You are on page 1of 85

Power

System Security
Assessment of the future
National Electricity Market



A report by the Melbourne Energy Institute at the University of
Melbourne in support of The Independent Review into the Future
Security of the National Electricity Market

June 2017


Power system security assessment of
the future National Electricity Market

A report by the
Melbourne Energy Institute
at the
University of Melbourne
in support of the

‘Independent Review into the


Future Security of the National Electricity Market’

June 2017

Lead author
Prof Pierluigi Mancarella, Chair of Electrical Power Systems, School of
Engineering

Co-authors
Mr Sebastian Puschel, School of Engineering
Mr Lingxi Zhang, Visitor, School of Engineering
Ms Han Wang, School of Engineering
Prof Michael Brear, FIEAust, FAIE, Director, Melbourne Energy Institute
Prof Terry Jones, School of Engineering and Melbourne Energy Institute
Dr Matthew Jeppesen, School of Engineering
Prof Robin Batterham, AO, School of Engineering and Melbourne Energy Institute
Laureate Prof Robin Evans, FAA FTSE FIEEE FIEAust, School of Engineering
Prof Iven Mareels, FTSE, FIEEE, FIEAust, CPEng, EngExec, Dean, School of
Engineering

1
© Copyright 2017 Department of the Environment and Energy

This report has been prepared by The University of Melbourne, on behalf


of, and for the exclusive use of the Department of the Environment and
Energy, and is subject to, and issued in accordance with, the provisions of
the contract between the University and the Department of the
Environment and Energy. The University accepts no liability or
responsibility whatsoever for, or in respect of, any use of, or reliance upon,
this report by any third party.

The concepts and information contained in this document are the property
of the Department of the Environment and Energy. Use or copying of this
document in whole or in part without the written permission of the
Department of the Environment and Energy constitutes an infringement of
copyright.

2
Table of Contents
Table of Contents ............................................................................................................................. 3
Executive Summary ........................................................................................................................ 6
Technical Summary......................................................................................................................... 8
List of acronyms ............................................................................................................................ 15
1. Introduction .......................................................................................................................... 16
2. Key definitions and concepts for frequency response security assessment 17
2.1. Key definitions and concepts ................................................................................ 17
2.1.1. Definition of “security” for the purpose of this work ............................ 17
2.1.2. Frequency response and system inertia ..................................................... 17
2.2. Frequency response in the current NEM .......................................................... 19
2.2.1. Current Frequency Control Ancillary Services (FCAS) arrangements ..
..................................................................................................................................... 19
2.2.2. Contingency FCAS requirements calculation ............................................ 20
2.3. Frequency response challenges in a low inertia power system .............. 21
2.4. Potential need for new operational requirements and constraints ....... 22
2.4.1. From static to dynamic frequency response requirements ................ 22
2.4.2. Mapping inertia-dependent frequency response requirements and
constraints ............................................................................................................................. 23
2.5. Further considerations on system security aspects not covered ........... 25
3. Methodology for frequency response security assessment................................ 26
3.1. Methodology for security assessment ............................................................... 26
3.1.1. General approach................................................................................................. 26
3.1.2. “Unconstrained” frequency Nadir case (base case) ............................... 26
3.1.3. “Constrained” frequency Nadir case ............................................................ 27
3.1.4. Generation/demand combination snapshots considered in the
analysis .................................................................................................................................... 28
3.1.5. Summary of the methodology for security assessment ........................ 29
3.2. Considerations on minimum synchronous inertia and interaction
among inertial, primary and secondary frequency response ................................. 29
4. Scenario analysis ................................................................................................................. 31
4.1. Scenarios overview ................................................................................................... 31
4.1.1. The modelling scenarios ................................................................................... 31
4.1.2. Input data from modelling scenarios ........................................................... 31
4.2. Scenario results overview ...................................................................................... 34
4.2.1. Nadir ......................................................................................................................... 34
4.2.2. ROCOF ...................................................................................................................... 35
4.2.3. Inertia ....................................................................................................................... 37
4.2.4. PFR ............................................................................................................................ 39
4.3. Analysis of specific results for CET&LL scenario .......................................... 40
4.3.1. Nadir assessment for different damping factors ..................................... 41
4.3.2. ROCOF assessment for different damping factors .................................. 41
4.3.3. Inertia assessment for different damping factors ................................... 42
4.3.4. Impact of different ROCOF limits................................................................... 43
5. Minimum synchronous generation analysis for CET&LL scenario .................. 45
5.1. General considerations and methodology ....................................................... 45
5.2. Static assessment of minimum synchronous generation level through
FRSC-OPF..................................................................................................................................... 46
3
5.2.1. Methodology .......................................................................................................... 46
5.2.2. MW value of minimum synchronous generation per year .................. 46
5.2.3. Comparison with the SNSP index .................................................................. 47
5.3. Dynamic assessment of minimum synchronous generation levels
through Frequency Response Security Constrained Unit Commitment (FRSC-
UC) ........................................................................................................................................... 48
5.3.1. Problem description and methodology....................................................... 48
5.3.2. Results ...................................................................................................................... 50
5.3.3. Assessment of the system non-synchronous penetration (SNSP) ... 55
6. Role of new and alternative technologies and services........................................ 57
6.1. More than synchronous generators ................................................................... 57
6.2. Synchronous condensers ........................................................................................ 57
6.3. Fast Frequency Response ....................................................................................... 59
6.3.1. Fast Frequency Response (FFR) and Synthetic/Virtual Inertia ........ 59
6.3.2. Some numerical results ..................................................................................... 59
6.4. Dynamic contingency rescheduling and regional security analysis ...... 62
6.4.1. Dealing with transmission contingencies .................................................. 62
6.4.2. Example of transmission contingency analysis for South Australia 62
6.5. Advanced protection schemes .............................................................................. 65
6.6. New inertial response services ............................................................................ 66
6.6.1. Valuing “differentiated inertia” levels ......................................................... 66
6.6.2. Example: annual generation dispatch and benefits ............................... 66
6.7. Further considerations on frequency response services and
technologies ............................................................................................................................... 68
7. Appendix: the National Electricity Market ................................................................ 69
7.1. General aspects of the NEM ................................................................................... 69
7.2. NEM operation ............................................................................................................ 69
8. Appendix: Emissions reduction modelling scenarios description and input
data .................................................................................................................................................... 71
8.1. Scenario description ................................................................................................. 71
8.1.1. Business as Usual Scenario (BAU)................................................................. 71
8.1.2. Limited Lifetime (LL) ......................................................................................... 71
8.1.3. Clean Energy Target (CET)............................................................................... 71
8.1.4. Clean Energy Target and Limited Lifetime (CET&LL) ........................... 71
8.1.5. Emissions Intensity Scheme (EIS)................................................................. 71
8.2. Input data ...................................................................................................................... 72
8.2.1. Electricity demand .............................................................................................. 72
8.2.2. Installed capacity by scenario ......................................................................... 73
8.2.3. Example of VRE generation profiles ............................................................. 75
9. Appendix: Main definitions, modelling assumptions and input data for
system security studies .............................................................................................................. 77
9.1. Security: general definition and definition for the purpose of this work ..
........................................................................................................................................... 77
9.2. Definition of credible and noncredible contingencies in the NEM ......... 78
9.3. Main assumptions for FCAS and frequency response modelling ............ 79
9.4. Main operational limits adopted in the base case security assessment
studies .......................................................................................................................................... 80
9.5. Definition of the System Non-Synchronous Penetration (SNSP) level
index ........................................................................................................................................... 81
4
9.6. Specific input data for FRSC-UC studies for dynamic assessment of
minimum synchronous generation levels ...................................................................... 81
9.7. Further information on main input data and assumptions used for the
security assessment studies for all scenarios ............................................................... 82
9.7.1. General inputs and assumptions ................................................................... 82
9.7.2. Snapshots scenarios ........................................................................................... 83

5
Executive Summary
This report has been written by the Melbourne Energy Institute (MEI) at the
University of Melbourne in support of the ‘Independent Review into the Future
Security of the National Electricity Market’ (the Review) chaired by Dr Alan Finkel
AO. The Review has commissioned Jacobs to model a number of emissions
reduction policy scenarios. The generation portfolios resulting from five
modelling scenarios have then been used as inputs to the University’s
assessment of power system security in the National Electricity Market (NEM).

Each of these scenarios features generation portfolios with increasing


penetration of Variable Renewable Electricity (VRE) sources out to 2050. Such
systems may raise security concerns since most VRE technologies do not
typically provide the so-called inertia, which has historically been one of the
main means of regulating the AC system frequency to (nominally) 50Hz.

This report therefore focuses on so-called frequency response adequacy, i.e.,


assessing whether the dynamic response of a given generation portfolio
following a so-called contingency event would be adequate to return the system
frequency to 50Hz without disruptions to energy supply. The main contingency
event considered in this report is the loss of the largest operating generator in
the network, which is standard practice in such studies.
The report first defines key terms and concepts, then presents the methodology
used to undertake these security assessment studies, and finally presents the
relevant results. The work performed includes:
1. Analysis of future power system security under each scenario, and advice on
what subsequent actions may be required to establish power system security,
if needed; and
2. Analysis of the minimum level of inertia required in the so-called Clean
Energy Target and Limited Lifetime (CET&LL) scenario.

Whilst a comprehensive analysis of system security is beyond the scope of this


work, these analyses highlight two main points:
1. Without implementation of appropriate operational measures, the NEM will
experience increasing issues of frequency control in all modelled scenarios,
including the Business As Usual case.
2. There is significant potential to use several operational measures and
electricity market designs to ensure frequency response adequacy in VRE-
rich power systems.

Such operational measures and market designs could drive the generation
portfolio’s operation so that the dynamic response of the system after a potential
contingency event is constrained to return the system frequency to 50Hz.
Indeed, frequency response adequacy appears achievable in all of the scenarios
studied in this report.

Furthermore, whilst inertia-bearing generation is likely to remain important, our


analyses also show that there are a number of other technologies and services
6
that can provide so-called Fast Frequency Response (FFR) and therefore also
legitimately play a significant role in supporting frequency regulation. This
includes demand response, energy storage of several forms, and the so-called
synthetic inertia in newer wind turbines, amongst others. Other means to
increase the system inertia, such as synchronous condensers, could also assist in
providing this support.

More broadly, numerous technologies and services could provide frequency


support given careful electricity market design that does not discriminate
unfairly among these technologies and services. This could also include
operational services using emerging techniques such as dynamically
rescheduling for the current largest contingency and simultaneously scheduling
inertia and frequency response. Such approaches may be particularly useful in
systems with potentially very low inertia, as might occur in a region of the power
system that is at risk of becoming islanded or has become islanded after an
interconnector failure. In these cases, specific arrangements for provision of
frequency response and system inertia should be made regionally, in addition to
simultaneously controlling frequency for the NEM as a whole.

Finally, it should be highlighted that frequency stability is not the only important
aspect of system security in a future NEM. These other aspects, which were out
of the scope of this work, require detailed further study.

7
Technical Summary
Given the concerns in running low inertia power systems that might result from
the widespread penetration of VRE sources, this report focuses on so-called
frequency response adequacy, i.e., assessing whether the frequency response of a
given generation portfolio would be adequate following a contingency event. The
main contingency event considered in this report is the loss of the largest
operating generator in the network, which is standard practice in such studies.
Figure 0.1 illustrates how system frequency could change after the same
contingency event in high and low inertia power systems with the same demand.
The contingency event is the loss of a large generator in both cases. The key
parameters involved are:
 the Rate Of Change Of Frequency (ROCOF): the initial slope of the frequency
change immediately following the contingency event;
 the Frequency Nadir: the minimum frequency that is reached following the
contingency event;
 the Quasi Steady-State Frequency (QSSF): the steady-state frequency that is
reached after the scheduled generators automatically provide the so-called
primary frequency response (PFR) to its nominal output.

The capacity of PFR scheduled in Figure 0.1 is deliberately the same for the two
inertia cases, and calculated on the basis of current static requirements. This
yields the same QSSF, which is independent of the amount of inertia in the
system. However, a lower inertia system experiences a lower frequency Nadir,
which also occurs earlier. This is partly due to the lower inertia system having a
higher ROCOF immediately after the contingency event.

Figure 0.1: Simulated effect of lower inertia on system frequency response, with key
parameters defined, after a given contingency event

8
Methodology

Traditionally, synchronous generators with physical inertia that is associated


with their rotational kinetic energy have proven effective in providing frequency
response adequacy. However, Figure 0.1 suggests that the provision of PFR, and
more generally so-called Frequency Control Ancillary Services (FCAS) that include
PFR, needs to be reconsidered as the proportion of synchronous generation in
the power system reduces. In order to operate a stable system, all three of the
parameters in Figure 0.1 that characterise frequency response - the ROCOF, the
frequency Nadir and the static requirement for QSSF - should be constrained in
order to avoid further consequences, such as additional generation
disconnection, which could end up in a cascaded event.

Figure 0.2 illustrates the need to constrain all three of these parameters in the
general frequency control problem. This frequency response security map shows
the (shaded) secure area in which all three of these constraints are observed for
systems with varying amounts of PFR and system inertia. The vertical line
depicts the level of inertia that corresponds to a desired ROCOF limit, whilst the
horizontal line corresponds to the static PFR requirement for a desired QSSF
limit. The red curve depicts the desired limit for the frequency Nadir. The secure
area resides above all three curves. Thus, high inertia systems are typically
constrained by static requirements that are independent of the system inertia,
i.e. the secure area is bounded by the horizontal line. However, when
transitioning towards lower inertia systems, both the frequency Nadir limit (the
red curve) and the ROCOF limit (the vertical line) may become binding. Since the
secure area for low inertia systems is bounded by both the frequency Nadir and
ROCOF requirements, the scheduled PFR must therefore be higher than the static
requirement whilst also taking into account the current system inertia.

Figure 0.2: Frequency response security map showing the secure area of operation for
different levels of primary frequency response (PFR) and system inertia

9
This report therefore considers two cases with different forms of generation
dispatch:
1. A base case: which approximates the current provision of FCAS in the
National Electricity Market (NEM) with static requirements on frequency
response and no explicit consideration of system inertia;
2. A Frequency Response Security Constrained Optimal Power Flow (FRSC-OPF)
case: which takes into account the on-line system inertia and dispatches
energy and FCAS so as to observe the desired limits on the ROCOF, the
frequency Nadir and the QSSF.

The FRSC-OPF tool used in this study thus performs a co-optimization of all the
resources that can meet all frequency response requirements, including inertial
response, and optimally schedules generating units and other available
resources to provide FCAS.

Key findings

While there are significant challenges in operating a low inertia power system,
the analyses performed demonstrate that there is significant potential to use
operational measures and electricity market designs to ensure frequency
response adequacy in VRE-rich power systems. Such operational measures and
market designs could co-optimize inertia and FCAS so that the system remains
secure after a potential contingency event. Indeed, frequency response adequacy
appears achievable in all of the scenarios studied in this report.

A summary of the results is shown in Figure 0.3. Generation portfolios from the
following five scenarios were used as inputs to the University’s assessment of
power system security:
1. Business As Usual (BAU)
2. Limited Lifetime (LL)
3. Clean Energy Target and Limited Lifetime (CET&LL)
4. Clean Energy Target (CET)
5. Emissions Intensity Scheme (EIS)

Further analysis of the CET&LL scenario was also performed to estimate the
minimum synchronous generation level that could ensure frequency adequacy.
This analysis was not performed for the other four scenarios as the results are
expected to be similar.

