You are on page 1of 26

ll

Article
The curtailment paradox in the transition to
high solar power systems
Bethany Frew, Brian Sergi, Paul
l
10.0 l
l
l
llll Denholm, Wesley Cole,
No Storage ll
Nathaniel Gates, Daniel Levie,
VRE curtailment (%)

Robert Margolis
l
7.5 bethany.frew@nrel.gov
l
l
l ll
l
l
l
l
l
l Highlights
l
l
ll l
l Curtailment varies by thermal
5.0 l
flexibility, operating reserve rules,
l and other factors
l
l l
l
ll
l Thermal generator flexibility
2.5 l matters most at mid-PV (25%–40%
l
l l penetration) levels
l
ll
l
l l
l l
l System cost and curtailment
0.0 l
decline when VRE and storage
20 25 30 35 40 45 provide operating reserves
PV penetration (%)
PV gens suppress their revenue
potential from operating reserves
Base 2x Ramp Rate
Zero Min Gen Copperplate when providing them
2x Min Gen 5 Min
Zero Up/Down Time DA−RT
1.1x Up/Down Time No Storage
10% Ramp Rate

The adoption of PV and battery storage has accelerated globally in recent years,
driven by rapid cost declines. A corresponding increase in curtailment is
anticipated as PV growth continues. This study explores the effect of system
flexibility options on curtailment across increasing PV penetration levels. Results
highlight a paradox where thermal generator flexibility matters most only at mid-
PV penetration levels (25%–40%), and a misalignment exists where PV provides
system value by providing operating reserves without sufficient opportunity for
monetary compensation.

Frew et al., Joule 5, 1143–1167


May 19, 2021 ª 2021 Elsevier Inc.
https://doi.org/10.1016/j.joule.2021.03.021
ll

Article
The curtailment paradox in the transition
to high solar power systems
Bethany Frew,1,3,* Brian Sergi,2 Paul Denholm,1 Wesley Cole,1 Nathaniel Gates,1 Daniel Levie,1
and Robert Margolis1

SUMMARY Context & scale


Rising penetrations of variable renewable energy (VRE) in power The adoption of PV and battery
systems are expected to increase the curtailment of these resources storage has accelerated globally
because of oversupply and operational constraints. We evaluate the in recent years, driven by rapid
effect on curtailment from various flexibility approaches, including cost declines. A corresponding
storage, thermal generator flexibility, operating reserve eligibility increase in curtailment is also
rules, transmission constraints, and temporal resolution, by using anticipated as PV growth
a highly resolved realistic system. Results reveal two aspects of a continues. Although excessive
curtailment paradox as the system evolves to higher solar penetra- curtailment can affect resource
tion levels. First, thermal generator parameters, especially in re- financial viability, economic
stricting minimum operating levels and ramp rates, affect VRE curtailment is a new normal in grid
curtailment more in mid-PV penetration levels (25%–40%) but operations by providing flexibility
much less at lower (20%) or higher (45%) PV penetration levels. to ensure grid reliability. This
Second, although allowing VRE and storage to provide operating study systematically explores the
reserve results in significant operating costs and curtailment bene- effect of other flexibility options
fits, the price suppression effect from these resources reduces in- on curtailment levels as PV
centives for PV to provide operating reserves with curtailed energy. penetrations grow. These include
battery storage, the operational
flexibility of thermal generators,
INTRODUCTION
transmission, and allowing VRE
A common attribute of power systems with increasing penetration levels of variable and storage to provide operating
renewable energy (VRE) is a corresponding—and often non-linear—increase in the reserves, among others. This
frequency and magnitude of curtailed energy.1–9 Curtailment is a reduction in sched- study provides specific insights to
uled energy delivery when generation exceeds demand and/or system conditions (1) grid planners and operators on
impose operational constraints that preclude all the available energy from being uti- the particular importance of
lized.3,5,9 Curtailment of renewables results from oversupply and a lack of system thermal generator operations at
flexibility, which can include transmission congestion, minimum generation levels mid-PV penetration levels (25%–
of thermal generators or hydropower, or back-feeding in the distribution sys- 40%) and (2) market designers on
tem.10–14 the potential need to revise
operating reserve eligibility rules
The large quantity of economic curtailment that is observed in grid integration and compensation structures as
studies with very high VRE penetration levels and is also suggested by empirical ev- PV penetrations increase.
idence reflects a possible paradigm shift in how future power systems will operate.
Curtailment might be a ‘‘new normal’’ for everyday operations,4,15 as the ability to
curtail output from VRE sources is an important source of flexibility by helping to
maintain supply-and-demand balance and system frequency. It is also increasingly
seen as a potential source for operating reserves.16 This new normal idea is sup-
ported by utility planning and operations that account for a certain level of curtail-
ment, or availability, of VRE resources.17,18

Although there is a general acknowledgment that a lack of system flexibility leads to


curtailment, there is no clear understanding of the importance of individual system

Joule 5, 1143–1167, May 19, 2021 ª 2021 Elsevier Inc. 1143


ll
Article

contributors to curtailment, particularly in systems with high VRE penetration levels.


Previous work has shown that curtailment can be substantially reduced by changing
the operational practices of solar photovoltaic (PV) plants or by changing the flexi-
bility of thermal plants.16,19 Several works have highlighted the importance of min-
imum generation levels of thermal units, demonstrating that reducing these levels
can result in favorable changes to not only VRE curtailment but also to plant oper-
ating hours, startups, maintenance costs, system-wide operating costs, and net rev-
enues.11,12,19 For example, reducing minimum generation levels of thermal plants by
25% and 50% (from baseline levels) allows those generators to stay online for more
hours across the year, resulting in an overall reduction in startup costs by 21% and
32%, respectively.20 Furthermore, coupling a reduced minimum generation level
with thermal energy storage yields similar curtailment, startup, and operating costs
benefits, while also reducing operating reserve violations and unserved load.21
Conversely, other work has shown how increasing these minimum generation levels
by a factor of about 2 results in non-linear increases in curtailment for systems with
moderate VRE penetrations (33%–40% VRE), and it also suggests that reducing min-
imum generation levels might be potentially more effective in reducing curtailment
than storage.11 Other work has demonstrated how minimum generation levels,
startup and shutdown cost, and must-run generators (potentially including self-
scheduled generators) can significantly increase curtailment, even when multiple
storage technologies are available.22 Among these studies, the annual PV energy
penetration levels were generally in the 5%–20% range, with some scenarios reach-
ing 28%.11,16,19,20 Other studies did not include any PV (i.e., only wind),21,22 and the
largest reported penetration was 50% PV in one scenario with limited exploration of
thermal generator sensitivity to only minimum generation levels and startup costs.12

More broadly throughout the literature, studies looking at both planning and oper-
ational considerations have made high-level observations about the interaction of
the available flexibility of the system and curtailment.23,24 For example, Sepulveda
et al. show that VRE curtailment can be reduced from nearly 60% to less than 15%
of the available resources by utilizing deployable low-carbon resources in a fully dec-
arbonized system.2 Others have noted significant investment decision effects from
the representation of operational flexibility (e.g., inter-hour, operating reserve,
and maintenance constraints), temporal resolution, and system curtailment, high-
lighting the important interplay of operational constraints and curtailment.25–27

This paper adds a novel contribution to the literature in two ways. First, it provides a
more extensive exploration of operational parameters that can affect curtailment by
using a highly detailed and resolved production cost modeling (PCM) database that
is informed by capacity expansion modeling (CEM). And second, it explores gener-
ally higher PV penetration systems that also include significant levels of wind and
storage with some demand response; the highest VRE penetration system satisfies
load with 44% PV penetration. This scenario also has 31% annual generation from
wind and a 36% storage penetration in relation to peak demand on a capacity basis.
Although the use of CEMs and PCMs is not new, the full suite of experiments we have
designed has not been reported in the literature for a realistic, highly resolved sys- 1Strategic
Energy Analysis Center, National
tem systematically across the PV penetration levels we are exploring. Our results Renewable Energy Laboratory, Golden, CO
80401, USA
reveal a novel framing of a solar curtailment ‘‘paradox’’ relating to the role of thermal 2Power Systems Engineering Center, National
generator flexibility on curtailment as a function of PV and thermal generator pene- Renewable Energy Laboratory, Golden, CO
tration levels, as well as a misalignment of the system value provided versus poten- 80401, USA
3Lead
tial compensation for PV providing operating reserves in a competitive wholesale contact
market context. Here, we focus on PV for two reasons: (1) it is poised to provide *Correspondence: bethany.frew@nrel.gov
the largest share of new VRE deployed in the United States28,29 and many other https://doi.org/10.1016/j.joule.2021.03.021

1144 Joule 5, 1143–1167, May 19, 2021


ll
Article

countries,30 and (2) it shows a more rapid increase of curtailment with penetration
because of its coincident nature. Furthermore, the growth of PV paired with battery
storage across our scenarios, as determined by the CEM, is consistent with the cur-
rent status of many U.S. interconnection queues, which have very large amounts of
these technologies.31

Our analysis leverages multiple modeling tools to mirror the investment-operations


sequence often used in utility planning practices. We begin with the use of a CEM to
establish least-cost future build-out scenarios with increasing penetration levels of
VRE resources that are dominated by PV across the study footprint. CEMs optimize
the build-out and retirement of generators, transmission, and storage resources
across multiple decades, subject to the economic outlook for each resource type
and a wide range of system constraints, including load balancing, operating re-
serves, planning reserve, and policy targets, among others. However, CEMs do
not provide a granular representation of system operations, so we then employ a
utility-grade PCM that optimizes the least-cost chronological economic unit commit-
ment and dispatch of generation and transmission to assess the detailed operation
of each future build-out scenario across the study footprint. PCMs optimize opera-
tions in a highly resolved temporal and spatial manner, and they account for the
many operational and cost considerations when scheduling the system. These
include generators-specific variable operations and maintenance costs, fuel costs
paired with a heat rate curve, startup and shutdown costs, minimum and maximum
operating levels, minimum up and down times, and ramping limits, among others.
Given that PCMs only evaluate system operations, each of the CEM build-outs is pro-
vided as an input. See experimental procedures for more details on these modeling
tools and methods.

