350 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 1, FEBRUARY 2007
A Survey of Frequency and Voltage Control Ancillary
Services—Part I: Technical Features
Yann G. Rebours, Student Member, IEEE, Daniel S. Kirschen, Fellow, IEEE, Marc Trotignon, and
Sébastien Rossignol
Abstract—This two-part paper surveys the frequency and
voltage control ancillary services in power systems from various
parts of the world. In this first part, the nomenclature used to
describe active power reserves across 11 systems is first reviewed
in order to facilitate the comparison of frequency control ancillary
services. The essential technical features of frequency and voltage
control ancillary services are then described. Finally, the technical
requirements adopted in eight jurisdictions (North America,
continental Europe, Germany, France, Spain, the Netherlands,
Belgium, and Great Britain) are compared. The companion paper
surveys the economic features of these ancillary services.
Index Terms—Ancillary services, frequency control, reactive
power, spinning reserve, system services, voltage control.
I. INTRODUCTION
C
ONTROLLING frequency and voltage has always been an
essential part of operating a power system. However, since
the liberalization of the electricity supply industry, the resources
required to achieve this control have been treated as services that
the system operator has to obtain from other industry partici-
pants. Because this liberalization has proceeded independently
in different parts of the world and because of the structural dif-
ferences in the underlying power systems, the technical defini-
tions of these services and the rules governing their trading vary
considerably. The objectives of this paper and its companion
[1] are to survey these differences and provide a sound basis for
comparisons. This paper focuses on the definitions and technical
characteristics of the frequency and voltage control ancillary
services in the 11 power systems shown in Table I. This table
also shows the abbreviation used in this paper to refer to each
system, the name of the regulatory authority, and the name of the
transmission system operator (TSO). The companion paper sur-
veys the economic features of trading in these ancillary services.
II. N
OMENCLATURE
The profusion of terms used for ancillary services may lead to
some misunderstandings and confusion when one tries to com-
pare services from different power systems or jurisdictions. This
section establishes a common framework, which is then used to
classify the services considered in this survey. This framework
Manuscript received April 6, 2006; revised August 9, 2006. This work was
supported by Electricité de France (EDF). Paper no. TPWRS-00197-2006.
Y. G. Rebours and D. S. Kirschen are with the University of Manchester,
Manchester M60 1QD, U.K. (e-mail: yann.rebours@ieee.org; daniel.kirschen@
manchester.ac.uk).
M. Trotignon and S. Rossignol are with EDF Research and Development,
Clamart 92 141 Cedex, France (e-mail: marc.trotignon@edf.fr; sebastien.
rossignol@edf.fr).
Digital Object Identifier 10.1109/TPWRS.2006.888963
TABLE I
S
YSTEMS
INCLUDED IN THE
SURVEY OF TECHNICAL
FEATURES
is based on the one used by the Union for the Co-ordination
of Transmission of Electricity (UCTE), which is the associa-
tion of the TSOs operating within the synchronous system of
mainland Europe. The UCTE establishes the security and relia-
bility standards for this interconnected system. Its activities are
thus comparable to those of the North American Electric Relia-
bility Council (NERC). However, like NERC, the UCTE is not a
system operator and hence does not intervene in the operational
working of the system.
A. Frequency Control Services
Maintaining the frequency at its target value requires that the
active power produced and/or consumed be controlled to keep
the load and generation in balance. A certain amount of active
power, usually called frequency control reserve, is kept available
to perform this control. The positive frequency control reserve
designates the active power reserve used to compensate for a
drop in frequency. On the other hand, the deployment of nega-
tive frequency control reserve helps to decrease the frequency.
Three levels of controls are generally used to maintain this
balance between load and generation [3]–[8], [37]. Primary fre-
quency control is a local automatic control that adjusts the ac-
tive power generation of the generating units and the consump-
tion of controllable loads to restore quickly the balance between
load and generation and counteract frequency variations [3]–[5].