Figure 0.3 shows that the frequency Nadir that could be experienced in the
presence of maximum and minimum demand in 2050 for each scenario,
following the chosen contingency of a loss of 700MW of synchronous generation.
Results for all scenarios with the base case, applying the current static
requirements (top), and the FSRC-OPF case (bottom) are shown. The frequency
Nadir that could be experienced depends strongly on different VRE penetrations
that occur in each scenario. These results also show that it is possible to operate
the system so as to confine the frequency Nadir to the desired 49.2Hz limit in all
cases. This limit is argued to be a reasonable security limit and is discussed
further in the report, but other limits could be applied as well.
10
Figure 0.3: Nadir level after contingency without explicit frequency Nadir and ROCOF
constraints (top) and with Nadir constraint of 49.2Hz and ROCOF constraint of 1Hz/s.
Comparison of all scenarios in different years, with load damping factor of 2%/Hz

Figure 0.4 shows the evolution of the associated amount of inertia that is
scheduled in the system at times of maximum and minimum demand for the base
case and FRSC-OPF cases in the CET&LL scenario. The FRSC-OPF schedules
higher levels of system inertia given the joint requirements of more kinetic
energy and FCAS provided by the synchronous units. The figure also shows how
the amount of system inertia could change due to different demand “damping
factor” values that might apply in the future, which describe how demand varies
following a frequency event.

It was also found that the system could securely run with a minimum
synchronous generation output in the order of 25% of the minimum demand
(approximately 4GW of around 17GW, with associated system inertia of about
48GWs) in 2050 for the CET&LL scenario. On the other hand, for higher demand
levels the system could securely run with even lower levels of minimum

11
synchronous generation output and inertia. It is emphasised that these values
are approximate and require simultaneous provision of suitable amount of
scheduled FCAS and inertia, as can be determined by dynamic optimization
studies as presented in this report.

Indeed, given the complex and dynamic interplay between inertia and FCAS,
‘rules of thumb’ approaches that statically determine the minimum level of
synchronous output are necessarily only estimates. Therefore, there may be
scope for improved system level outcomes if frequency response constraints in
the system dispatch engine, as done in the FRSC-OPF case above, are used to
dynamically co-optimise all the system resources, including the available
synchronous inertia. Figure 0.5 further demonstrates how inertia, PFR,
renewable (asynchronous) generation and demand might interact dynamically
using the FRSC-OPF approach. The PFR and inertia requirements are correlated,
with PFR being higher when the system has lower inertia and vice versa. It can
also be appreciated how lower inertia coincides with periods of higher VRE
output and therefore of lower net demand (demand minus renewable output)
that is to be covered by conventional generation.

Role of new technologies and services

Whilst inertia-bearing synchronous generation is likely to continue to play an


important role in supporting frequency stability in the future, there are a
number of technologies and services that could also play a significant part. This
is particularly because frequency Nadir requirements appear to be important in
future systems, at least at the NEM level.

New technologies and services that can operate on these timescales to provide
so-called Fast Frequency Response (FFR) are therefore expected to have an
important future role. These include demand response, energy storage of several

Figure 0.4: Evolution of system inertia for different demand damping factors and for
maximum and minimum demand for the CET&LL scenario: without constraints (i.e., base
case, left) and with a Nadir constraint of 49.2 Hz and ROCOF constraint of 1Hz/s (i.e., FRSC-
OPF case, right)
12
forms, and the so-called synthetic inertia in newer wind turbines, amongst
others. Other means to increase the system inertia, such as synchronous
condensers, could also assist in providing this support.
More broadly, numerous technologies and services could provide frequency
support given careful electricity market design that does not discriminate
unfairly among these technologies and services. This could also include further
operational services such as dynamically rescheduling for the current largest
contingency or use of advanced protection technologies such as system integrity
protection schemes that “inter-trip” specific contingencies and frequency
response activation. Such approaches may be particularly useful in systems with
potentially very low inertia, as might occur in a region that is at risk of becoming
islanded or has become islanded after an interconnector failure. In these cases,
specific arrangements for provision of frequency response and system inertia
should be made regionally, in addition to simultaneously controlling frequency
for the NEM as a whole.

Figure 0.5: Example of primary frequency response (PFR) and system inertia
requirements (top), as co-optimized by the FRSC-OPF algorithm with the frequency Nadir
constrained, and the corresponding system net demand and variable renewable
electricity (VRE) generation profiles (bottom) for a day in the CET&LL scenario in 2050

13
Finally, it should be pointed out that, whilst frequency stability boundaries are
likely to be more critical in a lower inertia NEM, a full system security
assessment will require consideration of a number of other aspects. These
include consideration of rotor angle stability, voltage stability and reactive
power support, short-circuit strength, new balancing reserves, and so on. These
other aspects, which were out of the scope of this work, require detailed further
study.

14
List of acronyms
General acronyms
AEMO Australian Energy Market Operator
CCGT Combined Cycle Gas Turbine
DFIG Doubly-Fed Induction Generator
FCAS Frequency Control Ancillary Services
FFR Fast Frequency Response
FRSC-OPF Frequency Response Security Constrained Optimal Power Flow
FRSC-UC Frequency Response Security Constrained Unit Commitment
NEM National Electricity Market
OCGT Open Cycle Gas Turbine
OPF Optimal Power Flow
PFR Primary Frequency Response
PV Photovoltaic
QSSF Quasi Steady-State Frequency
ROCOF Rate of Change of Frequency
SFR Secondary Frequency Response
SNSP System Non-Synchronous Penetration
UFLS Under-Frequency Load Shedding
UC Unit Commitment
VRE Variable Renewable Electricity

Emission Reduction Modelling Scenarios


BAU Business As Usual
LL Limited Lifetime
CET&LL Clean Energy Target and Limited Lifetime
CET Clean Energy Target
EIS Emission Intensity Scheme

15
1. Introduction
This report has been written by the Melbourne Energy Institute (MEI) at the
University of Melbourne in support of the ‘Independent Review into the Future
Security of the National Electricity Market’ (the Review) chaired by Dr Alan Finkel
AO.
The Review has commissioned Jacobs to model a number of emissions reduction
policy scenarios. The generation portfolios resulting from five modelling
scenarios have then been used as inputs to the University’s assessment of power
system security.
Within the project scope, this document illustrates the main modelling
assumptions and results from the security studies, specifically providing:
1. Analysis of and advice on the security1 of the future power system under
each of the policy scenarios and what subsequent actions may be required to
establish power system security, in case.
2. Analysis of the minimum level of synchronous generation to be applied to
the Clean Energy Target and Limited Lifetime (CET&LL) scenario.

Given the concerns in running low-inertia power systems that might result from
large penetration of Variable Renewable Electricity (VRE) sources, focus has
been placed on frequency response adequacy, i.e., on assessing whether the
portfolio of given resources is frequency-stable following a contingency event,
e.g., the loss of the largest operating generator.

To that end, a suite of tools has been specifically built. In particular, as a key
output, a new bespoke model has been developed to perform a Frequency
Response Security Constrained Optimal Power Flow (FRSC-OPF) that explicitly
takes into account change of system inertia levels and considers new operational
constraints on a key set of frequency response parameters. The results from the
FRSC-OPF are compared to a base case simulation platform resembling the
current provision of Frequency Control Ancillary Services (FCAS) with only static
requirements on frequency response and no consideration for system inertia.

This report is organised as follows. Section 2 introduces key definitions and


concepts for the analysis of low-inertia power systems, which are then used to
build the security assessment methodology presented in Section 3. Section 4
presents the main results from the analysis conducted for the emissions
reduction policy scenarios, while Section 5 focuses on the concept of minimum
synchronous generation with specific application to the CET&LL scenario.
Section 6 then discusses the potential role of new technologies and services.
Various Appendices finally summarize additional information, data and
assumptions used in the analysis.

1‘Security’ is primarily intended here in the sense of frequency response adequacy in a low-inertia
power system context.

16
2. Key definitions and concepts for frequency response
security assessment

2.1. Key definitions and concepts

2.1.1. Definition of “security” for the purpose of this work

For the purposes of this work, focus has been placed on frequency response
adequacy aspects, which may become a particularly severe issue with larger
penetration of non-synchronous renewable energy sources and therefore in
lower inertia systems, as already experienced in the recent Black System event in
South Australia in September 2016. More specifically, we have addressed this
general security issue by assessing whether, in each modelling scenario under
analysis, there were sufficient resources and facilities (to be operated with
appropriate operational requirements and strategies) to guarantee adequate
system frequency response following a disturbance (or “contingency”). However,
considerations on the other aspects of power systems security have also been
provided both quantitatively and/or qualitatively.

2.1.2. Frequency response and system inertia

For power systems to operate correctly, energy demand and supply need to be in
balance at all times, on a second by second basis. The system frequency is a
measure of this instantaneous balance and it can only deviate within a narrow
band around 50Hz to prevent equipment damage, disconnection, and potential
cascading effects that may lead to a black system.

Large disturbances or contingencies such as the sudden loss of a generator or the


trip of an interconnector between regions can alter the supply-demand balance
and lead to serious and potentially dangerous frequency excursions.

Suitable means are therefore in place to control the frequency within desired
levels and provide frequency response to contingencies. This is usually obtained
by procuring Frequency Control Ancillary Services (FCAS) in the market from the
available resources.

The main providers of FCAS have traditionally been conventional generating


units that provide automatic response2 to a change of frequency through the so-
called Primary Frequency Response (PFR). However, the demand side can also
provide frequency response support, for example through load disconnection.

A typical example of system frequency response following generation loss can be


seen in the simulation results shown in Figure 2.1. The automatic PFR from
generators usually starts after a few seconds following the contingency, when

2The automatic response of a generator providing primary frequency response (PFR) is usually
activated through its governor that senses a change in frequency and increases/decreases the
generator’s power output in response to a frequency drop/rise accordingly.

17
additional power is injected to stop the frequency drop and then stabilise to a
certain level (“Quasi Steady-State Frequency - QSSF”). Further units then increase
their output to bring the frequency back into the “normal” operational band,
which is often indicated as Secondary Frequency Response 3 (SFR), and
maintain it there until further notice before the 5 minutes dispatch takes over4.

However, before generators or other resources can respond automatically to any


frequency change (which requires inevitable delays due to frequency variation
sensing, activation of the relevant equipment to respond, etc.), there is a natural
response that take place in the system. This is called Synchronous Inertial
Response – or simply “inertial response” – and is associated to the kinetic
energy that is embedded in the spinning masses of the generators (and some
loads, primarily motors) that rotate “synchronously” with each other and at a
speed that corresponds to 50Hz in terms of system frequency. The very moment
that the supply-demand equilibrium is broken, for example due to the loss of a
generator, part of this kinetic energy is released into the system in order to
preserve the energy balance. This comes at the cost of the synchronous
generators slowing down, which means that the frequency also drops. Generally
speaking, the larger the rotating kinetic energy in the system (which depends on
the physical “inertia” of all the machines that are synchronously connected), the
slower the frequency changes; it is in fact this kinetic energy that prevents the
frequency from changing too fast, thus giving time to the PFR to respond.

The amount of kinetic energy in the system (measured in MWs) is thus generally
associated with “system inertia”5.

3 This is normally done following control signals by the system operator, but can also be done
independently, depending on the market and system operation environment.

4 Other frequency control means also interact at this level, such as the “Regulation” service
activated by Automatic Generation Control (AGC) by AEMO, so that the frequency can return to
its nominal 50Hz value.

5 Even if there may be different definitions of “inertia”, the fundamental physical concept is
nevertheless associated to the amount of kinetic energy in the system (in [MWs]) at any point in
time. For the purposes of this work, we will therefore use system “kinetic energy” and “inertia” as
synonyms.

18
Figure 2.1: Example of frequency response services timescales

2.2. Frequency response in the current NEM

2.2.1. Current Frequency Control Ancillary Services (FCAS) arrangements

There are eight markets in the NEM for procuring FCAS6, which are co-optimised
with energy dispatch every five minutes, namely, two for Regulation and six for
Contingency at different time scales of delivery. More specifically:
1. Regulation Raise: Regulation service used to correct a minor drop in
frequency.
2. Regulation Lower: Regulation service used to correct a minor drop in
frequency.
3. Fast Raise (6 Second Raise): 6 second response to arrest a major drop in
frequency following a contingency event.
4. Fast Lower (6 Second Lower): 6 second response to arrest a major rise in
frequency following a contingency event.
5. Slow Raise (60 Second Raise): 60 second response to stabilise frequency
following a major drop in frequency.
6. Slow Lower (60 Second Lower): 60 second response to stabilise frequency
following a major rise in frequency.
7. Delayed Raise (5 Minute Raise): 5 minute response to recover frequency to
the normal operating band following a major drop in frequency.
8. Delayed Lower (5 Minute Lower): 5 minute response to recover frequency to
the normal operating band following a major rise in frequency.

6AEMO Guide to ancillary Services: https://www.aemo.com.au/-/media/Files/PDF/Guide-to-


Ancillary-Services-in-the-National-Electricity-Market.ashx

19
Focusing on the FCAS contingency services, the Fast service aims at arresting a
change in system frequency within 6s following a frequency event, then orderly
transitioning to the 60s Slow service which in turns aims at stabilizing the
frequency following the event. Broadly speaking, these two combined services
can be associated to the provision of inertial response and Primary Frequency
Response (PFR) services discussed above 7 and are typically activated
automatically through generator governor response or load control through local
frequency measure. The Slow service then provides an orderly transition to the
Delayed service; the latter broadly corresponds to a Secondary Frequency
Response (SFR) service, aiming at restoring the frequency towards its nominal
level within 5 minutes and until further notification from the next dispatch
intervals. This is usually provided through rapid unit loading/unloading based
on delayed frequency threshold controllers.

2.2.2. Contingency FCAS requirements calculation

The general rule to calculate the contingency FCAS requirements8, in MW, is

Requirement = Contingency Risk – Load Relief

where the “Contingency Risk” is the potential MW change for a contingency


(largest generator, load block, etc.), while the “Load Relief” corresponds to the
demand variation due to the relevant frequency variation reference for the
specific service9, proportional to the frequency variation itself and the demand
level at the time of the contingency.

The load “relief”, also known in the literature as “damping”, always goes in the
direction of alleviating the frequency variation (hence its name), mainly caused
by electric motors connected as load drawing, for example, less active power as
the frequency drops.

Based on this discussion, it can be appreciated how the current FCAS


requirements are “static”, in the sense that they only depend on the largest
credible contingency and the incumbent demand level, but do not depend on
other specific system conditions (e.g., amount of inertia in the system).

7 It has to be highlighted that, for the sake of terminology, the inertial response is often implicitly
incorporated within the primary frequency response.

8The general rule applies for different types of contingency, including generation, network, etc.,
while specific numerical values to be used in the relevant parameters may change with the case.

9For both Slow and Fast services the reference frequency level for load relief calculation is
49.5Hz, while for Delayed service it is 49.85Hz (0.15 Hz is the “dead-band” currently adopted for
normal operation before a contingency service is to be activated).

20
2.3. Frequency response challenges in a low inertia power system

As previously mentioned, it is intuitive that the larger the amount of kinetic


energy in the system, the more there will be “inertia” to changes in the system
frequency. It is in fact the inertial response that, by slowing down frequency
changes, allows “buying” sufficient time for the controls available in the system
to have sufficient time to react and generators to provide PFR before reaching
frequency limits. Therefore, it is also intuitive to understand that if the level of
inertia in the system decreases, new challenges will arise in terms of keeping the
frequency within the predefined operational bands.

Most VRE sources are physically decoupled from the system frequency as they
are asynchronously connected to the grid through power electronic
converters/inverters and, therefore, do not contribute to system inertia. These
sources include, in particular, modern wind turbines based on Doubly-Fed
Induction Generator (DFIG) and on full power converter connected generators,
as well as Photovoltaic (PV) generators. Therefore, with more and more energy
output from wind and PV, the system level of inertia could inevitably drop.