Our analysis uses a system that is roughly based on the Los Angeles Department of
Water and Power (LADWP) generation and transmission system, and it leverages
data sets developed for the LA100 study.32 More details on this system are provided
in experimental procedures. This system is an area with rapidly growing levels of PV
and is therefore likely to experience increasing curtailment. Because LADWP acts as
an independent balancing authority, and for the sake of computational tractability,
we model the operations of this footprint as an islanded system with all energy
derived from utility-owned or contracted resources throughout the Western Inter-
connection, but with no additional short-term market purchases or sales. We eval-
uate power system operation under six scenarios of future PV deployment, which
are shown in Figure 1. Because PV is the dominant source of generation and a
main driver of curtailment, we refer to the different build-outs by their PV penetra-
tion level.

This study has two objectives: (1) determine which operational factor or factors most
contribute to curtailment in these high PV futures and (2) quantify the potential value
of PV providing operating reserves with this curtailed energy (i.e., reserves in the up
direction, or ‘‘raise’’) or in PV further curtailing in order to provide down direction
(i.e., ‘‘lower’’) operating reserves. We evaluate these questions by assessing several
sensitivities. For operational factors, we explore bookend sensitivities related to
thermal plant flexibility (including ramp rates, minimum generation levels, and min-
imum up-and-down times), as well as other operational constraints, such as transmis-
sion network constraints, the deployment of storage, and the use of 5-min resolution
dispatch. We also evaluate sensitivities related to operating reserve eligibility, spe-
cifically whether utility-scale VRE resources and/or storage can provide operating re-
serves (distributed PV is precluded from providing operating reserves in all

Joule 5, 1143–1167, May 19, 2021 1145


ll
Article

A B

Figure 1. Overview of the six future build-out cases explored in this analysis
Each case is referenced in the x axis labels by its PV penetration as a percentage of annual generation.
(A) The left shows installed capacity across the LA-like study footprint.
(B) The right illustrates the total annual generation in the Base scenario, along with total demand and demand from storage charging.

scenarios). The goal of these scenarios is to provide a guide regarding which class of
operational factors might be most important to enhance or better manage as sys-
tems progress to higher PV penetration levels.

The PCM determines the least-cost schedule for energy and operating reserves
across all eligible generators for each time step. Eligible generators might provide
operating reserves up to their operating limits (i.e., within ramping, maximum
resource availability, and online status constraints). In scenarios where reserves are
provided by VRE, the model ensures the upward reserve capacity is available on
the basis of the forecasted availability. VRE is allowed to only provide relatively short
duration reserve products, typically requiring only 30 min of response. In all sce-
narios except DA-RT, this scheduling is done assuming 24 h of perfect foresight of
load and VRE availability. Assuming perfect foresight might underestimate the flex-
ibility needs of the system caused by forecast errors. Our results, thus, likely under-
estimate curtailment levels and the effect of our sensitivity cases on curtailment. We
note that our DA-RT scenario is intended to explore the role of forecast errors; results
yielded very little difference from the Base case with perfect foresight, but future
work should explore varying lookahead horizons and resolutions as wells as varying
degrees of forecast accuracy.

These sensitivity scenarios are outlined in Table 1. We evaluate each sensitivity for
each of the six PV penetration levels, which yields 84 total scenarios.

RESULTS AND DISCUSSION


The PCM results for each scenario were analyzed for curtailment, operating costs,
generation and storage operations, and operating reserve price trends. Results
were also evaluated for unserved load and operating reserve violations, which are
key metrics of grid reliability and resource adequacy. The load was met in all time-
steps in all scenarios.

Table 2 summarizes key statistics for the Base sensitivity case across the six build-out
levels. The table reveals the three main dimensions analyzed in this study: PV

1146 Joule 5, 1143–1167, May 19, 2021


ll
Article

Table 1. Sensitivity scenarios grouped by category of the parameters they adjust (thermal plant flexibility, operating reserve provision, or
operational constraints)

Category Sensitivity Description


Baseline Base hourly resolution real-time operations with
Base values; utility-scale VRE eligible to
provide operating reserves (distributed PV
cannot provide reserves)
Thermal plant flexibility Zero Min Gen minimum generation levels for online
dispatchable generators set to zero
23 Min Gen minimum generation levels for dispatchable
generators increased to double the base value,
up to a maximum of 1 (as a fraction of
nameplate capacity)
Zero Up/Down Time minimum on/off times for dispatchable
generators set to zero
1.13 Up/Down Time minimum on/off times for dispatchable
generators set to 1.1 times Base value
10% Ramp maximum ramp Up/Down rates for
dispatchable generators set to 10% of Base
value
23 Ramp maximum ramp Up/Down rates for
dispatchable generators set to double Base
value
Eligibility of VRE and storage resources to No VRE Reserves utility-scale VRE (stand-alone and VRE portion
provide operating reserves of hybrid systems) ineligible to provide
operating reserves
No Storage Reserves battery storage (stand-alone and battery
portion of hybrid systems) and PSH ineligible to
provide operating reserves
No Storage or VRE Reserves all utility-scale VRE and battery and PSH
storage ineligible to provide operating
reserves
Other operational constraints 5-Min 5-min resolution real-time operations; other
cases use hourly resolution
DA-RT unit commitment for certain units occurs in day-
ahead (DA) simulation using forecasted wind
and solar time series, with final dispatch
determined by a real-time (RT) simulation with
actual wind and solar time series
No Storage all storage (battery, PSH, and CAES) replaced
with equivalent capacity of gas-CT; serves as
counterfactual case
Copperplate transmission limits not enforced
These scenarios are run for each of the six PV penetration levels, resulting in 84 model runs.
Dispatchable generators include natural gas combined cycle (Gas-CC), steam (Gas-ST), and combustion turbine (Gas-CT); coal; geothermal; and biopower. PSH,
pumped-storage hydropower; CAES, compressed air energy storage.

penetration level, storage penetration level, and the resulting curtailment. The six
build-out levels evaluated are based on least-cost build-outs from a CEM with
increasing levels of PV and storage, though we note that these build-outs represent
only a small subset of possible generator combinations and, thus, do not exhaustively
represent the full PV-storage-curtailment spectrum. We emphasize that, although
these build-outs are based on the least-cost CEM results for this test system under
the study assumptions, they do not necessarily represent the ideal combination of
PV and storage for any system under any set of policy or economic assumptions. In
other words, these build-outs reflect one plausible pathway to higher penetrations
of both PV and storage. Furthermore, the No Storage scenario provides a fixed level
of storage (i.e., no storage) to highlight how increasing levels of PV penetration affect
curtailment. As a point of reference, in real power systems, annual average curtail-
ment rates from 2013 to 2018 across the U.S. ISO market areas were approximately
6% for wind and 1.5% for solar.3

Joule 5, 1143–1167, May 19, 2021 1147


ll
Article

Table 2. Summary statistics for the Base scenarios, which are based on least-cost build-outs from
the CEM

Curtailment
Renewable penetration annual avg (max annual avg (max
instantaneous)% of total annual Capacity penetration instantaneous)
generation % of peak load % of available resource
PV VRE RE Storage VRE PV
21% (71%) 28% (79%) 38% (84%) 6% 0.1% (28%) 0.1% (21%)
26% (81%) 40% (90%) 52% (92%) 8% 1.1% (36%) 1.0% (43%)
30% (87%) 44% (92%) 56% (95%) 10% 2.8% (46%) 3.0% (50%)
33% (88%) 51% (93%) 63% (96%) 11% 6.7% (40%) 5.7% (50%)
38% (94%) 63% (96%) 76% (96%) 24% 5.8% (47%) 6.8% (54%)
44% (96%) 75% (96%) 86% (97%) 37% 8.2% (57%) 9.9% (65%)
Renewable penetration values are shown for PV resources only, which include utility-scale, stand-alone,
and hybrid PV systems, as well as distributed PV resources; VRE resources, which include wind and PV;
and all renewable energy (RE) resources, which include VRE, biopower, geothermal, and hydropower.
Instantaneous curtailment levels shown are for hours with available VRE generation of at least 1% of
installed capacity.

Results from the PCM analysis reveal two key findings that collectively capture the
solar curtailment paradox. By definition, a paradox involves seemingly contradictory
components that, when investigated, are shown to be true.

(1) Thermal generator flexibility—such as flexibility regarding ramp rate, mini-


mum generation levels, and minimum up-and-down time—has a much larger
effect on VRE curtailment in a ‘‘transition zone’’ at mid PV levels (roughly 25%–
40%), but less so at lower (about 20%) or higher (about 45%) penetration
levels. Another way to view this transition zone concept is as a ‘‘Goldilocks
zone’’ where it is just the right combination of PV and thermal generators to
result in thermal flexibility effects to curtailment. This first observation of ther-
mal plant flexibility only having a significant effect in the transition zone and
much less so at lower and higher PV penetrations is initially contradictory to
intuition, but it proves to be well-founded once further analyzed. By defini-
tion, this is a paradox. In comparison to these thermal operational factors,
much less significant curtailment effects are observed from transmission con-
straints, forecasting errors, or temporal resolution.
(2) At high PV penetrations, model scenarios in which VRE and storage resources
cannot provide operating reserves yield significant curtailment and operating
cost increases. This reflects the overall value of allowing these resources to
provide operating reserves. Despite this value, when VRE and storage pro-
vide operating reserves, operating reserves prices consistently remain nearly
zero during periods of PV curtailment. This misalignment of system value with
sufficient opportunity for monetary compensation is initially contradictory,
but it is well explained by further investigation. Thus, this constitutes a second
paradox. The low operating reserve prices are driven by storage, which has
near-zero cost and ample capacity to provide reserves, especially during pe-
riods of curtailment. The overall low prices could limit the incentive for PV to
provide operating reserves with curtailed energy in a market environment.

