penetration

13
Galen Barbose. US Renewable Portfolio Standards: 2018 Annual Status Report. Lawrence
Berkeley National Laboratory. Electricity Markets & Policy Group. November 2018.
14
Emma Foehringer Merchant. The Year of the Corporate PPA. Greentech Media, December 21,
2018.
15
Sarah Krulewitz. Corporates May Be Leaving Millions on the Table by Procuring Wind Over Solar
in ERCOT. Greentech Media. February 27, 2019.

6
 3 | Renewable Energy Penetration and Barriers

without storage, the duck curve becomes increasingly emphasized. Hawaii has
already experienced days where the duck curve has become so pronounced by
solar penetration that the load drops below zero at midday, forcing the grid operator
to backfeed energy, and as the director of Hawaii’s Electro Co’s RE planning states,
“circuits that send power back up to distribution transformers and substations
cause all sorts of technical and operational challenges.”16 In this particular scenario,
storage would enable time-shifting of dispatch so that sources like solar and wind
would not have to dispatch at the same time that they generate.

Broadening the perspective from solar duck curves to negative wholesale

markets presents an even starker picture across the country. According to
Greentech Media, multiple regional ISOs have logged patterns of renewable
saturation driving down energy prices to negative states. Even at negative
prices, renewable plants will often continue to generate when the Production
Tax Credit17 still allows for positive revenues.

Table 1 Negative Wholesales Market Prices by ISO

2017 congestion (negative
pricing) in several ISOs
Total Negative Total Opportunity
in the US. SPP had Price Hours Average (Annulal $/MW,
the largest number (All Price Nodes /Zone) Negative Average Across Price
of negative hours in 2017. ISO Nodes /Zone) Price Nodes /Zone)

CAISO 2,044 -$ 4.57 -$ 667

ERCOT 37 -$ 1.11 -$ 41

ISO-NE 59 -$ 2.62 -$ 77

MISO 71 -$ 2.58 -$ 36

NYISO 0 — —

PJM 38 -$ 1.31 -$8

SPP 3,784 -$ 5.36 -$ 1,269

 Source: GTM Research, ISO data

16
Dora Nakafuji. Jeff St. John. Hawaii’s Solar Landscape and the “Nessie” Curve. Greentech
Media, February 10, 2014.
17
A federal incentive providing financial support for the development of renewable energy.

7
 3 | Renewable Energy Penetration and Barriers

Here again, pushing the imbalance between generation and load hits such
extremes because renewable generators lack control over when they can
generate (i.e. when the wind blows and when the sun shines). This economic
representation of lost natural resource efficiency can be ameliorated with
storage, enabling the discharge of renewable energy to occur when there is
more balance between supply and demand.

Finally, in instances where continued dispatch of energy would realize negative

returns or is simply prohibited by the grid operator due to congestion, renewable
plants may have to curtail production, e.g. go offline, during their finite windows
of access to the sun or wind. As Figure 7 demonstrates, the incidence of
curtailment has been escalating each year, corresponding with increased
renewable penetration.

Figure 1 Wind an solar curtailment totals by month

250,000
Curtailment has been
growing rapidly in recent
225,000
years in CAISO, where
renewable penetration is
only approaching 40%.
200,000

175,000
Megawatt hour (MWh)

150,000

125,000

100,000

75,000

50,000

25,000

0
2015 2016 2017 2018 2019

The burgeoning C&I market is now coping with the above challenges and
their inherent risks largely without long duration storage as an option. While
this technology gap continues, C&I off-takers have been turning to increasingly
complex and exotic financial instruments in order to minimize these risks.
Companies like Microsoft, for example, have partnered with REsurety, Allianz,
and Nephila Climate to pioneer the concept of a Volume Firming Agreement
(VFA), which moves risk related to weather conditions and the market value of
renewables to insurers.18

18
Ibid, Emma Foehringer Merchant.

8
 3 | Renewable Energy Penetration and Barriers

Storage technology has the potential to change the intermittency drawback

of renewables by allowing renewable producers to capture their energy and
discharge it at times when they can fetch adequate returns. More generally,
pairing storage with renewable generators allows those renewable plants to
behave more like dispatchable - guaranteeing fixed quantities of electricity on
deliberate schedules, regardless of the wind or sun shine. This opens the way
for renewable energy and storage technologies to permeate the market at utility
scale, allowing the substitution of fossil fuel-based plants with more renewable
plus storage plants that behave like dispatchable generators. Chapters 5 and 6
provide a more detailed discussion of existing storage technologies and their
potential role in addressing renewables intermittency.

9
4 Commercial RES, US Example

4.1 | C&I evolution

As LODES solutions become increasingly commercialized and available, one
of the most promising market segments for renewable energy plus storage to
quickly penetrate is with a newer brand of customer (off-takers) in Commercial
& Industrial (C&I) entities. One of the fastest growing mechanisms of renewable
contracting in the United States is the C&I off-taker power purchase agreement.
C&I off-takers were responsible for a quarter of new solar and wind capacity in
2018 (Green Tech Media 2018). According to the Rocky Mountain Institute, 2018’s
volume of corporate off-take agreements reached 6.43 GW of capacity, thereby
shattering the 2015 record of 3.22 GW (Rocky Mountain Institute Business
Renewables Center 2018).

What has spurred the C&I off-taker phenomenon? From the seller’s perspective,
C&I off-takers represent a burgeoning market that promises rising opportunity
in developing more renewable generation projects. From the C&I buyer’s
perspective, renewable PPAs are not only beginning to make economic sense
but are also enabling corporate entities to achieve sustainability objectives, such
as offsetting emissions from electricity consumption, in part by taking ownership
of associated Renewable Energy Credits (RECs) from the project. The rise of C&I
PPAs provides benefits for both sides. For the developer, it presents a new avenue
for renewable energy project development with less risk. For the C&I off-taker, it
ensures increasingly economically attractive scenarios for cheaper and cheaper
electricity while also achieving sustainability and climate change objectives.

4.2 | Current C&I PPA Structures

and Limitations
As the pool of C&I off-takers continues to expand in markets that are seeing
increasing penetration of renewables, offtake arrangements will need to further
evolve and become more flexible in order to mitigate the associated risks.
Currently, C&I PPAs are usually construed as either Physical PPAs or Virtual PPAs.

Physical PPAs: Physical PPAs are most commonly used by organizations that
have large concentrated loads. The renewable energy seller builds, owns, and
operates the project, and sells the output to the C&I off-taker at a specified
delivery point (market hub, etc.). At that point, the off-taker takes ownership

10
 4 | Commercial RES, US Example

of the energy and gains the associated Renewable Energy Credits (RECs), in
exchange for the off-taker paying a fixed price. Inherently, the physical PPA
requires that the energy can be physically exported through the grid from the
seller to the C&I customer. As a consequence, Physical PPAs are not a viable
option over considerable distances, even when technically possible, where grid
congestion charges and multiple ISOs make them economically unsustainable.
For these reasons this study focuses on virtual PPAs, as they represent a more
general case.