In particular, it is designed to stabilize the frequency following
large generation or load outages. It is thus indispensable for the
stability of the power system. All the generators that are lo-
cated in a synchronous zone and are fitted with a speed gov-
ernor perform this control automatically. The demand side also
participates in this control through the self-regulating effect of
frequency-sensitive loads such as induction motors [3], [4] or
the action of frequency-sensitive relays that disconnect or con-
nect some loads at given frequency thresholds. However, this
0885-8950/$25.00 © 2007 IEEE
REBOURS et al.: SURVEY OF FREQUENCY AND VOLTAGE CONTROL ANCILLARY SERVICES I 351
TABLE II
N
AMES OF THE
FREQUENCY
CONTROL RESERVES IN THE
SYSTEMS INCLUDED IN THE
SURVEY
demand-side contribution is not always taken into account in
the calculation of the primary frequency control response [37].
The provision of this primary control is subject to some con-
straints. Some generating units that increase their output in re-
sponse to a frequency drop cannot sustain this response for an
indefinite period of time. Their contribution must therefore be
replaced before it runs out. It is also important that the contribu-
tors to primary control be distributed across the interconnected
network to reduce unplanned power transits following a large
generation outage and enhance the security of the system. In
addition, a uniform repartition helps maintain the stability of is-
landed systems in case of a power system separation.
Secondary frequency control is a centralized automatic con-
trol that adjusts the active power production of the generating
units to restore the frequency and the interchanges with other
systems to their target values following an imbalance [6], [7]. In
other words, while primary control limits and stops frequency
excursions, secondary control brings the frequency back to its
target value. Only the generating units that are located in the area
where the imbalance originated should participate in this con-
trol as it is the responsibility of each area to maintain its load and
generation in balance. Note that loads usually do not participate
in secondary frequency controls. Contrary to primary frequency
control, frequency secondary control is not indispensable. This
control is thus not implemented in some power systems where
the frequency is regulated using only automatic primary and
manual tertiary control. However, secondary frequency control
is used in all large interconnected systems because manual con-
trol does not remove overloads on the tie lines quickly enough.
Within the UCTE, secondary frequency control is also called
load-frequency control (LFC) [37], while the term automatic
generation control (AGC) is preferred in North America [25].
However, within the UCTE, the term AGC designates the com-
bination of dispatching and secondary frequency control [37].
Lastly, synchronous time is controlled by adjusting the target
frequency used in the secondary frequency control [37].
Tertiary frequency control refers to manual changes in the
dispatching and commitment of generating units. This control
is used to restore the primary and secondary frequency control
reserves, to manage congestions in the transmission network,
and to bring the frequency and the interchanges back to their
target value when the secondary control is unable to perform this
last task. Some aspects of tertiary control relate to the trading of
energy for balancing purposes. This paper does not deal with
these aspects because they do not represent a service provided
to the TSO by the market participants but a mechanism for the
participants to balance their financial positions.
Table II compares the nomenclature used to describe the re-
serves associated with each type of frequency control in the sys-
tems included in this survey. The names are shown in the orig-
inal languages to avoid misinterpretation. In this table, for each
type of control, the reserves are ordered from the fastest (left)
to the slowest (right). Reserves within a given column are thus
similar. Within a system, terms are classified from the most gen-
352 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 1, FEBRUARY 2007
eral (top) to the most specific (bottom). A dashed line means
that positive and negative reserves have different names. For in-
stance, in Great Britain, the terms “primary response” and “sec-
ondary response” are used for the positive primary frequency
control reserve and “high frequency response” for the negative
one. In some cases (e.g., Australia), the name of a service may
be used as the corresponding reserve does not have a name.
A number of observations can be drawn from this table. First,
it is clear that care must be taken when comparing services
defined in different systems. For example, the terms “secondary
response” (GB), “secondary control reserve” (UCTE), and
“secondary reserve” (PJM) describe three completely different
services. Even the term “reserve” may lead to some misun-
derstandings. For example, in Australia, “reserve” designates
investments in generation capacity and not the frequency
control reserves as defined in this paper. “Spinning reserve” is
another example: California and PJM clearly use this term in
different ways. Their definitions are also different from the one
given by Wood and Wollenberg [4]: “the total synchronized
capacity, minus the losses and the load.” Indeed, the objectives
considered, which are operational (California, PJM) or sched-
uling (Wood and Wollenberg), are different.