We discuss these two findings in detail in key finding 1: thermal generator flexibility
matters at mid PV penetration levels but not at low or high levels and key finding 2:
allowing PV and storage to provide operating reserves results in system-wide oper-
ating costs and curtailment reductions, but curtailed PV has limited revenue poten-
tial from operating reserve markets. Overall, our work highlights the very nuanced

1148 Joule 5, 1143–1167, May 19, 2021


ll
Article

nature of flexibility and its role in the solar curtailment paradox. Flexibility comes in
many forms, and our results indicate that the role of storage (penetration level, oper-
ating reserve eligibility rules, and configuration) and thermal generators (penetra-
tion level and operating flexibility capabilities) are significant drivers of the overall
flexibility needs of the system and, therefore, also of system curtailment levels.

Although these results reflect assumptions and attributes inherent to our LADWP
system, qualitative insights can still be transferred to other regions deploying large
shares of PV and storage. These insights include the presence of a transition zone
with a greater sensitivity to thermal generator flexibility, as well as the awareness
of a tradeoff in system value versus asset-level remuneration when allowing PV re-
sources to provide operating reserves. On the first key finding, system planners
and operators can anticipate providing non-linear attention—initially less, then
more, then less—to the capabilities and operational decisions of thermal generators
as more PV is added to the system. For example, a self-scheduled thermal generator
might have little effect on PV curtailment at low PV penetration, but then have a high
effect as the PV penetration increases, until the system is dominated by PV and other
VRE resources and that thermal generator is utilized far less. Therefore, a phased
approach to the operational paradigms for thermal generators would be prudent
for any system. For the second key finding, the system value enabled by PV
providing operating reserves is likely to be inconsistent with the compensation
that the PV resources receive given the low operating reserve prices observed in
our modeling during hours when PV can provide reserves. System market designers
might need to revise compensation structures to accurately and efficiently reflect in-
dividual resource contribution to system value.

Future work should continue to explore the sensitivity factors included in this study
on other power systems with additional storage penetration level and configuration
assumptions (e.g., storage durations and inclusion in hybrid resources). In addition,
given that this study focuses only on curtailment reductions and not the full set of
benefits resulting from thermal generator flexibility improvements, future work
should conduct a full cost-benefit analysis of thermal generator flexibility upgrades.
This would ideally involve incorporating the cost to enhance generator flexibility into
the CEM to assess the overall cost competitiveness with storage and other flexibility
options on the investment timescale, as well as subsequent PCM runs to assess oper-
ational cost and price outcomes. All of this would need to be done generator-by-
generator to account for the age and other design specifications. One previous
cost-benefit analysis, which explicitly documents retrofit costs, has shown that retro-
fitting natural gas- and coal-fired generators to have lower minimum generation
levels yields net-benefits to the system, though results vary from generator to gen-
erators and are sensitive to fuel prices and fleet composition.33 Other work has
calculated cycling and other costs resulting from more flexible operations of thermal
generators.34

Key finding 1: Thermal generator flexibility matters at mid PV penetration


levels but not at low or high levels
Our first key finding relates to the effect on VRE curtailment by various operational
factors across different PV penetration levels. We find that thermal plant flexibility
only affect VRE curtailment at mid PV penetration levels, where enough PV is present
to result in overgeneration and enough thermal capacity remains for operational
sensitivities to affect unit commitment and dispatch outcomes. This aspect of the so-
lar curtailment paradox reveals the importance of the evolution and interaction of
solar with the rest of the system.

Joule 5, 1143–1167, May 19, 2021 1149


ll
Article

15

No Storage
VRE curtailment (%)

l
37%
l
10 l
l
llll
l
l

24%
l
11% l
l l
l
l ll
l ll

l
lll l
5
10% l
l
l
l l
ll
ll
l
8% l
l l
6% l
ll
l
l l
l l
l
0 l

20 25 30 35 40 45
PV penetration (%)

Base Zero Up/Down Time 2x Ramp Rate DA−RT


Zero Min Gen 1.1x Up/Down Time Copperplate No Storage
2x Min Gen 10% Ramp Rate 5 Min

Figure 2. Comparison of total annual PV penetration and curtailment for all sensitivities related
to thermal plant flexibility and operational constraints
The percentages indicate the share of storage capacity installed (as a percentage of peak load) for
each PV penetration level.

Figure 2 provides a comparison of VRE curtailment across all the PV and system
build-out levels. The figure highlights the findings from sensitivities related to ther-
mal generator flexibility (applied to all coal, natural gas, geothermal, and biopower
by using a consistent multiplier adjustment to the applicable underlying base values)
and other operational constraints (copper-plate transmission, temporal resolution,
and forecasting errors). See Table S3 for a summary of the underlying base genera-
tors values and Figures S6–S8 for a summary of the flexibility parameters after the
sensitivity adjustments were applied.

We first note the dramatic increase in curtailment that results in the No Storage case, re-
flecting the declining marginal contribution of VRE resources and, in particular, PV. This
serves as a counterfactual case to all other scenarios with some amount of storage, and it
reflects the difficulty of using only VRE resources without storage to achieve higher PV
penetration levels. Although the simulated storage represents electricity storage (in
the form of batteries, pumped-storage hydropower (PSH), or similar technologies),
this could also be a proxy for demand response and load shifting with similar overall
physical capacity. We note that this study did include a small amount of demand
response (see Figure 1). Under the assumptions in this study, the system cannot continue
to increase the PV penetration past approximately 30% PV penetration without the abil-
ity to utilize storage; we caution that this exact threshold might be system-specific, and
future work is needed to confirm a generalizable value across systems with different
sizes, configurations, and inter-regional exchanges. The ability of storage to reduce
curtailment at increasing PV penetration is consistent with other literature2,4,24,35–37
and is one reason storage is deployed as PV penetration increases. We highlight this
important role of storage, which is not explored further in this study apart from the static
penetration levels embedded in our six system build-outs.

Beyond the No Storage case, we find that the magnitude and spread of curtailment
is small across all sensitivity cases at the lowest and highest PV penetration levels

1150 Joule 5, 1143–1167, May 19, 2021


ll
Article

(20% and 45% penetration, respectively). In the lowest penetration case, insuffi-
cient VRE exists for changes to thermal flexibility to have a meaningful effect on the
system. In the highest penetration case, the extensive use of VRE resources results in
higher curtailment levels, whereas the limited reliance on thermal resources (see Fig-
ure S3) means changes to thermal flexibility have limited ability to affect that
curtailment.

In the transition zone between these two endpoints, however, thermal flexibility param-
eters have a much larger effect on curtailment. This is reflected by the larger spread of
curtailment values across all scenarios in the 25%–40% PV penetration range, as shown
in Figure 2. Minimum generation levels of thermal generators are the largest drivers of
curtailment, with the sensitivities on this parameter yielding the largest changes. More-
stringent minimum up and down times or ramp rates also produce increased curtailment
but relaxing the ramp further yields no change over the Base case.

These results indicate the importance of thermal plant flexibility in systems that are
transitioning from thermal-dominant to VRE-dominant fleets. When PV penetrations
are low, there is insufficient curtailment to result in significant differences from modi-
fying operational factors. When PV penetrations are sufficiently high, there is not
enough incentive to use the remaining thermal capacity to yield significant curtail-
ment effects from adjusting thermal operating characteristics. The transition zone
of 25%–40% PV penetration range has sufficient levels of both thermal generators
and PV to yield noticeable curtailment effects from the level of thermal generator
flexibility. These results suggest a phased approach might be needed to support
the ongoing transformation of power systems, particularly as PV and storage pene-
trations grow and thermal resources decline.

To further explore the operational changes resulting from the most impactful scenarios,
which include the thermal generator minimum generation levels, up/downtimes, and
ramp rates scenarios, we compare the total annual generation differences between
the Base case and these sensitivity cases across a subset of the PV penetration levels
in Figure 3 (all scenarios and PV penetration levels are provided in Figure S1). These dif-
ferences are driven by key trade-offs between thermal generator cost and flexibility,
particularly when a mix of VRE and thermal generators is on the system. Overall, mini-
mum generation levels and ramp rates result in the largest differences in dispatch.

In scenarios with greater thermal generator flexibility (i.e., Zero Min Gen, Zero Up/
Down Time, and 23 Ramp Rate), there is generally more solar, Gas-CC, and coal
generation and less gas-CT generation. With the additional flexibility afforded to
the coal generators in these cases, the system can utilize more of the zero-mar-
ginal-cost solar and relatively low-marginal-cost coal and Gas-CC. This is especially
true in the Zero Min Gen scenario (which, on whole across all PV penetration levels,
has the greatest overall generation difference among any of the thermal generator
sensitivity cases, see Figure S1) and Zero Up/Down Time scenarios. As an example,
we show in Figure 4 4 days of dispatch for the Base and Zero Min Gen, among other
scenarios, for the 33% PV penetration level; the system-wide operating cost and
curtailment during these days are also displayed as Figure 4 annotations. For the
four days shown in the figure, the Zero Min Gen sensitivity results in 3.2% greater uti-
lization of solar and 20% less curtailment, which yields about $0.33 million (16.5%) in
operating cost savings relative to the Base case.

