1

Evolution of Operating Reserve Determination

in Wind Power Integration Studies
Erik Ela, Member, IEEE, Brendan Kirby, Senior Member, IEEE, Eamonn Lannoye, Student Member,
IEEE, Michael Milligan, Member, IEEE, Damian Flynn, Member, IEEE, Bob Zavadil, Member, IEEE
Mark O’Malley, Fellow IEEE

these are studies and that the results should be realized as

Abstract--The growth of wind power as an electrical power such, much controversy often occurs over these costs due to
generation resource has produced great benefits with reductions the many assumptions that are required to be made. For
in emissions and the supply of zero cost fuel. It also has created example, wind forecasts used in the study are usually based on
challenges for the operation of power systems arising from the
a model of how a forecast may be produced some time in the
increased variability and uncertainty it has introduced. A
number of studies have been performed over the past decade to future (looking at a past weather year). Since predicting the
analyze the operational impacts that can occur at high actual output of the wind resource from a model is a difficult
penetrations of wind. One of the most crucial impacts is the task in itself, predicting what the forecast error may be can
amount of incremental operating reserves required due to the add more uncertainty to the results.
variability and uncertainty of wind generation. This paper Secondly, the determination of operating reserves has been
describes different assumptions and methods utilized to calculate
analyzed using many different methodologies and can differ
the amount of different types of reserves carried, and how these
methods have evolved as more studies have been performed. significantly from study to study. Operating reserves are
subject to many different naming conventions in different
Index Terms--operating reserves, power system operation, regions throughout the world. This paper defines operating
power system reliability, power systems, wind power generation reserve as the real power capacity that can be called on at any
instance of imbalance between generation and load.1 Most
I. INTRODUCTION wind power integration studies run hourly simulations of bulk

W IND power has seen rapid growth in the past decade. Its
zero-cost fuel and emissions-free output provide great
benefits to consumers and society. Utility-scale wind is a new
power system operations for a particular study area, and
therefore the actual utilization of the designated operating
reserve capacity is not in fact realized in detail. Therefore,
resource and is increasing at such a rapid rate that utilities and operating reserve requirements are determined statistically, but
system operators are becoming concerned about the usually not validated in simulations. Two important objectives
integration issues and costs that it introduces. Wind power that often form part of these studies are the costs or savings of
integration studies have been performed by numerous entities integrating additional wind power and the operational changes
to help understand and quantify these impacts [1], [2]. The that are recommended at high penetrations of wind power.
studies typically simulate a future power system with high Many of the studies recommend the use of incremental
wind penetrations, and evaluate the impacts on the grid and operating reserves, which will also affect the total costs. This
the incremental operating costs that result [3]. These studies makes the assumptions used in the methodology to assess
have been maturing continuously as the state of the art operating reserve a very important component of the overall
advances, with each study generally building on previous study. Many areas will adopt these methodologies from the
studies. studies as wind penetrations increase and therefore it is
Some of these studies have compared the costs and additionally important for actual system operations in the
operational differences between a system with high wind future.
penetration and a system that does not bring the incremental This paper will focus on methods of determining operating
variability and uncertainty that wind presents. The additional reserves for power systems with high penetrations of wind
costs generally occur because the unit commitment is power. Section II will describe current practices and
inefficient due to forecast errors and because it is adjusted to definitions of operating reserves in North America, mainland
provide more flexibility, accommodating wind’s increased Europe, and Ireland. Section III will cover some recent wind
variability and uncertainty. Additional flexibility can be in the power integration studies that have been performed, and will
form of increased ramp rates, decreased minimum generation focus on the methods used in each study when calculating the
limits, and increased amounts of operating reserve. Noting that operating reserve requirements. In section IV, the authors will
provide insight into the strengths of different methods,
E. Ela and M. Milligan are with the National Renewable Energy including analysis on how the operating reserve determination
Laboratory (email: [email protected], [email protected]) problem may change in wind power integration studies and in
E. Lannoye, D. Flynn, and M. O’Malley are with the University College
Dublin (email: [email protected], [email protected],
actual system operations. Section V concludes the discussion.
[email protected])
1
B. Kirby is a private consultant (email: [email protected]) Additionally, operating reserves or ancillary services can include voltage
B. Zavadil is with Enernex Corporation (email: [email protected]) or reactive power support as well as black start service.