Besides differences in terminology, there are also significant
differences in implementation. In Sweden, Great Britain, New
Zealand, and Australia, the primary frequency control reserves
have been divided in different categories. On the other hand, in
the other systems, a single reserve is defined for this type of con-
trol. This difference may be explained by the fact that smaller
systems are subject to larger frequency deviations than the large
interconnected systems of North America and mainland Eu-
rope. Reserve that can be used for very fast primary frequency
control reserve is therefore more valuable in these smaller sys-
tems, and making a distinction between different categories of
primary reserve is technically and commercially worthwhile.
Only one type of reserve is defined for performing secondary
frequency control, except in Sweden and Great Britain, where
secondary control is not used. Since the whole of the British
system is operated by a single TSO (and is not synchronous with
the Western European interconnected system), secondary con-
trol is not needed to correct deviations from interchange sched-
ules. While Sweden is interconnected with other countries, its
TSO can rely on numerous fast manual tertiary control offers
provided by the large hydro generation capacity. Tertiary fre-
quency control reserves do not lend themselves to easy com-
parison because TSOs have adopted widely different limits for
deployment times and various approaches to the treatment of
non-synchronized generating units.
B. Voltage Control Service
From a system perspective, the overall task of regulating
the voltage is sometimes organized into a three-level hierarchy
[3], [9]–[11], [37]. Primary voltage control is a local automatic
control that maintains the voltage at a given bus (at the stator
in the case of a generating unit) at its set point. Automatic
voltage regulators (AVRs) fulfill this task for generating units
[9]. Other controllable devices, such as static voltage compen-
sators, can also participate in this primary control. Secondary
voltage control is a centralized automatic control that coor-
dinates the actions of local regulators in order to manage the
injection of reactive power within a regional voltage zone. This
uncommon technology is used in France and Italy [10], [11].
Tertiary voltage control refers to the manual optimization of
the reactive power flows across the power system. In practice,
because of the close link between voltage and reactive power
in transmission networks, these three levels of control require
that participating devices be able to generate or absorb reactive
power.
From the perspective of providers of voltage control services,
it is convenient to divide the production of reactive power into a
basic and an enhanced reactive power service. The basic or com-
pulsory reactive power service encompasses the requirements
that generating units must fulfill to be connected to the network.
The enhanced reactive power service is a non-compulsory ser-
vice that is provided on top of the basic requirements. The ter-
minology of voltage control is much more uniform than for fre-
quency control and does not need to be discussed further.
C. System Services and Ancillary Services
It is sometimes useful to make a distinction between system
services and ancillary services. System services are the services
provided by the system operator to all users of the network,
while ancillary services are the services supplied by some of
the users of the network to the system operator [2]. To provide
its system services, the system operator usually buys ancillary
services from generators and consumers.
III. T
ECHNICAL FEATURES OF
ANCILLARY
SERVICES
This section first describes the main technical features that
the various ancillary services listed in the previous section must
have. It then compares the actual values required in the sys-
tems included in this survey. The comparison is centered on the
UCTE interconnected system to illustrate the differences that
can coexist within a single synchronous zone. The values for
Great Britain (which is not synchronized with the UCTE) and
NERC put this comparison into perspective. Only large gener-
ating units connected to the transmission network are consid-
ered here. Exceptions granted to small units and distributed gen-
erators are beyond the scope of this survey. The issues related
to the transmission of data to and from the provider of ancillary
services are not included in this survey either.
A. Common Technical Features
The most important technical parameters for frequency-re-
lated ancillary services are the deployment times. The maximum
amount of time that can elapse between the request from the
TSO and the beginning of the response by the service provider
will be called the “deployment start.”“Full availability” is the
maximum time that can elapse between the moment when the
provider receives the request and the moment at which it de-
livers its full response. Lastly, “deployment end” is the max-
imum amount of time during which the service must be provided
starting from the time of the request. Note that the framework
described in the previous section does not mention deployment
times because these times vary too much between systems.