In scenarios with less thermal generator flexibility (i.e., 23 Min Gen, 1.13 Up/Down
Time, and 10% Ramp Rate), there is generally less coal and solar generation, with

Joule 5, 1143–1167, May 19, 2021 1151


ll
Article

21%

1000

0
Change in generation from base (GWh)

−1000

33%

Curtailment
1000
Other
0 PV
Gas CT
−1000
Gas CC
Coal
44%

1000

−1000

Zero Min 2x Min Zero Up/ 1.1x Up/ 10% Ramp 2x Ramp
Gen Gen Down Down Rate Rate
Time Time

Figure 3. Change in total annual generation and curtailment relative to the Base case for select PV
penetration levels for each of the sensitivity runs that adjust thermal generator flexibility (i.e., min
gen levels, up/down time, and ramp rates)
Each bar reflects the difference between the sensitivity and Base scenario for that PV penetration
level; positive values indicate an increase in the generation of that technology in the sensitivity
relative to the Base case, and negative values indicate a decrease. ‘‘PV’’ includes the PV portion of
PV-battery hybrids and stand-alone PV. ‘‘Other’’ includes wind, biopower, geothermal,
hydropower, demand response, oil-gas-steam (OGS), and all storage technologies, including the
storage portion of PV-battery hybrids, stand-alone batteries, PSH, and CAES. Technologies
grouped into the ‘‘other’’ category generally had small changes relative to the Base. ‘‘Curtailment’’
refers to total VRE curtailment. A comparable difference plot for all scenarios and PV penetration
levels with additional technology breakouts can be found in Figure S1.

nuanced trade-offs among natural gas-fired generation, as shown in Figure 3. In the


10% Ramp Rate scenario, we see more gas-CT and less Gas-CC generation at both
low (about 20%) and higher (about 45%) PV penetration levels. This is because Gas-
CTs have the fastest ramp rates of all generators modeled and are, thus, preferred
over the Gas-CCs in order to satisfy the system flexibility needs in both thermal-
dominant and VRE-dominant systems. However, in the transition zone (roughly
25%–40% PV penetration levels), the relatively larger share of VRE generation pre-
cludes the need for energy from very inflexible coal units in favor of running Gas-
CC units with their middle-tier costs and ramping capability, and there is still a
need for additional flexibility from the Gas-CTs. This transition zone interaction is
captured by the example shown in Figure 4 at the 33% PV penetration level, where
the 10% Ramp Rate scenario has 9 GWh more Gas-CC generation and 1 GW less coal
generation while maintaining similar levels of gas-CT generation, in relation to the
Base scenario across the four days. In this 10% Ramp Rate scenario, the reduced
ramp rates force units to turn on earlier to ensure availability in the evening,
increasing curtailment by about 13% in relation to the Base scenario for this 4-day
example. The 23 Min Gen scenario further reduces coal plant flexibility and results
in 16 GWh (150%) more Gas-CC generation than in the Base scenario.

Among the other operational sensitivities (i.e., 5 min, Copperplate, and DA-RT), we see
minimal effect to total VRE curtailment in relation to the Base case. Future work should

1152 Joule 5, 1143–1167, May 19, 2021


ll
Article

Figure 4. Dispatch plots for the Base, Zero Min Gen, 23 Min Gen, and 10% Ramp Rate scenarios
at the 33% PV penetration level
Dispatch is shown for 4 days total, which are centered on the hour of minimum net load in the Base
case surrounded by 2 days on each side.

continue to evaluate the effect of the operational resolution, forecasting errors (from
both load and VRE resources) and a broader set of transmission topologies and con-
straints (including a larger operating footprint with interregional transmission).

Furthermore, our work focuses almost entirely on flexibility options deployable on


the supply side of the system and notably does not consider a meaningful role of de-
mand flexibility beyond the very small amount of shiftable demand deployed in the
scenarios. Demand response potentially has lower implementation costs than elec-
tricity storage and, therefore, is potentially more cost-effective, particularly as the
cost of communication technologies decreases.38 A challenge of evaluating demand
response is the temporal nature of this resource. Much of the literature has focused
on the ability of demand response to reduce demand peaks, and the potential value
of this application.39,40 Less work has examined the availability of demand response
during periods of highest solar output, which often has limited coincidence with
space heating and cooling.41 As a result, a direct comparison of demand with other
flexibility options will require a more thorough analysis of the coincidence of specific

Joule 5, 1143–1167, May 19, 2021 1153


ll
Article

37%
l
15

24%
VRE curtailment (%)

10 l l l

11%
l
l

l
l

ll
l

5 10%
l

ll
l

8%
l
l
l
l
6%
0 l

20 25 30 35 40 45
PV penetration (%)

Base No Storage Reserves


No VRE Reserves No VRE or Storage Reserves

Figure 5. Comparison of total annual PV penetration and curtailment for all sensitivities related
to operating reserve provision by VRE and storage
The percentages indicate the share of storage capacity installed (as a percentage of peak load) for
each PV penetration level.

shiftable loads with periods of high solar output. This includes the potential for
controlled electric vehicle charging, which might provide a more consistent source
of flexible loads.42

Key finding 2: Allowing PV and storage to provide operating reserves results


in system-wide operating costs and curtailment reductions, but curtailed PV
has limited revenue potential from operating reserve markets
Our second key finding relates to (1) the system-wide value of VRE and storage
providing operating reserves and (2) the revenue potential for curtailed PV providing
operating reserves, as reflected by operating reserve prices. We find that allowing
VRE and storage to provide operating reserves at high PV penetration levels results
in system-wide operating cost and curtailment reductions. However, this paradoxi-
cally also leads to low prices for operating reserves, which reduces incentives for PV
to provide operating reserves with curtailed energy. This aspect of the solar curtail-
ment paradox reveals the importance of proper alignment of grid system value and
compensation.

First, allowing wind, utility-scale PV, and storage to provide operating reserves has
system-wide operating cost and curtailment benefits. Figure 5 provides a compari-
son of VRE curtailment versus PV penetration level across all VRE build-out levels for
the baseline and operating reserve eligibility scenarios, with a layout similar to that
previously shown in Figure 2. Although the No VRE Reserves and No Storage Re-
serves scenarios each track closely with the Base scenario across most PV penetra-
tion levels, the combined No VRE or Storage Reserves scenario has significantly
higher curtailment levels across mid and high VRE levels.

To better understand the drivers behind the stark difference between the Base and
No VRE or Storage scenario, Figure 6 shows a summary of available committed

1154 Joule 5, 1143–1167, May 19, 2021


ll
Article

Figure 6. Timeseries of total committed thermal capacity (dark lines) and total generation (light lines) by fuel type for the 44% PV penetration level
The left column shows results from the Base case, and the right column shows results from the No VRE or Storage Reserves sensitivity. The text in the
upper left of each panel is the total annual generation (GWh) for that technology type.

thermal capacity (dark lines) and resulting generation (light lines) in these two cases
at the 44% PV penetration level. Provision of regulation and spinning contingency
reserves from thermal generators requires them to be online and operating at the
partial output. As shown on the left, when VRE and storage can provide reserves
(i.e., Base scenario), the thermal capacity that is otherwise committed only to
meet operating reserve requirements is no longer needed, resulting in large por-
tions of the year where Gas-CC, Gas-CT, and biopower are not committed. This in
turn reduces generation across these technologies by about half. The avoided
commitment and generation of these units also enables greater utilization of
lower-cost VRE and storage resources for not only operating reserves but also en-
ergy, thereby reducing curtailment as shown by the difference between the Base
and No VRE or Storage Reserves scenarios in Figure 5. At this 44% PV penetration
level, allowing neither VRE nor storage to provide reserves (i.e., No VRE or Storage
Reserves scenario) results in 53% more curtailment in relation to the Base case.

Because these thermal generators, especially biopower, have high operating costs,
the avoided commitment and generation patterns demonstrated in Figure 6 trans-
late to significant system-wide operational savings when VRE and storage are al-
lowed to provide operating reserves. This can be seen in Figure 7, which compares

Joule 5, 1143–1167, May 19, 2021 1155


ll
Article

21% 26%

500

400

300

200

100

0
30% 33%
Total system costs (million $)

500 Pumped storage


CAES
400 Battery
Hybrid Battery
300 Biopower
Geothermal
200
Oil Gas Steam
100 Gas CT
Gas CC
0 Coal
Nuclear
38% 44%

500

400

300

200

100

0
Base No VRE No No Base No VRE No No
Reserves Storage VRE or Reserves Storage VRE or
Reserves Storage Reserves Storage
Reserves Reserves

Figure 7. Total system operating costs by fuel type for the operating reserve provision
sensitivities by PV penetration level

total operating cost by technology type and PV penetration for the sensitivities
related to operating reserve provision. At low PV penetrations, operating costs of
the system are generally high and the effect of having VRE or storage resources pro-
vide reserves is low. Although total operating costs of the system decline with the
integration of more PV (and wind), the difference in operating costs incurred across
the different operating reserve sensitivities rises sharply as well. This trend reflects
the increasing value of utilizing storage and VRE resources to provide
operating reserves with larger PV penetration levels. For the 44% PV penetration
case highlighted in Figure 6, which corresponds to the bottom right panel in Figure 7,
not allowing VRE and storage to provide operating reserves results in nearly double
the operating cost in relation to the Base scenario.

The second aspect of this key finding relates to the revenue potential for curtailed PV
providing operating reserves, as reflected by operating reserve prices. This is also
the second component of the solar curtailment paradox. Although our results
demonstrate curtailment and system-wide operating cost benefits associated with
allowing storage and VRE to provide reserves, they do not necessarily translate to
increased revenues for the VRE resources in a wholesale market environment. In
particular, the higher curtailment levels associated with increasing penetration of
PV resources lead to lower prices for reserves, especially during times of PV curtail-
ment, thereby deflating potential revenues from PV providing reserves in a whole-
sale market environment. Operating reserve prices in the transition zone (roughly
25%–40% PV penetration levels) are further affected by operating reserve shortage

1156 Joule 5, 1143–1167, May 19, 2021


ll
Article

No VRE or Storage Reserves No Storage Reserves


40 40
Mean reserve price ($ per MW)

30 30
VRE can
provide reserves
20 20

10 10

0 0
21% 26% 30% 33% 38% 44% VR pro 21% 26% 30% 33% 38% 44%
ca

E vid
n

an e
d
Storage can
st ese
or
provide reserves
r
ag ves
e
r

No VRE Reserves Base


40 40
Mean reserve price ($ per MW)

30 30

20 20

10 10

0 0
21% 26% 30% 33% 38% 44% 21% 26% 30% 33% 38% 44%
PV penetration (% of annual generation) PV penetration (% of annual generation)

All hours Hours with no PV curtailment Hours with PV curtailment

Figure 8. Average reserve prices by PV penetration level and operating reserve provision
sensitivity for the example of the spinning reserve product
Average prices are calculated separately for hours with and without PV curtailment. The full set of
operating reserve products is provided in Figure S2.

conditions, which generally only affect operating prices during non-curtailment pe-
riods (see Table S2).

Although we do not simulate a formal wholesale market, we demonstrate the poten-


tial effect of high PV scenarios and operating reserve eligibility rules on
reserve prices and associated revenues that results from the modeled least-cost sys-
tem operations. Figure 8 provides an example using spinning reserve prices from the
PCM runs across the six PV build-out levels (see Figure S2 for the full set of operating
reserve products). Three sets of operating reserve price metrics are shown in each
case: annual average across all hours (dark purple), the annual average for periods
without PV curtailment (teal), and the annual average for periods with PV curtailment
(yellow).