978-1-4244-6551-4/10/$26.00
Authorized ©2010
licensed use limited to: UNIVERSIDADE IEEEDO RIO DE JANEIRO. Downloaded on February 20,2024 at 02:56:44 UTC from IEEE Xplore. Restrictions apply.
FEDERAL
 2

II. OPERATING RESERVE DEFINITIONS AND STANDARDS capacities to account for other transmission contingencies,
Variability and uncertainty are not unique to wind similar to generator operating reserve.
generation: similar characteristics in aggregate electric Other standards and policies detail how much a balancing
demand and even supply resources have always posed area will require of each type of operating reserve [4]. For
challenges for power system operators. Future loads cannot instance, the NERC BAL-002 standard requires that a
be perfectly predicted, loads and generator outputs can vary balancing authority or reserve sharing group maintain at least
substantially in different time frames, and large power system enough contingency reserve to cover the most severe single
equipment can fail at any given time without notice. Power contingency. For the western interconnection, this is extended
system operators secure different amounts and types of by a proposal by WECC to state that the minimum amount of
operating reserves to compensate for these characteristics in contingency reserve should be the greater of the most severe
order to serve load reliably and maintain the system single contingency or the sum of 3% of the balancing area
frequency. There are many different definitions and rules load and 3% of the balancing area generation. Detailed
concerning what operating reserves entail. For example, the specifications of contingency reserve requirements, including
North American Electric Reliability Corporation (NERC) the amount of spinning compared to supplemental reserve, are
defines operating reserves as the following: established by each Regional Reliability Organization.
“That capability above firm system demand required to Regions typically require at least half of the contingency
provide for regulation, load forecasting error, equipment reserve to be spinning. An example of how reserves are
forced and scheduled outages and local area protection. It deployed following a contingency is provided in Fig. 1.
consists of spinning and non-spinning reserve.” [4] Regulating reserve usually does not include explicit
In most of North America, these reserves can be further requirements. Instead, balancing areas will maintain sufficient
placed in three categories: regulating reserves so that they meet their CPS1 and CPS2
• Spinning reserve – The portion of Operating Reserve performance requirements. In some areas that currently have
consisting of: Generation synchronized to the system high penetrations of wind power, such as the Electric
and fully available to serve load within the Disturbance Reliability Council of Texas (ERCOT), the forecasted wind
Recovery Period following the contingency event; or power output is considered when defining regulating and other
Load fully removable from the system within the types of operating reserve requirements [5].
Disturbance Recovery Period following the
Market Response
contingency event.
• Supplemental reserve – The portion of Operating Supplemental Operating Reserve
Reserve consisting of: Generation (synchronized or
capable of being synchronized) that is fully available to Contingency Spinning & Non-Spinning Reserve
Occurs
serve load within the Disturbance Recovery Period
following the contingency event; or Load fully
removable from the system within the Disturbance Reserves "Should" Reserves
Recovery Period following the contingency event. be Restored "Must" be
Restored
• Regulating Reserve – An amount of reserve responsive Frequency
Response
to Automatic Generation Control which is sufficient to
provide normal regulating margin.
-10 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Further definitions can separate the second category by Minutes

response time and response sustainability. In North America, Fig. 1. Reserve deployment as defined by NERC [6]
the spinning reserve and supplemental reserve described above
are often combined and referred to as “contingency reserve”, In Europe, broad guidelines are given by the former TSO
only being used for instances of generator or network groupings such as Nordel and the Union for Coordination of
contingency events. Though contrary to the NERC definition Transmission of Electricity (UCTE), now part of the European
of operating reserve described above, regulating reserve is Network for Transmission System Operators for Electricity
generally procured in both the upward and downward (ENTSO-E). ENTSO-E defines reserve in three categories;
directions (i.e. in cases of over-generation). Fast frequency primary, secondary and tertiary control [7]. Primary control is
response (governor response) is not yet explicitly addressed by activated when system frequency deviates by 20 mHz from the
NERC as a distinct operating reserve, but the Western set point value (nominally 50 Hz) and must be fully
Electricity Coordinating Council (WECC) has started to study operational within 30 seconds. The purpose of primary control
the need for a 30 second response: frequency responsive is to limit the deviation of system frequency following a
reserve (FRR). Transmission also impacts the need for system event.
operating reserves, since it can provide access to additional Secondary control consists of units controlled by automatic
reserve supplies and reduce the overall need for reserves by generation control (AGC) and fast starting units. These are
increasing the reserve sharing pool. Limited transmission may engaged 30 seconds after a contingency event and must be
also mean more localized reserve requirements. Transmission fully operational within 15 minutes. This category of control
lines may be operated at levels below their maximum attempts to restore the frequency to its nominal value and
reduce the area control error. Tertiary control has a slower

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO RIO DE JANEIRO. Downloaded on February 20,2024 at 02:56:44 UTC from IEEE Xplore. Restrictions apply.
 3

response and is engaged to restore primary and secondary

control units back to the reserve state.

