Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

The Electricity Journal xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

The Electricity Journal


journal homepage: www.elsevier.com/locate/tej

Energy Policy Institute's Seventh Annual Energy Policy Research Conference

The value of demand response in Florida



Brady Stolla, , Elizabeth Buechlerb, Elaine Halea
a
National Renewable Energy Laboratory, United States
b
Stanford University, United States

A R T I C L E I N F O A B S T R A C T

Keywords: Many electrical loads may be operated flexibly to provide grid services, including peaking capacity, reserves, and
Demand response load shifting. The authors model 14 demand end uses in Florida and analyze their operational impacts and
Power systems overall value for a wide range of solar penetrations and grid flexibility options. They find demand response is
Solar photovoltaics able to reduce production costs, reduce the number of low-load hours for traditional generators, reduce starting
Renewable integration
of gas generators, and reduce curtailment.

1. Introduction and peak load management services that demand response already
provides (Lee et al., 2015).
Demand response (DR) is a broad descriptor for any electric utility In this article we analyze the impact on grid operations of 14 dif-
or aggregator program that incentivizes or requires loads to reduce or ferent load types that could provide demand response in Florida. We
otherwise modify their energy consumption in support of grid opera- include two demand response penetrations, one approximately
tions. These types of programs have been in use for many years in the equivalent to 2015 participation in demand response programs, and
United States in the industrial sector, where customers are contracted to one assuming a high participation rate. We analyze the changing im-
reduce their demand during emergency periods. More recently, demand pacts of demand response as PV penetration increases from 5% to 45%,
response programs have been used to shift residential and commercial with a focus on the value of demand response to the system at higher
loads away from peak periods. As renewable generation sources in- penetrations in terms of system operation impacts.
crease, the ability of demand response to help mitigate challenges with
increased net load variability and uncertainty has become of higher 2. Methods
interest to utilities due to its ability to provide flexibility to the electric
grid. Additionally, rapidly developing communication and control We analyzed the grid operations of the Florida Reliability
technologies have made demand response more flexible and easier to Coordinating Council (FRCC) using the commercial software package
implement. PLEXOS (Energy Exemplar, 2014), and the FRCC model created for
In this article we study the potential impact of demand response in (Denholm et al., 2016). In this model, we include all operational con-
Florida assuming both low and high levels of solar photovoltaic (PV) straints on generators and enforce line limits on all lines above 200 kV.
deployment. Florida currently has a high level of demand response and It includes a connection to the only reliability region in the eastern
little PV generation, but the state is poised to see high growth in solar interconnection that is electrically connected to FRCC, the SERC Re-
energy in coming years due to its high solar resource potential and the liability Corporation in Georgia, which is modeled as a single load and
falling costs of solar installations (Gagnon et al., 2016; Lopez et al., supply curve of generation. We include additional utility scale photo-
2012; NREL, 2016; Lazard, 2016). Challenges with the “duck curve” voltaic generators to enable penetrations up to 45%. For more in-
that are starting to be seen in California, including low net loads during formation on the PLEXOS model, see (Hale et al., 2017; Denholm et al.,
the day and a steep ramp in net load during the evening, are likely to be 2016).
seen in Florida as well in the 5- to 15-year timeframe (Hale et al., 2017). We incorporate demand response resources for 14 different load end
Demand response can help mitigate these issues by shifting loads from uses into the model of FRCC, dispersing the demand response into the
high net load periods to low net load periods, providing reserves to two nodes with the highest load per region. We use hourly estimates of
meet increased variability from solar, and by providing capacity for the total potential demand response resource from (Olsen et al., 2013),
long-term system planning. This is in addition to the valuable capacity which uses several filters to calculate the fraction of a load that is


Corresponding author at: 15013 Denver West Pkwy, MS RSF300, Golden, CO 80401.
E-mail address: brady.stoll@nrel.gov (B. Stoll).

https://doi.org/10.1016/j.tej.2017.10.004

1040-6190/ © 2017 Elsevier Inc. All rights reserved.


Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted manuscript.
The published version of the article is available from the relevant publisher.
Please cite this article as: Stoll, B., The Electricity Journal (2017), http://dx.doi.org/10.1016/j.tej.2017.10.004
B. Stoll et al. The Electricity Journal xxx (xxxx) xxx–xxx

Table 1
Demand response end-use constraints by sector, including grid services that can be pro-
vided.