The upper left panel in the figure, which corresponds to the No VRE or Storage Re-
serves scenario, shows the general increase in all three price metrics as a function of
PV penetration. This trend is driven primarily by the need to commit additional ther-
mal resources to meet operating reserve requirements during periods of high
curtailment when the energy is unneeded. As a much smaller secondary effect, pri-
ces during periods without PV curtailment (i.e., the teal bars) are further elevated by
the presence of operating reserve shortage (also referred to as scarcity) events. Dur-
ing these events, operating reserve prices are temporarily increased to

Joule 5, 1143–1167, May 19, 2021 1157


ll
Article

administratively set values ($4,000/MW for this example of spinning reserves) when
insufficient capacity exists to satisfy the operating reserve requirement. Details
about the modeling of operating reserve shortage conditions are provided in
RPM-to-PLEXOS dData. As provided in Table S2, this shortage pricing effect has
the largest absolute effect on PV penetration levels in the transition zone, with the
largest absolute difference in mean operating reserve prices in this No VRE or Stor-
age Reserves scenario at the 30% PV penetration level. Although these shortage
events sometimes have a significant effect on operating reserve prices, we note
that they comprise only a small portion (<1%) of the overall system operating cost.

If we allow VRE to provide reserves (upper right in the figure), operating reserve pri-
ces generally drop because curtailed VRE can provide operating reserve capacity
with zero cost instead of higher-cost thermal resources that were previously needed
in the No VRE or Storage Reserves scenario. This price reduction is particularly prom-
inent during periods of PV curtailment (i.e., yellow bars in Figure 8). In addition, as in
the No VRE or Storage Reserves scenario, operating reserve shortage events play a
much smaller role in elevating prices during non-PV-curtailment periods (see the
supplemental information).

Alternatively, allowing storage to provide reserves (lower left in the figure) dramat-
ically reduces operating reserve prices. Storage built largely to provide capacity and
energy shifting services often has spare capacity to provide reserves, particularly
during periods of overgeneration, as fast-responding storage, such as batteries in
this study, can quickly stop charging and discharge. Because storage has near-
zero cost in these simulations, it becomes the marginal (i.e., price-setting) unit for
more time periods, resulting in lower overall operating reserve prices, especially
during periods of PV curtailment. As in the previous two cases, operating reserve
shortage events also play a small role in elevating prices during non-PV-curtailment
periods, particularly at mid-PV penetration levels (see the supplemental informa-
tion). Furthermore, allowing storage to provide operating reserves results not only
in a reduction in the total MWh of curtailment (Figure 5) but also up to a 36% reduc-
tion in the total number of annual hours of curtailment compared with the case when
neither storage nor VRE can provide operating reserves (see the Table S1). This im-
plies there are fewer periods when PV could use curtailed energy to provide oper-
ating reserves if it were permitted to do so.

Adding VRE to the storage case (lower right in the figure), which is the Base case,
further reduces prices, but with a somewhat limited effect because storage has
already reduced prices to nearly zero during periods of curtailment and has reduced
the number of time periods with curtailment in order for PV to provide operating re-
serves. This indicates that it is important to evaluate the role of storage for the pro-
vision of operating reserves under increased PV deployment.

These results suggest that, given the model formulation used in our simulations and
the assumed large deployment of energy storage, there is little revenue potential for
PV providing operating reserves with curtailed energy. This is particularly true if stor-
age is also providing operating reserves, as storage dramatically reduces operating
reserve prices, driving them to near zero during periods of curtailment. This is a sec-
ond aspect of the solar curtailment paradox: allowing VRE and storage to provide
operating reserves at high PV penetration levels results in low operating reserve pri-
ces, thereby reducing incentives for PV to provide operating reserves with curtailed
energy. In other words, as PV penetration levels increase, these results suggest that
operating reserve markets will not provide a significant revenue source to

1158 Joule 5, 1143–1167, May 19, 2021


ll
Article

compensate for reduced revenues resulting from greater levels of curtailment and
declining energy prices.

Additional disincentives might exist for down-direction reserves in cases where such
additional curtailment impedes the system’s ability to meet certain renewable pro-
duction goals, such as renewable portfolio standards.43 We also note that our sim-
ulations do not account for any additional operating reserve costs on top of the
default opportunity cost calculation in the market-clearing process (e.g., additional
costs for wear and tear for regulation reserves). Future work should incorporate this
factor, which might increase operating reserve prices and thereby affect the revenue
potential trends described here.

Furthermore, the role of shortage pricing and operating reserve eligibility on prices
and curtailment in these results suggests the need for investigation into operating
reserve market rules that support efficient, reliable, and cost-effective operations.
This includes technology eligibility and shortage pricing rules, among other factors.
For example, market structures and policies could allow all VRE resources to provide
operating reserves to ensure lower-cost operations and curtailment; this is currently
not common practice in competitive wholesale markets.43–45 Shortage pricing struc-
tures can be enhanced to be more transparent and better reflect the true reliability
and flexibility value of operating reserves. These enhancements could include
improved operating reserve demand curves (ORDCs) with higher prices as the sys-
tem is tighter on available capacity (as opposed to static or multi-step approaches
currently used by most U.S. market areas), a higher maximum price point on that
curve, and dynamic updating of the curve parameters.46,47

Improvements could also be made to the operating reserve bidding practices, spe-
cifically in how opportunity cost should be calculated. Many operating reserve prod-
ucts use only the energy offers to determine the operating reserve price, but the
actual value of providing that reserve or the missed opportunity for providing energy
might be higher due to complications from the market sequence and out-of-market
VRE production-based payments such as renewable energy credits.45

Additionally, market participation rules for VRE, storage, hybrid, and distributed so-
lar resources will play a significant role in how these devices are scheduled, which
could shape many of the aspects examined in this work, such as minimum generator
levels or operating reserve availability. Related to this, many resources that operate
through bilateral contracts outside of the pool-based market and others that do not
respond to price signals are self-scheduled.48 This means that the operator does not
have access to the physical flexibility of those resources, which can result in a larger
effective minimum generation level and system curtailment, as well as lost profit for
the generator and additional cost for end-use customers.48,49

Finally, competitive wholesale electricity markets might need to explore alternative


compensation mechanisms to capture the full set of benefits from PV, and in turn, the
value of avoiding curtailment. Curtailment occurs when the marginal cost of gener-
ation in a given time step is equal to or less than the variable cost of VRE (zero). In
competitive wholesale electricity markets, this usually means that the marginal
resource is VRE bidding in at zero or a negative value, for example, due to produc-
tion-based subsidies. However, it does not mean that generators with variable costs
greater than zero are not generating. Thermal plants might also bid zero (or nega-
tive) to avoid startup and shutdown costs, and if cleared at these low bid levels,
will take a temporary loss in revenue. As PV deployment increases, there might be

Joule 5, 1143–1167, May 19, 2021 1159


ll
Article

greater hours of prices at zero, increasing the number of hours for such cleared ther-
mal generators where operating costs that exceed revenue. Increased thermal
generator flexibility, including operating at a lower minimum generation level, re-
duces the likelihood of prices at zero and also the amount of generation where these
thermal generator net revenues are negative. This provides some level of incentive
for thermal generators to increase flexibility, and it offers a partial ‘‘self-correcting’’
mechanism in wholesale markets. However, this might not completely reflect the
benefits of avoiding curtailment for two reasons. First, uplift payments are available
to units committed for reliability purposes, even if a lack of generator flexibility re-
quires them to be committed.50 Second, and potentially more important, is the
lack of full monetization of PV benefits in locational marginal prices, including
avoided emissions; this is in addition to the compensation paradox described above
for PV providing operating reserves. When marginal costs are zero (resulting in
curtailment), this implies that curtailed PV has no value. However, this can occur
when thermal plants do not have full incentives to reduce output and provide the op-
portunity to avoid the externalities associated with fossil fuel generation.

The scope of the market dispatch should be deliberate to ensure that overall system
goals are met. These pricing dynamics are particularly relevant as systems evolve to
higher penetration levels of storage, motivated in part by greater deployment of
VRE resources, especially solar.

EXPERIMENTAL PROCEDURES
Resource availability
Lead contact
Further information and requests for resources and materials should be directed to
and will be fulfilled by the Lead Contact, Bethany Frew (bethany.frew@nrel.gov).

Materials availability
This study did not generate new unique materials.

Data and code availability


There are restrictions on the availability of the final PCM dataset and CEM model
code used in this study due to their proprietary nature. Components that are avail-
able in a generic form include the dGen model (https://github.com/NREL/dgen), the
System Advisor Model (SAM) (https://github.com/nrel/sam), and PIDG (https://
github.com/NREL/PIDG). Additionally, the final PCM dataset and CEM model are
further described in this section, and the PCM dataset is similar to that reported in
the LA100 study,32 which leveraged data from the Western Electricity Coordinating
Council’s Transmission Expansion Planning Policy Committee (TEPPC) 2020 data-
base,51 the National Solar Resource Database52 processed through NREL’s SAM
model,53 and hourly wind profiles from NREL’s Wind Integration National Dataset
(WIND) Toolkit.54

Summary of approach
This study uses multiple modeling tools following a sequence common in utility plan-
ning practices, such as integrated resource planning (see Figure S4 for a conceptual
illustration). This process starts with using a CEM to determine the cost-optimal
build-out and is followed by using a PCM to validate system operability. In this study,
we further modify the system representation within the PCM data set to enable a
more representative system. We then make additional adjustments to capture the
operational sensitivity cases.

1160 Joule 5, 1143–1167, May 19, 2021


ll
Article

Using this modeling workflow, we translate multiple systems build-out years from a
highly resolved CEM into a PCM database at a nodal resolution. This allows us to
characterize multiple potential future LADWP systems with very high VRE penetra-
tion levels with a high degree of operational detail, assuming an islanded system
with no power exchanges with adjacent regions. Eighty-four scenarios are evaluated
in total, which collectively captures 6 different PV penetration levels across a Base
scenario and 13 operational sensitivity cases.