TABLE I
NAMES AND DEFINITIONS OF RESERVES

Name Definition Common terms

Operating Used for an imbalance between Operating reserve
Reserve generation and load at any time. All (US), reserve,
other reserve categories are a subset balancing reserve
of this.
Spinning Any of the categories where the Spinning reserve,
Reserve resources are frequency responsive synchronous
and begin responding immediately. reserve, on-line
Generators providing this service reserve, responsive
must have been committed. reserve
Non-Spinning Similar to spinning reserve but the Non-spinning
Fig. 2. ENTSOE-UCTE control mechanisms [7] Reserve resources do not need to respond to reserve, non-
frequency autonomously or begin synchronous
The Irish system represents a much smaller, isolated power responding immediately. reserve, off-line
system, and has a more granular approach to its definition of Generators providing this service reserve, quick start
do not have to be committed. reserve
reserve [8]. There are five main types including: regulating, Upward only.
operating, replacement, substitute, and contingency reserves. Contingency Used during instantaneous failures Contingency
Regulating reserve acts within 30 seconds of a frequency reserve (e.g. generator failures). reserve (US),
deviation to restore the frequency to within 0.1 Hz of the set Contingency reserves may include operating reserve
other types (e.g. frequency (Ireland), primary
point value and controls inter-system transfers on the North- responsive, spinning, non-spinning, and secondary
South interconnector which joins the two systems on the and supplemental). Upward only reserve (Europe)
island. This is a subset of primary operating reserve. Operating Regulating Used for frequency control and Regulating reserve,
reserve, as defined in the Irish system, is divided into three Reserve maintaining area control error frequency control
during normal (non-event)
parts: primary, secondary, and tertiary operating reserve. conditions in a time frame that is
Primary reserve acts for the first 15 seconds to avoid transient faster than energy markets clear.
nadirs below 49 Hz. Secondary reserve then acts to avoid Due to random movements.
continuous system operation below 49.5 Hz, and it is fully Requires automatic generation
control. Upward and downward.
available from 15 seconds after an event for a further 75 Load Used for frequency control and Load following,
seconds. Tertiary operating reserve is used to replace the following maintaining area control error dispatch
primary and secondary reserve. It is split into Tertiary1, which Reserve during normal (non-event)
restores primary and secondary operating reserve for the first 5 conditions. Due to non-random
movements. Slower movements
minutes and Tertiary2 reserve which is available after 5 than regulation. Does not require
minutes for an additional 15 minutes. automatic generation control.
Replacement reserve acts as a longer term resource to Upward and downward.
restore secondary and tertiary operating reserve from Frequency Reserves that provide the initial Governor response,
Responsive autonomous response to a major primary control
operation. This is fully available within 20 minutes for a four
Reserve disturbance. Upward and (Europe), FRR
hour period. Substitute reserve is utilized to restore downward.
replacement reserve after 4 hours for a duration of 24 hours Ramping Used during failures and events that Variable generation
and is available for the replacement of emissions-restricted Reserves are not instantaneous but occur over event reserve,
plant. Contingency reserve is available to restore all reserves long time frames (e.g., wind ramps, forecast error
forecast errors). Upward and reserve
24 hours after the event. downward.
The names and definitions for operating reserve across Supplemental Reserve that replaces faster reserve Replacement
different parts of the world vary but the functionality is very Reserve so that the system is secure and the reserve,
similar. Systems of different sizes and with different level of pre-event reserve is supplemental
restored. Upward only. reserve, tertiary
generation or load characteristics may require different reserve (Europe),
response requirements or different qualifications on how the contingency reserve
operating reserve is provided. However, it is important to note (Ireland), substitute
that regardless of category, response time, and qualifications, reserve
operating reserves as discussed here are all secured and
utilized for essentially the same reason; imbalance between The objective behind the table is to define a consistent set of
generation and load. Thus, to offer clarity and reduce categories based on the purpose of their deployment and any
confusion between North American and European readers, differences in how they are provided. Note that any of these
Table I displays the terminology that will be used throughout categories could have numerous response time definitions as
the rest of this paper. Reference [9] also shows a detailed the current standards generally do. Fig. 3 shows a hierarchical
mapping and descriptions of a subset of these terms for active diagram of the defined operating reserves and how they may
power control as well as for voltage control services. be interrelated. Note that the definitions of spinning and non-

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO RIO DE JANEIRO. Downloaded on February 20,2024 at 02:56:44 UTC from IEEE Xplore. Restrictions apply.
 4

spinning reserve are characteristics of each of the other types, reserve requirement similarly evaluated the added variability
as seen in the diagram. of wind but calculated it to be a 2 MW standard deviation for
every 100 MW wind plant installed. This calculation was
based on operational data from existing wind plants. The ratio
was used to calculate the regulating reserve requirement as
seen in equation (1).