In addition, the secure provision of Primary Frequency Response (PFR),


traditionally provided by governor response of conventional generators, is
challenged too. In fact, now not only are there fewer conventional generators
online capable of providing PFR (as they are partly displaced by renewable
generators in system operation), but they also have less time to react since the
lower system inertia leads to faster frequency dynamics. Eventually, the overall
PFR available from these online conventional generators might not be sufficient
to cope with these faster frequency variations.

An example of the potential frequency challenges that could be found in lower


inertia systems with more VRE output is shown in Figure 2.2, whereby the
frequency change following a same large generation loss is analysed for two
different levels of inertia and same demand.

In this example the PFR amount is deliberately the same for the two inertia
cases, and calculated on the basis of the current static requirements on the
contingency and incumbent demand levels. This yields a same “static” level of
Quasi Steady-State Frequency (QSSF), which is in fact independent of the level
of inertia in the system as it only depends on the steady-state output from the
generators scheduled to provide PFR reserve.

However, a lower inertia system could experience a much lower minimum


frequency (technically called frequency “Nadir”); in addition, the frequency
Nadir occurs much earlier with lower inertia, which, as mentioned above, gives
less time to the scheduled PFR resources to react.

Another key parameter of relevance in assessing frequency response adequacy is


the Rate Of Change Of Frequency (ROCOF) [Hz/s] immediately after the
contingency. Clearly, the ROCOF is much higher for the lower inertia system,
which also contributes to reaching a faster and lower frequency nadir.

21
Figure 2.2: Illustration of the effects of lower inertia in system frequency response (with
focus on the inertia and primary frequency response timescales)

2.4. Potential need for new operational requirements and constraints

2.4.1. From static to dynamic frequency response requirements

As mentioned earlier, the current FCAS arrangements correspond to only “static”


conditions corresponding to steady-state frequency reference levels and not
changing with system conditions such as the level of inertia. There is, in fact, an
implicit assumption that there is sufficient inertia in the system to provide
adequate frequency response based only on the static requirements.

However, as seen from the above example, the frequency response of the system
may change significantly with system inertia even if the scheduled generation to
provide PFR (which determines the QSSF) is the same, and there is no guarantee
that ROCOF and/or frequency Nadir do not breach any operational limit.

In lower inertia systems there may therefore be scope for reconsidering the
operational strategies that have historically proven very effective in “traditional”
power systems where synchronous inertia is plentiful10.

10In fact, AEMO considers frequency response requirements that are inertia-dependent for
example for Tasmania (AEMO, “Proposed FCAS Calculation Changes in Tasmania”, June 2014).
Furthermore, prior to the September 2016 Black System event in South Australia, AEMO
implemented inertia constraints in SA under certain circumstances (AEMO and Electranet,
“Update to Renewable Energy Integration in South Australia”, Feb 2016), and these were
increased following that event (AEMO, “Black System South Australia 28 September 2016”,
March 2017).

22
Furthermore, as the level of VRE changes in time, so does the system inertia and
therefore the inertial response, so that, overall, new frequency response
requirements that change dynamically and in function of the incumbent system
operation characteristics may need to be introduced.

Considerations on ROCOF and frequency Nadir limits

ROCOF should not normally exceed certain values to prevent tripping of


embedded generation or damage to conventional generators. Typical values of
maximum allowed ROCOF in different countries varies between 0.125 and 1
Hz/s, and depends on several factors, including the specific embedded
generation protection settings, the size of the system and the penetration VRE
sources. The ROCOF might also be limited in order to indirectly prevent
unacceptably low frequency Nadir or Under-Frequency Load Shedding (UFLS). A
ROCOF limit of 1Hz/s is used in this work, which is considered a reasonable
upper bound for future operation.

The frequency Nadir should not fall to below 49 Hz, since this may also
jeopardise system operation in different ways. UFLS is therefore usually
activated at frequency levels around 49Hz.

2.4.2. Mapping inertia-dependent frequency response requirements and


constraints

A visual representation of new frequency response requirements that might be


required in low-inertia systems is provided in Figure 2.3, showing the different
relations between the required PFR and system inertia after contingency for the
set of static and dynamic security constraints that are used in the analysis. These
constraints, arising from the solution of the fundamental system frequency
response dynamic equation, guarantee the ability of the system to exhibit a
frequency response within a specified minimum frequency Nadir level, keep the
ROCOF within a maximum limit, and reach a predetermined quasi-steady state
frequency level.

In particular, the new requirements are mapped in a plot PFR vs Inertia, where,
for a given contingency level:

Vertical lines depict the maximum acceptable ROCOF. In fact, the minimum
inertia requirement to comply with ROCOF only depends on the size of the
contingency in the system. This means that, for a given contingency size,
each ROCOF limit will produce one level of minimum inertia requirement.
Therefore, by changing the ROCOF limit ceteris paribus there will be a family
of vertical lines in the plot.
Horizontal lines are associated to the static primary frequency response
requirement to reach quasi-steady state. The PFR requirement is calculated
considering the quasi-steady state frequency requirement, the size of the
contingency, the demand damping factor, and the demand level. In
particular, keeping all other parameters constant, different demand levels
will correspond to a family of horizontal lines. The higher the demand level,
23
the more load relief due to the damping factor will be provided, and the PFR
requirement will decrease.
Hyperbolic-like curves determine the response requirement to meet the
minimum allowed frequency Nadir for different demand levels. Since again
the demand level plays a substantial role in relieving the frequency change
after contingency, it also impacts on the joint requirements on PFR and
inertia levels to constrain the frequency Nadir. For higher demand levels and
given a target minimum frequency Nadir level, in particular, the minimum
required combined amount of PFR and inertia decreases.
For a given system operation point, that is, for given demand level, contingency
size, load damping factor, desired ROCOF, frequency Nadir and quasi-steady-
state frequency limits, then the security map will have only one vertical line
(ROCOF requirement), one horizontal line (static requirement) and one
frequency Nadir curve requirement. Since all constraints represent a minimum
requirement for PFR, inertia, or the combination of both, the area of secure
operation points is above the intersection of all constraints, as from depicted in
Figure 2.3.

Figure 2.3: Frequency response requirements map and secure area of operation for
different levels of primary frequency response and system inertia, for given demand and
contingency levels

In particular, when looking at the system inertia levels, it can be noticed how for
high-inertia systems the static PFR requirements dominate the dynamic
frequency Nadir requirements in spite of the inertia level. This explains why
inertia is generally not explicitly modelled in frequency response requirements
in systems where synchronous generators are plentiful, and there is no need for
co-optimizing inertia and FCAS. However, while moving towards lower inertia
systems, the frequency Nadir requirements could dominate the static
requirements, and for each inertia level there is a minimum PFR that needs to be
provided to meet the frequency Nadir, which calls for co-optimization of inertia
24
and FCAS. Furthermore, inertia would also need to be limited to prevent the
system from operating in a region with excessive ROCOF.

2.5. Further considerations on system security aspects not covered

It should be pointed out that, while frequency stability boundaries are likely to
be most critical in a lower inertia NEM, there are a number of other aspects that
should be considered for a full system security assessment11 (see also Appendix,
Section 8), especially in systems with high proportion of renewables, which were
out of the scope of this work.

These include issues such as: rotor angle stability (transient stability, inter-area
oscillations, etc., which might be exacerbated while inertia varies in both time
and space across regions with different VRE types, installed capacities, and
outputs); voltage stability and reactive power support (and in particular to
guarantee certain dynamic performance of asynchronously connected VRE
sources); availability of sufficient short-circuit strength from all generating units
or other resources (e.g., synchronous condensers) for adequate activation of
protection systems12; supply adequacy to guarantee that there are sufficient
resources (including assessment of the “capacity value” of renewables and of
“energy” availability from hydro-resources) in the system to meet the peak
demand; and so forth. All these issues could change the actual synchronous
generation output requirements, for which further specific dynamic and
geographically detailed studies are required.

Further aspects to be considered may also include additional frequency control


requirements (e.g., for regulation) as well as new balancing and reserve
requirements (e.g., for ramping flexibility) (see also Section 5.3) to deal with
larger uncertainty and variability due to VRE output.

11 This is also discussed in AEMO's Advice on the Integration of Energy and Climate Policy, which
is an addendum to the 2016 AEMC report titled Integration of energy and emissions reduction
policy.

12Asynchronously connected plants that rely on power electronics interface do typically provide
limited contribution to fault currents.

25
3. Methodology for frequency response security assessment

3.1. Methodology for security assessment

3.1.1. General approach

The fundamental research question to assess frequency response adequacy


given a certain portfolio of resources can be formulated as:

“Is a given portfolio of resources able to provide a stable frequency response


following any contingency (such as loss of a generator or interconnector) under
any system incumbent condition (of demand, VRE and inertia, in particular)?”

In order to address this question, and based on the above considerations, two
sets of studies have been performed to assess the security of the system dispatch
with the resources available from the results of the modelling scenarios
provided, considering first principles of dynamic system response:

1. “Unconstrained Nadir” case: this is the base case whereby schedule of FCAS
contingency services is formulated in line with current “static” requirements.
2. “Constrained Nadir” case: the schedule of FCAS contingency services is in this
case formulated as from a new set of requirements that are dynamically
quantified taking into account the time-changing system environment and
new operational constraints13, namely:
a) Maximum allowed ROCOF;
b) Minimum allowed Frequency Nadir;
c) Minimum allowed QSSF14 in correspondence of the reference frequency
values as from the current static requirements.

3.1.2. “Unconstrained” frequency Nadir case (base case)

In this case, a conventional reserve-constrained optimal power flow (OPF)


analysis is performed for every relevant snapshot where energy output and
static FCAS reserve requirements are co-optimised in a least-cost fashion subject
to generation constraints15 and interconnector flow constraints16.

13The modelling of the new constraints has been developed from first principles, based on a
suitable approximation of the solution of the system frequency response dynamic equations.

14Requirements on QSSF are usually adopted to guarantee that the frequency is not at
dangerously low levels for too long, e.g., up to a few minutes (in case a new event were to
happen), and to allow successive control schemes (such as secondary frequency response and
further reserves that could also be manually activated by the system operator) to bring back the
frequency to secure pre-contingency levels.

15Such as minimum stable generation, maximum operational capacity, transient and nominal
ramp rates, available headroom vs frequency response provision functions, etc.

26
The FCAS schedule corresponds to the current static frequency response
requirements for 6s, 60s and 5 minute services, which, as aforementioned, are a
function of the largest credible contingency and the incumbent demand level at
every dispatch interval. Therefore, in this “base case”, there is no explicit
consideration for frequency Nadir and ROCOF constraint, and inertia is not
explicitly modelled or captured.

After running the OPF according to the current FCAS requirements, the three
frequency response adequacy metrics mentioned above, namely, ROCOF,
Frequency Nadir and QSSF are checked for compliance with the predefined
limits. The relevant inertia level yielded as a “by-product” of the dispatch is also
recorded.

3.1.3. “Constrained” frequency Nadir case

Frequency Response Security Constrained Optimal Power Flow (FRSC-OPF)

A bespoke operational optimization model was developed so as to be able to run


a Frequency Response Security Constrained Optimal Power Flow (FRSC-OPF) that
is representative of possible operational strategies and algorithms that could be
implemented in the system dispatch engine in consideration of lower inertia
scenarios. The model is based on the same OPF tool used for the “unconstrained”
frequency Nadir case above, but also explicitly accounts for ROCOF and
frequency Nadir constraints in addition to the static FCAS constraints.

Dynamic calculation of FCAS requirements and co-optimization of FCAS and inertia

In the proposed FRSC-OPF, at every considered dispatch interval FCAS


requirements are dynamically calculated and co-optimised with the appropriate
level of system inertia17 from synchronous machines18 so as to satisfy all static,
ROCOF and frequency Nadir constraints (as well as of course the required energy
output).

The algorithm thus yields, for each dispatched unit, the operational setpoints
(corresponding to the energy output to satisfy the net demand), the amount of
scheduled FCAS (6s, 60s, and 5min) and the headroom required to deal with the
relevant ramp rate limits, and the overall online capacity that corresponds to the
required inertia in the system so that all the above frequency response
constraints are satisfied.

16A DC power flow model is assumed, for the sake of simplicity and given that the interaction
with reactive power flows and voltage control can be at first approximation neglected in the
problem under analysis and considering the planning nature of the study.

17Effectively, since inertia can be seen as another (instantaneous) form of frequency response
provided by synchronous kinetic energy, as mentioned earlier, the FRSC-OPF performs a co-
optimization of all frequency response services and resources available in the system.

18Primarily synchronous generators, which also provide FCAS, but they could also be the likes of
synchronous condensers or loads (e.g., pumped-hydro operating in pumping mode).

27
As in low inertia systems the speed of response (and effectively the magnitude
and extent of the overall system ramp) may have to increase significantly, it is
expected that faster PFR will be required alongside the co-optimised level of
inertia. To allow easy comparison with the current requirements, the PFR level
scheduled from the FRSC-OPF is then also calculated in correspondence of 6
seconds19. In fact, the fast FCAS requirement is effectively replaced by the
frequency Nadir constraint when these become active (“binding”), whilst
enabling frequency Nadir and ROCOF constraints does not modify the slow (60s)
and delayed (5min) FCAS static requirements.

It is also worth pointing out that in those cases when the ROCOF and frequency
Nadir constraints are not binding, the FRSC-OPF algorithm would schedule the
same resources as in the “unconstrained” case, meaning that the only actual
binding requirements are the static ones.

3.1.4. Generation/demand combination snapshots considered in the


analysis

For each scenario and year, a number of snapshots corresponding to a number of


plausible combinations of demand and generation outputs have been considered.
In particular, we looked into various possible outputs from VRE (e.g., high wind-
zero solar during the night, low wind-low solar or high wind-high solar during
the day, etc.) that could drive dispatch with less inertia in the system20, as well as
two extreme demand conditions, namely, maximum and minimum demand.

Times of minimum demand are of particular interest for system security


assessment for a few reasons, namely:

- It is more likely that the relative share of VRE covering demand is higher and
therefore the overall synchronous inertia in the system lower;
- There may as well be fewer online synchronous generators available to
provide inertia and reserves21;
- The load relief is proportional to the incumbent demand level, so at
minimum demand the “relief” effect is also minimal.

However, especially for large amount of installed VRE capacity, situations may
arise where VRE output could even cover the peak demand, and of course all
intermediate situations too. Hence, systematic analysis has been carried out for a
range of VRE outputs and both min and max demand levels.

19 However, as in a low inertia system the ROCOF can be much higher and the frequency Nadir
can be much lower and happen much earlier, it may happen that generators will have to provide
relevant output much before 6 seconds, even though the “official” scheduled PFR is calculated in
correspondence of 6 seconds.

20VRE has been assumed with priority dispatch. However, VRE output could be curtailed for
security reasons.

21 Unless dispatched with specific inertia-dependent constraints.

28
3.1.5. Summary of the methodology for security assessment

A summary of the security assessment methodology is reported in Table 3.1,


with indications of the relevant operational constraints, FCAS requirements and
(additional) security requirements used in the “unconstrained” (base reserve-
constrained OPF) and “frequency Nadir-constrained” (FRSC-OPF) cases as well
as allocation to one of the two “classical” reserves used, namely, automatic
primary frequency response PFR or secondary frequency response SFR.