The system build-outs in this study are from a CEM, but all operational results presented
are from the PCM. In this way, we were able to leverage the CEM for investment and
resource adequacy decisions and a chronological PCM for operational decisions. Each
of these modeling tools, as well as the translation process from various model databases
and additional adjustments, are described below.

Capacity expansion modeling


Power system planners, policymakers, and other stakeholders commonly use CEMs
to inform investment and retirement decisions of power system assets. CEMs range
in spatial scope from a single utility or region55,56 to national tools.57–59 These
models typically solve for the system-wide least-cost portfolio of generators, trans-
mission, storage, and other resources that are needed to serve load while satisfying
a set of reliability, economic, and policy constraints. Because expansion decisions
are usually made in interim steps across a multidecadal utility planning horizon, sys-
tem operations are often simplified and aggregated to a coarser temporal extent
and resolution.60–62

This study used one primary CEM, the National Renewable Energy Laboratory’s
(NREL’s) Resource Planning Model (RPM), for the bulk power system, supplemented
with input from NREL’s Distributed Generation Market Demand (dGen) model for
distributed PV (DPV) adoption.63

RPM is a CEM with a highly resolved spatial and temporal treatment for a regional
focus area, in this case, the LADWP balancing area. RPM differs from many CEMs
by having a nested spatial resolution that represents a focus region at the nodal-
and line-levels for transmission and at the unit-level for generators, with zonal and
aggregate treatment, respectively, in surrounding regions to capture power trans-
fers into and out of the focus region and between connected regions.64 The tempo-
ral treatment consists of representative weeks in order to model hourly chronology;
RPM currently can represent three representative weeks and one peak day at hourly
resolution.56,65 The hourly chronology and unit-level resolution, along with a mixed-
integer optimization approach that allows generation units of distinct sizes to be
built, all provide RPM with a highly detailed system operation representation.66
Resource adequacy is ensured through a planning reserve margin constraint in the
CEM and by considering the contribution of all resource types using all hours of
the year in the PCM in order to capture events outside of the three representative
weeks used for dispatch decisions. VRE resource potential and availability are also
represented at a high resolution across renewable resource zones, each with similar
output characteristics of resource potential, hourly profiles, and grid interconnection
costs, which are created using clustering techniques.67

RPM does not co-optimize customer adoption of distributed resources. Utility-side


investment decisions in RPM are based on least-cost investment decisions, subject
to meeting load, reliability requirements, transmission constraints, and environ-
mental and policy regulations. Rooftop solar investment decisions are based on a

Joule 5, 1143–1167, May 19, 2021 1161


ll
Article

very different set of criteria, including individual end-user choice, so a second model
(dGen) is needed to estimate possible DPV adoption. The dGen model produces
scenarios of market uptake of DPV considering the following: the electricity con-
sumption projections for various industrial, residential, and commercial customers;
resource and generation capacity potentials in their respective building rooftops;
regional electricity rate structures; and estimated cost savings.68 The dGen model
uses Bass diffusion theory and DPV market adoption data to generate projections.
It provides RPM with exogenously determined DPV projections for consideration
in the broader system build-out decision process.

The RPM model optimizes PV build decisions over two types of resources: stand-
alone PV and PV paired with storage (hybrid systems). Hybrid PV devices are
assumed to be co-located at a common point of grid interconnection. Constraints
are included in the CEM and PCM to ensure that the combined generation of the
PV and storage is less than the rated capacity of the inverter. The hybrid system is
modeled as loosely coupled, meaning that the storage device is able to charge
from either the PV system or the grid.

Production cost modeling


System planners, operators, and other decision-makers use PCMs to evaluate the opera-
tion of power systems, which can range in size from small test systems69 to national- or con-
tinental-scale systems.70 These tools solve for the least-cost unit commitment and
dispatch of a static fleet of generators and network topology, often across a full year of
chronologic operations at hourly or sub-hourly resolution. Operational decisions minimize
the system-wide operating (or production) costs, which can include variable operations
and maintenance, fuel, emissions, and startup/shutdown costs. Various commercially
available PCMs are used by different utility planners.

We used the commercial-grade PCM software PLEXOS to model a combination of


hourly and 5-min system operations.71 Capacity build-out decisions from RPM for
the six PV penetration levels were translated into a PLEXOS database to allow for
unit commitment and economic dispatch simulations in the PCM step. The PCM
uses a nodal resolution with 109 nodes across the LADWP study area and DC power
flow and transmission flow limits between those nodes. A map of the regional rep-
resentation is provided in Figure S5.

Mixed-integer programming simulations were performed for a single year at an


hourly resolution, except for the sensitivity case with a 5-min resolution and the
sensitivity case with hourly day-ahead (DA) and 5-min RT resolution. For the latter
(DA-RT) scenario, forecasted load and VRE data were used to determine the DA
unit commitment decisions, which were then passed down to a 5-min RT simulation
that dispatches any fast-starting units needed to balance RT load with any realized
forecast errors. For all runs, scheduling was done assuming 24 h of additional, per-
fect foresight, which was at a 2-h resolution for hourly runs and 30-min resolution for
5-min runs. To improve run time, each month of the year was run in parallel, and the
12 monthly runs were then combined to provide annual results, where an additional
24-h look-ahead informs but does not extend the solution beyond the last time step
in each month. For a discussion of temporal decomposition of the unit commitment
problem, see Barrows et al.72

RPM-to-PLEXOS data
RPM and PLEXOS share many input data values, which are passed directly to
PLEXOS in the translation process. The values used in RPM are derived from the

1162 Joule 5, 1143–1167, May 19, 2021


ll
Article

Western Electricity Coordinating Council’s TEPPC 2020 database,51 with retirement


adjustments made based on input from LADWP. Additional proprietary project data
from the LADWP system were added to improve the representation of the study
area, including updated heat rates and transmission representation, and to ensure
all planned builds were included. Timeseries data include hourly load (generated
as part of the LA100 study32), hourly PV generation profiles from the National Solar
Resource Database52 processed through NREL’s SAM model,53 and hourly wind
profiles from NREL’s WIND Toolkit,54 all for 2012 meteorological conditions. The
2012 year was chosen because it is the most representative of high-resolution
wind, solar, and load data.73

All available generator parameters from RPM are passed to PLEXOS during the RPM-
to-PLEXOS translation step. However, some generator parameters that are required
in PLEXOS are not used by RPM. To properly characterize the generators in PLEXOS,
these critical values were derived from the TEPPC database. They include mean time
to repair as well as average plant sizes used to disaggregate RPM builds in situations
when the RPM solution included large builds in a single solve year. To avoid model
degeneracy, several generator properties (e.g., heat rate, start cost, and variable
operation and maintenance costs) are randomly adjusted by less than 5% during
the RPM-to-PLEXOS translation process. All fuel prices are based on U.S. Energy In-
formation Administration Annual Energy Outlook regional data.74 The final values
are seen by PLEXOS for generators (in the 33% PV penetration case as an example)
and fuel costs are summarized in Tables S3 and S4.

Operating reserves were enforced for the entire footprint at a regional level, which is
similar to current utility practices. We enforce six reserve products: spinning contin-
gency reserves, non-spinning contingency reserves, regulating up reserves, regu-
lating down reserves, flexible ramping down reserves, and flexible ramping up the
reserve. Spinning and non-spinning reserves are used to address large generator
or transmission line failures, while regulation and flexibility reserves are used to
address both normal and unscheduled variability in net demand over sub-hourly
timescales. The operating reserve requirements enforced by PLEXOS in each time
step are based on the methodology from Phase 2 of NREL’s Western Wind and Solar
Integration Study75; all reserve calculation approaches are summarized in Table 3,
and the resulting values are provided in Table S5. Each of these types of reserves
can be held by partially loaded generators with sufficient ramping capability to
respond in a given time frame, including energy storage and curtailed VRE, except
where otherwise noted. Generators must be online in order to meet all operating re-
serves, except for non-spinning and flexibility in the up direction, which can be met by
either online or offline units. For all scenarios, generators located on the far ends of
the radially connected DC lines of the system, such that they could not reliably supply
load in the event of a contingency, were not eligible to provide contingency oper-
ating reserves (i.e., spinning and non-spinning reserves). This included a small subset
of wind, utility PV, battery storage, coal, Gas-CC and CT, PSH, and CAES resources.

Operating reserve shortage prices, which are administratively set in real markets and
are called the value of reserve shortage (VoRS) in PLEXOS, are applied any time an
operating reserve constraint is violated. Because energy and operating reserves are
co-optimized, both prices can be influenced by the operating reserve shortage pric-
ing in a shortage event, when the price rises significantly above the variable cost of
the marginal unit. In this study, VoRS is set to $3,900/MW for flexibility up-and-down
reserves, $4,000/MW for spinning and non-spinning reserves, and $4,100/MW for
regulation up-and-down reserves.

Joule 5, 1143–1167, May 19, 2021 1163


ll
Article

Table 3. Operating reserve calculations and characterizations


Reserve Calculation Time frame
Contingency (spinning and Variable across all time-steps and equal to the 10 min
non-spinning) largest of 738 MW or the flows across two large
DC transmission lines; only 50% is required
from online (spinning) generators; the other
50% is from non-spinning generators.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Regulation Up/Down 5 min
ð1% of loadÞ2 + ðWind ReqtÞ2 + ðPV ReqtÞ2
where (Wind Reqt) and (PV Reqt) will cover
95% of 10-min persistence forecast errors of
each technology
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Flexibility Up/Down 20 min
ðWind ReqtÞ2 + ðPV ReqtÞ2
where (Wind Reqt) and (PV Reqt) will cover
70% of 60-min persistence forecast errors of
each technology

Additional adjustments
After the RPM-to-PLEXOS linkage was complete, additional modifications were
made to enable a more representative system and to capture the operational sensi-
tivity cases in Table 1. To make the system more representative, we replaced 90% of
the PSH capacity with natural gas-CT generators, replaced some PV-hybrid systems
with stand-alone PV, and added additional oil-gas-steam (OGS) and gas-CT units
when needed to avoid unserved energy resulting from the previous adjustments.
NREL’s Probabilistic Resource Adequacy Suite (PRAS),76 which is a collection of tools
developed by NREL to study the unserved energy risk in electric power systems, was
used to help with these modifications to achieve a roughly uniform loss of load
expectation (LOLE) levels across all build-out levels. We note that reducing the
PSH to levels more reflective of typical systems also removed inherent system flex-
ibility; not making this adjustment would likely have diminished the effect of the flex-
ibility sensitivities.

