LBNL-1000917

Utility-Scale Solar 2014

An Empirical Analysis of Project Cost, Performance,
and Pricing Trends in the United States

Authors: Mark Bolinger and Joachim Seel

Energy Analysis & Environmental Impacts Division, Lawrence Berkeley National Laboratory

Table of Contents
Executive Summary ....................................................................................................................i
1. Introduction ...........................................................................................................................1
2. Technology Trends Among the Project Population ...........................................................5
PV (194 projects, 6,236 MWAC) ................................................................................................................................. 6
CSP (15 projects, 1,673 MWAC)...............................................................................................................................11
3. Installed Prices ....................................................................................................................12
PV (170 projects, 5,874 MWAC, including 2 CPV projects totaling 35 MWAC) .........................................................13
CSP (6 projects, 1,270 MWAC).................................................................................................................................19
4. Operation and Maintenance Costs ....................................................................................20
5. Capacity Factors .................................................................................................................22
PV (128 projects, 3,201 MWAC) ...............................................................................................................................22
CPV (2 projects, 35 MWAC) ......................................................................................................................................26
CSP (13 projects, 1,390 MWAC)...............................................................................................................................27
6. Power Purchase Agreement (“PPA”) Prices .....................................................................29
7. Conclusions and Future Outlook .......................................................................................38
References ................................................................................................................................41

List of Figures
Figure 1. Historical and Projected PV Capacity by Sector in the United States.............................................................. 1
Figure 2. Capacity Shares of PV Module and Mounting Configurations by Installation Year ......................................... 7
Figure 3. Map of Global Horizontal Irradiance (GHI) and Utility-Scale Solar Project Locations ..................................... 8
Figure 4. Trends in Global Horizontal Irradiance by Mounting Type and Installation Year ............................................. 9
Figure 5. Trends in Inverter Loading Ratio by Mounting Type and Installation Year ....................................................10
Figure 6. Installed Price of Utility-Scale PV and CPV Projects by Installation Year......................................................13
Figure 7. Installed Price of Utility-Scale PV and CPV Projects by Project Design and Installation Year ......................14
Figure 8. Installed Price of 2014 PV Projects by Size and Project Design ...................................................................16
Figure 9. Installed Price of Utility-Scale CSP Projects by Technology and Installation Year ........................................19
Figure 10. Empirical O&M Costs Over Time .................................................................................................................21
Figure 11. Cumulative PV Capacity Factor by Project Vintage: 2010-2013 Projects Only ..........................................23
Figure 12. 2014 PV Capacity Factor by Project Vintage: 2010-2013 Projects Only ....................................................24
Figure 13. Cumulative PV Capacity Factor by Resource Strength, Fixed-Tilt vs. Tracking, Inverter Loading Ratio, and
Module Type ................................................................................................................................................25
Figure 14. Capacity Factor of CSP Projects (Solar Portion Only) Over Time ...............................................................27
Figure 15. Levelized PPA Prices by Technology, Contract Size, and PPA Execution Date .........................................31
Figure 16. Levelized PPA Prices by Operational Status and PPA Execution Date ......................................................33
Figure 17. Generation-Weighted Average PV PPA Prices Over Time by Contract Vintage .........................................35
Figure 18. Average PV PPA Prices and Natural Gas Fuel Cost Projections Over Time ..............................................36
Figure 19. Levelized PV PPA Prices by Contract Vintage ............................................................................................37
Figure 20. Solar and Other Resource Capacity in 35 Selected Interconnection Queues .............................................39
 List of Acronyms
AC……………………………………………………. Alternating Current
c-Si…………………………………………………… Crystalline Silicon
COD………………………………………………….. Commercial Operation Date
CPV………………………………………………….. Concentrated Photovoltaics
CSP…………………………………………………... Concentrated Solar (Thermal) Power
DC……………………………………………………. Direct Current
DIF…………………………………………………….. Diffuse Horizontal Irradiance
DNI…………………………………………………… Direct Normal Irradiance
DOE………………………………………………….. U.S. Department of Energy
EIA…………………………………………………… Energy Information Agency
EPC…………………………………………………… Engineering, Procurement & Construction
FERC…………………………………………………. Federal Energy Regulatory Commission
GHI……………………………………………………. Global Horizontal Irradiance
FiT…………………………………………………….. Feed-in Tariff
ILR……………………………………………………. Inverter Loading Ratio
ISO……………………………………………………. Independent System Operator
ITC……………………………………………………. Investment Tax Credit
LBNL…………………………………………………. Lawrence Berkeley National Laboratory
LCOE…………………………………………………. Levelized Cost of Energy
MW…………………………………………………... Megawatt(s)
NCF………………………………………………….. Net Capacity Factor
NREL………………………………………………… National Renewable Energy Laboratory
O&M…………………………………………………. Operation and Maintenance
PII……………………………………………………. Permitting, Interconnection & Inspection
PPA…………………………………………………... Power Purchase Agreement
PV……………………………………………………. Photovoltaics
REC………………………………………………….. Renewable Energy Credit
RTO………………………………………………….. Regional Transmission Organization
SEGS………………………………………………… Solar Energy Generation Systems
TOD………………………………………………….. Time-Of-Delivery
 Executive Summary
Other than the nine Solar Energy Generation Systems (“SEGS”) parabolic trough projects built in the
1980s, virtually no large-scale or “utility-scale” solar projects – defined here to include any ground-
mounted photovoltaic (“PV”), concentrating photovoltaic (“CPV”), or concentrating solar thermal
power (“CSP”) project larger than 5 MW AC – existed in the United States prior to 2007. By 2012 – just
five years later – utility-scale had become the largest sector of the overall PV market in the United
States, a distinction that was repeated in both 2013 and 2014 and that is expected to continue for at least
the next few years. Over this same short period, CSP also experienced a bit of a renaissance in the
United States, with a number of large new parabolic trough and power tower systems – some including
thermal storage – achieving commercial operation.

With this critical mass of new utility-scale projects now online and in some cases having operated for a
number of years (generating not only electricity, but also empirical data that can be mined), the rapidly
growing utility-scale sector is ripe for analysis. This report, the third edition in an ongoing annual
series, meets this need through in-depth, annually updated, data-driven analysis of not just installed
project costs or prices – i.e., the traditional realm of solar economics analyses – but also operating costs,
capacity factors, and power purchase agreement (“PPA”) prices from a large sample of utility-scale solar
projects in the United States. Given its current dominance in the market, utility-scale PV also dominates
much of this report, though data from CPV and CSP projects are presented where appropriate.

Some of the more-notable findings from this year’s edition include the following:
• Installation Trends: Among the total population of utility-scale PV projects from which data
samples are drawn, several trends are worth noting due to their influence on (or perhaps
reflection of) the cost, performance, and price data analyzed later. For example, the use of
tracking devices (overwhelmingly single-axis, though a few dual-axis tracking projects entered
the population in 2014) continues to expand, particularly among thin-film (CdTe) projects, which
had almost exclusively opted for fixed-tilt mounts prior to 2014. The quality of the solar
resource in which PV projects are being built in the United States has increased on average over
time, as most of the projects in the population (>90% in MW terms) are located in the Southwest
where the solar resource is the strongest. That said, the market has also begun to expand outside
of the Southwest, most notably in the Southeast. The average inverter loading ratio – i.e., the
ratio of a project’s DC module array nameplate rating to its AC inverter nameplate rating – has
also increased among more recent project vintages, as oversizing the array can boost revenue,
particularly when time-of-delivery pricing is used. In combination, these trends should drive AC
capacity factors higher among more recently built PV projects (a hypothesis confirmed by the
capacity factor data analyzed in Chapter 5). Finally, 2014 also saw three new large CSP projects
– i.e., two 250 MW trough projects and one 377 MW solar power tower project – achieve
commercial operation; in contrast, no new CPV plants came online in 2014.
• Installed Prices: Median installed PV project prices within a sizable sample have steadily fallen
by more than 50% since the 2007-2009 period, from around $6.3/W AC to $3.1/W AC (or
$5.7/W DC to $2.3/W DC , all in 2014 dollars) for projects completed in 2014. The lowest-priced
projects among our 2014 sample of 55 PV projects were ~$2/W AC , with the lowest 20th
percentile of projects having fallen considerably from $3.2/W AC in 2013 to $2.3/W AC in 2014.

i
 The three large CSP projects that came online in 2014 were priced considerably higher than our
PV sample, ranging from $5.1/W AC to $6.2/W AC .
• Operation and Maintenance (“O&M”) Costs: What limited empirical O&M cost data are
publicly available suggest that PV O&M costs appear to have been in the neighborhood of
$20/kW AC -year, or $10/MWh, in 2014. CSP O&M costs are higher, at around $40-$50/kW AC -
year. These numbers include only those costs incurred to directly operate and maintain the
generating plant, and should not be confused with total operating expenses, which would also
include property taxes, insurance, land royalties, performance bonds, various administrative and
other fees, and overhead.
• Capacity Factors: The capacity-weighted average cumulative capacity factor across the entire
PV project sample is 27.5% (median = 26.5% and simple average = 25.6%), but individual
project-level capacity factors exhibit a wide range (from 14.8% to 34.9%) around these central
numbers. This variation is based on a number of factors, including (in approximate decreasing
order of importance): the strength of the solar resource at the project site; whether the array is
mounted at a fixed tilt or on a tracking mechanism; the inverter loading ratio; and the type of
modules used (e.g., c-Si versus thin film). Improvements in the first three of these factors have
driven capacity-weighted average capacity factors higher by project vintage over the last three
years – e.g., 29.4% among 2013-vintage projects, compared to 26.3% and 24.5% for projects
built in 2012 and 2011, respectively. In contrast, two of the new CSP projects built in recent
years – a trough project with storage and a power tower project – generated lower-than-expected
capacity factors in 2014, reportedly due to startup and teething issues. Performance has
subsequently improved at both projects during the first six months of 2015 (compared to the
same period in 2014). Likewise, the two CPV projects in our sample seem to be
underperforming, relative to both similarly situated PV projects and ex-ante expectations.
• PPA Prices: Driven by lower installed project prices, improving capacity factors, and – more
recently – the rush to build projects in advance of the scheduled reversion of the 30% investment
tax credit (“ITC”) to 10% in 2017, levelized PPA prices for utility-scale PV have fallen
dramatically over time, by a steady ~$25/MWh per year on average from 2006 through 2013,
with a smaller price decline of ~$10/MWh evident in the 2014 and 2015 samples. Some of the
most-recent PPAs in the Southwest have levelized PPA prices as low as (or even lower than)
$40/MWh (in real 2014 dollars). At these low levels – which appear to be robust, given the
strong response to recent utility solicitations – PV compares favorably to just the fuel costs (i.e.,
ignoring fixed capital costs) of natural gas-fired generation, and can therefore potentially serve as
a “fuel saver” alongside existing gas-fired generation (and can also provide a hedge against
possible future increases in fuel prices).

Looking ahead, the amount of utility-scale solar capacity in the development pipeline suggests continued
momentum and a significant expansion of the industry through at least 2016. For example, at the end of
2014, there was at least 44.6 GW of utility-scale solar power capacity making its way through
interconnection queues across the nation (though concentrated in California and the Southwest).
Though not all of these projects will ultimately be built, presumably those that are built will most likely
come online prior to 2017, given the scheduled reversion of the 30% ITC to 10% at the end of 2016.
Even if only a modest fraction of the solar capacity in these queues meets that deadline, it will still mean
an unprecedented amount of new construction in 2015 and 2016 – as well as a substantial amount of
new data to collect and analyze in future editions of this report.

ii
1. Introduction
The term “utility-scale solar” refers both to large-scale concentrating solar power (“CSP”)
projects that use several different technologies to produce steam used to generate electricity for
sale to utilities,1 and to large photovoltaic (“PV”) and concentrating photovoltaic (“CPV”)
projects that typically sell wholesale electricity directly to utilities, rather than displacing onsite
consumption (as has been the more-traditional application for PV in the commercial and
residential markets). Although utility-scale CSP has a longer history than utility-scale PV (or
CPV), 2 and has recently experienced a bit of a renaissance, 3 the utility-scale solar market in the
United States is now largely dominated by PV: there is currently significantly more PV than
CSP capacity either operating (6.4x), under construction (30.5x), or under development (12.1x)
in utility-scale projects (SEIA 2015). PV’s dominance follows explosive growth in recent years:
utility-scale PV has been the fastest-growing sector of the PV market since 2007, and since 2012
has accounted for the largest share of the overall PV market in terms of new MW installed (with
3,934 MW DC of new capacity added in 2014 alone – see Figure 1), a distinction that is projected
to continue through 2016 (GTM Research and SEIA 2015). 4
14,000 70,000
Utility
12,000 60,000
Commercial
Annual Capacity Additions (MWDC)

Residential

Cumulative Capacity (MWDC)

10,000 50,000
2,307

6,807
8,000 40,000
Columns show annual capacity additions, 1,651
area shows cumulative capacity.
6,000 4,825 1,399 30,000

1,082
3,934
4,000 20,000
2,855
1,803
2,000 10,000
267 760
22 70
9

0 0
2007 2008 2009 2010 2011 2012 2013 2014 2015E 2016E 2017E 2018E 2019E 2020E

Source: GTM/SEIA (2010-2015), Tracking the Sun Database

Figure 1. Historical and Projected PV Capacity by Sector in the United States

1
Operating CSP projects most commonly use either parabolic trough or, more recently, power tower technology.
CSP projects using other technologies, including compact linear Fresnel lenses and Stirling dish engines, have also
been built in the United States, but largely on a pre-commercial prototype basis.
2
Nine large parabolic trough projects totaling nearly 400 MWAC have been operating in California since the late
1980s/early 1990s, whereas it was not until 2007 that the United States saw its first PV project in excess of 5 MWAC.
3
More than twice as much CSP capacity came online in the United States in 2013/2014 as in the previous 28 years.
4
GTM/SEIA’s definition of “utility-scale” reflected in Figure 1 is not entirely consistent with how it is defined in
this report (see the text box – Defining “Utility-Scale” – in this chapter for a discussion of different definitions of
“utility-scale”). In addition, the capacity data in Figure 1 are expressed in DC terms, which is not consistent with
the AC capacity terms used throughout the rest of this report (the text box – AC vs. DC – at the start of Chapter 2
discusses why AC capacity ratings make more sense than DC for utility-scale projects). Despite these two
inconsistencies, the data are nevertheless useful for the basic purpose of providing a general sense for the size of the
utility-scale market (both historical and projected) and demonstrating relative trends between market segments.

1
This rapidly growing utility-scale sector of the solar market is ripe for analysis. Historically,
empirical analyses of solar economics have focused primarily on up-front installed costs or
prices, and principally within the residential and commercial PV sectors (see, for example,
Barbose and Darghouth 2015). But as more utility-scale projects have come online and begun to
acquire an operating history, a wealth of other empirical data has begun to accumulate as well.
Utility-scale solar projects can be mined for data on not only installed prices, but also project
performance (i.e., capacity factor), operation and maintenance (“O&M”) costs, and power
purchase agreement (“PPA”) prices ($/MWh) – all data that are often unavailable publicly, and
are also somewhat less meaningful, 5 within the residential and commercial sectors.

This report is the third edition in an ongoing annual series that, each year, compiles and analyzes
the latest empirical data from the growing fleet of utility-scale solar projects in the United States.
In this third edition, we maintain our definition of “utility-scale” to include any ground-mounted
project with a capacity rating larger than 5 MW AC (the text box below describes the challenge of
defining “utility-scale” and provides justification for the definition used in this report). Within
this subset of solar projects, the relative emphasis on different solar technologies within the
report largely reflects the distribution of those technologies in the broader market – i.e., most of
the data and analysis naturally focuses on PV given its large market share (78% of cumulative
installed capacity), but CPV (<1%) and CSP (21%) projects are also included where useful data
are available.