Table 3.1: Constraints and reserves used in the security assessment methodology

Constraint Base OPF FRSC-OPF Response

Generating units’ operational limits Y Y

Interconnectors’ flow limits Y Y

FCAS reserve requirement @5min Y Y SFR

FCAS reserve requirement @60s Y Y PFR

FCAS reserve requirement @6s Y Y PFR

ROCOF constraint N Y Inertial

Nadir constraint N Y Inertial + PFR

3.2. Considerations on minimum synchronous inertia and interaction


among inertial, primary and secondary frequency response

As discussed earlier, inertia is key to limit the ROCOF at the initial times
following the event, when the kinetic energy from the spinning synchronous
machines slows down the frequency change. Therefore, increasing the level of
inertia in the system may be seen as a means to improve frequency response, at
least in terms of slowing down the frequency drift through inertial response.

There is ongoing research into a minimum synchronous inertia or generation


level or, equivalently, the maximum System Non-Synchronous Penetration
(SNSP) allowed in the system, defined as a measure of the non-synchronous
generation in the system at an instant in time as a percentage of the demand22.

However, increasing system inertia on its own (which can be obtained by


committing more generators online and therefore providing an upper limit to the
SNSP metric) may not be sufficient to guarantee an acceptable frequency Nadir
level, for which also a suitable level of automatic PFR needs to be scheduled to
stop and stabilise the frequency drift.

22J. O’Sullivan, A. Rogers, D. Flynn, P. Smith, A. Mullane, and M. O’Malley, “Studying the Maximum
Instantaneous Non-Synchronous Generation in an Island System—Frequency Stability
Challenges in Ireland”, IEEE Trans. on Power Systems, Vol. 29, No. 6,Nov 2014, pp. 2943-2951.

29
Furthermore, the synchronous generation output is only a proxy of the actual
level of system inertia, which is associated to the online capacity 23 of a
synchronous unit rather than its current output. For example, assuming a power
factor equal to one and a same inertia constant (e.g., 5MWs/MVA) for different
units of a same technology but with different size24, a synchronous generating
unit of 500MW capacity operating at 250 MW output would contribute the same
kinetic energy as two 250MW capacity units each producing their full output;
however, the contribution of the two sets of generators to the SNSP index,
calculated on the basis of their energy output, would be different.

The proposed FRSC-OPF effectively co-optimizes all the different frequency


response options available in the resource portfolio, that is, by committing online
synchronous units that can provide an adequate frequency response following a
contingency, from very fast inertial response to PFR and SFR.

Therefore, as an output of the modelled FRSC-OPF, at every dispatch interval


there is a specific value of minimum synchronous output (or maximum SNSP
level) that is yielded alongside the corresponding inertia level (given by the
spinning synchronous capacity) and the required levels of primary and secondary
reserves to be scheduled to provide suitable frequency response. In other words,
there is a complex interplay between actual energy output and inertial, primary
and secondary reserves that need to be scheduled, and the FRSC-OPF optimizes
this interaction.

23 The kinetic energy of a synchronous unit is in fact defined with respect to its MVA capacity.

In reality, for a given technology the inertia constant may slightly change with the size.
24

However, this does not affect the generality of the discussion.

30
4. Scenario analysis
This Section presents main results from scenario analysis, providing first a
comparative overview of the scenarios and key results, and then discussing more
in detail the CET&LL scenario as an example and sensitivities around the main
parameters used in the study.

4.1. Scenarios overview

4.1.1. The modelling scenarios

The five modelling scenarios are listed in Table 4.1, while a more detailed
description is presented in the Appendix, Section 8.

An overview of the demand input data and technology investment resulting from
the modelling is also provided below.

Table 4.1: Modelling scenarios and relevant emission targets

Scenario Target in 2030

Business as Usual (BAU) NA

Limited Lifetime (LL) No target

Clean Energy Target and Limited Lifetime 28% below 2005’s level
(CET&LL)

Clean Energy Target (CET) 28% below 2005’s level

Emissions Intensity Scheme (EIS) 28% below 2005’s level

4.1.2. Input data from modelling scenarios

Electricity demand

A 50% probability-of-exceedance (POE) median peak demand is used for all


scenarios. The minimum level and maximum level of electricity demand (which
is the same for all scenarios) throughout the years up to 2050 is shown in Figure
4.1.

31
Figure 4.1: Range of electricity demand (min to max) for all scenarios to 2050

Installed capacity

The conventional and VRE installed capacity for the NEM varies in every
scenario and year. Hydro installed capacity increases slightly in the period 2017-
2025, but in general it is very similar in the years under analysis. In the following
figures, it is possible to appreciate the amount of total installed capacity, thermal
installed capacity and VRE installed capacity25.

Figure 4.2: Total installed capacity for all scenarios to 2050

25CCGT: Combined Cycle Gas Turbine; OCGT: Open Cycle Gas Turbine; Solar is overall composed
of Concentrated Solar Power (CSP) and solar Photovoltaic (PV); Hydro is composed of run-of-the-
river, hydro with reservoirs, and pumped-hydro.

32
Figure 4.3: Thermal Synchronous installed capacity for all scenarios to 2050

Figure 4.4: Renewables installed capacity for all scenarios to 2050

In terms of synchronous inertia contribution, it is assumed that CCGT, coal,


OCGT, hydro, biomass, and concentrated solar technologies are synchronously
connected, while wind 26 and solar PV (as well as batteries) have
nonsynchronous connection.

Both synchronous and non-synchronous installed capacity present similarities


among some scenarios. This will lead to similar results when comparing
different metrics for security among scenarios, which will be highlighted in the
coming sections.

26In reality, some older wind turbines might be based on synchronous machines, but their
amount is negligible and it is envisaged that all new wind turbines will be non-synchronously
connected through doubly-fed induction generators (DFIG) or full-converter interface.

33
4.2. Scenario results overview

This Section discusses the frequency Nadir, ROCOF, system inertia and Primary
Frequency Response (PFR) requirements 27 considering a (worst case)
generation contingency28 for all scenarios, with a demand frequency damping
factor of 2%/Hz. Both cases when only the current static frequency response
requirements are applied (but without explicit frequency Nadir and ROCOF
constraints, nor any consideration for inertia), and then with also frequency
Nadir constraint of 49.2Hz and ROCOF constraint of 1Hz/s as from the FRSC-OPF
tool discussed above, are considered.

4.2.1. Nadir

If the frequency Nadir is not formally constrained in the system (Figure 4.5), the
minimum level of frequency after contingency might naturally reach
unacceptable levels in each scenario, especially in the farther future with more
nonsynchronous VRE sources connected to the system.

Since the current synchronous capacity in the NEM is high and the amount of
VRE is relatively low, the 2017 “unconstrained” case, when only static PFR
requirements are applied, shows acceptable levels of frequency Nadir both for
high and low demand.

In the following 10 to 15 years, however, this situation might change radically for
minimum demand periods, mainly driven by the massive penetration of VRE for
all scenarios except from the base scenario (BAU). During minimum demand
periods, some extreme cases for frequency Nadir present an irregular evolution
along the years, which depends on specific combinations of technologies under
particular conditions, for instance, hydro capacity availability and must-run
minimum synchronous output due to minimum stable generation constraints,
among others.

On the other hand, during high demand periods the contribution of the natural
frequency response provided by the load damping effect leads to minimum
frequency Nadir levels that are within more acceptable limits (in the order of
49Hz).

27PFR requirements are based on static requirements (response capacity sufficient to meet the
current 6s and 60s FCAS requirements in correspondence of 49.5Hz), as well as on frequency
Nadir and ROCOF constraints in the cases when these are applied (response capacity and online
synchronous machines sufficient to meet the frequency Nadir and ROCOF constraints in addition
to static requirements that always apply). Secondary Frequency Response (SFR) requirements,
corresponding to the current 5 minutes contingency FCAS requirements, do always apply too.

28 Assumed equal to 700MW.

34
Figure 4.5: Nadir level after contingency without constraints for all scenarios, with
demand frequency damping factor of 2%/Hz

If Nadir is formally constrained (alongside the ROCOF) in the dispatch, in this


case to 49.2 Hz (Figure 4.6), the units are dispatched by the FRSC-OPF in a way
that prevents the system’s Nadir to depart from that level, given the maximum
contingency level. It is possible to check consistency among constrained and
unconstrained cases by comparing the Nadir limits where the Nadir constraint is
not active.

Figure 4.6: Nadir level after contingency with nadir constraint of 49.2 Hz for all scenarios,
with demand frequency damping factor of 2%/Hz

4.2.2. ROCOF

The rate of change of frequency immediately after the contingency is a direct


measure of the ratio between the size of the contingency and the aggregated
system inertia (after the potential inertia loss due to the contingency itself). As it
35
can be appreciated in Figure 4.7, at the level of the whole NEM out of the whole
range of possible combinations of demand and VRE outputs, there are only a few
cases, starting from 2030, where ROCOF exceeds the limit of -1 Hz/s. These cases
occur during periods of maximum demand in which the output of renewable
sources is very high and the system inertia rather low too. In general, the worst
level of ROCOF increases by year; this is due to the decrease of installed
synchronous capacity and the associated available synchronous inertia along
with the increase of renewables capacity that might be dispatched in the system
at time of high demand and high availability of VRE output.

Figure 4.7: ROCOF after contingency without constraints for all scenarios, with demand
frequency damping factor of 2%/Hz

When running the system dispatch with ROCOF (limited to -1Hz/s) and Nadir
(limited to 49.2 Hz) constraints, the worst case that is obtained for ROCOF is
close to -0.6 Hz/s (Figure 4.8). The reason for this level of ROCOF to be quite
lower than its actual imposed limit is the fact that it is the Nadir constraint that
determines the frequency response requirements at the whole NEM level. In
particular, the co-optimization of all frequency response services and inertia to
meet the Nadir constraint leads to the scheduling of a number of synchronous
generating units whose equivalent inertia is higher than what would be required
to meet the ROCOF limit of -1Hz/s. This applies in both maximum and minimum
demand conditions, suggesting that, in other words, the 1Hz/s ROCOF constraint
is inactive in all scenarios for all levels of demand due to the higher requirements
that the Nadir constraint imposes on the frequency response security constraint
dispatch.

36
Figure 4.8: ROCOF after contingency with nadir constraint of 49.2 Hz for all scenarios, with
demand frequency damping factor of 2%/Hz

4.2.3. Inertia

The results for the inertia level that is scheduled in the system when no ROCOF
and Nadir constraints are applied are somehow inversely proportional to the
previous results for ROCOF (Figure 4.9). In fact, since the value of the largest
considered contingency size is constant in all scenarios and years, all the results
for ROCOF and inertia are directly comparable. The average level of system
inertia in the considered VRE output/demand combinations is lower for the low
demand level, which is an intuitive result given the lower output requirements
from synchronous generators in that condition. The relevant inertia level for
minimum demand also slightly decreases with time, corresponding to larger
output availability from VRE. At the same time, the range of inertia levels that are
seen for minimum demand gets significantly larger, again due to an increase in
the potential range of VRE output while more renewables are installed in the
system. Similar trends of decreasing average level and increasing range of
operating synchronous inertia can also be appreciated, in an even more evident
way, for the maximum demand conditions. In particular, it is interesting to notice
that in the farther future, when VRE output could be plentiful in most cases so as
to cover significant shares of both maximum and minimum demand, the inertia
levels and ranges for the two demand conditions become much closer and more
similar.

37
Figure 4.9: System inertia after contingency without constraints for all scenarios, with
demand frequency damping factor of 2%/Hz

The amount of system inertia that is scheduled by the FRSC-OPF with


constrained Nadir and ROCOF (Figure 4.10) shows, for the maximum level of
inertia for both maximum and minimum demand, a trend that is very similar to
the unconstrained case. However, the minimum inertia level for both demand
conditions increases when compared with the unconstrained case, and
particularly for minimum demand. This clearly shows the effect of the imposed
minimum allowed Nadir requirement, which leads to the commitment of more
synchronous generating units to increase the system inertia (as well as the PFR
provided by generators) to be able to meet the frequency response constraint.
Further, the effect of increased system kinetic energy is more evident for
minimum demand, as under this condition the frequency response support from
load relief is lower and therefore the system needs to schedule more FCAS and
synchronous inertial response to maintain frequency response adequacy.

38
Figure 4.10: System inertia after contingency with nadir constraint of 49.2 Hz for all
scenarios, with demand frequency damping factor of 2%/Hz

4.2.4. PFR

The minimum PFR allocation that is scheduled when only the current static 6s
and 60s FCAS requirements are applied are limited to the quasi steady-state
frequency target and diminishes in time (Figure 4.11). This is driven by an
increase in both maximum demand and minimum demand levels throughout the
years, which increases the contribution of the load relief and therefore decreases
the PFR static requirements.

As the load relief contribution is directly proportional to the demand level, it can
also be clearly seen that PFR requirements are higher for minimum demand.
Further, as demand levels are the same throughout the scenarios, there is no
change in the minimum PFR requirements (and actual allocation) for the
different scenarios.

However, the maximum level of PFR in some cases is higher than the minimum
static requirement. This is explained by the technical characteristics of the set of
units providing the response and the assumptions on their operation during the
energy-FCAS dispatch, which can lead to an allocation of slightly more PFR than
the minimum requirement.

39
Figure 4.11: Primary frequency response requirements after contingency without
constraints for all scenarios, with demand frequency damping factor of 2%/Hz

When Nadir and ROCOF constraints are in place (Figure 4.12), in some cases the
amount of PFR increases well above the static PFR requirements due to the need
for meeting the Nadir constraint, particularly if the system is running with
relatively low levels of inertia. Once again, the amount of PFR and inertia that are
provided by online synchronous generators are optimally co-optimised and
allocated by the FRSC-OPF tool.

Figure 4.12: Primary frequency response requirements after contingency with nadir
constraint of 49.2 Hz for all scenarios, with demand frequency damping factor of 2%/Hz

4.3. Analysis of specific results for CET&LL scenario

In this Section, the CET&LL scenario is analysed in more detail in terms of Nadir,
ROCOF and synchronous inertia results, focusing in particular on the effect of
different load damping factors and more stringent ROCOF constraints.
40
4.3.1. Nadir assessment for different damping factors

When it comes to analyse the results for frequency Nadir in any scenario, the
Nadir-constrained and -unconstrained cases show clearly the differences in
terms of system response after contingency, as it was presented in the previous
Section. The new element presented in the following picture is thus the effect of
the load relief associated to different demand damping factors.

In particular, the results from the unconstrained case with only static
requirements clearly show the relieving effect of demand on the Nadir level after
contingency, which is much higher for 3%/Hz as opposed to 1%/Hz and for
maximum demand as opposed to minimum demand, as the load relief effect is
directly proportional to the demand level. This is particularly evident for the
1%/Hz and minimum demand case, whereby, due to both lower damping factor
and lower demand, the frequency Nadir could potentially plunge to very low
levels following the worst case contingency, and especially for farther scenarios
with potentially much larger shares of VRE outputs.

Figure 4.13: Nadir level after contingency without constraints and with nadir constraint of
49.2 Hz (CET&LL)

4.3.2. ROCOF assessment for different damping factors

The rate of change of frequency also varies depending on the level of relief
coming from demand. However, in this case a higher load damping leads to
worse ROCOF levels in the case when no formal Nadir constraint is applied. This
behaviour results from the fact that, in order to meet the static FCAS
requirements, the system needs fewer online synchronous units and therefore
schedules less inertia. This in turn means that higher ROCOF could occur. For the
Nadir-constrained case, as mentioned earlier the Nadir constraint is normally
more binding than ROCOF, and therefore the allocation of resources to limit
Nadir will also automatically limit the ROCOF. However, there may also be cases
for which the ROCOF constraint could become active to limit it to a maximum of
1Hz/s. This applies, in particular, for higher damping factor and maximum
41
demand level in future years, where the overall load relief effect (which is
proportional to both the demand level and the damping factor) would be such
that the number of synchronous units scheduled to provide an adequate Nadir
response might not bear sufficient inertia to also constrain the ROCOF to the
defined limit. However, as mentioned above, it is more likely that in the future a
damping factor lower than the 3%/Hz currently used would apply.