SUPPLEMENTAL INFORMATION
Supplemental Information can be found online at https://doi.org/10.1016/j.joule.
2021.03.021.

ACKNOWLEDGMENTS
The authors thank Mike Meshek (NREL) for careful editing and Daniel Steinberg
and Daniel Bilello (NREL) for feedback on the draft of this paper. The authors espe-
cially thank the Los Angeles Department of Water and Power (LADWP) for the use
of their data, as well as Matthew Irish, Brady Cowiestoll, and Jennie Jorgenson
(NREL) for assistance on the various modeling components of this work. This
work was authored by the National Renewable Energy Laboratory, operated by
Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) un-
der contract number DE-AC36-08GO28308. Funding provided by U.S. Depart-
ment of Energy Office of Energy Efficiency and Renewable Energy Solar Energy
Technologies Office. A portion of this research was performed by using computa-
tional resources sponsored by the Department of Energy’s Office of
Energy Efficiency and Renewable Energy and located at the National Renewable
Energy Laboratory. The views expressed in the article do not necessarily represent
the views of the DOE or the U.S. government. The U.S. government retains and the
publisher, by accepting the article for publication, acknowledges that the U.S. gov-
ernment retains a nonexclusive, paid-up, irrevocable, worldwide license to publish
or reproduce the published form of this work or allow others to do so, for U.S. gov-
ernment purposes.

1164 Joule 5, 1143–1167, May 19, 2021


ll
Article

AUTHOR CONTRIBUTIONS
B.F., B.S., and N.G. developed and ran the model and produced the figures. D.L.
developed the initial scripts for producing the figures. B.F., B.S., P.D., W.C., and
R.M. selected the scenarios, analyzed the results, and wrote the paper.

DECLARATION OF INTERESTS
The authors declare no competing interests.

Received: December 24, 2020


Revised: March 5, 2021
Accepted: March 29, 2021
Published: April 28, 2021

REFERENCES
1. Jenkins, J.D., Luke, M., and Thernstrom, S. renewable integration. Energy Policy 115, 21. Wang, Y., Lou, S., Wu, Y., and Wang, S. (2020).
(2018). Getting to zero carbon emissions in the 249–257. Flexible operation of retrofitted coal-fired
electric power sector. Joule 2, 2498–2510. power plants to reduce wind curtailment
12. Bistline, J.E. (2018). Turn down for what? The considering thermal energy storage. IEEE
2. Sepulveda, N.A., Jenkins, J.D., de Sisternes, economic value of operational flexibility in Trans. Power Syst. 35, 1178–1187.
F.J., and Lester, R.K. (2018). The role of firm electricity markets. IEEE Trans. Power Syst. 34,
low-carbon electricity resources in deep 527–534. 22. de Boer, H.S., Grond, L., Moll, H., and Benders,
decarbonization of power generation. Joule 2, R. (2014). The application of power-to-gas,
2403–2420. 13. Jorgenson, J., Mai, T., and Brinkman, G. (2017). pumped hydro storage and compressed air
Reducing wind curtailment through energy storage in an electricity system at
3. Sun, Y., Wachche, S.V., Mills, A., and Ma, O. transmission expansion in a wind vision future, different wind power penetration levels.
(2020). 2018 Renewable Energy Grid (National Renewable Energy Laboratory). Energy 72, 360–370.
Integration Data Book (National Renewable https://www.nrel.gov/docs/fy17osti/67240.
Energy Laboratory). pdf. 23. Mileva, A., Johnston, J., Nelson, J.H., and
Kammen, D.M. (2016). Power system balancing
4. Frew, B., Cole, W., Denholm, P., Frazier, A.W., 14. O’Shaughnessy, E., Cruce, J.R., and Xu, K. for deep decarbonization of the electricity
Vincent, N., and Margolis, R. (2019). Sunny with (2020). Too much of a good thing? Global sector. Appl. Energy 162, 1001–1009.
a chance of curtailment: operating the US grid trends in the curtailment of solar PV. Sol.
with very high levels of solar photovoltaics. Energy 208, 1068–1077. 24. de Sisternes, F.J., Jenkins, J.D., and Botterud,
iScience 21, 436–447. A. (2016). The value of energy storage in
15. E3. (2014) Investigating a Higher Renewables decarbonizing the electricity sector. Appl.
5. Bird, L., Lew, D., Milligan, M., Carlini, E.M., Portfolio Standard in California (Energy and Energy 175, 368–379.
Estanqueiro, A., Flynn, D., Gomez-Lazaro, E., Environmental Economics). https://www.
Holttinen, H., Menemenlis, N., Orths, A., et al. ethree.com/wp-content/uploads/2017/01/ 25. Palmintier, B.S., and Webster, M.D. (2016).
(2016). Wind and solar energy curtailment: a E3_Final_RPS_Report_2014_01_06_with_ Impact of operational flexibility on electricity
review of international experience. Renew. appendices.pdf generation planning with renewable and
Sustain. Energy Rev. 65, 577–586. carbon targets. IEEE Trans. Sustain. Energy 7,
16. Nelson, J., Kasina, S., Stevens, J., Moore, J., 672–684.
6. Frew, B.A., Becker, S., Dvorak, M.J., Andresen, Olson, A., Morjaria, M., Smolenski, J., and
G.B., and Jacobson, M.Z. (2016). Flexibility Aponte, J. (2018). Investigating the economic 26. Helistö, N., Kiviluoma, J., Holttinen, H., Lara,
mechanisms and pathways to a highly value of flexible solar power plant operation, J.D., and Hodge, B.-M. (2019). Including
renewable US electricity future. Energy 101, (Energy + Environmental Economics). https:// operational aspects in the planning of power
65–78. www.ethree.com/wp-content/uploads/2018/ systems with large amounts of variable
10/Investigating-the-Economic-Value-of- generation: a review of modeling approaches.
7. Hitaj, C. (2015). Location matters: the impact of Flexible-Solar-Power-Plant-Operation.pdf. WIREs Energy Environ 8, e341.
renewable power on transmission congestion
and emissions. Energy Policy 86, 1–16. 17. California, I.S.O. (2019). Managing oversupply, 27. Nicolosi, M. (2010). The importance of high
(California Independent System Operator). temporal resolution in modeling renewable
8. Golden, R., and Paulos, B. (2015). Curtailment http://www.caiso.com/informed/Pages/ energy penetration scenarios, Lawrence
of renewable energy in California and beyond. ManagingOversupply.aspx. Berkeley National Laboratory, LBNL-4197E.
Electr. J. 28, 36–50. https://escholarship.org/uc/item/9rh9v9t4.
18. APS. (2017). 2017 integrated resource plan.
9. Bird, L., Cochran, J., and Wang, X. (2014). Wind https://www.aps.com/-/media/APS/APSCOM- 28. Cole, W., Corcoran, S., Gates, N., Mai, T.,D.,
and solar energy curtailment: experience and PDFs/About/Our-Company/Doing-business- and Das, P. (2020). 2020 standard scenarios
practices in the United States, (National with-us/Resource-Planning-and-Management/ report: a U.S. electricity sector outlook,
Renewable Energy Laboratory). https://www. 2017IntegratedResourcePlan.ashx. (National Renewable Energy Laboratory).
nrel.gov/docs/fy15osti/63054.pdf. https://www.energy.gov/eere/analysis/
19. Palchak, D., and Denholm, P. (2014). Impact of downloads/2020-standard-scenarios-report-
10. Lew, D., Bird, L., Milligan, M., Speer, B., Wang, generator flexibility on electric system costs us-electricity-sector-outlook.
X., Carlini, E.M., Estanqueiro, A., Damian, F., and integration of renewable energy, (National
Emilio Gomez-Lazaro, N.M., et al. (2013). Wind Renewable Energy Laboratory). https://www. 29. EIA (2020). Annual energy outlook 2020, (U.S.
and solar curtailment. Preprint. In (NREL CP- nrel.gov/docs/fy14osti/62275.pdf. DOE Energy Information Administration).
5500-60245). https://www.eia.gov/outlooks/ieo/.
20. Schill, W.-P., Pahle, M., and Gambardella, C.
11. Denholm, P., Brinkman, G., and Mai, T. (2018). (2017). Start-up costs of thermal power plants in 30. International Energy Agency (2020). World
How low can you go? The importance of markets with increasing shares of variable energy outlook 2020. https://www.iea.org/
quantifying minimum generation levels for renewable generation. Nat. Energy 2, 17050. reports/world-energy-outlook-2020.