where k is a factor relating regulation capacity requirement to

the standard deviation of the regulation variations (assumed to
be 5 in this study reflecting current practices); σload is the
Fig. 3. Operating Reserve Diagram standard deviation of regulation variations from load; σW100 is
the standard deviation of regulation variations from a 100 MW
III. RESERVE DETERMINATION IN WIND INTEGRATION wind plant; and N is the wind generation capacity in the
STUDIES scenario divided by 100. The results showed increases of 12,
In the past several years, various organizations have 16, and 20 MW for the 15, 20, and 25% cases, respectively.
participated and/or initiated wind power integration studies. The Minnesota study quantified two other defined
The studies evaluate a power system in the future with high categories that the New York study did not. In the study these
wind power penetrations and simulate the impacts that occur. are defined as load following and operating reserve margin
A general process for wind power integration studies can be (load following reserve and ramping reserve in Table I). Load
found in [10]. following was calculated as twice the standard deviation of the
A major component of each study is to evaluate the five minute changes in the net load, and increases ranged from
incremental need for additional operating reserves for the 10 to 24 MW for the three cases. The operating reserve margin
future system that result from high wind penetration. The was allocated specifically for hourly forecast errors in the net
study teams usually consider traditional definitions and load. This analysis assumed a dynamic requirement, one that
requirement determinations and propose needed changes to was not constant for all hours but in fact was a function of the
maintain reliability while accommodating the variability and amount of expected wind capacity. The consideration was that
uncertainty present in the wind power. This value is generally the variability of wind is highest when the wind capacity is in
calculated via statistical methods analyzing wind power time the middle range (i.e. 40-60% of total capacity) due to the
series that are modeled for the study. The methodologies used wind turbines being on the steepest parts of their wind speed
to compute these values have evolved as each study learns to wind power conversion curves. Therefore, more reserve
from past studies. The most recent studies evaluating very was needed for the middle range compared to times of very
high penetrations are using sophisticated methodologies that low wind generation or times of very high wind generation.
are diverging further from the traditional methods used today These studies paved the way for future thinking of how
in actual system operations. operating reserves should be handled with high penetrations of
In the US, the first two major wind power integration wind. More recent studies have evolved and taken the ideas
studies were performed in the states of New York and from those preceding studies with increasing sizes,
Minnesota [1], [2]. In New York, the study evaluated 3,300 penetrations, and scopes. The traditional definitions and
MW of wind power on the 33,000 MW peak load NYISO methods used in current operations were simply not feasible
system. The study concluded that no incremental contingency with penetrations of 20% wind energy and more. We will
reserves would be needed since the largest single severe focus on two recently completed studies and describe the
contingency would not change. The study concluded that an operating reserve determination methods explored in each.
additional 36 MW of regulating reserve was required on top of
A. All Island Grid Study (Ireland)
the current 175 - 250 MW procured today. This is a result of
analyzing the standard deviation of 6-second changes in load The All Island Grid Study in Ireland was published in 2007
net of wind compared with that of load alone. The standard and examined, among other things, the Irish system’s ability to
deviation with wind increased from 71 MW to 83 MW, integrate various penetrations of wind generation [11]. Six
presenting a 12 MW increase. As was the current guideline in plant portfolios were examined to meet the load forecasted for
New York, the total standard deviation is multiplied by three 2020. Portfolio 1 contained 2 GW of wind; 2, 3 and 4
to ensure that the total regulation requirement is sufficient to contained 4 GW; portfolio 5 contained 6 GW; and portfolio 6
cover 99.7% of all instances, thus giving the 36 MW increase. contained 8 GW of wind generation. This is in the context of
In Minnesota, the study evaluated 15, 20, and 25% wind a projected peak load of 9,618 MW and a load factor of
energy as a percentage of total annual demand (3441 MW, 63.9%. The study involved hourly scheduling of the system
4582 MW, and 5688 MW on a system with a peak demand of with the WILMAR system planning tool [12].
roughly 20,000 MW). Similar to New York, it was concluded The study incorporated a refined implementation for
that there would be no impact on the contingency reserve reserve provision with only two categories specified in the
requirement with the added wind penetrations. The regulating model: spinning and replacement reserve. The definition of a

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO RIO DE JANEIRO. Downloaded on February 20,2024 at 02:56:44 UTC from IEEE Xplore. Restrictions apply.
 5

unit capable of meeting the replacement reserve standard was horizons. In Fig. 6, where portfolio 5 contains 6 GW of wind
an off-line unit with a start up time of less than 60 minutes and generation, the requirement for replacement reserve is seen to
online units whose capacity was not allocated to the spinning exceed 3 GW in one instance. This is due to a 1 GW load rise
reserve requirement. This is a highly simplified model given at the same time as a 1 GW decrease in wind, combined with a
the existing structure of reserve provision in the Irish system. forecast error.
The requirements for spinning and replacement reserve were
based on a mixture of existing and proven requirements and
newer techniques for the provision of reserve for wind
generators.
The spinning reserve requirement is calculated as being the
size of the largest on-line unit plus an additional contribution
for wind generation, calculated based on the work in [13].
Ireland is an island system with one 500 MW interconnector
in operation and a 500 MW interconnector under construction.
System modeling for the year 2020 assumed that 100MW of
spinning reserve can be obtained through interconnection.
Another 50MW of reserve is assumed to be provided from
interruptible contract loads. Of the remainder, a constraint of a
maximum of 50% of reserve demand can be provided by Fig. 5. Average requirement for replacement reserve by time horizon
pumped storage. Wind generators are allowed to provide
spinning reserve through curtailment. 4000

