You are on page 1of 104

TECHNICAL REPORT

IEEE Power & Energy Society

Oct 2017 PES-TR-24

Measurement, Monitoring,
and Reliability Issues
Related to Primary
Governing Frequency
Response
PREPARED BY THE
Power System Dynamic Performance Committee
Power System Stability Controls Subcommittee
Task Force on Measurement, Monitoring and Reliability
Issues Related to Primary Governing Frequency
Response

© IEEE Oct 2017 The Institute of Electrical and Electronics Engineers, Inc.
No part of this publication may be reproduced in any form, in an electronic retrieval system or otherwise, without the prior written permission of the publisher.
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

THIS PAGE LEFT BLANK INTENTIONALLY

ii
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

TASK FORCE ON MEASUREMENT, MONITORING, AND


RELIABILITY ISSUES RELATED TO
PRIMARY GOVERNING FREQUENCY RESPONSE
Chair: Howard Illian Secretary: James Feltes
Members and Contributors

Brook Abegaz Hanna Gwan Sergay Pustarit


Baj Agrawal Les Hajagos Frank Puyleart
Aftab Alam Sam Hirsi D. Ramasubramanian
Eric Allen Shun-Hsien Huang Shruli Rao
Georgios Anagnostou Joseph Hurley Barry Rawn
P. Banerjee Chavdar Ivanov Jim Reilly
Mathieu Bergeron S. Karagiannopoulos Juan J. Sanchez-Gasca
Anjan Bose John Kehler Walter Sattinger
Roy Boyer Kyung Soo Kook John Schmall
Richard Bravo Dimitry Kosterev Dave Schooley
H. E. Brown Derek Kou Uwe Seeger
Neil Burbure Marc Langevin Marino Sforna
Claudio Canizares Pierre Laraviere Sanjiv Shah
Hao Chen Peijie Li Dinemayer Silva
Joe Chow Leonardo Lima Stefan Sterpu
Woojeeng Chsi Lin Liu Kurt Sullivan
Jose Conto Xiaochuan Luo Lee Taylor
Bob Cummings Hongtuo Ma Denden Tekeste
Sohom Datta Sujit Mandal Kjetil Uhlen
Donald Davies Gaetan Marcotte Edi Von Engeln
C. L. DeMarco Arturo Messina Petr Vorobev
Ruisheng Diao Edward Miska Costas Vournas
Matt Donnelly Jay Murphy David Vowles
Pengwei Du Ravi P. Mutukutti Yawei Wei
Reza Ebrahimian Ram Nath Anthony C. Williams
Eric Ela Hung Nguyen Zongyu Joanna Wu
Pavel Etingov Pramila Nirbhavane Steve Yang
Xinghao Fang Mohamed Osman Yang Yi
Bernardo Fernandes Jean-Nicolos Paquin Guangyuan Zhang
Vahan Gevorgian A Paranietharan Qing Zhang
Mike Gibbard Mahendra Patel Yu Zhang
Carlos Grande-Moran LeRoy Patterson Qing-Chang Zhong
Sven Granfors Les Pereira Yaming Zhu
Pouyan Pourbeik

iii
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

ACKNOWLEDGMENTS (Optional)

The Task Force is truly grateful for the support of our sponsoring subcommittee,
committee and Task Force members.
The Task Force gratefully acknowledges the principal authors/contributors (listed in
alphabetical order) of the following sections:
Section 1: Eric Allen, Carlos Grande-Moran, H. F. Illian, and Ravi P. Mutukutti.
Section 2: Aftab Alam, Eric Allen, Carlos Grande-Moran, H. F. Illian, S.
Karagiannopoulos, Ravi P. Mutukutti, Ram Nath, and D.
Ramasubramanian.
Section 3: Eric Allen, H. E. Brown, Carlos Grande-Moran, H. F. Illian, and Ravi
P. Mutukutti.
Section 4: Bob Cummings, H. F. Illian, Marino Sforna, and Stefan Sterpu.
Section 5: H. F. Illian, and D. Ramasubramanian.
Section 6: Eric Ela, and H. F. Illian.
Section 7: H. F. Illian, and Qing-Chang Zhong.
Section 8: H. F. Illian, and Stefan Sterpu.
Appendix A: H. F. Illian.
The Task Force gratefully acknowledges the contributions of Jeanie Illian for aid with
editing the final report.

KEYWORDS

Autonomous power systems, demand-side management, eigenvalues and eigenfunctions,


frequency measurement, primary governing frequency response, phasor measurement
units, power system economics, power system reliability, power system stability, systems
operation, virtual inertia, and virtual synchronous machines.

iv
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

CONTENTS

1 INTRODUCTION ......................................................................................................... 1
1.1 Primary Governing Frequency Response (PGFR) ............................................. 2
1.1.1 Primary Frequency Control – PGFR Illustration .......................................... 3
1.2 Summary ............................................................................................................ 8
2 Reliability Studies, Results, & Measurement Methods ................................................ 9
2.1 Adequacy of Present Levels of PGFR ................................................................ 9
2.1.1 Adequacy of Present Levels of PGFR in North America ............................. 9
2.2 Distribution of PGFR .......................................................................................... 9
2.2.1 European Study ......................................................................................... 10
2.2.2 North American Western Interconnection (WI) Study ................................ 13
2.3 PGFR Withdrawal ............................................................................................. 17
2.4 Recommended Methods to Measure PGFR .................................................... 22
2.4.1 Settled PGFR as a PGFR Measure ........................................................... 22
2.4.2 Arrested PGFR as a PGFR Measure ........................................................ 23
2.4.3 Comparing Settled and Arrested PGFR .................................................... 24
2.5 Summary .......................................................................................................... 25
3 PGFR Components ................................................................................................... 26
3.1 Governor Droop, Deadband, Time Delays & Dynamic Characteristics ............ 27
3.1.1 Governor Droop ......................................................................................... 27
3.1.2 Governor Deadband .................................................................................. 27
3.1.3 Governor Time Delays............................................................................... 28
3.1.4 Dynamic Characteristics ............................................................................ 30
3.2 Balancing Inertia, Kinetic and Synthetic ........................................................... 31
3.2.1 Kinetic Balancing Inertia ............................................................................ 31
3.2.2 Synthetic Balancing Inertia ........................................................................ 32
3.3 Load Damping Synthetic PGFR ....................................................................... 36
3.4 Synthetic PGFR ................................................................................................ 37
3.4.1 Synthetic Step PGFR – Manual Restoration ............................................. 37
3.4.2 Synthetic Step PGFR – Time Delay Restoration ....................................... 37
3.4.3 Synthetic Step PGFR – Frequency Based Restoration ............................. 37
3.4.4 Synthetic Continuous PGFR – Automatic Restoration .............................. 38

v
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

3.5 Summary .......................................................................................................... 39


4 PGFR for Reliability Standards ................................................................................. 40
4.1.1 Requirements for PGFR, Reserves and Backdown Margin ...................... 40
4.1.2 ENTSO-E Reliability Standards ................................................................. 44
4.2 Reliability Standard Design Based on Settled PGFR ....................................... 44
4.2.1 ERCOT Reliability Standard ...................................................................... 45
4.2.2 NERC Reliability Standard ........................................................................ 45
4.3 Reliability Standard Design Based on Arrested PGFR .................................... 47
4.4 Summary .......................................................................................................... 47
5 Economics Associated with PGFR ............................................................................ 48
5.1 PGFR Economics ............................................................................................. 48
5.1.1 PGFR Costs – Supply Side ....................................................................... 48
5.1.2 PGFR Costs – Demand Side..................................................................... 50
5.1.3 Requirements for PGFR ............................................................................ 50
5.1.4 Economic Cost of Reserves for PGFR ...................................................... 53
6 Incentive Design for PGFR Provision ........................................................................ 55
6.1 Existing Compensation Schemes for PGFR .................................................... 55
6.2 Industry Proposals for PGFR Compensation ................................................... 56
6.3 Potential Options for PGFR Sufficiency and Compensation ............................ 58
6.3.1 PGFR Capability Individual Resource Mandate ........................................ 58
6.3.2 Energy Market Design and Elimination of Off-basepoint Penalties ........... 59
6.3.3 Requiring PGFR Capability in other Ancillary Services ............................. 59
6.3.4 Market Design for PGFR Capability .......................................................... 60
6.3.5 Market Design for PGFR Capability Based on Settled PGFR ................... 60
6.3.6 Including Other PGFR Characteristics as Part of the Market Design ........ 62
6.3.7 Including non-traditional resources in the PGFR market design ............... 63
6.3.8 Market Design Based on Arrested PGFR .................................................. 64
6.3.9 Monitoring of Performance ........................................................................ 65
6.4 Summary and Best Practices ........................................................................... 65
7 Developing a Path to the Future ............................................................................... 66
7.1 Paradigm Change of Power Systems .............................................................. 66
7.2 Fundamental Challenge Faced by Power Systems ......................................... 67
7.3 Unified Architecture for Future Power Systems ............................................... 67
7.4 PGFR in Future Power Systems ...................................................................... 69
7.4.1 PGFR Contributions from both Supply and Load ...................................... 69

vi
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

7.4.2 Droop ......................................................................................................... 70


7.4.3 Fast Action Without Delay ......................................................................... 70
7.4.4 Reconfigurable Virtual Inertia .................................................................... 70
7.4.5 Continuous PGFR ..................................................................................... 70
7.5 Summary ......................................................................................................... 71
8 Conclusions and Recommendations ......................................................................... 72
8.1 Conclusions ...................................................................................................... 72
8.2 Recommendations ........................................................................................... 75
Appendix A – Market Designs for PGFR......................................................................... 77
A.1. Market Design Based on Settled PGFR ........................................................... 77
A.1.1. Nomenclature .............................................................................................. 77
A.1.2. Market Scheduling ....................................................................................... 80
A.1.3. Market Pricing.............................................................................................. 82
A.1.4. Bidding Structure and Settlements .............................................................. 84
A.2. Market Design Based on Settled PGFR Including Reliability Changes ............ 85
A.2.1. Nomenclature Including Reliability Changes ............................................... 86
A.2.2. Market Scheduling Including Reliability Changes ........................................ 87
A.2.3. Market Pricing Including Reliability Changes .............................................. 88
A.2.4. Bidding Structure and Settlements Including Reliability Changes ............... 88
A.3. PGFR Market Design Based on Arrested PGFR ............................................. 88
A.3.1. Nomenclature Based on Arrested PGFR .................................................... 89
A.3.2. Market Scheduling Based on Arrested PGFR ............................................. 90
A.3.3. Market Pricing Based on Arrested PGFR .................................................... 93
A.3.4. Bidding Structure and Settlements Based on Arrested PGFR .................... 94
A.4. Summary .......................................................................................................... 96

vii
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

THIS PAGE LEFT BLANK INTENTIONALLY

viii
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

1 INTRODUCTION
This Task Force recommends ways to resolve the issues identified in the IEEE Special
Publication 07TP180 “Interconnected Power System Response to Generation Governing:
Present Practice and Outstanding Concerns”, May 2007, as they relate to primary
frequency regulation. The report viewed Primary Governing Frequency Response
(PGFR) primarily as a service provided by traditional generation. The word “governing”
in PGFR refers to speed governing as opposed to the generator governor. Therefore, the
scope includes the following issues raised by that report as listed below:
(1) Technical analysis should confirm the current security adequacy of PGFR
assumed by expert judgment.
(2) “If some portions of an interconnection require greater PGFR due to their
topology and risk of separation, how does this greater PGFR affect
reliability when the interconnection experiences a disturbance and remains
intact? Can imbalances in PGFR result in increased risk during
disturbances?”
(3) The TF should make recommendations on possible ways of performing on-
line monitoring to estimate the governor steady state droop and response of
turbine-generator governor action.
Penetration of non-traditional generation resources and the development of synthetic
components as contributors to PGFR expand the scope of this report to include
consideration of additional issues not included in the initial report including:
(4) Understanding governor droop settings;
(5) Understanding the effect of governor deadband settings;
(6) Understanding the effect of time delays in governor response;
(7) Maintaining sufficient balancing inertial response to assure reliability;
(8) Limitations of synthetic balancing inertia based on frequency deviation;
(9) Limitations of synthetic balancing inertial response proportional to rate of
change of frequency;
(10) Limitations of step load interruption based on frequency without restoration;
(11) Limitations of step load interruption based on frequency with restoration
based on time delay;
(12) Limitations of step load interruption based on frequency with restoration
based on frequency recovery;

1
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

(13) Limitations of synthetic continuous PGFR proportional to frequency error;


(14) Understanding the impact of PGFR withdrawal;
(15) Providing reliability standards using arrested PGFR to replace settled PGFR;
(16) Assuring reliability by including balancing inertial response and settled
PGFR and as an alternative arrested PGFR in market designs;
(17) What are the costs and benefits of having demand-side participation?
(18) Would demand-side participation contribute to more reliable systems?
(19) Economics of synthetic balancing inertial PGFR and synthetic PGFR;
(20) How would loads be rewarded for providing this service?
(21) What types of load could provide primary governing frequency response?
(22) What work needs to be completed to enable demand-side participation?
(23) How would demand-side participation affect interconnection modeling?
(24) How would the provision of this service be measured? and,
(25) Developing a feasible path to future system operations.
Before addressing the issues raised above, an in-depth review of PGFR follows.

1.1 Primary Governing Frequency Response (PGFR)1


There are three types of overlapping frequency control:
Primary Frequency Control (PGFR) is any actions locally provided to arrest and
stabilize frequency in response to a frequency deviation. Primary control comes from
automatic generator governor response (also known as speed regulation), load response
(Load Damping) typically from motors and other devices providing an immediate
response based on device-level, frequency responsive control or device characteristics.
Secondary Frequency Control is any action provided by an individual control area
(CA), Balancing Authority2 (BA) or its reserve sharing group to correct the resource-load
imbalance causing a frequency deviation, restoring both scheduled frequency and
primary frequency responsive reserves. Secondary control comes from either manual or
automated dispatch from a centralized control system to correct frequency error.

1
Adapted from comments by H. F. Illian to Federal Energy Regulatory Commission, Docket No. RM16-
6-000, Essential Reliability Services and the Evolving Bulk Power System – Primary Frequency
Response, April 25, 2016.
2
The North American Reliability Corporation has redefined control areas as Balancing Authorities. The
term control area is used throughout this document.
2
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Tertiary Frequency Control is any action provided by control areas on a balanced basis
that are coordinated so there is a net zero effect on Area Control Error (ACE). Examples
of tertiary control include the dispatching of generation to serve native load, economic
dispatch, to affect interchange, and the re-dispatching of generation. Tertiary control
actions are intended to restore secondary control reserves by reconfiguring reserves.
These terms are used in both North America and Europe. There is a slight difference in
definitions between these two regions. Secondary frequency control in Europe includes
only AGC and excludes manual control to restore frequency to schedule, while secondary
frequency control in North America includes manual control to restore frequency to
schedule. The North American definition is used throughout this report, because it is
technology neutral and based only on the result accomplished not on the technology used.
1.1.1 Primary Frequency Control – PGFR Illustration3
PGFR is the first stage of overall frequency control. It is the response of resources and
load to a locally sensed change in frequency that arrest the change in frequency. PGFR is
autonomous, is not driven centrally, and begins immediately (typically within seconds).
Local governor response and active power change due to frequency dependent loads are
typical examples. Different resources and loads provide PGFR with differing response
times, based on system conditions, and their respective resource/load mix. The
discussion of PGFR is typically presented in the context of a large generator, because this
loss is much more likely than a sudden loss of an equivalent amount of load. PGFR
includes many components. Understanding PGFR requires an understanding of each
component and how that component contributes to the total response. The following
series of graphs show how these components contribute to the total PGFR.
Fig. 1 presents a sudden loss of generation of 1000 MW. Although a large event of 1000
MW illustrates the response components, even small events will result in similar
responses. The magnitude of the event only affects the shape of the curves on the graph;
it does not eliminate the need for PGFR. The graph presents the frequency and PGFR
components relative to the horizontal time axis in seconds. This simplified example
assumes a disturbance event, the sudden loss of 1000 MW of generation resulting from a
breaker trip instantaneously removing 1000 MW from the interconnection. The power
deficit line, shown in black, using the MW scale on the left illustrates this. The
frequency line shown in red using the Hertz scale on the right illustrates interconnection
frequency, assuming starting frequency is 60 Hz. Even though the generation tripped and
power injected by it removed from the interconnection, loads continue to use the same
amount of power as before the generation was removed. The “Law of Conservation of
Energy” requires the 1000 MW must be supplied as power, first derivative of energy,
dE/dt. The system extracts 1000 MW of power from kinetic energy stored in the rotating
mass of all synchronized generators and motors on the interconnection with this
equipment acting as a giant flywheel.

3
Adapted from slides by H. F. Illian for a NERC Webinar presenting the NERC Reliability Standard
BAL-003-1.1 Frequency Response and Frequency Bias Setting to the North American Electric Power
Industry, May 22 & 24, 2012.
3
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

3000 60.100

2500 60.000
Power Deficit

2000 Frequency 59.900


Change in Power (MW)

1500 59.800

Frequency (Hz)
1000 59.700

500 59.600

0 59.500

-500 59.400
-10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60
TIme (Seconds)

Fig. 1. Primary Frequency Control – PGFR – 1

The kinetic energy extracted from the interconnection energy provides the “balancing
inertia” (expressed in MW) to supply the required power (rate of change of energy) to
maintain the power balance. The term “balancing inertia” is coined from the terms
“balancing energy” (a common market term) and “inertial response”. Balancing power
(rate of change of energy) describes the power used to restore the system active power
balance. Inertial response describes the power supplied from the kinetic energy stored in
the rotating mass of both generators and motors synchronized to the system.
This consumption of “balancing inertia” causes the speed of the rotating equipment to
decline, reducing the frequency. Fig. 2 shows this with orange dots representing the
balancing inertia power that exactly offsets the power deficit and change in system losses.
As the frequency decreases, synchronized and frequency dependent motors slow, and the
work they are providing declines resulting in a decrease in load called “load damping.”
This load damping (load reduction) is the main reason for the initial decline in power
deficit. Only synchronously operated and frequency dependent motors contribute to the
load damping. Variable speed drives decoupled from the interconnection frequency do
not contribute to load damping. Any load that does not change with interconnection
frequency including resistive load does not contribute to the load damping. The power
deficit exactly equals the balancing inertia plus the load damping contribution thus
indicating that there is no power or energy imbalance at any time during this process.

4
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

3000 60.100

Power Deficit
2500 60.000
Balancing Inertia

Load Damping
2000 59.900
Frequency

1500 59.800
Power (MW)

Frequency (Hz)
1000 59.700

500 59.600

0 59.500

-500 59.400
-10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60
TIme (Seconds)

Fig. 2. Primary Frequency Control – PGFR – 2

What is normally considered as “balancing power or energy” is specifically the power or


energy required to correct the frequency error. The balancing inertia power or energy
extracted from the interconnection instantaneously corrects any apparent power or energy
imbalance. The balancing function is a frequency control function described as a
balancing function because ACE is calculated in MW instead of Hertz.
During the initial seconds of the disturbance, the governors do not respond until the
frequency decline is detected and processed in various elements in the governor and
prime mover. The blue line on Fig. 3 showing governor response illustrates this. This
time delay results from the time it takes for frequency deviation to surpass the governor
dead band and the controller to adjust the equipment. This allows the working fluid mass
to flow from the source of the energy (main steam control valve for steam turbines, the
combustor for gas turbines, or the gate valve for hydro turbines) to the turbine-generator
blades, where the power is converted to mechanical energy and then to electricity in the
generator. The frequency continues to decline due to the ongoing extraction of balancing
inertia power and energy from rotating turbine-generators and synchronous and
frequency dependent motors on the interconnection. The reduction due to load damping
continues as load and frequency declines. During this time delay before the governor
active power response begins, the balancing inertia limits the rate of change of frequency.

5
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

3000 60.100

Power Deficit
2500 60.000
Balancing Inertia

Load Damping

Governor Response
2000 59.900
Frequency

1500 59.800
Power (MW)

Frequency (Hz)
1000 59.700

500 59.600

0 59.500

-500 59.400
-10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60
TIme (Seconds)

Fig. 3. Primary Frequency Control – PGFR – 3


3000 60.100

Power Deficit
2500 60.000
Balancing Inertia

Load Damping

Governor Response
2000 59.900
Frequency

1500 59.800
Power (MW)

Frequency (Hz)
Point A

1000 59.700

500 59.600

0 59.500

-500 59.400
-10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60
TIme (Seconds)

Fig. 4. Primary Frequency Control – PGFR – 4

6
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

After a short time-delay, the governor response begins to increase the active power in
response to the initial rapid decline in frequency, as illustrated on Fig. 4. The governor
response exactly offsets the load damping reduced power deficit at the point in time that
the frequency decline is arrested called the frequency nadir (Point C). At this point in
time, the governor response replaces the balancing inertia completing the balancing
inertia’s contribution to reliability and reducing its power contribution to zero. Reducing
the time delay associated with the delivery of governor response also reduces the amount
of balancing inertia required to limit the change in frequency for the disturbance event.
This supports the conclusion that balancing inertia manages the time delays associated
with the delivery of PGFR. Not only is the rapid delivery of PGFR important, shortening
the time delay associated with its delivery is also important.
Therefore, two important components of PGFR are related to the length of the time delay
before the initial delivery of PGFR begins and how much PGFR is delivered. The length
of the time delay depends on parameters such as deadband, dead time, governor tuning,
sluggishness, time constants of governor and prime mover components etc. The
magnitude of delivered response depends on the droop setting of individual governors
and available frequency responsive reserve at the operating point. This “Point C”, at
which the frequency is first arrested, is the value used to calculate arrested PGFR. This is
the nadir or minimum frequency (maximum frequency for load loss events) during a
disturbance event. The change in MW between the A-Value MW and MW at the nadir
divided by the change in frequency between A-Value frequency and the nadir frequency
is the “arrested PGFR”.
3000 60.100

Power Deficit
2500 60.000
Balancing Inertia

Load Damping

Governor Response
2000 59.900
Frequency

1500 59.800
Power (MW)

Frequency (Hz)

Point A

1000 59.700

500 59.600

0 59.500

-500 59.400
-10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60
TIme (Seconds)

Fig. 5. Primary Frequency Control – PGFR – 5


7
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Fig. 5 shows the post-disturbance transient period. The governors continue to respond
after arresting the frequency decline due to the time delay in their response. This results
in the frequency partially recovering from the minimum arrested value and sometimes
results in an oscillating transient after the minimum (arrested) frequency until it settles
about 20 seconds after the disturbance. Note that a permanent frequency error remains in
the system because of the proportional control characteristic of the speed governors.
3000 60.100

Power Deficit Balancing Inertia


2500 60.000
Load Damping Governor Response

A-value B-value

2000 Frequency 59.900

Frequency (Hz)
1500 59.800
Power (MW)

A-Value Averaging Period Point A B-Value Averaging Period

1000 59.700

500 59.600

0 59.500

-500 59.400
-20 -15 -10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60
TIme (Seconds)

Fig. 6. Primary Frequency Control – PGFR – 6

Fig. 6 shows the total disturbance event. Frequency and power contributions stabilize
after the transient period. “settled PGFR” is the best estimator for the “Frequency Bias
Setting,” used in the ACE equation. NERC BAL-003-1.1 – Frequency Response and
Frequency Bias Setting reliability standard currently uses the settled PGFR. The pre-
disturbance (A-Value) is the average power or average frequency for an averaging period
prior to a disturbance. The post disturbance (B-Value) is the average power or average
frequency for an averaging period after the disturbance transient settles. Adequate
reliability requires minimum frequency, the C-Value, to remain above under-frequency
relay settings so as not to trip firm load.

1.2 Summary
This section presents the basics of primary frequency control. The following sections use
these basics to build an in-depth understanding of PGFR. This report provides a solid
basis to move the electric industry forward as non-synchronous resources with little or no
inertia comprise larger proportions of the interconnections.
8
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

2 Reliability Studies, Results, & Measurement Methods


Section 2 presents a set of case studies and detailed accounts of experience of various
investigations from around the world on the PGFR performance and estimates of the
reliability effects of current practices and measurement methods in the present day
interconnected power systems. At present, under-frequency load shedding is the only
hazard identified with frequency deviations across an entire interconnection. Other
hazards, such as damage to rotating equipment, are regarded as negligible for frequency
deviations that are less than the difference between scheduled frequency and the highest
UFLS threshold. Consideration of PFGR for islanding conditions is beyond the scope of
this report.

2.1 Adequacy of Present Levels of PGFR


This section confirms by technical analysis the risk to security, previously assumed by
expert judgment, to be adequate for present levels of PGFR.
2.1.1 Adequacy of Present Levels of PGFR in North America4
Dynamics simulations of the Eastern, Western, and Texas Interconnections for the
recommended Interconnection Frequency Response Obligations (IFRO) showed levels of
PGFR to be adequate to avoid tripping of the first stage of the interconnection under-
frequency load shedding (UFLS) systems. All three of these analyses use light‐load
cases.

NERC performed off‐peak dynamics analysis of the recommended IFROs for the
Eastern, Western, and Texas Interconnections to determine if those levels of PGFR are
adequate to avoid tripping of the first stage of regionally approved UFLS systems in the
interconnection. Each of the interconnections prepared light load cases as the starting
root case for the analyses. In each case, the dynamic governor or load responses of the
PGFR of the interconnection closely matched the recommended IFRO value for the
prescribed resource loss. All three simulations did not model automatic generation
control (AGC), which typically starts to influence PGFR in the 30‐45 second timeframe.

In all three interconnections analyzed, frequency remained above the highest UFLS set
point even with Interconnection PGFR degraded to the IFRO value. The 2016 Frequency
Response Annual Analysis presents further analysis of each interconnection.

2.2 Distribution of PGFR


Section 2.1 answers the following question, “If some portions of an interconnection
require greater PGFR due to their topology and risk of separation, how does this greater

4
Adapted from 2016 Frequency Response Annual Analysis, Prepared jointly by System Analysis and
Performance Analysis, NERC, September 2016,
http://www.nerc.com/comm/oc/documents/2016_fraa_report_2016-09-30.pdf.