Virtual PPAs: When Physical PPAs are not possible or not economically
viable, Virtual PPAs can be an effective alternative. Like with physical PPAs,
the renewable energy seller builds, owns, and operates the project, and
delivers the output to a specified delivery point or node. Unlike physical
PPAs, the seller liquidates the energy locally (i.e. where power is generated)
at the market price; furthermore, the seller and the C&I off-taker enter a
contractual arrangement called contract for differences (CfD), whereby the
off-taker pays the seller the difference between the market price and a fixed
price (the strike price), when such difference is negative, and the payment is
reversed, when the difference is positive. Uncertainty in the drivers of that
market price and how much it will fluctuate constitutes a considerable risk to
potential C&I customers. This is partly what would make shorter-term PPAs
more attractive, as that uncertainty is less pronounced the closer the time
horizon is. Virtual PPAs most often still allow for the C&I customer to own
the associated RECs from the project, allowing them to take credit for using
renewable energy.

While Virtual PPAs have in particular opened up vast opportunity for C&I entities
to “go renewable” even when not physically receiving the power directly from
the renewable energy developer, these contracts do carry risks. The C&I off-
taker must have a strong grasp of the various factors that can reduce the local
market price so that the fixed price that they pay does not exceed it. Among
the myriad factors that drive the market price is the level of local renewable
penetration and how much its intermittency can affect the grid.

Despite their many advantages, today’s renewables PPAs include a number

of risks to both parties involved in the long-term commitment. The PPA is a
flexible contract structure that assigns risk across the parties involved in the
transaction. Depending on the PPA structure, the risks fall on either or both
parties. Although no hard rules exist, commercial trends have converged over
time to require parties to typically take on risks they are best positioned to
manage, or where the bargaining power favors the counterparty. These risks
stem from the intermittency of the resources, the uncertainty of demand and
the volatility of prices. The major categories of risk are:

• Basis risk: the spread in prices between where renewable power is generated
and where it is delivered (or in the case of virtual PPAs, settled) can be significant.
This spread, referred to as “basis”, is a function of the transmission capacity

11
 4 | Commercial RES, US Example

between the generation node and, for example, the settlement hub. With many
renewable projects co-located and, by nature of the resource, correlated, large
bursts of generation can result in locally depressed prices, in contrast with a large
settlement hub with much more power liquidity and exposure to a diversified
generation pool. Consider the case where the off-taker settles the virtual PPA at
the hub: in that case, the asset owner must sell power at the generation node and
settle with the off-taker at the hub, incurring a loss if the node prices are lower.
This exposure is the essence of basis risk.

• Volume risk: the difference between the forecasted renewable asset

volume and its actual production volumes can be a source of downside in
a number of ways. In its simplest form and in the context of a virtual PPA,
consider an asset owner bidding into the day-ahead market. At the time of
the bid, a decision is made based on a resource forecast for the next-day.
If actual volumes fall short, any gap between the volumes committed day-
ahead and the actual production volumes must be purchased from the market,
often at a penalty that captures the cost to the market of suboptimal marginal
producer entering the market to cover the shortage. This exposure to forecast
uncertainty is one form of volume risk.

• Shape risk: if the output of the renewable asset is drastically different from
the load shape of the C&I user, the value of a PPA as a hedge for the buyer
against increases in future electricity markets is severely reduced. In the case
of high local renewable penetration for a buyer and seller in the same hub,
the output of a renewable asset can drive low nodal prices, while the buyer is
exposed to high prices in the market where they purchase power coincident
with their load. This exposure is a form of shape risk.

In the early days of corporate PPAs, and with relatively low renewables
penetration on the grid, the impact of such risks on the profitability and value of
these long-term commitments was small. However, with increased renewables
penetration and larger price volatility, risk management is, more than ever, a
key consideration. As with growing congestion and curtailment (showcased
earlier in this paper) renewable PPA risks will increase drastically with deeper
renewables penetration, threatening the march towards a decarbonized grid.
Corporate procurement of renewables rests on more effective mechanisms for
risk management, for both parties.

Although PPAs are inherently flexible in how risk is allocated, general market
trends have emerged and evolved over time to reflect the needs of the various
parties. Current trends in virtual PPA procurement by corporate off-takers tend
to assign basis risk to the project owner / developer. Forecast uncertainty
and associated volume risks is again mainly borne by the asset operator &
developer, given their deeper understanding of the asset market conditions
and control of its scheduling & bidding. Shape risk is at present borne by both
parties – the asset operator / developer is exposed to the hub pricing volatility,
whereas the off-taker is exposed to the mismatch of their load costs with the

12
 4 | Commercial RES, US Example

PPA revenue. As renewables penetration drives higher volatility in markets,

managing shape risk is likely to become a more central piece of both virtual
and physical PPAs. Some recent PPAs have factored in premiums for energy
delivered during peak demand hours for off-takers.

One potential approach to PPA risk management is an increased trend towards

securitization, financial derivatives and complex financial structures. These
structures - by definition - involve the entry of intermediaries and additional
overhead costs. In this paper, we focus on a physical storage enabled alternative,
which acts to fundamentally reduce risk by addressing the intermittency of
renewables.

4.3 | Charting the Pathway for Future

C&I Off-taker Agreements
From the perspective of off-takers and developers alike, future long-term off-take
agreements must effectively manage risk exposure for both parties.

In general, customers do not want long-run risk exposure, and unless risk
factors can be tightly managed, shorter-term off-take commitments will become
prevalent. With shorter long-term off-take agreements, more projects will rely
in part on future cash flows from merchant revenue which are inherently riskier
from a revenue certainty perspective. As a result, the cost of borrowing will
increase and equity financing will constitute a larger fraction of the capital stack,
increasing the cost of capital for renewable projects. In addition, customers do
not want to take on basis risk as the complexity of forecasting and hedging
against basis risk requires specialized knowledge, that falls outside the core
business for most C&I energy buyers.

Volume and shape risk are arguably inevitable. Renewable assets production will
remain uncertain, and the correlation of customer load and prices will always result
in some level of risk exposure. These risk factors can be effectively managed
through a number of tools, like better forecasting, and, in liquid markets, financial
hedges. Customers need a clear and simple formulation that aligns incentives and
enables effective exchange of risk & return with developers. Many C&I energy
buyers are fairly sophisticated, but ultimately, off-take contract complexity will act
to slow down the adoption of corporate PPAs and limit the growth of the segment.

From a developer’s standpoint, off-take agreements must match the right level
of risk appetite of commercial off-takers, while allowing for sufficient revenue
certainty. In markets with capacity value, renewable energy projects can unlock
significant value if they are able to provide dispatchable power. Unlocking capacity
value can result in higher certainty for at least a portion of the project revenue. In
addition, dispatchability can allow for the renewable output to be better matched
with customer load and market prices, curbing the sources of risk.