Replacement Reserve
The demand for spinning reserve is illustrated in Fig. 4 on a
2000
weekly averaged basis. Spinning reserve is required more

Demand (MW)
frequently as the amount of wind increases in the portfolio, 0
significantly so in portfolio 6. The scheduled outage of the

1441
1921
2401
2881
3361
3841
4321
4801
5281
5761
6241
6721
7201
7681
8161
8641
1
481
961
largest unit on the system (480MW CCGT unit) is seen to
reduce the spinning reserve requirement significantly during
weeks 31 to 34. While the variable generation requires Hour
additional spinning reserve, the largest contributing factor
remains the loss of the largest conventional unit. P5 P3 P1
Fig. 6. Hourly requirement for replacement reserve

B. Eastern Wind Integration and Transmission Study (US)

The Eastern Wind Integration and Transmission Study
(EWITS) evaluated the operational impacts of various wind
penetrations, locations, and transmission build-out options for
most of the US eastern interconnection. The study included
three scenarios of 20% wind energy with each representing
different primary locations of the wind, and one 30% wind
energy scenario [14]. The majority of this region is currently
operated by Independent System Operators (ISO) and
Fig. 4. Weekly demand for spinning reserve for each generation portfolio
Regional Transmission Organizations (RTO) who administer
the wholesale electricity markets. These markets have evolved
Replacement reserve is calculated as a function of the since their inception in the late 1990s. The further evolution of
possible forced outages of units and an additional margin the rules and procedures that the markets will follow is a key
which is a function of the 90th percentile of the net load (load- assumption on how operating reserve requirements are
wind) forecast for each particular scenario. The 90th percentile determined in the study, with the boundary between operating
was chosen as it most closely matches experience with proven reserves and what is extracted from sub-hourly energy markets
reserve standards. WILMAR implements rolling unit also having an impact on the method used.
commitment and has stochastic optimization functionality, The first procedure of the study was to determine the
which requires the forecast data to be an input to the scenario contingency reserves required. As many previous US studies
tree tool and thus, replacement reserve is activated accordingly have done, these assumed the current rule and determined that
by the scheduling tool. Demand for replacement reserve is a the largest contingency was not affected by the large amounts
function of the installed wind power and the forecast error of wind generation. One and a half times the single largest
over longer timelines, as shown in Fig. 5 and Fig.6. Fig. 5 hazard in each operating region determined the amount of
shows how the replacement reserve requirement is a function contingency reserves for that region.
of how far ahead the optimization is evaluating. In other Many prior studies in the US concluded a slight, but not
words, generally, errors will be larger further out and therefore insignificant, increase in the amount of required regulating
more replacement reserve is required than more immediate

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO RIO DE JANEIRO. Downloaded on February 20,2024 at 02:56:44 UTC from IEEE Xplore. Restrictions apply.
 6

reserve due to the increased variability of wind added to that

of the load. In EWITS, a similar methodology to the prior
studies was performed. The minute to minute variability
separated from a 20 minute rolling average of a 100 MW wind
plant was used for the analysis and the standard deviation was
determined to be 1 MW. It was assumed that there is no
correlation between wind plants for power output deltas in this
time frame, and therefore the total standard deviation for a
balancing area was calculated by geometrically adding the 1-
MW standard deviation for all 100-MW wind plants on the
system. For load-only, the regulating reserve requirement was
assumed to be 1% of the total load, and assumed to be equal to
three times the standard deviation of the load variability. Since Fig. 7. 10-min ahead wind generation forecast errors as function of production
load and all wind variability on this timeframe were also
considered to be independent of one another, the standard A similar approach was used for the hour-ahead wind
deviations of all wind and all load were then geometrically forecast error. However, in this case it was assumed that the
added together by calculating the square root of the sum of errors that were not occurring often could be compensated for
their squares. The total standard deviation was increased by with off-line non-spinning reserve. Therefore, one standard
less than 1 MW when the wind was added to the load, and deviation of the hour-ahead forecast error was required to be
therefore the variability of wind was not considered as part of spinning, and two standard deviations could be non-spinning.
the regulating reserve for the study.2 Also, since the reserves were used in the production cost
Different from most studies, however, it was determined simulations for the study, it was ensured that if the reserves
that the uncertainty in the wind forecasts used for economic had to be used for the hour-ahead forecast error of the hour in
dispatch would impact the regulating reserve much more than question, those reserves did not have to be kept in real-time. In
what was shown for the variability. Economic dispatch other words, if reserves were needed because less wind was
programs that run every five minutes would use information available than forecast, the model would release that amount
from at least ten minutes before the operating interval. Since it of reserves in real-time since the reserves were used for the
is too late to adjust the economic dispatch for any deviations, forecast error and not needed further. The total amounts of all
these deviations would all be met by units providing reserves used in the study are shown in Fig. 8. The reserve
regulating reserve. Assuming a 10-min ahead persistence requirement was an hourly value that was a function of both
forecast, the additional regulating reserve was determined by wind and load levels.
looking at the standard deviation of ten-minute changes in
wind output (load forecast for 10-min ahead was assumed to
be quite good and load forecast error was ignored). Fig. 7
shows the standard deviation of the 10-min ahead wind
forecast errors as a function of the average hourly production
of the total wind. The highest variability is near 50%
production, where the anticipated 10-min change can be up or
down and also relates to wind turbines being at the steepest
part of the power conversion curve. The function was used for
the hourly wind-related standard deviation of the regulating
reserve requirement and was geometrically added to the load
regulating reserve requirement discussed above. Equation (2)
is shown below, where σst is the standard deviation of wind
forecast errors described in Fig. 7.