The report proceeds as follows. First, Chapter 2 describes key characteristics of the overall
utility-scale solar project population from which the data samples that are analyzed in later
chapters are drawn, with a goal of identifying underlying technology trends that could potentially
influence trends in the data analyzed in later chapters. The remainder of the report analyzes the
cost, performance, and price data samples in a logical order: up-front installed costs or prices are
presented in Chapter 3, followed by ongoing operating costs and performance (i.e., capacity
factor) in Chapters 4 and 5, all of which influence the PPA prices that are reported and analyzed
in Chapter 6. Chapter 7 concludes with a brief look ahead.

Data sources are diverse and vary by chapter depending on the type of data being presented, but
in general include the Federal Energy Regulatory Commission (“FERC”), the Energy
Information Administration (“EIA”), state and federal incentive programs, state and federal

5
For example, even if performance data for residential systems were readily available, they might be difficult to
interpret given that residential systems are often partly shaded or otherwise constrained by roof configurations that
are at sub-optimal tilt or azimuth. Utility-scale projects, in contrast, are presumably less constrained by existing site
conditions and better able to optimize these basic parameters, thereby generating performance data that are more
normalized and easier to interpret. Similarly, even if known, the price at which third-party owners of residential PV
systems sell electricity to site hosts is difficult to interpret, not only because of net metering and other state-level
incentives that can affect the price, but also because residential PPAs are often priced only as low as they need to be
in order to present an attractive value proposition relative to retail electricity prices (this is known as “value-based
pricing”). In contrast, utility-scale solar projects must often compete (policy incentives notwithstanding) for PPAs
against other generating technologies within competitive wholesale power markets, and therefore tend to offer PPA
prices that reflect the minimum amount of revenue needed to recoup the project’s initial cost, cover ongoing
operating expenses, and provide a normal rate of return (this is known as “cost-plus” pricing). Whereas cost-plus
pricing data provide useful information about the amount of revenue that solar needs in order to be economically
viable in the market, value-based PPA price data are somewhat less useful in this regard, in that they often reflect
the “price to beat” more than the lowest possible price that could be offered.

2
 regulatory commissions, industry news releases, trade press articles, and communication with
project owners and developers. Sample size also varies by chapter, and not all projects have
sufficiently complete data to be included in all data sets. All data involving currency are
reported in constant or real U.S. dollars – in this edition, 2014 dollars 6 – and all PPA price
levelization uses a 7% real annual discount rate.

Defining “Utility-Scale”
Determining which electric power projects qualify as “utility-scale” (as opposed to commercial- or residential-scale) can be a
challenge, particularly as utilities begin to focus more on distributed generation. For solar PV projects, this challenge is exacerbated
by the relative homogeneity of the underlying technology. For example, unlike with wind power, where there is a clear difference
between utility-scale and residential wind turbine technology, with solar, the same PV modules used in a 5 kW residential rooftop
system might also be deployed in a 100 MW ground-mounted utility-scale project. The question of where to draw the line is,
therefore, rather subjective. Though not exhaustive, below are three different – and perhaps equally valid – perspectives on what is
considered to be “utility-scale”:

• Through its Form 860, the Energy Information Administration (“EIA”) collects and reports data on all generating plants
larger than 1 MW, regardless of ownership or whether interconnected in front of or behind the meter (note: this report
draws heavily upon EIA data for such projects).

• In their Solar Market Insight reports, Greentech Media and SEIA (“GTM/SEIA”) define utility-scale by offtake arrangement
rather than by project size: any project owned by or that sells electricity directly to a utility (rather than consuming it
onsite) is considered a “utility-scale” project. This definition includes even relatively small projects (e.g., 100 kW) that sell
electricity through a feed-in tariff (“FIT”) or avoided cost contract (Munsell 2014).

• At the other end of the spectrum, some financiers define utility-scale in terms of investment size, and consider only those
projects that are large enough to attract capital on their own (rather than as part of a larger portfolio of projects) to be
“utility-scale” (Sternthal 2013). For PV, such financiers might consider a 20 MW (i.e., ~$50 million) project to be the
minimum size threshold for utility-scale.

Though each of these three approaches has its merits, this report adopts yet a different approach: utility-scale solar is defined
herein as any ground-mounted solar project that is larger than 5 MWAC.

This definition is grounded in consideration of the four types of data analyzed in this report: installed prices, O&M costs, capacity
factors, and PPA prices. For example, setting the threshold at 5 MWAC helps to avoid smaller projects that are arguably more
commercial in nature, and that may make use of net metering and/or sell electricity through FiTs or other avoided cost contracts
(any of which could skew the sample of PPA prices reported in Chapter 6). A 5 MWAC limit also helps to avoid specialized (and
therefore often high-cost) applications, such as carports or projects mounted on capped landfills, which can skew the installed price
sample. Meanwhile, ground-mounted systems are more likely than roof-mounted systems to be optimally oriented in order to
maximize annual electricity production, thereby leading to a more homogenous sample of projects from which to analyze
performance, via capacity factors. Finally, data availability is often markedly better for larger projects than for smaller projects (in
this regard, even our threshold of 5 MWAC might be too small).

Some variation in how utility-scale solar is defined is natural, given the differing perspectives of those establishing the definitions.
Nevertheless, the lack of standardization does impose some limitations. For example, GTM/SEIA’s projections of the utility-scale
market (shown in Figure 1) may be useful to readers of this report, but the definitional differences noted above (along with the fact
that GTM/SEIA reports utility-scale capacity in DC rather than AC terms) make it harder to synchronize the data presented herein
with their projections. Similarly, institutional investors may find some of the data in this report to be useful, but perhaps less so if
they are only interested in projects larger than 20 MWAC.

Until consensus emerges as to what makes a solar project “utility-scale,” a simple best practice is to be clear about how one has
defined it (and why), and to highlight any important distinctions from other commonly used definitions – hence this text box.

6
Conversions between nominal and real dollars use the implicit GDP deflator. Historical conversions use the actual
GDP deflator data series from the U.S. Bureau of Economic Analysis, while future conversions (e.g., for PPA
prices) use the EIA’s projection of the GDP deflator in Annual Energy Outlook 2015.

3
Finally, we note that this report complements several other related studies and ongoing research
activities, all funded as part of the Department of Energy’s (“DOE”) SunShot Initiative, which
aims to reduce the cost of PV-generated electricity by about 75% between 2010 and 2020. For
reference, this related work is briefly described in the text box below.

Related National Lab Research Products

Utility-Scale Solar is produced in conjunction with several related and

ongoing research activities:

• Tracking-the-Sun is a separate annual report series produced by

LBNL that focuses on residential and commercial solar and
includes trends and analysis related to PV project pricing.

• The Open PV Project (openpv.nrel.gov) is an online data-

visualization tool developed by the National Renewable Energy
Laboratory (NREL) that incorporates data from Tracking the Sun
and Utility-Scale Solar.

• Photovoltaic System Pricing Trends: Historical, Recent, and Near-

Term Projections is an annual briefing produced jointly by NREL
and LBNL that provides a broad overview of PV pricing trends,
based on ongoing research activities at both labs.

• In-Depth Statistical Analyses of PV pricing data by researchers at

LBNL and several academic institutions seek to further illuminate
PV pricing dynamics and the underlying drivers, using more-
refined statistical techniques.

These and other solar energy publications are available at:

http://emp.lbl.gov/projects/solar

4
2. Technology Trends Among the Project Population

Before diving into project-level data on installed prices, operating costs, capacity factors, and
PPA prices, this chapter analyses trends in utility-scale solar project technology and
configurations among the entire population of projects from which later data samples are drawn.
This population consists of 209 ground-mounted PV, CPV and CSP projects, each larger than 5
MW AC and with an aggregate capacity of 7,910 MW AC , that had achieved full commercial
operation within the United States by the end of 2014. 7 The intent is to explore underlying
trends in the characteristics of this fleet of projects that could potentially influence the cost,
performance, and/or price data presented and discussed in later chapters. As with the data
samples explored in later chapters, the total project population is broken out and described here
by technology type – first PV (including CPV) and then CSP. For reasons described in the text
box below, all capacity numbers (as well as other metrics that rely on capacity, like $/W installed
prices) are expressed in AC terms, unless otherwise noted.

AC vs. DC: AC Capacity Ratings Are More Appropriate for Utility-Scale Solar

Because PV modules are rated under standardized testing conditions in direct current (“DC”) terms, PV project capacity is also
commonly reported in DC terms, particularly in the residential and commercial sectors. For utility-scale PV projects, however,
the alternating current (“AC”) capacity rating – measured by the combined AC rating of the project’s inverters – is more
relevant than DC, for two reasons:

1) All other conventional and renewable utility-scale generation sources (including concentrating solar power, or “CSP”) to
which utility-scale PV is compared are described in AC terms – with respect to their capacity ratings, their per-unit installed
and operating costs, and their capacity factors.

2) Utility-scale PV project developers have, in recent years, increasingly oversized the DC PV array relative to the AC capacity
of the inverters (described in more detail in this chapter, and portrayed in Figure 5). This increase in the “inverter loading
ratio” boosts revenue and, as a side benefit, increases AC capacity factors. In these cases, the difference between a project’s
DC and AC capacity ratings will be significantly larger than one would expect based on conversion losses alone, and since the
project’s output will ultimately be constrained by the inverters’ AC rating, the project’s AC capacity rating is the more
appropriate rating to use.

Except where otherwise noted, this report defaults to each project’s AC capacity rating when reporting capacity (MWAC),
installed costs or prices ($/WAC), operating costs ($/kWAC-year), and AC capacity factor.

7
With the exception of Chapter 6, which examines PPA prices for both online and planned projects, we do not
include projects that have not yet achieved full commercial operation, unless multiple years lie between consecutive
phases (in which case project development is more akin to the development of separate projects). One implication
of this approach is that projects are attributed in their entirety to the year in which their last phase comes online,
even though they may have been under construction (and even partially operating) for several years. We chose this
approach because certain important project characteristics (such as project prices) are usually only reported for a
project as a whole, rather than for its individual phases.

5
PV (194 projects, 6,236 MWAC)
At the end of 2014, 194 PV projects totaling 6,236 MW AC were fully online in the United States
and met the definition of utility-scale used in this report (ground-mounted and larger than 5
MW AC ). 8 These 194 projects, the first of which were installed in 2007, make up the total
population of PV projects from which data samples are drawn in later chapters of this report.
More than half of this capacity – i.e., 63 projects totaling 3,218 MW AC – achieved commercial
operation in 2014.

Figure 2 breaks out this capacity by module type and project configuration – i.e., projects that
use crystalline silicon (“c-Si”) versus thin-film modules, 9 and projects mounted at a fixed tilt
instead of on a tracking device that follows the position of the sun.10 Though thin-film modules
powered two-thirds of the new utility-scale PV capacity installed in 2010, c-Si projects
dominated in 2011, 2012, and 2013, accounting for 70% of all new utility-scale PV capacity
installed in those three years. This trend reversed yet again in 2014, however, when the 6 largest
projects built all used thin-film modules, resulting in a 70% market share.

Among the entire project sample that came online in 2014 (including both c-Si and thin-film
projects) the number of projects using solar tracking technologies increased slightly from 55% in
2013 to 58% in 2014. In capacity terms, however, tracking projects decreased to 41% of new
2014 capacity (from 56% in 2013) as the three largest 2014 projects (Topaz, Agua Caliente and
Desert Sunlight) all used fixed-tilt racking.

Notably, 12 of the 16 thin-film projects that came online in 2014 use single-axis tracking – a
significant departure from just 2 tracking thin-film projects built prior to 2014. This shift is
largely attributable to First Solar’s acquisition of RayTracker’s single-axis tracking technology
back in 2011; First Solar deployed this technology in all but its four largest projects in 2014. 11
Tracking has historically not been as common among thin-film projects, largely because the
lower efficiency of thin-film relative to c-Si modules requires more land area per nameplate MW
– an expense that is exacerbated by the use of trackers (that said, the efficiency of First Solar’s
CdTe modules has been increasing over time).

8
Because of differences in how “utility-scale” is defined (e.g., see the text box on page 3), the total amount of
capacity in the PV project population described in this chapter cannot necessarily be compared to other estimates
(e.g., from GTM Research and SEIA 2015) of the amount of utility-scale PV capacity online at the end of 2014.
9
Module manufacturer First Solar, which produces CdTe modules, accounts for all new thin-film capacity added to
the project population in 2014.
10
All but two of the PV projects in the population that use tracking systems use single-axis trackers (which track the
sun from east to west each day). In contrast, two recently built PV projects in Texas, along with the two CPV
projects and one CSP power tower project (described later), use dual-axis trackers (i.e., east to west daily and north
to south over the course of the year). For PV, where direct focus is not as important as it is for CPV or CSP, dual-
axis tracking is a harder sell than single-axis tracking, as the roughly 10% boost in generation (compared to single-
axis, which itself can increase generation by ~20%) often does not outweigh the incremental costs (and risk of
malfunction), depending on the PPA price.
11
The very large Topaz, Agua Caliente, and Desert Sunlight projects had all executed PPAs and were well under
development (and perhaps even construction) prior to the acquisition of RayTracker. The large Antelope Valley
project was in a similar position, but did manage to incorporate tracking in roughly 20% of the project.

6
 3,000
Annual and Cumulative Capacity (MWAC)

Tracking Thin-Film Columns show annual capacity additions,

lines show cumulative capacity.
2,500 Fixed-Tilt Thin-Film 2,431
Tracking c-Si
2,000 2,069
Fixed-Tilt c-Si

1,500

1,000
865

500 609

0
2007-2009 2010 2011 2012 2013 2014
Installation Year
Figure 2. Capacity Shares of PV Module and Mounting Configurations by Installation
Year

Figure 2 also breaks down the composition of cumulative installed capacity as of the end of
2014. Fixed-tilt thin-film (2,431 MW AC ) held a slight lead over tracking c-Si (2,069 MW AC , but
spread across more than twice as many projects), while fixed-tilt c-Si (865 MW AC ) and tracking
thin-film (609 MW AC ) followed more distantly. Overall, the total project population as of the
end of 2014 was split fairly evenly (in capacity terms) between fixed-tilt (55%) vs. tracking
(45%) projects, and thin-film (53%) vs. c-Si projects (47%).

Figure 3 overlays the location of every utility-scale solar project in the LBNL population
(including CPV and CSP projects) on a map of solar resource strength, as measured by global
horizontal irradiance (“GHI”). 12 Not surprisingly, most of the projects (and capacity) in the
population are located in the southwestern United States, 13 where the solar resource is the
strongest and where state-level policies (such as renewable portfolio standards, and in some
cases state-level tax credits) encourage utility-scale solar development. As shown, however,
utility-scale solar projects have also been built in various states along the east coast and in the
Midwest, where the solar resource is not as strong; these installations have largely been driven
by state renewable portfolio standards. Though there are obviously some exceptions, Figure 3
also shows a preponderance of tracking projects (both c-Si and, more recently, thin-film) in the
high-GHI Southwest, compared to primarily fixed-tilt c-Si in the lower-GHI East.

12
Global Horizontal Irradiance (GHI) is the total solar radiation received by a surface that is held parallel to the
ground, and includes both direct normal irradiance (DNI) and diffuse horizontal irradiance (DIF). DNI is the solar
radiation received directly by a surface that is always held perpendicular to the sun’s position (i.e., the goal of dual-
axis tracking devices), while DIF is the solar radiation that arrives indirectly, after having been scattered by the
earth’s atmosphere. The GHI data represent average irradiance from 1998-2009 (Perez 2012).
13
As of the end of 2014, the Southwest (defined rather liberally here to include CA, NV, AZ, UT, CO, NM, and TX)
accounted for 90% of the population’s cumulative PV capacity, and 96% of its CSP capacity.