13
 4 | Commercial RES, US Example

Our perspective in this paper is that dispatchable renewable power enabled by

long duration storage can significantly reduce risk for both parties, and enable
a more effective and clear exchange that simplifies off-take agreements.
With long duration storage, a dispatchable renewable asset can have higher
revenue certainty, can be better matched with customer load and structural
prices regimes (e.g. peak and off-peak), address volume risk through stored
power, and act as a low-cost conduit for power across high congestion times.
Quantitative analysis in Chapter 6 will offer a detailed analysis to demonstrate
the value of storage in managing both risk and return for long-term contracted
renewable assets.

14
5 Energy Storage Systems

Energy storage is poised to play a critical role in modernizing grids, making

them more scalable and resilient, solving emerging energy problems in the
burgeoning contexts of climate change and rapid urbanization, as well as
recasting the way that we approach the economics of electricity. The grid of
the future will require all sorts of storage applications from the micro (home/
EV to grid) to the medium/distribution scale to large transmission scale and
hybrid applications with utility-scale storage.

To a first order, batteries can be characterized by energy rating, charge and

discharge power ratings and charge and discharge efficiencies. Energy rating
indicates how much energy can be held by the battery when it’s fully charged;
charge and discharge power ratings indicate the maximum power that a
battery can draw from the grid or supply into the grid; finally, charge and
discharge efficiencies measure the power losses while the battery is charging
or discharging. A derivative and useful metric often used to characterize
batteries is nominal duration, which is the ratio between energy rating and
discharge power rating; alternatively, an effective duration metric can be
used, which is calculated by de-rating nominal duration by the discharge
efficiency. Some technologies, such as flow batteries, allow plant designs
where energy and power ratings are independent variables, allowing, in
theory, for flexible duration. Other technologies, such as Li-ion cells, impose
stringent constraints on battery durations, mostly related to manufacturing
processes and cell architectures.

5.1 | Short Duration Storage

Of the battery types that have seen recent growth in the marketplace,
lithium-ion particularly stands out as a “yardstick” to compare against for
other emerging storage technologies. Lithium-ion offers high energy and
power density, high cycling efficiency, low self-discharge rates, fast response
times, and low cost of maintenance (Argyrou et al, 2018). While the cost
of lithium-ion has historically been prohibitive, limiting their application to
high power and energy density ones, such as electric vehicles, these costs
have been dramatically decreasing. Vehicle packs, which are currently the
highest volume lithium-ion product, have dropped from $1,160/kWh in 2010
to $176/kWh in 2018, with projections of under $100/kWh in 2024 (Goldie-
Scot, Bloomberg NEF 2019). In contrast with BNEF’s cell and pack level cost
estimates, NREL provides plant level cost projections that are nearly twice
the energy cost for the pack as forecasted by BNEF. Plant level storage costs

15
 5 | Energy Storage Systems

include additional overhead such as electrical and structural components as

well as land and construction costs which nearly double the pack level energy
cost for a 4-hour project and more than triple the pack level cost for a 1-hour
system. Driven in part by falling costs, Li-ion has seen a dramatic uptake over
recent years, dominating 95% of all new energy storage capacity in the US
since 2013 and seeing a 43% increase in installed capacity from 2017 to 2018
(IHS Markit, 2019).

However, lithium-ion storage suffers a number of limitations, such as relatively

quick degradation rates, safety concerns tied to high-profile explosions and
fire incidents, concerns over the sustainability (and cost) of lithium and other
essential material mining, among others. Most importantly, project developers
do not typically use lithium-ion for dispatch applications longer than 4 hours in
duration because of the modular cost structure of the technology, which today
make longer duration projects (>4hrs) economically infeasible. This means
that in the near-term (2-5 years), the value of Li-ion technology is maximized in
shorter duration applications and especially where the technology really shines
such as frequency control response. In this study, we also explore the possibility
of future Li-ion costs, using forecasts for 2020, 2030 and 2040 Li-ion batteries
to explore longer-duration deployments of the technology.

For renewable energy to fully penetrate electricity markets, supply rapidly

growing demand among C&I customers, and supplant the need for fossil-fuel
based systems, either significant advances in existing technologies or new
long duration storage technologies will be needed to close the gap. LODES
will enable renewable energy to provide grid resilience, enable multi-hour time-
shifting and arbitrage, as well as replace intermittency with stability, savings,
and sustainability and, for C&I customers, to buy utility-scale renewable energy
on-demand (specific times of day).

5.2 | Long Duration Storage

LODES can be broadly defined as electrical storage systems with durations
greater than 10 hours. For example, LODES would be well in the scope of the
recent ARPA-E federally funded grant program to develop energy storage systems
that provide power to the electric grid for durations of 10 to approximately 100
hours with the scope of “opening significant new opportunities to increase grid
resilience and performance”. Because of the long duration, a critical requirement
of LODES is a very low $/kWh energy capital cost.

Traditionally, power assets are categorized by their role in serving the supply
stack. Baseload assets (like coal & nuclear reactors) are meant to operate at very
high capacity factors (>80%) to serve the majority of the load requirements.
Mid-merit assets (like modern CCGT plants) are more flexible and can typically
provide parts of the baseload function in addition to ramping over days and
seasons to match cyclical components of load. Peaker plants (like open-cycle

16
 5 | Energy Storage Systems

combustion turbines, diesel engines or more modern reciprocating engines)

operate at very low capacity factors (<10%) and are primarily intended to
address rapid load fluctuations and contribute to higher grid reliability. In a
future with deep renewables, asset roles shift as large intermittent low
marginal cost assets become the new source of baseload power, while the
need for flexible fast-ramping assets increases. This latter role is where long
duration storage will be most needed.

An order of magnitude analysis of costs can be instructive here. For example,

if the installed capital cost of natural gas plant $1,000/kW was used as a
benchmark, a 10-hour LODES would have to have an energy capital cost of
less than or equal to $100/kWh and a 100-hour LODES of less than or equal
to $10/kWh to respectively provide 10 and 100 hour output services cost
competitively. These prices appear to be beyond the most optimistic price
projections of currently available technologies at the plant level;19 in particular,
$10/kWh is well below the sole material cost of Li-ion cells.