The accuracy of the measurements is another important issue
because it affects the efficiency of the control and the payments
REBOURS et al.: SURVEY OF FREQUENCY AND VOLTAGE CONTROL ANCILLARY SERVICES I 353
TABLE III
T
ECHNICAL
COMPARISON OF
PRIMARY
FREQUENCY CONTROL
PARAMETERS IN
VARIOUS
SYSTEMS
to the providers. For example, if the instrumentation at a gener-
ating unit overestimates the frequency, its response to frequency
deviations will be inadequate, and the generating unit may be
paid more than what it deserves. However, it is generally in the
interest of producers to measure accurately so they can argue
more persuasively with the TSO in case of a dispute.
B. Technical Features of the Primary Frequency Control
For a steady-state frequency deviation
from the nominal
frequency
, a generator participating to the primary control
will change its generation by
. In the case of a generator,
note that frequency is generally not measured from the network
but through the rotation speed of the shaft. The droop
of this
generator, which is the gain of the feedback loop in the primary
frequency controller, is then defined as follows [37]:
(1)
where
is the nominal generator output power. A lower droop
increases the response of a unit but would cause more stress in
the generating unit as it would react more strongly to each devia-
tion. Moreover, a unit with a low droop is more likely to succeed
in switching to islanded mode in case of a major disturbance.
Adjusting the droop is not always easy because it often requires
that the plant be shut down.
The frequency deviation for which the entire primary fre-
quency reserve has to be deployed is also an important param-
eter. Indeed, using this information, the value of the droop, and
(1), one can calculate the maximal share of the nominal power
output that must be kept in reserve to provide the required pri-
mary frequency control power.
The frequency characteristic
of a control area is defined
as follows [37]:
(2)
where
is the actual power exchange from the zone to all
neighboring zones (a positive value represents an overall ex-
port) following a steady-state frequency deviation
from the
nominal frequency.
is the scheduled power exchange from
the zone to all neighboring zones (a positive value represents
exports). Therefore, the frequency characteristic represents
the total action of the primary frequency control provided
by generators and the self-regulating effect of the load. In
North America, the terms of “Beta” or “frequency governing
characteristic” are preferred to designate
[5].
The insensitivity of a primary controller is the frequency band
within which the controller does not change its output. Two in-
sensitivities should be distinguished: the non-intentional insen-
sitivity, which is intrinsic to the controller, and the intentional in-
sensitivity, which is added on purpose. In Europe, the non-inten-
tional insensitivity issometimes simply called insensitivity while
the intentional insensitivity is called dead band. The sum of these
two insensitivities gives the total insensitivity. If two generators
have different total insensitivities, it means that the one with the
smallest insensitivity will participate to the primary frequency
control before and probably more frequently than the other one.
Table III summarizes the primary frequency control param-
eters in eight different systems. Since units providing primary
354 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 1, FEBRUARY 2007
frequency control must respond immediately to a change in fre-
quency, the “deployment start” does not appears in this table.
Deployment times are the same all across the systems within
the UCTE because it is important to have a homogeneous re-
sponse in the synchronous zone. On the other hand, NERC does
not make any recommendations on the value of this parameter.
A faster primary frequency control is used in Great Britain be-
cause the size of this system makes it more susceptible to fre-
quency variations.
The frequency characteristic of the Eastern interconnection
of North America was around 31 000 MW/Hz in 2004 [5], [26].
The requirements for the national systems within the UCTE are
estimated from their annual electricity production [12] and the
total UCTE’s requirement (20 570 MW/Hz [29]). Lastly, the fre-
quency characteristic in Great Britain is estimated from require-
ments in the tendering process for primary frequency control
service [23].
Within the UCTE, the full deployment of the primary fre-
quency reserve must occur before a deviation of
mHz has
happened. Hence, TSOs with a tight requirement for the droop
(i.e., a small droop, such as in France) would need to reserve a
large headroom for their generating units if they want to provide
continuously their primary frequency control until
mHz.