7
Figure 3. Map of Global Horizontal Irradiance (GHI) and Utility-Scale Solar Project
Locations

While Figure 3 provides a static view of where and in what type of solar resource regime utility-
scale solar projects within the population are located, knowing when each of these projects was
built – and hence how the average resource quality of the project fleet has evolved over time – is
also useful, for example, to help explain any observed trend in project-level capacity factors by
project vintage (explored later in Chapter 5).

Figure 4 addresses this question by showing the capacity-weighted average GHI (in
kWh/m2/day) among PV projects built in a given year, both for the entire PV project population
(solid black line) and broken out by fixed-tilt vs. tracking projects. Across the entire population,
the average GHI has increased steadily over time, suggesting a relative shift in the population
towards projects located in the high-GHI Southwest. Although the capacity-weighted averages
for fixed-tilt and tracking projects are not too dissimilar, the 20th percentiles are markedly
different, with fixed-tilt projects stuck around 4 kWh/m2/day, in contrast to much higher (and
generally increasing by vintage) 20th percentile values for tracking projects. The wide
distribution of fixed-tilt projects reflects the fact that – as shown previously in Figure 3 – most
8
 projects in the lower-GHI regions of the United States are fixed-tilt, yet very large fixed-tilt
projects are also present in the high-GHI Southwest (often using CdTe thin-film technology,
perhaps due to its greater tolerance for high-temperature environments 14). Tracking projects,
meanwhile, are concentrated primarily in the Southwest.
6
Annual GHI (kWh/m2/day)

4
Fixed-Tilt PV
Tracking PV
Markers represent capacity-weighted averages, with 20th and 80th percentiles. All PV
3
2010 2011 2012 2013 2014
n=10 (175 MW-AC) n=35 (485 MW-AC) n=43 (948 MW-AC) n=38 (1,335 MW-AC) n=63 (3,218 MW-AC)
Installation Year
Figure 4. Trends in Global Horizontal Irradiance by Mounting Type and Installation Year

A second project-level characteristic that influences both installed project prices and capacity
factors is the inverter loading ratio (“ILR”), which describes a project’s DC capacity rating (i.e.,
the sum of the module ratings under standardized testing conditions) relative to its aggregate AC
inverter rating. 15 With the cost of PV modules having dropped precipitously in recent years (and
more rapidly than the cost of inverters), and with some utilities (particularly in California)
offering time-varying PPA prices that favor generation during certain daylight hours, including
late afternoon, many developers have found it economically advantageous to oversize the DC
array relative to the AC capacity rating of the inverters. As this happens, the inverters operate
closer to (or at) full capacity for a greater percentage of the day, which – like tracking – boosts
the capacity factor, 16 at least in AC terms (this practice will actually decrease the capacity factor
in DC terms, as some amount of power “clipping” will often occur during peak production

14
The vast majority of thin-film capacity in the project population uses CdTe modules from First Solar. On its web
site (First Solar 2015), First Solar claims that its CdTe technology provides greater energy yield (per nameplate W)
than c-Si at module temperatures above 25° C (77° F) – i.e., conditions routinely encountered in the high-insolation
Desert Southwest region.
15
This ratio is referred to within the industry in a variety of ways, including: DC/AC ratio, array-to-inverter ratio,
oversizing ratio, overloading ratio, inverter loading ratio, and DC load ratio (Advanced Energy 2014; Fiorelli and
Zuercher - Martinson 2013). This report uses inverter loading ratio, or ILR.
16
This is analogous to the boost in capacity factor achieved by a wind turbine when the size of the rotor increases
relative to the turbine’s nameplate capacity rating. This decline in “specific power” (W/m2 of rotor swept area)
causes the generator to operate closer to (or at) its peak rating more often, thereby increasing capacity factor.

9
periods 17). Particularly under time-varying PPA prices that extend peak pricing into the morning
and/or evening hours, the resulting boost in generation (and revenue) during the shoulder periods
of each day outweighs the occasional loss of revenue from peak-period clipping (which may be
largely limited to just the high-insolation summer months).

Figure 5 shows the capacity-weighted average ILR among projects built in each year, both for
the total PV project population (solid black line) and broken out by fixed-tilt versus tracking
projects. Across all projects, the average ILR has increased significantly over time, from around
1.2 for projects built in 2010 to 1.31 in 2013. In 2014, the capacity-weighted average declined
slightly to 1.28, as a number of very large projects that had been under construction for several
years finally came online; some of these projects have lower ILRs than their more-recently
designed counterparts. But the 2014 median ILR (not shown) remained unchanged from 2013, at
1.29.

Figure 5. Trends in Inverter Loading Ratio by Mounting Type and Installation Year

With the exception of 2014 (again, influenced by these few large fixed-tilt projects with lower
ILRs), fixed-tilt projects generally feature higher ILRs than tracking projects. This finding is
consistent with the notion that fixed-tilt projects have more to gain from boosting the ILR in
order to achieve a less-peaky, “tracking-like” daily production profile.

17
Power clipping, also known as power limiting, is comparable to spilling excess water over a dam (rather than
running it through the turbines) or feathering a wind turbine blade. In the case of solar, however, clipping occurs
electronically rather than physically: as the DC input to the inverter approaches maximum capacity, the inverter
moves away from the maximum power point so that the array operates less efficiently (Advanced Energy 2014,
Fiorelli and Zuercher‐Martinson 2013). In this sense, clipping is a bit of a misnomer, in that the inverter never really
even “sees” the excess DC power – rather, it is simply not generated in the first place. Only potential generation is
lost.

10
All else equal, Figure 4 and Figure 5 suggest that project-level capacity factors should increase
among more recently built PV projects. This hypothesis is explored further (and confirmed) in
Chapter 5.

CSP (15 projects, 1,673 MWAC)

After the nearly 400 MW AC SEGS I-IX parabolic trough build-out in California in the 1980s and
early 1990s, no other utility-scale CSP project was built in the United States until the 68.5
MW AC Nevada Solar One trough project in 2007. This was followed by the 75 MW AC Martin
project in 2010 (also a trough project, feeding steam to a co-located combined cycle gas plant in
Florida), and the 250 MW AC Solana trough project in Arizona in 2013 (which also includes 6
hours of molten salt storage capacity).

In 2014, three additional CSP projects came online in California: two more trough projects
without storage (Genesis and Mojave, each 250 MW AC ) and the first large-scale “solar tower”
project in the United States (Ivanpah at 377 MW AC ). A second 110 MW AC solar tower project
with 10 hours of built-in thermal storage – Crescent Dunes in Nevada – has finished major
construction activities but, at the time of writing, was still in the commissioning phase and not
yet commercially online, and is thus excluded from this report. In the wake of this
unprecedented buildout – totaling 1,127 MW AC – of new CSP capacity in the past two years,
there are currently no other major CSP projects moving towards construction in the United
States.

11
3. Installed Prices
This chapter analyzes installed price data from a large sample of the overall utility-scale solar
project population described in the previous chapter. 18 Specifically, LBNL has gathered
installed price data for 176 utility-scale (i.e., ground-mounted and larger than 5 MW AC ) solar
projects totaling 7,145 MW AC and built between 2007 and 2014. The price sample is dominated
by 170 PV projects (including 2 CPV projects) that total 5,874 MW AC (i.e., PV accounts for 97%
of all projects and 82% of all capacity in the installed price sample). It also includes 6 CSP
projects totaling 1,270 MW AC , consisting of the more recently built projects described in the
previous chapter (rather than the older SEGS projects).

In general, only fully operational projects for which all individual phases were in operation at the
end of 2014 are included in the sample 19 – i.e., by definition, our sample is backward-looking
and therefore may not reflect installed price levels for projects that are completed or contracted
in 2015 and beyond. Moreover, reported installed prices within our backward-looking sample
may reflect transactions (e.g., entering into an Engineering, Procurement, and Construction or
“EPC” contract) that occurred several years prior to project completion. In some cases, those
transactions may have been negotiated on a forward-looking basis, reflecting anticipated future
costs at the time of project construction. In other cases, they may have been based on
contemporaneous costs (or a conservative projection of costs), in which case the reported
installed price data may not fully capture recent reductions in component costs or other changes
in market conditions. 20 For these reasons, the data presented in this chapter may not correspond
to recent price benchmarks for utility-scale PV (Feldman et al. 2015), and may differ from the
average installed prices reported elsewhere (Bloomberg New Energy Finance 2015; Fu et al.
2015; GTM Research and SEIA 2015). A text box later in this chapter (see Bottom-Up vs. Top-
Down) explores this issue in more detail.

This chapter analyzes installed price trends among the sample of utility-scale projects described
above. It begins with an overview of installed prices for PV (and CPV) projects over time, and
then breaks out those prices by module type (c-Si vs. thin-film vs. CPV), mounting type (fixed-
tilt vs. tracking), and system size. The chapter then provides an overview of installed prices for
the six CSP projects in the sample. Sources of installed price information include the Treasury
Department’s Section 1603 Grant database, data from applicable state rebate and incentive
programs, state regulatory filings, FERC Form 1 filings, corporate financial filings, interviews
with developers and project owners, trade press articles, and data previously gathered by the
National Renewable Energy Laboratory (NREL). All prices are reported in real 2014 dollars.

18
Installed “price” is reported (as opposed to installed “cost”) because in many cases, the value reported reflects
either the price at which a newly completed project was sold (e.g., through a financing transaction), or alternatively
the fair market value of a given project – i.e., the price at which it would be sold through an arm’s-length transaction
in a competitive market.
19
In contrast, later chapters of this report do present data for individual phases of projects that are online, or (in the
case of Chapter 6 on PPA prices) even for phases of projects or entire projects that are still in development and not
yet operating.
20
This reasoning may partially explain why the decline in installed prices presented in this chapter has seemingly
not kept pace with the decline in PPA prices reported later in Chapter 6.

12
 PV (170 projects, 5,874 MWAC, including 2 CPV projects totaling 35 MWAC)
LBNL’s sample of 170 PV (and CPV) projects totaling 5,874 MW AC for which installed price
estimates are available represents 87% of the total number of PV projects and 94% of the amount
of capacity in the overall PV project population described in Chapter 2. Focusing just on those
PV projects that achieved commercial operation in 2014, LBNL’s sample of 55 projects totaling
3,052 MW AC represents 87% and 95% of the total number of 2014 projects and capacity in the
population, respectively.

Figure 6 shows installed price trends for PV (and CPV) projects completed from 2007 through
2014 in both DC and AC terms. Because PV project capacity is commonly reported in DC terms
(particularly in the residential and commercial sectors), the installed cost or price of solar is often
reported in $/W DC terms as well (Barbose and Darghouth 2015; GTM Research and SEIA 2015).
As noted in the text box (AC vs. DC) at the beginning of Chapter 2, however, this report analyzes
utility-scale solar in AC terms. Figure 6 shows installed prices both ways (in both $/W DC and
$/W AC terms) in an attempt to provide some continuity between this report and others that
present prices in DC terms. The remainder of this chapter, however, as well as the rest of this
document, report data exclusively in AC terms, unless otherwise noted.
10 Capacity-Weighted Average (DC)
9 Median (DC)
Individual Projects (DC)
8
Installed Price (2014 $/W)

Capacity-Weighted Average (AC)

7 Median (AC)
Individual Projects (AC)
6

0
2007-2009 2010 2011 2012 2013 2014
n=5 (75 MW-AC) n=10 (175 MW-AC) n=29 (428 MW-AC) n=38 (875 MW-AC) n=33 (1,269 MW-AC) n=55 (3,052 MW-AC)
Installation Year
Figure 6. Installed Price of Utility-Scale PV and CPV Projects by Installation Year

As shown, the median utility-scale PV prices (solid lines) within our sample have declined fairly
steadily in each year, to $3.1/W AC (or $2.3/W DC ) in 2014. This represents a price decline of
more than 50% since the 2007-2009 period (and 37% since 2010). The lowest-priced projects
among our 2014 sample of 55 PV projects were ~$2/W AC , with the lowest 20th percentile of
projects having fallen considerably, from $3.2/W AC in 2013 to $2.3/W AC in 2014.

In contrast, capacity-weighted average prices (dashed lines) have declined more slowly through
2013, and even increased slightly in 2014 to $3.8/W AC (or $2.9/W DC ). The divergence between
median and capacity-weighted average prices in 2014 can be explained by a number of very
large PV projects that have been under construction for several years but that only achieved final

13
 commercial operation in 2014 (and so only entered our sample in 2014). These projects may
have signed EPC contracts several years ago, perhaps at significantly higher prices than some of
their smaller and more-nimble counterparts that started construction more recently. 21 Although
in general we prefer capacity-weighted averages over medians, 22 the next graph will focus on
medians rather than capacity-weighted averages in order to avoid the apparent distortion seen in
Figure 6 for 2014.

While median prices in the sample have generally declined over time, there remains a
considerable spread in individual project prices within each year. The overall variation in prices
may be partially attributable to differences in module and mounting type – i.e., whether PV
projects use c-Si or thin-film modules, and whether those modules are mounted at a fixed tilt or
on a tracking system.
9
Fixed-Tilt c-Si Tracking c-Si
8
Fixed-Tilt Thin-Film Tracking Thin-Film
Installed Price (2014 $/WAC)

7 CPV All PV

2
Markers represent medians, with 20th and 80th percentiles.
1

0
2007-2009 2010 2011 2012 2013 2014
n=5 (75 MW-AC) n=10 (175 MW-AC) n=29 (428 MW-AC) n=38 (875 MW-AC) n=33 (1,269 MW-AC) n=55 (3,052 MW-AC)
Installation Year
Figure 7. Installed Price of Utility-Scale PV and CPV Projects by Project Design and
Installation Year

Figure 7 breaks out installed prices over time among these four combinations (and also includes
the two CPV projects in the sample – but excludes several “hybrid” projects that feature a mix of

21
For example, within our PPA price sample (described later in Chapter 6), the longest span between PPA execution
date (as a proxy for EPC contract execution date) and commercial operation date for projects that came online in
2014 is 5 ¾ years, with the average lag for systems larger than 100 MWAC being 3 ¾ years, compared to 2¼ years
for systems smaller than 100 MWAC. Because of their size, very large projects dominate the capacity-weighted
average price in 2014 (eight projects larger than 100 MWAC represent 74% of the capacity additions, but only 12.5%
of new projects, in 2014).
22
Whereas medians (and simple means) tell us about the typical project, capacity-weighted averages tell us more
about the typical unit of capacity (e.g., the typical MW). Throughout most of this report, we are interested in
analyzing the U.S. solar market in its entirety – e.g., deriving a representative installed price per unit of capacity
(rather than per project), or a representative capacity factor or PPA price per MWh for the US fleet as a whole – and
therefore tend to favor capacity-weighted averages over medians (or simple means). Given the apparent distortion
noted above, however, as well as our increasing sample size over time (which lends itself more readily to medians),
the use of medians seems more appropriate for this chapter – and will also align this report more closely with
reported median prices for the residential and commercial PV systems in LBNL’s companion Tracking the Sun
series (e.g., see Barbose and Darghouth 2015).

14
module and/or mounting types, and so do not fit neatly into these four combinations). In 2014,
the median price was $2.8/W AC for fixed-tilt c-Si projects, $3.1/W AC for tracking c-Si projects,
$3.3/W AC for fixed-tilt thin-film projects, and $3.2/W AC for tracking thin-film projects.

