556

Power System Technical Performance Issues

Related to the Application
of Long HVAC Cables

Working Group
C4.502

October 2013
POWER SYSTEM TECHNICAL
PERFORMANCE ISSUES RELATED
TO THE APPLICATION OF LONG
HVAC CABLES
WG C4.502

Copyright © 2013
“Ownership of a CIGRE publication, whether in paper form or on electronic support only infers right
of use for personal purposes. Are prohibited, except if explicitly agreed by CIGRE, total or partial
reproduction of the publication for use other than personal and transfer to a third party; hence
circulation on any intranet or other company network is forbidden”.

Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept
any responsibility, as to the accuracy or exhaustiveness of the information. All implied warranties
and conditions are excluded to the maximum extent permitted by law”.

ISBN : 978-2-85873-251-7
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

WORKING GROUP CONVENER:

WOJCIECH WIECHOWSKI (POLAND)

WORKING GROUP SECRETARY:

TERUO OHNO (JAPAN)

WORKING GROUP TASK FORCE LEADERS:

ZIA EMIN (UK), TOM MCDERMOTT (USA), PAUL WILSON (CANADA)

WORKING GROUP MEMBERS AND CONTRIBUTORS

Vladislav Akhmatov Denmark Akihiro Ametani Japan

Claus Leth Bak Denmark Roberto Benato Italy
Luigi Colla Italy Eric Dorison France
David Dubois France Marta Val Escudero Ireland
Flavio Fernandez Germany A.M. Golé Canada
Unnur Stella Gudmundsdottir Denmark Sebastien Henry France
Jan de Jong Netherlands Jakob Kessel Denmark
Stefano Lauria Italy Edwin Lerch Germany
Frederic Lesur France Luis Marti Canada
Francesco Palone Italy Marjan Popov Netherlands
Filipe Faria da Silva Denmark Jeewantha de Silva Canada
Lou Van der Sluis Netherlands Frans Jonas Sollerkvist Sweden
Stefan Thorburn Sweden

Page i
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

POWER SYSTEM TECHNICAL

PERFORMANCE ISSUES RELATED TO
THE APPLICATION OF LONG HVAC
CABLES
TABLE OF CONTENTS
EXECUTIVE SUMMARY ......................................................................................................................................... 1

CHAPTER 1. INTRODUCTION ............................................................................................................................ 2

CHAPTER 2. ANALYSIS OF THE TECHNICAL PERFORMANCE OF POWER SYSTEMS SPECIFIC FOR

SYSTEMS WITH LONG HVAC CABLE LINES...................................................................................................... 4
2.1 Introduction ................................................................................................................................................. 4
2.2 Studies and Tools ....................................................................................................................................... 5
2.2.1 Planning study scope .......................................................................................................................... 5
2.2.2 System impact study scope ................................................................................................................ 6
2.2.3 Equipment and system design study scope .......................................................................................6
2.2.4 System technical performance issues and study tools ....................................................................... 7
2.3 Example Test Systems ............................................................................................................................... 9
2.3.1 Many-cable example test system ....................................................................................................... 9
2.3.2 Long-cable example test system ........................................................................................................ 9
2.4 System Planning Issues ........................................................................................................................... 13
2.4.1 Reactive power compensation .......................................................................................................... 13
2.4.2 Example case study of reactive power compensation ...................................................................... 15
2.5 System Impact Issues............................................................................................................................... 23
2.5.1 Harmonics and network resonances ................................................................................................ 23
2.5.2 TOV - Resonance ............................................................................................................................. 26
2.5.3 TOV – Ground faults and clearing .................................................................................................... 31
2.5.4 TOV – System islanding ................................................................................................................... 33
2.5.5 Temporary overvoltages versus cable tests recommended by IEC ................................................. 35
2.5.6 Generator self-excitation ................................................................................................................... 35
2.5.7 Transient stability .............................................................................................................................. 36
2.6 Equipment Design Issues ......................................................................................................................... 37
2.6.1 Cable energization ............................................................................................................................ 37
2.6.2 Auto-reclosing overvoltages.............................................................................................................. 41
2.6.3 Cable discharge ................................................................................................................................ 41
2.6.4 Zero-miss effect ................................................................................................................................ 42
2.6.5 Couplings to other conductive infrastructures................................................................................... 43
2.7 Modelling of Other Equipment .................................................................................................................. 45
2.7.1 Extent of cable network to be modelled in detail .............................................................................. 46

Page ii
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

CHAPTER 3. CABLE MODELLING FOR POWER FREQUENCY STUDIES................................................... 48

3.1 Introduction ............................................................................................................................................... 48
3.2 Cable Data ................................................................................................................................................ 49
3.3 Example Model ......................................................................................................................................... 50
3.4 Practical Considerations ........................................................................................................................... 51
3.4.1 Semi-conducting layers..................................................................................................................... 51
3.4.2 Screen or sheath considerations ...................................................................................................... 51
3.4.3 Effects of earth .................................................................................................................................. 53
3.4.4 AC resistance of conductors ............................................................................................................. 53
3.4.5 Influence of temperature on skin effect............................................................................................. 54
3.4.6 Influence of bonding and transposition ............................................................................................. 55
3.4.7 Long transmission line correction ..................................................................................................... 55
3.5 Sequence Impedance Calculations according to Published Standards ................................................... 56
3.5.1 Calculation of AC resistance (R) ....................................................................................................... 57
3.5.2 Calculation of positive sequence impedance.................................................................................... 58
3.5.3 Calculation of zero sequence impedance ......................................................................................... 60
3.5.4 Calculation of capacitance (C) .......................................................................................................... 61
3.6 Verification of the Nominal Frequency Cable Model ................................................................................ 62
3.7 Construction of the Nominal Frequency Cable Model .............................................................................. 63