Sector End-Use Grid Load Balancing Daily


Services Recovery Freq. Duration
Restrictions (days) Restriction
(h)

Residential Cooling R, C, E 5 am–6 pm 1 1


Heating R, C, E 3 am–7 pm 1 1
Water heating R, C, E – 1 –
Commercial Cooling R, C, E 5 am–6 pm 1 2
Heating R, C, E 3 am–7 pm 1 2
Ventilation R, C – – –
Lighting R, C – – –
Municipal Wastewater E – 1 3
pumping
Water E – 1 2
pumping Fig. 1. Scenario framework for the analysis. Each scenario consists of one PV level, one
Outdoor R, C – – – DR option and one flexibility option.
Lighting
Industrial Datacenters C, E 4 am–8 pm 1 4
Manufacturing C, E – 1 – systems, any resource that can potentially shift generation from one
Refrigerated E – 1 4 time period to another in rhythm with the diurnal cycles of the sun and
warehouses
Agricultural C, E – 7 8
electricity demand becomes more valuable and can be used to perform
pumping this service on days of significant curtailment. Additionally, resources
that can quickly increase generation or decrease loads during periods of
rapidly changing net-load, such as during the afternoon when solar
controllable, sheddable, and acceptable to consumers. In this and our output decreases and total demand is increasing, can reduce stress on
companion work we adjust these filters to better represent the current the grid and limit the use of high-cost generators during these times.
status of demand response (Low DR scenario) and a more realistic fu- These roles, and a description of how the presence of battery storage
ture scenario with high DR participation (High DR scenario) (Hale influences DR operations and vice versa, are covered in the second set
et al., 2017). These data provide an estimate of the total load a parti- of results.
cular end use can change in response to a grid signal, including po-
tential contributions to energy shifting and reserves provision. Each end
use is subject to different constraints based on assumptions about how 3.1. Overall impacts of DR on FRCC
much that end use could reasonably be used. These constraints, listed in
Table 1, are expanded from similar work done in (Hummon et al., 3.1.1. Production costs
2013). The total production cost of serving a region’s electricity demand
These constraints consist of requirements on allowable time ranges consists of all direct costs incurred through generation, including the
for shifting energy, including when load shifting can be recovered based fuel costs, variable operation and maintenance costs (VO&M), and start
on building occupancy and comfort levels, restrictions on the amount of and shutdown costs. This represents the direct costs of operating the
time a single load may be shifted, and a requirement that a demand electric system modeled. The total production cost of the combined
response provider may not simultaneously provide reserves and energy FRCC-SERC system ranged from $11.6 million to $14.2 million for
shifting for more than its total capacity. different PV penetrations before the addition of any flexibility options,
For this work, we analyze a set of scenarios to study the impact of as in Table 2. Increasing PV penetration reduces the production cost
demand response in comparison with or in addition to other measures due to lower usage of fuel, reducing the overall fuel costs in the system.
of flexibility. This includes battery capacities of 1 GW and 4 GW with Demand response reduces the total production costs in all cases;
6 hours of energy storage, and a set of conditions, called the Flex however, the Low DR scenario does not significantly impact costs at low
System, to increase operational flexibility: reducing minimum genera- PV penetrations. Table 2 shows the total reduction in cost for a range of
tion levels of gas combined-cycle (CC) generators, enabling reserves PV penetrations in the base flexibility scenario. The Low DR scenario
sharing between regions, and allowing PV to provide reserves. The reduces costs by 0.1% at 5% PV, and by 0.8% at 45% PV. The benefits
scenario framework can be seen in Fig. 1. Each scenario consists of a of the Low DR scenario are more operational in nature, and will be
particular PV penetration, demand response option, and flexibility op- discussed in subsequent sections. The High DR scenario has a higher
tion. impact, reducing costs by 0.5% at 5% PV, 1.0% at 25% PV and 2.2% at
45% PV. This is a much more significant impact on the overall costs, in
addition to operational changes seen due to DR.
3. Results