Trends of particular note include:

• Although projects using c-Si modules were more expensive than projects using thin-film
modules (e.g., by ~$1.1/W AC on average in 2010 for fixed-tilt projects), the average
installed price of fixed-tilt c-Si and thin-film projects has converged over time, and even
reversed in 2014 when c-Si held a ~$0.6/W AC advantage over thin-film projects
completed in the same year (although some smaller fixed-tilt thin-film projects are
offered at prices similar to the cheaper c-Si projects). This convergence has been led by
the falling price of c-Si modules over time. As the price of c-Si projects has converged
with thin-film, the predominance of c-Si projects has grown in both the installed price
sample and the broader population (although this is not necessarily true for total
interconnected capacity, given several very large thin-film projects that came online in
2014).

• Tracking systems remain slightly more expensive than fixed-tilt systems within the
sample – a difference of about $0.3/W AC in 2014 among c-Si projects. As shown later in
Chapter 5, however, this higher up-front expenditure results in greater energy production.
In contrast, fixed-tilt thin-film projects do not appear to have a similar cost advantage
over tracking thin-film projects, though this may be attributable to the previously noted
price lags associated with several very large fixed-tilt thin-film projects (as well as
perhaps to the vertical integration of First Solar and RayTracker).

• The two high-concentration CPV projects built in 2011 and 2012 exhibit installed prices
that are comparable to the average PV pricing in the sample (yet, as shown later in
Chapter 5, these two CPV projects have not performed as well as the average PV
project). One or more low-concentration CPV projects (e.g., SunPower’s new C7
technology powering an Apple server farm in Nevada) will enter the sample in 2015,
providing additional data points.

Differences in project size may also explain some of the variation in installed prices, as PV
projects in the sample range from 5.1 MW AC to 585 MW AC . Figure 8 investigates price trends
by project size. To minimize the potentially confounding influence of price reductions over
time, Figure 8 focuses on just those PV projects in the sample that became fully operational in
2014.

15
 5
Markers represent capacity-weighted averages, with 20th and 80th percentiles.
Figure only includes 2014-vintage projects.
4
Installed Price (2014 $/WAC)

1
All PV Fixed-Tilt c-Si Tracking c-Si Fixed-Tilt Thin-Film Tracking Thin-Film

0
5-10 MW 10-20 MW 20-100 MW 100-1,000 MW
n=18 (141 MW-AC) n=23 (401 MW-AC) n=6 (251 MW-AC) n=7 (2,029 MW-AC)
Project Size (MWAC)
Figure 8. Installed Price of 2014 PV Projects by Size and Project Design

As shown, no consistent evidence of economies of scale can be found among the PV systems in
our pricing sample that achieved commercial operation in 2014. 23 For example, there are no
clear trends – either among the various mounting/module combinations (e.g., fixed-tilt c-Si) or
for all projects in aggregate – among the first three project size bins shown in Figure 8, which
range from 5 MW AC up to 100 MW AC . One possible explanation for this lack of trend is that
economies of scale may be limited primarily to projects smaller than 5 MW AC – which are
excluded from our sample – given that the standardized and modular “power blocks” of module
manufacturers like SunPower and First Solar are sized below this 5 MW AC threshold. Another
possibility is potential inconsistency in what costs or prices are captured among projects; e.g.,
some of the larger projects may include interconnection and transmission costs that are not
present (or at least not reported) for smaller projects.

More notable in Figure 8 are the price penalties for projects larger than 100 MW AC ; two factors
may contribute to these apparent diseconomies of scale for very large projects. As discussed
earlier, most of these very large projects have been under construction for several years and may
therefore reflect higher module and EPC costs from several years ago. Moreover, these mega-
scale projects – some of which involve more than 8 million modules and project sites of nearly
10 square miles – may face greater administrative, regulatory, and interconnection costs than do
smaller projects.

23
These empirical findings more or less align with recent modeling work from NREL (Fu et al. 2015), which also
finds only modest scale economies for a 100 MW project compared to a 10 MW project, and no additional scale
economies for projects larger than 100 MW.

16
 Bottom-Up versus Top-Down: Different Ways to Look at Installed Project Prices
The installed prices analyzed in this chapter generally represent empirical top-down price
estimates gathered from sources (e.g. corporate financial filings, FERC filings, the Treasury’s
Section 1603 grant database) that typically do not provide more granular insight into component
costs. In contrast, several recent publications (Fu et al. 2015; GTM Research and SEIA 2015;
Bloomberg New Energy Finance 2015) take a different approach of modeling total installed
prices via a bottom-up process that aggregates modeled cost estimates for various project
components to arrive at a total installed price. Each type of estimate has both strengths and
weaknesses – e.g., top-down estimates often lack component-level detail but benefit from an
empirical reality check, while bottom-up estimates provide more detail but rely on modeling.
This text box explores to what extent the two different types of price estimates are in alignment,
and where any differences lie. To aid in this comparison, LBNL obtained a detailed project cost
breakdown for one of the PV projects in its price sample: a 20 MW AC (25 MW DC ) single-axis
tracking c-Si project that came online in the Southwest in 2014. The reported total installed
price of this project – $2.37/W DC or $2.97/W AC – is comparable to other similar 2014 projects in
the LBNL sample, suggesting that this project’s detailed cost breakdown may be representative
of other similar projects.
$3.0 Other (Land, Enviro, Permit, Insurance, Legal, PPA)
0.20
$2.5 0.63 Construction Equipment and Labor
0.16
Civil Engineering/Grading
$2.0 0.50 0.33
2014 $/W

0.09 Design/Engineering
$1.5 0.26 0.26
0.08 Transmission
0.20 0.38
$1.0 0.30 0.22 Interconnection/Substation/IT
0.18
Tracker
$0.5 0.82
0.66 Inverter
$0.0
Modules
$/W-DC $/W-AC
Representative Bottom-up Price of 2014 20 MW AC Single-Axis Tracking System
The original cost breakdown for this project reported costs in 67 different categories that, for
ease of presentation, are grouped into 9 larger cost bins in the figure above. As shown, the three
major hardware components account for almost half of total costs, with 28% ($0.66/W DC /
$0.82/W AC ) coming from the modules, 13% ($0.30/W DC / $0.38/W AC ) from the tracking/racking
system, and 7.5% ($0.18/W DC / $0.22/W AC ) from the inverters. Construction equipment and
labor accounts for another 21% ($0.50/W DC / $0.63/W AC ), while 11% ($0.26/W DC / $0.33/W AC )
is attributable to civil engineering and grading.
The figure on the next page compares the cost breakdown for this seemingly representative
project with modeled bottom-up estimates from NREL (Fu et al. 2015), BNEF (Bloomberg New
Energy Finance 2015), and Greentech Media (GTM Research and SEIA 2015). Because each of
these publications reports costs slightly differently, we had to create fairly broad (and hence
rough) cost bins that reflect the “lowest common denominator” in order to compare them. In
contrast to the rest of this report, costs in the next graph are shown exclusively in $/W DC to align
with how they are reported in these other publications.

17
 $2.5 2.37 2.25
System Cost or Price (2014 $/W-DC)

0.14
1.99 1.99
$2.0 0.46 1.88
0.82 0.33 0.36
0.36
$1.5 0.52 0.34
0.59 0.26
0.58 0.43 0.48 0.37
$1.0
0.27
0.18 0.13 0.13 0.18
0.11
$0.5
0.66 0.71 0.71 0.66 0.72

$0.0
LBNL 2014 25 MW-DC Single- NREL 2014 20 MW-DC Single- NREL 2014 100 MW-DC Single- BNEF Q4 2014 Utility Project in GTM H2 2014 10 MW-DC
Axis Tracking Project Price in Axis Tracking Project Cost in Axis Tracking Project Cost (U.S. California (incl. Developer Single-Axis Tracking Project
Southwest Southwest with Union Labor National Average) Margins) Cost in California (excl.
Development)
Modules Inverter Tracker, Structural BOS, Interconnection, Transmission Design, EPC, Labor, PII Other

Comparison of Bottom-Up Utility-Scale PV Project Cost Estimates

As shown, the sample LBNL project has the highest installed price – despite reporting among the
lowest module costs. That said, the total installed price of $2.37/W DC is not too dissimilar from
NREL’s modeled bottom-up estimate of $2.25/W DC for a similar project (i.e., a 20 MW DC
tracking c-Si project located in the Southwest and built with union labor). The other three
estimates are all lower, with the NREL national and the BNEF model both arriving at about
$2/W DC . The GTM estimate is the lowest as it excludes development costs (captured by the
LBNL empirical breakdown); meanwhile, GTM’s relatively high inverter costs include the AC
subsystem, which other estimates include within interconnection costs. Finally, there are
probably other differences in costs captured by the various estimates (e.g., financing costs,
developer profit margins, transaction costs) that impede straightforward comparisons.

Among cost categories, the largest discrepancy between the sample LBNL project and the
modeled bottom-up prices comes from the category that includes project design, EPC, labor, and
permitting, interconnection and inspection (“PII”). One potential explanation for this
discrepancy is that the bottom-up models may be modeling current EPC (or other) costs for
projects that will be built in the future, whereas the sample LBNL project achieved commercial
operation in 2014 and may therefore reflect, for example, EPC costs from some time ago (e.g.,
from before the project entered the construction phase).

Although it’s difficult to pin down the exact reason for the discrepancy in installed prices shown
in the figure above, this analysis nevertheless highlights the potentially substantial variation
between empirical top-down and modeled bottom-up installed price estimates (and even among
the various modeled bottom-up price estimates themselves), as well as the importance of
understanding what each price estimate represents.

18
 CSP (6 projects, 1,270 MWAC)

The CSP installed price sample excludes the nine SEGS projects built several decades ago, but
includes all other concentrated solar thermal power (CSP) projects, totaling 1,270 MWAC, that
were commercially operational at the end of 2014 and larger than 5 MW AC . Five of these six
projects feature parabolic trough technology, while the sixth uses power tower technology
(consisting of a total of 3 solar towers). Another large solar tower project that had finished major
construction activities in early 2014 but that had not yet entered commercial operation by the end
of 2014 has been excluded from the sample.

Figure 9 breaks down these various CSP projects by size, technology and commercial operation
date (from 2007 through 2014), 24 and also compares their installed prices to the median installed
price of PV (from Figure 6) in each year from 2010 through 2014. The small sample size makes
it difficult to discern any trends. In 2014 alone, for example, two equal-sized trough systems
using similar technology (and both lacking storage) had significantly different installed prices
($5.10/W vs. $6.16/W). Meanwhile, the 2013 Solana trough system with six hours of storage
was (logically) priced above both 2014 trough projects (at $6.76/W), while the 2014 power tower
project was priced at the higher end of the range of the two trough projects. In general, CSP
prices do not seem to have declined over time to any notable extent, in stark contrast to the
median PV prices included in the figure.

8
6.76
250 MWAC with
6.16
7 75 MWAC 6 hours of storage
Installed Price (2014 $/WAC)

6.01
6
5.67
68 MWAC
250 MWAC each
5
5.10
4.48
4 377 MWAC

2 CSP Trough
CSP Tower
1
Median PV (for reference)
0
2006 2007 2008 2009 2010 2011 2012 2013 2014
Installation Year
Figure 9. Installed Price of Utility-Scale CSP Projects by Technology and Installation Year

24
The installed CSP prices shown in Figure 9 represent the entire project, including any equipment or related costs
to enable natural gas co-firing.

19
4. Operation and Maintenance Costs
In addition to up-front installed project costs or prices, utility-scale solar projects also incur
ongoing operation and maintenance (“O&M”) costs, which are defined here to include only those
direct costs incurred to operate and maintain the generating plant itself. In other words, O&M
costs – at least as reported here – exclude payments such as property taxes, insurance, land
royalties, performance bonds, various administrative and other fees, and overhead (all of which
contribute to total operating expenses). This section reviews and analyzes the limited data on
O&M costs that are in the public domain.

Empirical data on the O&M costs of utility-scale solar projects are hard to come by. Very few of
the utility-scale solar projects that have been operating for more than a year are owned by
investor-owned utilities, which FERC requires to report on Form 1 the O&M costs of the power
plants that they own. 25 Even fewer of those investor-owned utilities that do own utility-scale
solar projects actually report operating cost data in FERC Form 1 in a manner that is useful (if at
all). It also appears that most investor-owned utilities (with the exception of Florida Power &
Light) do not report empirical O&M costs for individual solar projects, but instead report
average O&M costs across their entire fleet of PV projects, pro-rated to individual projects on a
capacity basis. This lack of project-level granularity requires us to analyze solar O&M costs on
an aggregate utility level rather than an individual project level. Table 1 describes our O&M cost
sample and highlights the growing cumulative project fleet of each utility.

PG&E 26 PNM APS 27 FP&L

Year
MW AC # projects MW AC # projects MW AC # projects MW AC # projects
2011 N/A N/A N/A N/A 51 3 110 3
2012 50 3 20 4 96 4 110 3
2013 100 6 42 4 136 6 110 3
2014 N/A N/A 65 6 168 7 110 3
predominant
fixed-tilt c-Si fixed-tilt thin-film primarily tracking c-Si mix of c-Si and CSP
technology
Table 1. Operation and Maintenance Cost Sample

Despite these limitations, Figure 10 shows average utility fleet-wide annual O&M costs for this
small sample of projects in $/kW AC -year (blue solid line) and $/MWh (red dashed line) 28. The

25
FERC Form 1 uses the “Uniform System of Accounts” to define what should be reported under “operating
expenses” – namely, those operational costs of supervision and engineering, maintenance, rents, and training (and
therefore excluding payments for property taxes, insurance, land royalties, performance bonds, various
administrative and other fees, and overhead).
26
As PG&E does not report operating costs for its solar projects on FERC Form 1, we turned to O&M costs
reported in a CPUC compliance report (Middlekauff and Mathai-Jackson 2015) that unfortunately did not include
usable cost data for 2014.
27
APS reports O&M costs in FERC Form 1 only in an aggregated manner across customer classes (residential,
commercial, and utility-scale). For lack of better data, we use their 168 MWAC of total PV capacity (including
residential and commercial) as a proxy for the 7 utility-scale solar plants with a combined capacity of 158 MWAC.

20
whiskers represent both the lowest and the highest utility fleet-wide cost in each year. The dotted
line refers to FP&L’s project-specific annual O&M costs of its 75 MW CSP plant.

Average O&M costs for the PV plants within this sample have steadily declined from about
$30/kW AC -year (or $19/MWh) in 2011 to about $17/kW AC -year ($8/MWh) in 2014. This
decline could potentially indicate that utilities are capturing economies of scale as their PV
project fleets grow over time, although the most recent drop from 2013 to 2014 may simply be a
result of missing PG&E’s costs for 2014 (PG&E’s reported costs for 2012 and 2013 were above
average). In 2014, all but one PV project had O&M costs of less than $20/kW AC -year (or
$11/MWh), which is lower than recent medium-term projections by bond rating agencies (see the
O&M cost section of Bolinger and Weaver (2014)).

The only CSP plant in our sample reports higher O&M costs, in the $40-$50/kW AC -year range
for 2013 and 2014.

Figure 10. Empirical O&M Costs Over Time

As utility ownership of operating solar projects increases in the years ahead (and as those utilities
that already own substantial solar assets but do not currently report operating cost data hopefully
begin to do so, as required in FERC Form 1), the sample of projects reporting O&M costs should
grow, potentially allowing for more interesting analyses in future editions of this report.

28
O&M costs for the single CSP project (a 75 MW parabolic trough project) are only shown in $/kW-year terms
because this project provides steam to a co-located combined cycle gas plant.