CHAPTER 4. CABLE MODELLING FOR TRANSIENT STUDIES ................................................................... 64

4.1 Introduction ............................................................................................................................................... 64
4.2 Available Computer Models ...................................................................................................................... 64
4.2.1 Bergeron transmission line model .................................................................................................... 65
4.2.2 Modal domain models ....................................................................................................................... 66
4.2.3 Phase domain models ...................................................................................................................... 66
4.3 Selection of Cable Models for Various Types of Studies ......................................................................... 67
4.3.1 Bergeron model ................................................................................................................................ 67
4.3.2 Recommended cable models ........................................................................................................... 69
4.3.3 Guideline for setting up phase-domain FD models .......................................................................... 70
4.4 Calculation of Electrical Parameters of a Cable ....................................................................................... 72
4.4.1 General formulation .......................................................................................................................... 72
4.4.2 How to deal with non-standard configurations .................................................................................. 72
4.5 Installation and Environmental Effects on Long Cable Models ................................................................ 72
4.5.1 Bonding of cable screens.................................................................................................................. 72
4.5.2 Installation and layout of cables ........................................................................................................ 73
4.5.3 Effects of temperature....................................................................................................................... 73
4.5.4 Effects of earth resistivity .................................................................................................................. 73
4.6 Example of Sensitivity Analysis of Input Parameters on Study Results ................................................... 73
4.7 Verification of Line Constant Solver Results against Manual Calculations .............................................. 80
4.8 Transient Cable Model Validation Procedure ........................................................................................... 81

CHAPTER 5. CONCLUSIONS AND RECOMMENDATIONS ........................................................................... 85

BIBLIOGRAPHY/REFERENCES.......................................................................................................................... 88

ANNEX 1. CABLE MODEL IMPLEMENTATIONS........................................................................................... 92

ANNEX 2. MEASUREMENT OF CABLE ELECTRICAL PARAMETERS ....................................................... 96

ANNEX 3. MEASUREMENTS ON THE 100 KM LONG 150 KV AC CABLE CONNECTION OF THE HORNS
REV 2 OFFSHORE WIND FARM ......................................................................................................................... 98

Page iii
EXECUTIVE SUMMARY
This Technical Brochure describes system technical performance study issues important for a power system with
(a large share of) long AC cable lines. The study areas are divided into planning studies, system impact studies
and equipment design studies. The brochure contains examples and explanations of system technical performance
phenomena occurring as a result of various system operations, guidelines for simulation methods, interpretation of
the results and possible suggested countermeasures. As some of the presented examples produce overvoltages of
significant duration and / or amplitude, consideration is given to the adequacy of cable tests recommended by IEC.
The focus of this Technical Brochure is system technical performance issues and modelling of cables and other
equipment is not its main topic. Nevertheless, in order to retain the ability to stand on its own, relevant
recommendations concerning modelling of cables and other pieces of equipment as well as the required extent of
network equivalents are given. Guidelines and recommendations for adequate modelling of AC cable lines for
different types of power frequency studies and for studying phenomena that involve other frequencies are
presented. Practical methods for gaining confidence in the accuracy of cable line models are described.
Calculations of power frequency impedances based on published norms are shown. Available models for transient
studies are identified and their selection for various studies is discussed including the required level of detail,
accuracy and other implementation details. Sensitivity of study results due to variation of cable input parameters is
analysed. Line constant solver results are benchmarked against manual calculations. Also a transient cable model
validation procedure is shown. Derivation of basic computer cable line models, setups for measurement of cable
impedances and results of real-life measurements performed on 100 km long AC cable of an offshore wind farm
are also presented.

Page 1
Chapter 1. INTRODUCTION

In recent years many electricity transmission network operators have been faced with the necessity of upgrading
their transmission networks. This has been due to a variety of reasons, including an increase of cross-border
trades, integration of large scare renewable energy sources, smart grid projects, the replacement of aging facilities,
or a growing demand. Traditionally such upgrades would have been accomplished through the installation of
overhead transmission lines (OHLs). Underground cable systems (UGC) on the transmission voltage level have
been used as well, but their applications have been mainly restricted to densely populated areas. UGC have so far
been widely used on the lower voltage (distribution) levels, typically as radial lines of shorter lengths. This tendency
of not using UGC on the transmission level has been slowly changing over the past ten years due to various factors.
The purpose of high voltage transmission lines is to transmit bulk of electrical energy over longer distances and this
function must be fulfiled by UGC transmission lines as well. As the usage of such long high voltage AC cables on
the transmission level has not been common, network planners, network operators and equipment suppliers do not
typically possess much experience with the technical performance of transmission systems with long AC cables nor
with modelling of cables for studies on the transmission level. The few examples where long EHV AC cables are in
service include Japan, where a length of 40 km at 500 kV is installed, Connecticut, where Northeast Utilities and
United Illuminating have installed 37 km and 15 km double-circuit 345 kV cables into a relatively weak system, or
Denmark, where a 150 kV connection to the offshore wind farm Horns Rev 2 is installed and has a length of 100
km. Denmark has unprecedented plans for underground cabling of the entire existing 132 kV and 150 kV grids as
well as most of the planned new 400kV lines. In 2030 this plan should be completed.
Fundamentally there is nothing new in terms of the technical performance of a transmission system with long cable
lines compared to one with long overhead lines. However, due to the difference in their physical properties cables
have a much larger shunt capacitance compared to overhead lines. This results in a much greater need for reactive
power compensation, observation of resonances at the lower end of the frequency spectrum as well lower
frequency components within current and voltage waveforms generated during transients. Furthermore, modelling
of cables, especially for frequencies other than the system frequency is more complicated and demanding in
comparison to modelling of OHLs.
The main aspects where a system of AC cable lines differs from a system of AC overhead lines at the same
voltage level and transmission capacity are:
1. Level of shunt compensation required is much greater, because the cable capacitance is higher
2. Lower system resonance frequency
3. Lower series reactance, because the phase conductors are closer together
4. Special concerns related to cross-bonding and cable armour
5. For transients, the surge impedance and travelling wave velocity are both lower
6. Electromagnetic field strength can be higher along the line axis or at the edge of a right-of-way, but
decays more quickly with distance from the cable.
7. Cable faults are permanent in nature and usually require long repair time.