Figure 4.14: ROCOF after contingency without constraints and with nadir constraint of
49.2 Hz (CET&LL)

4.3.3. Inertia assessment for different damping factors

Inertia allocation in the system exhibits a behaviour that is in line with the
previous ROCOF analysis. For higher relief (coming from either the damping
factor of demand or the demand level or both) the system generally tends to
commit fewer synchronous generating units, which has a direct impact on
synchronous inertia. For the Nadir-constrained case, in particular, it can be
appreciated how the minimum after-contingency system inertia that is allowed
in the system goes systematically up. This is particularly evident in the 1%/Hz
damping factor case, where, as the load relief effect is less, more online
synchronous units are required to provide inertial response and the other FCAS.
Also, in those cases where ROCOF constraint is active (high demand, damping
factor 3%/Hz, years 2035-2050), it is possible to see that the inertia level is
limited to 17.5GWs, which corresponds to the largest credible contingency of
699.32 MW and the ROCOF limit of -1 Hz/s.

42
Figure 4.15: System inertia after contingency without constraints and with nadir
constraint of 49.2 Hz (CET&LL)

4.3.4. Impact of different ROCOF limits

This Section shows the impact on Nadir and ROCOF from constraining the
system operation with different levels of ROCOF limits, namely, the base case of
-1Hz/s as well as -0.5Hz/s, with a demand frequency damping factor of 2%/Hz.

In Figure 4.16 it can be appreciated that the Nadir constraint of 49.2Hz is always
active, so that the system schedules an adequate and appropriate amount of
inertia and FCAS to meet the Nadir target for both ROCOF limit cases.

On the other hand, Figure 4.17 shows the ranges of ROCOF levels for both
maximum and minimum demand for both the unconstrained case as well as the
cases whereby the Nadir is constrained to 49.2 Hz and the ROCOF constrained to
-1 and -0.5 Hz/s.

From comparison of the two figures, it can be seen how for the case with ROCOF
limited to -1 Hz/s the Nadir constraint is driving the synchronous inertia level to
an extent that the ROCOF level does not activate; in other words, more online
synchronous units are needed in the system to provide suitable frequency
response to meet the minimum allowed Nadir requirement than to meet the
maximum ROCOF requirement.

In contrast, for the case with a ROCOF limit of -0.5 Hz, there are situations in the
farther years (2035-2050) where the ROCOF constraint becomes active for
maximum demand. As also the Nadir constraint activation occurs for the same
situations, the joint analysis of the results indicate that the FRSC-OPF algorithm
optimally schedules the online synchronous units so as to provide the
appropriate amount of inertial response and FCAS so as to limit the minimum
frequency Nadir while also limiting the maximum ROCOF at the desired levels.

43
Figure 4.16: Nadir level after contingency with different ROCOF constraints, with demand
frequency damping factor of 2%/Hz (CET&LL)

Figure 4.17: ROCOF after contingency with different ROCOF constraints, with demand
frequency damping factor of 2%/Hz (CET&LL)

44
5. Minimum synchronous generation analysis for CET&LL
scenario

5.1. General considerations and methodology

The main purpose of this study was to analyse the concept of minimum
synchronous generation to make the NEM secure and provide indication as to
what this level would be in the CET&LL scenario.

As discussed earlier, there may be different approaches and indices that can be
used to provide some constraint to system operation in terms of inertia, such as
limiting the level of nonsynchronous generation operating in the system at any
time. For example, this limit could be enforced through a mechanism that relies
on the System Non-Synchronous Penetration (SNSP) index29. In Ireland, this
mechanism currently applies a 55% limit on non-synchronous generation at any
instant. This limit is to be increased to reach 75% in the future.

As the SNSP% limit is applied to variable system conditions, and particularly to


variable levels of demand and renewables, the maximum amount of
nonsynchronous generation output allowed in the system, measured in MW,
actually varies in time.

Another potential approach could be to identify a minimum amount of


synchronous generation that should always be in the system, for example in MW
or in % of a demand reference (e.g., the peak demand or the actual demand at a
given point in time).

However, all these approaches are only proxy of the actual requirements, since
there is a tight interplay between inertial response and other frequency control
mechanisms to provide adequate frequency response, and in particular what
matters is the synchronous kinetic energy (inertial response) that is associated
to the spinning capacity (not output) of the synchronous machines, as well as the
schedule of the other FCAS that is associated with having a certain headroom to
provide reserves.

The explicit co-optimization of all response services (including inertia) that can
be carried out through the proposed FRSC-OPF is thus able to schedule for every
system condition the right amount of each service, and in simple terms the actual
synchronous and nonsynchronous generation outputs (that may be used to
calculate the SNSP index or the likes) as well as the spinning capacity (which
intrinsically bears system inertia, but also brings adequate headroom to provide
the required FCAS).

29 http://www.eirgridgroup.com/site-
files/library/EirGrid/OperationalConstraintsUpdateVersion1_47_December_2016.pdf

45
Based on the above considerations, two approaches have therefore been
followed to assess a potential minimum level of synchronous generation:

- A static assessment of one conservative level of minimum synchronous


generation level, in MW, to be always online in the system, for each year; this
analysis has been carried out through the FRSC-OPF tool.
- A detailed dynamic assessment of a minimum synchronous generation level
at every half-hour interval for the years 2020 and 2050; a bespoke frequency
control security constrained unit commitment (FRSC-UC) has been used in
this case, which also considers intertemporal constraints and flexibility
reserves and could be more relevant in future years.

5.2. Static assessment of minimum synchronous generation level through


FRSC-OPF

5.2.1. Methodology

The minimum level of synchronous generation (in MW) to be online at all times
to maintain security has been determined by running the FRSC-OPF tool
constraining the system to comply with both a maximum value of ROCOF (set
equal to 1Hz/s) and a minimum frequency level (“nadir”), set equal to 49.2Hz,
and considering a damping factor equal to 2%/Hz.

The same maximum contingency level equal to 700MW (largest generator, as


from the modelling scenario results) has been considered in all studies.

For every year, we have performed analysis for both minimum and maximum
demand and for different levels of renewables outputs (relative to their installed
capacity in the specific year considered), as from the input data in the modelling
scenarios as well as for extreme cases to stress test the system under credible
situations.

This refers in particular to a condition with minimum level of demand and high
output of wind (e.g., 90% of installed capacity). In fact, under these conditions,
the low-demand load relief effect due to the damping parameter is fairly limited
and therefore frequency response is more challenged, as already discussed.

5.2.2. MW value of minimum synchronous generation per year

The resulting requirements for minimum synchronous generation capacity


(in MW, by year) that should always be online have been provided in Figure 5.1.

The plot was drawn under the assumption of one single MW figure for the
minimum synchronous generation to be always online in the system. This
corresponds to a worst-case condition, yielding a certain MW level that generally
corresponds to a low demand/high renewable case30. This is also a generally

30 A wind output equal to 90% of its installed capacity has been considered.

46
conservative number and only a proxy of the actual requirements that change
with the system condition, as discussed, and that could be fully captured through
the dispatch results of the FRSC-OPF tool. However, such a number could be
useful for instance for applications in long-term expansion tools as the one used
to model generation investment in the scenarios analysed here, where the
modalities to provide security constraint inputs may be limited.

From the picture, it can be appreciated how the minimum synchronous


generation capacity decreases in different year and then tends to saturate from
about 2035. In fact, in 2017 to 2030 the wind penetration is not yet substantial,
and the synchronous capacity is needed for adequacy purposes too to cover the
minimum demand. In contrast, from 2035 there is significant wind which
increases up to 2050. However, no matter how much wind there is in the system,
from 2035 the minimum synchronous generation level constraint becomes
active in the system, and thus does not change significantly with time (minimum
demand level is about constant throughout those years).

Figure 5.1: Minimum online synchronous generation capacity and generation output
(calculated in a min demand/high VRE case) by year, with Nadir constraint of 49.2 Hz and
load damping factor of 2%/Hz (CET&LL)

5.2.3. Comparison with the SNSP index

The SNSP index is calculated based on the actual VRE output relative to demand.
Therefore, taking the 2050 case as reference, for a low demand level of about
17GW, 4GW of synchronous output corresponds to about 24% (i.e., SNSP=76%).
This is aligned with the 75% limit that Ireland could consider as future target,
even though one needs be careful to compare systems given the complexity of all
the different parameters that interplay (size of the system, size of the
contingency, demand level, load damping factor, etc.).

47
However, it also needs to be highlighted that under worst case conditions (min
demand/high wind) a secure frequency response with synchronous generators
providing 4GW output also needs to be accompanied by relevant provision of
inertia and FCAS that require in total 8GW of output, which corresponds to about
48GWs of inertia and a spare capacity of around 4GW. Without information on
the actual online synchronous capacity (which is the real physical inertia driver),
the output per se does not provide sufficient information.

On the other hand, for high demand level in the order of 35GW, 4GW of
synchronous output corresponds to about 12% (i.e., SNSP=88%). Once again,
this per se does not provide sufficient information on the frequency response
adequacy of the system and requires information on the actual spinning capacity.
If the same 8GW of spinning capacity were assumed, it is very likely that the
corresponding 48GWs of inertia could be overconservative in many cases. This
can for example be seen from Figure 5.1, in which the absolute minimum system
inertia level for maximum demand, while still meeting all frequency response
constraints, is in the order of 25GWs. Similarly, 4GW of headroom would often
be overconservative also in terms of provision of the required FCAS for the
maximum demand case.

In summary, indicators that refer to relative levels of synchronous and


nonsynchronous output need to be treated carefully, as the physical
requirements of inertia (as kinetic energy) and interplay between different
frequency response services (inertial response, primary and secondary) may not
be properly acknowledged by such indicators.

5.3. Dynamic assessment of minimum synchronous generation levels


through Frequency Response Security Constrained Unit Commitment
(FRSC-UC)

5.3.1. Problem description and methodology

In this section, a detailed dynamic assessment of a minimum synchronous


generation level at every half-hour interval for the years 2020 and 2050 is
presented. The analysis is performed through a bespoke frequency control
security constrained unit commitment (FRSC-UC) which is an adaptation of a
flexibility assessment-oriented unit commitment model 31 with inclusion of
ROCOF and Nadir constraints. The model also considers intertemporal
constraints and flexibility reserves and could be more relevant in future years

31L. Zhang, T. Capuder and P. Mancarella, “Unified unit commitment formulation and fast multi-
service LP model for flexibility evaluation in sustainable power systems”, IEEE Transactions on
Sustainable Energy, vol. 7, no. 2, pp. 658-671, April 2016.

The Unit Commitment model includes all most relevant operational constraints and cost
functions, such as minimum stable generation, maximum generating capacity, minimum up and
down times, maximum transient and nominal ramp rates, part-load marginal cost functions,
start-up costs, etc.

48
with significant amount of VRE output whose variability and uncertainty might
pose substantial balancing challenges to the system.

The detailed analysis throughout a year aims at providing insights into the actual
amount of time when specific FCAS requirements may be needed, and
particularly in terms of having to activate ROCOF and Nadir constraints.

Furthermore, a techno-economic assessment of the potential costs and benefits


from adopting certain operational strategies/technologies is provided
For brevity, only the cases of 2020 (closer to today’s situation) and 2050 (closer
to extreme cases in the future) for the CET&LL scenarios are considered.

For the purpose of assessing the implications of having additional security


constraints relative to the current conditions, three cases are considered in each
case study example:

- “Nadir-unconstrained”, which represents the base case, where the three 6s,
60s and 5 minutes FCAS static requirements are modelled.
- “Nadir-constrained” case, with both ROCOF and frequency Nadir constraints
added to the static requirements of the base case, as for the FRSC-OPF.
- “Flexibility-constrained” case. In this case, additional reserve requirements
are added to deal with expected very large penetration level of VRE,
particularly to actively tackle its potentially large uncertainty and variability
that might lead to insufficient ramping capacity of the system and
consequently higher operational costs (if more standing reserves or load
shedding where required) or even infeasible scheduling of the system32.

A summary of the constraints and requirements for FCAS and reserves is


provided in Table 5.1.

32Following (B. Kirby, E. Ela, M. Milligan, “Analysing the impact of variable energy resources on
power system reserves”, in Renewable Energy Integration, Elsevier, 2014) a 3σ rule has been
applied to consider variability of load, solar and wind in the 5-minutes timescale as well as
forecast uncertainty for load, solar and wind in the one-hour timescale, with σ being the standard
deviation of the relevant variability/forecast uncertainty functions, estimated from current
information.

49
Table 5.1: Constraints used in different UC modelling approaches

“Nadir- “Nadir- “Flexibility-


unconstrained constrained” constrained”
Static PFR @6s Y Y Y

Static PFR @60s Y Y Y

ROCOF constraint Y Y Y

Nadir constraint -> “dynamic” PFR N Y Y

FCAS @5min (secondary response) Y Y Y

FCAS @5min + variability reserve N N Y

one-hour ahead uncertainty reserve N N Y

5.3.2. Results

Impact of Nadir constraint

The first study performed is to understand the performance of the system


with/without Nadir and ROCOF 33 constraints as well as considering the
additional flexibility reserve. In general, very similar results can be found for the
“Nadir-constrained” and “flexibility-constrained” cases.

The half-hourly duration curves of the system Nadir with/without Nadir


constraint and with flexibility constraint in 2020 and 2050 are shown in Figure
5.2. From Figure 5.2, it can be seen that for 2020 the system frequency does not
drop below 49.2Hz following a contingency, suggesting frequency response
adequacy of the NEM in the next few years. However, if the Nadir constraint is
not imposed by 2050, the Nadir can be lower than 49.2 Hz in some 40% of the
time throughout the year, with a minimum value lower than 49 Hz.

It can be appreciated how the resulting Nadir values for all cases are higher than
in the studies with FRSC-OPF. In fact, a unit commitment with inter-temporal
constraint naturally “forces” more power plants to be online to minimise start-
up and shut-down costs and abide by minimum up and down time constraints, so
that more inertia is automatically scheduled in the system, resulting in improved
frequency response performance.

In this respect, the duration curves of system kinetic energy for the different
cases are depicted in Figure 5.3. It can be noticed that the range of kinetic energy
is between 70GWs to 167GWs in 2020. In 2050, the minimum level of the kinetic
energy could instead drop to as low as 29 GWs without Nadir constraint, while
this minimum level increases to 34GWs when the constraint is activated, as more

33However, the 1Hz/s ROCOF constraint is basically never active, that is why the focus is put on
analysing the frequency Nadir results.

50
generators are needed online to provide extra inertial response and PFR (as well
as the flexibility reserve). It is also evident how the Nadir constraint activates
additional inertia in the system for again about 40% of the time.

The duration curves of the system PFR are then plotted in Figure 5.4. Same as
above, there is no difference between Nadir-constrained and Nadir-
unconstrained cases in the 2020 scenario; this is because the minimum value of
Nadir is above 49.2Hz also in the unconstrained scenario based on the PFR
yielded by the static FCAS requirements. However, in order to maintain the Nadir
above 49.2Hz, in 2050 the maximum level of PFR needs to increase from 550
MW to 850 MW due to the lower system kinetic energy and a substantial
increase can be generally noticed.