However, in practice, units with a small droop attain their max-
imum before the limit of
mHz, so they do better than what
the UCTE recommends. In Belgium and Germany, no general
condition is applied to the droop as this parameter is agreed
between the generators and the TSO during the procurement
process. Lastly, the full deployment of the British primary fre-
quency reserve happens for larger deviations than in the UCTE
because the British power system has a smaller inertia.
Policies on controller insensitivity are similar in all systems,
except in Great Britain, where the requirement is less strict.
NERC has recently changed its policy and no longer recom-
mends any insensitivity. Before, a total insensitivity of 36 mHz
(30 mHz if a 50/60 coefficient is applied to account for the dif-
ference in nominal frequency) was required [24], which is three
times the UCTE’s requirements (10 mHz). However, the UCTE
requirement applies only to the controller [37], while NERC
former recommendation was for the “governor” [24]. This last
term is not formally defined, but in the common understanding,
it includes both primary frequency controller and actuators (e.g.,
jacks and valve). Therefore, in this table, the insensitivity of
the controller only and the combination of the controller and
the generator unit are not distinguished. Finally, the accuracy of
the frequency measurement is not addressed explicitly in most
systems.
C. Technical Features of the Secondary Frequency Control
Secondary frequency control can be organized in three
manners: centralized, pluralistic, or hierarchical [37]. In a
centralized organization, the control is performed by a single
controller for the whole control area. In a pluralistic organiza-
tion, the system is split into independent zones, each of which
having its own controller and regulating capacity. In a hierar-
chical organization, the organization is similar to the pluralistic
approach, but a main controller coordinates the action of all the
other controllers.
The area control error (ACE) of a zone is calculated as follows
according to the UCTE
or NERC
(3)
(4)
where
is the K-factor of the control area (in MW/Hz and
positive) and
the frequency bias setting (in MW/0.1 Hz and
negative).
and are an overestimate of the frequency char-
acteristic of the zone. An underestimate would indeed lead to a
conflict between the primary and secondary frequency controls.
is the measured value of the total power exchanged by the
zone with other zones (a positive value represents exports). Note
that
is slightly different from because of measurement
errors.
is the measured network frequency. is the target
frequency, which can differ from the nominal frequency when
controlling the synchronous time [37]. The additional term
introduces a very small correction factor. This factor compen-
sates the difference between the integration of the instantaneous
power exchanged and the demand’s energy measurements [25].
Secondary frequency control usually relies on a proportional
integral (PI) controller, filters, and heuristics to bring the ACE
back to zero. The exact configuration varies from a system to
another [6], [7], [32], [37], so no specific algorithm is given in
this paper.
Because of data management problems, the frequency and
the power flows through interconnections are measured with
discrete time steps. Therefore, the time steps of frequency and
power exchanges measurements, as well as the cycle time of the
controller, should be considered carefully. Shorter cycle times
help to get a more efficient secondary control but also imply
higher data management costs.
The speed of the effective frequency correction by the sec-
ondary control depends on the error size. Nevertheless, the fre-
quency normally should begin to come back to its target value
(neglecting the frequency oscillations) no later than the full de-
ployment of the primary frequency reserve. Indeed, at that time,
the balance between consumption and production should have
been re-established using the primary frequency control, and the
secondary frequency control should have started its action.
Table IV shows the parameters adopted for secondary fre-
quency control in seven different systems. Great Britain is not
included in this table because secondary frequency control is not
used in that country.
Deployment times in European countries are generally
smaller than what the UCTE recommends, probably to provide
some additional margin or to remain consistent with former
local policies. For its part, NERC does not give any specific di-
rect recommendation. However, it imposes some performance
criteria through its Disturbance Control Standard (DCS) and its
Control Performance Standard (CPS) [25]. To give an example
of U.S. practice, within PJM, the secondary frequency control
power should be deployed within five minutes and for at least
one hour [31].
With a requirement of 0.8 mHz (coefficient of 50/60 applied),
NERC recommends a more precise frequency measurement