Different choices of cable system types may influence the need for shunt compensation or other remedial actions.
For instance, 1-Core and 3-Core solutions, onshore and offshore respectively will give varying needs of additional
equipment depending on parameters like phase separation, transpositions etc. This might give varying constraints
for the cable system depending on the connections points. Further discussions about cable system types can be
found in Chapter 3 of [8].

Aim and focus of this technical brochure

The aim of this technical brochure is to serve as a practical guide for preparing models and performing studies
necessary during the assessment of the technical performance of HV/EHV systems with a large share of (long) AC
cables with a nominal voltage of 72.5 kV and higher. The brochure is particularly focused on the differences in the
technical performance of an electric power system with long AC cables as compared to one with AC overhead lines.
These differences in performance are due to the different electrical properties of AC cables as compared to OHLs.
To interpret the results of studies and identify how they are affected by the distinctive characteristics of the cable, it
is necessary to be familiar with the characteristics of various cable models in terms of their performance and
resultant range of applicability, the effect of input data and their practical availability at various study phases, and
the effect of physical properties of cables on their electrical performance. The reader is provided with some
essential information about requirements for cable models for system technical performance studies over a range

Page 2
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

of frequencies. This ranges from data gathering to model accuracy, effect of data on model performance,
evaluation of available cable models and their applicability for various studies, methods of validation, etc. However
it must be emphasised that modelling of cables is covered in this Technical Brochure only to the extent as required
to perform the described system studies. Information such as derivations of mathematical equations that constitute
various cable models are not presented. Such information can be found in [8], which is referred where applicable.
Summary of earlier work
This Technical Brochure contains many references to earlier work, particularly in Chapters 2-4. Some of the IEC,
CIGRE, and IEEE publications offer information of broad interest in cable studies. For example, IEC/TR 60071-4 [9]
describes guidelines and test cases for a variety of transient overvoltage studies, over time frames ranging from
temporary overvoltage to very fast-front overvoltage. Some of those test cases involve long OHL or GIL with shunt
compensation, so those results would be similar to those from cable system studies. System islanding overvoltages,
line energization, lightning transients, resonance between line and shunt reactor, and arc extinction following
reactor switching are all covered in [9].
Many CIGRE documents cover cable modelling and studies. Much of this earlier material addresses lightning
transients and the impact of cable bonding methods, so this Technical Brochure focuses on other areas. TB 110 [2]
compares typical impedance and admittance values for OHL and cables. TB 189 [3] covers insulation coordination
for HVAC cables, including the overvoltage types of interest, calculation of the highest theoretically possible
switching overvoltage, and a lightning overvoltage example. TB 189 suggested more work on cables with extruded
insulation, and on developing guidelines for the use of simulation tools, which this Technical Brochure addresses.
TB 250 [4] discusses cable electrical characteristics and critical length, switching overvoltages, lightning
overvoltages, and sheath overvoltages during faults. TB 250 includes an example of EMT simulation of lightning
overvoltage on a mixed OHL/UGC (Underground Cable) line. TB 268 [5] provided more analysis and examples of
lightning transients on long cables, and recommended future work on long-cable switching transients, which this
Technical Brochure addresses. TB 283 [6] discussed the effect of cable bonding techniques on steady-state sheath
overvoltages during faults, and transient sheath overvoltages on mixed OHL/UGC lines during lightning strikes and
auto-reclose operations. It compared sheath overvoltage calculations from theory with various numerical and EMT
models. TB 347 [7] addressed the effect of cable bonding methods on earth potential rise during faults.
From CIGRE General Sessions, [38] described studies for a 39 km, 500 kV UGC in Tokyo with load shedding
overvoltage examples, and [88] described studies for a 43 km, 400 kV UGC in Italy with focus on shunt
compensation and parallel resonance. An IEEE panel session presentation [89] described how to prepare cable
data for use in EMT studies, including model responses to a step voltage excitation. This material was later
expanded into a book chapter [90]. Two IEEE references [13][14] address the special protective relaying issues
that may arise with UGC. All of these references would be helpful to engineers beginning a cable system study.

Definition of a “Long Cable”

In this brochure, “a long cable” is a cable that normally requires shunt reactor compensation and / or other
components installed with the cable. It often involves a number of cross-bonding points and continuous grounding
of the sheath, except for pipe-type cables and submarine cables. A single-circuit, cross-bonded EHV cable line that
is longer than 20 km typically requires shunt compensation [10][11][12].