21
5. Capacity Factors
At the close of 2014, more than 140 utility-scale solar projects (again, ground-mounted projects
larger than 5 MW AC ) had been operating for at least one full year (and in some cases for many
years), thereby enabling the calculation of capacity factors. 29 Sourcing net generation data from
FERC Electronic Quarterly Reports, FERC Form 1, EIA Form 923, and state regulatory filings,
this chapter presents net capacity factor data for 128 PV projects totaling 3,201 MW AC , two CPV
projects totaling 35 MW AC , and thirteen CSP projects (a mix of parabolic trough and power
tower projects, with and without thermal storage) totaling 1,390 MW AC (and for which only the
solar generation is reported here – no gas or oil augmentation is included). The PV sample size
of 128 projects totaling 3.2 GW is double the amount analyzed in last year’s edition of this
report, and should once again increase significantly in next year’s edition (along with more CSP
as well), as the record amount of new utility-scale solar capacity that came online in 2014 will
have its first full operating year in 2015.

PV (128 projects, 3,201 MWAC)

Project-level capacity factors for utility-scale PV projects can vary considerably, based on a
number of factors, including (in approximate decreasing order of importance): the strength of
the solar resource at the project site (measured in GHI with units kWh/m2/day); whether the
array is mounted at a fixed tilt or on a tracking mechanism; the DC capacity of the array relative
to the AC inverter rating (i.e., the inverter loading ratio, or ILR); and the type of modules used
(e.g., c-Si versus thin-film). Other factors such as tilt and azimuth will also play an obvious role,
though since we focus only on ground-mounted utility-scale projects, our operating assumption
is that these fundamental parameters will be equally optimized to maximize energy production
across all projects.

One might also expect project vintage to play a role – i.e., that newer projects will have higher
capacity factors because the efficiency of PV modules (both c-Si and thin-film) has increased
over time. As module efficiency increases, however, developers simply either use fewer
modules to reach a fixed amount of capacity (thereby saving on balance-of-system and land costs
as well) or, alternatively, use the same number of modules to boost the amount of capacity
installed on a fixed amount of land (directly reducing at least $/W DC costs, if not also $/W AC
costs). In other words, for PV more than for other technologies like wind power, efficiency
improvements over time show up primarily as cost savings rather than as higher capacity factors.
Any increase in capacity factor by project vintage is therefore most likely attributable to a time
trend in one of the other variables noted above – e.g., towards higher inverter loading ratios or
greater use of tracking.

29
Because solar generation is seasonal (generating more in the summer and less in the winter), capacity factor
calculations should only be performed in full-year increments.

22
 Figure 11 illustrates and supports this hypothesis, by breaking out the average net capacity factor
(“NCF”) by project vintage across the sample of projects built from 2010 through 2013 (and by
noting the relevant average project parameters within each vintage). The capacity factors
presented in Figure 11 represent cumulative capacity factors – i.e., calculated over as many years
of data as are available for each individual project (a maximum of four years, from 2011 to 2014,
in this case), rather than for just a single year (though for projects completed in 2013, only a
single year of data exists at present) – and are expressed in net, rather than gross, terms (i.e., they
represent the output of the project net of its own use). Notably, they are also calculated in AC
terms (i.e., using the MW AC rather than MW DC nameplate rating), 30 yielding higher capacity
factors than if reported in DC terms, 31 but allowing for direct comparison with the capacity
factors of other generation sources (e.g., wind energy or conventional energy), which are also
calculated in AC terms.

As shown, the average capacity factor increases only slightly from 2010- to 2011-vintage
projects, due primarily to a higher proportion (in capacity terms) of projects using tracking
among 2011-vintage projects, given virtually no change in the average ILR or GHI across these
two vintages. Projects built in 2012 and especially 2013, however, have progressively higher
capacity factors on average, driven by an increase in both average ILR and GHI in each year.
Cap-Wgtd Avg Cumulative Net Capacity Factor

30%

25%

20% Cap-Wgtd Avg: Cap-Wgtd Avg: Cap-Wgtd Avg: Cap-Wgtd Avg:

NCF = 23.8% NCF = 24.5% NCF = 26.3% NCF = 29.4%
15% ILR = 1.19 ILR = 1.19 ILR = 1.27 ILR = 1.30
GHI = 5.22 GHI = 5.16 GHI = 5.36 GHI = 5.57
Tracking = 13% (MW) Tracking = 51% (MW) Tracking = 38% (MW) Tracking = 48% (MW)
10%

5%

0%
2010 Vintage 2011 Vintage 2012 Vintage 2013 Vintage
7 Projects 32 Projects 36 Projects 47 Projects
144 MW 464 MW 834 MW 1,676 MW

Figure 11. Cumulative PV Capacity Factor by Project Vintage: 2010-2013 Projects Only

Because Figure 11 analyzes cumulative capacity factors, one other possible explanation for the
upward trend by vintage could be if the solar resource across the United States were significantly
stronger in 2014 than in 2011-2013. If this were the case – which seems unlikely based on ex-
post annual solar resource data (3Tier 2013; Vaisala 2014; Vaisala 2015) – then 2013-vintage
projects might be expected to exhibit higher cumulative capacity factors than 2010-2012

30
The formula is: Net Generation (MWhAC) over Single- or Multi-Year Period / [Project Capacity (MWAC) *
Number of Hours in that Same Single- or Multi-Year Period].
31
For example, a project with a 30% capacity factor in AC terms would have a 25% capacity factor in DC terms at
an inverter loading ratio of 1.20, and a 20% capacity factor in DC terms at an inverter loading ratio of 1.50.

23
 projects, given that 2014 is the only applicable performance year for a 2013-vintage project. To
check against this possibility, Figure 12 replicates Figure 11, but based on single-year 2014
capacity factors rather than cumulative capacity factors. In other words, each vintage is
measured based on its performance during the same single year – 2014 – rather than over a one-
to four-year period, depending on vintage. As shown, the upward trend still holds, suggesting
that ILR, GHI, and tracking are the true drivers. 32

30%
Cap-Wgtd Avg 2014 Net Capacity Factor

25%

20% Cap-Wgtd Avg: Cap-Wgtd Avg: Cap-Wgtd Avg: Cap-Wgtd Avg:

NCF = 22.9% NCF = 24.5% NCF = 26.3% NCF = 29.4%
15% ILR = 1.19 ILR = 1.19 ILR = 1.27 ILR = 1.30
GHI = 5.22 GHI = 5.18 GHI = 5.36 GHI = 5.57
Tracking = 13% (MW) Tracking = 52% (MW) Tracking = 38% (MW) Tracking = 48% (MW)
10%

5%

0%
2010 Vintage 2011 Vintage 2012 Vintage 2013 Vintage
7 Projects 31 Projects 36 Projects 47 Projects
144 MW 454 MW 834 MW 1,676 MW

Figure 12. 2014 PV Capacity Factor by Project Vintage: 2010-2013 Projects Only

To the extent that this observable time trend in net capacity factor by project vintage is, in fact,
attributable to a time trend in one or more of the other variables noted, it is perhaps best to
measure the effect of those other variables directly. Figure 13 does just that, by categorizing the
entire data sample in four different ways: by solar resource strength (in GHI terms), by fixed-tilt
versus tracking systems, by the inverter loading ratio, and by module type (c-Si versus thin-film).
The capacity-weighted average net capacity factor across the entire sample is 27.5%, the median
is 26.5%, and the simple average is 25.6%, but there is a wide range of individual project-level
capacity factors (from 14.8% to 34.9%) around these central numbers.

32
There is one less project in the sample for Figure 12 than for Figure 11, due to 2014 net generation data not yet
being available for one project in New Jersey.

24
 35% Capacity-Weighted Average (all projects)
Individual Project (c-Si)
Cumulative Net Capacity Factor

30%
Individual Project (thin-film)
25%

20%

15%
Sample includes 128 projects totaling 3,201 MWAC that came online from 2007-2013
10%

14 projects
10 projects
11 projects

13 projects
15 projects
2 projects
3 projects

4 projects

5 projects

9 projects
4 projects

6 projects
6 projects

8 projects

9 projects

8 projects

ILR≥1.275 377 MW
1 project
ILR<1.2 156 MW

1.2-1.275 100 MW

1.2-1.275 492 MW
ILR≥1.275 150 MW

1.2-1.275 465 MW
ILR≥1.275 125 MW

ILR<1.2 122 MW

ILR≥1.275 648 MW

ILR<1.2 182 MW
ILR<1.2 19 MW
5%
1.2-1.275 89 MW

1.2-1.275 36 MW

ILR<1.2 59 MW
ILR<1.2 67 MW
ILR≥1.275 89 MW

1.2-1.275 26 MW
0% ILR≥1.275

Fixed-Tilt Tracking Fixed-Tilt Tracking Fixed-Tilt Tracking

GHI of <4.75 kWh/m2/day GHI of 4.75-5.5 kWh/m2/day GHI of ≥5.5 kWh/m2/day

Figure 13. Cumulative PV Capacity Factor by Resource Strength, Fixed-Tilt vs. Tracking,
Inverter Loading Ratio, and Module Type

Each of the four variables explored in Figure 13 is discussed in turn below.

• Solar Resource: Each project in the sample is associated with a global horizontal
irradiance (GHI) value derived from the map shown earlier in Figure 3. Solar resource
bin thresholds (<4.75, 4.75-5.5, and ≥5.5 kWh/m2/day GHI) were chosen to ensure that a
sufficient number of projects fall within each bin. 33 Not surprisingly, projects sited in
stronger solar resource areas have higher capacity factors, all else equal. The difference
can be substantial: the capacity-weighted average net capacity factors in the highest
resource bin, for example, average 8% higher (in absolute terms) than their counterparts
in the lowest resource bin (with the range extending 5-9% depending on fixed-tilt versus
tracking and the inverter loading ratio).
• Fixed-Tilt vs. Tracking: Tracking (all single-axis in this sample) boosts average
capacity factor by 3-4% on average (in absolute terms), depending on the resource bin
(4% on average across all three resource bins).
• Inverter Loading Ratio (ILR): Figure 13 breaks the sample down further into three
different inverter loading ratio bins: <1.2, 1.2-1.275, and ≥1.275. 34 The effect on
average capacity factor is noticeable: across all resource bins and fixed/tracking bins, the
absolute difference in capacity factor between the highest and lowest inverter loading
ratio bin ranges from 1% to 6% (for an average of 4%).

33
Thirty-four projects totaling 436 MW fall into the lowest resource category of less than 4.75 kWh/m2/day, 33
projects totaling 582 MW fall into the middle resource category of between 4.75 and 5.5 kWh/m2/day, and 61
projects totaling 2,183 MW fall into the highest resource category of at least 5.5 kWh/m2/day.
34
These ILR bins were chosen to ensure a roughly equal number of projects in each bin. The lowest ILR bins
include 45 projects totaling 605 MWAC, the middle bins include 44 projects totaling 1,208 MWAC, and the highest
bins include 42 projects totaling 1,388 MWAC.

25
 • Module Type: Figure 13 differentiates between projects using c-Si and thin-film
modules by the shape and color of the markers denoting individual projects (the capacity-
weighted averages include both c-Si and thin-film projects). Though somewhat difficult
to tease out, the differences in project-level capacity factors by module type are generally
small (smaller than for the other variables discussed above), and do not appear to exhibit
any sort of pattern. That said, the prevalence of fixed-tilt thin-film projects within the
highest resource bin is noticeable. As mentioned in an earlier section of this report,
however, many of the new thin-film projects completed in 2014 (which will enter our
capacity factor sample in next year’s report) have deployed single-axis trackers.

CPV (2 projects, 35 MWAC)

The two CPV-only projects in the sample (the 5 MW AC Hatch and the 30 MW AC Cogentrix
Alamosa projects) use virtually the same high-concentration technology (from Amonix), and
both appear to be underperforming – both relative to publicly stated expectations and to how a
single-axis tracking PV project would probably have performed in similar conditions. In
November 2011, a few months after Hatch came online and a few months before Alamosa went
online, a conference presentation from Amonix suggested a 31.5% capacity factor for Hatch and
a 32.5% capacity factor for Alamosa (Pihowich 2011). 35 This 31.5-32.5% range is consistent
with the empirical PV capacity factors seen in the second column from the right in Figure 13,
which – except for the fact that they feature dual-axis, rather than single-axis, tracking – is where
these two CPV projects would otherwise fall based on resource strength and inverter loading
ratio. Actual experience to date, however, has been below this range: Hatch’s 20.9% capacity
factor in 2012 dropped to 18.5% in 2013 and 18.1% in 2014, while Cogentrix Alamosa posted a
24.9% capacity factor in 2013, followed by 24.2% in 2014. 36 The 2013 and 2014 capacity
factors may have been reduced somewhat by the reportedly below-average insolation levels in
the southwestern United States during the summers of 2013 and 2014 (3Tier 2013; Vaisala 2014;
Vaisala 2015). 37

35
The Amonix slide deck (Pihowich 2011) contains conflicting information: it lists expected generation numbers
that equate to a 31.5% capacity factor for Hatch, yet also states a slightly lower capacity factor estimate of 29.4% –
either of which is higher than actual experience. Meanwhile, documents from El Paso Electric (the offtaker) list
9,189 MWh, or a 20.8% capacity factor, as the expected output of Hatch (the project met this expectation in 2012,
but fell short in 2013 and 2014). For Cogentrix Alamosa, an April 2011 environmental assessment prepared for the
DOE’s Loan Program Office assumed a 29% capacity factor, although the current Loan Program Office project
description notes annual generation of 58,000 MWh, equivalent to just 21.9% (well below the >24% achieved to
date). The reason for these disparate expectations (for both projects) is not clear, though one potential explanation
might have to do with timing – i.e., the current El Paso Electric and Loan Program Office numbers might be more
recent, therefore potentially reflecting some degree of actual experience.
36
A third project that includes a mix of PV and CPV technologies – the 6.9 MWAC SunE Alamosa project – has
performed better than the two CPV-only projects, having logged a 28.7% cumulative capacity factor over six full
years of operation (from 2008-2013). The CPV portion of the project, however, only accounts for about 12% of the
project’s total capacity (the rest being PV with diurnal (~80%) or seasonal (~7%) tracking), and unfortunately, the
project-level net generation data are not granular enough to enable a determination of how the CPV and PV portions
of this project have performed independently.
37
The entire year 2014 was an average to slightly above average solar year in the West, and an average to slightly
below average solar year elsewhere in the continental United States. These annual averages mask important
seasonal divergences, however – e.g., the above-average insolation tended to be concentrated in the less-important

26
 CSP (13 projects, 1,390 MWAC)
Three new CSP projects totaling 892 MW AC achieved commercial operation in late 2013,
providing a significant boost to this year’s CSP capacity factor sample. Solana is a 250 MW AC
(net) parabolic trough project with six hours of molten salt storage located in Arizona; Genesis is
a 250 MW AC (net) parabolic trough project without storage located in California; and Ivanpah is
a 377 MW AC (net) power tower project without storage located in California.

30%
Genesis
Net Capacity Factor (solar portion only)

Solana
25% SEGS III-IX and Nevada Solar One (dashed)

20%

15%
Ivanpah
10%
SEGS I & II
5%

0%
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Figure 14. Capacity Factor of CSP Projects (Solar Portion Only) Over Time

Figure 14 shows the net capacity factors by calendar year from just the solar portion (i.e. no
augmentation with natural gas or fuel oil is included in Figure 14 38) of our CSP project sample.
The two new trough projects performed at roughly 28-29% capacity factors in 2014, while the
Ivanpah power tower project performed at ~12% capacity factor. For at least Solana (with 6
hours of storage) and Ivanpah, these first-year numbers are below long-term expectations of 41%
and 27%, respectively, and are projected to improve in future years as these projects overcome
typical start-up challenges and are fine-tuned for optimal performance (Danko 2015; Stern
2015). 39 Indeed, the performance of these two projects has already improved somewhat in the

non-summer months, while the critical months of May through September were generally below average across
much of the United States, including the Southwest (Vaisala 2014; Vaisala 2015).
38
Many of these projects also use gas-fired turbines to supplement their output (e.g., during shoulder months, into
the evening, or during cloudy weather). In the case of Nevada Solar One, for example, gas-fired generation has
boosted historical capacity factors by twenty to forty basis points depending on the year (e.g., from 19.4% solar-only
to 19.8% gas-included in 2014), with gas usage most often peaking in the spring and fall (shoulder months). The
SEGS projects use relatively more gas-fired generation, which boosted their aggregate capacity factors by 60-200
basis points in 2014, depending on the project. The Ivanpah power tower project also burns gas – and reportedly
more than originally anticipated (Danko 2015) – though data on its gas-fired generation in 2014 were not available
at the time of writing.
39
Ivanpah documentation suggests that this initial ramp-up could last as long as four years (Danko 2015).