Table 2
The traditional role of demand response is to reduce load at peak Production cost reduction from demand response in the baseline flexibility scenario.
times, or in times of system emergency, so as to improve system re-
liability and economic efficiency. The outcomes of the fulfillment of Flexibility DR Scenario Total Production Cost (million $)
such roles include avoided new capacity builds, and reduced production Scenario
5% PV 15% PV 25% PV 35% PV 45% PV
costs. Production cost savings can accrue from reduced prices in peak
times, reduced reserve costs (especially to cover contingencies), and Base No DR 14,232 13,148 12,372 11,841 11,607
reduced startups and shutdowns of other generators. These types of Reduction in Total Production Cost (million $) from
benefits are summarized both monetarily and from an operational point Base, no DR scenario
5% PV 15% PV 25% PV 35% PV 45% PV
of view in the first part of this section.
Base Low DR 20.98 31.92 40.60 67.63 95.21
In emerging power systems that include more wind and solar, DR High DR 76.35 93.43 127.75 181.72 259.15
may assume additional roles. In particular for high penetration PV

2
Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted
manuscript. The published version of the article is available from the relevant publisher.
B. Stoll et al. The Electricity Journal xxx (xxxx) xxx–xxx

Fig. 2. Contingency reserve provision by generator type for each PV penetration and flexibility scenario.

Generally, demand response has a greater impact on costs at higher variation and uncertainty in net load. Many fewer types of demand
PV penetrations due to its ability to reduce curtailment, enabling higher response are able to provide regulation reserves because of this con-
utilization of a zero-fuel-cost resource. At low PV penetrations, demand tinuous signal-following requirement.
response is typically used to switch between coal and natural gas fuel In our modeling, demand response was able to provide nearly all
types. required contingency reserves in all cases, even for the Low DR sce-
nario, as in Fig. 2. Demand response predominantly displaced gas CC
3.1.2. Reserves units from providing reserves, as well as a small amount of coal and gas
Many types of demand response are well suited for providing re- CT units. Contingency reserves do not increase with rising PV pene-
serves, particularly contingency reserves. Contingency reserves are held tration, so their quantity does not change with PV penetration. Reg-
to cover potential generator or transmission line outages, and are ty- ulation reserves increase in line with rising PV penetrations such that
pically called infrequently. Regulation reserves are used to maintain demand response in the High DR scenario can provide nearly all reg-
system frequency at short time scales, and are called on throughout the ulation reserves at 5% PV, but as the requirements rise it cannot
day to change their output levels every 4–6 seconds to balance out maintain the same proportion of the acquired reserves, as in Fig. 3.

Fig. 3. Regulation reserve provision by generator type for each PV penetration and flexibility scenario.

3
Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted
manuscript. The published version of the article is available from the relevant publisher.
B. Stoll et al. The Electricity Journal xxx (xxxx) xxx–xxx

Fig. 4. Mean number of hours two generator types


spent operating at their minimum stable levels for
each PV penetration and several flexibility sce-
narios.