9
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

PGFR affect reliability when the interconnection experiences a disturbance and remains
intact? Can imbalances in PGFR result in increased risk during disturbances?”
2.2.1 European Study5
In the European power system, the current total primary control reserve is 3000 MW
(reference incident), and fully activated in case of a frequency excursion of ±200 mHz.
The European network of transmission system operators for electricity (ENTSO-E)
annually determine the required size of primary control contributions for each control
area (CA) based on the annual net electricity production of each area. ENTSO-E
determines the primary control requirements and the reallocation possibilities based on
the concept of proportional reserve allocation for each area within the network, rather
than technical considerations. However, there is an increasing incentive for many
countries to initiate ancillary service trading activities with other countries to reduce their
procurement costs. Therefore, investigating reallocation strategies of Primary Control
Reserves6 (PCRs) considering power system security is of great interest. Due to the PCR
nature, “reallocation” is a transfer of the PCR location based on an agreement between
two parties; however, PCR activation should not lead to a physical exchange of energy
between these parties.
Optimal reserve allocation is the subject of several studies in the literature. Previous
work focuses mainly on small test systems with very limited research on large scale
power systems. In this case, computationally expensive optimization-based approaches
are likely to be intractable using the nonlinear AC power flow equations. Network
reduction techniques could be used to reduce the size of the network, but constraint
satisfaction cannot be guaranteed for all lines, and loop flows cannot be captured.
This study proposes a simulation-based method to identify PCR reallocation strategies of
interest from a technical perspective in large interconnected power systems. The basis of
this method is N-1 security assessment in case of generator outages, on AC power flow
analysis, and considers the variability of Renewable Energy Sources (RES) by applying
copula theory. Furthermore, the design is tractable for realistic network sizes up to
several thousands of buses and lines, and with large RES shares.
Fig. 7 & 8 summarize the proposed method for PCR reallocation in large power systems.
First, apply copula theory on a large set of official weather data generating a reasonable
number of synthetic weather data. Second, convert the synthetic weather data into RES
in-feed scenarios, and then combine them with real network snapshots. The resulting
snapshots represent different grid states in terms of topology, conventional
generation/load patterns and RES in-feeds. Third, perform an N-1 security assessment on
different PCR allocations using the resulting snapshots to look at the post-contingency

5
S. Karagiannopoulos, E. Vrettos, P. Centeno Lopez, M. Vrakopoulou, F. Oldewurtel, G. Andersson and
M. Zima, "On Geographical Allocation of Primary Frequency Control Reserves in Large Interconnected
Power Systems," in Power Systems Computation Conference (PSCC), Wroclaw, Poland, Aug 2014.
6
Primary Control Reserves is another name for Primary Governing Frequency Response reserves.

10
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

line loadings. Analyze these results to evaluate the possible PCR allocations and suggest
the ones preferable from a technical point of view.

Fig. 7. Flowchart of the proposed methodology for reallocation of primary reserve

This method focuses on pairwise PCR reallocations, and excludes the splitting the reserve
requirements among more than two control areas. The proposed procedure is tractable
even for large networks for two reasons: (a) the number of synthetic snapshots can be
arbitrarily chosen depending on the network size, and (b) the simulations for each
snapshot and each PCR allocation are independent and comparable.

Fig. 8. Procedure for pairwise reallocation of primary control reserves.

This European Study explains the procedure to derive the final synthetic snapshots to be
examined in terms of PCR reallocation. Fig 8 illustrates the last step of the method,
defines the candidate PCR allocation strategies, and fixes generators with active primary
controllers. This step changes the different allocations with respect to their PCR
contributions. Following a disturbance, if a new equilibrium point can be reached after
the primary control is activated, security assessment is disregarded. This approach
focuses on steady-state analysis while neglecting the dynamics of the grid. Furthermore,
the analysis does not consider network splitting constraints. Finally, the evaluation
11
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

process identifies the best PCR allocation among the candidates. First, the method
assumes the current PCR allocation (PCR-alloc-0) secure, and uses it as a reference to
indicate whether other allocations are also secure. This means that, if another allocation
(PCR-alloc-k) leads to a post-contingency overloading not present in PCR-alloc-0, then
PCR-alloc-k is unsecure and is discarded. The final step chooses the best allocation
among the secure allocations using the maximum and the average post-contingency
loading values for each element of the network and for each contingency.
The proposed methodology for PCR reallocation uses the European interconnected power
system as a case study. The authors first identify an empirical upper bound on the PCR
that can be reallocated among all countries in ENTSO-E. This is done via simulations
using the four grid snapshots that became available by the ENTSO-E, but without
considering the synthetic RES in-feed scenarios. Although this approach is quite simple,
it still provides useful insight. The second part focuses explicitly on three pairs of CAs in
the European network, highlighting the importance of the geographic position of the two
parties of a PCR reallocation agreement. MATPOWER calculated the AC power flows.
Parallel computing decreased the computational time up to 85%.

Fig. 9. Maximum reserve reallocation between any two control areas, based on
the ENTSO-E winter peak snapshot. The colormap expresses the reallocated
reserve as a percentage of the 2012 reserve requirement of each control area.

Fig. 9 presents the results for the winter 2013 off-peak snapshot. It identifies several
areas that cannot reduce their reserves without causing additional line over-loadings in
case of a generator contingency. Those are CAs with high PCR requirements such as
Germany, France, Spain and Italy, but also Czech Republic, Denmark, Netherlands, and
12
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Portugal. On the other hand, CAs located in central Europe such as Switzerland,
Germany, Belgium, and the Netherlands can host more reserves coming from most of the
other CAs without endangering the network.
According to these results, there are CAs that could reduce their PCR requirements, and
others that could increase them, while maintaining system security. For example,
Switzerland seems to be an attractive option for hosting PCRs from other CAs, favoring
neighboring CAs. Interestingly, certain reallocation strategies might be beneficial not
only for the involved CAs, by reducing their PCR procurement costs, but also for the
entire system by reducing the post-contingency line loadings.
2.2.2 North American Western Interconnection (WI) Study
NERC determined the WI’s IFRO based on the largest potential generation loss, the loss
of two Palo Verde generating units (2,626 MW). NERC allocates this IFRO to each CA
in the interconnection based on each CA’s portion of the interconnection’s annual
generation and annual load. Each CA must plan on having an adequate amount of PGFR
capability available to respond to actual frequency events. Given the WI’s surplus
PGFR, CAs may enter bilateral agreements to re-allocate obligation based on an expected
PGFR performance deficiency or surplus while continuing to support grid reliability.
This study7 uses the California Independent System Operator (CAISO) CA as an example
to explore the primary question of whether imbalances in PGFR may result in an
increased risk during disturbances. Traditionally, CAs have not explicitly procured
PGFR capability. CAs in the WI assumed it was provided through adherence to the WI
standard for governor droop settings and the automatic governor response of
interconnected generating resources. However, the compliance calculation method from
the new standard measures the actual response of generators and load. Therefore, any
solution adopted by the CA must ensure obligated units respond appropriately. To insure
compliance, various CAs such as the CAISO have provisions to procure transferred
PGFR from external CAs along with tariff provisions that include requirements for
generators to ensure governor operation. Longer term approaches in CAs, such as
CAISO, consider potential efficiency improvements from implementing more
comprehensive market solutions. Implementing a reserve procurement constraint or new-
market products, defined in a technology neutral way, allows all certified resources,
including non-conventional resources, to have PGFR capability. Procuring transferred
PGFR from external CAs can lead to imbalances in PGFR. The following studies seek to
determine if these imbalances in PGFR can lead to increased reliability risks during
disturbances.
2.2.2.1 Methodology
A real-time snapshot from CAISO’s online transient stability assessment tool analyzed
the following four scenarios to determine the impact of imbalances in PGFR:

7
A. Alam, H. F. Illian, D. Ramasubramanian, R. Nath, U.S. Western Interconnection Study, previously
unpublished.
13
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Archive Scenario: The Archive Scenario, retrieved from the archived snapshots,
ensured that a no-fault run produces a relatively flat frequency for about 60 seconds of
simulation. The scenario then tested for the loss of 1300 MW of generation in Arizona
and the loss of 1200 MW of generation within California to determine the normal
frequency plots and PGFR.
Blocked Scenario: The Blocked Scenario blocked all the governors within California
to determine if there would be any stability impacts when PGFR was completely blocked
in California and not transferred to external entities. The blocking of PGFR was
achieved by disabling the governor models in the input dynamic data for all machines in
California.
Transfer Scenario: The Blocked Scenario was extended to create a Transfer
Scenario. In the Transfer Scenario, all the PGFR within California remained blocked.
Turning on additional units at 0 MW output and ensuring that they have active governors
established the transfer of PGFR. Turning on additional units until a significant
improvement in the frequency performance of each area was observed compared to the
Blocked Scenario created the Transfer Scenario.
Balanced Scenario: For a true comparison of the impact of transfer of PGFR on
reliability, the Blocked Scenario was modified. Within California governors were
unblocked, until the MW and the MW/Hz response of the entire WI for the contingencies
under consideration was comparable with the Transfer Scenario.
The study evaluated each of the above scenarios for the following two contingencies:
(1) 1300 MW Loss of Generation in Arizona, and
(2) 1200 MW Loss of Generation in California.
The methodologies to calculate the necessary values for each combination of scenario
and contingency follow:
(1) Evaluate each scenario to assure no negative stability impacts occurred due to the
contingencies;
(2) Determine the total MW PGFR of each unit in the WI by:
a. Calculating the average output of each individual unit before T0, where T0
is the time instance where generation is tripped,
b. Calculating the average output of each individual unit from T0+20
seconds to T0+52 seconds, and
c. Calculating the difference between the average output of each individual
unit from T0+20 seconds to T0+52 seconds and the average output of each
individual unit before T0, the unit output prior to the disturbance;

14
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

(3) Determine for each combination of scenario and contingency, the total MW
response of each area by summing the MW PGFR of each unit belonging to the
individual areas in the WI;
(4) Determine the total MW PGFR of the entire interconnection by summing the MW
PGFR for each area;
(5) Determine the MW/Hz PGFR of each area and the interconnection using the
frequency of each area in the WI by:
a. Calculating the difference between the average frequency from T0+20
seconds to T0+52 seconds and the frequency prior to T0; and
b. Calculating the MW/Hz PGFR for each area by dividing the total MW
PGFR of each area by the frequency difference for each area;
(6) Determine the MW/Hz response for the WI by summing the individual MW/Hz
responses for each area.
Fig. 11 and 12 show the PGFR in MW/Hz for all scenarios for the 1300 MW loss in
Arizona and the 1200 MW loss in California respectively.
20,000 20,000

15,000 15,000
PGFR (MW/Hz)

PGFR (MW/Hz)

10,000 10,000
MW/Hz Archived MW/Hz Archived
MW/Hz Blocked MW/Hz Blocked
MW/Hz Transferred MW/Hz Transferred
5,000 5,000
MW/Hz Balanced MW/Hz Balanced

0 0

Fig. 11. PGFR for 1300 MW Loss Fig. 12. PGFR for 1200 MW Loss in
in Arizona. California.

2.2.2.2 Results
Fig. 13 and 14 show the PGFR characteristic for the “archive scenarios” for the 1300
MW loss in Arizona and the 1200 MW loss in California respectively. The scenarios
demonstrate no negative stability impacts due to the contingencies. The WI regions vary
with all demonstrating reasonable variability.
Fig. 15 and 16 show the PGFR characteristic for the “blocked scenarios” for the 1300
MW loss in Arizona and the 1200 MW loss in California respectively.
Fig. 17 and 18 show the PGFR characteristic for the “transfer scenarios” for the 1300
MW loss in Arizona and the 1200 MW loss in California.
15
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

60.05 60.05

60.00 60.00

59.95 59.95
Frequency (Hz)

Frequency (Hz)
59.90 59.90

59.85 59.85
BPA BPA
BC Hydro BC Hydro
59.80 SCE 59.80 SCE
APS APS
59.75 PGAE 59.75 PGAE
MONTANA MONTANA
Mean Mean
59.70 59.70
0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5 0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5
Time (Seconds) Time (Seconds)

Fig. 13. Archived Scenario PGFR Fig. 14. Archived Scenario PGFR
for 1300 MW Loss in Arizona. for 1200 MW Loss in California.
60.05 60.05

60.00 60.00

59.95 59.95
Frequency (Hz)

Frequency (Hz)
59.90 59.90

59.85 59.85
BPA BPA
BC Hydro BC Hydro
59.80 SCE 59.80 SCE
APS APS
59.75 PGAE 59.75 PGAE
MONTANA MONTANA
Mean Mean
59.70 59.70
0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5 0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5
Time (Seconds) Time (Seconds)

Fig. 15. Blocked Scenario PGFR Fig. 16. Blocked Scenario PGFR for
for 1300 MW Loss in Arizona. 1200 MW Loss in California.
60.05 60.05

60.00 60.00

59.95 59.95
Frequency (Hz)

Frequency (Hz)

59.90 59.90

59.85 59.85
BPA BPA
BC Hydro BC Hydro
59.80 SCE 59.80 SCE
APS APS
59.75 PGAE 59.75 PGAE
MONTANA MONTANA
Mean Mean
59.70 59.70
0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5 0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5
Time (Seconds) Time (Seconds)

Fig. 17. Transferred Scenario for Fig. 18. Transferred Scenario for
1300 MW Loss in Arizona. 1200 MW Loss in California.

Fig. 19 and 20 show the PGFR characteristic for the “balanced scenarios” for the 1300
MW loss in Arizona and the 1200 MW loss in California.

16
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

60.05 60.05

60.00 60.00

59.95 59.95
Frequency (Hz)

Frequency (Hz)
59.90 59.90

59.85 59.85
BPA BPA
BC Hydro BC Hydro
59.80 SCE 59.80 SCE
APS APS
59.75 PGAE 59.75 PGAE
MONTANA MONTANA
Mean Mean
59.70 59.70
0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5 0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5
Time (Seconds) Time (Seconds)

Fig. 19. Balanced Scenario for 1300 Fig. 20. Balanced Scenario for 1200
MW Loss in Arizona. MW Loss in California.

Fig. 21 and 22 show the PGFR characteristic for the mean values of the WI regions for
the 1300 MW loss in Arizona and the 1200 MW loss in California.
60.05 60.05

60.00 60.00

59.95 59.95
Frequency (Hz)

Frequency (Hz)

59.90 59.90

59.85 59.85
Balanced Mean Transferred Mean
59.80 59.80
Transferred Mean Balanced Mean
59.75 59.75

59.70 59.70
0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5 0.0 4.5 9.0 13.5 18.0 22.5 27.0 31.5 36.0 40.5 45.0 49.5 54.0 58.5
Time (Seconds) Time (Seconds)

Fig. 21. Mean PGFR characteristic Fig. 22. Mean PGFR characteristic
for 1300 MW Loss in Arizona. for 1200 MW Loss in California.

The difference between the mean values is not significant for either the Arizona or the
California loss. This indicates that there is no reliability reason to limit the transfer of
PGFR on the WI.

2.3 PGFR Withdrawal


As part of the evaluation of the PGFR in North America, NERC4 initiated PGFR
withdrawal investigations. These investigations drive the efforts to modify the
measurement methods and reliability concerns. Markets designed without inclusion of
adjustments for PGFR are one of the main reasons for this phenomenon.
Withdrawal of PGFR is an undesirable characteristic associated most often with digital
turbine-generator control systems using set-point output targets for generator output.
These are typically outer-loop control systems that defeat the primary frequency response
of the governors after a short time delay, and return the unit to operating at a requested
MW output.
17
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Some of the typical causes of the withdrawal are:

Plant outer-loop control systems – driving the units to MW set-points;

Unit characteristics:
o Plant incapable of sustaining primary frequency response;
o Governor controls overridden by other turbine/steam cycle controls; and,

Operating philosophies – operating characteristic choices by plant operators;


o Desire to maintain highest efficiencies for the plant.
Sustainability of primary frequency response becomes more important during light load
conditions (nighttime) when there are generally fewer frequency-responsive generators
online.
Several of the governor survey questions addressed the operational status and parameters
of the governor fleet. The results of the survey reported:

About 90% of the generators have governors;

Virtually all (95–99% by interconnection) of the GOs and GOPs governors are
operational;

80–85% (by interconnection) of the governors can sustain primary frequency


response for longer than 1 minute if the frequency remains outside of their
deadband; and,

Roughly 50% of the governors had unit-level or plant-level control systems that
override or limit governor performance.
Even though most generators reported having operational turbine governors, half have
unit- or plant-level control systems that override governor responses. These control
systems allow the units to return to scheduled output (MW set-point) or an optimized
operating point for economic reasons. These factors heavily influence the sustainability
of primary frequency response, contributing to the withdrawal symptom often observed.
This is often evident during light load periods in the middle of the night when high-
efficiency, low-cost units that operate on MW set-points comprise most the generators
dispatched to serve load.
In one weekend in the spring of 2012, two events exhibited generator trips. The first
event (Fig. 23) tripped 1,711 MW of generation with a typical -23,690 MW/Hz frequency
response.

18
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Value A = 60.021 Hz

ΔF = 0.0722 Hz
FR = -23,690 MW/Hz

Value B = 59.948 Hz

Fig. 23. PGFR Characteristic 3:30 pm Saturday 1,711 MW Resource Loss.

The second event (Fig. 24) occurred late Sunday night when load in the Eastern
Interconnection was much lighter, and the generators dispatched—probably the most
efficient units—were of a different character. The resource loss of almost 700 MW less,
exhibited the “lazy L” of primary frequency response withdrawal with a significant
reduction in frequency response of the interconnection. Point C defined to occur during
the first 8 seconds (at that time) was 59.962 Hz, while a lower frequency point of about
59.939 Hz occurred about 1 minute after the event.
These two events point to the composition of the dispatch and the characteristics of the
units on-line as primary elements in the frequency response strength, as well as the key
elements in creating withdrawal. Therefore, when calculating an Interconnection
Frequency Response Obligation (IFRO), it is important for operational planners and
operators to recognize the potential for withdrawal and consequentially the lower
frequency one to two minutes after the beginning of the event.

19
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Value A = 60.026 Hz

ΔF = 0.0799 Hz
FR = -1,312 MW/0.1 Hz

Value B = 59.946 Hz

Point C’

Fig. 24. PGFR Characteristic 11:21 pm Sunday 1,049 MW Resource Loss.


The measurement of Value B (20 to 52 seconds) does not capture the lowest frequency.
That frequency point is the true frequency event nadir, hereafter referred to as Point C’
(“Point C Prime”), and is normally equal to Point C for events that do not exhibit the so-
called “lazy L” effect. It is important that the phenomenon be recorded and trended to
determine if it is improving or deteriorating.
NERC Recommendation – Measure and track frequency response sustainability trends
by observing frequency between T+45 seconds and T+180 seconds. Calculating a pair of
indices comparing that value to both Point C and Value B provides for gauging
sustainability. A later alternative comprises the use of arrested PGFR as the PGFR
measure.
“NERC should embark immediately on the development of a NERC Frequency
Response Resource Guideline to define the performance characteristics expected of those
resources for supporting reliability. That guideline should address appropriate
parameters for the following:

20
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Existing conventional generator fleet – to retain or regain frequency response


capabilities of the existing generator fleet, adopt:
o deadbands of ±16.67 mHz,
o droop settings of 3%–5% depending on turbine type,
o continuous, proportional (non-step) implementation of the response,
o appropriate operating modes to provide frequency response, and
o appropriate outer-loop controls modifications to avoid primary frequency
response withdrawal at a plant level.

Other frequency-responsive resources – augment existing generation response


with fast-acting, electronically coupled frequency responsive resources,
particularly for the arresting and rebound periods of a frequency event:
o contractual high-speed demand-side response,
o wind and photo-voltaic – particularly for over-frequency response,
o storage – automatic high-speed energy retrieval and injection, and
o variable-speed drives – non-critical, short-time load reduction.”
Withdrawal of primary frequency response caused by outer-loop control systems must be
addressed. The Frequency Response Withdrawal section of this report shows the mix of
dispatched resources highly influences frequency response during light load periods.
Economics of the dispatch dictates that the most efficient, cost-effective generation will
remain on-line during those periods. Such generation often employs set-point controls
that return generation to AGC-prescribed or efficiency-prescribed generation levels
regardless of system frequency. This results in “squelching” of any primary frequency
response that the governors may have provided during a frequency event. This
withdrawal of primary response before secondary frequency response from AGC
becomes effective starting at about T+45 to T+60 seconds, creating the “lazy L” event
response prevalent in the Eastern Interconnection. This resulted in the following Industry
Advisory:8
“As a result of the Eastern Interconnection Frequency Initiative, the NERC
Resources Subcommittee has determined that a significant portion of the Eastern
Interconnection generator deadbands or governor control settings inhibit or prevent
frequency response. While this specific work was based on the Eastern
Interconnection, in the absence of more stringent regional requirements the

8
Industry Advisory: Generator Governor Frequency Response, Initial Distribution: February 5, 2015,
http://www.nerc.com/pa/rrm/bpsa/alerts%20dl/2015%20alerts/nerc%20alert%20a-2015-02-05-
01%20generator%20governor%20frequency%20response.pdf.

21
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

following good practice and guidance is applicable to all interconnections. The


proper setting of deadbands, droop, and other controls to allow for primary
frequency response is essential for reliability of the Bulk Electric System (BES) and
critical during system restoration. Further, the accuracy of Transmission Planning
models are impacted by incorrect governor data. The purpose of this Advisory is to
alert the industry of recommended governor dead band and droop settings that will
enable generators to provide better frequency response to support the reliable
operation of the Bulk Electric System.”

2.4 Recommended Methods to Measure PGFR


The TF recommends possible ways to perform on-line monitoring to estimate the
governor steady state droop and response of turbine-generator governor action.
A discussion of the measurement of PGFR immediately follows. The discussion includes
consideration of the factors that affect the methods chosen to measure PGFR for
implementation in the NERC BAL-003-1.1 – Frequency Response and Frequency Bias
Setting reliability standard.
2.4.1 Settled PGFR as a PGFR Measure
Fig. 6 shows the classic PGFR profile including estimates for A, C, and B, used in the
Frequency Response Measure (FRM) specified in NERC BAL-003-1.1 – Frequency
Response and Frequency Bias Setting reliability standard. Settled PGFR provides the
best FRM to estimate the Frequency Bias Setting in Tie-line Bias Control based
Automatic Generation Control Systems. However, the industry recognizes considerable
variability in measurement resulting from the selection of Point A and Point B. This
makes the traditional measurement method unsuitable as the basis for an enforceable
reliability standard in the real world setting of multiple control area interconnections.
Measuring the settled PGFR of an Interconnection is straightforward and accurate. All
that is needed to make the calculation is the active power size of a given contingency
(MW), divided by the change in frequency to express PGFR in MW/Hz.
Measuring a CA’s contribution to settled PGFR is more challenging. Prior to BAL-003-
1.1, NERC’s Frequency Response Characteristic Survey Training Document provided
guidance to calculate PGFR. In short, it identified the CA’s interchange values
“immediately before” and “immediately after” the disturbance event and use the
difference to calculate the MW the CA deployed for the event. There are two challenges
with this approach:
(1) Two people looking at the same data would come up with different values
when assessing which exact points immediately before and after the event.
(2) In practice, the actual response provided by the CA can change significantly
in the window of time between point B and when secondary and tertiary
control assist in recovery.

22
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Therefore, NERC standardized the measurement of settled PGFR to limit the variability
in measurement resulting from the poorly specified selection of Point A and Point B. The
first scan value that shows a deviation in frequency of some significance, usually
approaching 10 mHz, is t-0. The goal is to select the first scan prior to t-0 unaffected by
the deviation and appropriate for one of the averaging periods.
The A-value averaging period of approximately the previous 16 seconds prior to t-0
allows for an average of at least 2 scans for entities utilizing 6 second scan rates.
The B-value averaging period of approximately (t+20 to t+52 seconds) is the average of
the data after PGFR deployment and a completed (settled) transient, but before
significant influence of secondary control.
Settled PGFR only requires the values associated with the frequency change between the
initial frequency, A-value, and the settled frequency, B-value. No reasonable or
consistent calculations can be made related to the arrested frequency, C-value, using
Energy Management System (EMS) scan rate data for as long as 6-seconds. The same
applies to tie-line flow values associated with the minimum value of the PGFR
characteristic (C-value) as currently measured at the CA level.
Availability of an accurate frequency metering source and knowledge of the magnitude of
the resource/load imbalance in MW, makes calculation of the settled PGFR on an
interconnection basis straight forward.
Measurement on a CA basis can be a challenge, since the change in the individual CA's
metered tie lines that are not measured at the same time determines the change in MW.
Algebraic summation of tie line flows (net tie flow) obtained by time stamped, digital
scanning of meters would improve the accuracy of the process.
Measurement of settled PGFR on an individual resource or load basis requires analysis of
energy amounts that are often small and difficult to measure using current methods. In
addition, the number of resources and loads providing their response on an
interconnection could be problematic when compiling results for multiple events.
An appropriate sample size in the analysis provides for consistent measurement of PGFR
for a selected number of events and produces representative PGFR values. Available
research investigating the minimum sample size to provide consistent measurements of
PGFR shows that a minimum sample size of 20 events should be adequate.
2.4.2 Arrested PGFR as a PGFR Measure
The PGFR of interest when developing a measure to assure reliability by bounding
instantaneous interconnection frequency and preventing the operation of under-frequency
relays is “arrested PGFR.” The balancing inertia, governor response and the time delays
associated with the delivery of governor response significantly affect the arrested PGFR.
The arrested PGFR is calculated using the change in frequency between the initial
frequency, A-estimate, and the maximum frequency change during the event, C-estimate,
instead of the B-estimate. Arrested PGFR is the correct response for determining the
23
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

minimum PGFR associated with under-frequency relay operation and the support of
interconnection reliability. This is because it provides a direct estimate of the maximum
frequency deviation an interconnection will experience for an initial frequency and event
size in MW.
As measurement infrastructure improves, the Frequency Response Obligation (FRO) is
expected to transition to a measurement based directly on the arrested PGFR while the
Frequency Bias Setting will continue to be based on the settled PGFR. However
currently, EMS measurement methods in use do not support the necessary level of
accuracy to estimate arrested PGFR contribution for an individual CA or resource.
Current technology suggests that Phasor Measurement Units (PMUs) must eventually be
used.
2.4.3 Comparing Settled and Arrested PGFR
Simulations performed with a simple model, to show the effects that changes in selected
individual components, provide the basis for all the graphs. This model is an expansion
of the model used to provide the graphs included in the Primary Frequency Control –
PGFR Illustration section. This expanded model includes additional PGFR attributes. It
does not include any of the effects of the electrical system, it simulates only the power
and energy balances.

60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6 Settled PGFR = -9,750 MW / Hz Arrested PGFR = -5,059 MW / Hz

Settled PGFR = -8,125 MW / Hz Arrested PGFR = -4,677 MW / Hz

Base Settled PGFR = -6,500 MW / Hz Arrested PGFR = -4,236 MW / Hz


59.5
Settled PGFR = -4,875 MW / Hz Arrested PGFR = -3,695 MW / Hz

Settled PGFR = -3,500 MW / Hz Arrested PGFR = -3,040 MW / Hz

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 25. Settled PGFR versus Arrested PGFR

24
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

The settled PGFR is the base measure used in the FRO. Fig. 25 demonstrates that settled
PGFR and arrested PGFR are not necessarily equivalent or proportional. The droop
setting on a generator or the final PGFR delivered determines the settled PGFR, while
arrested PGFR results from the interaction of all the components from all resources
contributing to PGFR. In this specific case, a settled PGFR of -6,500 MW/Hz only
results in an arrested PGFR of -4,236 MW/Hz. As settled PGFR changes, the arrested
PGFR also changes, but not necessarily proportionally. Arrested PGFR correctly
measures contributions to reliability, when premature withdrawal of PGFR results in a
secondary nadir, and the arrested PGFR calculation is based on the secondary nadir.