27
first half of 2015. 40 Even despite these teething issues, however, the two new trough projects
performed significantly better in 2014 than the existing fleet of ten older trough projects in the
sample, including the nine SEGS plants (totaling 392 MW AC ) that have been operating in
California for more than twenty years, and the 68.5 MW AC Nevada Solar One trough project that
has been operating in Nevada since mid-2007. 41

These ten older trough projects tend to fall into two groupings, with SEGS I and II set apart from
the rest by significantly lower capacity factors, perhaps attributable to some combination of
separate ownership from SEGS III-IX as well as different plant characteristics (such as the size
of the collector field relative to the capacity and efficiency of the steam turbine). Nearly all of
these projects experienced lower solar-only capacity factors in 2013 and 2014 than in other
recent years. This decline is potentially attributable in part to inter-year variations in the solar
resource, which was below average in the southwestern United States (where these projects are
located) during the summers of 2013 and 2014 (3Tier 2013; Vaisala 2014; Vaisala 2015), with
summer being particularly important for CSP projects.

Looking ahead, another 250 MW AC (net) parabolic trough project in California without storage
(Mojave) achieved commercial operation in late 2014, and so will enter our capacity factor
sample in 2015. A second power tower project – the 110 MW Crescent Dunes project in
Nevada, with 10 hours of storage – is expected to be placed in service later in 2015 after a
prolonged commissioning process. Along with the three new projects added to the sample this
year (which should continue to mature over the next few years), these two new additions will
expand the CSP performance data set in future years.

40
For example, Ivanpah generated 309,913 MWh in the first six months of 2015 (for an annualized capacity factor
of 18.2%), compared to 173,138 MWh in the first six months of 2014 (an annualized capacity factor of 10.2%). For
Solana, the corresponding numbers are 352,569 MWh (32.5%) vs. 314,906 MWh (29.0%).
41
One additional parabolic trough project – the 75 MWAC Martin project in Florida – is excluded from the analysis
due to data complications. Specifically, since 2011, the Martin project has been feeding steam to a co-located
combined cycle gas plant, and a breakdown of the amount of generation attributable to solar versus gas is not readily
available.

28
6. Power Purchase Agreement (“PPA”) Prices

The cost of installing, operating, and maintaining a utility-scale solar project, along with its
capacity factor – i.e., all of the factors that have been explored so far in this report – are key
determinants of the price at which solar power can be profitably sold through a long-term power
purchase agreement (“PPA”). Relying on data compiled from FERC Electronic Quarterly
Reports, FERC Form 1, EIA Form 923, and a variety of regulatory filings, this section presents
trends in PPA prices among a large sample of utility-scale solar projects in the U.S. The sample
includes a total of 109 contracts totaling 8,578 MW AC and broken out as follows: 100 PV PPAs
totaling 7,234 MW AC , two CPV PPAs totaling 35 MW AC , one 7 MW AC PPA that is a mix of PV
and CPV, and 6 CSP PPAs (four parabolic trough, two power tower) totaling 1,301 MW AC .

The population from which this sample is drawn includes only those utility-scale projects that
sell electricity (as well as the associated capacity and renewable energy credits or “RECs”) in the
wholesale power market through a long-term, bundled PPA. Utility-owned projects, as well as
projects that benefit from net metering or customer bill savings, are therefore not included in the
sample. We also exclude those projects that unbundle and sell RECs separately from the
underlying electricity, because in those instances the PPA price alone does not reflect the
project’s total revenue requirements (at least on a post-incentive basis). PPAs resulting from
Feed-in Tariff (“FiT”) programs are excluded for similar reasons – i.e., the information content
of the pre-established FiT price is low (most of these projects do not exceed the 5 MW AC utility-
scale threshold anyway). In short, the goal of this chapter is to learn how much post-incentive
revenue a utility-scale solar project requires to be viable. 42 As such, the PPA sample comes
entirely from utility-scale projects that sell bundled energy, capacity, and RECs to utilities (both
investor-owned and publicly-owned utilities) or other offtakers through long-term PPAs resulting
from competitive solicitations or bilateral negotiations. 43 As a practical matter, this means that
we exclude “avoided cost” contracts – discussed in the text box on the next page – from our PPA
price sample as well.

42
Using PPA prices for this purpose reflects an implicit assumption that PPA prices will always be sufficient to
cover all costs and provide a normal rate of return. This may not always be the case, however, if projects
underperform relative to expectations or have higher-than-anticipated operating costs. In general, the project
sponsor and investors bear these risks (to varying degrees, depending on the specifics of their contractual
arrangements).
43
Because all of the PPAs in the sample include RECs (i.e., transfer them to the power purchaser), we need not
worry too much about REC price trends in the unbundled REC market. It is, however, worth noting that some states
(e.g., Colorado) have implemented REC “multipliers” for solar projects (whereby each solar REC is counted as
more than one REC for RPS compliance purposes), while others have implemented solar “set-asides” or “carve-
outs” (requiring a specific portion of the RPS to be met by solar) as a way to encourage specifically solar power
development. In these instances, it is possible that utilities might be willing to pay a bit more for solar through a
bundled PPA than they otherwise would be, either because they need to in order to comply with a solar set-aside, or
because they know that each bundled solar REC has added value (in the case of a multiplier). So even though REC
prices do not directly affect the analysis in this report, policy mechanisms tied to RECs might still influence bundled
PPA prices in some cases – presumably to the upside.

29
 Trend to Watch: The Rise (and Fall?) of “Avoided Cost” Markets

As discussed in the text, virtually all of the PPAs analyzed in this chapter result from competitive solicitations or some other form of
bilateral negotiation. Yet as the cost of solar has fallen to more-competitive levels, a “new” market for utility-scale solar (which is
actually one of the oldest markets for renewables in the United States, in existence ever since the Public Utility Regulatory Policies
Act, or PURPA, was signed into law in 1978) has emerged over the past year or so. Specifically, PURPA requires utilities to purchase
electricity from “qualifying facilities” (including solar and wind projects) at prices that represent their “avoided cost” – i.e., what they
would pay for the same amount of electricity generated by a non-qualifying facility. As a matter of policy, we exclude these “avoided
cost” contracts from our PPA price sample, because they are FiT-like and, in some states, also involve unbundling RECs; yet, as
discussed below, this growing market is not to be ignored.

Solar developers have been capitalizing on these avoided cost contracts for several years now in North Carolina, though the 5 MW
capacity limit in that state means that most of the more than 150 projects that are operational in North Carolina fall below our
threshold of what is considered to be “utility-scale.” But in the past year, numerous avoided cost contracts for larger projects have
been announced in other states that had not previously seen any solar development to speak of. For example:

• In Utah, at least two new 80 MWAC PV projects (in addition to a number of smaller projects) should begin commercial operation
by the end of 2015, selling electricity to PacifiCorp through 20-year avoided cost contracts. More than 700 MW of additional
solar capacity is under development in Utah and could come online in 2016 under these same avoided cost contracts.
Although the prices for these larger contracts are subject to negotiation, they are based loosely on PacifiCorp’s published
avoided cost rates, which are in the neighborhood of $50-$60/MWh when averaged over the 20-year contract term.

• Just to the north in Idaho, Idaho Power announced in late 2014 that it had recently entered into avoided cost contracts for 461
MW of utility-scale PV, and that another 885 MW was actively seeking such contracts. Idaho Power’s avoided cost contracts
feature pricing that varies by time of day and season, but averages out to about $60-70/MWh over the 20-year term.

This recent onslaught of applications for avoided cost contracts has prompted the utilities involved and their state utility regulators to
re-evaluate these contracts and the utilities’ PURPA requirements. In early 2015, for example, Idaho Power requested that its
standard avoided cost contract term be reduced from 20 to just 2 years; regulators subsequently reduced the term to 5 years while
they examined the issue, and in August 2015 agreed to impose a 2-year term. Around the same time, regulators in North Carolina
rejected a similar utility request to lower the capacity threshold, shorten the contract term, and reduce contract pricing for solar
projects. Meanwhile in Utah, PacifiCorp is also pushing back (as developers with Idaho projects that were stranded by the reduction
in contract term are now looking south to PacifiCorp as a more-viable market), and would like to reduce solar compensation via a
reduction in the capacity credit assigned to solar within the avoided cost calculation. How these various proceedings play out could
significantly affect the future of utility-scale solar within these states. In the near term, however, grandfathered contracts will fuel a
frenzy of new PV project construction in these emerging states through 2016.

For each of the contracts in the sample, 44 we have collected the contractually locked-in PPA
price data over the full term of the PPA, 45 and have accounted for any escalation rates and/or
time-of-delivery (“TOD”) pricing factors employed. 46 The PPA prices presented in this section,
therefore, reflect the full revenue available to (and presumably in many cases, the minimum

44
In general, each PPA corresponds to a different project, though in some cases a single project sells power to more
than one utility under separate PPAs, in which case two or more PPAs may be tied to a single project.
45
The minimum PPA term in the sample is 10 years (though the two 10-year contracts in the sample are effectively
4-year “bridge” PPAs with a California municipality, whereby the buyer takes 100% of the output for the first four
years and then, once a long-term contract with an investor-owned utility begins in 2019, just 1% of the output in the
last six years). The maximum is 34 years, the mean is 22.8 years, the median is 25 years, and the capacity-weighted
average is 23.4 years.
46
In cases where PPA price escalation rates are tied to inflation, the EIA’s projection of the U.S. GDP deflator from
Annual Energy Outlook 2015 is used to determine expected escalation rates. For contracts that use time-of-delivery
pricing and have at least one year of operating history, each project’s average historical generation profile is
assumed to be replicated into the future. For those projects with less than a full year of operating history, the
generation profiles of similar (and ideally nearby) projects are used as a proxy until sufficient operating experience
is available.

30
 amount of revenue required by47) these projects over the life of the contract – at least on a post-
incentive basis. In other words, these PPA prices do reflect the receipt of federal tax incentives
(e.g., the 30% investment tax credit or cash grant, accelerated tax depreciation) 48 and state
incentives (e.g., grants, production incentives, various tax credits), and would be higher if not for
these incentives. 49,50 As such, the levelized PPA prices presented in this section should not be
equated with a project’s unsubsidized levelized cost of energy (“LCOE”).

$300
Levelized PPA Price (Real 2014 $/MWh)

32 MW (New York)
$250
550 MW
$200
250
MW
$150

$100 PV (7,234 MW, 100 contracts)

CPV (35 MW, 2 contracts)
$50 Mix of PV/CPV (7 MW, 1 contract)
CSP (1,301 MW, 6 contracts)
$0
Jan-02

Jan-03

Jan-04

Jan-05

Jan-06

Jan-07

Jan-08

Jan-09

Jan-10

Jan-11

Jan-12

Jan-13

Jan-14

Jan-15

Jan-16
PPA Execution Date

Figure 15. Levelized PPA Prices by Technology, Contract Size, and PPA Execution Date

47
In a competitive “cost-plus” pricing environment – where the PPA price is just sufficient to recoup initial capital
costs, cover ongoing operating costs, and provide a normal rate of return – PPA prices will represent the minimum
amount of revenue required by a project. In contrast, “value-based” pricing occurs when the project developer or
owner is able to negotiate a higher-than-necessary PPA price that nevertheless still provides value to the buyer.
48
In addition to the other federal incentives listed, eleven projects within the sample also received DOE loan
guarantees through the Section 1705 program. In all eleven cases, however, the projects had already executed PPAs
by the date on which the loan guarantee was awarded, suggesting that the guarantee didn’t affect the PPA price.
49
For example, taking a simplistic view (i.e., not considering financing effects), the average PPA price could be as
much as 50% higher (i.e., 30%/(1 minus the federal tax rate)) if there were no federal investment tax credit (“ITC”).
Without the ITC, however, the resulting increase in PPA prices would be limited by the fact that sponsors with tax
appetite could then leverage up their projects more heavily with cheap debt, while sponsors without tax appetite
would be able to forego expensive third-party tax equity in favor of cheaper forms of capital, like debt. Because of
these financing shifts, the PPA price would not increase by 50%, but rather more like 35-40% in the case of a
sponsor with tax appetite, and by roughly 20% in the case of a sponsor without tax appetite that currently relies on
third-party tax equity to monetize the ITC (Bolinger 2014).
50
Though there is too much variety in state-level incentives to systematically quantify their effect on PPA prices
here, one example is New Mexico’s refundable Production Tax Credit, which provides a credit of varying amounts
per MWh (averaging $27/MWh) of solar electricity produced over a project’s first ten years. One PPA for a utility-
scale PV project in New Mexico allows for two different PPA prices – one that is $43.50/MWh higher than the
other, and that goes into effect only if the project does not qualify for the New Mexico PTC. Based on New
Mexico’s top corporate tax rate of 7.6%, a $43.50/MWh price increase due to loss of New Mexico’s PTC seems
excessive (a more appropriate 20-year adjustment would seemingly have been roughly half that amount), but
nevertheless, this is one tangible example of how state incentives can reduce PPA prices.

31
Figure 15 shows trends in the levelized (using a 7% real discount rate) PPA prices from the
entire sample over time. Each bubble in Figure 15 represents a single PPA, with the area of the
bubble corresponding to the size of the contract in MW and the placement of the bubble
reflecting both the levelized PPA price (along the vertical y-axis) and the date on the which the
PPA was executed (along the horizontal x-axis). 51 Different solar technologies (e.g., PV versus
CPV versus CSP) are denoted by different colors and patterns.

Figure 15 provides a number of insights:

• PPA pricing has, in general, declined over time, to the point where recent PPAs have been
priced as aggressively as $40/MWh levelized (in real, 2014 dollars), or even lower. In the
Southwest (where these low-priced projects are primarily located), pricing this low is, in
some cases, competitive with in-region wind power. 52 This is particularly the case when
considering solar’s on-peak generation profile, which can provide ~$25/MWh of TOD
value relative to wind. 53
• Although at first glance there does not seem to be a significant difference in the PPA prices
required by different solar technologies, it is notable that all of the recent PPAs in the
sample employ PV technology. Back in 2002 when the Nevada Solar One (CSP) PPA was
executed, PV was too expensive to compete at the wholesale level, but by 2009-2011 when
the other five CSP PPAs in the sample were executed, PV pricing had closed the gap.
Since then, virtually all new contracts have employed PV technology, while a number of
previously-executed CSP contracts have been either canceled or converted to PV
technology. CPV was seemingly competitive back in 2010 when the two contracts in the
sample were executed, but lack of any new contracts since then (at least within the sample)
prevents a more-recent comparison – and is perhaps telling in its own right. 54
• Smaller projects (e.g., in the 20-50 MW range) feature PPA prices that are just as
competitive as larger projects. Very large projects often face greater development
challenges than smaller projects, including heightened environmental sensitivities and
more-stringent permitting requirements, as well as greater interconnection and transmission
hurdles. Once a project grows beyond a certain size, the costs of overcoming these
incremental challenges may outweigh any benefits from economies of scale in terms of the
effect on the PPA price.