While it is fairly unrealistic that DR could provide all or nearly all spent at minimum stable level, until 35% PV penetration, at which
reserves of either type since many balancing areas limit the amount of point it rises. A large part of the difference in this behavior between
non-conventional capacity that can provide reserves, this does de- coal and gas CC units is that gas units have a lower startup cost, so are
monstrate the degree to which DR is capable of filling the reserves more likely to turn off during the daytime as solar penetrations increase
market. Other types of flexibility, such as batteries or allowing PV to (Denholm et al., 2016). DR is able to reduce the number of hours spent
provide reserves, also quickly fill the market. This is discussed further in at the minimum stable level for CCs in all scenarios, with the High DR
Section 3.2.2. scenario having a higher impact than the 1 GW battery scenario. DR has
a median reduction of 22 hours in the Low DR scenario and 50 hours in
the High DR scenario, with a range of −1 to 123 hours and −4 to
3.1.3. Operational impacts on traditional generators
250 hours, respectively. When other flexibility options are added DR
An important aspect of demand response is how it impacts other
has less of an impact than when DR is the only flexibility option. In-
generators and their operations. Traditional generators have a
creased flexible generation, including the large battery and the High DR
minimum stable level they must operate above; however, they operate
scenario, are able to eliminate the increase in CC hours at minimum
much less efficiently at this level compared to full capacity. As solar
generation seen at high PV penetrations.
penetrations increase, coal generators tend to reduce their output sig-
Another important operational metric is the number of starts and
nificantly during the day so they are available during the evening peak-
shutdowns an average plant undergoes per year. This is another in-
load hours (Denholm et al., 2016). However, this can be costly for
dicator of the cycling of these units, which can increase wear on the
generators in terms of efficiency losses and there are concerns about
plants. Fig. 5 shows the number of starts in the No DR scenario for each
equipment wear. Demand response can help mitigate this issue by
generator type and flexibility scenario. In all scenarios, the number of
shifting energy to times of high solar output, allowing these generators
starts increases for all generator types at higher PV penetrations,
to operate above their minimum stable level.
though coal starts remain relatively constant. Gas CC units see a high
Fig. 4 shows the number of hours coal and gas CC generators spend
increase in the number of starts, rising from 50 starts/year at 5% PV
operating at their minimum stable level in the base and battery only
penetration to 150 starts/year at 45% PV penetration. Gas CT plants see
scenarios, with and without DR. Coal generators spend an increasing
a similar three-fold rise in the number of starts in the base scenario.
number of hours at their minimum stable level as the PV penetration
Demand response has a negligible impact on coal at low penetra-
increases in all scenarios. The presence of batteries and demand re-
tions, and may increase the number of starts per year slightly at higher
sponse, individually or together, reduces the number of hours coal
penetrations. The ability of DR to reduce the number of starts of gas and
generators spend at their minimum levels. The Low DR scenario reduces
CC plants varies based on the other flexibility scenarios included, Fig. 6.
the mean number of hours at minimum level by −11 to 112 hours per
DR is most effective at reducing starts of gas generators, which are
year and the High DR scenario reduces the mean number of hours by
typically used for their quick-start capabilities during evening ramping
51–214 hours per year. The median values across all scenarios are 37
and high net-load periods. DR can reduce nearly half of the starts of
and 102 hours of reduction, for the Low and High DR scenarios re-
these generators in the High DR scenario. The impact on CC generators
spectively. The High DR scenario and the 1 GW battery scenario reduce
varies by PV penetration; the impact of DR levels off or even decreases
the number of hours at minimum stable level for coal generators by
drastically at higher PV penetrations. The addition of flexibility options
approximately the same amount.
tends to increase the ability of DR to reduce starts in CC generators at
Gas CC generators show a more complicated pattern of time spent at
high PV penetrations.
their minimum stable levels as a function of rising PV penetration
(Fig. 4). In the baseline scenario, a decreasing number of hours are

Fig. 5. Average number of times each generator


type was started per year in the no-DR scenario, for
each flexibility option and PV penetration. In gen-
eral, higher PV penetrations lead to increased gen-
erator starts, whereas flexibility options reduce
starts.

4
Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted
manuscript. The published version of the article is available from the relevant publisher.
B. Stoll et al. The Electricity Journal xxx (xxxx) xxx–xxx

Fig. 6. Reduction in the mean generator starts per year for each generator category by demand response, for all flexibility options.