2.5 Summary
Dynamics simulations of the North American Eastern, Western, and Texas
Interconnections for the recommended IFRO showed levels of PGFR to be adequate to
avoid tripping of the first stage of the interconnection UFLS systems.
A European based study indicated there are CAs that could reduce their PCR
requirements, and others that could increase them, while maintaining system security.
Interestingly, certain reallocation strategies might be beneficial not only for the involved
CAs, by reducing their PCR procurement costs, but also for the entire system by reducing
the post-contingency line loadings.
In a North American WI PGFR transfer study(Section 2.2.2), the difference between the
PGFR characteristics is not significant. This indicates that there is no reliability reason to
limit the transfer of PGFR on this interconnection.
The North American Eastern Interconnection must address withdrawal of primary
frequency response caused by outer-loop control systems. Such generation employs set-
point controls that return generation to AGC-prescribed or efficiency-prescribed
generation levels regardless of system frequency. This results in “squelching” of any
PGFR that the governors provide during a frequency event. This resulted in the
following Industry Advisory:8
“As a result of the Eastern Interconnection Frequency Initiative, the NERC Resources
Subcommittee has determined that a significant portion of the Eastern Interconnection
generator deadbands or governor control settings inhibit or prevent frequency response.
… The proper setting of deadbands, droop, and other controls to allow for primary
frequency response is essential for reliability of the Bulk Electric System (BES) and
critical during system restoration. Further, the accuracy of Transmission Planning
models are impacted by incorrect governor data. … to support the reliable operation of
the Bulk Electric System.”
Without a significant improvement in measurement infrastructure accuracy, the industry
cannot measure the arrested PGFR for less than a full interconnection accurately. If
measurement of PGFR is to be effective, arrested PGFR is the measure that must
eventually be used. This can only be accomplished effectively with frequency, power
and energy measurements that are time stamped and sampled at much higher rates than
25
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

currently supported at the EMS level. Current technology suggests that Phasor
Measurement Units (PMUs) must eventually be used at CA tie line boundaries and at
points of interconnection for generators, loads and storage participating in a PGFR
market or receive compensation for PGFR.
When there is no secondary nadir, only arrested PGFR correctly measures contributions
to reliability. When there is premature withdrawal of PGFR resulting in a secondary
nadir, the arrested PGFR calculation based on the secondary nadir also correctly
measures contributions to reliability.

3 PGFR Components9
This chapter presents discussions related to each of the components necessary to assure
adequate PGFR to a change in grid frequency. Example calculations with frequency
profiles in the following sections compare these components. Discussion along with
estimates of the reliability effects includes studies of each component and synthetic
alternatives.
In the simplest case, PGFR includes any automatic response to changes in local
frequency. If the response contributes to decrease the change in frequency, it is
beneficial to reliability. If the response contributes to increase the change in frequency, it
is detrimental to reliability. However, this definition does not address the relative value
of one response as compared to other responses that a specific case may provide. There
are numerous components associated with the PGFR that affect the reliability value and
economic value of the response.
The list below includes the components of interest and the associated issues:
(1) Governor droop settings;
(2) Governor deadband settings;
(3) Time delays in governor response;
(4) Dynamic characteristics of governor response;
(5) Kinetic balancing inertial response10;
(6) Synthetic step inertial response based on change in frequency and synthetic
balancing inertial response proportional to rate of change in frequency;
(7) Load damping;

9
Adapted from comments by H. F. Illian to Federal Energy Regulatory Commission, Docket No. RM16-
6-000, Essential Reliability Services and the Evolving Bulk Power System – Primary Frequency
Response, April 25, 2016.
10
Kinetic balancing inertial response is also known as synchronous inertial response.

26
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

(8) Step load interruption based on frequency without restoration;


(9) Step load interruption based on frequency with time delay based restoration;
(10) Step load interruption and restoration based on frequency; and,
(11) Synthetic continuous PGFR proportional to frequency error.
In all examples, the simulation used for Fig. 25 provides the sensitivities associated with
the PGFR characteristics that represent the example variations of components.

3.1 Governor Droop, Deadband, Time Delays & Dynamic Characteristics


Governor settings have significant effects on the PGFR provided by governors on
generation equipment. The following sections discuss these effects.
3.1.1 Governor Droop
Governor droop is the traditional way of specifying the generator’s PGFR to a change in
grid frequency. Droop indicates the relative change in frequency that would cause the
governor to change the generator output by the full capacity of the generator. It only
specifies the slope of the governor response in terms of the final steady state; it does not
indicate when that response will begin or the timing of the response. For example, a 5%
droop for a generator governor indicates that a change in frequency of 5% would cause a
generator to change its output by its defined capacity. A generator with a 5% droop and a
defined capacity of 100 MW would change its output by 33.3 MW / Hz = 100 MW / (.05
x 60 Hz). This method of specifying PGFR makes it easy to specify the desired PGFR of
a generator in a way that makes all generators respond in a similar proportion to their
defined capacity with respect to steady state contribution but ignores the dynamic
performance.
When including demand side or storage resources providing PGFR, the specification of
droop fails to provide the desired ability to effectively specify PGFR with droop alone.
As the industry moves forward in the use of demand side and storage supplying PGFR, it
may be desirable to specify PGFR as a specific MW range to be delivered over a
specified frequency range. This clarifies the responsibilities of those participating in the
PGFR market and provides a common specification for loads and generators.
3.1.2 Governor Deadband
Today, governor deadbands are a carryover from mechanical governor controls. They
accounted for backlash in mechanical systems as part of control system design. The new
electronic controls reduce the need to represent backlash in these control systems, unless
intentional deadbands are needed. However, experience demonstrates that intentional
deadbands transfer frequency control responsibilities from one generator to another. For
example, a generator on the Eastern Interconnection with a deadband setting of +/-50
mHz will respond to less than 1% of the frequency deviations while a generator with a +/-
36 mHz deadband will respond to about 5% of the frequency deviations, and a generator

27
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

with a deadband of +/-18 mHz will respond to a little over 30% of the frequency
deviations. This transfer of responsibility is one of the reasons that primary frequency
control includes rules to limit the deadbands on generators.
Fig. 26 shows the effects of changing the governor deadband on both the arrested PGFR
and the settled PGFR. The governor deadband on generators affects the frequency profile
in multiple ways. Although, the initial rate of change of frequency is the same for all
deadbands, the additional change in frequency and time required, before governor
response is initiated, changes both the minimum frequency and the time at which the
minimum frequency is reached. The change in deadband changes the amount of
balancing inertia required because of the additional time before the governor response is
initiated. The deadband also affects the settled frequency and the settled PGFR by
changing the settled frequency by an amount equal to the deadband. Additionally, it
causes a change in minimum frequency and arrested PGFR, but that change is not a
proportional change.

60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6 0.0000 Hz Governor Deadband Arrested PGFR = -4,739 MW / Hz

0.0167 Hz Governor Deadband Arrested PGFR = -4,492 MW / Hz

0.0360 Hz Governor Deadband Arrested PGFR = -4,236 MW / Hz


59.5
0.0500 Hz Governor Deadband Arrested PGFR = -4,069 MW / Hz

0.1000 Hz Governor Deadband Arrested PGFR = -3,564 MW / Hz

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 26. Arrested PGFR – Governor Deadband

3.1.3 Governor Time Delays


In addition to time delays caused by the deadband setting, there are two types of time
delays associated with governor response. The first is the time delay in the initiation of
the response. It is usually the result of the time delays associated with the initiation of
working fluid flows to increase or decrease the energy supply to the turbine-generator.

28
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

The second is the result of specific control system instructions to limit the rate of change
of the output of the generator to maintain generator control stability.
3.1.3.1 Time Delay Before Response
Turbine-generators that convert heat and kinetic energy to mechanical energy and
eventually electrical energy have time delays related to the time that the control valves
are moved to initiate the change in power and the time that the power is delivered. These
time delays are usually associated with the time it takes a change in working fluid flow to
travel from the control valve to the first blades of the turbine in the turbine-generator.
Fig. 27 shows the effect of this type of time delay.

60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6 1 Sec Governor Response Delay Arrested PGFR = -5,500 MW / Hz

2 Sec Governor Response Delay Arrested PGFR = -4,977 MW / Hz

3 Sec Governor Response Delay Arrested PGFR = -4,236 MW / Hz


59.5
4 Sec Governor Response Delay Arrested PGFR = -3,710 MW / Hz

5 Sec Governor Response Delay Arrested PGFR = -3,325 MW / Hz

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 27. Arrested PGFR – Governor Response Delay

3.1.3.2 Time Delay During Response


Some turbine-generators have control systems that limit the rate of change in input to the
turbine-generator and, therefore, the rate of change of generator output. These controls
have the effect of delaying the delivery of power and energy as called for by the speed
governor. They are simulated here by delaying the delivery of the power and energy, as
indicated by the governor, linearly over a time interval from 2 to 8 seconds. Fig. 28
shows these limits by percent change per second, with 50% representing a 2 second
distributed delay and 12.5% representing an 8 second distributed delay. This delay
results in a change in the arrested PGFR without a change in the settled PGFR.

29
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6 None Governor Rate of Change Limit Arrested PGFR = -4,236 MW / Hz

50.0 % / Sec Governor Rate of Change Limit Arrested PGFR = -4,208 MW / Hz

25.0 % / Sec Governor Rate of Change Limit Arrested PGFR = -3,587 MW / Hz


59.5
16.7 % / Sec Governor Rate of Change Limit Arrested PGFR = -3,211 MW / Hz

12.5 % / Sec Governor Rate of Change Limit Arrested PGFR = -2,956 MW / Hz

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 28. Arrested PGFR – Rate of Change Limit

3.1.4 Dynamic Characteristics


A governor dynamic characteristic depends on the tuning of the governor controller. This
could be related to dashpot timing of older mechanical governors, gain parameters on
electronic and digital governors, dynamic gain changing or damping change by governors
etc. In general, offline gains of the governors (dashpot timing and temporary droop) are
well tuned with proper damping ratio for optimal speed control. However, in the absence
of well-defined guidelines for governor online performance, typically governor online
gains are tuned in an arbitrary manner. This arbitrary online tuning is more prevalent
with digital and electronic governors. Some of them are tuned with higher gains for
faster response and many others are with conservatively low gains and hence slow or
sluggish. Governors with higher gains tend to contribute more to arrest the frequency
(nadir) but also may contribute to an overshoot of the frequency or even could lead to an
oscillatory instability. Alternatively, governors with lower gains make the frequency
recovery slow with longer settling time.
As the industry moves forward to commercialize the ability to respond to a frequency
change, it may be desirable to specify dynamic response of the system to a frequency
change in terms of step response control specifications such as damping ratio and settling
time. This would clarify the responsibilities of those participating in the PGFR market.

30
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

3.2 Balancing Inertia, Kinetic and Synthetic


This section includes studies concerning balancing inertial response and recently
identified limitations associated with the provision of synthetic balancing inertial
response. Balancing inertial response is an important PGFR component and is a concern
as additional non-synchronous generation is added to an interconnection replacing
traditional synchronous generation. The replacement of the traditional synchronous
generation with non-synchronous generation results in a reduction in system inertia. This
section considers the following two conditions:
(1) Kinetic balancing inertia to assure interconnection reliability.
(2) Synthetic balancing inertia proportional to “rate of change of frequency.”
3.2.1 Kinetic Balancing Inertia
Kinetic energy from rotating mass synchronized with electric system frequency provides
kinetic balancing inertial response. All equipment synchronously connected to the
interconnection, both turbine-generators and rotating loads provide kinetic balancing
inertia. Kinetic balancing inertia provides power and energy without delay in an amount
proportional to the rate of change in frequency. As the system frequency changes, the
rotational speed of this synchronized rotating mass changes. This change in rotational
speed releases kinetic energy as it slows or absorbs energy that it converts to kinetic
energy as its speed increases. It is a physical characteristic of the system dependent on
the mass moment of inertia and the speed of rotation. Kinetic balancing inertia does not
affect the settled PGFR; it only affects the arrested PGFR. Contributions to kinetic
balancing inertia will not affect the settled PGFR measurement because as frequency
settles to a new value; the kinetic balancing inertia contribution settles to zero.
However, kinetic balancing inertia has a significant effect on the arrested frequency
value. As the kinetic balancing inertia increases, the rate of change of frequency
decreases, and the arrested PGFR and minimum frequency increase. As the kinetic
balancing inertia decreases, the rate of change of frequency increases, and the arrested
PGFR and minimum frequency decrease. Kinetic balancing inertia has this effect
because it slows the decline in frequency occurring during any time delays associated
with the initiation of PGFR. The initial slope of the frequency profile and the minimum
value reached because of the rate of change in frequency show this effect.
Fig. 28 shows how the frequency characteristic changes as the amount of kinetic
balancing inertia on the interconnection changes. A reduction in the balancing inertia
results in a greater change in the frequency nadir (lower nadir) due to the increase in the
initial change in frequency before the governors begin their response. This increased
initial change appears as the greater slope of the frequency characteristic immediately
after the disturbance occurs at time 0. This graph also indicates that, if the balancing
inertia is reduced sufficiently, the frequency will eventually drop below the under-
frequency relay limit resulting in the interruption of firm load. An interruption of firm
load is an indication of insufficient security margin on an interconnection.

31
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6 +50% Kinetic Balancing Inertia Arrested PGFR = -4,979 MW / Hz

+25% Kinetic Balancing Inertia Arrested PGFR = -4,654 MW / Hz

Base Kinetic Balancing Inertia Arrested PGFR = -4,236 MW / Hz


59.5
-25% Kinetic Balancing Inertia Arrested PGFR = -3,709 MW / Hz

-50% Kinetic Balancing Inertia Arrested PGFR = -3,028 MW / Hz

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 28. Arrested PGFR – Kinetic Balancing Inertia

3.2.2 Synthetic Balancing Inertia


The availability of new resources provides an opportunity to partially replace kinetic
balancing inertia with a synthetic means to supplement or even partially replace kinetic
balancing inertia. The need for balancing inertia resulted in the development of some
wind generators designed to provide synthetic balancing inertia. Initially, additional
energy withdrawn from the wind turbine provided additional energy to the wind
generator. The amount of energy available from the turbine limits this energy, and thus,
synthetic balancing inertia is only available for a short time following a disturbance. This
same limitation also prevents the energy provided by these wind generators from being
proportional to the rate of change in frequency.
These initial attempts to provide synthetic balancing inertia expand both the form of the
synthetic balancing inertia provided and the resources available to provide that response.
Additional resources considered as providers of synthetic balancing inertia include both
energy storage and loads. In the case of both energy storage and load, there is sufficient
energy available to provide synthetic balancing inertia proportional to rate of change in
frequency. The synthetic balancing inertia supplied would be much closer in form to the
kinetic balancing inertia supplied by traditional synchronous generation and load. While
synthetic balancing inertia may be useful, it may also lead to instability. Moreover, the
implementation often involves measuring the frequency and the rate of change of
frequency, which inherently amplifies measurement noises making the delivery of
synthetic balancing inertia problematic.
32
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

3.2.2.1 Simple Simulation


60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6 No Synthetic Balancing Inertia Arrested PGFR = -4,236 MW / Hz

+15% Synthetic Balancing Inertia Arrested PGFR = -4,240 MW / Hz

+30% Synthetic Balancing Inertia Arrested PGFR = -4,243 MW / Hz


59.5
+45% Synthetic Balancing Inertia Arrested PGFR = -4,247 MW / Hz

>=50% Synthetic Balancing Inertia Arrested PGFR = Unstable

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 29. Arrested PGFR – Synthetic Balancing Inertia <50%

60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6

59.5
>=50% Synthetic Balancing Inertia Arrested PGFR = Unstable

No Synthetic Balancing Inertia Arrested PGFR = -4,236 MW / Hz

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 30. Arrested PGFR – Synthetic Balancing Inertia >=50%


33
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Fig. 29 indicates that with synthetic PGFR simulated at various percentages, synthetic
balancing inertia may be a very good substitute for kinetic balancing inertia. However,
there is a possibility of instability when the amount of synthetic balancing inertia exceeds
the amount of kinetic balancing inertia. We investigated this problem in greater depth.
The first step investigated the frequency characteristic when the system is unstable.
Fig. 30 shows the instability in the system frequency. When a disturbance occurs, the
frequency appears to stabilize, but then becomes unstable. This characteristic alone fails
to provide sufficient information to understand why the frequency is unstable. However,
it does supply information about where to investigate to determine the cause. The solid
red region on the plot indicates the frequency varies at very short periods. This additional
information provides the necessary information about how to identify the reason for the
instability. Expanding the time domain identifies the reason for rotor angle and
frequency instability.

2000.0

1500.0
Imbalance Contribution (MW)

1000.0

500.0

0.0

-500.0

Kinetic Balancing Inertia Contribution MW


Synthetic Balancing Inertia Contribution MW

-1000.0
-1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Seconds

Fig. 31. Synthetic Balancing Inertia >=50% - Contribution Detail

Fig. 31 plots the imbalance as two separate plots and how the instability results from a
phase shifting of the imbalance and limit cycling between kinetic balancing inertia and
synthetic balancing inertia that increases over time resulting in the frequency instability.
The first red plot shows the MW contribution to the imbalance by the kinetic balancing
inertia, and the second black plot shows the MW imbalance contribution by the synthetic
balancing inertia. These plots indicate a significant rapid change in power between
kinetic balancing inertia and synthetic balancing inertia that could be highly detrimental
to equipment. This kind of instability could be of significant concern to equipment
providing this size varying output at a high frequency. Since this variability occurs with
34
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

synthetic balancing inertia proportions greater than 50%, suggests investigation of output
variability for synthetic balancing inertia with proportions less than 50%.

2000.0

1500.0
Imbalance Contribution (MW)

1000.0

500.0

0.0

-500.0

Kinetic Balancing Inertia Contribution MW


Synthetic Balancing Inertia Contribution MW

-1000.0
-1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Seconds

Fig. 32. Synthetic Balancing Inertia 45% – Contribution Detail

Fig. 32 shows the energy contributions by kinetic balancing inertia and synthetic
balancing inertia in megawatts. This indicates that even though the resulting frequency is
stable, there is a short period, fraction of a second, when there is a significant phase
shifting of balancing inertia contributions between kinetic balancing inertia and synthetic
balancing inertia. This is due to the control time delay for the synthetic balancing inertia.
Although synthetic balancing inertia is stable at contributions less than 50%, there may
still be detrimental effects for synthetic balancing inertia. A very simple energy model
illustrates these conclusions. More detailed studies investigate the additional effects on
electrical grid. The following section describes one such study for a small test system,
including representation of the electric grid.
3.2.2.2 Synthetic Balancing inertia and Small Signal Instability11
This work draws attention to the fact that when considering small signal behavior adding
synthetic balancing inertia to a power system may not always increase system stability. It
explores the choice of synthetic balancing inertia and damping constants for control
exercised at load points using one-axis generator models with fast excitation systems in
the IEEE 14-bus test system. Modifying the dynamic generator parameters produced two
topologically similar, but dynamically distinct models, which were both stable before the
addition of synthetic balancing inertia controls. While increasing the synthetic balancing
inertia and damping for each control point, the eigenvalues of the linearized system tracks
the small signal stability locus. The work assumes the synthetic balancing inertia control
signal produces a change in power proportional to the time rate of change of frequency.

11
H. E. Brown and C. L. DeMarco, "Synthetic inertia and small signal instability," 2016 North American
Power Symposium (NAPS), Denver, CO, 2016, pp. 1-6. doi: 10.1109/NAPS.2016.7747848

35
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

In one test system, adding synthetic balancing inertia alone induced small signal
instability (i.e., the linearized system displayed one or more eigenvalues with strictly
positive real part). This instability could be corrected if the locations exercising synthetic
balancing inertia control also applied simple damping control (i.e., a linear feedback
proportional to frequency error). However, the benefit of added damping was not
monotonic. The damping gains could not be increased indefinitely. If damping gains
increased sufficiently, critical eigenvalues became less well damped, and ultimately again
induced instability. These case studies suggest when considering the addition of
synthetic balancing inertia to a power system, the possibility of adverse effects on small
signal stability should be considered in addition to evaluating improved performance in
the frequency nadir. In other words, these results underscore the point made in the
previous section—requiring detailed studies to verify the stabilizing or de-stabilizing
effects of synthetic balancing inertia control that is proportional to time rate of change of
frequency.

3.3 Load Damping Synthetic PGFR


Load damping is another component of PGFR. As residential, commercial and industrial
customers convert many of their processes from synchronously connected equipment to
systems controlled independently of system frequency, load damping will decline. One
example is higher efficiency air conditioning using variable speed drives to improve
efficiency replacing induction motor driven air conditioning, eliminating load damping.

60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6 +50% Load Damping Arrested PGFR = -4,769 MW / Hz

+25% Load Damping Arrested PGFR = -4,501 MW / Hz

Base Load Damping Arrested PGFR = -4,236 MW / Hz


59.5
-25% Load Damping Arrested PGFR = -3,974 MW / Hz

-50% Load Damping Arrested PGFR = -3,715 MW / Hz

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 33. Arrested PGFR – Load Damping

36
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Fig. 33 shows the effect of changes in this component. Load damping currently provides
a minority of the PGFR, but it is an important contributor to reliability. This reduction in
load damping due to efficiency improvements provides an opportunity to use the loads
with improved efficiency to contribute to synthetic PGFR by installing PGFR controls on
this new higher efficiency equipment. However, these modifications require new efforts
to improve the PGFR from these resources.

3.4 Synthetic PGFR


As variable energy resources (VERs) grow, PGFR from traditional generation resources
needs to be supplemented. PGFR available from some existing and new resources can
provide synthetic products to supplement or even partially replace the current PGFR.
3.4.1 Synthetic Step PGFR – Manual Restoration
Interruptible load triggered by frequency relays, set above the involuntary under-
frequency relay set points (step PGFR) supplements PGFR in some areas. Compensation
for providing this PGFR service incentivizes its use. The demand side frequency relay
triggered by interruptible load is simply a step response without an automatic reset. It is
only available for use once as frequency declines, and provides no response as frequency
recovers.
Step PGFR with manual restoration of the load used to provide the response through
frequency based interruption proves to be an effective method of supplementing PGFR.
ERCOT has come to depend upon this type of response for as much as one third of their
PGFR.
3.4.2 Synthetic Step PGFR – Time Delay Restoration
The automatic step interruption could control some loads with automatic reconnect. This
provides a step response after a time delay to allow for frequency recovery during the
delay (step PGFR with time delay restoration). However, if implemented, the amount of
continuous response limits supplemental response to a maximum amount equal to the
underlying continuous response. Implementing step response, with automatic reconnect
greater than the amount of the underlying continuous response, could result in an
immediate interruption upon reconnection. Therefore, this method of determining when
to reconnect is not reliable, and could result in additional frequency events.
3.4.3 Synthetic Step PGFR – Frequency Based Restoration
The preferred method controls load with automatic step interruption and automatic
reconnect based on a recovery frequency. Setting the recovery frequency at a value
sufficiently different from the interruption frequency prevents the immediate re-
interruption of load due to the reconnect. However, this type of reset limits step PGFR to
an amount less than the continuous PGFR available between the frequency limits for
interruption and restoration. Therefore, this limits the step PGFR with restoration to less
than 50% of the total response over any frequency range.

37
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

3.4.4 Synthetic Continuous PGFR – Automatic Restoration


Wind generation first implemented new control systems using synthetic continuous
PGFR capabilities. Continuous indicates no discontinuities in the delivery of the PGFR
(no step changes) and is both proportional to the frequency error and supplied on a
continuous basis. However, without the retention of reserve capacity on the wind
generation, this synthetic PGFR was only available for loss of load disturbances and high
frequency conditions. This response is referred to as fast frequency response (FFR) in
ERCOT and elsewhere. The tax laws, implemented to encourage the installation of new
wind generation, discouraged the holding of reserve capacity on wind. Reserve held on
wind does not have the same availability as reserve from traditional generation because
of the variable nature of wind.
Controlling some non-essential loads provides a similar response to continuous PGFR
from generation. The same is true for electric storage resources and other non-
synchronous generators like photovoltaic solar. This synthetic response provides
minimal time delays and minimal deadbands, like generation governor response.
Reducing the time delays and the deadbands associated with the response, significantly
reduces the need for kinetic balancing inertial response. Since the delay time associated
with the provision of this type of response is like the delay time associated with the
provision of synthetic balancing inertial response, this type of synthetic response is
superior to both generation governor response and synthetic balancing inertia.

60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6 +97% Synthetic PGFR Arrested PGFR = -6,349 MW / Hz

+75% Synthetic PGFR Arrested PGFR = -5,329 MW / Hz

+50% Synthetic PGFR Arrested PGFR = -4,803 MW / Hz


59.5
+25% Synthetic PGFR Arrested PGFR = -4,490 MW / Hz

No Synthetic PGFR Arrested PGFR = -4,236 MW / Hz

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 34. Arrested PGFR – Synthetic Continuous PGFR

38
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Fig. 34 shows the effect of replacing traditional governor response with synthetic
continuous PGFR. It assumes the percentage kinetic balancing inertia retained on the
interconnection is the inverse of the percentage of synthetic PGFR. As the effective time
delay associated with providing PGFR decreases, because of the provision of synthetic
PGFR, the difference between settled PGFR and arrested PGFR declines. It declines
until the kinetic balancing inertia is insufficient to prevent the nadir from falling below
the settled PGFR. This condition occurs above 97% synthetic PGFR and 3% kinetic
balancing inertia in this example. Eliminating additional kinetic balancing inertia by
replacing traditional PGFR with synthetic PGFR causes the frequency nadir to decline
below the settling frequency. Reaching this point ends the advantage of synthetic PGFR.

60.1

60.0

59.9
Frequency (Hz)

59.8

59.7

59.6
+99% Synthetic PGFR Arrested PGFR = -4,415 MW / Hz (Hz)

59.5

59.4
-5.0 -2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0 47.5 50.0 52.5 55.0 57.5 60.0
Seconds

Fig. 35. Arrested PGFR – 99% Synthetic Continuous PGFR

Fig. 35 shows a frequency profile with insufficient kinetic balancing inertia. In this case,
the PGFR characteristic transitions from critical damping in the PGFR characteristic to
under-damping in the PGFR characteristic. This transition in the PGFR characteristic is
an indicator that kinetic balancing inertia is approaching a region where the
interconnection could fail to maintain operation in a reliable frequency range.

3.5 Summary
This section provides information about how a change in each component of PGFR
affects the total frequency characteristic when a frequency event occurs. It demonstrates
that reductions in sensitivity to frequency such as increased droop, increased deadbands,
increased time delays, reductions in kinetic balancing inertia, and reductions in load
damping are detrimental to limiting frequency deviations resulting from system
39
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

imbalances between load and generation. It also demonstrates that in many cases
synthetic components provided by both load and generation may be superior to their
natural counter parts. This is because, they can be implemented with greater sensitivities
such as smaller droops, smaller deadbands, and reduced time delays on most generation
and load. However, it also demonstrates that care must be taken when attempting to
replace some natural components with synthetic components, such as synthetic balancing
inertia, because small changes can lead to instability. Finally, it indicates that the use of
synthetic continuous PGFR may be superior to attempting to use synthetic balancing
inertia because of the instability issues associated with the later.