Fig. 8 Summary of reserve methodologies for EWITS

1%
3 2
3 IV. EVOLVING METHODS FOR DETERMINING RESERVE
REQUIREMENTS WITH HIGH PENETRATIONS OF WIND
GENERATION
The treatment of the reserve determination problem has
evolved substantially for both wind power integration studies
and in actual system operations. The authors believe that this
is a continuing trend and that there are still a number of
inconsistencies with the data and methodologies used, where
2 improvements can be made. The key issue to recognize is that
Calculations based on a balancing area with 100 GW load and 60 GW
wind, which was about the average for the largest ISO balancing areas that it is generally not difficult to be overly conservative and hold
were a part of the study. much more reserves than is needed. The real issue is knowing

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO RIO DE JANEIRO. Downloaded on February 20,2024 at 02:56:44 UTC from IEEE Xplore. Restrictions apply.
 7

what is causing the need for different types of reserves today, storage resources from providing operating reserves, when in
how high penetrations of wind power will change the needs of fact, in some cases, they could provide the service with as
reserves, and how to use as much information as is available at good or better performance than the generating resources that
any given time to determine an optimal and efficient amount had currently provided them. The amount of spinning reserve,
of operating reserves. compared to non-spinning reserve, for all of the different
The data quality used in wind power integration studies is categories should be evaluated both economically and
very crucial in the reserve determination problem. Studies are operationally. The response times for different categories
evaluating high penetrations of wind power that do not should be evaluated regarding what is optimal, and may also
currently exist so the estimated power output of the wind change depending on predicted system conditions. The sharing
generation must be modeled. Different modeling techniques between reserves should also be evaluated. For instance, if all
are usually employed where issues can often occur that differ ramping reserves are used up but a net load event continues to
the modeled output from what would be realistic. Data ramp, what are the consequences from taking contingency
quantity is also important as longer data sets and higher reserves for that purpose? Also, how should reserves be
resolution data can give much more information on the accounted for with a stochastic scheduling system that
anticipated behavior of wind power and how it affects reserve inherently schedules reserves without explicitly calling for
requirements. Lastly, the wind power forecasts used in studies them? This team has built a high resolution power system
are yet another model that may not be totally representative of model similar to those used in wind power integration studies.
wind power forecasts used in actual operations. The The model will focus on the utilization of reserves, however,
requirements for ramping reserves, load following reserves, at a very fine timescale (4-6 seconds for one day) and attempt
and regulating reserves may be highly dependent on how good to capture all of the different contributing factors that would
a wind forecast is, so it is important that the error cause the utilization of different types of operating reserves.
characteristics of the synthetic forecasts closely match those in This research should help highlight how different operating
operations today and in the future. reserves could be determined based on a function of the
Dynamic reserve requirements have been proposed in many predicted operating condition inputs. Table II shows further
of the recent studies [2], [5], [11], and [14]. In operations additional research ideas for each of the operating reserve
today, most reserves are static and even those that vary hour categories defined in Table I.
by hour are usually based on hourly rules, not on forecasted
conditions. It is important that if more reserves are needed TABLE II
OPERATING RESERVE DETERMINATION TRENDS
because of certain operating conditions, those conditions must
be taken into account when the control area operator decides
Name Trends and future research questions
the operating reserve requirement. In the future, we see it Spinning Reserve For each operating reserve category, the percentage of
possible that each balancing area operator will have a reserve reserve that is spinning should be based on reliability
requirement that is a function of load forecast, variable and economics. Quicker responses need to be spinning.
generation forecast, net load variability forecast, uncertainty For event reserves, events that occur frequently should
have more spinning reserve, based on a tradeoff between
predictions (i.e. confidence of forecasts), and possibly even spinning reserves having a higher standby cost, but non-
information on the predicted behavior of conventional spinning reserves having a higher utilization cost. Should
generation. demand response or generation that can be started in
In addition to reserve requirements that vary by time, they extremely quick times (e.g. < 1 minute) be considered
spinning?
may also vary by time horizon. Today, the majority of Non-Spinning See above. How quickly must these resources start up for
operating reserves are dominated by contingency and Reserve the different categories?
regulating reserves (considering the first tier of Fig. 3). These Contingency Should this only be used for major failures? Can we
reserves are mostly used today due to generation and Reserve share this with other categories? Can we use a fully
probabilistic approach with forced outage rates of all
transmission failures, and load and generation variability. generators and facilities, rather than simply using the
These phenomena generally are as likely to happen in the next largest hazard (for instance if many smaller units
five minutes as they are tomorrow (i.e. their likelihood does dominate a region rather than larger ones)?
not change with look-ahead horizon). On the other hand, Regulating Is this based on variability or forecast errors? Can
Reserve uncertainty or variability predictions (rather than energy
events that are caused by wind may be much better predicted
forecasts) be used for determination of the requirement?
as the operating time gets closer. This was introduced in the How good does regulating reserve have to be?
All Island Grid Study [11] for the use of replacement reserves. Load following Unit commitments are performed on an hourly resolution
When operators are more confident in the outcome of Reserve and normally load trends monotonically in one direction
operating conditions the need to hold reserves is reduced. This within an hour. High wind penetrations may change this
assumption so that load following reserves are set aside
applies mostly to ramping and load following reserves but for the hour to meet changes within the hour. Conditions
may apply to other categories as well. where the short-term economic generation and load
The last point that the authors would like to make concerns response stack has insufficient response capability need
the traditional requirements that areas throughout the world to be identified and addressed with a dedicated response
service.
have set for operating reserves. The requirements should Frequency What type of response is needed (appropriate droop)?
always be set towards the needs of the power system, not Responsive
towards the resources that are currently available to provide Reserve
those needs. This issue has arisen recently due to requirements Ramping Should the response time requirement be a function of
Reserves the net load ramp prediction? Should the spinning and
that originally precluded many demand response resources and non-spinning contribution be a function of the