Page 3
Chapter 2. ANALYSIS OF THE TECHNICAL PERFORMANCE OF POWER
SYSTEMS SPECIFIC FOR SYSTEMS WITH LONG HVAC CABLE LINES

2.1 Introduction
The application of cables in power grids and mixed line-cable-line sections will influence the behaviour of the power
system in several aspects. The steady state, transient and dynamic responses of cables for both small and large
disturbances differ from overhead lines. This has implications for the operation of systems with cables in terms of
technical performance and security of supply. Therefore, all steady state and transient phenomena should be
subject to study before integration of cables is realised in practice.
From the planning and the operational points of view, overhead lines cannot simply be replaced by cables. The
electrical behaviour of cables is different from that of overhead lines. An overhead line is surrounded by air without
other insulation and has a dominant inductive behaviour. A cable consists of an inner and a sheath conductor with
insulating material and semiconductor layers in between and is therefore predominantly capacitive. Cable
capacitance and inductance per unit length differ from the values for overhead lines. The series inductance of a
cable is about five times smaller and the shunt capacitance is around 20 times larger than the corresponding value
for an overhead line. As a result, the surge impedance of cables is roughly 10 times lower and the travelling wave
velocity is about half that of overhead lines. The value of the surge impedance lies between 30 and 70 Ω. This
means the surge impedance loading (SIL) for a cable is several times larger. The cable current rating is usually well
below its SIL and hence behaves like a shunt capacitor.

Reactive power surplus in any operating condition causes a power-frequency voltage rise, not only at the cable
terminations but also at the adjacent nodes in the grid. In normal operating conditions, a certain voltage step is
usually allowed while connecting or disconnecting a cable as prescribed in the grid code or the
planning/operational criteria of the system operator. To keep voltages within acceptable margins, reactive power
compensation is usually necessary. This compensation can be achieved using shunt reactors typically installed at
both ends of a cable.

The capacitive current of a cable depends on the applied voltage and the capacitance per unit length. At the critical
cable length, the cable current rating is completely consumed by the capacitive current and no active power can
flow through the cable. Shunt reactive power compensation installed along the cable route will correct this.

The amount and location of shunt compensation influences the voltage profile along the cable. Theoretically,
uniformly distributed shunt compensation may produce the best voltage profile, but at a high cost. The external
system also plays a role. System studies, as illustrated later in this chapter, identify the optimal compensation
scheme to maintain acceptable voltage profile at a reasonable cost.

Cables and overhead lines can create parallel transmission paths. The lower impedance of cables can lead to an
inequality in the power flow, which could even result in overloading the cable connections. This implies that during
steady state operation, compensation is needed in two different ways: shunt compensation for reactive power
surplus and series compensation for the difference in impedance to control the power flow on parallel paths.

The higher capacitance of a cable also has consequences for the system’s resonant frequencies. Cables are more
likely than overhead lines to create system resonances at low-order harmonic frequencies. Resonances may lead
to overvoltages or steady state harmonic problems. Cable switching transients are in principle similar to overhead
line switching transients. Shunt reactor switching should also be considered. Series and parallel resonance with
transformers may exacerbate cable switching overvoltages.

Another transient issue to consider is the zero-missing phenomenon that can occur while energizing shunt
compensated cables. The decay of the DC component of the current depends on the resistances that are, in
general, small. This results in a current not crossing zero for several cycles. Countermeasures can be taken to
avoid failure of circuit breakers in the event of a fault.

Page 4
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

From the installation point of view, high voltage underground cables longer than 1-2 km require cross-bonding to
reduce induced sheath currents. For shorter lengths, single-point bonding is generally applied. Solid bonding
(earthing in several points) is generally applied to submarine cables but single core cable laying may require
special studies. See Figure 3-2 for an illustration of cross-bonding, and [8] for illustration of other bonding methods.

2.2 Studies and Tools

This chapter follows the planning and analysis tasks of a typical underground transmission project. The main
considerations are power flow, voltage control, and operational criteria that include reliability and system security.
Once the high-level specification (number and type of cables, voltage level, current rating, reactive compensation)
has been determined, a more detailed design phase follows to specify equipment, such as circuit breakers, shunt
reactors, and surge protection. No foreseeable problem should affect the reliable and safe operation of the system.
Example test systems are used to illustrate study techniques, including interpretation of results. The studies can be
divided into four types, corresponding to chronological phases of a project’s planning, design, and implementation:
1. System planning study, which determines where new lines are needed, the voltage and current ratings,
and major auxiliary equipment such as shunt compensation. At this stage, system contingencies are
considered and overhead vs. underground options may also be considered. The basic study tool is a power
flow program, and positive sequence, power frequency models are adequate.

2. System impact study or detailed system design study, which determines the impact of a cable on the rest
of the power system. At this stage, the high-level specification has been determined. The system impact
study may result in some adjustments, or mitigations applied to the cable or other parts of the system.
Study topics include harmonic resonance, short circuit currents, transient stability, voltage stability, and
system relaying. The study tools include short-circuit, stability, and harmonic analysis programs, and in
some cases an EMT program to explore resonant overvoltages. The modelling needs vary from lumped
parameter to distributed parameter, from positive sequence to three-phase unbalanced representation, and
from DC to a few kHz, depending on the subject. Models are often generic in early studies, later
progressing to specific models for particular equipment.