The only commercially available option of low-cost, long-duration storage

today is pumped hydro; unfortunately, as much as the technology is desirable,
its deployment is geographically limited because of the very specific site and
environmental requirements. A large class of electrochemical systems, which
have long been neglected because of their low-rate capabilities and lower
than Lithium-ion round-trip efficiencies, may very well become the enabling
technology at the heart of LODES. Systems based on cheap, abundant
materials such as water, air and certain metals can achieve extremely low $/
kWh cost, at the expense of low round trip efficiencies (in the range 40%-70%)
and high self-discharge rates (>5%/month). However, these drawbacks should
be weighed against emerging grid conditions. If future grids do involve large
amounts of inexpensive renewable electricity, low round-trip efficiencies can
be tolerated. Moreover, LODES will necessarily have a much smaller yearly
cycle count than a short-duration storage system and cycle life requirements
can be correspondingly relaxed from the several thousands of a Lithium-ion
system to a few hundreds or a few tens for a system with durations between
10 and 100 hours.

Several new classes of electro-chemical, thermal and mechanical LODES are

being researched and developed today by academic groups and start-ups across
the globe and will reach full commercial bankability by the end of the next decade.
Their maturity will be timely with the next phase of large-scale deployment of
renewables on the grid. The many cheap hours of storage capacity will allow
to store excess renewables and avoid congestion during extended periods
of over-generation and to supply market needs and maintain a high quality of
electrical service during extended periods of renewable under-generation. The

19
It is important to note that although cell-level future Li-ion costs can be as low as $70/ kWh, pack
and plant total cost of ownership for a 12 year lifetime exceeds $100/kWh in future forecasts.

17
 5 | Energy Storage Systems

main question of this study is whether inexpensive LODES can make renewable
generation truly dispatchable at a reasonable cost, at the power plant as well as
at the grid level, considering here a deregulated market environment. Specifically,
the study focuses on quantitatively assessing whether LODES can substantially
reduce basis and volume risk of a renewable farm and make its output more
dispatchable at a reasonable cost.

The quantitative analysis outlined In Chapter 6 shows how LODES can be

conveniently integrated in a renewable rich system to neutralize the intermittency
problem, and offer a quantitative framework to evaluate the value of various
embodiments of LODES in managing risk in virtual PPAs.

18
6 Quantitative Analysis

6.1 | Overview
In this section, we explore the value of long duration storage in addressing
market risk factors through a quantitative analysis of long-term contracted
wind farms on a virtual PPA contract. We explore the effect of basis risk and
forecast uncertainty on the asset’s revenue distribution, with and without
storage. The following sections provide an overview of scope & methodology,
followed by a survey of results from the analysis. The quantitative analysis
here was performed through FormWare™, Form Energy’s proprietary asset
management analytics platform.

6.1.1 | Objectives

• Demonstrate the value of more than 4 hours of storage in merchant risk

management
• Provide a quantitative framework to assess the value of long duration
storage technologies in merchant risk management

6.1.2 | Scope

Virtual PPAs (described above) are one class of long-term contracted renewable
assets with large exposure to merchant risks. We focus this quantitative study
on wind farms on hub-settled virtual PPAs, operating in the Southwestern
Power Pool (SPP) footprint in a day-ahead and real-time market structure. The
historical data used for the analysis refer to two years of operations of three
wind farms operating in SPP. Day-ahead bidding of the wind farms is based
on state-of-the-art weather forecasts, resulting in an average energy output
uncertainty of about 15%. The average total energy availability of the farms
is 90%, and average energy curtailment for congestions is about 5%, with a
maximum of about 24%.

We also limit our focus on risk factors to two primary sources that are most
critical today: volume and basis risk. We capture volume risk exposure through
wind production and market price forecast uncertainty, and resulting penalties
when production falls short of day-ahead commitments. Basis risk is directly
reflected in the difference in price between the node and the hub, usually a
result of local congestion due to correlated wind production. Basis risk exposure
can be managed through physical storage by arbitering dispatch around nodal
congestion, effectively smoothing out wind delivery over time. Volume risk

19
 6 | Quantitative Analysis

exposure is managed through using storage to fill the shortages where wind
production is lower than forecasted and resulting DA bids.

In this study, we take a technology-agnostic view to long duration storage.

We model a wide range of long duration storage technology specifications
(+200 possibilities) as per the permutations from the following table.

Table 2
Roundtrip Efficiency
The full range of energy (RTE) [%] 30 40 50 60 70 80 90
storage technology
specifications modeled Energy CapEx
in this work. ($/kWh) 5 9 17 31 56 104 190
All combinations of these
specifications were Power CapEx
analyzed in this work.
($/kW) 100 178 316 562 1000

Many of the data points on this table are not commercially available today.
However, given the novel nature of this asset class and the extensive activity
in R&D and venture funding for new long duration storage technologies (e.g.
the United States ARPA-E DAYS program), we opted for a broad survey of the
design space, with the objective of providing an understanding of the value of
various technologies in addressing risk-management challenges. For illustrative
purposes, we provide a sample of detailed results for a storage specification that
corresponds to pumped-hydro storage, with an energy and power capex of $5/
kWh and $562/kW, and 80% RTE.

Finally, with deeper renewables penetration, we expect a rise in intermittency

and associated costs. These costs will in part be internalized by the market. In
one future trajectory, intermittency costs will manifest in larger basis and higher
penalties for missing DA (Day Ahead) commitments. To explore these future
scenarios, and given the difficulty of forecasting, we take a simple approach
where basis and penalties are increased linearly (i.e. a linear hourly increase
across the entire year).

For each case, we analyze the revenue distribution of each farm as is, then contrast
with the optimized revenue for the same asset with a co-located storage plant.

6.1.3 | Methodology

To model the wind farms in a hub-settled virtual PPA environment, we use a

two-step optimization approach that seeks to mirror the information available to
the operator trading the output into the markets, and to replicate the uncertainty
around future conditions as decisions are made. Energy storage sizing and
operation are focused on managing real-time risk factors only, in particular:

20
 6 | Quantitative Analysis

1_ Output shortages corresponding to times when wind generation falls below

DA commitments predicated on DA forecasts.
2 _ Congestion (basis) and other real-time price dynamics unfavorable correlated
with wind generation.

The end-to-end optimization process is implemented in Form Energy’s

proprietary asset management software, FormWare™, which optimizes asset
buildout and dispatch with an hourly resolution for a whole year based on a
linear programming framework.

A schematic of the process is shown below in figure 2.

Figure 2
• Wind, Forecasts • Wind, Updated Forecasts
The dynamic scheduling
• Day-Ahead Price, Forecasts • RT, Actual
framework used in this
work to capture the
• RT, Forecast Day-Ahead
trader’s viewpoint in a Commitment
day-ahead / Real-time
environment, with
• Wind, Forecasts (P50) • Wind, Multiple Forecasts
imperfect foresight.
• Day-Ahead Price, Actual • RT, Actual
• RT, Forecast • Multi-Scenario Optimization

There are two rounds of optimization with two decision horizons. In the first
round, which occurs 12 hours ahead of real time, DA commitments for the wind
farm are calculated under limited knowledge of wind production and real-time
prices: P5020 hourly daily forecasts are used for the former and a proprietary
regression of real-time prices is used for the latter. The output of that optimization
are DA bids for the wind farm. These bids are assumed to be all committed by
the market and the corresponding commitments can either be satisfied through
available production, or through market purchases with a penalty that reflects
DA/RT spreads (DART spread) and potential regulatory imposed penalties.