3.2. Interactions between DR, PV, and batteries most operational flexibility – manufacturing and wastewater pumping.
We assumed that both manufacturing and pumping had enough excess
3.2.1. Impact of DR on PV curtailment capacity to increase their output during any hour of the day, making
One way in which DR interacts with PV capacity is by reducing them valuable for curtailment reduction. On the other hand, many re-
curtailment on high PV penetration systems. Curtailment decreases the sidential DR end uses have more limited flexibility during daytime
value of PV and increases system costs, since the low-cost energy from hours due to occupancy and comfort restrictions, limiting their ability
PV must be replaced by higher-cost generation from other fuel sources. to impact curtailment.
Curtailment increases significantly at higher penetrations (Fig. 7). Very The capability to increase load grows at higher PV penetration for
little curtailment occurs before 20% PV penetration, but curtailment all DR types due to the higher number of hours during which curtail-
rises rapidly from 8% at 30% PV to 28% at 45% PV in the base scenario. ment occurs. As PV penetrations rise, curtailment tends to occur during
Nearly all flexibility scenarios are able to reduce curtailment, with the more hours of the day, making it more likely that DR providers will be
Flex System scenario being the only one to increase curtailment at PV able to increase their load during those times.
penetrations less than 30%. The large battery is most effective at re-
ducing curtailment, reducing the overall curtailment by nearly half in 3.2.2. Reserves impacts
the highest-penetration systems studied. Demand response can provide a large fraction of all required re-
Demand response is able to reduce curtailment in all scenarios. The serves, as discussed in Section 3.1.2. How this changes with rising PV
High DR scenario is able to reduce curtailment by an amount ap- penetration and the addition of other flexibility options (Fig. 9) pro-
proximately equivalent to a 1 GW battery, enabling up to 6 TWh of vides interesting insights into the reserves market and the value of
additional PV generation in the Base scenario at 45% PV. demand response. In the base scenario, contingency reserve provision is
The different DR end-uses have varying potentials to reduce cur- relatively constant or decreases at increasing PV due to higher use of
tailment based on whether or not they have operational constraints that these resources for energy shifting as curtailment increases. Regulation
limit their ability to shift load to hours in which curtailment occurs. reserve provision increases steadily due to an increase in the reserve
Fig. 8 shows the ability of each end use to reduce curtailment based on requirement as the PV penetration increases and excess availability to
the total amount of load they increased during hours when curtailment provide regulation reserves for many end uses in the High DR scenarios.
occurred in the base scenario. The end uses with the highest ability to The High DR scenario is able to provide more reserves, but a much
increase load, and thus impact curtailment, are those end uses with the smaller fraction of the DR resource is used to provide reserves than in

Fig. 7. Total curtailed energy per year, TWh, for each flexibility op-
tion and penetration. Total curtailment increases sharply at rising PV
penetrations.

5
Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted
manuscript. The published version of the article is available from the relevant publisher.
B. Stoll et al. The Electricity Journal xxx (xxxx) xxx–xxx

Fig. 8. Increased load by demand response end-uses


during hours of curtailment in the no-DR scenario.
This indicates the ability of DR end-use to reduce
curtailment at each PV penetration. Some DR types
have higher availability and flexibility during times
of curtailment, making them more valuable for load
shifting at higher PV penetrations.

Fig. 9. Total annual reserves provision by demand


response for all flexibility scenarios.

the Low DR scenario. The Low DR scenario itself can provide nearly all when both are present. Fig. 10 shows the mean daily charge and dis-
contingency reserves in the model, so there is little additional need for charge patterns of batteries (a) and demand response (b). The overall
reserves provision in the High DR scenario. The presence of other differences between battery and DR usage stem from the different
flexibility options also reduces the amount of reserves provided by DR. constraints on how the technologies operate. Batteries have fewer re-
The Flex scenario suite allows PV to provide reserves, which it does at strictions on their usage compared to DR, which enables batteries to
higher PV penetrations in all scenarios including that flexibility option more fully shift load away from evening peak hours.
(Figs. 2–3). Batteries, which also have very low operational costs, We find, however, little to no interaction between battery and DR
compete with DR to provide low-cost reserves, and typically take away dispatch patterns at any PV penetration (Fig. 10). While the use pat-
some market share from DR, reducing the overall amount of reserves terns of both DR and batteries changes substantially at increasing PV
DR provides at all penetrations. penetrations, the addition of either DR or batteries to a system already
These factors indicate that market saturation of the reserves market containing capacity of the other technology does not substantially alter
is reached fairly quickly. The market itself is fairly shallow, with reserve the charging or discharging patterns of either resource.
prices ranging from 0 to 226 $/MWh in the 5% PV base scenario and 0 Interestingly, we found that the presence of both resources in-
to 250 $/MWh in the 30% PV base scenario. The median reserve price is creased the usage of both for generation. Fig. 11 shows the annual
41 $/MWh for both scenarios. As the number of resources that can generation from batteries and DR, which tends to increase at higher PV
cheaply provide reserves, including batteries and PV in addition to DR, penetration. Usage of batteries (Fig. 11a) increases as more demand
increases, the reserves price will decrease, as will the value each can get response is included in the system. This also occurs for DR at low PV
from providing reserves. As renewable penetrations increase, reserve penetrations; however, at high PV penetrations the addition of batteries
markets may change. For example, more jurisdictions may add reserve has no effect or a slightly negative impact (Fig. 11b). This increase with
products similar to the flexible ramping product recently implemented additional flexible resources is largely tied to the reserves market. When
in the California ISO1 (California I.S.O., 2015). only batteries or only DR are included, much of the capacity ends up
providing reserves. As discussed in Section 3.2.2, the addition of flex-
ible resources eventually saturates the reserves market, leading to an
3.2.3. Impacts of batteries on DR generation increased use of both resources to provide energy shifting. The total
Batteries and demand response provide very similar services: the provision of generation and reserves reduces for both resources when
ability to shift energy usage in time and provide reserves. Here we in- the other resource is included.
vestigate the degree to which these resources compete with each other
4. Discussion
1
As of November 1, 2016, see https://www.caiso.com/informed/Pages/
StakeholderProcesses/FlexibleRampingProduct.aspx. Demand response resources have the ability to shift generation