51
Because PPA prices reflect market expectations at the time a PPA is executed – which could be two years or more
in advance of when the project achieves commercial operation – the PPA execution date is more relevant than the
commercial operation date when analyzing PPA prices.
52
See, for example, the text box in Bolinger and Weaver (2013) that compares the economics of the co-located
Macho Springs wind and solar projects.
53
For further explanation, see the text box titled Estimating PV’s TOD Value in the 2013 edition of this report
(Bolinger and Weaver 2014). Also note that the levelized PPA prices shown in Figure 15 (and throughout this
chapter) already incorporate all applicable TOD factors. Not all PPAs, however, use explicit TOD factors, though in
those instances where they are not used, PV’s on-peak generation profile still presumably provides higher implicit
value (compared to wind) to the buyer.
54
That said, SunPower has been quietly rolling out its new low-concentration (7 suns) C7 CPV technology, with a 1
MWAC pilot project at Arizona State University (online in early 2013); a contract with Apple for the 20 MWAC Fort
Churchill Solar Project (scheduled to come online in 2015) to power its data center near Reno, NV; and a sale of
technology to a project in China. At present, no cost or price information is available for these projects.

32
 • Not surprisingly, the highest-priced contract in the sample comes from Long Island, which
does not enjoy the abundant sunshine of the Southwest (where most of our sample is
located – 93% of the total capacity within the PPA sample is located in CA, NV, AZ, or
NM), and where wholesale power prices are high due to transmission constraints.

Not all of the projects behind the contracts shown in Figure 15 are fully (or even partially)
operational, though all of them are still in play (i.e., the sample does not include PPAs that have
been terminated). Figure 16 shows the same data as Figure 15, but broken out according to
whether or not a project has begun to deliver power. 55 Understandably, most of the more-
recently signed PPAs in the sample pertain to projects that are still in development or under
construction, and have not yet begun to deliver electricity under the terms of the PPA. Given
that many of these same PPAs are also the lowest-priced contracts in the sample, it remains to be
seen whether all of these projects can be profitably built and operated under the aggressive PPA
price terms shown here. 56 That said, a recent and related modeling analysis (Bolinger, Weaver,
and Zuboy 2015) finds that today’s aggressive PPA prices can indeed pencil out using modeling
assumptions that are based on best-in-class PV data presented in other sections of this report.
Moreover, as described in the text box on the next page, a survey of recent solicitation responses
reveals a deep field of projects bidding into solicitations at these low prices – i.e., the recent low
prices shown in Figure 15 and Figure 16 do not appear to be one-off anomalies.

$300
Levelized PPA Price (Real 2014 $/MWh)

32 MW (New York)
$250
550 MW
$200
250
MW
$150

$100
Operating (6,668 MW, 80 contracts)
$50
Planned (1,910 MW, 29 contracts)

$0
Jan-02

Jan-03

Jan-04

Jan-05

Jan-06

Jan-07

Jan-08

Jan-09

Jan-10

Jan-11

Jan-12

Jan-13

Jan-14

Jan-15

Jan-16

PPA Execution Date

Figure 16. Levelized PPA Prices by Operational Status and PPA Execution Date

55
If a project had begun to deliver power by August 2015 – even if not yet fully operational or built out to its
contractual size – it is characterized as “operating” in Figure 16. Only those projects that were still in development
or were under construction but not yet delivering power are characterized as “planned.”
56
There is a history of solar project and PPA cancellations in California, though in many cases these have involved
projects using less-mature technologies (e.g., Stirling dish engines, compact linear Fresnel reflectors, and power
towers). For PV projects, price revisions are perhaps a more likely risk – e.g., if the solar trade dispute with China
were to harm existing module supply contracts.

33
 Solicitation Responses Reveal Deep Market at Low Prices

Although our sample of low-priced solar PPAs signed in 2014 and so far in 2015 is relatively small (21 contracts totaling 1.33
GW), a survey of developer responses to several recent utility solicitations suggests that there is considerable depth in the
market at these low price levels, at least in the Southwest. For example:

• Southwestern Public Service’s 2014 request for proposals (“RFP”) for 200 MW of solar received 53 transmission-level
project bids totaling 4,040 MW and 59 distribution-level project bids totaling 1,210 MW (for a total of 5,250 MW –
more than 26 times the 200 MW target). Of the transmission-level proposals, 2,718 MW bid levelized prices ranging
from $40-$50/MWh, while another 1,182 were priced between $50-$60/MWh. Of the distribution-level proposals
(typically featuring smaller projects), 240 MW bid levelized prices ranging from $40-$50/MWh while 600 MW were
priced from $50-60/MWh. Many of these projects are located in New Mexico, which provides a 10-year state
production tax credit that helps developers to lower PPA prices (though not all bidders assumed full receipt of the
state PTC).

• In late 2014 and early 2015, NV Energy issued two 100 MW renewable energy RFPs; bidders in the 2014 RFP were
allowed to re-bid into the 2015 RFP, and two 100 MW PV projects were ultimately selected. One of the winning
projects is priced at $46/MWh flat over 20 years (i.e., $38.6/MWh levelized in real 2014 dollars), while the other
starts at $38.70/MWh (nominal) and escalates at 3%/year over 20 years (i.e., $40.1/MWh levelized in real 2014
dollars). Of the 2,537 MW of renewable resources that bid (or re-bid) into the 2015 RFP, more than 90% were solar,
while wind and geothermal accounted for just 8% and 2%, respectively. Though no pricing information is available for
the non-winning bids, several hundred additional MW of shortlisted capacity were reportedly bid at prices very
similar to the winning bids. Moreover, NV Energy noted that the solar bids were priced lower than the wind or
geothermal bids, and also better matched its load profile.

• Austin Energy’s 2015 RFP for 600 MW of solar received 149 unique proposals totaling 7,976 MW (more than 13 times
coverage) from 33 different bidders. Almost 1,300 MW were reportedly bid at levelized prices of $45/MWh or less.

Taken as a whole, these responses suggest that there is a significant amount of utility-scale PV capacity capable (at least
with the 30% ITC) of selling electricity at very low prices in the Southwest. Moreover, at these low price levels, solar can
compete head on with wind power in terms of both price and generation profile (for more on comparisons of solar and
wind in the Southwest, see the text box titled “Estimating PV’s TOD Value” in the 2013 edition of this report, or the text box
on the co-located Macho Springs wind and solar projects in the 2012 edition).

More than two-thirds of the PV contracts in the sample feature pricing that does not escalate in
nominal dollars over the life of the contract – which means that pricing actually declines over
time in real dollar terms. Figure 17 illustrates this decline by plotting over time, in real 2014
dollars, the generation-weighted average price among all PPAs executed within a given year
(i.e., including both escalating and non-escalating contracts).

34
 250
Gen-Weighted Average PPA Price (2014 $/MWh)

2006 (7 MW, 1 PPA)

2007
(5 MW, 1 PPA)
200
2009 (956 MW, 13 PPAs)
2008 (770 MW, 3 PPAs)
150

2010 (1,676 MW, 28 PPAs)

100
2013 (492 MW, 15 PPAs) 2011 (1,177 MW, 11 PPAs)
2012 (861 MW, 10 PPAs)
2014 (449 MW, 9 PPAs)
50

2015 (885 MW, 12 PPAs)

0
2006
2008
2010
2012
2014
2016
2018
2020
2022
2024
2026
2028
2030
2032
2034
2036
2038
2040
2042
2044
2046
2048
2050
Figure 17. Generation-Weighted Average PV PPA Prices Over Time by Contract Vintage

By offering flat or even declining prices in real dollar terms over long periods of time, solar (and
wind) power can provide a long-term hedge against the risk of rising fossil fuel prices (Bolinger
2013). Figure 18 illustrates this potential value by plotting the future stream of average PV PPA
prices from contracts executed in 2014 and 2015 (i.e., the same two lines as the 2014 and 2015
vintage PPA lines in Figure 17 above) against a range of projections of just the fuel costs of
natural gas-fired generation. 57 Focusing on the 2015 PPA vintage in particular, average PPA
prices from PV contracts executed in 2015 start out higher than the range of fuel cost projections
in 2017, but decline (in real 2014 $/MWh terms) over time and eventually fall below the
reference case gas price projection by 2021 (and below the entire range of gas price projections
by 2037). On a levelized basis from 2017 through 2040, the 2015-vintage PV PPA prices come
to $42.1/MWh (real 2014 dollars) compared to $48.1/MWh for the reference case fuel price
projection, suggesting that PV may be able to compete with even just the fuel costs of existing
gas-fired generators (i.e., not even accounting for the recovery of fixed capital costs incurred by
new gas-fired generators).

Moreover, it is important to recognize that the PV PPA prices have been contractually locked in,
whereas the fuel cost projections to which they are compared are highly uncertain – actual fuel
costs could end up being either lower or potentially much higher. Either way, as evidenced by
the widening range of fuel cost projections over time, it becomes increasingly difficult to
forecast fuel costs with any accuracy as the term of the forecast increases.

57
The national average fuel cost projections come from the Energy Information Administration’s Annual Energy
Outlook 2015 publication, and increase from around $4.67/MMBtu in 2015 to $8.83/MMBtu (both in 2014 dollars)
in 2040 in the reference case. The range around the reference case is bounded by the high oil and gas resource case
on the low end, and the greater of the high oil price or high economic growth cases on the high end (since AEO
2015 does not include a low oil and gas resource case), and ranges from $4.98/MMBtu to $10.75/MMBtu (again, all
in 2014 dollars) in 2040. These fuel prices are converted from $/MMBtu into $/MWh using the heat rates implied by
the modeling output (these start at roughly 8,100 Btu/kWh and gradually decline to around 7,200 Btu/kWh by
2040).

35
 80

70

60
2014 $/MWh

50

40

30
Overall range of AEO 2015 gas price projections (converted to $/MWh terms)
20 AEO 2015 reference case gas price projection (converted to $/MWh terms)
10 Average price over time among sample of PV PPAs signed in 2014 (9 contracts, 449 MW)
Average price over time among sample of PV PPAs signed in 2015 (12 contracts, 885 MW)
0
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
Figure 18. Average PV PPA Prices and Natural Gas Fuel Cost Projections Over Time

In addition to the declining real prices over time within each PPA vintage shown in Figure 17
and Figure 18, the steady march downward across vintages is also evident in Figure 17,
demonstrating substantial reductions in pricing by PPA execution date. 58 To provide a clearer
look at the time trend, the blue-shaded columns in Figure 19 simply levelize the price streams
shown in Figure 17. Based on this sample, levelized real PPA prices for utility-scale PV projects
consistently fell by almost $25/MWh per year on average from 2006 through 2013, with a
smaller price decline of ~$10/MWh evident in the 2014 and 2015 samples. With levelized real
PPA prices now below $50/MWh on average (based on the combined 2014/2015 sample), future
price declines are likely to be much smaller than in the past.
Figure 19 also shows that the overall spread in pricing has narrowed over time – e.g., the 2013-
2015 samples show a tighter range of levelized prices than do the 2009-2011 samples –
suggestive of an increasingly mature and transparent market. Moreover, this narrowing has
occurred despite the fact that the geographic scope of the market (and sample) has broadened
with time. Although the PPAs in our sample are still heavily concentrated in the Southwest, the
market is beginning to expand to new parts of the country – notably the Southeast (see the text

58
This strong time trend complicates more-refined analysis of other variables examined in earlier chapters, such as
resource strength (though again, 93% of the capacity in the PPA price sample is in the high-insolation states of CA,
NV, AZ, and NM), tracking versus fixed-tilt, and c-Si versus thin-film. To try and control for the influence of time,
one could potentially analyze these variables within a single PPA vintage, but doing so might divide the sample to
the point where sample size is too small to reliably discern any differences. Furthermore, it is not clear that some of
these variables should even have much of an effect on PPA prices. For example, several of the PV contracts in the
sample note uncertainty over whether or not tracking systems will be used, or whether c-Si or thin-film modules will
be deployed. Yet the executed PPA price is the same regardless of the ultimate project configuration, suggesting
that the choice of tracking versus fixed-tilt or c-Si versus thin-film is (at least in these cases) not a critical
determinant of PPA pricing. This makes sense when one considers that tracking systems, for example, add up-front
costs to the project (see Chapter 3) that are recouped over time through greater energy yield (see Chapter 5), thereby
potentially leaving the net effect on PPA prices largely a wash. In support of this theory, the Public Service
Company of New Mexico estimated (based on a review of 216 solar responses to its 2012 Renewable RFP) that the
average PPA price benefit of single-axis tracking was just $3/MWh, or less than 4% of a levelized PPA price in the
mid-$70/MWh range (O’Connell 2013).

36
 box below) but also states like Utah and Idaho where utilities have had attractive avoided cost
rates (see earlier text box on “avoided cost” markets). For example, the 2015 PPA price sample
includes contracts not only in the usual Southwestern states, but also in Florida, Arkansas, and
Alabama – and at prices not too far above those seen in the Southwest.
300
Levelized PPA Price (Real 2014 $/MWh)

Generation-Weighted Average
250
Individual Contract
200

150

100

50

0
PPA Year: 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Contracts: 1 1 3 13 28 11 10 15 9 12
MW: 7 5 770 956 1,676 1,177 861 492 449 885
Figure 19. Levelized PV PPA Prices by Contract Vintage

Trend to Watch: The Rise of the South

Although the sample of PPA prices analyzed in this chapter is highly concentrated in the southwestern United States – i.e., 97% of the
8.7 GW of capacity in the sample is located in CA (68%), NV (11%), AZ (11%), TX (4%), NM (3%), and CO (1%) – there have been a
number of notable announcements over the past year about new utility-scale solar PPAs being signed at competitive prices in several
southeastern states that have not previously seen much development. The following non-exhaustive list of new contracts (only three
of which are currently included in our PPA price sample, due to lack of sufficient information on the others) illustrates this expansion
of the market to the Southeast:

• In October 2014, Georgia Power announced long-term PPAs with four “smaller” PV projects totaling 76.5 MW and six “larger”
projects totaling 439 MW. Although pricing for individual projects has not been disclosed, the average PPA price among the
“smaller” projects is reportedly $65/MWh.
• In February 2015, the Tennessee Valley Authority announced that it had signed a 20-year PPA with an 80 MW PV project in
Alabama at a price of $61/MWh.
• In April 2015, NextEra and Entergy Arkansas announced a PPA for the 81 MW Stuttgart Solar Project in Arkansas; the price is
reportedly just north of $50/MWh.
• Highlighting yet another notable trend towards direct corporate purchases of renewable power, in June 2015, Community
Energy and Amazon Web Services announced a PPA for an 80 MW PV project in Virginia (pricing was not disclosed).
• In July 2015, the Orlando (Florida) Utilities Commission announced a 20-year PPA with a 13 MW PV project priced at
$70/MWh, which is less than half the $194/MWh it is paying for a similar 5.5 MWAC project that came online in late 2011.

This trend – also evident in regional interconnection queues, as shown later in Figure 20 – is all the more notable because the
Southeast has historically not seen much renewable energy development at all (other than in North Carolina, which has had an active
solar market for a number of years, primarily featuring “avoided cost” PURPA contracts for projects of 5 MW or less that fall below
our utility-scale size threshold), due in part to fewer state-level policies like renewable portfolio standards, as well as wind resource
constraints. For example, unlike in much of the rest of the country, wind power has yet to gain much of a foothold in the Southeast –
and may find it hard to compete with solar at the price levels evident in some of these solar contracts.