Figure 5.2: Frequency Nadir duration curve

51
Figure 5.3: System kinetic energy (“inertia”) duration curve

Figure 5.4. PFR requirement duration curve

52
In order to demonstrate further the relation between system kinetic energy and
PFR requirements, their half-hourly time series profiles in two consecutive days
are shown in the top picture of Figure 5.5, while the corresponding renewable
generation profile and system net demand are depicted in the bottom picture. It
can be seen how the PFR requirement and inertia are highly correlated, with PFR
higher when the system has a lower inertia and vice versa. Furthermore, lower
inertia is coincident with periods of high renewable generation and lower net
demand.

Figure 5.5: Example of PFR requirement and system kinetic energy profiles (top) and
corresponding system net demand and renewable generation profiles (bottom), for a
typical day in the “Nadir-constrained” scenario in 2050

53
System parameters sensitivity studies and cost implications

In this section we analyse the cost implication of different operational settings as


sensitivity studies (for load damping factor and Nadir limit), whose results for
2050 are synthetically depicted in the “tornado plot” of Figure 5.6. The reference
case for the cost assessment is taken as the Nadir-unconstrained and 2%/Hz
load damping.

First of all, it can be appreciated how the frequency Nadir constraints, that are
active some 40% of the time, only add some 2.8% of costs, while adding further
reserve constraints for flexibility brings about some additional 1% of costs,
relative to the base case when these constraints do not apply.

If the system damping rate decreases from 2%/Hz to 1%/Hz, the operating cost
increases by some 3% even without Nadir constraint, as more frequency
response capability needs to be procured if less load relief can support the
system frequency after the contingency. The increase in cost reaches up to 8.5%
when the flexibility constraints are added. However, the cost difference is
relatively small between Nadir-constrained and flexibility-constrained cases.
This is because the Nadir constraint with 1%/Hz damping factor requires more
generators online for inertia purposes, which can also provide the reserve
required in flexibility constrained scenario. When the load damping increases to
3%/Hz, the operational costs of the three scenarios are reduced by 2.2% (Nadir-
unconstrained case), 4.6% (Nadir-constrained case) and 1% (flexibility-
constrained case), respectively, relative to their corresponding scenarios in the
reference case.

For the sensitivity studies on Nadir allowed limit, it can be noticed that
increasing the frequency Nadir limit to 49.5Hz leads to a substantial additional
11.5% system operational cost for both Nadir-constrained and flexibility-
constrained case, as more synchronous generators need to be committed online
for inertia purposes and more PFR needs to be procured, at the cost of VRE
curtailment with zero marginal cost. However, if the Nadir allowance is lowered
to 49 Hz, there is no extra operational cost, as the lowest Nadir in the
unconstrained scenario is 49 Hz, as from Figure 5.2. Consequently, relaxing the
Nadir allowance to 48.5 Hz brings no extra benefits relative to the cost of the
scenario with 49 Hz limit. Finally, when analysing the system operational cost
with 49.5Hz Nadir and 1%/Hz damping, the annual system operation costs can
increase by up to 17%.

These results indicate that the inclusion of the proposed additional operational
constraints allows the secure integration of VRE generation at moderate system
cost increases, with the same overall mix of generating technologies. Although
the cost increases are moderate, the results also suggest that further work may
be useful to assess how these constraints affect the revenues to each generating
plant. Finally, it may be useful to re-optimise the mix of generating technologies,
so that these additional costs and revenue impacts can be included.

54
Figure 5.6. “Tornado plot” of system annual operational cost changes (relative to the base
“Nadir-unconstrained” case in 2050)

5.3.3. Assessment of the system non-synchronous penetration (SNSP)

Further to the above FRSC-UC studies, in this Section we also perform an analysis
of the SNSP level for the CET&LL scenario, with the aim of getting further
insights in this index.

Figure 5.7 shows the SNSP% duration curve for the years 2020 and 2050 for the
base case (damping factor 2%/Hz and Nadir limit 49.2Hz) and for the different
unconstrained and constrained cases. From the results, it can be appreciated
how the Nadir and flexibility constraints are basically never active in 2020, and
the maximum SNSP is in order of 60%, causing no issue.

For 2050, first of all the penetration level increases significantly due to the rise of
VRE in the scenario, and it can be appreciated that the maximum SNSP is
significant, in the order of 90%. This is in line with our previous analysis in
Section 5.2, indicating that a fairly high SNSP could potentially be obtained
depending on the system conditions, and that a fixed % limit to be applied at any
time might not be straightforward to impose (although it would be a better
option than a fixed, static value in MW).

There is no significant difference between the cases, even though it is noticeable


that for some 35% of the time Nadir and flexibility constraints bring down the
SNSP by some 5% to allow secure operation, which is obtained by operating
more synchronous generation at the cost of curtailing some VRE output. With the
more stringent flexibility constraints, in particular, the SNSP could still reach up
to some 87%.

55
Figure 5.7: System non-synchronous penetration level duration curve for CET&LL scenario
(2%/Hz damping rate, 49.2Hz Nadir)

56
6. Role of new and alternative technologies and services

6.1. More than synchronous generators

While inertia and PFR from synchronous generators are key to deliver an
adequate frequency response, there are other ways to influence the system
response, particularly in the timeframe of the inertial response. This can be
basically done by either adding external sources of inertia that can contribute to
the synchronous kinetic energy (for example, synchronous condensers that do
not produce active power as generators but still provide natural, automatic
inertial response following a contingency; or pumped-hydro plants operating in
pumping mode) or through very fast injection of energy into the system in the
inertial response timescale of a few seconds, that is, before governor-driven PFR
from conventional units can respond.

The latter options are associated with the so-called Fast Frequency Response
(FFR), which in some cases is also named synthetic inertia or virtual inertia as
it corresponds to very fast energy injection (usually through power electronics)
as kinetic energy would naturally do.

A high level techno-economic assessment of some of these options for selected


cases is provided below. It is important to note that this Section only considers
the change in operating costs, and does not consider the construction and other
establishment costs of these alternatives technologies. When assessing these
new technologies, the overall impact on system costs and benefits should be fully
considered.

6.2. Synchronous condensers

While various synchronous generators are expected to retire, some of these


could be utilised as synchronous condensers to maintain certain levels of inertia
in the system and at the same time potentially provide other services such as
reactive power and voltage support and short circuit capacity.

In this Section, we have thus performed an assessment of the potential benefits


from adding synchronous condensers equivalent to 5 GWs of kinetic energy
considering the CET&LL scenario in 2050 (with base case 2%/Hz load damping
and 49.2Hz Nadir limit), and using the FRSC-UC model introduced in Section 5.3.
Based on the simulation results, the annual operational cost reduction is in the
order 0.5% in the Nadir-constrained scenario relative to the case without
synchronous condensers, resulting from higher inertia and lower PFR
requirements.

This can be appreciated further from the duration curves of system kinetic
energy and PFR requirement depicted in Figure 6.1 and Figure 6.2, respectively.
It can be noticed that the system inertia increases overall, as expected, but with
57
the difference in inertia between the two cases becoming smaller in the low
kinetic energy range. This suggests that fewer synchronous generating units are
committed when inertia is lower, which should bring system benefits. This is
confirmed by the duration plot of the PFR requirement, which can be reduced for
some 35% of the time (when PFR requirements are driven by the Nadir
constraint rather than the static requirements, so that having higher inertia
helps the system response overall) and by up to some 70 MW.

Figure 6.1: System kinetic energy in base and synchronous condenser cases (49.2Hz nadir-
constrained case, damping factor 2%/Hz, CET&LL 2050 scenario)

Figure 6.2: PFR requirement in base and synchronous condenser cases (49.2Hz nadir-
constrained case, damping factor 2%/Hz, CET&LL 2050 scenario)

58
These results thus suggest that synchronous condensers may have larger
benefits during the lower inertia periods (some 35% of the time, in the studies
carried out here), which could provide indications as to whether they might be
run (if for frequency support purposes rather than reactive power support or
short circuit capacity contribution).

It is also likely that the benefits in the future might actually be higher, if
assuming an increase in load damping factor which challenges frequency
response more. In this respect, simulations with 1%/Hz load damping rate
indicate that the annual operating cost savings for 2050 could be about double
than in the 2%/Hz base case.

6.3. Fast Frequency Response

6.3.1. Fast Frequency Response (FFR) and Synthetic/Virtual Inertia

There are different potential sources of FFR, among which it is possible to cite:

- Storage devices, such as batteries, supercapacitors and flywheels: these have


response timescales in the region of a few tens of milliseconds.
- PV: this may require pre-curtailing of energy output to create headroom for
FFR that is then injected in the system through inverter control, and with
response timescales in the region of several tens of milliseconds.
- Wind plant connected to the grid through DFIG or full converters: the kinetic
energy physically embedded in the turbine blades may in this case be
extracted through the power electronic converter control system, which
would not require any energy output pre-curtailment as it would come, as
for synchronous generators, at the cost of slowing down the machine;
however, consideration for energy recovery that is needed to re-establish
the turbine speed is also required. The response time is in the order of
several tens of milliseconds.
- Demand response through fast load disconnection: this can typically be
obtained with response times in the regions of several tens to few hundreds
of milliseconds by disconnecting different types of loads, especially if
equipped with intrinsic storage (for example, air conditioning systems that
can exploit the thermal inertia of building fabric) or with very fast control
(for example, electrolyzers for production of hydrogen).
- Static Compensators (STATCOM) and High Voltage DC (HVDC) links that have
the possibility to provide very fast response within a few tens of
milliseconds.

It is worth pointing out that in most cases FFR might only have to be required for
a few to several seconds and up to a few minutes, so that the “energy” impact
might not be substantial (especially in the case of load disconnection).

6.3.2. Some numerical results

In this Section, we present an example study based on the CET&LL 2050 scenario
(base case with Nadir limit of 49.2Hz and damping factor of 2%/Hz) to assess the

59
benefits and performance of FFR for the system. The study was carried out with
the FRSC-UC tool and performed with half-hourly resolution for one year.

More specifically, three different technologies are simulated to provide two


different levels of FFR (100MW and 200MW) in order to limit the frequency
Nadir and reduce the PFR provision requirement of conventional generators.
These technologies are battery storage and flywheels, with response times
(including detection out of the normal frequency deadband, signalling and full
activation) assumed in the order of 150ms, as well as demand response
(modelled as fast load disconnection) with response time in the order of 500ms.

The relative annual operational cost comparison is shown in the “tornado plot”
of Figure 6.3 for the same cases simulated in Section 5.3.

In the Nadir-constrained case, 100MW FFR can lead to a range of 1.5% to 2%


cost reduction, while 200MW FFR can substantially reduce the system
operational cost by 5.5% relative to the base case. However, the cost reduction of
FFR application is much less in the flexibility-constrained cases, which is only
1.1% even with 200MW FFR. This is because flexibility constraints enforce more
generators online in any case to provide spinning reserve, which brings along
more inertia and ability to provide PFR too.

Interestingly, it can also be noted that faster FFR does not bring substantial cost
reduction, as all three technologies show a similar cost reduction range for a
same FFR level. However, of course from a security perspective it may be
desirable to have in place faster technologies, for example if FFR could be used to
help in situation of islanding risk of some NEM region. Also, further initial studies
suggest that there might be higher value in adopting faster technologies if the
load damping factor were to decrease, as the frequency response requirements
would become more challenging.

The duration curves of system kinetic energy and PFR requirement for
conventional generators are also shown in Figure 6.4 and Figure 6.5,
respectively. It can be noticed that the minimum value of system kinetic energy
is reduced with extra FFR provision. More specifically, the minimum secure level
of kinetic energy is now as low as 17.5 GW.s in the case with 200MW FFR from
battery: this minimum inertia level is actually triggered by an activation of the
ROCOF limit, otherwise even lower inertia might in principle have been reached.
The maximum PFR requirement for conventional generators also substantially
decreases from 820 MW to 520 MW with 200MW FFR provision from battery.

However, it is important to highlight that, especially for the lower levels of


inertia considered, the results provided here are only indicative of the potential
system benefits and full dynamic simulations should be carried out to assess the
frequency response performance of the system when multiple technologies and
services interacts, including detailed control schemes for the activation of FFR
and so forth. As noted above, it is also important to consider the construction
and other costs of these alternative technologies when evaluating their overall
techno-economic performance.

60
Figure 6.3: Relative cost comparison of various FFR applications (49.2Hz nadir-
constrained case, damping factor 2%/Hz, CET&LL 2050 scenario)

Figure 6.4: System kinetic energy duration curves in base and FFR cases (49.2Hz nadir-
constrained case, damping factor 2%/Hz, CET&LL 2050 scenario)

61
Figure 6.5: PFR provision requirement duration curve for conventional generators in base
and FFR cases (49.2Hz nadir-constrained case, damping factor 2%/Hz, CET&LL 2050
scenario)

6.4. Dynamic contingency rescheduling and regional security analysis

6.4.1. Dealing with transmission contingencies

A tool such as the FRSC-OPF used for the NEM-level studies can be extended to
deal with transmission and interconnector contingencies that can put a region at
risk of islanding, as well as to islanded region themselves. Such contingencies
might be extremely severe, as for example experienced in the September 2016
South Australia blackout, due to the intrinsically low level of inertia relative to
the contingency size.

Under some specific operating conditions, the loss of an interconnector should


therefore be included in the security constrained dispatch of the system. An
example of this operation condition could be the maintenance of one circuit of a
double-circuit interconnector and the subsequent loss of the available circuit.
Loss of a full interconnector could also happen in specific cases, whereby a
normally non-credible contingency such as the loss of both Heywood circuits
while operating could be reclassified as credible, In such a situation, the system
could then be made more resilient to extreme events.

To deal with transmission contingencies, a further operational measure was then


incorporated in the FRSC-OPF tool, namely, dynamic “contingency size
rescheduling”, which helps alleviating the potential security issues.

6.4.2. Example of transmission contingency analysis for South Australia

To exemplify the effect of a transmission contingency, let us focus on the loss of


the Heywood AC interconnector between VIC and SA in 2030, CET scenario.
62
Assuming that only one circuit is operating, its original secure maximum transfer
capacity is halved from 650 MW to 325 MW. The scenario under analysis
considers high demand and medium level of VRE (40% Wind, 10% Solar), so that
there is transfer from VIC to SA. The frequency parameters used for the analysis
are a demand damping of 2%/Hz and a minimum allowed frequency nadir of
49Hz34.

Base FRSC-OPF with NEM-level constraints

Under these conditions, the transfers in the system, not considering the
occurrence of the potential transmission contingency, as yielded by our FRSC-
OPF algorithm are presented in Figure 6.6. As in the main studies carried out
above, the system’s FCAS have been scheduled to withstand the potential
700MW loss of Kogan Creek in QLD. The resulting transfer through the available
circuit of the Heywood Interconnector is 313.4 MW. However, under these
dispatch conditions, if the Heywood circuit were lost, the FCAS reserves available
in SA would not be sufficient to prevent the frequency from dropping below
49Hz and so UFLS or another measure would be needed.

1669.64 MW 8405.05 MW 12442.3 MW 9115.91 MW


G G G G
125.0 MW
678.7 MW 107.7 MW

2387.22 MW
313.4 MW
G

2108.04 MW 7917.96 MW 13013.2 MW 9223.65 MW

1757.22 MW
Figure 6.6: Pre-contingency system operation status considering the Heywood
Interconnector has only one available circuit (max. transfer capacity equal to 50%). For
each region, the demand level is presented in brown, generation level in blue and
transmission flows between adjacent states in green (with the Heywood ones in red)

34A relaxed frequency Nadir of 49Hz is assumed in this case, which might potentially activate
some UFLS, provided that no resetting is carried out in the future.