3. Equipment and system design study, which establishes detailed protection and operating procedures for
the cable, sheath, switchgear, shunt compensation, and related equipment. The basic study tool is an
electromagnetic transients (EMT) program, which can also handle the grounding and bonding connections.
Accurate frequency dependent models are preferable and sometimes necessary for many of these studies.

4. Forensic study, which becomes necessary if a cable or equipment failure occurs during or after
commissioning. This type of investigation is not unique to cables. Information in this Technical Brochure on
modelling and phenomena may help fulfil this type of study.

There may be some iterations between study phases. For example, in an extreme case, resonance problems
identified in phase 2 may change the choice from high-pressure fluid-filled (HPFF) to cross-linked polyethylene
(XLPE) cable, in order to reduce the total capacitance. In another example, zero-miss phenomena may influence
the shunt reactor arrangement. However, usually, the output of one study phase fixes the input to the next phase.

2.2.1 Planning study scope

The planning study addresses:
1. Cable types to use

2. Parallel transmission paths

3. Shunt compensation schemes

4. Phase transposition and cross-bonding

5. Electromagnetic Fields (EMF) and all other environmental issues , as they affect the overhead vs.
underground choice

Page 5
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

2.2.2 System impact study scope

The impact study addresses:
1. Fault levels, which affect switchgear duty and safety. Increased penetration of cable circuits in transmission
systems can increase the prospective fault levels, because series impedance of cables is typically 2 (EHV)
to 5 (HV) times smaller than the equivalent overhead line circuit of the same length. Also, changes in the
system X/R ratios can affect short-circuit DC current time constants.

2. Self-excitation in synchronous generators may occur when the machine’s internal reactance is greater than
the equivalent external reactance. A system with cables is more likely to meet this condition than a system
with overhead lines, as described in section 2.5.6.

3. Voltage stability – Underground cables (UGCs) provide more reactive power than overhead lines (OHLs).
Switched or variable shunt reactors associated with UGCs may provide more flexibility in reactive power
control. However, the loss of these components must be considered as a contingency in voltage stability
analyses.

4. Transient stability - The lower series reactance of cable lines is favourable to angle stability. On the other
hand, underexcited operation of nearby synchronous generators is detrimental due to smaller internal
electromotive force. and higher load angle. Usually the net effect is favourable.

5. Small signal stability - For due diligence, the small-signal stability of the system should be checked to
determine whether cable installation has any significant impact due to changes in system impedances.

6. Temporary Overvoltages (TOV) are more likely with cables, due to:

a. Series and parallel resonance

b. Transformer energization

c. Load shedding or system islanding

7. Ferroresonance is sensitive to the amount of capacitance isolated with a transformer winding. Cables may
provide enough, or too much, capacitance for classical ferroresonance to occur. Transformers and long
cables should not be switched together.

8. Protection issues. References [13]–[15] provide a starting point for considering the impact of cables on
relaying and protection.

2.2.3 Equipment and system design study scope

The design study addresses:
1. Switching transients (energization, fault clearing, etc).

2. Energization of long HVAC cables from already energized cables (similar to back-to-back capacitor
switching).

3. Circuit breaker capacitive current ratings.

4. TOV and energy duty will influence arrester specifications.

5. Circuit-breaker restrike after interrupting small inductive current (i.e. disconnecting a shunt reactor).

6. Lightning transients and arrester placement at cable junctions, open points and other locations. Arresters
may not be needed at all within the cable system [5].

Page 6
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

7. Auto-reclosing. Impact on system availability and reliability could be considered in the Planning Study. If
reclosing is to be used on a mixed OHL/UGC line, impacts on sheath requirements for current carrying
capability and the effect of mutual couplings of shunt reactors should be considered.

8. Cable discharge of trapped charges through inductive voltage transformers (VTs) and their energy
capability requirements.

9. DC offset in cable and shunt reactor energization and zero-miss phenomenon when high compensation
degrees are deployed.

10. Shunt reactor switch reignition and current chopping.

11. Saturation of current transformers from DC offset currents when energizing shunt reactors and cable.

12. Synchronized (also known as point-on-wave) switching to mitigate inrush currents or overvoltages.

13. Energization of transformer at the end of a cable.

14. Sheath and core overvoltages [16].

15. Effect of cables on circuit breaker transient recovery voltage (TRV). Because cables have lower surge
impedance and higher capacitance, the TRV rate-of-rise is usually lower than with overhead lines. Fault
currents may be higher with cables due to lower series impedance, but in general, the TRV is less severe
with cables. However, any associated current-limiting reactors (CLR) would require their own TRV study.

16. Effect of cables on incoming surges to the substation. The substation insulation coordination and arrester
placement are usually based on a design incoming surge due to lightning. If the connected circuits are all
underground cables, there will be no direct incoming lightning surges. A different design base for
substation insulation coordination must be selected. This can be an attenuated lightning surge from a
mixed OHL/UGC line, or some type of high-frequency switching surge.

2.2.4 System technical performance issues and study tools

Table 2-1 summarises system technical performance issues associated with long or many AC cables. The issues
are grouped by study phase, and then sorted by weight (column “Wt.”) within each group to emphasise their
relative importance and uniqueness (the higher the weight, the more important and unique the issue).

For example, the “active power transmission” issue in the fourth row of Table 2-1 is important, but the concerns are
nearly the same as for overhead lines, resulting in a low weight for the purpose of this Technical Brochure. On the
other hand, “reactive power compensation” in the first row is both important, and of more concern for cables than
for overhead lines, resulting in a high weight.
The study tools mentioned in Table 2-1 are:
 Cable Constants – a program that calculates and represents electrical RLC parameters in a matrix form for
a general system of cables and other conductors, over a range of frequencies, and using either transposed
or full unbalanced assumptions. This function may be bundled with another tool, or used separately.