In the second round of optimization, which occurs in the day of operation,

energy storage sizing and dispatch are optimized to minimize penalties
from missing wind day-ahead commitments and to maximize real-time
price arbitrage value. Energy storage sizing is optimized considering several
possible real-time wind and price scenarios to make sure that the storage
system deployed is robust across several possible operating conditions.

The distribution of wind farm and storage revenue and costs across possible
wind and price scenarios are subsequently used to estimate average revenue
and maximum downside, as proxies for risk and return.

20
Refers to a forecast where the expected volumes have an exceedance probability of 50%.

21
 6 | Quantitative Analysis

In detail, the following steps are used:

1_ We first model the wind asset without any storage, and optimize bidding
decisions between the DA and real-time market. To simulate forecast
uncertainty, we handicap the optimization algorithm by providing as inputs
the P50 forecasts for wind volumes and market prices (and not the real data
sets) that would be available to a trader bidding for that power plant, so at
the time when DA commitments are decided. This approach reflects the fact
that operators place bids in the day-ahead market with a necessarily
imperfect and limited knowledge of the future production and prices.

2_ The DA commitments computed by the model in the first optimization step

are locked in as inputs to the second optimization step. These commitments
can either be satisfied through available production, or through market
purchases with a penalty that reflects DA/RT spreads (DART spread) and
potential regulatory imposed penalties.

3_ We expose the wind farm in the second case to 3 statistically representative

wind scenarios that are statistically representative of the volume forecast
used in the first optimization step. We explore the distribution of wind farm
revenues and compute estimates of average revenue and maximum
downside as proxies for return and risk.

When storage is added to the wind farm, an additional set of decision

variables are added to the optimization problem, corresponding to the storage
build size and hourly dispatch. Steps (1) and (2) of the optimization remain the
same, while in step (3), the storage build and dispatch profile are co-optimized
across the modelled scenarios. The distribution of wind farm and storage
revenue and costs are subsequently used to estimate average revenue and
maximum downside, as proxies for risk and return.

While storage could affect DA bid strategy, the present analysis focuses
on managing real-time risk factors only, in particular wind shortage and
corresponding penalties and real-time price dynamics. Future work may
incorporate storage in DA bid decisions.

6.1.4 | Data Sets

The data of the wind farms modeled in this work was provided by the generous
courtesy of the Enel Green Power team, and are here anonymized to mask
confidential information. They represent different project vintages across the
geography of SPP, capturing the growth in the PPA market, increase in wind
turbine capacity factor and evolution of contract terms. We model all the
farms to be on a simple contract for differences with no hedges, collars or
other financial instruments.

22
 6 | Quantitative Analysis

6.2 | Results
In this section, we explore the quantitative results from the modeling exercise.
First, we survey the performance of the wind farms without storage, to
understand how forecast uncertainty and congestion act to reduce wind farm
revenue. Next, we take a closer look at the temporal results of how storage
integrates with wind farms, when specifications correspond to “pumped-
hydro like” storage. Finally, we extend our evaluation to survey aggregate
results for a variety of storage specifications across a number of scenarios
corresponding to future penetration of renewables.

6.2.1 | Results without Storage

Figure 3 contrasts the 24-hour wind forecast available for trading decisions
with actual wind production for the same wind farm. The wind forecast tracks
production closely on average, in aggregate. However, the hourly forecast’s
ability to correctly predict hourly production is relatively low, and at times,
production is not just at odds with average forecast, but even with the +/-
40% probability envelope (see for example Jan 4th or Jan 9th).

+/- 40% Wind Probability Actual Wind Expected Wind

Figure 3
200
Contrasting wind
production with day-ahead
forecasts shows the
limits of hourly resolution
150
of current weather
forecasting data.
MW

100

50

0
Jan 03 Jan 04 Jan 05 Jan 06 Jan 07 Jan 08 Jan 09 Jan 10
2018

Forecast uncertainty manifests in missed day-ahead commitments, which

must be covered by market purchases of energy at a premium, or a penalty,
that captures the social costs of last-minute activation of expensive marginal
producers. We model penalties in the form:

P = (Qbid - QDA ) (PRT + α|PRT - PDA|)

where Q bid is the volume bid in day-ahead, QDA is the volume available for
DA at the time of dispatch, PRT and PDA are the real-time and day-ahead
prices, and α is a tuning parameter that captures an increase in penalties. The
structure of the penalties here is that any shortage in production against day-

23
 6 | Quantitative Analysis

ahead commitments is covered at the real-time market price and an additional

payment capturing the DA-RT spread (the DART), which is a reasonable hourly
proxy for the tightness of supply.

In the case of this wind farm with the forecasts and production in figure 3,
the mean annual revenue from the farm under the virtual PPA is around $18.6
mn, whereas the worst case scenario revenue is $9.2 mn, and the best case
revenue is $26.5 mn.

To capture the impact of storage on the risk & returns for merchant-exposed
wind farms, we use the mean and standard deviation of net present value
(NPV) as indices of profitability. For two projects with identical mean NPVs,
lower volatility of returns would indicate lower risk, and therefore should
result in a lower cost of capital. Although investors certainly care about both
downside and upside risk, infrastructure developers are generally risk-averse
and more focused on downside risk scenarios.

Therefore, we focus in this work on the normalized downside risk spread,

which corresponds to the difference in revenue between the mean and worst
case scenario, normalized by mean revenue. Given the value of storage in
addressing both downside and upside risk, this assumption is conservative.
For this farm, this value corresponds to a normalized downside risk of 51%.
The caveat is that the downside risk scenario modelled here was an aggressive
one that corresponds to a scenario where consistently low wind persists for
an entire year. With this context in mind, the results for downside risk can
be roughly interpreted to mean that almost half the project’s revenue is not
guaranteed due to forecast uncertainty and basis risk, under base conditions
(current market basis and penalty of ~25% of DART). For the two other wind
farms considered in this analysis, the normalized downside risk spread is 52%
and 56% under the base conditions.

6.2.2 | The Impact of Storage - Pumped Hydro Storage like Example

As shown in the previous section, a significant amount of revenue from the

wind farms is at risk in the virtual PPA environment, when exposed to basis and
forecast uncertainty. The co-optimization of energy storage will be shown here
to be an effective risk management tool. Before exploring the aggregate results,
it may be instructive to showcase an example of how the combined wind plus
storage behaves. In the following set of figures, we zoom in on a week in the
winter, with a focus on pumped-hydro like storage performance.