4 PGFR for Reliability Standards


This section reviews general definitions of PGFR in many regions around the world and
examines recent progress in the development of reliability standards applying to PGFR.
It reviews suggested forward looking definitions of reliability related ancillary services.
Finally, it investigates examples using arrested PGFR and balancing inertial response
instead of settled PGFR and how balancing inertial response affects reliability standards.
(1) It assures reliability by including PGFR in reliability standards.
(2) It provides standards based on arrested PGFR instead of settled PGFR.
4.1.1 Requirements for PGFR, Reserves and Backdown Margin
Table 112 provides and overview of the parameters associated with PGFR in many
regions around the world. Many differences exist from region to region. The expectation
is these parameters will change as PGFR approaches a limiting parameter with respect to
maintaining reliability on the interconnections.
In North America changes in the definitions of ancillary services results from recent
changes in the NERC reliability standards promulgated since Table 1 was developed.
U.S. Federal Energy Regulatory Commission (FERC) Ancillary Services currently
defines two types of reserves: Schedule 5 – Operating Reserve–Spinning Reserve
Service; and, Schedule 6 – Operating Reserve–Supplemental Reserve Service. The
industry is at the point where these FERC Ancillary Service definitions require change.
The terms Operating Reserve, Operating Reserve–Spinning, and Operating Reserve–
Supplemental are obsolete. Operating Reserve, initially defined by NERC as reserve
available within 10 minutes, consisted of two types, spinning and non-spinning. Many
places further required 50% of the Operating Reserve consist of Operating Reserve–
Spinning. NERC modified “non-spinning” to “supplemental.” The time frame changed
from 10 minutes to the Disturbance Recovery Period, 15 minutes. The NERC Reliability
Guideline – Operating Reserve Management excludes Operating Reserve–Spinning.

12
Yan Rebours, A comprehensive Assessment of Markets for Frequency and Voltage Control Ancillary
Services, Electric Power, University of Manchester, 2008.

40
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Table 1. Technical Comparison of Primary Frequency Control Parameters12

41
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Formal markets, approved by FERC, linked the secondary reserves providing for
regulation with the regulation reserve service. This resulted in the definition of
regulation as a service that included two aspects, a rate of change, and a capacity
requirement (reserve and backdown margin). Contingency reserve, currently the last of
the time dependent reserves, includes two aspects, a fifteen-minute delivery period, the
Disturbance Recovery Period, and a secondary reserve amount Most Severe Single
Contingency, MSSC. This links the reserve to a single service, Contingency Response,
defined by BAL-002-2 - Disturbance Control Standard – Contingency Reserve for
Recovery from a Balancing Contingency Event. This specific service now defines a
reserve amount and an associated delivery time frame.
The development of frequency responsive capacity requirement (reserve and backdown
margin), resulting from the implementation of the BAL-003-1 – Frequency Response and
Frequency Bias Setting standard, caused a similar change to occur with PGFR. The
change in output closely follows the change in frequency, and the capacity requirement
(reserve and backdown margin) links this change in output creating a single service,
PGFR. The FRO times the frequency range over which it is delivered, defines the
amount of required reserve to assure meeting the FRO.
Because of these changes, there is no longer a need for the reserve categories of
Operating Reserve–Spinning and Operating Reserve–Supplemental. Capacity categories
linked to specific performance services, PGFR, regulation and contingency management
replaces these reserve categories. If the industry develops new categories of capacity,
they should link to a specific service, such as flexibility service to manage long term
delivery of multiple hours in duration.
In the future, PGFR and regulation are likely to be separated in the pro forma ancillary
service tariffs. This will define the services explicitly. This greatly simplifies these
tariffs and assures their alignment with the reliability standards they support. The
services to be redefined are likely to be Regulation and Frequency Response, Operating
Reserve–Spinning, and Operating Reserve–Supplemental. The three new service
definitions are likely to be: PGFR, Regulation, and Contingency Recovery. These
definitions of the ancillary services will become clearer and will better support the
necessary market development.
Additionally, it may be an appropriate time to investigate the need to include Balancing
Inertia as an ancillary service. There are only small differences between the services
evaluated in this document and the services suggested in the ERCOT Concept Paper,
Future Ancillary Services in Electrical Reliability Council of Texas (ERCOT), 2013.13
Table 2 compares the services and makes a guess at the final definition for services that
are likely to be developed in the future.

13
ERCOT Concept Paper, Future Ancillary Services in Electrical Reliability Council of Texas (ERCOT),
2013, http://www.ferc.gov/CalendarFiles/20140421084800-ERCOT-ConceptPaper.pdf.
42
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Table 2. Comparison of Selected Ancillary Services

Initial ERCOT Eventual Description of Differences


Services Services Services

Kinetic Synchronous Kinetic Same


Balancing Inertial Balancing
Inertia Response Inertia
(SIR)

Synthetic Synthetic SIR Synthetic In both the initial services and


Balancing & Fast Balancing eventual services, synthetic
Inertia Frequency Inertia (limited balancing inertia is always
Response to an amount delivered in proportion to rate of
less than change of frequency. Therefore,
(FFR) Kinetic synthetic balancing inertia is like
Balancing kinetic balancing inertia except for
Inertia) the time delay before the energy is
delivered.

Governor Primary
Settled Frequency
Frequency Response
Response
(PFR)
(with limits)

Synthetic Fast Frequency


Continuous Response
PGFR
(FFR)

Regulating Regulating Regulating Same


Reserve Reserve Reserve
Service Service Service

Contingency Contingency Contingency Same


Reserve Reserve Reserve
Service Service Service

43
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

4.1.2 ENTSO-E Reliability Standards


European codes generally address the requirements for PGFR and associated reserves by
defining how the generators must provide response to frequency deviations and the
requirements for the system operator to schedule associated reserves to enable their
delivery.14,15,16 The associated reserves meet a system reserve requirement distributed
over the interconnection in proportion to load. These codes generally do not measure the
delivery of this required response for each disturbance event. They instead initially
determine the correctness of the response with off-line tests. General requirements for
most systems require a sensitivity (deadband) of +/- 10 mHz and full delivery at +/- 200
mHz for PGFR. However, these systems lack specific performance requirements for
actual events and only evaluate selected events after they occur. For example, the full
delivery time requirements do not assure a desired frequency nadir will not be exceeded.
The European systems are investigating rate of change of frequency measures (RoCoF).
Studies show RoCoF is proportional to the event size in MW. RoCoF is the ratio of event
size to balancing inertia, while the nadir is proportional to the ratio of event size to
arrested PGFR. Therefore, RoCoF is a surrogate for arrested PGFR when balancing
inertia and PGFR are not changing. Finally, Section 3.4.4 Synthetic Continuous PGFR –
Automatic Restoration demonstrates that on a system with changing balancing inertia and
synthetic continuous PGFR, RoCoF provides little additional knowledge. Fig. 34 shows
the initial slope of the frequency profiles. Arrested PGFR – Synthetic Continuous PGFR
demonstrates the failure of RoCoF to provide useful knowledge as synthetic continuous
PGFR replaces traditional PGFR and RoCoF increases.

4.2 Reliability Standard Design Based on Settled PGFR


While the IEEE Special Publication 07TP180 “Interconnected Power System Response to
Generation Governing: Present Practice and Outstanding Concerns”, May 2007, was
under development, NERC supported a similar project providing justification for the
development of a reliability standard to insure sufficient PGFR to assure reliability on the
North American interconnections. This document17 provided justification for the
development of a PGFR standard for NERC. It included examination of the trends in
PGFR on the interconnections and the implications resulting from those trends. It
followed with an analysis considering the form a standard should take to resolve the
problems that analysis of the trends revealed. It also recommended that the primary goal

14
Network Code on Load-Frequency Control and Reserves, ENTSO-E, June 28, 2013,
https://www.entsoe.eu/fileadmin/user_upload/_library/resources/LCFR/130628-NC_LFCR-Issue1.pdf.
15
Test for Primary Control Capability, Swissgrid Ltd, Version 1.0, November 2009,
https://www.swissgrid.ch/dam/swissgrid/experts/ancillary_services/prequalification/D110426_test-for-
primary-control-capability_V1R1_EN.pdf.
16
Code for Transmission, Dispatching, Development and Security of the Grid, Terna Electric System, May
11, 2004, http://www.terna.it/en-gb/sistemaelettrico/codicedirete.aspx.
17
Frequency Response Standard Whitepaper, Prepared by the Frequency Task Force of the NERC,
Resources Subcommittee, NERC, April 6, 2004,
http://www.nerc.com/docs/oc/rs/Frequency_Response_White_Paper.pdf.

44
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

of a reliability standard be based on preventing the interconnection frequency from


causing the tripping of firm load through the operation of under frequency relays. One of
the important conclusions from this document is “both responsiveness and depth of
response must be assured.” Issues presented in the whitepaper contributed to the
development and approval of two reliability standards.
4.2.1 ERCOT Reliability Standard
ERCOT developed a reliability standard to address the provision of PGFR based on a
settled PGFR measure in the ERCOT region.18 This standard assigns responsibility to the
CA to assure that the PGFR for the ERCOT region is maintained above the minimum
PGFR to assure reliability. It includes options to change the requirements for generators.
It also assigns responsibility for providing PGFR to most generators in the ERCOT
region based on the measurement of settled PGFR. This approach requires the standard
to include highly complex analysis and definitions for each category of generator with
respect to initial response, sustained response, governor operability, deadband, and droop
to meet specifications. The standard also includes a measurement that confirms the
PGFR is sustained until secondary control restores frequency to pre-disturbance values.
This approach results in a standard including requirements, definitions, equations and
flow charts extending for 49 pages.
4.2.2 NERC Reliability Standard
NERC developed a reliability standard to address the provision of settled PGFR in North
America.19 To assure reliability, this standard assigns responsibility to the CA to assure
the settled PGFR for each interconnection remains above the minimum settled PGFR. It
assigns responsibility to the CA to avoid requiring a specific assignment to generation or
load for providing PGFR. This approach reduces the complexity of the standard by only
requiring it to include only settled PGFR measurements. This approach results in a
standard including requirements, definitions, equations and two spread sheets extending
for a total of 12 pages plus two spreadsheets to automate the collection and analysis of
the necessary data. Additionally, by assigning responsibility for both the PGFR and the
PGFR reserve to the CA, responsibility for reliability is appropriately consolidated in a
single party assuring effective enforcement.
This approach is a significant improvement over the ERCOT standard with respect to the
complexity of the standard. It requires less complexity but it still has weaknesses. The
NERC Frequency Response Initiative Report.20 discusses the parameters used in the
standard. The parameter adjustments include:

18
BAL-001-TRE-1, Primary Frequency Response in the ERCOT Region, May 21, 2015,
http://www.nerc.com/pa/Stand/Reliability%20Standards%20Complete%20Set/RSCompleteSet.pdf.
19
BAL-003-1.1 — Frequency Response and Frequency Bias Setting, November 13, 2015,
http://www.nerc.com/pa/Stand/Reliability%20Standards%20Complete%20Set/RSCompleteSet.pdf.
20
Frequency Response Initiative Report - The Reliability Role of Frequency Response, NERC, October 30,
2012, http://www.nerc.com/pa/stand/project%20200712%20frequency%20response%20dl/fri_report_10-
30-12_master_w-appendices.pdf
45
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Adjusting the starting frequency from the assumption of 60 Hz to a reduced


frequency with a 95% probability that it will be equal to or greater than at the
beginning of a randomly timed frequency event;

Setting the minimum frequency limit to a value consistent with under-frequency


relay setting on the interconnection used with the starting frequency to get an
initial delta frequency available to manage frequency events;

Adjusting for the difference between frequency measured using one second
averages and frequency measured using sub-second values (This adjustment has
since been eliminated by better measurement methods.);

Adjusting the maximum acceptable delta frequency for the difference in settled
PGFR and arrested PGFR with a C to B ratio, CBR;

Adjusting the acceptable delta frequency for a nadir (C’) occurring after the initial
arresting point (C) for only the Eastern Interconnection to represent the
withdrawal of PGFR immediately after governor response;

Selecting a resource contingency protection criteria for each interconnection


determined by the largest MW loss for reliability planning purposes;

Using the credit for load resources to adjust the resource protection criteria by
reducing it by the amount of contractual high speed demand side response;

Calculating the Frequency Response Obligation (FRO) for the interconnection by


dividing the adjusted resource protection criteria by the adjusted delta frequency;

Distributing the FRO among the CAs on the interconnection in proportion to a


load plus generation ratio share; and,

Restricting the obligation for providing this FRO so that it does not require
measurement below the minimum frequency limit.
Thus, the standard requires both responsiveness and depth of response as recommended
in the Frequency Response Standard Whitepaper. Using a primary measurement weakly
related to the frequency nadir diminishes the quality of this standard in its current form.
Improving this relationship significantly improves the standard.
Many have contributed to the development of this standard and the supporting systems
that aid with the analysis of the data necessary to support this standard. For more
information on this issue see Frequency Response Analysis using Automated Tools and
Synchronized Measurements.21

21
R. D. Quint, P. V. Etingov, D. Zhou and D. N. Kosterev, "Frequency response analysis using automated
tools and synchronized measurements," 2016 IEEE Power and Energy Society General Meeting, Boston,
MA, 2016, pp. 1-5. doi: 10.1109/PESGM.2016.7741441

46
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

4.3 Reliability Standard Design Based on Arrested PGFR


This section addresses the problem recognized in the previous section that settled PGFR
is weakly related to the frequency nadir during frequency events. It investigates the
result that would be realized, if arrested PGFR were used in place of settled PGFR in the
NERC standard. To insure reliability, the modified standard continues to assign
responsibility to the CA to assure that the PGFR for each interconnection is above the
minimum PGFR. It is still appropriate to assign responsibility to the CA to avoid
requiring a specific assignment to generation or load for providing PGFR. This approach
reduces the complexity of the standard by replacing settled PGFR with arrested PGFR.
However, this still requires the measurement of settled PGFR because it is the correct
basis to determining the frequency bias setting for tie-line bias control.
The complexity of the standard using arrested PGFR in place of settled PGFR provides
significant improvement. It requires less complexity while eliminating the weaknesses
resulting from the use of settled PGFR. The modification of the parameters used in the
standard and improvements resulting from those modifications follow.

This would not change the starting frequency.

This would not change the minimum frequency limit.

This eliminates the C to B ratio, CBR, since C represents the nadir.

This eliminates C’, since C and C’ become the same value, representing the nadir.

The resource contingency protection criteria for each interconnection determines


the largest MW loss for reliability planning purposes and remains the same.

The credit for load resources remains the same.

The Frequency Response Obligation (FRO) for interconnections remain the same.

The distribution of the FRO among CAs on an interconnection remains the same.

The minimum frequency limit still restricts the obligation to provide the FRO.
The standard still requires both responsiveness and depth of response as recommended in
the whitepaper. Making this single change in measurement from settled PGFR to
arrested PGFR eliminates two parameters and the weakness resulting from the use of
settled PGFR.

4.4 Summary
Numerous approaches to assure sufficient PGFR to maintain reliability are in effect
across the interconnections. Only the format based on arrested PGFR offers the
opportunity to relate reliability directly to the measurement and provide both generators
and loads with a level playing field.

47
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

5 Economics Associated with PGFR22


This section analyzes the economics of PGFR including those provided synthetically by
load and non-synchronous resources.

5.1 PGFR Economics


In every economic system, there are two sides of the market, the supply side and the
demand side. The supply side provides the services used by the demand side. In the case
of PGFR, the supply side includes all providers of PGFR or components of PGFR. The
demand side includes all participants that create the need for PGFR or components of
PGFR. This definition does not refer to generation as the supply side and load as the
demand side. One of the basic mistakes made in the industry is assuming the supply side
of the PGFR market is generation and that the demand side is load. At times load can be
the supply side and generation can be the demand side. By discarding this basic
assumption, the problem and possible solutions change significantly.
The consideration of what is the supply side and what is the demand side should be
obvious from the start. Load damping is PGFR provided by load. Although it currently
contributes significantly less than half of the PGFR, it is a potential source of additional
synthetic PGFR. Including synthetic products such as synthetic PGFR supplied by
electronic control systems, it becomes obvious that the supply side includes all load,
generation and storage, and the demand side also includes all load, generation and
storage.
5.1.1 PGFR Costs – Supply Side
Costs associated with providing PGFR require some method of appropriate compensation
to those resources providing PGFR. Without compensation providers of PGFR will incur
additional costs unless they reduce or eliminate the response they provide. PGFR
acquired in a formal Regional Transmission Organization/Independent System Operator
(RTO/ISO) market or in a traditional CA subject to the tariffs will incur these costs.
It is the responsibility of the CA or the RTO/ISO to acquire the necessary amount of
PGFR to support reliability in the most cost effective manner. This efficiency function
performs best when suppliers are appropriately compensated for PGFR based on the
value of the PGFR they provide. Assuring fair compensation also assures suppliers will
provide PGFR.
Numerous factors affect the cost of providing PGFR from resources. Understanding
these costs before considering how to perform this economic evaluation and provide
compensation is a necessary part of the process. Some cost factors for providing PGFR
include the following.

22
Adapted from comments by H. F. Illian to Federal Energy Regulatory Commission, Docket No. RM16-
6-000, Essential Reliability Services and the Evolving Bulk Power System – Primary Frequency
Response, April 25, 2016.
48
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

(1) Metering and Measurement Costs support installation of the necessary


metering and measurement equipment to confirm delivery from a specific
resource.
(2) Control System Costs support the installation of the necessary controls to
provide PGFR from the resource.
(3) Capacity Opportunity Costs associated with providing reserve capacity for
PGFR support the foregone profits associated with alternative uses of the
capacity including energy or other ancillary services. There may also be
capacity opportunity costs associated with the loss in average capacity by a
load providing PGFR.
(4) Fuel Costs provide energy for PGFR. The costs for fuel to provide PGFR
result in energy costs significantly different from the system marginal energy
cost, both higher and lower. This is the case when resources that are not at the
system marginal cost provide PGFR.
(5) Energy Efficiency Penalty Costs include the loss in efficiency when the
resource is operated in a mode that supports the delivery of PGFR. This cost
is usually in the form of additional fuel to provide the same amount of energy.
An example is the difference between operating a steam turbine in valve
control mode with an active governor, and sliding pressure mode with valves
wide open and no active governor control except for over-speed. All the
energy provided by the resource incurs this cost, not just the energy provided
for PGFR. There may be additional energy costs or benefits associated with a
load providing PGFR from loss or gain in efficiency of their process when
load varies with the provision of PGFR.
(6) Capacity Efficiency Penalty Costs results from the loss of capacity
associated with the loss in energy efficiency. When efficiency is lost, capacity
may be lost at the same time because of limitations in the amount of input
energy available to the resource.
(7) Maintenance Costs result from additional maintenance incurred during the
operation of the resource in a manner necessary to provide PGFR.
(8) Emissions Costs result from additional costs incurred to manage any
additional emissions that result when the resource is providing PGFR or
stands ready to provide PGFR.
A good contract for the acquisition of PGFR from a resource compensates the resource
for all costs the resource incurs to provide PGFR. It also includes a method to evaluate
the least cost mix of resources necessary to provide the minimum required PGFR for
maintaining reliability.

49
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

5.1.2 PGFR Costs – Demand Side


Participants creating the need for PGFR cause the costs associated with acquiring PGFR
from the supplying resources. Reducing the amount of PGFR needed reduces the amount
of PGFR required, and the costs associated with acquiring PGFR. Assigning the cost of
acquiring PGFR from the supply resources to those parties that create the need for PGFR,
reduces or minimizes the amount of PGFR required to maintain reliability. These
considerations are the same as those driving the development of “real time pricing” and
“dynamic pricing”. Passing these costs on to those contributing to the need for PGFR
creates incentives to reduce the need for PGFR. This makes the interconnection less
expensive and more reliable. The problem is to balance both cost and complexity against
reliability on both the supply side and the demand side.
The industry should perform the necessary actuarial studies to assign responsibility for
providing and paying for the costs of, not only energy but, the additional services
required to support reliability.
5.1.3 Requirements for PGFR
It is surprising that the industry has been looking at the need for PGFR for many years
without investigating how much PGFR is used on the interconnections. A simple
analysis of the quantity and frequency of use for PGFR provides insight to solving the
problem of where it can be acquired in the most efficient manner.

1.E+00
2006 - Eastern I - Frequency Error Risk - 2 second Actual Probability Density Function
1.E-02
1.E-04
1.E-06
1.E-08
1.E-10
1.E-12
1.E-14
1.E-16
1.E-18
1.E-20
1.E-22
Probability

1.E-24
1.E-26
1.E-28
1.E-30
1.E-32
1.E-34
1.E-36
1.E-38
1.E-40
1.E-42
1.E-44
1.E-46
1.E-48
1.E-50
-0.800
-0.750
-0.700
-0.650
-0.600
-0.550
-0.500
-0.450
-0.400
-0.350
-0.300
-0.250
-0.200
-0.150
-0.100
-0.050

0.450
0.000
0.050
0.100
0.150
0.200
0.250
0.300
0.350
0.400

0.500
0.550
0.600
0.650
0.700
0.750
0.800

Frequency Error (Hz)

Fig. 36. Discrete Frequency Error Density

50
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Fig. 36 shows the 2-second frequency error for the Eastern Interconnection for the year
2006. It includes 14,446,000 data points out of a possible of 15,768,000 2-second
intervals, and is the most recent analysis available for this report. The tails are a little
fuzzy because the data tends to become sparse approaching the minimum and maximum
frequency errors. This plot indicates the number of seconds the frequency error is equal
to the indicated value on the horizontal axis. Dividing the count associated with each
frequency error by the total number of data samples in the data set converts the count
values to probabilities. The data plots the verticle axis on a log scale. The minimum
value on the verticle scale is 1/14,446,000, 6.9x10-8. This plot cannot indicate
probabilities smaller than this value.

1.E+00
2006 - Eastern I - Frequency Error Risk - 2 second Actual Cumulative Probability Density
1.E-02
1.E-04
1.E-06
1.E-08
1.E-10
1.E-12
1.E-14
1.E-16
1.E-18
1.E-20
1.E-22
Probability

1.E-24
1.E-26
1.E-28
1.E-30
1.E-32
1.E-34
1.E-36
1.E-38
1.E-40
1.E-42
1.E-44
1.E-46
1.E-48
1.E-50
-0.800
-0.750
-0.700
-0.650
-0.600
-0.550
-0.500
-0.450
-0.400
-0.350
-0.300
-0.250
-0.200
-0.150
-0.100
-0.050

0.200
0.000
0.050
0.100
0.150

0.250
0.300
0.350
0.400
0.450
0.500
0.550
0.600
0.650
0.700
0.750
0.800
Frequency Error (Hz)

Fig. 37. Discrete Cumulative Frequency Error

Fig 37 shows the accumulation of the frequency error counts from each tail limit to the
zero-frequency error point at the approximate center of the graph. This significantly
reduces or eliminates the fuzzy nature of the tails. The resulting graph for high frequency
indicates the cumulative probability of the frequency error is greater than or equal to the
value plotted (or less than or equal to for low frequency).
Fig 37 also indicates the probability that the frequency error will be greater than or less
than the indicated point on the plot. Seletion of a point on this graph provides an accurate
estimate of the probability that the frequency error will be greater or less than the selected
point. This provides a reasonable estimate of probabilities associated with frequency
errors that actually occur, but it does not answer the questions about the probabilities of

51
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

frequency errors greater than the probability associated with a single data point or the
probabilities associated with frequency errors beyond the limits of the graph.
Evaluation of many years of frequency data indicates the tails, when plotted on a semi-
log plot of the cumulative probability distributions, are linear. This simple plot indicates
that the tails are exponential. Performing a linear regression on the tails with a force fit
through the .054 Hz point, starting at three standard deviations from the center of the
distribution as set by the CPS1 limit (18 mHz on the Eastern Interconnection), 0.054 Hz,
results in a line showing the expected probability of reaching the under-frequency relay
limits for the interconnection assuming the probability tail remains linear.

1.E+00
2006 - Eastern I - Frequency Error Risk - 2 second Actual Cumulative Regression Line
1.E-02
1.E-04 1/10 Y Points 3 SD @ +/- 0.054 Hz
1.E-06
1.E-08
1.E-10
1.E-12
1.E-14
1.E-16
1.E-18
1.E-20
1.E-22
1.E-24
Probability

1.E-26
1.E-28
1.E-30
1.E-32
1.E-34
1.E-36
1.E-38
1.E-40
1.E-42
1.E-44
1.E-46
1.E-48
1.E-50
-0.800
-0.750
-0.700
-0.650
-0.600
-0.550
-0.500
-0.450
-0.400
-0.350
-0.300
-0.250
-0.200
-0.150
-0.100
-0.050
0.000
0.050
0.100
0.150
0.200
0.250
0.300
0.350
0.400
0.450
0.500
0.550
0.600
0.650
0.700
0.750
0.800
Frequency Error (Hz)

Fig. 38. Cumulative Tail Probability

Fig. 38 shows a reasonable estimate of the risk that a specific frequency error will be
exceeded under the linear assumption. Although this data is a little dated, it can estimate
preliminary costs associated with providing reserves and backdown margins necessary to
maintain reliability. The first step in this process determines how much reserve and
backdown margin to retain on each interconnection to assure adequate reliability. BAL-
003-1 Frequency Response and Frequency Bias Setting standard provides these values.
The NERC State of Reliability 201723 states the relevant required frequency responsive
reserves on the Eastern, Western and ERCOT Interconnections as 4,500 MW, 2,626 MW,

23
State of Reliability 2017, NERC, June 2017,
http://www.nerc.com/pa/RAPA/PA/Performance%20Analysis%20DL/SOR_2017_MASTER_20170613.
pdf

52
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

and 2,750 MW respectively. That same report provides estimates of the current actual
PGFR of -24,829 MW / Hz for the Eastern Interconnection, -15,446 MW / Hz for the
Western Interconnection and -8,067 MW / Hz for the ERCOT Interconnection. Finally,
This report estimates the starting frequency for the determining the Frequency Response
Obligation (FRO) for the interconnections as 59.974 Hz, 59.967 Hz and 59.966 Hz
respectively.
5.1.4 Economic Cost of Reserves for PGFR
With this information, the Eastern Interconnection estimates the cost of providing the
PGFR for the region when the frequency error is less than -0.054 Hz. Assuming the
starting frequency error is less than the frequency defined by the tails on Fig. 38, the
Eastern Interconnection determines that it will need the full 4,500 MW of reserves.
Assuming an initial frequency of 59.974 Hz leaves a frequency range of 0.474 Hz of
which at least 94%, 0.446 / 0.474, is required beyond 3 times Epsilon1. If the average
deadband is zero, the proportion is 94%. Larger deadbands will lead to proportions
greater than 94%. Further, assuming the cost of frequency responsive reserve, including
opportunity costs is about $10 / MWh, the total cost per year for frequency responsive
reserve will approach $400 million dollars per year, 8766 hours in the average year
multiplied by $10 / hour multiplied by 4500 MW.