 Power flow – calculate steady-state voltages and currents based on a positive sequence model, with non-
linear loads. The cable model is symmetric and transposed. Power flow is the basic tool for transmission
planning.

 Short-Circuit – a program that solves for voltage and current during faults, especially three-phase and
single-phase-to-ground faults. The model is linear, symmetric, and assumes phase transposition. An
auxiliary Protection function simulates the response of relays to fault current and voltage.

Page 7
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

 Dynamics – a time-domain simulator based on numerical integration of differential equations. It differs from
an EMT program in focusing on (slower) electromechanical and control system transients, rather than
electromagnetic transients. The models are sometimes linear and balanced. The program usually includes
eigenvalue analysis, or other functions for small-signal stability.

 Harmonics – a frequency domain program that solves voltage and current over a range of frequencies,
using linear or non-linear load and source models, and balanced or unbalanced impedances. The
frequency-scan function outputs driving point impedance, as obtained from the bus voltage for a unit
current injection.

 Electromagnetic Transients (EMT) program – a time-domain or transient simulator based on numerical

integration of differential equations, including non-linear component models, unbalanced impedances, and
frequency-dependent RLC parameters. An EMT program can also perform frequency scans, and may
include a Cable Constants function.

Phase representations in Table 2-1 are:

 + Seq – positive sequence lumped parameters. (Note: unlike overhead lines, the positive sequence lumped
parameters of a cable can vary significantly up to 1 kHz).

 +/0 Seq – positive and zero sequence lumped parameters.

 3Φ – three-phase unbalanced lumped or distributed parameters, with earth returns reduced.

 3Φ-sh – three-phase unbalanced distributed parameters, retaining the cable sheaths as separate
conductors.

Frequency classes in Table 2-1 are:

 50/60 – parameters calculated at power frequency.

 DC-kHz – parameters used up to switching surge frequencies, and may also be frequency-dependent.
Typical for slow-front overvoltages.

 MHz – parameters used up to lightning / fast switching surge frequencies, and may also be frequency-
dependent. Typical for fast-front overvoltages.

Custom tools can also be built on general matrix solvers [17]. These work well in focused studies of different cable
phenomena, but they may be less convenient for system operators to use with existing large-scale models.

Table 2-1: Performance and Model Issues related to Study Phase

Study Issue Wt. Tool(s) Phase Freq.

Planning Reactive power compensation 5 Power flow + Seq. 50/60
Planning Ferranti effect 3 Power flow + Seq. 50/60
Planning Active power transmission (limits and 1 Power flow + Seq. 50/60
contingencies)
Planning Phase unbalance and transposition 1 Cable Const. 3 50/60
Impact Harmonic frequency scans 5 Harm. / EMT 3 DC-kHz
Impact TOV – Ground faults and clearing 5 EMT 3 DC-kHz
Impact TOV – Resonance 5 EMT 3 DC-kHz
Impact TOV – System islanding and load rejection 5 Dyn. / EMT 3 DC-kHz
Impact Generator self-excitation in steady state 4 Power flow + Seq. 50/60
Impact Ferroresonance 3 EMT 3 DC-kHz
Impact Harmonic sources and flows 3 Harm. / EMT 3 DC-kHz
Impact Generator stability (power factor) 3 Dynamics + Seq. 50/60
Impact Short-circuit levels 3 Short Cct. +/0 Seq. 50/60

Page 8
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Study Issue Wt. Tool(s) Phase Freq.

Impact Short-circuit, low X/R ratio 3 Short Cct. +/0 Seq. 50/60
Impact Relaying 2 Short Cct. / +/0 Seq. 50/60
Prot.
Impact Telecommunications over cable 1 Cable Const. 3 DC-kHz
Impact Voltage, RMS dynamic, small-signal stability 1 Dynamics + Seq. 50/60
Equipment Cable energization 5 EMT 3-sh DC-kHz
Equipment Auto-reclosing overvoltages 5 EMT 3-sh DC-kHz
Equipment Discharge of the cable 5 EMT 3 DC-kHz
Equipment Sheath overvoltages and protection 5 EMT 3-sh DC-kHz
Equipment DC offset / zero-miss effect 4 EMT 3 DC-kHz
Equipment Current transformer (CT) saturation 3 EMT 3 DC-kHz
Equipment Influence on substation surges and TRV 2 EMT 3-sh MHz
Equipment Capacitive current interruption 1 Cable Const. 3 50/60
Equipment Lightning overvoltage 1 EMT 3-sh MHz
Equipment Shunt reactor restrike 1 EMT 3 MHz
Equipment Inductive coordination and EMI 1 Cable Const. 3 50/60

2.3 Example Test Systems

Two example test systems are defined to illustrate system technical performance studies. Explanations of basic
phenomena use reduced sets of components from one of these two test systems.

2.3.1 Many-cable example test system

Figure 2-11 shows the IEEE 24-bus reliability test system [18]. This test system, denoted RTS-79, has two 138 kV
cables, one 3 miles (4.83 km) and the other 16 miles (25.74 km) long. There is also a 230 kV portion of the test
system, and several generators. The system power frequency is 60 Hz. A later version called RTS-96 [19] adds
phase-shifting transformers, HVDC lines, and overhead inter-area ties, but no new cables. Because of its familiarity,
the many-cable test system is based on Figure 2-1. Planning and system impact studies are illustrated by replacing
overhead lines with cables.