Figure 4 shows the dispatch for the wind farm and storage asset. Tracking the
DA commitments first, it is clear that they are a complex function of expected
production volumes and prices. On the 8th of January for example, large DA
commitments are made while production volumes drop significantly almost
abruptly (possibly due to a drop in temperature). This “white” gap corresponds
to a shortage where a penalty is paid. The storage asset does intervene and

24
 6 | Quantitative Analysis

provide parts of the shortage, entering the 9 th of January with a significantly

reduced state of charge.

In addition to covering missed commitments, the storage asset also actively

arbiters around congestion and prices. For example, the storage asset charges up
during the 7th of January when there is excess wind production over committed
in DA, accompanied by a drop in real-time prices. A similar but shorter cycle
can be tracked between January 3rd and January 5th, with a series of charge &
discharge cycles corresponding to congestion and price arbitrage activity.

Wind Discharge Charge DA Commitment

Figure 4
250
Sample results for 1 week
of dispatch from a wind 200
farm in SPP, operating in
a day-ahead / real-time
market structure, with 150
a co-optimized storage
MW

asset (specifications 100

corresponding to a
pumped-hydro-like energy
storage asset). 50

-50
Jan 03 Jan 04 Jan 05 Jan 06 Jan 07 Jan 08 Jan 09 Jan 10
2018

Figure 5 600
Sample results for 1
week showing the state
500
of charge evolution of
a co-optimized storage 400
asset (specifications
MWh

corresponding to a
300
pumped-hydro-like energy
storage asset) attached
to a wind farm in SPP, 200
operating in a day-ahead /
real-time market structure.
100

0
Jan 03 Jan 04 Jan 05 Jan 06 Jan 07 Jan 08 Jan 09 Jan 10
2018

Given the rich and multi-faceted nature of the wind-plus-storage co-

optimization, we switch gears to tracking the mean and downside risk
management impact of adding storage. Table 3 summarizes the results with
and without storage for the three wind farms. Pumped hydro-like storage
results in a modest impact of risk and return for farm (1), a sizeable impact on
farm (2), and a significant impact on farm (3).

25
 6 | Quantitative Analysis

Table 3 Wind Storage Optimal Optimal Mean virtual Worst Case Downside Downside
Farm technology Duration Power PPA NPV NPV risk risk
Summary of results for (hrs) (MW) (MM USD) (MM USD) variation
three wind farms in SPP,
operating with and without 1 No Storage N/A N/A 18.6 9.2 51% -
a storage asset (whose
Pumped
specifications correspond
Hydro 13 36 19.2 (+3%) 9.7 49% -2%
to a pumped-hydro like
storage technology),
2 No Storage N/A N/A 16.9 8.2 51% -
demonstrating the impact
of storage on the risks and Pumped
returns of the projects. Hydro 17 108 19.0 (+12%) 10.2 46% -6%

3 No Storage N/A N/A 13.9 6.1 56% -

Pumped
Hydro 23 106 20.7 (+49%) 11.1 46% -10%

For comparison, we also modeled one version of short duration storage

technology using state-of-the-art Lithium ion, with costs estimated for 2025.
We estimate the costs to be around $190/kWh and $178/kW at a 90% round-
trip efficiency (based on NREL’s plant-level cost estimates). As expected, the
optimal durations were short (<1.5 hrs) while the impact on mean returns and
downside risk was minimal for all farms, except for farm 3, where a short cycle
arbitrage unlocked 27% in mean returns though had a limited impact on risk.
These findings are consistent with our hypothesis that wind farm merchant risk
management requires longer durations of storage because of the fundamental
structure of forecast uncertainty and wind correlated congestion patterns.

Table 4 Wind Storage Optimal Optimal Mean virtual Worst Case Downside
Farm technology Duration Power PPA NPV NPV risk
Summary of results (hrs) (MW) (MM USD) (MM USD)
for three wind farms
in SPP, operating 1 2020 Li 1 2 18.9 (+2%) 9.2 51% (-0%)
with a storage asset
whose specifications 2 2020 Li 1 13.5 17.0 (+1%) 8.3 51% (-1%)
correspond to currently
available short duration 3 2020 Li 1.4 137 17.7 (+27%) 8.5 52% (-6%)
storage technology
using state-of-the-art
lithium ion technology,
demonstrating the limited
impact of this storage 6.2.3 | Sensitivity Analysis
technology on the
risks and returns of the The results above represent for the most part the current state of the world, where
projects.
wind farms are subject to relatively moderate congestion conditions and market
penalties for missing commitments. To explore how these results might change
as more renewables come online, we repeat the same analysis for a variety of
scenarios capturing a linear increase in basis over time (i.e. with the same shape,
the difference in prices between hub and node goes up by a linear fraction across
all hours). We also model scenarios with larger penalties, one mechanism markets
may use to price-in the externalities of intermittency onto market players.

26
 6 | Quantitative Analysis

Figure 6 Penalties None 0.25 - 1.0x DART 2.0x DART

Sensitivity analysis 100,000

contrasting the impact
of a pumped-hydro like 95,000 2018
storage technology (solid level basis
Mean Returns, $/MW

triangles) to without 90,000

storage (hollow triangle)
on the risk and returns
85,000
of a wind farm in SPP,
1.25x - 1.5x
operating in a day-ahead /
real-time market structure
80,000
under a contract-for-
differences arrangement. 75,000
Results show the impact
of storage across a 70,000 2x baseline
range of scenarios of
penalties and basis risks, 65,000
corresponding to potential
0.45 0.55 0.65 0.75 0.85
future scenarios of
renewables deployment. Normalized Downside Risk

With storage No storage

Figure 6 shows the sensitivity of the mean returns and normalized downside
risk to an increase in basis and penalties for wind farm (1). The impact of higher
penalties is clear: they act to increase the cost of forecast uncertainty, and
therefore, increase risks. On the other hand, the primary impact of basis is to
reduce mean profits by eating at wind production revenue. Storage (solid labels)
consistently improves mean returns over baseline and reduces downside risk.
Here, storage provides backup energy to mitigate forecast errors when they
are most expensive, and provides a time-shifting mechanism for production to
circumvent high congestion costs. Across the data points shown in this figure,
the optimal duration of storage is between 13 and 14 hours, with a strong
and positive monotonic relationship between penalty levels and total energy
storage required (in MWhs of capacity).

The impact of duration on returns can be further highlighted by computing the

returns for a range of durations, while fixing other cost and efficiency variables.
Figure 7 shows how incremental returns (the improvement in returns through
the addition of storage) are impacted by duration for a merchant-risk exposed
wind farm. For durations less than 5 hours, the impact of storage on returns
remains limited, then increases linearly to around 10 and plateaus at around
13 hours. Beyond 15 hours, incremental returns drop linearly, reflecting the
limited marginal value of additional storage and the pressure the storage CapEx
imposes on returns. In the simulations shown in this paper, the optimization
algorithm automatically determines the optimal duration that maximizes
returns, and we report results based on this optimal configuration.