63
Dynamic “contingency size rescheduling”

A further iteration of the FRSC-OPF can be carried out to reschedule the FCAS
reserves and inertia level in SA as well as the interconnector flow 35, so that SA is
able to withstand the potential sudden loss of the imports. Practically, this
corresponds to a co-optimization of FCAS, inertia and contingency size.

The new FRSC-OPF is solved through an optimization-by-simulation bi-section


search with new constraints so that SA can securely cope with a contingency size
equivalent to the variable import until the system can find a feasible point of
operation. The solution thus reschedules the system (see Figure 6.7) so that the
transfer of cheaper resources from VIC to SA is decreased to 137.4 MW (thereby
reducing the contingency size) and synchronous output and FCAS reserves as
well as inertia within SA are increased to comply with the desired Nadir limit.

1845.64 MW 8405.05 MW 12266.25 MW 9115.91 MW


G G G G
125.0 MW
854.7 MW 107.7
MW
2387.22 MW
137.4 MW
G

2108.04 MW 7917.96 MW 13013.2 MW 9223.65 MW

1757.22 MW
Figure 6.7: Pre-contingency system operation status considering the Heywood
Interconnector has only one available circuit (max. transfer capacity equal to 50%) with a
maximum transfer that would allow SA to withstand its sudden loss.

The specific results for SA in each operation condition are presented in Table 6.1.
As expected, the synchronous generation output increases in SA. This both
reduces the amount of transfers from VIC to SA (contingency reduction) and also
increases the inertia in the area (more synchronous units committed within the
region). The primary frequency response in SA also becomes higher.

35No reschedule is assumed for the Murraylink DC link, which under the given conditions is
operated at maximum capacity and doesn’t transfer inertia or FCAS from the rest of the system.

64
Table 6.1: Results for the system dispatch without and with consideration of Heywood
interconnector contingency.
SA SA VRE SA SA SA SA SA
Sync Output Import n. of n. of PFR SFR
Ouput (MW) (MW) online online (MW) (MW)
(MW) CCGT OCGT
units Units
Base case 472 1197.6 438.4 2/2 2/30 97.7 288.5

With 648 1197.6 262.4 2/2 4/30 130.7 278.8


transmission
contingency
consideration

The total amount of primary frequency response (PFR) and secondary frequency
response (SFR) at the NEM level does not change when considering the
transmission contingency, and is equal to 359.2MW and 597.3MW, respectively.
However, SA is now allocated more PFR to withstand the transmission
contingency and the primary reserves in the rest of the NEM (primarily in NSW)
decrease accordingly to compensate the reallocation in SA. Since it is assumed
that the frequency response due to largest generation contingency will be
supported by all states, including SA, the system is thus prepared to withstand
any generation contingency as well as the loss of the remaining circuit of the
interconnector between SA and VIC.

It has also to be pointed out that the rest of the NEM after SA gets islanded would
experience a high-frequency event due to loss of a large equivalent load.
Therefore, adequate “Lower” FCAS services also need to be scheduled, as in this
case.

Based on the above analysis, it can be concluded that there is significant


potential for operational and possibly market-based measures such as co-
optimization of inertia, FCAS and contingency size to improve system security. In
particular, this should apply at a local level, whereby suitable requirements are
scheduled in all of the potential regions at risk as well as for the whole NEM. This
is because there may be particular challenges in dealing with security in smaller
regions with lower inertia, which would require adequate provision of local
FCAS too. Besides covering credible contingency cases, there is also an
opportunity to use these options to deal with noncredible events, such as those
driven by extreme weather, to make the system not only more reliable, but also
more resilient.

6.5. Advanced protection schemes

Following up on the previous Section, it should be noted that the loss of an


interconnector may be a particularly severe contingency event for an area that
may experience very low inertia, with potentially very high ROCOF that could
also cause Nadir issues, if not already breaching ROCOF limit requirements. In
this respect, besides operational and technological measures that may be put in
place, such as dynamic contingency rescheduling or FFR technologies, advanced
protection schemes that can react very fast may also play a key role in
65
supporting provision of security. In particular, dedicated System Integrity
Protection Schemes (SIPS) could be adopted that perform an “inter-trip”
between the contingency event (e.g., loss of the interconnector) and the
technology response (e.g., FFR through disconnection of demand-responsive
loads or grid-connected batteries). The viability and design of such schemes
need to be assessed case by case and alongside all the other alternative and
complementary options.

6.6. New inertial response services

6.6.1. Valuing “differentiated inertia” levels

As widely discussed throughout this report, since inertia has always been
plentiful in power systems there has not historically been any specific associated
requirement. However, with the possibility of facing kinetic energy scarcity in
future low-carbon systems, and with the growing and increasingly more complex
interplay between inertial response, various FCAS, FFR, balancing services, etc.,
cases and discussions are being put forward for having the likes of “inertia
markets”, formal “inertial response services”, and so forth.

While a comprehensive analysis of this issue is outside the scope of this work,
some studies have been performed to demonstrate potential benefits and
opportunities from formally “valuing” inertia, apart from the specific market-
based or other type of mechanism to do so.

More specifically, the following modelling has been carried out. In the studies
performed, also given the lack of information about such parameter, especially
for future power plants, the inertia constant of all synchronous generators has
been considered equal to an average value of 5MWs/MVA36. However, different
generators may have different inertia constants, typically in the range of
2MWs/MVA to 10MWs/MVA. It is the inertia constant of the online units the
parameter that eventually determines the system kinetic energy, which in turn
affects the PFR requirement of the system and interacts with the other services
as well. Therefore, a few study cases have been analysed to assess if allocating
more specific inertia values to different synchronous generators could change
the scheduling of the system, and then also provide indications as to the
potential value and benefits of doing so.

6.6.2. Example: annual generation dispatch and benefits

A first analysis has been carried out with the FRSC-UC tool with reference to the
“Nadir-constrained” 2050 CET&LL base case (2%/Hz damping factor, 49.2Hz
Nadir limit). Two cases have then been considered:

- “Average inertia” case, whereby the inertia constant H of all generators is set
to 5MWs/MVA, as in all studies.

36 The measuring unit of the generators’ inertia constant is also often briefly indicated in seconds.

66
- “Differentiated inertia” case, whereby the inertia constant H values for coal,
hydro and OCGT plants have been set to 4, 3 and 3 MWs/MVA, respectively.
With regards to CCGT, the inertia constant values of relatively high-cost
plants have been set to 8MWs/MVA, while for lower-cost ones the value has
been set to 6MWs/MVA37.

Studying the implications of these settings for the CCGT plants is of particular
interest to demonstrate the potential of an inertial market or service. In this
respect, the annual energy output for the two differentiated CCGT clusters is
shown in Table 6.2 for the two cases considered. The output share of higher cost
CCGT units increases from 17.6% to 18.7% in the “differentiated inertia” case,
while for the lower cost CCGT unit it reduces from 13.7% to 12.4%38.

Since the higher cost units have been (arbitrarily) assigned a higher inertia
constant value and are now being dispatched more in the “differentiated inertia”
case, the results highlight that there is an intrinsic economic value in the
generators’ inertia. This should not be surprising, since, by interacting with the
rest of frequency response services, kinetic energy also impacts the dispatch of
all generating units and the system schedule. At the system level, there are also
both economic benefits and environmental benefits39.

Table 6.2: Generation output share for two CCGT clusters in “average” and “differentiated”
inertia cases

Average Differentiated
inertia inertia

Low cost CCGT (5.9 GW, H=6s) 13.7% 12.4%

High cost CCGT (9.3 GW, H=8s) 17.6% 18.7%

37The resulting average inertia constant of the whole generating portfolio is equal to 5.06
MWs/MVA, thus very close to 5 MWs/MVA in the “average inertia” case. All the “differentiated”
inertia constant values considered are in line with information available in the literature,
indicating in particular relatively higher values for CCGT plants.

38In the modelling exercise performed, the installed capacity of higher-cost CCGT units is 9.3GW,
while for lower-cost ones it is 5.9GW, which also explains the higher initial energy output from
high-cost units.

39In the specific example provided, the system economic benefits are fairly limited but material,
in the order of 0.25%, while the environmental benefits assessed as reduction in VRE curtailment
are in the order of 2%. However, further studies suggest that the benefits might be much more
substantial in the presence of lower load damping.

67
6.7. Further considerations on frequency response services and
technologies

It also has to be pointed out that frequency response adequacy assessment in


low-inertia systems is a territory largely unexplored and, while the simplified
modelling performed here provide insights into future system operation and has
been checked with dynamic simulation tools, much more detailed analysis,
which were out of the scope of the present work, are needed to understand the
complex interactions among: low-inertia frequency dynamics; old and new
sources of inertial, fast, primary and other frequency responses; voltage stability
and reactive power response, especially in the presence of renewables; need for
further security requirements such as to guarantee minimum short-circuit
capacity levels in the system; rotor angle stability issues associated to
geographical and temporal inertia changes across the NEM; and so on.

In particular, on top of the frequency response requirements discussed in this


work, there might be other reasons why synchronous resources would be
deployed (e.g., for reactive power support or for short-circuit power provision),
and the interaction between all technologies and multiple services should be
analysed in detail in order to provide system security at minimum cost.

68
7. Appendix: the National Electricity Market

7.1. General aspects of the NEM

The National Electricity Market (NEM)40 is the wholesale electricity market that
supplies New South Wales, Queensland, Victoria, South Australia, Tasmania, and
the Australian Capital Territory (ACT). The Australian Energy Market Operator
(AEMO) is the Independent System Operator (ISO) of the NEM. The NEM
currently supplies around 200TWh of electricity annually, or roughly 80% of
Australian electricity consumption to more than 9 million customers. Over 300
generators are registered in the NEM, with around 47 GW of electrical generating
capacity as of December 2016.

The NEM has markets for electrical energy and ancillary services41. Contingency
Frequency Control Ancillary Service (FCAS) and Regulation FCAS are provided
by the market, and Network Support and Control Ancillary Services (NSCAS) and
System Restart Ancillary Service (SRAS) are provided under contracts. The NEM
does not have a capacity market. Generators and retailers can also manage
market risk using financial instruments such as over the counter (OTC) contracts
and futures markets, and these can provide additional revenue to generators.

7.2. NEM operation

The NEM operates as a gross pool with five-minute dispatch intervals. Every five
minutes, offers from generators are matched against real time demand to clear
the market using a dispatch engine that co-optimises energy and FCAS, and
relevant dispatch instructions are then issued to the market actors. This is
carried out by a linear optimisation engine that operates close to real time.
Differently from many other markets worldwide, no time ahead formal decisions
and commitment are required. Instead, generators make their own commitment
decisions and structure their offers accordingly. Prices for energy and for
ancillary services in each NEM region are calculated using the outcomes from
this dispatch optimisation. Offers can be submitted more than a day ahead of
time, and every half hour a pre-dispatch process calculates and publishes
demand, dispatch, and price forecasts for each half hour of the following day 42.
AEMO also publish market notices, e.g., lack-of-reserve forecasts. Unlike other

40AER, "State of the energy market 2015", http://www.aer.gov.au/publications/state-of-the-


energy-market-reports

41AEMO, “Guide to the Ancillary Services Market”, https://www.aemo.com.au/-


/media/Files/PDF/Guide-to-Ancillary-Services-in-the-National-Electricity-Market.ashx

42This is the closest mechanism that there is in the NEM to some form of time-ahead unit
commitment that is performed in other markets worldwide (for example by the system operator
to formally check on system security); however, as a key difference, in the NEM the pre-dispatch
bids are not binding.

69
markets and differently to what happens in most markets worldwide, this
process is informational only, and there are no payments to or obligations on
generators arising from the pre-dispatch process. Participants can change their
offers up to the time of dispatch. AEMO nonetheless does have the ability to
direct generators to operate under certain circumstances.

70
8. Appendix: Emissions reduction modelling scenarios
description and input data

8.1. Scenario description

8.1.1. Business as Usual Scenario (BAU)

In the BAU scenario, when coal plants reach 35 years of age, a modest
refurbishment is applied if the costs can be recovered within a 5-year payback
period, otherwise the availability and the efficiency of the plant will decrease by
1 percentage point each year. All the coal plants are assumed to stop operating
once they reach 60 years, or if they incur net trading losses.

8.1.2. Limited Lifetime (LL)

In the LL scenario, coal generating units close after 50 years, while major
refurbishments are assumed to occur when they are needed. No emissions
reduction target is applied for this scenario.

8.1.3. Clean Energy Target (CET)

In the CET scenario, coal generating units are not applied explicit closure rules;
instead, the longevity will depend on their profitability.

The emissions-intensity threshold is set to be 0.6t CO2-e/MWh. New generators


below the emissions-intensity threshold can earn certificates, while existing
generators that do not receive Renewable Energy Targets (RET) certificates can
earn certificates for generation above their historical baseline level. The CET
trajectory grows from zero in 2020 to 2050, in order to achieve the emission
reduction target.

The emissions reduction target of CET is set to reduce 28% of the emission of
2005’s level by 2030.

8.1.4. Clean Energy Target and Limited Lifetime (CET&LL)

In the CET&LL scenario, coal generating units’ closure rules operate as indicated
in the LL scenario, while a low emission target scheme operates as indicated in
the CET scenario. The emission reduction target of CET&LL is set to reduce 28%
of the emissions of 2005’s level by 2030.

8.1.5. Emissions Intensity Scheme (EIS)

In the EIS scenario, coal generating units are not applied explicit closure rules;
instead, their longevity will depend on its profitability.

The emissions intensity baseline is set for the electricity sector as a whole. This
baseline is consistent with the emissions reduction target, which starts at grid

71
average emissions intensity in 2020 and linearly declines to meet the emission
reduction target in 2030. The gradient will then be maintained through to 2050.

The emissions reduction target of EIS is set to reduce 28% of the emissions of
2005’s level by 2030.

8.2. Input data

8.2.1. Electricity demand

A 50% probability-of-exceedance (POE) median peak demand is used for all


scenarios. The minimum level and maximum level of electricity demand (which
is the same for all scenarios) throughout the years up to 2050 is shown in Figure
8.1, while Figure 8.2 shows an example of load profiles that have been used.

Figure 8.1: Range of electricity demand (min to max) for all scenarios to 2050

Figure 8.2: Hourly demand profile for two representative weeks in year 2040, including
minimum and maximum demand periods

72
8.2.2. Installed capacity by scenario

Figure 8.3 to Figure 8.7 show the evolution of installed capacity of different types
of resources for the different scenarios.

Figure 8.3: Installed capacity (business as usual – BAU)

Figure 8.4: Installed capacity (limited lifetime – LL)

73
Figure 8.5: Installed capacity (clean energy target and limited lifetime – CET&LL)

Figure 8.6: Installed capacity (clean energy target – CET)

74
Figure 8.7: Installed capacity (emissions intensity scheme – EIS)

8.2.3. Example of VRE generation profiles

Figure 8.8 and Figure 8.9 show examples of hourly generation profiles in two
different weeks for solar energy and wind energy, respectively, for different
regions of the NEM in two different years.

Figure 8.8: Hourly solar available output for two representative weeks in year 2020 and
2040

75
Figure 8.9: Hourly wind available output for two representative weeks in year 2020 and
2040

76
9. Appendix: Main definitions, modelling assumptions and
input data for system security studies

9.1. Security: general definition and definition for the purpose of this
work

Power system security is generally concerned with the strength of the electrical
power system to tolerate and respond to system disturbances without suffering
equipment damage while maintaining quality of electricity supply to
customers43.

Power system security is achieved by operating the power system in a


satisfactory and stable operating state, by maintaining the system frequency,
voltages at various points, power flows in all transmission lines, and all power
stations’ outputs within predefined limits, as well as ensuring that the system
operational points are such as to activate circuit breakers and disconnect the
faulty equipment.