2.3.2 Long-cable example test system

Figure 2-2 2 illustrates a simple network model for applications of a “long-cable” system. The system power
frequency is 50 Hz. The shunt reactor sizes may be adjusted, moved, made variable, or switched to cover different
scenarios. The parallel cables can also be switched in or out of service to cover different scenarios. Cable charging
capacity is compensated to 94.0 % for both 108 km and 54 km cables.
Table 2-2, Table 2-3 and
Table 2-4 contain the cable and transformer data. The cable conductors are transposed at 1.5 km intervals and
buried at a depth of 1.5 m in horizontal (flat) configuration with 0.5 m between the middle and outer phase centres.
The sheaths are cross-bonded at 1.5 km intervals and grounded at 4.5 km intervals. The soil resistivity is 100 Ω-m.
The resistivity values in Table 2-2 were taken from IEC 60287-1-1 Table 1 at 20°C to yield less damping. All
relative permeability values are 1. To account for the semi-conducting screens, an effective insulation permittivity
was estimated from:
R3 R
 i1   i 0 ln( ) / ln( so ) , (2.1)
R2 Rsi
where i0 = 2.4 is the insulation material’s actual relative permittivity and Rso, Rsi, R2, R3 are defined in Table 2-2.

1
Notation applied in this drawing is the following: numbers next to generator symbols are their MW generation,
numbers next to load symbols are their MW consumption, numbers next to reactive power compensation symbols
are their MVar generation, and numbers next to transformer symbols are their MVA rating.
2
Numbers next to transformer symbols are their MVA rating (for example 3*500 means 3 transformers of 500 MVA
rating).

Page 9
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

With one cable in each corridor and simple equivalent 230 kV source inductances of 30 mH at Sub A and 25 mH at
Sub C, a frequency scan from Sub B shows a resonant frequency near 100 Hz. This resonance is at the 2nd
harmonic, suggesting possible resonant overvoltages after energizing the transformer at Sub B. RLC ladder
networks may be used to synthesise more complicated network equivalents of the system shown in Figure 2-2.
28.96 km

km
.71
54

96.54 km
107.80 km
19.31 km

43
.4
4
km
54.71 km

46.66 km
km
88.50 km

.44
43

37.01 km

km
.45
80

25.74 km

Figure 2-1: Many-cable Example System [18], © IEEE

Page 10
POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure 2-2: Long-cable Test System

Page 11
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Table 2-2: 400 kV Cable Data for Figure 2-2

Component Parameter Description Units Value
Conductor R1 Inner radius of tubular core mm 0
R2 Outer radius of tubular core mm 28.3
ρc Resistivity of copper Ω.m 1.724e-8
Conductor Screen Tcs Thickness of conductor screen mm 2
Rsi Outer radius of screen mm 30.3
Insulation Ti1 Thickness of insulation mm 26
εi1 Effective relative permittivity 2.8
Rso Outer radius of insulation mm 56.3
Insulation Screen Tis Thickness of insulation screen mm 2
R3 Outer radius of insulation screen mm 58.3
Metallic Sheath Ts Thickness of sheath mm 2
R4 Average outer radius of sheath mm 60.3
ρs Resistivity of aluminium tape Ω.m 2.840e-8
Outer Covering Ti2 Thickness of cover mm 5
εi2 Relative permittivity of cover 2.4
Complete Cable R5 Average outer radius of cable mm 65.3
S Minor section length km 1.5
Lcw Cross-bond wire inductance, 20 m mH 0.02
Lgw Ground wire inductance, 20 m mH 0.01
d Burial depth m 1.5
h Phase horizontal separation m 0.5
Normal Joint Rjoint Ground resistance of normal joint Ω 5
Substation Rsub Ground resistance at substation Ω 1
Cable Charging C’ Capacitance F/km 0.216
B’ Charging at 400 kV MVar/km 10.832

Table 2-3: 500-MVA Example Transformer Data

Parameter Primary Secondary Tertiary
Line-Line Voltage [kV] 400 230 66
Connection Y Y d1
Leakage Inductance [mH] 150 25 10
Resistance [Ω] 0.50 0.20 0.05
Stray Capacitance [pF/] 5000 1000 8000

Table 2-4: Transformer Saturation Data

I [A, peak] Phi [Wb-T, peak]
0 0
0.05 600
1.01 1200
150 1350
400 1450
650 1530
1250 1700

Page 12
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

2.4 System Planning Issues

Adding a cable into a system mainly consisting of overhead lines, or replacing some overhead lines with cable, will
have some implications on the power flow distribution, losses, and fault levels. The first aspect is the series
reactance X, which is lower for cables, resulting in different power flows for equal terminal conditions. The second
aspect is the X/R ratio (or loss angle), which can be an order of magnitude larger for an overhead line compared to
a cable.
In a situation of parallel transmission paths a cable may carry more load than intended, due to its lower impedance
(a similar issue would occur in the case of parallel overhead lines with unequal impedances). Likewise, unbalanced
flows occur when XLPE and paper-insulated cables are in parallel. There are several solutions to this issue, such
as the use of phase shifting transformers and carefully selected series reactors. All of these situations with parallel
paths must be studied carefully. However the methods and tools needed for these studies are practically the same
as in case of systems based on overhead lines, and therefore will not be discussed in this document any further.