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO RIO DE JANEIRO. Downloaded on February 20,2024 at 02:56:44 UTC from IEEE Xplore. Restrictions apply.
 8

confidence of the anticipated ramp event? [14] Enernex Corporation, “Eastern Wind Integration and Transmission
Supplemental How quickly should each of the categories be replaced? Study,” Prepared for the National Renewable Energy Laboratory,
Reserve Is there any reason that downward reserves should be January 2010. Available:
considered so that if too much are used that they can be http://www.nrel.gov/wind/systemsintegration/pdfs/2010/ewits_final_rep
replaced? ort.pdf.

V. SUMMARY VII. BIOGRAPHIES

Erik Ela (M ’05) received the B.S.E.E degree from Binghamton University
This paper describes new methods of determining the and the M.S. degree in Power Systems at the Illinois Institute of Technology.
optimal amount of operating reserves for systems with high He joined the NREL grid integration team to work on different wind
wind penetrations. Wind power integration studies are integration issues. His experience lies mostly in different topics relating to
analyzing innovative methods on this issue, but different grid operations and market operations. Erik previously worked for the New
York ISO developing and improving products in the energy markets and
methods can produce substantially different results. The All operations areas. Erik is a member of IEEE and the Power and Energy
Island Grid Study and the Eastern Wind Integration and Society.
Transmission Study are two of the most recent studies of their
kind and both have made key contributions to the optimal Brendan Kirby (SM ‘76) is a private consultant with numerous clients
operating reserve requirement determination problem. including AWEA, National Renewable Energy Laboratory, Oak Ridge
National Laboratory, EPRI and others. He recently retired as a senior
Most of the wind power integration studies that have been researcher with the Oak Ridge National Laboratory’s Power Systems
performed to date run hourly simulations of a security Research Program. He has 34 years of electric utility experience and has
constrained unit commitment and dispatch of the system. Few published over 120 papers, articles, and reports on ancillary services, wind
are truly capturing the details at a level that can replicate integration, restructuring, the use of responsive load as a bulk system
reliability resource, and power system reliability. He has a patent for
system operations, whereby operating reserves are being responsive loads providing real-power regulation and is the author of a NERC
utilized in detail. Therefore, the statistical methods used in the certified course on Introduction to Bulk Power Systems: Physics / Economics
studies have not been validated extensively. The authors finish / Regulatory Policy.
by discussing different assumptions and evolving methods on
Eamonn Lannoye (S’08) received the B.E (Hons.) degree in mechanical
determining the optimal amount of operating reserves with engineering from University College Dublin, Dublin, Ireland in 2009. He
high penetrations of wind power both in future studies and joined the Electricity Research Centre in UCD as an Irish Research Council
ultimately in actual system operations. for Science, Engineering and Technology Embark Initiative doctoral student.
His research is concerned with flexibility in power systems with high
penetrations of variable generation.
VI. REFERENCES
[1] Richard Piwko, Xinggang Bai, Kara Clark, Gary Jordan, Nicholas Michael Milligan (M ’98) is part of the Transmission and Grid Integration
Miller, and Joy Zimberlin. “The Effects of Integrating Wind Power on Team at the National Renewable Energy Laboratory. He has authored or
Transmission System Planning, Reliability, and Operations: Report on coauthored more than 100 papers and book chapters and has served on
Phase 2,” Prepared for The New York State Energy Research and numerous technical review committees for wind integration studies around the