Security assessment requires power systems engineering analysis tools (e.g.,


load flow and stability modelling), which help in defining a secure system
envelope within which the market can be dispatched and customer demand can
be met.

These types of studies, fundamentally operational in their nature, have been


performed within this work through state of the art and bespoke modelling tools
able to simulate and optimize the economic operation of the system subject to
predefined limits (“constraints”).

In particular, for the purposes of this work, focus has been put on frequency
response adequacy aspects, which may become a particularly severe issue with
larger penetration of non-synchronous renewable energy sources and therefore
in lower inertia systems, as already experience in the recent Black System event
in South Australia in September 2016. More specifically, we have addressed this
general security issue by assessing whether, in each modelling scenario under
analysis, there were sufficient resources and facilities (to be operated with
appropriate operational requirements and strategies) to guarantee adequate
system frequency response following a disturbance (or “contingency”).

Considerations on the other aspects of power systems security listed above have
also been provided both quantitatively and/or qualitatively.

43Further details about power systems security and related matters are provided in Chapter 4 of
the National Electricity Rules:
http://www.aemc.gov.au/getattachment/d1308438-44e2-42d0-8582-1ead65c93f3d/National-
Electricity-Rule-Version-3.aspx

77
9.2. Definition of credible and noncredible contingencies in the NEM

Contingencies are classified in the NEM as credible and non-credible under the
NER44.

According to the current practices, the power system is operated in such a way
that it will remain in a satisfactory operating state for the loss of single elements
in the transmission network. These events are defined as credible contingency
events and include:

 Unexpected loss of a single transmission line, transformer or reactive plant;


 Unexpected loss of a single generating unit.
AEMO considers the occurrence of these events to be reasonably possible and
ensures contingency plans are in place to minimise impact on the power system
following a credible contingency event.

A non-credible contingency is a contingency event other than a credible


contingency. Examples include:

 Three phase electrical faults;


 The trip of any busbar in the transmission network;
 The trip of more than one transmission element;
 The trip of transmission plant in a manner not considered likely (e.g., a
transmission line that trips at one end only);
 The trip of multiple generating units;
 The trip of more than one load block in a region where the combined load
lost exceeds what would normally be considered a credible contingency
event in that region.
AEMO does not normally operate the power system such that it will remain in a
satisfactory operating state following a non-credible contingency event, as the
likelihood of this occurrence is low. This is also aligned with the practices of
most system operators worldwide.

AEMO may reclassify a non-credible contingency event as a credible contingency


event if the risk of this event impacting the power system is increased due to
abnormal conditions 45 . Abnormal conditions may include severe weather
conditions, lightning and bushfires.

44 National Electricity Rules, http://www.aemc.gov.au/energy-rules/national-electricity-


rules/current-rules

45 AEMO, “Review of Power System Reclassification Events: 1 November 2015 to 30 April 2016
for the NEM”, September 2016.
https://www.aemo.com.au//media/Files/Electricity/NEM/Market_Notices_and_Events/Reclassi
fication-events/Reclassification-report-Nov-2015-to-April-2016.pdf

78
9.3. Main assumptions for FCAS and frequency response modelling

A number of consistent assumptions for FCAS modelling, including on numerical


values of key study parameters, have been applied to all cases.

First of all, without loss of generality for the purposes of assessing the high level
security characteristics of each scenario the focus has been put on the “Raise”
services (corresponding to the responds to a sudden substantial frequency drop,
typically due to a generation or network loss contingency)46.

Specific generation headroom and ramp constraints for the different FCAS have
been considered. The synchronous generators providing the fast response
service (6 seconds), in particular, require the definition of transient ramp
capability, modelled as response slopes vs headroom capacity 47. Relevant
information was gathered through the analysis of bidding data in the FCAS
market by the different generators in NEM. Those results were also cross-
checked with international references. This exercise yielded high levels of
correlation between international parameters and average parameters for NEM
units. The 60 seconds and 5 minutes reserves have then been modelled by using
the generators’ nominal ramp rates, thus assuming that the relevant spinning
headroom could be fully deployed in these time frames.

For the purposes of analysing the system’s response to event that could cause
extreme frequency instability, the maximum credible contingency in the system
(as from the information provided from the scenario modelling outputs) has
been taken as a synchronous generation loss of 700 MW48. It has also been
assumed that loss of multiple units connected to the main system is not a
credible contingency.

Further to that, and especially considering that for extreme situations in future
scenarios with more and more renewables smaller units might be scheduled, we
have also considered a dynamic assessment of the largest contingency being co-
optimised with FCAS and synchronous inertia (see also Section 6.4 for specific
applications).

46Similar conditions would be applied to the “Lower” service, which is generally less demanding
in terms of response as in most cases there is some option to decrease generation output in a
sufficiently fast way (including curtailment of renewable energy output) in the case of a
frequency rise.

47As in general the frequency control requirements are a function of both the level of inertia and
the ramping capability of different power plants, this ramping capability needs to be modelled in
detail, especially in terms of transient ramps that generators can deliver within the 6s service. We
have modelled these as close as possible in terms of actual characteristics and constraints as
from data available for the current generators from the NEM.

48The unit corresponds to “Kogan Creek 1”, in Queensland South. In this worst case analysis it is
thus assumed that Kogan Creek 1 is always operated, for the sake of simplicity, which happens in
most cases considering its relatively low marginal cost.

79
The response of the generators providing reserves is considered to take place
with the delay associated to a frequency response dead-band of 0.150Hz as from
current standards, after which ramping of conventional generators take place.

A frequency damping factor equal to 3%/Hz is currently considered by AEMO for


a loss of generation in the NEM (1.5% per 1% frequency change)49. However,
while the 3% case corresponds to current practices, a range of values might
apply in the future, so that sensitivities considering variations of between 1%/Hz
and 3%/Hz have been analysed. In fact, the 1% (or potentially even less) case
might be more representative of future scenarios with smaller volume of
frequency-dependent load in the system (e.g., due to larger volumes of
electronically interfaced devices, electric vehicles, inverter-driven air
conditioning systems, etc.). This is also is in line with the figures used elsewhere,
for example in the UK. As discussed during the project, an intermediate damping
value of 2%/Hz has been considered as realistic base case compromise for
analysis of future scenarios.

The inertia constant for all generators has been assumed equal to
H=5MWs/MVA.

9.4. Main operational limits adopted in the base case security assessment
studies

Since in the FRSC-OPF algorithm the response requirements are dynamically


calculated subject to the relevant constraints and the incumbent conditions of
VRE output and demand level, the resulting generating units’ dispatch will
depend on the specific operational limits that imposed for ROCOF, frequency
Nadir and QSSF, as well as other parameters such as the load damping factor.

From discussions within the project, it was agreed that the following numerical
values would be taken as base case:

- Frequency Nadir limit = 49.2Hz


- ROCOF limit = -1 Hz/s
- QSSF60s = 49.5Hz
- QSSF5min = 49.85Hz
- Load damping factor = 2%/Hz

However, sensitivities have also been performed for various parameters, in


particular, for the damping factor (given its uncertain value in the future)
between 1%/Hz and 3%/Hz, and the Nadir (between 48.5Hz, corresponding to a
more relaxed frequency control in a low-inertia system, and 49.5Hz,
corresponding to tighter requirements as today).

A maximum allowed ROCOF limit of 1Hz/s was deemed appropriate as the main
case in that, as discussed, it could prevent large changes of frequency that could

49 AEMO, Frequency Control Ancillary Services, March 2012

80
trip off embedded generation (for example whose islanding schemes were to be
set through ROCOF measurements) or potentially even conventional generators
that could undergo excessive physical stresses. This is line with developments in
other countries such as UK and Ireland. Sensitivity studies have also been
performed to check the operational implications of different ROCOF limits when
lowered to 0.5Hz/s.

9.5. Definition of the System Non-Synchronous Penetration (SNSP) level


index

For the mainland NEM region as a whole, the SNSP, expressed in % of the
incumbent demand, can be simply defined as:

𝑁𝑜𝑛𝑠𝑦𝑛𝑐ℎ𝑜𝑛𝑜𝑢𝑠 𝑜𝑢𝑡𝑝𝑢𝑡𝑡
𝑆𝑁𝑆𝑃𝑡 % =
𝐷𝑒𝑚𝑎𝑛𝑑𝑡
𝑁𝑜𝑛𝑠𝑦𝑛𝑐ℎ𝑜𝑛𝑜𝑢𝑠 𝑜𝑢𝑡𝑝𝑢𝑡𝑡
=
𝑁𝑜𝑛𝑠𝑦𝑛𝑐ℎ𝑜𝑛𝑜𝑢𝑠 𝑜𝑢𝑡𝑝𝑢𝑡𝑡 + 𝑆𝑦𝑛𝑐ℎ𝑜𝑛𝑜𝑢𝑠 𝑜𝑢𝑡𝑝𝑢𝑡𝑡

where the subscript t indicates that the index needs to be calculated at every
relevant time step t.

9.6. Specific input data for FRSC-UC studies for dynamic assessment of
minimum synchronous generation levels

The maximum credible generation loss is set to 700 MW.

The dead-band of frequency response activation is set to 150 mHz, as for the
FRSC-OPF tool used in the main studies, whilst also a response delay of 0.5s has
been added50.

The default settings of load damping and frequency Nadir allowance are 2%/Hz
and 49.2 Hz, respectively.

The maximum allowed ROCOF is set to 1 Hz/s.

The static FCAS reserve constraints are calculated at 49.5 Hz at both 6 and 60
seconds.

The 5 minutes FCAS is assessed for 50±0.15Hz.

The one-hour flexibility reserve is calculated based on uncertainty and


variability of wind and solar generation extrapolating current data.

50This is an average value that has been assessed by dynamic simulations of different types of
power plants according to standard dynamic models. It should be noted that in the FRSC-OPF
studies previously performed this delay has not been modelled, and a comparison between
relevant results could bring important insights.

81
All synchronous generators are assumed to be characterised by an average
inertia constant of H=5MWs/MVA.

9.7. Further information on main input data and assumptions used for
the security assessment studies for all scenarios

The security assessment analysis has been primarily conducted for snapshots of
the system in various conditions, as from the emissions reduction modelling
scenario input data received, and based on various relevant assumptions that are
described below.

9.7.1. General inputs and assumptions

System

The system that has been modelled corresponds to a single busbar equivalent of
the NEM - this means that the effects of the transmission network are not being
taken into consideration, as discussed during the project51.

Timeframe

The analysis has been conducted for snapshots of the system in various
conditions. In particular, for each scenario a set of 8 years has been analysed,
namely every 5 years between 2020 and 2050, as well as year 2017.

Generation

The generation capacity available in each scenario has been divided in two
general sets, namely, synchronous capacity and non-synchronous capacity.

 Synchronous capacity: includes thermal technologies using coal and gas as


fuels, as well as hydro and pumped storage units. Following previous
modelling52, for computational speed purposes, the corresponding units of
the system are clustered (using a K-means algorithm) for each considered
year by technology and by state (mindful of potential future regional
analysis) based on the units’ capacity and variable cost. This leads to a set of
clusters with a representative capacity, incremental cost of generation and
number of units, whose aggregate capacity is equal to the input capacity by
technology. Additionally, synchronous generation clusters are represented
by a minimum stable generation level, inertia constant, maximum transient
and static ramp rates, maximum reserve headroom, and relation between

51However, considerations for interconnector contingencies and possible islanded operations of


regions such as South Australia o Tasmania have also been made on several case studies.

52L. Zhang, T. Capuder and P. Mancarella, Unified unit commitment formulation and fast multi-
service LP model for flexibility evaluation in sustainable power systems, IEEE Transactions on
Sustainable Energy, vol. 7, no. 2, pp. 658-671, April 2016.

82
headroom and ramping limit when applicable (e.g., in the case of “fast raise”
response service). Due to their low levels of penetration and for the sake of
simplicity, biomass units are assumed to provide all their inertia and 50%
output relative to installed capacity in each year.
 Non-synchronous capacity: includes Variable Renewable Electricity (VRE)
sources such as wind and solar (large and small scale) technologies. These
technologies are represented by means of various levels of available
generation output (relative to installed capacity) for each of them. These
output availability levels are calculated from the generation profiles given as
input to this study, in order to represent relevant cases in the assessment of
security.

Demand

Demand profiles have been calculated from the scenario modelling information
provided, using the relevant hourly profiles and maximum demand for each state
separately. Maximum and minimum system demand levels have then been
determined for each year. Additionally, maximum and minimum levels of
demand have been established for potential regional analysis.

Inertia provision by synchronous machines

It has been assumed that when a synchronous machine reaches its minimum
stable generation it provides an amount of kinetic energy that is conventionally
represented through its inertia constant. An average inertia constant value equal
to 5MWs/MVA has been taken as indicative for each machine providing inertia in
the system, also based on different sources.

9.7.2. Snapshots scenarios

Demand

Demand levels have been analysed for snapshots of maximum and minimum
values for each year. The exact date and time for maximum and minimum
demand might slightly change if a different method for energy rescaling from the
data provided is used. However, the general results would still hold. In this
particular case, the hourly demand trend was built considering the peak demand,
total energy consumption and demand profiles given for each state and year. To
fit the seasonal trends of the NEM, additional data from AEMO NEMWEB was
used to determine the weekly level of energy from real operation data for years
2015 and 2016.

VRE output

The main levels of available energy output (relative to the installed capacity)
from renewables that have been considered were chosen based on the scaled
profiles for each technology in each state and scenario. Further, more extreme
cases of resource availability have been considered to stress test the system. It
has also to be pointed out that the final actual dispatched VRE output might be
83
lower than the available VRE output if frequency stability constraints require
curtailment of renewable generation as a means to provide security.

The scenarios considered for VRE available output depend on the level of
demand as follows:

 Minimum Demand
 Very high wind and no solar (night): 90% output of wind, 0% output of
solar technologies, and 50% output of biomass is available.
 Medium wind and no solar (night): 50% output of wind, 0% output of
solar technologies, and 50% output of biomass is available.
 Medium-low wind and no solar (night): 40% output of wind, 0% output
of solar technologies, and 50% output of biomass is available.
 Low wind and no solar (night): 10% output of wind, 0% output of solar
technologies, and 50% output of biomass is available.
 Maximum Demand
 VRE 50%: 50% of output (relative to the installed capacity of VRE) is
available.
 VRE 80%: 80% of output is available.
 VRE 90%: 90% of output is available.
 Medium-high wind and low solar: 60% output of wind, 10% output of
solar technologies, and 50% output of biomass is available.
 Medium-high wind and medium-low solar: 60% output of wind, 20%
output of solar technologies, and 50% output of biomass is available.
 Medium-low wind and low solar: 40% output of wind, 10% output of
solar technologies, and 50% output of biomass is available.
 Very high wind and low solar: 90% output of wind, 10% output of solar
technologies, and 50% output of biomass is available.
Hydro-resources were also modelled separately and more specifically:

 Hydro output: To explore the effect of availability of hydro resources, four


scenarios for hydro output availability (relative to installed capacity) were
considered:
 Hydro 100%: Full hydro output is available.
 Hydro 50%: 50% of hydro output is available.
 Hydro energy: Hydro output availability is modelled through the
capacity factor of the technology in every year, as given by the yearly
energy yield in the scenario results.
 No hydro: an extreme case with no hydro available in the system.
 Pumped-Storage: To explore the effect of pumped-storage, two operation
modes were considered:
 Generation mode: Pumped-storage is considered with the relevant
generating constrained used above for Hydro output level.
 Load mode: Pumped-storage is in pumping mode, and the relevant
inertia is added to the system.

84

You might also like