1.E+50
2006 - Eastern I - 2 second Based on Regression
1.E+48
1.E+46
1.E+44
1.E+42
1.E+40
1.E+38
1.E+36
Cost of Reserve Used - $ / MWh

1.E+34
1.E+32
1.E+30
1.E+28
1.E+26
1.E+24
1.E+22
1.E+20
1.E+18
1.E+16
1.E+14
1.E+12
1.E+10
1.E+08
1.E+06
1.E+04
1.E+02
1.E+00
-0.800

-0.750

-0.700

-0.650

-0.600

-0.550

-0.500

-0.450

-0.400

-0.350

-0.300

-0.250

-0.200

-0.150

-0.100

-0.050

0.000

Frequency Error (Hz)

Fig. 39. Expected Cost of Reserves Below Frequency Error

Of greater interest is the cost per MWh of load interruption. Fig. 39 shows these
equivalent opportunity costs for the Eastern Interconnection. The graph illustrates an

53
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

average cost of avoiding load interruption beginning at -0.054 Hz and extending to an


average cost of avoiding load interruption to -0.500 Hz, the UFLS. The expected amount
of interruption that will occur at -0.054 Hz is a total of about 7.5 hours per year with a
cost of about $11,500 per MWh of load interruption. Comparing this value to estimates
of the value of lost load (VOLL), about $1,500 to $3,000 per MWh for residential
customers and from $10,000 per MWh and up for commercial and industrial customers,
indicates the cost per MWh of avoided load interruption is greater than the VOLL. These
values are within a range that justifies the use of load for reserves below a frequency of
59.946 Hz, 3 times Epsilon1, for the North American Interconnections instead of
continuing to provide PGFR from generation alone.
Considering the costs for resources to provide backdown margin, the equivalent for
reserves for high frequency conditions, yield similar results. Governor response from
generation is limited to a small percentage of the output of the generating unit, while
synthetic PGFR applies to much larger percentages of VERs and DERs, loads or storage
devices providing such response. Therefore, this analysis strongly suggests that the
PGFR market is not a single homogeneous market, but instead, divided into frequency
regions that are best served by differing resources. Initially generation, including
traditional generation resources, provides the PGFR reserves and backdown margin for
the frequency range between plus three times Epsilon1 and minus three times Epsilon1
on each of the interconnections. Load resources and VERs should provide reserves for
PGFR for frequency errors below minus three times Epsilon1. Generators and VERs
should provide backdown margin for PGFR for frequency errors greater than plus three
times Epsilon1. For example, the ERCOT Interconnection currently operates with wind
generation providing significant PGFR for high frequency errors. The three
recommended frequency regions, greater than plus 3 times Epsilon 1, less than minus 3
times Epsilon 1, and between plus and minus 3 times Epsilon 1, are based on assumed
economics. As the industry learns from implementing a program to assure PGFR
resources through voluntary participation with compensation, the ranges should be
adjusted to improve PGFR economics based on market conditions and prices.
It is unreasonable to expect the provision of PGFR over greater ranges will significantly
reduce or eliminate the need for regulation. The electric industry will achieve success to
the degree it is able to implement programs to reduce regulation requirements that are
unobservable to the customer or implemented in a manner that requires little customer
effort. The expectation that customers will make significant changes in behavior to
improve system operation is unreasonable when one considers that the average cost of
electricity is in the range of ¼% to 1% of the value of electricity to the customers. This
suggests that customers will be unwilling to change electric use patterns to achieve very
small changes in costs relative to the value for the product. Consequently, one must
assume either the costs savings resulting from changes in custormer behavior are large,
justifying compensaton to the customer. Or one must assume the changes must be
implemented without affecting the customer’s behavior. In other words, any changes in
energy use must occur with the consent of the customer or, without the customer being
aware of, or able to observe, those changes.

54
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

6 Incentive Design for PGFR Provision


PGFR has explicit and implicit costs involved in both the capability (installation of
control equipment) and the provision (reserving capacity in real-time and responding to
frequency events). In the 2007 Task Force Report, one of the primary conclusions was
the industry (particularly the North American Interconnections) lacked incentives for
providing this response. With incurred costs and lack of incentives, individual resources
may choose to block governors, set excessive governor dead bands, or operate in more
efficient modes with poor or no PGFR provision. Whether the CA is a ISO/RTO or a
vertically integrated regulated balancing area which owns the majority of generation
within the footprint, the set of rules in place impacts the system. This section discusses
some of the compensation and incentive designs in place or being proposed. It also
considers designs to integrate reliability, economic efficiency, and incentive
compatibility for the PGFR service.

6.1 Existing Compensation Schemes for PGFR


Historically, in North American restructured electricity markets, PGFR service has not
been included among the ancillary service market products. However, in some regions,
there are requirements that resources selling capacity into spinning and synchronized
reserve must be frequency responsive. This combines the two services and ensures an
incentive that resources earning revenue in these ancillary service markets are also
providing PGFR. A few areas of the U.S., particularly in ERCOT and the Western
Interconnection use this method. However, combining the compensation PGFR with
spinning/synchronized reserve, where the requirements have been specifically laid out as
secondary frequency control reserve (a 10-15 minute service), can lead to reliability
deficiencies and economic inefficiencies.24
Outside of North America, few electric markets include specific services and
compensation for the PGFR.25 Both the Australian National Electricity Market in
Australia and Transpower in New Zealand include ancillary services that correspond to
primary frequency response for contingency events. Australia includes eight different
frequency control ancillary services: regulation raise, regulation lower, six-second raise,
six-second lower, sixty-second raise, sixty-second lower, five-minute raise, and five-
minute lower.26 The six-second service arrests frequency decline and the sixty-second
service stabilizes system frequency. The six-second service corresponds to balancing
inertia and arrested PGFR, while the sixty-second service is closer to settled PGFR.

24
H. F. Illian, “Relating primary governing frequency response to future operating reliability,” in Proc.
IEEE PES General Meeting, Jul. 2007, pp. 1–5.
25
The Evolution of Ancillary Services to Facilitate Integration of Variable Renewable Energy Resources: A
survey of some changes to the ancillary services and ancillary service markets. EPRI, Palo Alto, CA:
2016. 3002008987.
26
J. Riesz, I. MacGill, “Frequency Control Ancillary Services: Is Australia a Model
Market for Renewable Integration?”, 2013.
55
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

In both cases these services are not tied to technology, and they do not require the service
to be online. It requires only that the injection delivery within the associated time frame.
Monitoring and compliance ensures the resources providing these services delivers
sufficient response. The response for six-second service must be automatic and local at
the resource without any direction from the Australian Electricity Market Operator
(AEMO). Each of these services have ancillary service market constructs comparable to
U.S. markets, with resources offering price-quantity bids for the service with prices based
on the marginal cost of the service.
In New Zealand, an “instantaneous reserve” is a required service to arrest the declining
frequency during disturbance events.27 It includes a fast instantaneous reserve and a
sustained instantaneous reserve. The fast instantaneous reserve needs to sustain output
for sixty seconds, while the sustained instantaneous reserve must respond for long
enough for the frequency to be brought back to nominal level. The rules allow generating
units or interruptible loads responding automatically to these events to provide these
services. Although comparable to the U.S. ancillary service markets, the generators and
HVDC owners, rather than loads, pay for the costs of these services. The overall cost of
this service in 2015 was about $19M New Zealand dollars (about $13M USD with
current exchange rates). This is a relatively high cost for the service for a system that
consumes about 10% of the energy that ERCOT does.

6.2 Industry Proposals for PGFR Compensation


In 2016, the U.S. Federal Energy Regulatory Commission (FERC) released a Notice of
Inquiry (NOI)28 on provision and compensation of PGFR posing these questions:
Should newly interconnecting resources be required to have the capability to
provide PGFR?
Should existing resources be required to have the capability to provide PGFR?
If so, what parameters should guide the capability (including droop and dead
band)?
Should resources that provide capability be compensated?
Should resources providing PGFR capacity and deployment be compensated?
How should compensation (pricing or markets) be determined?
How should a resources performance be monitored through instrumentation?
Ultimately, FERC realized that the NERC BAL-003-1 and FRO guided the amount of
PGFR that a CA must ensure, but did not have any mechanism for the CA to ensure it
was getting enough PGFR from individual resources. Based on this NOI, the two options

27
https://www.transpower.co.nz/system-operator/electricity-market/instantaneous-reserve
28
Essential Reliability Services and the Evolving Bulk-Power System—Primary Frequency Response, 154
FERC ¶ 61,117, (2016) (NOI).
56
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

that FERC considered to ensure enough PGFR from resources to meet the CA
requirement were:
(1) Requiring PGFR capability from the full set of individual resources, or
(2) Compensating and incentivizing resources that provide PGFR.
The follow-up Notice of Proposed Rulemaking29 from FERC suggested that a starting
point would be to enforce requirements for newly connecting resources to have the
capability of providing PGFR.
Current Proposed
Regulation Up Regulation Up
Fast-Responding Regulation Up Fast-Responding Regulation Up
Mostly unchanged
Regulation Down Regulation Down
Fast-Responding Regulation Down Fast-Responding Regulation Down

Fast Frequency Response 1 59.8 Hz, Limited duration


Fast Frequency Response 2 59.7 Hz, Longer duration
Responsive Primary Frequency Response

Contingency Reserves 1 SCED-dispatched


Contingency Reserves 2 Manually dispatched

Supplemental Reserves 1 SCED-dispatched


Non-Spin
Supplemental Reserves 2 Manually dispatched

Synchronous Inertial Response Ongoing development

Fig. 40. Current and proposed ancillary services as part of ERCOT’s ancillary
service redesign. The proposal was put on hold in 2016.

Other regions proposed targeted PGFR ancillary service markets to ensure sufficient
capacity and response as well as incentivize resources providing the service. Fig. 40
presents a proposed ERCOT redesign of their ancillary service markets. ERCOT
proposed the disaggregation of its current responsive reserve service into several new
services. The goal was to ensure technology neutral, needs-driven, unbundled services.
Part of the redesign unbundled the responsive reserve service into several different
services. The requirements for the services of PGFR linked fast frequency response 1
and fast frequency response 2. FFR1 and FFR2 are typically loads responding via under-
frequency relays, and differ based on what frequency triggers their response and for how
long the response is sustained. These services link to the proposed synchronous

29
Essential Reliability Services and the Evolving Bulk-Power System—Primary Frequency Response, 157
FERC ¶ 61,122, at PP 3-7 (2016) (NOPR).
57
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

balancing inertial response service.30 The requirement for PGFR may depend on how
much FFR and synchronous balancing inertia the system anticipates.
The ancillary service market design is similar to existing ancillary service market designs
for each of the services. Resources would bid the cost of the service and the quantity to
be provided. Using co-optimization with the energy market in the day ahead market
(DAM), ERCOT would select the least-cost set of resources for each service. Prices from
each service would be based on the marginal cost of providing that service. Some
services have relationships with other services that add complexities to the pricing of the
services. In May 2016, the ERCOT stakeholders decided against the proposed redesign
package. However, they continue to work with ERCOT to consider changes to remove
barriers for new resource types and to align with changing needs of the power system.
The CAISO has also recently begun discussions on potential compensation for PGFR
service. The ISO currently sees challenges in meeting its NERC FRO. It began
procuring the service from outside CAs to ensure sufficient response. In the CAISO’s
frequency response Phase 2 initiative31, the ISO is evaluating whether compensation
should be made for capital expenses for equipment needed, opportunity costs associated
with holding PGFR capacity, and/or any operating expenses associated with providing
PGFR when a frequency event occurs.

6.3 Potential Options for PGFR Sufficiency and Compensation


This section describes challenges and potential options regarding the scheduling and
pricing to ensure sufficient PGFR, and includes the task force’s recommendations for
best practices.
6.3.1 PGFR Capability Individual Resource Mandate
One option for ensuring sufficient PGFR requires all resources to provide PGFR. With
this option, there is no payment for PGFR. If all competitors are equal, they can build
anticipated PGFR costs into offered costs for other services, as in capacity markets.
There are two disadvantages to this proposal. First, requiring all resources to have PGFR
capability does not necessarily guarantee a level of PGFR in real-time, as the requirement
is only a capability requirement. If a resource is offline or at maximum capacity, it
cannot provide PGFR. Secondly, it may not be economic for all resources to have the
capability to contribute PGFR in real-time. The system may need only a subset of
resources contributing to assure that capability. One last challenge with this proposal is
the difficulty defining the requirement for non-traditional technologies, for example
demand response or energy storage. In addition, there may be new generating
technologies that provide energy very efficiently and cannot provide PGFR. If a mandate
to provide PGFR prevents them from providing energy, this may lead to economic
inefficiencies.

30
Synchronous inertia response service is called balancing inertia in this document.
31
California ISO, Frequency Response Phase 2 Issue Paper, December 2016. Available:
http://www.caiso.com/Documents/IssuePaper_FrequencyResponsePhase2.pdf .
58
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

6.3.2 Energy Market Design and Elimination of Off-basepoint Penalties


This section describes impacts the energy market design has on the incentives for
resources to provide PGFR.32 A properly functioning governor will respond
autonomously to frequency deviations by adjusting its output to assist the interconnection
in arresting the frequency decline or frequency increase, without the intervention of a
control room operator. This response is vital for power system reliability. Its energy
output will deviate from its real-time energy schedule. Few U.S. ISO manuals, tariffs,
and user guides, incorporate frequency as input into the energy settlements system. This
often results in financial penalties to the generator for deviating from its energy schedule.
As an example, assume the ISO has a 3% tolerance band around its financial penalties.
This means that if a generator deviates from its schedule by more than 3% of its operating
capacity above or below its energy schedule, it will receive a financial penalty for not
following schedule. Assuming, that a generator with a properly functioning governor has
a 5% droop curve, it follows that:
1 𝑝. 𝑢. 𝑝𝑜𝑤𝑒𝑟 0.03 𝑝. 𝑢. 𝑝𝑜𝑤𝑒𝑟
=
0.05 𝑝. 𝑢. 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑋 𝑝. 𝑢. 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦

𝑋 = 0.0015 𝑝. 𝑢. 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 = 90 𝑚𝐻𝑧 𝑓𝑜𝑟 𝑎 60𝐻𝑧 𝑠𝑦𝑠𝑡𝑒𝑚


If frequency is not incorporated within the energy settlements system, any time the
frequency deviates more than 90 mHz (i.e., if frequency dips below 59.91 Hz or above
60.09 Hz), a generator with a properly functioning governor and a 5% droop curve will
automatically be penalized under this rule. This is a significant disincentive. If the
generator has no other incentive to provide PGFR, it will have a disincentive to enable its
governor or operate in modes to provide PGFR. Eliminating this disincentive and these
penalties would be an important step forward. FERC Order No. 890 allowed CAs to
adjust for PGFR eliminating the penalty, but did not require them to do so. It is uncertain
how many CAs have adjusted their tariffs to eliminate the penalty when it occurs.
6.3.3 Requiring PGFR Capability in other Ancillary Services
This proposal would require that any resource providing spinning reserve and
participating in the existing spinning reserve market enable its governors and provide
PGFR reserve with enforcement. There is a connection with spinning reserve and PGFR
reserve since both may be responsive to disturbances. The spinning reserve would be
required to respond both to system operator commands and autonomously to frequency
deviations. This offers an additional incentive for resources to provide PGFR reserve in
contrast with Proposal 1 or Proposal 2. The resources would have to provide PGFR to
earn revenue in this spinning reserve market.
This is a relatively easy market design change to implement. However, it is not clear that
this change by itself will acquire the appropriate amount of PGFR. Resources differ in

32
E. Ela, A. Tuohy, M. Milligan, B. Kirby, and D. Brooks, “Alternative approaches for a frequency
responsive reserve ancillary service market,” Electricity J., vol. 25, no. 4, pp. 88–102, May 2012.
59
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

the amount of PGFR they are capable of providing. Scheduling sufficient spinning
reserve might not result in sufficient PGFR reserve being available. Further, the amount
of available PGFR reserve could change as the mix of resources providing the spinning
reserve changed. Imposing specific PGFR capability requirements on spinning reserve
resources would likely limit the spinning reserve supply and increase the price.24 Energy
prices could increase since spinning reserve resources also supply energy. Locational
constraints for spinning reserve and frequency response often differ, further complicating
a simple joint supply requirement. Spinning reserves in many areas have locational
requirements to avoid overloading the transmission system during contingency events.
Frequency responsive reserves may have different location-based requirements, or may
have none, which adds further complexity when the two are paired together in one
market. Other than these issues, the proposal is relatively easy to implement, making an
incentive available for resources to provide frequency responsive reserve.
6.3.4 Market Design for PGFR Capability
FERC defines PGFR capability as having a governor or equivalent equipment with the
ability to sense changes in frequency and to physically move the resource quickly enough
to provide useful levels of PGFR, measured in MW/Hz. This definition does not include
the capacity (MW) required to deliver a given number of MW to arrest frequency by
rebalancing the system. A market design that selects and compensates resources for
PGFR capability may have a similar effect as the PGFR capability individual resource
mandate. Determining the total capability requirement leads to a more efficient selection
of resources installing PGFR capabilities. This requires less cost to meet the PGFR
capability need. However, it is still a challenge to determine a PGFR capability
requirement, because the number of resources with this capability may not have
significant impact on actual PGFR capacity (MW) available in real-time. PGFR
capability (MW/Hz) only requires the rate of delivery to a frequency deviation, not the
reserve amount (MW). The market could also incentivize innovation to provide PGFR
capability at lower cost.
6.3.5 Market Design for PGFR Capability Based on Settled PGFR
There are benefits for reliability and economic efficiency by having an explicit market
that meets the PGFR response capability and compensates resources directly for holding
reserve capacity for PGFR in day-ahead and real-time markets. The need for PGFR
capability and capacity is based on the regional requirement (e.g., the FRO in North
America). The capacity cost, including lost opportunity costs, is the primary cost of
PGFR in real-time. Paying for capacity has advantages. This design33 is similar to the
ERCOT design laid out in their recent proposal (described above). It selects the required
amount of capacity at least cost to meet the CA requirement. Co-optimizing the costs of
PGFR with energy has benefits as well, as in U.S. markets for other ancillary services.

33
E. Ela, V. Gevorgian, A. Tuohy, B. Kirby, M. Milligan and M. O'Malley, "Market Designs for the
Primary Frequency Response Ancillary Service—Part I: Motivation and Design," IEEE Transactions on
Power Systems, vol. 29, no. 1, pp. 421-431, Jan. 2014.

60
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Equation (1) shows the market security-constrained unit commitment and economic
dispatch models can include this requirement for the capacity in a rather straightforward
manner. In the example of (1), P10 is the PGFR capacity, P10Req the PGFR capacity
requirement, D the load damping, L the load, fmax the maximum frequency deviation
(e.g., based on highest UFLS setting), f0 the nominal frequency, and IR the insufficient
reserve penalty. The marginal cost of meeting the PGFR capacity requirement
determines prices. Pricing hierarchy is also part of the market design. If PGFR is higher
quality than other capacity services (regulation or spinning reserve), then the price of the
service is guaranteed to be equal to or greater than those services.
𝑁𝐺
0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
∑ 𝑃10𝑖,𝑡 ≥ 𝑃1𝐴 − 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃10𝑅𝑒𝑞 ∀𝑡 ∈ 𝑁𝑇 (3)
ƒ0
𝑖=1

Two case studies provide an example of the benefits: (WC1) a base case with no PGFR
market or PGFR requirement and only a spinning reserve requirement that all online
resources assumed to be frequency response, and (WC2) a second case with a PGFR
requirement of 44 MW for a 0.2 Hz max deviation (22MW/0.1Hz) and separate
compensation for that service.34 The cases assume about a 15% wind power penetration
and these resources are not providing PGFR. The examples examine a 24-hour period.
Table 3. Comparison of Production Costs, Units on Line, and Spin
WC1 WC2
Production Costs ($) 401,287 403,616
Avg. Units Online 17 17
Avg. P1 (MW) 36.75 48.1

Table 3 shows that just relying on spinning reserve does not ensure the system meets the
PGFR requirement every hour. Since the system is co-optimizing PGFR with energy and
the other ancillary services, the cost of the PGFR capacity only increases by 0.5%.
Interestingly, Table 4 shows the overall payments are actually reduced because of a
reduction in energy payments occurs based on the added constraint.
Table 4. Total Payments for Energy, PGFR, and Spin

Total Energy Total PGFR Total Spin Total


Payments Payments Payments Payments
($) ($) ($) ($)
WC1 736,618 0 24 736,642
WC2 722,229 2,359 0 724,588

34
E. Ela, V. Gevorgian, A. Tuohy, B. Kirby, M. Milligan and M. O'Malley, "Market Designs for the
Primary Frequency Response Ancillary Service—Part II: Case Studies," IEEE Transactions on Power
Systems, vol. 29, no. 1, pp. 432-441, Jan. 2014.
61
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Numerous mathematical approaches can describe and solve this market design. On
isolated systems where the CA is in total control of the market to ensure keeping the
arrested frequency above some minimum level, the market operator can explicitly do
this.3536 Other designs and pricing methods can be used for CAs that are part of large
interconnections. The primary goal is to ensure the holding of sufficient capacity in
reserve to meet the PGFR requirement; and, the resources providing this response receive
compensation that aligns with their provision of providing this service, as well as other
services.
6.3.6 Including Other PGFR Characteristics as Part of the Market Design
Providing a market for PGFR capacity ensures meeting the PGFR capacity requirement
and compensating resources for providing capacity for that service. The lost opportunity
cost in the PGFR ancillary service price incentivizes the provision of PGFR service in
place of energy. However, there are many other characteristics beyond the capacity of
PGFR that are important to overall PGFR and reliability assurance. Listed below are a
few of these characteristics and possible ways to incentivize and ensure sufficient quality
based on many of the characteristics as shown in Section 3.
Balancing inertia can be incentivized in the market clearing as well, including
payments to resources for the amount they provide. Since balancing inertia is a
characteristic of the rotating machine, not of the dispatch, a capability
requirement may be sufficient to incentivize the characteristic.
Time delay in the initiation of PGFR is dis-incentivized to provide the response
quickly after the frequency event occurs.
Insensitivity to events incentivizes resources to respond to more events and to
reduce excessively high deadbands.
Sustainability incentivizes resources to maintain a response after the arresting
frequency until secondary frequency control returns the frequency to nominal.
Stable response incentivizes resources to provide a stable non-oscillatory
response by requiring proportional droop curves instead of stepped droop curves.
Response before frequency is arrested incentivizes response as the frequency
nadir occurs to ensure maintaining response above UFLS settings.
Incentives may include performance-based market designs that provide greater payment
for greater contribution to a reliable PGFR service. They also may include rules or
minimum requirements, which are simpler to implement and accomplish the desired
result without great inefficiencies. Market designers, system operators, and the

35
J. F. Restrepo and F.D. Galiana, “Unit commitment with primary frequency regulation constraints,”
IEEE Trans. Power Syst., vol. 20, no. 4, pp. 1836-1842, Nov. 2005.
36
R. Doherty, G. Lalor, and M. O’Malley, “Frequency control in competitive electricity market dispatch,”
IEEE Trans. Power Syst., vol. 20, no. 3, pp. 1588-1596, Aug. 2005.

62
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

stakeholders of the market are likely to debate the options to find common designs that
accomplish reliability, economic efficiency, incentive-compatibility, and simplicity goals.
6.3.7 Including non-traditional resources in the PGFR market design
In North America, reliability requirements for PGFR include demand side resources
along with generation, and require all participating resources to meet common reliability
delivery definitions. Other non-traditional resources, wind turbine generators, solar
photovoltaic power, and the many forms of electric energy storage have the capability to
contribute to overall PGFR as well. PGFR responsibility is not assigned to traditional
generation alone. There are many characteristics of these technologies that impact the
scheduling and pricing of the service that require consideration.
Demand response for PGFR typically includes loads that curtail, usually automatically
via under-frequency relays, when the local frequency falls below some level. The level at
which this occurs is often regional, but is at a point above the highest setting of
involuntary UFLS in the interconnection. Typically, the total response is independent of
frequency deviation, as long as the deviation is greater than the threshold. This is like a
governor deadband followed by vertical droop curve. However, future demand response
may include load providing synthetic continuous PGFR similar to turbine-generator
governor response with shorter delay times. Maximum and minimum power become the
high and low limits for frequency response in each direction for both loads and
generators. In addition, regional rules rather than the frequency deviation provides the
basis for sustainability of the requirement.
Other technologies like wind power, solar power, and electric energy storage are
converter-based resources not synchronous to the grid. In addition, they may have
additional limits on the ability to provide PGFR. This includes capacity limits, in the
case of wind or solar, and energy limits, in the case of energy storage. If wind speed or
solar irradiance is reduced, the amount of PGFR provided by wind and solar is reduced
respectively. Likewise, if an energy storage resource has limited energy charged, it may
not be able to sustain its PGFR response for the required time frames. It is important to
consider these limitations while also valuing the benefits of these resources. As shown
earlier in this report, these converter-based resources have much lower time delays than
thermal or hydro generation when providing PGFR. This shows the importance of
incentivizing responses for the full timeframe and set of characteristics regarding PGFR.
The ERCOT proposal accomplished this by adding a separate but related service, fast
frequency response, incentivized for providing a response, related but not identical to
settled PGFR, in its market.
Additionally, NERC modified its reliability standards to remove any reliability
requirements that differentiate between spinning and non-spinning reserves. This
modification insures demand side resources are on a level playing field with generators
with respect to the provision of reliability services. Only on the technical requirements of
the service differentiates resources (generation and load). There are no services reserved
solely for generators. Consequently, specifications for secondary reserve combine all
equations differentiating between spinning and non-spinning reserve.
63
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

6.3.8 Market Design Based on Arrested PGFR


This report recommends the use of arrested PGFR for CA compliance, since it has a more
direct relationship with avoiding UFLS as opposed to the settling frequency. The use of
arrested PGFR as the measure rather than the settling frequency changes the market
design. Appendix A of this document presents a market design based on arrested PGFR.
The PGFR at the frequency nadir becomes the required measure. This change requires
PMUs to measure the power flows for providers of the arrested PGFR service and on the
tie lines for CAs on an interconnection.
In this design, market participants offer power capabilities deliverable at the time of the
nadir following a frequency event. This would require market participants to estimate the
time of the nadir when making their offers, and they would be held responsible for
meeting their offered commitments at the time of the nadir. The design basis for this
market assumes the nadir time is dependent on the size of the event and is relatively
stable from event to event; and, that market participants understand their PGFR
sufficiently to estimate this delivery time.
This change passes some of the complexity associated with estimating the PGFR timing
for their equipment to market participants. This significantly simplifies the solution from
the market operator perspective. This market design eliminates the need for the market
clearing software to set limits on deadbands, time delays, or any other parameters that
affect the response delivered at the nadir. These parameters are under the control of the
market participant. This eliminates the need of the market to measure these parameters to
assure that the participant’s parameters conform to market rules. It relates directly to
reliability since reliability is directly tied to the frequency at the nadir. If there is
withdrawal of PGFR that results in a secondary nadir after the primary nadir, those
participants that fail to maintain their PGFR until frequency is restored will have their
measurement of PGFR reduced. Due to their withdrawal, their contribution at this
secondary nadir will affect their measurement of arrested PGFR contribution.
Finally, this market design enables the definition of all real energy ancillary services as
capacity capable of being delivered over a stated time. PGFR requires its capacity
delivery prior to reaching the frequency nadir. Secondary reserves are required to be
delivered within recovery times (e.g., 10-15 minutes), and replacement reserves during
the replacement time frames (30-105 minutes). This creates a standard methodology for
defining reserve type services that can extend to services such as flexibility services
currently under investigation. The design is similar to Australia’s which focuses on a
standard 6-second reserve service.
Equation (2) from Appendix A shows the requirement changes. When a limiting event
occurs on a single CA interconnection equal to most sever single contingency (MSSC),
the nadir occurs at tnadir. The new inequality (2) shows that at tnadir the balancing inertia is
equal to zero. Therefore, the PGFR contribution from generation and participating load,
plus the load damping contribution, plus the insufficient reserve contribution must be
greater than or equal to the PGFR requirement for the area MSSC. Note that in CAs that

64
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

are part of larger interconnections this MSSC may be a pro-rated portion of the Resource
Contingency Protection Criteria37 (RCPC).
𝑁𝐺
∆ƒ 𝑚𝑎𝑥 0𝑅𝑒𝑞
∑ 𝑃10𝑖,𝑡 𝑛𝑎𝑑𝑖𝑟 + 𝐷 ∗ 𝐿𝑡 𝑛𝑎𝑑𝑖𝑟 ∗ + 𝐼𝑅𝑡 𝑛𝑎𝑑𝑖𝑟,𝑃10𝑅𝑒𝑞 ≥ 𝑃1𝐴 ≥ 𝑀𝑆𝑆𝐶 (2)
ƒ0
𝑖=1

From Appendix A inequality (46b) is an alternative form of inequality (3b) that addresses
the specific time of the nadir. It also indicates how MSSC relates to the total settled
PGFR when an event of MSSC magnitude occurs. The ratios in (47b) indicate at the limit
of MSSC the required PGFR is equal to the settled PGFR. This shows the steady state
components of PGFR bring no additional value to the solution.
0𝑅𝑒𝑞
𝑃1𝐴 𝑃1𝑠𝑠
≥ (47b)
𝑀𝑆𝑆𝐶 𝑀𝑆𝑆𝐶
6.3.9 Monitoring of Performance
Current energy management system data collection methods will not support control area
measurements necessary to determine the contributions to the frequency nadir. This
market design requires better measurement infrastructure such as phasor measurement
units for implementation. Current NERC standards resolve this limitation by using a
static ratio of frequency at steady state to the frequency at the nadir and settled PGFR.
This determines the amount of PGFR and reserve to support the response at steady state
inferring the required response and reserve at the nadir. More precise measurement may
be required to provide a direct incentive for response before and during the nadir. If the
provided response is inferior to the offered response, penalties can be determined by
reviewing the monitored data during frequency events. In addition, payments previously
made to these lower performing resources for holding capacity for frequency events that
did not occur, may be subject to claw-back of revenue for the more appropriate nadir
response.