Development Authority, Mar. 2005. Available: U.S. Michael is a member of the leadership team of the NERC Variable
http://www.nyiso.com/public/webdocs/services/planning/special_studies Generation Task Force, member of WECC’s Variable Generation
/wind_integration_report.pdf. Subcommittee, member of the IEEE Wind Power Coordinating Committee,,
[2] Enernex Corporation, “2006 Minnesota Wind Integration Study Volume and served on the Western Governors’ Association Clean and Diverse Energy
I,” Nov. 2006. Available: http://www.uwig.org/windrpt_vol%201.pdf. Wind Task Force. Michael has M.A. and Ph.D. degrees from the University of
[3] J. C. Smith, M. Milligan, E. Demeo, and B. Parsons, “Utility Wind Colorado, and a B.A. from Albion College.
Integration and Operating Impact State of the Art,” IEEE Trans. Power
Syst., vol. 22, pp. 900-908, Aug. 2007. Damian Flynn (M’96) is a senior lecturer in power engineering at University
[4] North American Electric Reliability Corporation, “Reliability Standards College Dublin. His research interests involve an investigation of the effects
for the Bulk Electric Systems of North America,” November 2009. of embedded generation sources, especially renewables, on the operation of
[5] “ERCOT Methodologies for Determining Ancillary Service power systems. He is also interested in advanced modeling and control
Requirements,” 2009. techniques applied to power plant. He is a member of the IEEE.
[6] B. Kirby, 2006, Demand Response For Power System Reliability: FAQ,
ORNL/TM 2006/565, Oak Ridge National Bob Zavadil (M) is co-founder of EnerNex and is responsible for developing
Laboratory, December 2006. and overseeing the company’s power system engineering consulting business.
[7] Report: Entsoe, UCTE Operational Handbook Policy 1, [Online] He has worked on electric power system issues for wind generation for over
http://www.entsoe.eu/index.php?id=57 15 years. From 1989 to the summer of 2003, Mr. Zavadil served in various
[8] Report: Eirgrid, All Island Reserve Review, Dublin, Ireland, 2003. consulting and product development capacities for Electrotek Concepts and its
[9] Y. Rebours, D. Kirschen, M. Trotignon, and S. Rossignol, “A Survey of parent company, WPT. Mr. Zavadil began his career in the electric power
Frequency and Voltage Control Ancillary Services – Part I: Technical industry in 1982 as a special studies engineer in the Transmission and
Features,” IEEE Trans. Power Syst., vol. 22, no. 1, pp. 350-, Feb 2007. Distribution Engineering Division of the Nebraska Public Power District. He
[10] E. Ela, M. Milligan, B. Parsons, D. Lew, and D. Corbus, “The evolution is a member of the IEEE Power Engineering, Power Electronics, and
of wind power integration studies: past, present, and future,” Industrial Applications Societies, and serves as Secretary of the IEEE PES
Proceedings of IEEE PES General Meeting, Calgary, CA, July 2009. Wind Power Coordinating Committee.
[11] P. Meibom, R. Barth, H. Brand, B. Hasche, D. Swider, H. Ravn, and C.
Weber, “All Island Grid Study Workstream 2B: Wind Variability
Management Studies,” July 2007. Available: Mark O’Malley (F’07) received B.E. and Ph. D. degrees from University
http://www.uwig.org/Irish_All_Island_Grid_Study/Workstream_2B.pdf College Dublin in 1983 and 1987, respectively. He is the professor of
[12] “Wind Power Integration in Liberalised Electricity Markets Electrical Engineering in University College Dublin and is director of the
(WILMAR),” www.wilmar.risoe.dk Electricity Research Centre with research interests in power systems, grid
[13] R. Doherty and M. O’Malley, “New approach to quantify reserve integration of renewable energy, control theory and biomedical engineering.
demand in systems with significant installed wind capacity,” IEEE He is a fellow of the IEEE.
Trans. Power Syst., vol. 20, no. 2, pp. 587–595, May 2005.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO RIO DE JANEIRO. Downloaded on February 20,2024 at 02:56:44 UTC from IEEE Xplore. Restrictions apply.