Joule 5, 1143–1167, May 19, 2021 1165


ll
Article

31. Wiser, R.H., Bolinger, M., Gorman, W., Rand, J., https://www.nrel.gov/docs/fy19osti/72578. deployment system (ReEDS) model
Jeong, S., Seel, J., Warner, C., and Paulos, B. pdf. documentation: version 2016, (National
(2020). Hybrid power plants: status of installed Renewable Energy Laboratory). https://www.
and proposed projects. https://emp.lbl.gov/ 44. Chernyakhovskiy, I., Koebrich, S., Gevorgian, nrel.gov/docs/fy17osti/67067.pdf.
publications/hybrid-power-plants-status- V., and Cochran, J. (2019). Gird-friendly
installed. renewable energy: solar and wind participation 58. Energy Information Administration (EIA) (2020).
in automatic generation control systems, The electricity market module of the national
32. National Renewable Energy Laboratory (2021). (National Renewable Energy Laboratory). energy modeling system: model
In LA100: the los angeles 100% renewable https://www.nrel.gov/docs/fy19osti/73866. documentation 2020. https://www.eia.gov/
energy study, J. Cochran and P. Denholm, eds. pdf. outlooks/aeo/nems/documentation/
(National Renewable Energy Laboratory), electricity/pdf/m068(2020).pdf
Report number: NREL/TP-6A20-79444. https:// 45. EPRI (2019). Ancillary Services in the United
maps.nrel.gov/la100/report. States: Technical Requirements, Market 59. Electric Power Research Institute (EPRI) (2020)
Designs and Price Trends,. https://www. US-REGEN Model Documentation (EPRI).
33. Venkataraman, S., Jordan, G., O’Connor, M., offshorewindadvisory.com/wp-content/ https://www.epri.com/research/products/
Kumar, N., Lefton, S., Lew, D., Brinkman, G., uploads/2019/07/EPRI-Ancillary-Services.pdf. 000000003002016601
Palchak, D., and Cochran, J. (2013). Cost-
benefit analysis of flexibility retrofits for coal 46. Nicholson, E. (2019). Procuring flexibility in 60. Cole, W., Frew, B., Mai, T., Sun, Y., Bistline, J.,
and gas-fueled power plants: August 2012– wholesale electricity markets. Curr. and Blanford, G. (2017). Variable renewable
December 2013, (National Renewable Energy Sustainable. Renewable Energy Rep. 6, energy in long-term planning models: a multi-
Laboratory). https://www.nrel.gov/docs/ 100–106. model perspective. https://www.nrel.gov/
fy14osti/60862.pdf. docs/fy18osti/70528.pdf.
47. P.J.M.. (2018). Energy price formation senior
34. Kumar, N., Besuner, P., Lefton, S., Agan, D., task force. https://www.pjm.com/committees- 61. Blanford, G., Bistline, J., Young, D., and
and Hilleman, D. (2012). Power plant cycling and-groups/task-forces/epfstf.aspx. Merrick, J. (2016). Simulating annual variation in
costs, (National Renewable Energy load, wind, and solar by representative hour
Laboratory). https://www.nrel.gov/docs/ 48. Ela, E., Milligan, M., Bloom, A., Botterud, A., selection (EPRI). http://www.iaee.org/en/
fy12osti/55433.pdf. Townsend, A., Levin, T., and Frew, B.A. (2016). publications/ejarticle.aspx?id=3083.
Wholesale electricity market design with
35. Denholm, P., and Margolis, R. (2016). Energy increasing levels of renewable generation: 62. Pfenninger, S. (2017). Dealing with multiple
storage requirements for achieving 50% incentivizing flexibility in system operations. decades of hourly wind and PV time series in
penetration of solar photovoltaic energy in Electr. J. 29, 51–60. energy models: A comparison of methods to
California, (National Renewable Energy reduce time resolution and the planning
Laboratory). https://www.nrel.gov/docs/ 49. Daniel, J., Sattler, S., Massie, A., and Jacobs, M. implications of inter-annual variability. Appl.
fy16osti/66595.pdf? (2020). Used, but how useful? How electric Energy 197, 1–13.
gathStatIcon=true#::text=With%20this% utilities exploit loopholes, forcing customers to
20highly%20flexible%20system,to%20be% bail out uneconomic coal-fired power plants, 63. Cole, W., Frew, B., Gagnon, P., Reimers, A.,
20built%20by%202030. (Union of Concerned Scientists). https://www. Zuboy, J., and Margolis, R. (2018). Envisioning a
ucsusa.org/resources/used-how-useful. low-cost solar future: exploring the potential
36. Denholm, P., and Mai, T. (2019). Timescales of impact of achieving the. SunShot 2030 targets
energy storage needed for reducing 50. Saucer, W. (2014) Staff Analysis of Uplift in RTO for photovoltaics. Energy 155, 690–704.
renewable energy curtailment. Renew. Energy and ISO Markets (Federal Energy Regulatory
130, 388–399. Commission). http://www.ferc.gov/legal/staff- 64. Hale, E., Stoll, B., and Mai, T. (2016). Capturing
reports/2014/08-13-14-uplift.pdf the impact of storage and other flexible
37. Mallapragada, D.S., Sepulveda, N.A., and technologies on electric system planning,
Jenkins, J.D. (2020). Long-run system value of 51. WECC (Western Electricity Coordinating (National Renewable Energy Laboratory).
battery energy storage in future grids with Council) (2011). Study report: 10-year regional https://www.nrel.gov/docs/fy16osti/65726.
increasing wind and solar generation. Appl. transmission plan (TEPPC 2010 study program), pdf.
Energy 275, 115390. 2020. https://www.wecc.org/Reliability/
2011Plan_2020%20Study%20Report.pdf. 65. Getman, D., Lopez, A., Mai, T., and Dyson, M.
38. Gerke, B., Gallo, G., Smith, S., Liu, J., Alstone, (2015). Methodology for clustering high-
P., Raghavan, S., Schwartz, P., Piette, M.A., Yin, 52. SenGupta, M., Xie, Y., Lopez, A., Habte, A., resolution spatiotemporal solar resource data,
R., and Stensson, S. (2020). The California Maclaurin, G., and Shelby, J. (2018). The (National Renewable Energy Laboratory).
demand response potential study, phase 3: national solar radiation data base (NSRDB). https://www.nrel.gov/docs/fy15osti/63148.
final report on the shift resource through 2030. Renew. Sustain. Energy Rev. 89, 51–60. pdf.
https://emp.lbl.gov/publications/california-
demand-response-potential. 53. Freeman, J.M., DiOrio, N.A., Blair, N.J., Neises, 66. Mai, T., Drury, E., Eurek, K., Bodington, N.,
T.W., Wagner, M.J., Gilman, P., and Janzou, S. Lopez, A., and Perry, A. (2013). Resource
39. Hledik, R., Faruqui, A., Lee, T., and Higham, J. (2018). System advisor model (SAM) general planning model: an integrated resource
(2019). The national potential for load flexibility: description (version 2017.9.5), (National planning and dispatch tool for regional electric
value and market potential through 2030. Renewable Energy Lab.). https://doi.org/10. systems, (National Renewable Energy
https://brattlefiles.blob.core.windows.net/ 2172/1440404. Laboratory). https://www.nrel.gov/docs/
files/16639_national_potential_for_load_ fy13osti/56723.pdf.
flexibility_-_final.pdf. 54. Draxl, C., Clifton, A., Hodge, B.-M., and
McCaa, J. (2015). The wind integration national 67. Hale, E.T. (2018). NREL’s resource planning
40. Jordehi, A.R. (2019). Optimisation of demand dataset (WIND) toolkit. Appl. Energy 151, model, (National Renewable Energy
response in electric power systems, a review. 355–366. Laboratory). https://www.nrel.gov/analysis/
Renew. Sustain. Energy Rev. 103, 308–319. models-rpm.html.
55. WECC (2013). Interconnection-wide plan tools
41. Müller, T., and Möst, D. (2018). Demand and models 2013. https://www.wecc.biz/ 68. Sigrin, B., Gleason, M., Preus, R., Baring-Gould,
response potential: available when needed? Reliability/2013Plan_ToolsandModels.docx I., and Margolis, R. (2016). The distributed
Energy Policy 115, 181–198. generation market demand model (dGen):
56. Mai, T., Barrows, C., Lopez, A., Hale, E., Dyson, documentation, (National Renewable Energy
42. Denholm, P., Kuss, M., and Margolis, R.M. M., and Eurek, K. (2015). Implications of model Laboratory). https://www.nrel.gov/analysis/
(2013). Co-benefits of large scale plug-in hybrid structure and detail for utility planning: dgen/.
electric vehicle and solar PV deployment. scenario case studies using the resource
J. Power Sources 236, 350–356. planning model, (National Renewable Energy 69. RTS (2018). GMLC: reliability test system - grid
Laboratory). https://doi.org/10.2172/1215187. modernization lab consortium. https://github.
43. Denholm, P.L., Sun, Y., and Mai, T.T. (2019). An com/GridMod/RTS-GMLC.
introduction to grid services: concepts, 57. Eurek, K., Cole, W., Bielen, D.A., Blair, N.,
technical requirements, and provision from Cohen, S., Frew, B., Ho, J., Krishnan, V., Mai, T., 70. Martinek, J., Jorgenson, J., Mehos, M., and
wind, (National Renewable Energy Laboratory). and Steinberg, D. (2016). Regional energy Denholm, P. (2018). A comparison of price-

1166 Joule 5, 1143–1167, May 19, 2021


ll
Article

taker and production cost models for 73. Bloom, A., Novacheck, J., Brinkman, G., 75. Lew, D., Brinkman, G., Ibanez, E., Hodge,
determining system value, revenue, and McCalley, J., Figueroa-Acevedo, A.L., B.M., Hummon, M., Florita, A., and Heaney,
scheduling of concentrating solar power Jahanbani-Ardakani, A., Nosair, H., M. (2013). The western wind and solar
plants. Appl. Energy 231, 854–865. Venkatraman, A., Caspary, J., Osborn, D., and integration study phase 2, (National
Lau, J. (2020). The value of increased HVDC Renewable Energy Laboratory). https://
71. Energy Exemplar Simulation. (2021). PlEXOS capacity between eastern and western U.S. figliodellafantasia.files.wordpress.com/2013/
Software Energy Exemplar. Energy Exemplar. grids: the interconnections seam study, 10/rapporto_nrel.pdf.
https://energyexemplar.com/products/ (National Renewable Energy Laboratory).
plexos-simulation-software/. https://www.nrel.gov/docs/fy21osti/76850.
pdf. 76. Stephen, G.; USDOE Office of Energy
Efficiency and Renewable Energy (2019).
72. Barrows, C., Hummon, M., Jones, W., and Hale, Probabilistic resource adequacy suite
E. (2014). Time domain partitioning of electricity 74. EIA (2019). Annual energy outlook 2019
(PRAS), National Renewable Energy Lab.
production cost simulations, (National (U.S. DOE Energy Information
(NREL). https://doi.org/10.11578/dc.
Renewable Energy Laboratory). https://www. Administration). https://www.eia.gov/
20190814.1.
nrel.gov/docs/fy14osti/60969.pdf. outlooks/archive/aeo19/pdf/aeo2019.pdf.

Joule 5, 1143–1167, May 19, 2021 1167

You might also like