6.4 Summary and Best Practices


This Task Force recognizes that combination of technical and policy-based decisions
determines market designs. Each market has different characteristics that require
different designs to achieve the best solution to their system. The Task Force
recommends, through this report and the 2007 report, that incentives for the PGFR
service are beneficial for ensuring sufficient PGFR in a fair, reliable, and economically
efficient manner. We believe a PGFR ancillary service market is the best option for
ensuring sufficient capacity and response. This market option compensates resources
based on incentives that drive the market participants to provide the individual
contributions leading to reliable and economic PGFR. All characteristics of PGFR must

37
In North America the equivalent of the MSSC for an interconnection with multiple CAs is called the
Resource Contingency Protection Criteria. The RCPC is apportioned among the CAs for an
interconnection while each CA has its own MSSC. The apportioned value would be used as the MSSC in
equations (2) and (47b).
65
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

be incentivized to ensure a reliable system. This can either be done through specific
incentives for improved attributes and minimum rules for characteristics deemed
satisfactory, or through a market design that incentivizes all by using arrested PGFR. For
this response to be properly incentivized and scheduled in the long-term, it requires
sufficient monitoring, most likely with PMU technology.

7 Developing a Path to the Future


This chapter describes a path to the future based on a lateral system architecture.38,39 This
architecture unifies the integration and interaction of most participants with the grid
through the synchronization mechanism of synchronous machines. Primary Frequency
Control of Future Power Systems40 discusses this approach to control. With this, most
participants, including both supply and load, connect to the grid through synchronous
machines (SM) or through power electronic converters controlled as virtual synchronous
machines (VSM), also called cyber synchronous machines (CSM).38,39 They all become
active PGFR contributors, significantly improving system stability.

7.1 Paradigm Change of Power Systems


Currently, large centralized facilities dominate the generation of electricity and only a
small number of large generators maintain system stability by regulating the balance
between generation and demand. Most loads in the system do not actively take part in
system regulation, either. But, the landscape of power systems is rapidly changing.
Various non-synchronous distributed energy resources (DER), including renewables,
electric vehicles and energy storage systems, are being connected to power systems.
Moreover, the expectation is most loads that do not currently contribute to the regulation
of system stability will take part in system regulation in the future. The number of active
participants in a power system, including both load and supply, is rapidly growing and
could easily reach millions, even hundreds of millions. If these participants fail to
contribute to control, this could lead to a significant reduction of PGFR.
Apparently, the current centralized control paradigm is no longer feasible for future
power systems with millions of relatively small generators and flexible loads. The
industry needs a new paradigm. Adding a communication and information network into
power systems, hence the birth of smart grids, emerged as a potential solution to make
power systems more efficient, more resilient to threats, and friendlier to the environment.
However, this does not solve the problem of how these generators and flexible loads
interact with the grid at the physical low-level. Low-level control relying on the

38
Q. C. Zhong, "Power Electronics-enabled Autonomous Power Systems: Architecture and Technical
Routes," in IEEE Transactions on Industrial Electronics, vol. 64, no. 7, pp. 5907-5918, July 2017,
doi: 10.1109/TIE.2017.2677339.
39
Q. C. Zhong, Power Electronics-Enabled Autonomous Power Systems: Next Generation Smart Grids,
Wiley – IEEE Press, 2017.
40
Q. C. Zhong, "Primary Frequency Control of Future Power Systems," in IEEE Smart Grid Newsletter,
June 2017, available at http://smartgrid.ieee.org/newsletters/june-2017/primary-frequency-control-of-
future-power-systems.
66
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

communication infrastructure could lead to serious concerns about reliability. If the


communication network breaks down, the whole power system could crash. Moreover,
when the number of participants reaches a certain level, managing the communication
network is itself a challenge. While it is obvious that a communication network brings
many benefits to the operation and management of power systems, there is a need to
clarify its boundary.

7.2 Fundamental Challenge Faced by Power Systems


Facilitating the paradigm change for power systems creates a need to understand the
fundamental challenge behind the scene.
On the supply side, many DERs are connecting to power systems, with most connecting
through power electronic converters. For example, wind turbines generate electricity at
variable speeds, which means the electricity generated is not compatible with the grid.
This requires power electronic converters to control the generation and interaction with
the grid. Solar panels generate DC electricity requiring its conversion into AC electricity
to make it compatible with the grid as well. Similarly, electric vehicles and energy
storage systems require power electronic converters to interact with the grid, also.
More and more power electronic converters, such as HVDC (high-voltage DC) links and
FACTS (flexible AC transmission systems) devices, connect to transmission and
distribution networks too. These are used to electronically rather than mechanically
control future power systems, reducing power losses and improving controllability.
On the load side, although there are many different loads, according to the US Electric
Power Research Institute, there are mainly four types. These are: motors that consume
over 50% of the electricity, Internet devices that consume over 10%, lighting devices that
consume about 20%, and other loads that consume the remaining 20%. It is well known
that variable-speed motor drives need power electronic converters to convert AC into DC
at the front-end. In the future, to improve efficiency, the 50% of electricity currently
consumed by motors will be consumed by power electronic rectifiers supplying those
motors. Internet devices consume DC with power electronic converters providing the
10% of electricity consumed by Internet devices. As to the 20% of electricity consumed
by lighting devices, there is a clear trend in the lighting market to adopt LED lights,
which also have power electronic converters at the front end. Looking to the future,
power electronic converters will consume most of the electricity, whatever the end use.
Consequently, future power systems will be power electronics-based, instead of electric
machines-based, with millions of relatively small, non-synchronous participants on the
supply side, inside the network, and on the load side.

7.3 Unified Architecture for Future Power Systems


Although these participants are heterogeneous, they are all integrated with the
transmission and distribution network through power electronic converters that convert
electricity between AC and DC. Moreover, the power electronic converters can control

67
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

their connected devices using the intrinsic synchronization mechanism of synchronous


machines (SM). Such power electronic converters are virtual synchronous machines
(VSM) or cyber synchronous machines (CSM). Fig 41 illustrates this unified architecture
for future power systems.

Fig. 41. Architecture for Future Power Systems Unified by the Synchronization
Mechanism of Synchronous Machines38,39

Integrating all conventional power plants, including coal-fired, hydro and nuclear power
plants, to the transmission and distribution network through SM requires no major
changes. At the same time, all DER, with controlling power electronic inverters
interfacing with the grid, behave as VSM, more specifically, as virtual synchronous
generators. All loads that have rectifiers at the front-end also behave as VSM, more
specifically, as virtual synchronous motors. For HVDC links, the power electronic
converters at both ends perform as VSM, one as a virtual synchronous generator and the
other as a virtual synchronous motor. As a result, all active participants in a future power
system41 make contributions to system stability through their intrinsic synchronization

41
Q.-C. Zhong, "Synchronized and Democratized Smart Grids To Underpin The Third Industrial
Revolution", in Proceeding of the 2017 IFAC World Congress, Toulouse, France, July 2017.

68
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

mechanism in a synchronized and democratized manner, achieving the paradigm change


of power systems from centralized control to democratized interaction.
Since all the SM and VSM have the intrinsic synchronization mechanism, there is no
need to rely on additional communications to achieve low-level control. The
communication network in a power system is no longer neeeded for low-level control and
can be released to focuse on high-level functions, e.g. SCADA and market operations.
This also enhances the cybersecurity of the system because of reduced access to low-
level controllers by potential malicious attackers. It is worth highlighting that this
architecture turns all players into active and responsible players to maintain system
stability and achieves continuous demand response. This prevents some customers from
suffering complete loss of electricity. Instead, all players make a small, often negligible,
contributions in the event of frequency disturbances, which improves the quality of
service (QoS). This synchronization mechanism of SM is the fundamental mechanism
underpinning the growth and operation of power systems for over 100 years. Therefore,
the paradigm change of power systems from today’s centralized control to tomorrow’s
democratized interaction can be achieved gradually and smoothly.
Two technical routes currently are available to implement VSM.38,39 The basis for one is
the synchronverter technology and the other is the robust droop control technology. A
synchronverter embeds the mathematical model of synchronous machines into its
controller designed for converters with inductive impedance. The robust droop
controller38,39,42 also implements the intrinsic synchronization mechanism of synchronous
machines with the clear advantage of being universal to converters with different types of
impedances.

7.4 PGFR in Future Power Systems


This section discusses the components necessary to assure adequate PGFR in future
power systems.
7.4.1 PGFR Contributions from both Supply and Load
For a future power system shown in Fig 41, the intrinsic synchronization equips all
previously non-synchronous active participants with the mechanism of synchronous
machines, both generators and loads. The generators are all turned into frequency-
responsive synchronous participants and provide (virtual) kinetic balancing inertia to
improve frequency stability, in the same way as conventional synchronous machines.
This stops the trend of decreasing balancing inertia due to the penetration of DER. The
loads interfaced through power electronic converters also become frequency-responsive
and automatically respond to the change of frequency, quickly regulating power
consumption without impacting user experience. This frequency-responsive
characteristic makes these customer-side loads play a similar role of PGFR on the supply
side. It stops the trend of decreasing load damping because of the increasing adoption of
42
Q.-C. Zhong, “Robust droop controller for accurate proportional load sharing among inverters operated
in parallel”, IEEE Trans. on Industrial Electronics, vol.60, no.4, 1281-1290, 2013.

69
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

systems controlled independently of frequency. As shown in Fig 28, increasing kinetic


balancing inertia reduces the arrested frequency change, although, it does not reduce the
settled frequency change, and therefore, contributes to PGFR. A VSM can also
automatically control the change in power consumption according to the change in
voltage, for example, by adopting the robust droop control technology.38,39,42
7.4.2 Droop
Section 3 discusses how droop plays an important role in PGFR. The droop specifies the
slope of the PGFR in terms of the final steady state and the magnitude of delivered PGFR
depends on the droop settings of individual participants. For the system in Fig 41, the
VSM associated with different types of supply and load, according to their nature, can be
configured with different droop coefficients addressing critical levels, economic benefits,
frequency conditions and other factors. For example, a wind generator may not be able
to provide enough PGFR for a low frequency condition but can easily provide PGFR for
a high frequency condition. It is normally not a problem to completely shut down HVAC
systems for several minutes or to shift washing machine and dishwasher use by a couple
of hours, even up to 24 hours. For economic reasons, it is also possible to set a large
droop, e.g. 10% for small frequency conditions and a small droop, e.g. 3%, for large
frequency conditions. In this way, achieving the maximum PGFR without significantly
affecting the quality of service is possible in many cases. Moreover, shifting the peak
load reduces the peak/normal load ratio and the PGFR needed as well.
7.4.3 Fast Action Without Delay
VSMs are inherently power electronics converters and act upon frequency changes
without delay. Since any delayed response increases the maximum frequency change in
the event of disturbances, as illustrated in Fig 27, the fast action of VSM reduces the
amount of balancing inertia required before the frequency change is arrested in the event
of disturbances. For short frequency spikes, the impact on the system is small because of
the relatively large system balancing inertia. VSM will act upon the frequency spikes
immediately but will return to normal very quickly after the frequency spikes, as well.
7.4.4 Reconfigurable Virtual Inertia
The kinetic balancing inertia of a conventional synchronous machine does not vary. To
some extent, the virtual inertia of a VSM is reconfigurable. This provides more
flexibility for PGFR. Moreover, the virtual inertia of a VSM does not involve estimation
of system frequency or rate of change of frequency. Therefore, the instability caused by
synthetic balancing inertia does not occur. Another important factor related to instability
is damping. The damping of a VSM can be reconfigured to improve stability.
7.4.5 Continuous PGFR
A prominent feature of the future power systems illustrated in Fig 41 is that some active
participants, both supply and load, can provide PGFR. They can continuously adjust the
output or the intake according to the system frequency in an autonomous manner. Fig 42
illustrates this capability of a VSM connected to the U.K. grid. The real power output
70
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

changed autonomously according to the changing frequency, demonstrating excellent


PGFR. Some non-essential loads can be configured to provide this continuous PGFR to a
very high level. PGFR supplied by operating air conditioning systems without affecting
the comfortability and pumping systems without affecting the overall system
performance provides an example.

Fig. 42. Continuous PGFR capability of a VSM connected to the U.K. grid38,39

7.5 Summary
Section 7 presents a path to the future for PGFR with a lateral system architecture,
enabling all conventional generators, renewable generators and flexible loads to actively
participate in system PGFR. While unprecedented challenges face current power
systems, releasing the potential of most participants could address the PGFR issue
holistically. The key is to unify the integration and interaction with the grid through the
synchronization mechanism of synchronous machines so they all become active PGFR
contributors. The conventional power plants contribute to PGFR in the usual way
without much change. The non-synchronous DER and flexible loads behave as VSM
contributing to PGFR as well. The active contribution of flexible loads to PGFR will
significantly enhance PGFR and system stability. Practically, the contributions to
balancing inertia and PGFR likely increase rather than decrease as normally perceived.
This significantly reduces the need for secondary frequency control. As a matter of fact,
although the paradigm change for power systems, from centralized generation to
distributed generation and from centralized control to democratized interaction, brings
unprecedented challenges, it also brings numerous opportunities to the power industry.

71
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

For example, the fast speed and re-configurability of power electronic converters (VSM)
could significantly enhance the flexibility of future power systems. The droop
coefficients for both suppliers and loads can take into consideration different factors and
the virtual inertia can be reconfigured as well. Another prominent feature is that both
suppliers and loads can act as continuous PGFR themselves, adjusting the power
exchanged with the grid autonomously according to the grid condition as represented by
frequency. Ultimately, this will likely lead to power-electronics-enabled autonomous
power systems.38,39

8 Conclusions and Recommendations


Section 8 contains two subsections. The first section reports on the conclusions and
findings of the Task Force based on material presented in this report and discussed during
Task Force meetings. The second section presents the recommendations of the Task
Force and addresses some of the outstanding issues reported in our conclusions.

8.1 Conclusions
PGFR contains significantly greater complexity than presented in the initial IEEE Special
Publication 07TP180. This complexity along with changes in the mix of generation
resources requires this document to address a wider scope than initially envisioned.
(1) Technical analysis confirms the current security adequacy of PGFR,
previously assumed by expert judgment.
Technical analysis performed on the North American interconnections
confirms there is sufficient PGFR on these interconnections to assure an
adequate level of security and reliability.
(2) “If some portions of an interconnection require greater PGFR due to their
topology and risk of separation, how does this greater PGFR affect
reliability when the interconnection experiences a disturbance and remains
intact? Can imbalances in PGFR result in increased risk during
disturbances?”
Studies of the European power system demonstrate limits to the
redistribution of PGFR among the CAs without endangering the network
while the CAs achieve the goals of reducing PGFR procurement costs and
reducing post-contingency line loadings.
Studies of the U.S. Western Interconnection transferring PGFR from one
region of the interconnection to another indicate no significant change in the
frequency nadir for contingency events or negative stability impact. The
results indicate no apparent limits on the location of PGFR. This confirms
the ability of the CAs on the interconnection to implement an unrestricted
exchange of PGFR and an unrestricted market in PGFR.

72
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Penetration of non-traditional generation resources and the development of synthetic


components as contributors to PGFR expands the scope of this report. It includes
consideration of additional issues not contained in the initial special publication.
Section 3 investigates sensitivities and problems associated with the implementation of
synthetic alternatives including understanding of:
(3) Governor droop settings;
(4) The effect of governor deadband settings;
(5) The effect of time delays in governor response;
(6) The effect of kinetic balancing inertia on interconnection reliability; and,
(7) Synthetic balancing inertial response proportional to rate of change of
frequency.
Additionally, Section 3 addresses potential stability issues associated with
synthetic balancing inertia. It presents two studies to address the problem.
The simple power and energy balancing model performing the sensitivity
simulations indicates synthetic balancing inertia based on RoCoF with
penetrations greater than 50% could result in frequency instability. The
interaction between kinetic balancing inertia and synthetic balancing inertia
causes this potential instability and large power swings. These large power
swings also occur for proportions of synthetic balancing inertia below 50%
but instead of increasing to instability, they dampen out.
In one test system, adding synthetic balancing inertia alone induced small
signal instability (i.e., the linearized system displayed one or more
eigenvalues with strictly positive real part). If the locations exercising
synthetic balancing inertia control also apply simple damping control (i.e., a
linear feedback proportional to frequency error) this corrects the instability.
However, the benefit of added damping was not monotonic; simply put, the
damping gains could not be increased indefinitely. If damping gains
increased sufficiently, critical eigenvalues became less well damped, and
ultimately again induced instability. These case studies suggest that the
addition of synthetic balancing inertia to a power system, possibility causes
adverse effects on small signal stability in addition to improving
performance of the frequency nadir. In other words, these results
underscore the point made in the previous section—requiring detailed
studies to verify the stabilizing or de-stabilizing effects of synthetic
balancing inertia control that is proportional to time rate of change of
frequency.
Section 3 continues with investigations of sensitivities including the understanding of:
(8) Step load interruption based on frequency without restoration;
73
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

(9) Step load interruption based on frequency with restoration based on time
delay;
(10) Step load interruption based on frequency with restoration based on
frequency recovery; and,
(11) Synthetic continuous PGFR proportional to frequency error.
This discussion also includes an example how the reduction in kinetic
balancing inertia, that occurs as synthetic continuous PGFR replaces PGFR
from synchronous generators, affects the frequency profile. This example
highlights the transition from over-damping through critical-damping to
under-damping of the resulting frequency characteristic.
Section 4 presents examples of on-line monitoring of PGFR and implementation of
PGFR reliability standards with specific measurement methods. Additional issues
addressed are:
(12) The TF should make recommendations on possible ways of performing on-
line monitoring to estimate the governor steady state droop and response of
turbine-generator governor action; and,
(13) Consider assuring reliability by including arrested PGFR as the basis for
reliability standards.
Methods used around the world are summarized but generally fail to provide
specific measurement methods that apply to on-line operations. However,
reports from the Swiss Grid and Terna indicate the status of work
developing new measurement methods.
This report addresses three methods to measure PGFR during actual system
operation. The ERCOT Interconnection implemented a method to measure
settled PGFR from generators in a reliability standard. NERC implemented
a method to measure settled PGFR, including all resources, both generation
and load, in North America in a reliability standard. Finally, this report
considers the advantages of implementing a measurement method based on
the frequency nadir, arrested PGFR, in future reliability standards.
Section 5 presents a discussion of the economics associated with providing PGFR from
alternative resources. It then discusses the costs associated with providing PGFR from
both generation and load including:
(14) The costs associated with providing PGFR,
(15) The demand side of PGFR effects on costs incurred,
(16) The demand-side participation effects on interconnection economics,
(17) The rewarding of loads for providing this service,
74
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

(18) The measuring the provision of this service, and,


(19) The economics of synthetic PGFR.
Section 5 concludes, since the price of electricity is 1% or less than the value of
electricity, changes in customer use must be implemented in a manner unobservable to
the customer or takes little effort and does not require significant changes in customer
lifestyle.
Section 6 presents compensation and incentives for PGFR including examples from
markets around the world. These examples include many ways of including PGFR into
markets and other compensation methodologies. Market designs for the inclusion of
PGFR outside of North America add from four to six new services to represent the PGFR
ancillary service. Previous North American market designs add from two to six new
services to represent PGFR. A new market design presented in Appendix A presents
arrested PGFR to the market by replacing two previous services with two new services to
represent PGFR. Appendix A presents alternative wholesale market designs with specific
measurement methods including:
(20) Balancing inertia, and settled PGFR,
(21) Balancing inertia, and settled PGFR with simplified reserve categories, and,
(22) Balancing inertia, and arrested PGFR with simplified reserve categories.
These designs indicate that simplifying the reserve categories and using
arrested PGFR as the basis provides a significantly simpler market design.
It also allows each party supplying PGFR to this market to individually
optimize the parameters for the resources providing the response.
Section 7 provides one alternative path to future system design and operations. It
discusses advantages associated with providing PGFR from both generation and load
base on this alternative path including:
(23) Developing a feasible path to future system operations; and,
(24) Determining the types of load that could provide PGFR.

8.2 Recommendations
Based on the results, this task force recommends the following:
(1) The industry should move forward with the implementation of new
measurement methods (PMUs or equivalent) that would enable the direct
measurement of the frequency nadir and arrested PGFR at the CA level or
lower. This will provide information not currently available about the need
for and the provision of PGFR.

75
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

(2) PGFR should be measured using arrested PGFR directly indicating the
contribution to limiting the nadir (for loss of supply responses) in those
regions that are experiencing or are likely to experience PGFR withdrawal.
This will provide a direct measure indicating accurate contributions to
PGFR in those cases where PGFR withdrawal is likely to or is occurring.
(3) When implementing synthetic PGFR products, synthetic continuous PGFR
should be favored over synthetic balancing inertia. This approach will
mitigate risks of small signal instability resulting from the inclusion of
synthetic balancing inertia based on RoCoF.
(4) The industry should look for and evaluate any instances of PGFR that
indicate significant underdamping in the frequency characteristic. As the
penetration of non-synchronous devices increases, these instances may
provide a precursor to indicate an insufficiency of balancing inertia.
(5) The arrested PGFR measure is the preferred measure for determining the
adequacy of PGFR on an interconnection. Arrested PGFR is the preferred
measure for use in reliability standards.
(6) New regulations should not affect the existing installations. It is very
expensive to modify the existing installations for compliance to new rules.
(7) Arrested PGFR should be considered in market designs that include PGFR
as a market product. The use of arrested PGFR significantly simplifies
market designs including PGFR in the market.
(8) The compensation method or market design should incentivize all
characteristics of PGFR to ensure a reliable system.
(9) The industry should continue to develop additional regulation and flexibility
resources to insure the availability of sufficient secondary control resources
to maintain appropriate ranges of frequency operations until new alternative
methods of operation are established and confirmed to be adequate.
(10) As the industry resource mix changes from SM traditional resources to VSM
renewable resources, the industry should proceed with the development and
implementation of device based VSMs to insure sufficient market
penetration of these devices.
(11) New regulations should be consistent with the manufacturers' capacity in
building complex equipment.

76
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Appendix A – Market Designs for PGFR


This Appendix reviews the status of markets for ancillary services and investigates how
using arrested PGFR with balancing inertial response instead of settled PGFR with
balancing inertial response affects market designs. It includes:

Balancing inertial response and PGFR in market designs;

Adjusting market designs for recent changes in reliability definitions; and,

Providing market designs based on arrested PGFR instead of settled PGFR.

A.1. Market Design Based on Settled PGFR


Summary of Market Designs for the Primary Frequency Response Ancillary Service—
Part I: Motivation and Design43 presents a market design based on settled PGFR. The
following equations summarize the market design suggested in the technical paper. They
will be used to compare designs.