27
 6 | Quantitative Analysis

Figure 7
2.5
Incremental returns fall
off rapidly when more
than ~10% away from the
optimal ESS size.
Incremental Returns, k$/MW

2.0

1.5

1.0

5 10 15 20
Hours

It is instructive here to contrast the above results with those of short duration
storage, exemplified by 2025 Lithium ion technology, shown in figure 8. The
impact of basis and penalty on the baseline results remains the same. However,
it is evident in this case storage has a weaker impact on both mean returns and
downside risk. The optimal duration of storage deployed across the cases here
is around 1 hour. The long duration nature of these applications is well captured
by the difference in impact on both risk and returns by the “pumped-hydro like”
storage as compared with 2025 Lithium ion technology.

Figure 8 Penalties None 0.25 - 1.0x DART 2.0x DART

Sensitivity analysis 95,000

contrasting the impact
of today’s lithium ion 2018
90,000
short duration storage level basis
Mean Returns, $/MW

technology (solid triangles)

to without storage (hollow 85,000
triangle) on the risk and
returns of a wind farm
80,000 1.25x - 1.5x
in SPP, operating in a
day-ahead / real-time
market structure under a 75,000
contract-for-differences
arrangement. Results show
70,000 2x baseline
the limited ability currently
available short duration to
manage volume and basis 65,000
risks, corresponding to
0.45 0.55 0.65 0.75 0.85
potential future scenarios
of renewables deployment. Normalized Downside Risk

With storage No storage

28
 6 | Quantitative Analysis

To explore whether these findings were consistent across potential storage

technologies, we repeat the above analysis for all the combinations explored in
section 6.1.2 In each case, we optimize the wind farm dispatch in day-ahead and
real-time, accounting for penalties and the basis risk arising from price differences
between the settlement hub and the production node, computing the optimal
storage duration (if any) in the process. We run this analysis for the current state
of a wind farm in SPP, as well as an extrapolated future case where basis risk
increases by 50% and penalties for missing forecasts rise to 0.5 DART. Figures
9 and 10 show the variation of returns and risks versus optimal durations for a
variety of energy capex levels. Two clear trends emerge. The first trend is that
longer duration storage is correlated with an improvement in both returns and
risks, with a tendency for the impact to be maximized around 9-11 hours. The
second trend is that longer durations are optimal when energy capex costs are
lower, as one might expect.

Figure 9 E CapEx, $/MWh

The correlation between

optimal duration and 20 150,000
incremental profits and
risks, as well as the 100,000
Inc. Profit, k$/MW

underlying energy capex

costs (in $/kWh) and
15
resulting project size 50,000
(in MW), for an SPP
wind farm under today’s 10
conditions of basis risk ESS Power, MW
and penalties. 50
100
5 150

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Duration, H

0.00

-0.05
Inc. Risk

-0.10

-0.15

1 2 3 4 5 6 7 8 9 10 11 12 13 14
Duration, H

29
 6 | Quantitative Analysis

Figure 10
25 E CapEx, $/MWh

The correlation between

optimal duration and 150,000
incremental profits and 20
risks, as well as the 100,000
Inc. Profit, k$/MW

underlying energy capex

costs (in $/kWh) and
15 50,000
resulting project size
(in MW), for an SPP
wind farm under future
conditions of basis risk 10 ESS Power, MW
and penalties. 50
100
150
5

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Duration, H

0.00

-0.05
Inc. Risk

-0.10

-0.15
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Duration, H

To further resolve the value of storage in risk management, and the design trade-
offs involved, the following set of surface plots provide a different view of energy
capex, roundtrip efficiency, again for the same wind farm under current and future
basis & penalty trajectories, with depth of color representing the incremental
profits and risks associated with combining storage with a merchant risk exposed
wind farm. The contours plots show results corresponding to the lowest power
capex level included in the design space, $100/kW. Consistently, the results show
a clear relationship between lower energy capex costs (which drive longer optimal
durations) and higher incremental profits as well as lower incremental risks, with
roundtrip efficiency contributing on the margin. This trend is consistent for current
and future scenarios of basis risk & penalties.

30
 6 | Quantitative Analysis

Figure 11 Max Inc Profits k$/MW

Contour plot of the impact

of energy capex costs 180 22
(in $/kWh) and roundtrip
efficiency on incremental
profits for an SPP wind 20
160
farm under current 4
conditions of basis risk 18
and penalties. 140 2
16
ESS Energy Capex, $/kWh

120

Inc Profits k$/MW

14

100 12
6
10
80
10 8
60
6

40 8 4
12 14 16
20 2
18 20
0.4 0.5 0.6 0.7 0.8 0.9
Round Trip Efficiency

Figure 12 Min Inc Risk

Contour plot of the impact

of energy capex costs 180
(in $/kWh) and roundtrip
efficiency on incremental −0.02
risks for an SPP wind farm 160
04
under current conditions −0.
of basis risk and penalties. −0.04
140 −0.02
ESS Energy Capex, $/kWh

−0.06
120
Inc Risk

100 −0.08

6
80 −0.0 −0.08 −0.1

60 −0.12

40 −0.14

1 2
20 −0. −0.1 −0.14 −0.16

0.4 0.5 0.6 0.7 0.8 0.9

Round Trip Efficiency

31
 6 | Quantitative Analysis

Figure 13 Max Inc Profits k$/MW

Contour plot of the impact

of energy capex costs 180 24
(in $/kWh) and roundtrip
efficiency on incremental 2 4 22
profits for an SPP wind 160
farm under future
20
conditions of basis risk
and penalties. 140
18
ESS Energy Capex, $/kWh

120 16

Inc Profits k$/MW

14
100 8
6 12
80
10

60 8
12 14
6
40
10 4
20 18 20
16 2

0.4 0.5 0.6 0.7 0.8 0.9

Round Trip Efficiency

Figure 14 Min Inc Risk

Contour plot of the impact

of energy capex costs 0
180
(in $/kWh) and roundtrip
−0.01
2
efficiency on incremental
01 −0.0
risk for an SPP wind farm 160 −0. −0.02
under future conditions of
basis risk and penalties. −0.03
140
−0.04
ESS Energy Capex, $/kWh

4
120 −0.0 −0.05
−0.06
Inc Risk

100
−0.07
3
−0.0 −0.06 −0.08
80 −0.05
−0.09
60 −0.1
−0.08 −0.09 −0.11
40
−0 .07 −0.12

20 −0.13
−0.11 −0.12
−0.1 −0.14
0.4 0.5 0.6 0.7 0.8 0.9
Round Trip Efficiency

32
 6 | Quantitative Analysis

The trends are also consistent when higher power capex surfaces are showcased
in the contour plots. Figures 15 and 16 show the contour plots selected for a
specific power capex level, $562/kW. As expected, the impact on incremental
profits and risks is much less pronounced given the higher total cost of the storage
embodiments highlighted in this surface slice.