6
Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted
manuscript. The published version of the article is available from the relevant publisher.
B. Stoll et al. The Electricity Journal xxx (xxxx) xxx–xxx

Fig. 10. Battery (a) and Demand Response (b) operations on average for each hour of the day for 10%, 20% and 30% PV penetrations.

Fig. 11. Annual generation from batteries (a) and


demand response (b). Total generation tends to in-
crease with the addition of flexible resources, lar-
gely due to additional availability from providing
fewer reserves.

away from high-load periods, enabling more efficient use of resources low net load, particularly at higher PV penetrations, enables lower-cost
and lowering the total cost of operating the electric grid. While the resources to meet demand. This is also demonstrated by reducing
overall impact of demand response on total costs is small, typically less number of times gas turbines are started, a decline of 20–29% in the
than a 1% cost reduction, the addition of this resource is able to im- Low DR scenario and 47–61% in the High DR scenario. Gas turbines are
prove other aspects of grid operations. Demand response was able to more expensive units that can rapidly start and ramp quickly, so are
reduce the number of hours coal generators spent at their minimum typically deployed during peak periods. Demand response reduces the
stable level by 37 and 102 hours for the Low DR and High DR scenarios, reliance on these units by shifting small amounts of load for a few
respectively. The shifting of energy from peak load hours to hours of hours, which lessens the need for quick-start units and allows baseload

7
Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted
manuscript. The published version of the article is available from the relevant publisher.
B. Stoll et al. The Electricity Journal xxx (xxxx) xxx–xxx