A.1.1. Nomenclature
Generator Parameters:
Pmax Maximum generating capacity (MW)
Pmin Minimum generating capacity (MW)
Smax Maximum apparent power capacity (MVA)
RR Ramp rate for secondary reserve (MW/min)
H Inertia constant (s)
DB Governor deadband (Hz)
QSC Quick-start capability available in trec (MW)
R Droop curve (p.u. Hz/p.u. MW)
Req Equivalent droop curve (Hz/MW)
SF Shift factor contribution to line flow (unitless)
CE(*) Cost function for providing energy

CR(*) Cost function for providing reserve

43
E. Ela, V. Gevorgian, A. Tuohy, B. Kirby, M. Milligan and M. O'Malley, "Market Designs for the
Primary Frequency Response Ancillary Service—Part I: Motivation and Design," in IEEE Transactions
on Power Systems, vol. 29, no. 1, pp. 421-431, Jan. 2014. doi: 10.1109/TPWRS.2013.2264942

77
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

System Parameters and Requirements:


L Load (MW)
R Reserve demand (MW)

ƒ0 Nominal frequency (Hz)


D Load damping constant (p.u. MW/p.u. Hz)
𝑒𝑞
𝐻𝐿 Equivalent load inertia (s)
VOLL Value of lost load ($/MW-h)
VOIR Value of insufficient reserves ($/MW-h)
tnadir Time associated with frequency nadir (s)
tss Time associated with steady-state frequency (s).
trec Time required to recover to nominal frequency (min)
IReq Inertia requirement (MVAs)
P10Req PGFR capacity requirement (MW)
P1ssReq PGFR requirement at tss (MW)
P1nadirReq PGFR requirement at tnadir (MW)
P2spinReq Secondary spin reserve requirement (MW)
P2nspinReq Secondary non-spin reserve requirement (MW)

ƒ max Maximum frequency deviation (Hz)

DBmax Maximum deadband allowed (Hz)


F max Maximum amount of flow on line (MW)

ƒ ss Steady state frequency (Hz)

ƒ nadir Minimum frequency (Hz)

Variables:
P Energy Schedule (MW)
P10 Full PGFR capacity availability (MW)
P1nadir PGFR availability at nadir time (MW)

78
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

P1ss PGFR availability at steady-state time (MW)


P2spin Secondary spin reserve available (MW)
P2nspin Secondary non-spin reserve available (MW)
Pres General reserve schedule (MW)
u Integer variable of unit being on line [0,1]

Integer variable of whether governor is enabled [0,1]

Integer variable of having full head room for PGFR [0,1]

Sensitivity factor of PGFR capacity with time (unitless)


LD Load damping amount (MW)
ENS Energy not served (MW)
IR Insufficient reserve (MW)
Indices and Sets:
i, NG Generator index, set of generators
t, NT Time interval index, time interval set
n, NB Bus index, set of buses
r, NR Reserve type index, set of reserve types
l, NL Branch/transmission line, set of lines
Gnogov Set of generators with no governors enabled
A Area or market region
Dual Variables:

Dual variable for energy balance equality constraint

Dual variable for line flow inequality constraint

Dual variable for reserve balance inequality constraint


max min
, Dual variables for capacity constraints

79
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

A.1.2. Market Scheduling


The following 23 equations define the market scheduling. Equation (1) shows the
objective function, the minimization of total cost including the value of lost load and
insufficient reserve.
𝑁𝑇 𝑁𝑅

𝑀𝑖𝑛 ∑ {𝑉𝑂𝐿𝐿 ∗ 𝐸𝑁𝑆𝑡 + ∑ 𝑉𝑂𝐼𝑅𝑟 ∗ 𝐼𝑅𝑡,𝑟


𝑡=1 𝑟=1
𝑁𝐺 𝑁𝑅
𝑟𝑒𝑠
+ ∑ [𝐶𝑖𝐸 (𝑢𝑖,𝑡 , 𝑢𝑖,𝑡−1 , 𝑃𝑖,𝑡 ) + ∑ 𝐶𝑖𝑅 (𝑃𝑖,𝑡,𝑟 , 𝛾𝑖,𝑡 )]} (1)
𝑖=1 𝑟=1

Inequality (2) constrains balancing inertia to a minimum required to limit rate of change
of frequency (RoCoF) and under frequency load shedding (UFLS).
𝑁𝐺
𝑒𝑞 𝑅𝑒𝑞
∑{𝑢𝑖,𝑡 ∗ 𝐻𝑖 ∗ 𝑆𝑖𝑚𝑎𝑥 } + 𝐻𝐿 ∗ 𝐿𝑡 ≥ 𝐼𝐴 − 𝐼𝑅𝑡,𝐼𝑛𝑒𝑟𝑡𝑖𝑎 ∀𝑡 ∈ 𝑁𝑇 (2)
𝑖=1

Inequality (3) constrains PGFR capacity to a minimum necessary to limit the nadir to a
value above UFLS.
𝑁𝐺
0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
∑ 𝑃10𝑖,𝑡 ≥ 𝑃1𝐴 − 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃10𝑅𝑒𝑞 ∀𝑡 ∈ 𝑁𝑇 (3)
ƒ0
𝑖=1

Equation (4) creates an equivalent droop curve to convert to units of Hz/MW.

𝑒𝑞 𝑅𝑖 ∗ ƒ0
𝑅𝑖 = ∀𝑖 ∈ 𝑁𝐺 (4)
𝑃𝑖𝑚𝑎𝑥

Inequality (5) creates PGFR capacity from the droop curve.


𝜒𝑖,𝑡
𝑃10𝑖,𝑡 ≤ 𝑒𝑞 (∆ƒ
𝑚𝑎𝑥
− 𝐷𝐵𝑖 ) − 𝑃𝑖𝑚𝑎𝑥 ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (5)
𝑅𝑖

Inequality (6) sets the integer variable, , indicating head room is available to provide all
the PGFR.
1
𝑃𝑖,𝑡 + 𝑒𝑞 (∆ƒ
𝑚𝑎𝑥
− 𝐷𝐵𝑖 ) ≥ 𝑃𝑖𝑚𝑎𝑥 𝛾𝑖,𝑡 − 𝑃𝑖𝑚𝑎𝑥 𝜒𝑖,𝑡 ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (6)
𝑅𝑖

Inequality (7) ensures that PGFR of resources with head room and enabled PGFR will
equal their droop curve relationship.
1
𝑃10𝑖,𝑡 ≥ + 𝑒𝑞 (∆ƒ
𝑚𝑎𝑥
− 𝐷𝐵𝑖 ) − 𝑃𝑖𝑚𝑎𝑥 (1 − 𝛾𝑖,𝑡 ) ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (7)
𝑅𝑖

80
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Inequality (8) requires that a unit be on-line to enable its governor.

𝛾𝑖,𝑡 ≤ 𝑢𝑖,𝑡 ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (8)

Equation (9) sets equal to zero for units that are in the Gnogov set.

𝛾𝑖,𝑡 = 0 ∀𝑖 ∈ 𝐺𝑛𝑜𝑔𝑜𝑣 , ∀𝑡 ∈ 𝑁𝑇 (9)

When the governor deadband is too large, inequality (10) sets equal to zero.

𝐷𝐵𝑖 ≤ 𝐷𝐵𝑚𝑎𝑥 + ƒ0 (1 − 𝛾𝑖,𝑡 ) ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (10)

The above information drives a system simulation to insure the provision of sufficient
PGFR at tnadir and tss.
𝑁𝐺
𝑠𝑠 𝑠𝑠𝑅𝑒𝑞
∑ 𝑃1𝑖,𝑡 ≥ 𝑃1𝐴 − 𝐿𝐷 𝑠𝑠 ∀𝑡 ∈ 𝑁𝑇 (11)
𝑖=1

𝑁𝐺
𝑛𝑎𝑑𝑖𝑟𝑅𝑒𝑞
∑ 𝑃1𝑛𝑎𝑑𝑖𝑟
𝑖,𝑡 ≥ 𝑃1𝐴 − 𝐿𝐷𝑛𝑎𝑑𝑖𝑟 ∀𝑡 ∈ 𝑁𝑇 (12)
𝑖=1

The fourth characteristic insures ss and nadir meet minimum requirements during the
simulation.
𝑠𝑠
𝑠𝑠
𝑃1𝑖,𝑡
𝛼𝑖,𝑡 =[ ] from Dynamic Sim2 ∀𝑖 ∶ 𝛾̂𝑖,𝑡 = 0, ∀𝑡 ∈ 𝑁𝑇 (13)
𝑃10𝑖,𝑡

𝑛𝑎𝑑𝑖𝑟
𝑃1𝑛𝑎𝑑𝑖𝑟
𝑖,𝑡
𝛼𝑖,𝑡 =[ ] from Dynamic Sim2 ∀𝑖 ∶ 𝛾̂𝑖,𝑡 = 0, ∀𝑡 ∈ 𝑁𝑇 (14)
𝑃10𝑖,𝑡
𝑠𝑠
𝑠𝑠
𝑃1𝑖,𝑡
𝛼𝑖,𝑡 =[ ] from Dynamic Sim1 ∀𝑖 ∶ 𝛾̂𝑖,𝑡 = 1, ∀𝑡 ∈ 𝑁𝑇 (15)
̂ 0𝑖,𝑡
𝑃1

𝑛𝑎𝑑𝑖𝑟
𝑃1𝑛𝑎𝑑𝑖𝑟
𝑖,𝑡
𝛼𝑖,𝑡 =[ ] from Dynamic Sim1 ∀𝑖 ∶ 𝛾̂𝑖,𝑡 = 1, ∀𝑡 ∈ 𝑁𝑇 (16)
̂
𝑃10𝑖,𝑡

If the initial simulation does not meet reliability requirements and additional simulations
are required, Equations (17) and (18) apply.
𝑁𝐺
𝑠𝑠 𝑠𝑠𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
∑ 𝛼𝑖,𝑡 ∗ 𝑃10𝑖,𝑡 ≥ 𝑃1𝐴 𝑠𝑠
− 𝛼𝐿,𝑡 ∗ 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃1𝑠𝑠 ∀𝑡 ∈ 𝑁𝑇 (17)
ƒ0
𝑖=1

81
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

𝑁𝐺
𝑛𝑎𝑑𝑖𝑟 𝑛𝑎𝑑𝑖𝑟𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
∑ 𝛼𝑖,𝑡 ∗ 𝑃1𝑛𝑎𝑑𝑖𝑟
𝑖,𝑡 ≥ 𝑃1𝐴 − 𝑛𝑎𝑑𝑖𝑟
𝛼𝐿,𝑡 ∗ 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃1𝑛𝑎𝑑𝑖𝑟 ∀𝑡 ∈ 𝑁𝑇 (18)
ƒ0
𝑖=1

Stability, the fifth characteristic, only allows proportional droop curves.


The sixth characteristic assumes that there is no PGFR withdrawal prior to the recovery
of frequency. In alternative words, it assumes a constant frequency recovery to
scheduled frequency.
The seventh and last characteristic insures sufficient secondary response and reserve.
𝑠𝑝𝑖𝑛
𝑃2𝑖,𝑡 ≤ 𝑢𝑖,𝑡 ∗ 𝑡 𝑟𝑒𝑐 ∗ 𝑅𝑅𝑖 ∀𝑖 ∈ 𝑁𝐺 ∀𝑡 ∈ 𝑁𝑇 (19)
𝑁𝐺
𝑠𝑝𝑖𝑛 𝑠𝑝𝑖𝑛𝑅𝑒𝑞 0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
∑ 𝑃10𝑖,𝑡 + 𝑃2𝑖,𝑡 ≥ 𝑃2𝐴 − 𝑃1𝐴 − 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃2𝑠𝑝𝑖𝑛 ∀𝑡 ∈ 𝑁𝑇 (20)
ƒ0
𝑖=1

𝑛𝑠𝑝𝑖𝑛
𝑃2𝑖,𝑡 ≤ (1 − 𝑢𝑖,𝑡 ) ∗ 𝑄𝑆𝐶𝑖 ∀𝑖 ∈ 𝑁𝐺 ∀𝑡 ∈ 𝑁𝑇 (21)
𝑁𝐺
𝑠𝑝𝑖𝑛 𝑛𝑠𝑝𝑖𝑛 𝑛𝑠𝑝𝑖𝑛𝑅𝑒𝑞 𝑠𝑝𝑖𝑛𝑅𝑒𝑞 0𝑅𝑒𝑞
∑ 𝑃10𝑖,𝑡 ∗ 𝑃2𝑖,𝑡 ∗ 𝑃2𝑖,𝑡 ≥ 𝑃2𝐴 + 𝑃2𝑖,𝑡 − 𝑃1𝐴
𝑖=1
∆ƒ 𝑚𝑎𝑥
−𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃2𝑛𝑠𝑝𝑖𝑛 ∀𝑖 ∈ 𝑁𝐺 ∀𝑡 ∈ 𝑁𝑇 (22)
ƒ0
The last equation (23) insures there is enough capacity to supply the required reserves.
𝑠𝑝𝑖𝑛
𝑃𝑖,𝑡 + 𝑃10𝑖,𝑡 + 𝑃2𝑖,𝑡 ≤ 𝑢𝑖,𝑡 ∗ 𝑃𝑖𝑚𝑎𝑥 ∀𝑖 ∈ 𝑁𝐺 ∀𝑡 ∈ 𝑁𝑇 (23)

A.1.3. Market Pricing


The following 20 equations define the market pricing. Equation (24) shows the
Lagrangian function.
L=
𝑁𝑇 𝑁𝑅 𝑁𝐺 𝑁𝑅
𝑟𝑒𝑠
∑ {𝑉𝑂𝐿𝐿 ∗ 𝐸𝑁𝑆𝑡 + ∑ 𝑉𝑂𝐼𝑅𝑟 ∗ 𝐼𝑅𝑡,𝑟 + ∑ [𝐶𝑖𝐸 (𝑃𝑖,𝑡 ) + ∑ 𝐶𝑖𝑅 (𝑃𝑖,𝑡,𝑟 )]
𝑡=1 𝑟=1 𝑖=1 𝑟=1
𝑁𝐺 𝑁𝐵 𝑁𝐿 𝑁𝐵

− 𝜆𝑡 ∗ (∑ 𝑃𝑖,𝑡 − ∑ 𝐿𝑛,𝑡 − 𝐸𝑁𝑆𝑡 ) − ∑ (𝑢𝑖,𝑡 ∗ ∑ 𝑆𝐹𝑛,𝑙 ∗ (𝑃𝑛,𝑡 − 𝐿𝑛,𝑡 ) − 𝐹𝑙𝑚𝑎𝑥 )


𝑖=1 𝑛=1 𝑙=1 𝑛=1
𝑁𝑅 𝑁𝐺
𝑟𝑒𝑠 𝑚𝑎𝑥 𝑠𝑝𝑖𝑛
− ∑ 𝛽𝑟,𝑡 ∗ 𝐻𝑟 (𝑃𝑖,𝑡,𝑟 , 𝑅𝑡,𝑟 , 𝐼𝑅𝑡,𝑟 , 𝐿) − ∑ 𝜋𝑖,𝑡 ∗ (𝑃𝑖𝑚𝑎𝑥 − 𝑃𝑖,𝑡 − 𝑃10𝑖,𝑡 − 𝑃2𝑖,𝑡 )
𝑟=1 𝑖=1

𝑚𝑖𝑛
− 𝜋𝑖,𝑡 ∗ (𝑃𝑖,𝑡 − 𝑃𝑖𝑚𝑎𝑥 )} (24)

82
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

The Hr(*) equations represent the ancillary services inequality constraints.


𝑁𝐺
𝑛𝑎𝑑𝑖𝑟 𝑛𝑎𝑑𝑖𝑟𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
𝐻1 = ∑ 𝛼𝑖,𝑡 ∗ 𝑃10𝑖,𝑡 − 𝑃1𝐴 𝑛𝑎𝑑𝑖𝑟
+ 𝛼𝐿,𝑡 ∗ 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃1𝑛𝑎𝑑𝑖𝑟 (25)
ƒ0
𝑖=1

𝑁𝐺
𝑠𝑠 𝑠𝑠𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
𝐻2 = ∑ 𝛼𝑖,𝑡 ∗ 𝑃10𝑖,𝑡 − 𝑃1𝐴 𝑠𝑠
+ 𝛼𝐿,𝑡 ∗ 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃1𝑠𝑠 (26)
ƒ0
𝑖=1

𝑁𝐺
0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
𝐻3 = ∑ 𝑃10𝑖,𝑡 − 𝑃1𝐴 + 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃10 (27)
ƒ0
𝑖=1

𝑁𝐺
𝑠𝑝𝑖𝑛 𝑠𝑝𝑖𝑛𝑅𝑒𝑞 0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
𝐻4 = ∑ 𝑃2𝑖,𝑡 − 𝑃2𝐴 − 𝑃1𝐴 + 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃2𝑠𝑝𝑖𝑛 (28)
ƒ0
𝑖=1

𝑁𝐺
𝑛𝑠𝑝𝑖𝑛 𝑛𝑠𝑝𝑖𝑛𝑅𝑒𝑞 𝑠𝑝𝑖𝑛𝑅𝑒𝑞 0𝑅𝑒𝑞
𝐻5 = ∑ 𝑃2𝑖,𝑡 − 𝑃2𝐴 − 𝑃2𝐴 − 𝑃1𝐴
𝑖=1
∆ƒ 𝑚𝑎𝑥
+𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃2𝑠𝑝𝑖𝑛 (29)
ƒ0
Equations (30) and (31) show the Locational Marginal Price (LMP) and Ancillary Service
Clearing Price (ASCP) respectively.
𝑁𝐿
𝜕𝐿 𝑛𝑎𝑑𝑖𝑟
∆ƒ 𝑚𝑎𝑥 𝑠𝑠
∆ƒ 𝑚𝑎𝑥
𝐿𝑀𝑃𝑛 = = 𝜆 + ∑ 𝜇𝑙 ∗ 𝑆𝐹𝑛,𝑙 − 𝜆 ∗ 𝐿𝐹𝑖 − 𝛽1 ∗ 𝛼𝐿,𝑡 ∗ − 𝛽2 ∗ 𝛼𝐿,𝑡 ∗
𝜕𝐿𝑛 ƒ0 ƒ0
𝑙=1
∆ƒ 𝑚𝑎𝑥 ∆ƒ 𝑚𝑎𝑥 ∆ƒ 𝑚𝑎𝑥
−𝛽3 ∗ − 𝛽4 ∗ − 𝛽5 ∗ (30)
ƒ0 ƒ0 ƒ0
𝜕𝐿
𝐴𝑆𝐶𝑃𝑟 = = 𝛽𝑟 (31)
𝜕𝑅𝑟
Equations (32) – (36) show the ASCP for each service.
𝜕𝐿
𝐴𝑆𝐶𝑃𝑃1𝑛𝑎𝑑𝑖𝑟 = 𝑛𝑎𝑑𝑖𝑟𝑅𝑒𝑞
= 𝛽1 (32)
𝜕𝑃1𝐴

𝜕𝐿
𝐴𝑆𝐶𝑃𝑃1𝑠𝑠 = 𝑠𝑠𝑅𝑒𝑞 = 𝛽2 (33)
𝜕𝑃1𝐴

𝜕𝐿
𝐴𝑆𝐶𝑃𝑃10 = 0𝑅𝑒𝑞 = 𝛽3 + 𝛽4 + 𝛽5 (34)
𝜕𝑃1𝐴

83
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

𝜕𝐿
𝐴𝑆𝐶𝑃𝑃2𝑠𝑝𝑖𝑛 = 𝑠𝑝𝑖𝑛𝑅𝑒𝑞
= 𝛽4 + 𝛽5 (35)
𝜕𝑃2𝐴

𝜕𝐿
𝐴𝑆𝐶𝑃𝑃2𝑛𝑠𝑝𝑖𝑛 = 𝑛𝑠𝑝𝑖𝑛𝑅𝑒𝑞
= 𝛽5 (36)
𝜕𝑃1𝐴

Equations (37) – (41) show the payments for each service.

𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡,𝑃1𝑛𝑎𝑑𝑖𝑟 = 𝐴𝑆𝐶𝑃𝑡,𝑃1𝑛𝑎𝑑𝑖𝑟 ∗ 𝑃1𝑛𝑎𝑑𝑖𝑟


𝑖,𝑡 (37)
𝑠𝑠
𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡,𝑃1𝑠𝑠 = 𝐴𝑆𝐶𝑃𝑡,𝑃1𝑠𝑠 ∗ 𝑃1𝑖,𝑡 (38)

𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡,𝑃10 = 𝐴𝑆𝐶𝑃𝑡,𝑃10 ∗ 𝑃10𝑖,𝑡 (39)


𝑠𝑝𝑖𝑛
𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡,𝑃2𝑠𝑝𝑖𝑛 = 𝐴𝑆𝐶𝑃𝑡,𝑃2𝑠𝑝𝑖𝑛 ∗ 𝑃2𝑖,𝑡 (40)
𝑛𝑠𝑝𝑖𝑛
𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡,𝑃2𝑛𝑠𝑝𝑖𝑛 = 𝐴𝑆𝐶𝑃𝑡,𝑃2𝑛𝑠𝑝𝑖𝑛 ∗ 𝑃2𝑖,𝑡 (41)

Equations (42) and (43) show the pricing and payment for balancing inertia.
𝐿𝑃
𝐴𝑆𝐶𝑃𝑖𝑛𝑒𝑟𝑡𝑖𝑎 = 𝛽𝑖𝑛𝑒𝑟𝑡𝑖𝑎 (42)
𝑖𝑛𝑒𝑟𝑡𝑖𝑎
𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡 = 𝐴𝑆𝐶𝑃𝑡𝑖𝑛𝑒𝑟𝑡𝑖𝑎 ∗ {𝑢𝑖,𝑡 ∗ 𝐻𝑖 ∗ 𝑆𝑖𝑚𝑎𝑥 } (43)

A.1.4. Bidding Structure and Settlements


The following equations define the Bidding Structure and Settlements. Equation (44)
shows the total payment and equation (45) the profit for each market participant.

TotalPmnt =
𝑁𝑇 𝑁𝑅
𝐷𝐴 𝑅𝑇 𝐷𝐴 𝑟𝑒𝑠𝐷𝐴
∑ {𝐿𝑀𝑃𝑡𝐷𝐴 ∗ 𝑃𝑖,𝑡 + 𝐿𝑀𝑃𝑡𝑅𝑇 ∗ (𝑃𝑖,𝑡 − 𝑃𝑖,𝑡 ) 𝐷𝐴
+ ∑ 𝐴𝑆𝐶𝑃𝑡,𝑟 ∗ 𝑃𝑖,𝑡,𝑟
𝑡=1 𝑟=1
𝑟𝑒𝑠𝑅𝑇 𝑟𝑒𝑠𝐷𝐴
𝑅𝑇
+ 𝐴𝑆𝐶𝑃𝑡,𝑟 ∗ (𝑃𝑖,𝑡,𝑟 − 𝑃𝑖,𝑡,𝑟 𝐷𝐴
) + 𝐴𝑆𝐶𝑃𝑡,𝑖𝑛𝑒𝑟𝑡𝑖𝑎 𝐷𝐴
∗ 𝑢𝑖,𝑡 ∗ 𝑆𝑖𝑚𝑎𝑥 ∗ 𝐻𝑖
𝑅𝑇
+ 𝐴𝑆𝐶𝑃𝑡,𝑖𝑛𝑒𝑟𝑡𝑖𝑎 ∗ 𝑆𝑖𝑚𝑎𝑥 ∗ 𝐻𝑖 ∗ (𝑢𝑖,𝑡
𝑅𝑇 𝐷𝐴
− 𝑢𝑖,𝑡 )} (44)

𝑁𝑇 𝑁𝑅
𝑟𝑒𝑠
𝑃𝑟𝑜𝑓𝑖𝑡𝑖 = 𝑇𝑜𝑡𝑎𝑙𝑃𝑚𝑛𝑡 − ∑ {𝐶𝑖𝐸 (𝑃𝑖,𝑡 , 𝑢𝑖,𝑡 , 𝑢𝑖,𝑡−1 ) + ∑ 𝐶𝑖𝐸 (𝑃𝑖,𝑡,𝑟 , 𝛾𝑖,𝑡 )} (45)
𝑡=1 𝑟=1

This market design comes with a suggested set of bidding and settlement rules. The
market design recommends:

Allocating payments for PGFR services to the loads based on their load share
or the allocation of other reserves;

84
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Relaxing market mitigation of PGFR in day-ahead markets as there are no


locational constraints;

Placing no external bid-in costs for kinetic balancing inertia;

Waiving under- and over-generation penalties for individuals participating in


the PGFR ancillary service market during events where frequency deviates by
more than the individual’s DB;

Including PGFR and synchronous kinetic balancing inertia payments for day-
ahead and real-time payments; if the total profit is negative, the resource
receives a make-whole payment to force the profit to zero;

Monitoring PGFR suppliers to ensure they are supplying what they have
offered to the PGFR market; PMU technology could make this monitoring
feasible; and,

Penalizing market participants without appropriate excuse for poor or no


performance after persistent violations.
The second paper Market Designs for the Primary Frequency Response Ancillary
Service—Part II: Case Studies44 presents case studies for this market design.

A.2. Market Design Based on Settled PGFR Including Reliability Changes


This section presents a modified market design based on settled PGFR including recent
reliability changes. In North America, reliability requirements now include demand side
resources along with generation, and require all participating resources to meet reliability
delivery definitions. Frequency control responsibility is no longer assigned to generation
alone. Maximum generation capacity becomes maximum generating capacity plus
minimum load limit for loads participating in the ancillary services market. Minimum
generating capacity becomes minimum generating capacity plus maximum load capacity
for loads participating in the ancillary services market. Governor deadband and
maximum power becomes high and low limits for PGFR in each direction for both loads
and generators.
Additionally, the modified NERC reliability standards remove any reliability
requirements that differentiate between spinning and non-spinning reserves. This
modification insures demand side resources are offered a level playing field with
generators with respect to the provision of reliability services. The technical
requirements of the service provided are the only basis for any differentiation between
resources (generation or demand). Consequently, secondary reserve combines all
equations that differentiate between spinning and non-spinning reserve.

44
E. Ela, V. Gevorgian, A. Tuohy, B. Kirby, M. Milligan and M. O'Malley, "Market Designs for the
Primary Frequency Response Ancillary Service—Part II: Case Studies," in IEEE Transactions on Power
Systems, vol. 29, no. 1, pp. 432-440, Jan. 2014. doi: 10.1109/TPWRS.2013.2264951

85
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Finally, current energy management system data collection methods will not support
control area measurements necessary to determine contributions to the frequency nadir.
The NERC standards resolve this limitation by using the ratio of frequency at steady state
to the frequency at the nadir. This ratio determines the amount of PGFR and reserve to
support that response at steady state necessary to infer the required response and reserve
at the nadir. This market implementation uses ratios that achieve a similar result in
equations (13) – (18). The following modified design presents only changed equations.

A.2.1. Nomenclature Including Reliability Changes


Generator & Load Parameters:
Pmax Maximum generating capacity or minimum load available for
response (MW)
Pmin Minimum generating capacity or maximum load available for
response (MW)
RR Ramp rate for secondary reserve including load providing
secondary reserve (MW/min)
DB Governor deadband and equivalent load deadband (Hz)
QSC Quick-start and interruptible load capability for secondary reserve
in trec (MW)
System Parameters and Requirements:
Eliminating the difference between spinning reserve and non-spinning reserve
results in the consolidation of two parameters, P2spin and P2nspin, into one, P2Req.
P2Req Secondary reserve requirement (MW)
Variables:
Eliminating the difference between spinning reserve and non-spinning reserve
results in consolidation of these two parameters, P2spin and P2nspin, into one, P2.
P2 Secondary reserve available (MW)
u Integer variable of unit or a load being on line [0,1]
Indices and Sets:
i, NG Generator and load index, set of generators and participating loads
Dual Variables:
No modifications

86
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Generator parameters include loads providing PGFR with synthetic continuous PGFR.
Loads providing PGFR by interruption at a specific frequency are not in the design.
These loads directly reduce the primary reserve requirement. Loads providing secondary
reserve are in the secondary reserve equations.

A.2.2. Market Scheduling Including Reliability Changes


Section 5.2.2 presents the equations adjusted for reliability changes as accepted by NERC
and applied to North America. This section also includes the market scheduling
equations adjusted for changes in reliability requirements and labeled with an “a”. The
following presents only those modified equations.
The objective function does not change.
There is no change in Inequality (2).

𝑃10𝑖,𝑡 includes the capacity adjustment available from load providing PGFR.
𝑁𝐺
0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
∑ 𝑃10𝑖,𝑡 ≥ 𝑃1𝐴 − 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃10𝑅𝑒𝑞 ∀𝑡 ∈ 𝑁𝑇 (3a)
ƒ0
𝑖=1

Equation (4a) creates an equivalent droop curve convertingi to units of Hz/MW and
includes load providing PGFR. The high and low load limits constrain equivalent droop,
with 𝑃𝑖𝑚𝑎𝑥 set by the low load limit minus the high load limit.

𝑒𝑞 𝑅𝑖 ∗ ƒ0
𝑅𝑖 = ∀𝑖 ∈ 𝑁𝐺 (4a)
𝑃𝑖𝑚𝑎𝑥

Inequality (5a) creates PGFR capacity from the droop curve. The starting frequency for
load PGFR sets the equivalent deadband for load.
𝜒𝑖,𝑡
𝑃10𝑖,𝑡 ≤ 𝑒𝑞 (∆ƒ
𝑚𝑎𝑥
− 𝐷𝐵𝑖 ) − 𝑃𝑖𝑚𝑎𝑥 ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (5a)
𝑅𝑖

Inequality (6) and (7) are not required to be applied to load participating in PGFR since
(4a) and (5a) insure that the full PGFR is available.
Inequalities (8), (9), and (10) do not change since they only apply to generators.
There is no change in the system simulation with the exception that it now includes loads
providing PGFR. Otherwise there are no changes to equations (11) – (18).
Since there is no difference between spinning and non-spinning reserves, (19) includes
the reserves from (21), and eliminates (21).