Figure 15 Max Inc Profits k$/MW

Contour plot of the

impact of energy capex 16 4
costs (in $/kWh) and
roundtrip efficiency on
3.5
incremental profits for
an SPP wind farm under 14
0.5
current conditions of basis 3
ESS Energy Capex, $/kWh

risk and penalties, with a

Inc Profits k$/MW

surface slice at the $562/
12 2.5
kW power capex level.

2
10 2

1.5
1.5
8 1
3
1
0.5
6 2.5
3.5

0.7 0.75 0.8 0.85 0.9

Round Trip Efficiency
Figure 16 Min Inc Risk

Contour plot of the impact 0

of energy capex costs 16
(in $/kWh) and roundtrip
−0.005
efficiency on incremental
risks for an SPP wind farm
under current conditions 14 −0.01
of basis risk and penalties, 5
00
ESS Energy Capex, $/kWh

with a surface slice at the .

−0 −0.015
$562/kW power capex
level. 12
Inc Risk

−0.02
15
.0
10 −0
−0.025

−0.03
8
5
.02 −0.035
1 −0
0.0 2
6 −
−0.0
−0.04

0.7 0.75 0.8 0.85 0.9

Round Trip Efficiency

33
 6 | Quantitative Analysis

The results are persistent for the other wind farms modeled in this work,
although the optimal durations, the risk / return trends and their exact levels
vary with the price signal, the strike price of the virtual PPA and the correlation
of wind production with the basis and price signals. Figures 17 – 20 show the
results for the other two wind farms across the storage specification space
demonstrating similar trends, with varying risk and return levels, depending on
the wind farm conditions.
Figure 17 Max Inc Profits k$/MW

350
Contour plot of the impact
of energy capex costs
20 30 110
(in $/kWh) and roundtrip
efficiency on incremental 300
100
risks for SPP wind
farm (2) under current
250 90
conditions of basis risk
ESS Energy Capex, $/kWh

40
and penalties.
80

Inc Profits k$/MW

200
70

150 60

50
100 60
50 40

30
50 70 80

90 20
100
0.4 0.5 0.6 0.7 0.8 0.9
Round Trip Efficiency
Figure 18 Min Inc Risk

350
Contour plot of the impact
−0.08
of energy capex costs
.1
(in $/kWh) and roundtrip −0 −0.1
efficiency on incremental 300 .14
−0 −0.12
risks for SPP wind −0.1
2
farm (2) under current −0.14
conditions of basis risk 250
ESS Energy Capex, $/kWh

and penalties. −0.18 −0.16

−0.18
200 16
−0.
Inc Risk

−0.2

−0.22
2
150 −0. −0.22 −0.24

−0.26
100
−0.26 −0.28

−0.3
50
.24 −0.32
−0 −0.3
−0.28 −0.34
0.4 0.5 0.6 0.7 0.8 0.9
Round Trip Efficiency

34
 6 | Quantitative Analysis

Figure 19 Max Inc Profits k$/MW

Contour plot of the impact

of energy capex costs 180 30
(in $/kWh) and roundtrip
efficiency on incremental
returns for SPP wind farm 160
(3) under current conditions 25
of basis risk and penalties.
140
5
ESS Energy Capex, $/kWh

20
120

Inc Profits k$/MW

100
15

80 10

10
60

40
5
15 20
20
25

0.4 0.5 0.6 0.7 0.8 0.9

Round Trip Efficiency

Figure 20 Min Inc Risk

Contour plot of the impact

of energy capex costs 180
(in $/kWh) and roundtrip
02 −0.02
efficiency on incremental −0.
risks for SPP wind 160
farm (3) under current −0.04
conditions of basis risk
6
and penalties. 140 04 −0.0
−0. −0.06
ESS Energy Capex, $/kWh

120
−0.08
Inc Risk

100
−0.1

80 8
−0.0 −0.12
−0.1
60
−0.14

40
−0.16

20
−0.12 −0.18
−0.14 −0.16
0.4 0.5 0.6 0.7 0.8 0.9
Round Trip Efficiency

35
 6 | Quantitative Analysis

6.3 | Summary of Storage Impact

In summary, the quantitative analysis presented in this section provides a
methodology for dispatch optimization for real-time and day-ahead operation
and quantifies asset exposure to volume risk (due to forecast uncertainty) and
basis risk. Building on this methodology, we provide a quantitative framework
to evaluate the ability of storage to improve utility scale project risk and
return, across a variety of storage technologies. We use this framework to
understand storage specifications necessary to achieve target levels of risk
management for a variety of wind farms. We show how a pumped-hydro
like storage technology will result in optimal durations between 13 and 23
hours and offer a significant improvement in risk and returns, in contrast with
today’s short duration storage which results in 1-2 hours of optimal durations
and much more limited ability to modulate risk and returns. The specific target
performance criteria for storage will depend on a specific investor’s risk and
return preferences.

36
7 Conclusions

In this work, we highlight the impact of corporate PPAs as a major driver for
renewables penetration, and their value in reducing the cost of capital for new
projects. With deeper renewables penetration, the costs of intermittency are
increasing. In a deregulated market environment, these costs manifest as
increased volume and basis risks that are borne by developers and off-takers.
Effective management of increased intermittency risks is critical, if we are to
see continued expansion of corporate PPAs as a driving force for renewable
deployment.

We focus here on the use of physical storage technologies for risk management;
specifically, we look at effective risk management via novel long duration storage
technologies. We utilize Form’s asset optimization software, FormWare™, to
offer a flexible and technology-agnostic methodology to simulate financially-
settled and long-term contracted wind farms operating in a day-ahead / real-time
market structure and we explore the distribution of asset returns, using real data
from anonymized wind farms owned and operated by Enel Green Power.

Through this analysis, we provide a quantitative framework to demonstrate

the ability of storage to manage risk and return for wind farms exposed to
volume and basis risk factors. We explore a range of representative storage
specifications, using a pumped-hydro like storage technology to showcase
the risk management capabilities of this emerging asset class. The results are
persistent across the wind farms modeled and demonstrate the impact of bulk
energy storage technologies to effectively manage risks and maximize returns.
While general trends emerge, the specific performance thresholds for storage
will vary by project configuration and technology. To our knowledge, this paper
represents the first attempt in the literature to jointly quantify risk and return of
renewable and storage assets operated in a realistic day-ahead and real-time
market structure with imperfect foresight, in an attempt to more accurately
capture the trader’s perspective.

The future of renewable energy is bright and windy, and though the rising
variability risks present a roadblock on the path to complete decarbonization,
we are confident that the next wave of innovation in storage technologies, and
specifically in ultra-low-cost, long duration technologies, will address emerging
concerns and enable the rapid pace needed to manage the greatest challenge
of our generation: matching economic prosperity and environment protection.

37
enelfoundation.org

38