37
7. Conclusions and Future Outlook

Other than the SEGS I-IX parabolic trough CSP projects built in the 1980s, virtually no utility-
scale PV, CPV, or CSP projects existed in the United States prior to 2007. By 2012 – just five
years later – utility-scale had become the largest sector of the overall PV market in the United
States, a distinction that was repeated in 2013 and 2014 and that is expected to continue for at
least the next few years. Over this same short period, CSP also experienced a renaissance in the
United States, with a number of large new parabolic trough and power tower systems – some
including storage – either achieving commercial operation or entering the commissioning phase.
Although the operating history of many these newer PV, CPV, and CSP projects is still very
limited, a critical mass of data nevertheless enables empirical analysis of this rapidly growing
sector of the market.

This third edition of LBNL’s annual Utility-Scale Solar series paints a picture of an increasingly
competitive utility-scale PV sector, with installed prices having declined significantly since
2007-2009 (but perhaps showing signs of slowing), relatively modest O&M costs, solid
performance with improving capacity factors, and record-low levelized PPA prices of around
$40/MWh in some cases and under $50/MWh on average (again, with the steady decline over the
years perhaps showing signs of slowing). Meanwhile, the other two utility-scale solar
technologies – CPV and CSP – have also made strides in recent years, but are finding it difficult
to compete in the United States with increasingly low-cost PV. 59

Looking ahead, the amount of utility-scale solar capacity in the development pipeline suggests
continued momentum and a significant expansion of the industry – both in terms of volume and
geographic distribution – over the next few years. Specifically, Figure 20 shows the amount of
solar power (and, in the inset, other resources) working its way through 35 different
interconnection queues administered by independent system operators (“ISOs”), regional
transmission organizations (“RTOs”), and utilities across the country as of the end of 2014. 60

These data should be interpreted with caution: although placing a project in the interconnection
queue is a necessary step in project development, being in the queue does not guarantee that a
project will actually be built. 61 That said, efforts have been made by the FERC, ISOs, RTOs,

59
Avian mortality has also emerged as an unexpected potential challenge to power tower technology in particular,
but also to large PV projects that, from a distance, can reportedly resemble bodies of water and attract migrating
waterfowl that are injured or killed while attempting to land in the solar field.
60
The queues surveyed include the California ISO, Los Angeles Department of Water and Power, Electric
Reliability Council of Texas (ERCOT), Western Area Power Administration, Salt River Project, PJM
Interconnection, Arizona Public Service, Southern Company, NV Energy, PacifiCorp, Midcontinent Independent
System Operator (MISO), Southwest Power Pool (SPP), Duke/Progress Energy, Public Service Company of
Colorado, Public Service Company of New Mexico, and 20 other queues with lesser amounts of solar. To provide a
sense of sample size and coverage, the ISOs, RTOs, and utilities whose queues are included here have an aggregated
non-coincident (balancing authority) peak demand of about 86% of the U.S. total. Figure 20 only includes projects
that were active in the queue at the end of 2014 but that had not yet been built; suspended projects are not included.
61
It is also worth noting that while most of the solar projects in these queues are probably utility-scale in nature, the
data are not uniformly (or even commonly) consistent with the definition of “utility-scale” adopted in this report.
For example, some queues are posted only to comply with the Large Generator Interconnection Procedures in FERC
Order 2003 that apply to projects larger than 20 MW, and so presumably miss smaller projects in the 5-20 MW

38
 and utilities to reduce the number of speculative projects that have, in recent years, clogged these
queues.

Even with this important caveat, the amount of solar capacity in the nation’s interconnection
queues still provides at least some indication of the amount of planned development. At the end
of 2014, there were 44.6 GW of solar power capacity (of any type – e.g., PV, CPV, or CSP)
within the interconnection queues reviewed for this report – more than five times the installed
utility-scale solar power capacity in our entire project population at that time. These 44.6 GW
(19.5 GW of which first entered the queues in 2014) represented nearly 14% of all generating
capacity within these selected queues at the time, in third place behind natural gas at 45% and
wind at 30% (see Figure 20 inset). The end-of-2014 solar total is also more than 5 GW higher
than the 39.5 GW of solar that were in the queues at the end of 2013, suggesting that the solar
pipeline has been more than replenished over the past year, despite the record amount of new
solar capacity that came online (and therefore exited these queues) in 2014, as well as the
impending reversion of the 30% ITC to 10% scheduled for the end of 2016.

20 Entered queue in 2014 Total in queue at end of 2014

150
Nameplate Capacity (GW)
Nameplate Solar Capacity (GW)

15
100

10 50

0
Gas Wind Solar Nuclear Coal Other
5

0
California Southwest Texas Southeast Central Northeast Northwest
(NV, AZ, UT, CO, NM)
Source: Exeter Associates review of interconnection queue data
Figure 20. Solar and Other Resource Capacity in 35 Selected Interconnection Queues

The larger graph in Figure 20 breaks out the solar capacity by state or region, to provide a sense
of where in the United States this pipeline resides. Perhaps not surprisingly (given the map of
solar resource and project location shown in Figure 3, earlier), 60% of the total solar capacity in
the queues at the end of 2014 is within California (42%) and the Southwest region (18%). This
combined 60% is down from 80% at the end of 2013, however, and is yet another indication that
the utility-scale solar market is spreading to new states and regions beyond California and the
Southwest. For example, 14% of the solar capacity in the queues at the end of 2014 resides in
Texas, followed by 10% in the Southeast and 6% in each of the Central and Northeast regions.

range. Other queues include solar projects of less than 5 MW (or even less than 1 MW) that may be more
commercial than utility-scale in nature. It is difficult to estimate how these two opposing influences net out.

39
Moreover, in terms of new solar capacity entering the queue in 2014, Texas ranked first (22%),
followed by the Southeast (18%), Southwest (16%), California (15%), and Central (14%)
regions. As the competitiveness of solar continues to improve, the market is spreading to still-
untapped parts of the country.

Though not all of the 44.6 GW of planned solar projects represented within Figure 20 will
ultimately be built, presumably most of what is built will come online prior to 2017, given the
scheduled reversion of the 30% ITC to 10% at the end of 2016. To that end, as of the end of
2014, GTM/SEIA (2015) projected a utility-scale solar pipeline of 26.7 GW in 2015-2016 (9.1
GW in 2015 and 17.6 GW in 2016), 14.1 GW of which was already contracted (6 GW in 2015
and 8.1 GW in 2016). Even if only this 26.7 GW – or, for that matter, even just the contracted
14.1 GW portion – came online prior to 2017, it would still mean an unprecedented amount of
new solar construction in 2015 and 2016. Of course, accompanying all of this new capacity will
be substantial amounts of new operational data, which we will collect and analyze in future
editions of this report.

40
 References

Data Sources

Much of the analysis in this report is based on primary data, the sources of which are listed
below (along with some general secondary sources), broken out by data set:

Technology Trends (Chapter 2):

Form EIA-860, FERC Form 556, state regulatory filings, the National Renewable Energy
Laboratory (“NREL”), the Solar Energy Industries Association (“SEIA”), trade press
articles

Installed Prices (Chapter 3):

Section 1603 grant data from the U.S. Treasury, FERC Form 1, data from applicable state
rebate and incentive programs, state regulatory filings, company financial filings, interviews
with developers and owners, trade press articles, and data previously gathered by NREL

O&M Costs (Chapter 4):

FERC Form 1 and state regulatory filings (empirical data)

Capacity Factors (Chapter 5):

FERC Electronic Quarterly Reports, FERC Form 1, Form EIA-923, state regulatory filings

PPA Prices (Chapter 6):

FERC Electronic Quarterly Reports, FERC Form 1, Form EIA-923, state regulatory filings,
company financial filings, trade press articles

In addition, the individual reference documents listed below provided additional data and/or
helped to inform the analysis.

Literature Sources

3Tier. 2013. “Solar Variance from Average: June-August 2013.”

Advanced Energy. 2014. “DC Loading of AE TX Inverters.” http://solarenergy.advanced-
energy.com/upload/File/Application%20Notes/ENG-DCLoadingTX-260-02_Web.pdf.
Barbose, Galen, and Naim Darghouth. 2015. “Tracking the Sun VIII: The Installed Price of
Residential and Non-Residential Photovoltaic Systems in the United States.” Berkeley,
CA: Lawrence Berkeley National Laboratory (LBNL). trackingthesun.lbl.gov.
Bloomberg New Energy Finance Solar Insight Service. 2015. “American PV Market Outlook,
Q1 2015.” Bloomberg New Energy Finance.

41
Bolinger, Mark. 2013. “Revisiting the Long-Term Hedge Value of Wind Power in an Era of Low
Natural Gas Prices.” LBNL-6103E. Berkeley, CA: Lawrence Berkeley National
Laboratory. http://emp.lbl.gov/publications/revisiting-long-term-hedge-value-wind-
power-era-low-natural-gas-prices.
———. 2014. “An Analysis of the Costs, Benefits, and Implications of Different Approaches to
Capturing the Value of Renewable Energy Tax Incentives.” LBNL-6350E. Berkeley, CA:
Lawrence Berkeley National Laboratory. http://emp.lbl.gov/publications/analysis-costs-
benefits-and-implications-different-approaches-capturing-value-renewable.
Bolinger, Mark, and Samantha Weaver. 2013. “Utility-Scale Solar 2012: An Empirical Analysis
of Project Cost, Performance, and Pricing Trends in the United States.” LBNL-6408E.
Berkeley, CA: Lawrence Berkeley National Laboratory.
http://emp.lbl.gov/publications/utility-scale-solar-2012-empirical-analysis-project-cost-
performance-and-pricing-trends.
———. 2014. “Utility-Scale Solar 2013: An Empirical Analysis of Project Cost, Performance,
and Pricing Trends in the United States.” LBNL-6912E. Berkeley, CA: Lawrence
Berkeley National Laboratory. http://emp.lbl.gov/publications/utility-scale-solar-2013-
empirical-analysis-project-cost-performance-and-pricing-trends.
Bolinger, Mark, Samantha Weaver, and Jarett Zuboy. 2015. “Is $50/MWh Solar for Real?
Falling Project Prices and Rising Capacity Factors Drive Utility-Scale PV toward
Economic Competitiveness: Technical and Economic Improvements Drive Utility-Scale
PV to $50/MWh.” Progress in Photovoltaics: Research and Applications, May, n/a – n/a.
doi:10.1002/pip.2630.
Danko, Pete. 2015. “Ivanpah Solar Production Up 170% in 2015.” Breaking Energy. June 17.
http://breakingenergy.com/2015/06/17/ivanpah-solar-production-up-170-in-2015/.
Feldman, David, Galen Barbose, Robert Margolis, Mark Bolinger, Donald Chung, Ran Fu,
Joachim Seel, Carolyn Davidson, and Ryan Wiser. 2015. “Photovoltaic System Pricing
Trends. Historical, Recent and Near-Term Projections. 2015 Edition.” presented at the
U.S. Department of Energy, Washington D.C., August.
https://emp.lbl.gov/publications/photovoltaic-system-prici.
Fiorelli, J, and M Zuercher - Martinson. 2013. “Array Oversizing.” White Paper. Solectria
Renewables.
First Solar. 2015. “First Solar Advanced Thin-Film Module Features.” First Solar PV Modules.
May 1. http://www.firstsolar.com/Home/Technologies-and-Capabilities/PV-
Modules/First-Solar-Series-3-Black-Module.
Fu, Ran, Ted James, Donald Chung, Douglas Gagne, Anthony Lopez, and Aron Dobos. 2015.
“Economic Competitiveness of U.S. Utility-Scale Photovoltaics Systems: Regional Cost
Modeling of Installed Cost ($/W) and LCOE ($/kWh).” In IEEE PVSC 42 - Sub-Area 9.5
Power Electronics, Energy Storage and Other BOS Components. New Orleans.
GTM Research, and SEIA. 2015. “U.S. Solar Market Insight Report: 2014 Year in Review.”

42
Middlekauff, Charles, and Grady Mathai-Jackson. 2015. “Pacific Gas and Electric Company’s
(U 39E) Third and Final Photovoltaic Program Compliance Report.” California Public
Utilities Commission Compliance Report Application 09-02-019. San Francisco, CA:
Pacific Gas and Electric Company.
O’Connell, Patrick. 2013. Direct Testimony of Patrick J. O’Connell In The Matter Of Public
Service Company Of New Mexico’s Renewable Energy Portfolio Procurements Plan for
2014 And Proposed 2014 Rider Rate Under Rate Rider No. 36. Case No.13-00183-UT.
Perez, Richard. 2012. “Solar Direct Normal Irradiance Data 1998-2009 with 10km Resolution
(WGS_1984).” State University of New York/Albany , National Renewable Energy
Laboratory (NREL). http://www.nrel.gov/gis/data_solar.html.
Pihowich (Amonix), C. 2011. “Concentrated Photovoltaic (CPV) Solar Power Solutions in the
Southwest.” presented at the Texas Renewables 2011 (Texas Renewable Energy
Industries Association), November 7.
Solar Energy Industries Association (SEIA). 2015. “Major Solar Projects in the United States:
Operating, Under Construction, or Under Development.” http://www.seia.org/research-
resources/major-solar-projects-list.
Stern, Ray. 2015. “Solana Solar Plant’s Weak Output Will Improve, Officials Say.” Phoenix
New Times, June 16. http://www.phoenixnewtimes.com/news/solana-solar-plants-weak-
output-will-improve-officials-say-7414760.
Vaisala. 2014. “Summer 2014 Solar Performance Map.”
http://www.vaisala.com/Vaisala%20Documents/Vaisala%20News%20Articles/Misc/201
4%20Solar%20Performance%20Map.pdf.
———. 2015. “2014 US Solar Performance Maps.”
http://www.vaisala.com/Vaisala%20Documents/Vaisala%20News%20Articles/Misc/201
4_US_Solar_Performance_report.pdf.

43
Key Report Contacts Acknowledgments
Lawrence Berkeley National Laboratory: The authors thank Minh Le, Elaine Ulrich, Odette Mucha, Christina Nichols,
Josh Honeycutt, Ammar Qusaibaty, Dan Boff, and Craig Connelly of the U.S.
Mark Bolinger
Department of Energy's Solar Energy Technologies Office for their support of
603-795-4937; [email protected]
this work. The authors would also like to thank Justin Baca and Aaron Holm of
Joachim Seel the Solar Energy Industries Association (“SEIA”) for maintaining and sharing
510-486-5087; [email protected] SEIA’s list of major solar projects in the U.S., and Joe Rand (Lawrence
Berkeley National Laboratory) for his work on the project location and insolation
map. Thanks also to the following individuals for reviewing earlier drafts of this
Download the Report report: Amory Lovins (Rocky Mountain Institute), Greg Wagoner (Black and
http://emp.lbl.gov/reports/re Veatch), Jim Baak (Vote Solar), Rachel Golden (Large Scale Solar
Association), Cara Marcy (Energy Information Administration), David Feldman
and Ran Fu (National Renewable Energy Laboratory), Ryan Wiser and Naim
Darghouth (Lawrence Berkeley National Laboratory).Of course, any
remaining errors or omissions are the sole responsibility of the authors.

Cover photos courtesy of Sempra U.S. Gas & Power, NREL, and BrightSource
Limitless.

The work described in this report was funded by the U.S. Department of Energy’s
Solar Energy Technologies Office, within the Office of Energy Efficiency and
Renewable Energy, under Contract No. DE-AC02-05CH11231.

Disclaimer
This document was prepared as an account of work sponsored by the United States
Government. While this document is believed to contain correct information, neither
the United States Government nor any agency thereof, nor The Regents of the
University of California, nor any of their employees, makes any warranty, express or
implied, or assumes any legal 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 its 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,
or The Regents of the University of California. The views and opinions of authors
expressed herein do not necessarily state or reflect those of the United States
Government or any agency thereof, or The Regents of the University of California.

Ernest Orlando Lawrence Berkeley National Laboratory is an equal opportunity

employer.