P2𝑖,𝑡 ≤ 𝑢𝑖,𝑡 ∗ 𝑡 𝑟𝑒𝑐 ∗ 𝑅𝑅𝑖 + (1 − 𝑢𝑖,𝑡 ) ∗ 𝑄𝑆𝐶𝑖 ∀𝑖 ∈ 𝑁𝐺 ∀𝑡 ∈ 𝑁𝑇 (19a)

87
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

Since there is no difference between spinning and non-spinning reserves, (20) includes
the reserves from (22), and eliminates (22).
𝑁𝐺
𝑅𝑒𝑞 0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
∑ 𝑃10𝑖,𝑡 + 𝑃2𝑖,𝑡 ≥ 𝑃2𝐴 − 𝑃1𝐴 − 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃2 ∀𝑡 ∈ 𝑁𝑇 (20a)
ƒ0
𝑖=1

Finally, since secondary reserve includes fast start units and interruptible load, checking
for sufficient capacity requires the removal of these units and loads.

𝑃𝑖,𝑡 + 𝑃10𝑖,𝑡 + 𝑃2𝑖,𝑡 − (1 − 𝑢𝑖,𝑡 ) ∗ 𝑄𝑆𝐶𝑖 ≤ 𝑢𝑖,𝑡 ∗ 𝑃𝑖𝑚𝑎𝑥 ∀𝑖 ∈ 𝑁𝐺 ∀𝑡 ∈ 𝑁𝑇 (23a)

Consolidation of spinning and non-spinning reserve allows the elimination of one system
parameter, one variable and two inequalities. These changes result in a simplification of
the market scheduling.

A.2.3. Market Pricing Including Reliability Changes


The equations from Section 5.2.3 do not require adjustments for reliability changes.
They only need modified subscripts.
Since spinning and non-spinning reserve are no longer separated, there is no need to price
them separately. It requires only a price for secondary reserve. Therefore, this eliminates
(29), and (36). Additionally, the payment for separate secondary reserve services is no
longer required, thus eliminating (41).

A.2.4. Bidding Structure and Settlements Including Reliability Changes


For this market design, based on settled PGFR including modifications for reliability
changes, these rules remain unchanged.

A.3. PGFR Market Design Based on Arrested PGFR


This section presents a market design based on arrested PGFR including recent reliability
changes. Using arrested PGFR as the measure, the market design changes significantly.
This design retains the changes from Section 5.2. The only measure required is PGFR at
the frequency nadir. This market simplification requires PMUs to measure the power
flows on providers of the arrested PGFR service and on the tie lines for CAs on an
interconnection.
In this design, market participants would offer power capabilities deliverable at the time
of the nadir following a frequency event. This requires market participants to estimate
the time of the nadir when making their offers and holding them responsible for meeting
their offered commitments at the time of the nadir. The design basis for this market
assumes the nadir time is dependent on the size of the event and is relatively stable from
event to event and that market participants understand their PGFR sufficiently to estimate
this delivery time.

88
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

This change passes some of the complexity associated with estimating the PGFR timing
for their equipment to market participants. This significantly simplifies the market. This
market design eliminates the need to set limits on deadbands, time delays, or any other
parameters that affect the response delivered at the nadir. Since these parameters are
under the control of the market participant, this eliminates the need to measure them
directly to assure that the participant delivered what was offered. It relates directly to
reliability since reliability is directly tied to the frequency at the nadir. If there is
withdrawal of PGFR resulting in a secondary nadir after the primary nadir, those
participants failing to maintain their PGFR until frequency is restored will have their
measurement of PGFR reduced at this secondary nadir of arrested PGFR affected by their
withdrawal.
Although, this increases market complexity for the market participants, it comes with
additional flexibility for the market participants. Instead of requiring the same limits for
each market participant with respect to deadbands, time delays, or other parameters that
are set by the market operator, it affords the opportunity for individual participants to
optimize these limits to maximize their own profits. This offers significant benefits not
only to the participants but also the market because of improved market efficiency.
Finally, this market design enables the definition of all real-energy ancillary services as
capacity capable of being delivered over a stated time. It requires PGFR delivery prior to
reaching the frequency nadir. It also requires the delivery of secondary reserves within
10 minutes for regulation and associated regulating reserve, and within 15 minutes for
contingency recovery and associated contingency recovery reserve. This creates a
standard methodology for defining reserve type services that extends to services such as
flexibility services currently under investigation. The following presents only the design
changes.

A.3.1. Nomenclature Based on Arrested PGFR


Generator & Load Parameters:
Pmax Maximum generating capacity or minimum load available for
response (MW)
Pmin Minimum generating capacity or maximum load available for
response (MW)
Smax Maximum apparent power capacity or maximum apparent load
available (MVA)
RR Ramp rate for secondary reserve including load providing
secondary reserve (MW/min)
DB Governor deadband and equivalent load deadband (Hz)
QSC Quick-start and interruptible load capability for secondary reserve
in trec (MW)
89
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

System Parameters and Requirements:


tss NR Time associated with steady-state frequency (s)
P1ssReq NR PGFR requirement at tss (MW)

ƒ ss NR Steady state frequency (Hz)

Variables:
P1ss NR PGFR availability at steady-state time (MW)
u Integer variable of unit or a load being on line [0,1]
Indices and Sets:
i, NG Generator and load index, set of generators and participating loads
t, NT Time interval index, time interval set
NR indicates a parameter that is no longer required

A.3.2. Market Scheduling Based on Arrested PGFR


Section 5.3.2 presents the equations adjusted for reliability changes as accepted by NERC
and arrested PGFR as a measure. The market scheduling equations, labeled with a “b”,
adjust for changes in reliability requirements and arrested PGFR. The following presents
only those equations modified.
The objective function does not change; only the parts making up the objective function.
𝑁𝑇 𝑁𝑅

𝑀𝑖𝑛 ∑ {𝑉𝑂𝐿𝐿 ∗ 𝐸𝑁𝑆𝑡 + ∑ 𝑉𝑂𝐼𝑅𝑟 ∗ 𝐼𝑅𝑡,𝑟


𝑡=1 𝑟=1
𝑁𝐺 𝑁𝑅
𝑟𝑒𝑠
+ ∑ [𝐶𝑖𝐸 (𝑢𝑖,𝑡 , 𝑢𝑖,𝑡−1 , 𝑃𝑖,𝑡 ) + ∑ 𝐶𝑖𝑅 (𝑃𝑖,𝑡,𝑟 , 𝛾𝑖,𝑡 )]} (1b)
𝑖=1 𝑟=1

Inequality (2b) constrains inertia to a minimum required to limit rate of change of


frequency (RoCoF) and under frequency load shedding (UFLS). However, when
defining the most severe single contingency (MSSC) as the basis for the reliability limit,
RoCoF and UFLS are be directly related to each other. This occurs because the RoCoF is
a function of the maximum imbalance, MSSC, and the system balancing inertia. The only
other variable is the function describing the timing of delivery of PGFR. This allows the
development of a solution to a single limit, UFLS.
𝑁𝐺
𝑒𝑞 𝑅𝑒𝑞
∑{𝑢𝑖,𝑡 ∗ 𝐻𝑖 ∗ 𝑆𝑖𝑚𝑎𝑥 } + 𝐻𝐿 ∗ 𝐿𝑡 ≥ 𝐼𝐴 − 𝐼𝑅𝑡,𝑖𝑛𝑒𝑟𝑡𝑖𝑎 ∀𝑡 ∈ 𝑁𝑇 (2b)
𝑖=1

90
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

The capacity requirement in (3) that insures sufficient capacity to deliver required PGFR
is available for each area.
𝑁𝐺
0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
∑ 𝑃10𝑖,𝑡 ≥ 𝑃1𝐴 − 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃10𝑅𝑒𝑞 ∀𝑡 ∈ 𝑁𝑇 (3b)
ƒ0
𝑖=1

When a limiting event occurs equal to MSSC, the nadir occurs at tnadir. At tnadir the
balancing inertia is equal to zero. Therefore, the PGFR contribution from generation and
participating load, plus the load damping contribution, plus the insufficient reserve
contribution must be greater than or equal to the PGFR requirement for the area MSSC.
(46b) shows the new inequality.
𝑁𝐺
∆ƒ 𝑚𝑎𝑥 0𝑅𝑒𝑞
∑ 𝑃10𝑖,𝑡 𝑛𝑎𝑑𝑖𝑟 + 𝐷 ∗ 𝐿𝑡 𝑛𝑎𝑑𝑖𝑟 ∗ + 𝐼𝑅𝑡 𝑛𝑎𝑑𝑖𝑟,𝑃10𝑅𝑒𝑞 ≥ 𝑃1𝐴 ≥ 𝑀𝑆𝑆𝐶 (46b)
ƒ0
𝑖=1

Inequality (46b) is an alternative form of inequality (3b), that addresses the specific time
of the nadir. It also indicates how MSSC relates to the total settled PGFR when an event
of MSSC magnitude occurs as given by the ratios in (47b), and indicates that at the limit
of MSSC, the required PGFR is equal to the settled PGFR. From this we can conclude
that the steady state components of PGFR bring no additional value to the solution.
0𝑅𝑒𝑞
𝑃1𝐴 𝑃1𝑠𝑠
≥ (47b)
𝑀𝑆𝑆𝐶 𝑀𝑆𝑆𝐶
Equation (4b) creates an equivalent droop curve to convert to units of Hz/MW.

𝑒𝑞 𝑅𝑖 ∗ ƒ0
𝑅𝑖 = ∀𝑖 ∈ 𝑁𝐺 (4b)
𝑃𝑖𝑚𝑎𝑥

Inequality (5b) creates PGFR capacity from the droop curve. Participating loads use this
inequality and do not need (4b) and (5b) to calculate this capacity. 𝑃10𝑖,𝑡 equals the load
scheduled minus the minimum load limit for participating loads. The frequency at which
the load begins to participate sets the deadband, DB.
𝜒𝑖,𝑡
𝑃10𝑖,𝑡 ≤ 𝑒𝑞 (∆ƒ
𝑚𝑎𝑥
− 𝐷𝐵𝑖 ) − 𝑃𝑖𝑚𝑎𝑥 ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (5b)
𝑅𝑖

Inequality (6b) sets the integer variable, , indicating head room is available to provide
the PGFR. This test is not necessary for participating loads.
1
𝑃𝑖,𝑡 + 𝑒𝑞 (∆ƒ
𝑚𝑎𝑥
− 𝐷𝐵𝑖 ) ≥ 𝑃𝑖𝑚𝑎𝑥 𝛾𝑖,𝑡 − 𝑃𝑖𝑚𝑎𝑥 𝜒𝑖,𝑡 ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (6b)
𝑅𝑖

Inequality (7b) ensures that PGFR of resources with head room and PGFR enabled will
equal their droop curve relationship. This test does not apply to participating loads.

91
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

𝜒𝑖,𝑡
𝑃10𝑖,𝑡 ≥ + 𝑒𝑞 (∆ƒ
𝑚𝑎𝑥
− 𝐷𝐵𝑖 ) − 𝑃𝑖𝑚𝑎𝑥 (1 − 𝛾𝑖,𝑡 ) ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (7b)
𝑅𝑖

Inequality (8b) requires a unit be on-line to enable its governor. This test also applies to
participating loads.

𝛾𝑖,𝑡 ≤ 𝑢𝑖,𝑡 ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (8b)

Equation (9b) sets equal to zero for units that are in the Gnogov set. This test is not
applied to participating loads.

𝛾𝑖,𝑡 = 0 ∀𝑖 ∈ 𝐺𝑛𝑜𝑔𝑜𝑣 , ∀𝑡 ∈ 𝑁𝑇 (9b)

Inequality (10b) sets equal to zero when the governor deadband is too large.

𝐷𝐵𝑖 ≤ 𝐷𝐵𝑚𝑎𝑥 + ƒ0 (1 − 𝛾𝑖,𝑡 ) ∀𝑖 ∈ 𝑁𝐺, ∀𝑡 ∈ 𝑁𝑇 (10b)

The above information is used to perform a system simulation. Checking the results
insures enough PGFR is provided at tnadir. The steady state frequency has little value
from a reliability perspective. Therefore, the design eliminates (11). Inequality (12)
insures the simulated nadir will not be exceeded.
𝑁𝐺
𝑛𝑎𝑑𝑖𝑟𝑅𝑒𝑞
∑ 𝑃1𝑛𝑎𝑑𝑖𝑟
𝑖,𝑡 ≥ 𝑃1𝐴 − 𝐿𝐷𝑛𝑎𝑑𝑖𝑟 ∀𝑡 ∈ 𝑁𝑇 (12b)
𝑖=1

(13), (15) and (17) are no longer required and are eliminated.

𝑛𝑎𝑑𝑖𝑟
𝑃1𝑛𝑎𝑑𝑖𝑟
𝑖,𝑡
𝛼𝑖,𝑡 =[ ] from Dynamic Sim2 ∀𝑖 ∶ 𝛾̂𝑖,𝑡 = 0, ∀𝑡 ∈ 𝑁𝑇 (14b)
𝑃10𝑖,𝑡

𝑛𝑎𝑑𝑖𝑟
𝑃1𝑛𝑎𝑑𝑖𝑟
𝑖,𝑡
𝛼𝑖,𝑡 =[ ] from Dynamic Sim1 ∀𝑖 ∶ 𝛾̂𝑖,𝑡 = 1, ∀𝑡 ∈ 𝑁𝑇 (16b)
̂ 0𝑖,𝑡
𝑃1
𝑁𝐺
𝑛𝑎𝑑𝑖𝑟 𝑛𝑎𝑑𝑖𝑟𝑅𝑒𝑞
∑ 𝛼𝑖,𝑡 ∗ 𝑃1𝑛𝑎𝑑𝑖𝑟
𝑖,𝑡 ≥ 𝑃1𝐴
𝑖=1
𝑛𝑎𝑑𝑖𝑟
∆ƒ 𝑚𝑎𝑥
−𝛼𝐿,𝑡 ∗ 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃1𝑛𝑎𝑑𝑖𝑟 ∀𝑡 ∈ 𝑁𝑇 (18b)
ƒ0
The last characteristic assures sufficient secondary response and reserve.

P2𝑖,𝑡 ≤ 𝑢𝑖,𝑡 ∗ 𝑡 𝑟𝑒𝑐 ∗ 𝑅𝑅𝑖 + (1 − 𝑢𝑖,𝑡 ) ∗ 𝑄𝑆𝐶𝑖 ∀𝑖 ∈ 𝑁𝐺 ∀𝑡 ∈ 𝑁𝑇 (19b)

Since there is no difference between spinning and non-spinning reserves, (19) includes
the reserves from (21), and eliminates (21).

92
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

𝑁𝐺
𝑅𝑒𝑞 0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
∑ 𝑃10𝑖,𝑡 ∗ 𝑃2𝑖,𝑡 ≥ 𝑃2𝐴 − 𝑃1𝐴 − 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃2 ∀𝑡 ∈ 𝑁𝑇 (20b)
ƒ0
𝑖=1

Preventing instabilities during recovery after the nadir is an important issue. Not only is
preventing step responses by governors important, excluding step responses by load with
restoration before restoring frequency to a point that will not result in re-interruption is
also important. In addition, the recommendation is to replace or eliminate synthetic
balancing inertia with synthetic continuous PGFR until clarifying the instability issues
identified in Section 3.2.2 of this document.
Since there is no difference between spinning and non-spinning reserves, (20) includes
the reserves from (22), and eliminates (22). In North America, the NERC Standard BAL-
𝑅𝑒𝑞
002 requires the value for 𝑃2𝐴 to be greater than or equal to MSSC for the control area
and 𝑡 𝑅𝑒𝑞 is set at 15 minutes by that same standard.

𝑃𝑖,𝑡 + 𝑃10𝑖,𝑡 + 𝑃2𝑖,𝑡 − (1 − 𝑢𝑖,𝑡 ) ∗ 𝑄𝑆𝐶𝑖 ≤ 𝑢𝑖,𝑡 ∗ 𝑃𝑖𝑚𝑎𝑥 ∀𝑖 ∈ 𝑁𝐺 ∀𝑡 ∈ 𝑁𝑇 (23b)

Finally, since secondary reserves include fast start units, remove these units when
checking for sufficient capacity.

A.3.3. Market Pricing Based on Arrested PGFR


Equation (24b) shows the Lagrangian function. The changes identified in these equations
fail to be observable for the most part because subscripting specifies the changes. The
basis for these changes are the elimination of the steady state analysis and the
consolidation of spinning and non-spinning reserve categories. The elimination of steady
state frequency as a reliability requirement, eliminates equations (26), (33) and (38).
L=
𝑁𝑇 𝑁𝑅 𝑁𝐺 𝑁𝑅
𝑟𝑒𝑠
∑ {𝑉𝑂𝐿𝐿 ∗ 𝐸𝑁𝑆𝑡 + ∑ 𝑉𝑂𝐼𝑅𝑟 ∗ 𝐼𝑅𝑡,𝑟 + ∑ [𝐶𝑖𝐸 (𝑃𝑖,𝑡 ) + ∑ 𝐶𝑖𝑅 (𝑃𝑖,𝑡,𝑟 )]
𝑡=1 𝑟=1 𝑖=1 𝑟=1
𝑁𝐺 𝑁𝐵 𝑁𝐿 𝑁𝐵

− 𝜆𝑡 ∗ (∑ 𝑃𝑖,𝑡 − ∑ 𝐿𝑛,𝑡 − 𝐸𝑁𝑆𝑡 ) − ∑ (𝑢𝑖,𝑡 ∗ ∑ 𝑆𝐹𝑛,𝑙 ∗ (𝑃𝑛,𝑡 − 𝐿𝑛,𝑡 ) − 𝐹𝑙𝑚𝑎𝑥 )


𝑖=1 𝑛=1 𝑙=1 𝑛=1
𝑁𝑅 𝑁𝐺
𝑟𝑒𝑠 𝑚𝑎𝑥
− ∑ 𝛽𝑟,𝑡 ∗ 𝐻𝑟 (𝑃𝑖,𝑡,𝑟 , 𝑅𝑡,𝑟 , 𝐼𝑅𝑡,𝑟 , 𝐿) − ∑ 𝜋𝑖,𝑡 ∗ (𝑃𝑖𝑚𝑎𝑥 − 𝑃𝑖,𝑡 − 𝑃10𝑖,𝑡 − 𝑃2𝑖,𝑡 )
𝑟=1 𝑖=1

𝑚𝑖𝑛
− 𝜋𝑖,𝑡 ∗ (𝑃𝑖,𝑡 − 𝑃𝑖𝑚𝑎𝑥 )} (24b)

The ancillary services inequality constraints use Hr(*).


𝑁𝐺
𝑛𝑎𝑑𝑖𝑟 𝑛𝑎𝑑𝑖𝑟𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
𝐻1 = ∑ 𝛼𝑖,𝑡 ∗ 𝑃10𝑖,𝑡 − 𝑃1𝐴 𝑛𝑎𝑑𝑖𝑟
+ 𝛼𝐿,𝑡 ∗ 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃1𝑛𝑎𝑑𝑖𝑟 (25b)
ƒ0
𝑖=1
93
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

𝑁𝐺
0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
𝐻2 = ∑ 𝑃10𝑖,𝑡 − 𝑃1𝐴 + 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃10 (27b)
ƒ0
𝑖=1

𝑁𝐺
𝑅𝑒𝑞 0𝑅𝑒𝑞 ∆ƒ 𝑚𝑎𝑥
𝐻3 = ∑ 𝑃2𝑖,𝑡 − 𝑃2𝐴 − 𝑃1𝐴 + 𝐷 ∗ 𝐿𝑡 ∗ − 𝐼𝑅𝑡,𝑃2 (28b)
ƒ0
𝑖=1

Since spinning and non-spinning reserve are no longer separated, there is no need to price
them separately. The only requirement is a price for secondary reserve. This eliminates
(29) and (36).
𝑁𝐿
𝜕𝐿
𝐿𝑀𝑃𝑛 = = 𝜆 + ∑ 𝜇𝑙 ∗ 𝑆𝐹𝑛,𝑙 − 𝜆 ∗ 𝐿𝐹𝑖
𝜕𝐿𝑛
𝑙=1
∆ƒ 𝑚𝑎𝑥 ∆ƒ 𝑚𝑎𝑥
−𝛽2 ∗ − 𝛽3 ∗ (30b)
ƒ0 ƒ0

𝜕𝐿
𝐴𝑆𝐶𝑃𝑟 = = 𝛽𝑟 (31b)
𝜕𝑅𝑟
𝜕𝐿
𝐴𝑆𝐶𝑃𝑃1𝑛𝑎𝑑𝑖𝑟 = 𝑛𝑎𝑑𝑖𝑟𝑅𝑒𝑞
= 𝛽1 (32b)
𝜕𝑃1𝐴

𝜕𝐿
𝐴𝑆𝐶𝑃𝑃10 = 0𝑅𝑒𝑞 = 𝛽2 (34b)
𝜕𝑃1𝐴

𝜕𝐿
𝐴𝑆𝐶𝑃𝑃2 = 𝑅𝑒𝑞 = 𝛽3 (35b)
𝜕𝑃2𝐴

This eliminates equations (38) and (41) for payments.

𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡,𝑃1𝑛𝑎𝑑𝑖𝑟 = 𝐴𝑆𝐶𝑃𝑡,𝑃1𝑛𝑎𝑑𝑖𝑟 ∗ 𝑃1𝑛𝑎𝑑𝑖𝑟


𝑖,𝑡 (37b)

𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡,𝑃10 = 𝐴𝑆𝐶𝑃𝑡,𝑃10 ∗ 𝑃10𝑖,𝑡 (39b)

𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡,𝑃2, = 𝐴𝑆𝐶𝑃𝑡,𝑃2 ∗ 𝑃2𝑖,𝑡 (40b)

Equations (42) and (43) are retained to price balancing inertia.


𝐿𝑃
𝐴𝑆𝐶𝑃𝑖𝑛𝑒𝑟𝑡𝑖𝑎 = 𝛽𝑖𝑛𝑒𝑟𝑡𝑖𝑎 (42b)
𝑖𝑛𝑒𝑟𝑡𝑖𝑎
𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖,𝑡 = 𝐴𝑆𝐶𝑃𝑡𝑖𝑛𝑒𝑟𝑡𝑖𝑎 ∗ {𝑢𝑖,𝑡 ∗ 𝐻𝑖 ∗ 𝑆𝑖𝑚𝑎𝑥 } (43b)

A.3.4. Bidding Structure and Settlements Based on Arrested PGFR


The following equations define the Bidding Structure and Settlements. Equation (44b)
shows the TotalPmnt.
94
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

TotalPmnt =
𝑁𝑇 𝑁𝑅
𝐷𝐴 𝑅𝑇 𝐷𝐴 𝑟𝑒𝑠𝐷𝐴
∑ {𝐿𝑀𝑃𝑡𝐷𝐴 ∗ 𝑃𝑖,𝑡 + 𝐿𝑀𝑃𝑡𝑅𝑇 ∗ (𝑃𝑖,𝑡 − 𝑃𝑖,𝑡 ) + ∑ 𝐴𝑆𝐶𝑃𝑡,𝑟 𝐷𝐴
∗ 𝑃𝑖,𝑡,𝑟
𝑡=1 𝑟=1
𝑟𝑒𝑠𝑅𝑇 𝑟𝑒𝑠𝐷𝐴
𝑅𝑇
+ 𝐴𝑆𝐶𝑃𝑡,𝑟 ∗ (𝑃𝑖,𝑡,𝑟 − 𝑃𝑖,𝑡,𝑟 𝐷𝐴
) + 𝐴𝑆𝐶𝑃𝑡,𝑖𝑛𝑒𝑟𝑡𝑖𝑎 𝐷𝐴
∗ 𝑢𝑖,𝑡 ∗ 𝑆𝑖𝑚𝑎𝑥 ∗ 𝐻𝑖
𝑅𝑇
+ 𝐴𝑆𝐶𝑃𝑡,𝑖𝑛𝑒𝑟𝑡𝑖𝑎 ∗ 𝑆𝑖𝑚𝑎𝑥 ∗ 𝐻𝑖 ∗ (𝑢𝑖,𝑡
𝑅𝑇 𝐷𝐴
− 𝑢𝑖,𝑡 )} (44b)

𝑁𝑇 𝑁𝑅
𝑟𝑒𝑠
𝑃𝑟𝑜𝑓𝑖𝑡𝑖 = 𝑇𝑜𝑡𝑎𝑙𝑃𝑚𝑛𝑡 − ∑ {𝐶𝑖𝐸 (𝑃𝑖,𝑡 , 𝑢𝑖,𝑡 , 𝑢𝑖,𝑡−1 ) + ∑ 𝐶𝑖𝐸 (𝑃𝑖,𝑡,𝑟 , 𝛾𝑖,𝑡 )} (45b)
𝑡=1 𝑟=1

Some simple changes improve the bidding and settlement rules presented in the original
design presented in 5.1.4.

Instead of allocating the cost of ancillary services to the loads, allocate the
cost of the services to those market participants that create the need for those
services. This is an actuarial problem that needs a solution. With this solution
come significant market benefits. Proportionally allocating the costs to those
causing the need for the services, encourages those parties to reduce their
contribution to the need for the services. This action results in improved
market efficiency.

Since there are no locational constraints, the market mitigation of PGFR in


day-ahead markets should be relaxed.

Kinetic balancing inertia should incur no external bid-in costs.

Eliminate waiving the under- and over-generation penalties during events


where frequency deviates by more than the individual’s DB. Adjust the base
schedules to include scheduled PGFR of individual participants providing
PGFR. Base any under- and over-generation penalties on these adjusted
schedules.

All payments include PGFR and balaning inertia payments, for day-ahead and
real-time. If the total profit is negative, the resource receives a make-whole
payment to force the profit to zero.

Instead of penalizing PGFR suppliers for compliance with the services they
offered, monitor them to ensure they are supplying what they offered and sold
to the PGFR market. Charge consequential damages for not holding or
delivering the services offered and sold. PMU technology is necessary to
make this monitoring feasible and to effectively measure a market based on
arrested PGFR. Penalize market participants, for poor or no performance, and
after persistent violations without an appropriate excuse.

95
TR-24 — MEASUREMENT, MONITORING, AND RELIABILITY ISSUES RELATED TO PRIMARY
GOVERNING FREQUENCY RESPONSE

A.4. Summary
Section A.3 specifies a market design based on arrested PGFR and kinetic balancing
inertia. When comparing this design to the market design for settled PGFR suggested in
the referenced papers, the market design based on arrested PGFR appears to be
significantly simpler. The basis of the design, the frequency nadir, is directly related to
the reliability of the interconnection. It simplifies the necessary measurements required
to support the market design. It intrinsically includes most extraneous parameters that
affect the quality of the service acquired, simplifying the measurement structure and
minimizing the effort required for implementation. It fits into a structure that is
compatible with a market that includes other services such as secondary control and
secondary reserve services.

96

You might also like