generators, such as coal units, to provide that demand. National Renewable Energy Laboratory, Golden, Colorado Technical Report NREL/
Another important potential role of DR is providing operating re- TP-6A20–65298, Jan.
Lopez, A., Roberts, B., Heimiller, D., Blair, N., Porro, G., 2012. U.S. Renewable Energy
serves. These resources are particularly well suited for providing con- Technical Potentials: A GIS-Based Analysis. National Renewable Energy Laboratory,
tingency reserves; we found that in both the Low and High DR sce- Golden, Colorado Technical Report NREL/TP-6A20–51946, Jul.
narios, nearly all contingency reserves can be provided by demand NREL, 2016. Annual Technology Baseline. National Renewable Energy Laboratory
(NREL), Golden, CO Sep.
response. While this may be unrealistic due to requirements for firm Lazard, 2016. Lazard’s Levelized Cost of Energy Analysis—Version 10.0. Dec.
capacity to provide some fraction of reserves, it demonstrates the Hale, E.T., Stoll, B.L., Novacheck, J.E., 2017. Enabling More Solar in Florida: Potential
capability of DR to provide this service, even given low penetrations of Roles for Power System Flexibility. Appl. Energy [Forthcoming].
Lee, M.P., et al., 2015. Assessment of Demand Response & Advanced Metering. Federal
the resource. However, we found that the current reserves market is Energy and Regulatory Commission (FERC), Washington, D.C Staff Report, Dec.
fairly shallow, and easily satisfied by a variety of low-cost resources Energy Exemplar, PLEXOS, 2014. https://energyexemplar.com/software/plexos-desktop-
such as batteries and PV. Demand response providers looking to re- edition/.
Denholm, P., Novacheck, J., Jorgenson, J., O’Connell, M., 2016. Impact of Flexibility
serves for a majority of their revenue should take this into considera-
Options on Grid Economic Carrying Capacity of Solar and Wind: Three Case Studies.
tion, particularly as batteries and other flexible resources decrease in National Renewable Energy Laboratory, Golden, Colorado Technical Report NREL/
cost and regulations change to allow variable generators to participate TP-6A20-66854, Dec.
in reserves markets. Olsen, D.J., et al., 2013. Grid Integration of Aggregated Demand Response, Part 1: Load
Availability Profiles and Constraints for the Western Interconnection. Lawrence
We see a similar saturation of other values provided by DR as larger Berkeley National Laboratory, Berkeley, CA Technical Report LBNL-6417E, Sep.
amounts of flexible generators are included in the model, though to a Hummon, M., Palchak, D., Denholm, P., Jorgenson, J., 2013. Grid Integration of
much smaller degree than the saturation of the reserves market. While Aggregated Demand Response, Part 2: Modeling Demand Response in a Production
Cost Model. NREL (National Renewable Energy Laboratory (NREL), Golden, CO
the presence of energy storage capacity increases the amount of gen- Technical Report NREL/TP-6A20-58492, Dec.
eration provided by DR, particularly at low penetrations, this increase is California I.S.O, 2015. Flexible Ramping Product, Revised Draft Final Proposal.
smaller than the reduction in reserves provision, leading to an overall December.
lower use of the resource. The impacts of demand response on starting
Brady Stoll is an engineer in the Grid Systems Analysis Group at the National Renewable
of gas and CC generators, as well as the number of hours at minimum Energy Laboratory. Her research interests include the potential for demand response in
stable level for coal units, are not greatly affected by increasing flexible the United States and abroad, in particular using production cost models to analyze the
generation, indicating that demand response can continue to provide impacts of demand response and other flexible technologies on operations of the electric
grid. She also is interested in capacity expansion modeling with the Resource Planning
operational value to the grid even as overall flexibility increases. Model, a tool that analyzes future challenges for the electric grid including policy impacts
and falling renewable energy costs. Dr. Stoll received her Ph.D. in Mechanical
Acknowledgements Engineering from the University of Texas at Austin. She also holds a master’s degree in
Mechanical Engineering and a bachelor’s degree in Physics from University of Texas at
Austin.
The authors thank Josh Novacheck and Paul Denholm for providing
the underlying model and providing their insight for analyzing flex- Elizabeth Buechler is a graduate student at Stanford University studying Mechanical
ibility in Florida. Thanks to Laurel Dunn (Lawrence Berkeley National Engineering. She studied demand response at the National Renewable Energy Laboratory
Laboratory) for providing the demand response resource dataset used in during the summer of 2017 with a focus on the operational impacts and value of demand
response end-uses. Ms. Buechler has also performed probabilistic, hygrothermal modeling
this work, and to Galen Maclaurin for shepherding the dataset across of the indoor climates of residential buildings using Monte Carlo methods and the si-
cyberspace. The authors also thank Steve Capanna and Ookie Ma (U.S. mulation engine EnergyPlus at Oak Ridge National Laboratory. She received her B.S. in
Department of Energy) for supporting this work. This research was Mechanical Engineering from Tufts University.

funded by the U.S. Department of Energy under contract number DE-


Elaine Hale is a senior engineer in the Forecasting & Modeling Group at the National
AC36-08GO28308. Any errors or omissions are the sole responsibility of Renewable Energy Laboratory. She has over 15 years of experience in computational
the authors. systems engineering, numerical optimization, and software development; including six
years developing software for building energy modeling. Her current work focuses on
modeling and analyzing power system flexibility options, including demand response, in
References production cost and capacity expansion models under high renewables futures. Dr. Hale
received her Ph.D. and M.S. degrees in Chemical Engineering from the University of Texas
Gagnon, P., Margolis, R., Melius, J., Phillips, C., Elmore, R., 2016. Rooftop Solar at Austin. She also holds a bachelor’s degree in Chemical Engineering from the Georgia
Photovoltaic Technical Potential in the United States: A Detailed Assessment. Institute of Technology.

8
Pursuant to the DOE Public Access Plan, this document represents the authors' peer-reviewed, accepted
manuscript. The published version of the article is available from the relevant publisher.

You might also like