2.4.1 Reactive power compensation

The planning studies related to the reactive compensation of underground cables are similar to those carried out
for long overhead line circuits. Extensive information on this subject can be found in [20][21]. However, due to the
particular features of underground cables, some differences in the reactive compensation planning for overhead
lines and cable circuits can be identified:
1. Minimum length of circuit requiring shunt compensation. As a general rule, the shunt capacitance of an
underground cable is in the order of 20 to 30 times the capacitance of an equivalent overhead line circuit.
Therefore, shunt compensation may need to be considered in underground circuits in the order of 20 to 30
times shorter than the equivalent overhead line. Typical examples of critical overhead line circuit lengths
requiring shunt compensation can be found in [22].

2. Degree of shunt compensation. Overhead lines are typically restricted to 60% to 80% shunt compensation
due to risks of line resonance during open-phase conditions [23]-[24]. This phenomenon is caused by
capacitive coupling between the disconnected phase and the remaining energized phases. Given that
there is no capacitive coupling between insulated cables, the restriction on low degree of shunt
compensation does not apply. Compensation degrees close to 100% are typical for underground cables. In
the case of mixed cable-overhead line circuits, inter-phase capacitive coupling will be present in the
overhead line section. Special attention must be taken in these configurations if single-phase autoreclose is
to be applied.

3. Voltage rise at the receiving (open) end must not exceed the maximum continuous operative voltage
(MCOV) or TOV capabilities of shunt reactors, surge arresters, or other equipment. The voltage rise is
often higher in the presence of cables.

4. No-load current of a compensated cable must not exceed the rated capacitive switching current capability
of circuit breakers. If necessary, special-purpose or higher-voltage circuit breakers may be used.

The main questions the system planner needs to answer when faced with a new long cable circuit are:
1. Does the cable circuit need shunt compensation?

2. Does the shunt compensation need to be connected to the cable circuit or can it be placed at the
substation ends (i.e. busbars or tertiary windings of transformers)?

3. Does the shunt compensation need to be fixed, switchable, or variable?

4. What is the optimum shunt compensation scheme for the selected cable circuit?

The following step-by-step procedure helps the system planner answer those questions by using standard power
flow analysis techniques. The following section provides an example. This procedure is based solely on steady-
state analysis looking at voltage rise due to Ferranti effect and voltage step during switching procedures. It applies
to the analysis of a single long HVAC cable circuit. Other limitations on the size and location of the shunt reactors

Page 13
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

may arise from time-domain analysis of resonance conditions. The system planner must be aware of both
limitations and some level of iteration may be required.
Step 1. Establish acceptance criteria for receiving end voltage rise, sending end voltage step, cable active
power reduction due to charging current, and capacitive breaking current of circuit breakers. If
these conditions are not fulfilled then line connected shunt compensation is necessary. Notably, if
the active power reduction is not acceptable, compensation is needed at intermediate locations.
The minimum shunt compensation degree has to be set accordingly, e.g. following the procedure in
[25].

Step 2. Create power flow model representing a minimum demand scenario.

a. Analyse voltage profile using intact network topology. This test may identify the need for shunt
compensation.

b. Analyse voltage profile along cable length assuming open circuit-breaker at one end. This test
determines the need and required location for the shunt reactors – i.e. at the cable ends or at the
substation busbar/transformer tertiary winding.

c. Define various shunt compensation schemes with reactors at either end of the circuit, at both ends
or at different locations along the circuit. Open circuit-breaker at each cable end must be
considered.

d. Assess the preferred compensation scheme using intact network topology. Circuit thermal loading
and system voltages should be within acceptable limits.

e. Assess the preferred compensation scheme using applicable planning criteria (e.g. single circuit
contingency). Post-disturbance circuit thermal loading and system voltages should be within
acceptable limits.

f. If either Step 2d or Step 2e fails, then the scheme is not satisfactory. A new compensation scheme
must be investigated with different sizes of reactors in different locations along the cable circuit.
Step 2 must be repeated with the new preferred option. A number of iterations may be required.

Step 3. Create power flow model representing a maximum demand scenario and verify the scheme
resulting from Step 2.

a. Intact network analysis as in Step 2d.

b. Contingency analysis as in Step 2e.

c. Iterative refinement as in Step 2f.

Step 4. Assess the maximum unit size of shunt reactor that can be switched individually to comply with the
required voltage step limits. This test will also give an indication of requirements for static or
dynamic shunt compensation.

Points that require special attention:

1. Voltage profile output includes at least three points along the cable length.

2. Open-circuit voltage profile along the cable circuit under minimum system loading conditions.

3. Normal operating voltage profile under maximum and minimum system loading conditions.

4. Different source impedances that may exist when switching the cable from either end.

5. Reactive power operating point of nearby synchronous generators, SVC or STATCOM.

Page 14
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

This procedure applies to the analysis of a single long cable. In areas with high concentration of short cables or
lower voltage levels, it may be more appropriate to carry out area-wide analysis to determine the optimum number,
size and location of shunt reactors, taking technical constraints into consideration and with the objective of
minimising system losses and costs.

2.4.2 Example case study of reactive power compensation

The proposed planning procedure for the selection of a suitable compensation scheme for a long cable circuit is
illustrated in this example. Throughout this example, results for an overhead line and the equivalent cable
configuration are shown for comparison purposes.
The IEEE reliability test system in Figure 2-1 has been used as the base case model. In this example, the longest
overhead line circuit in the system (230 kV circuit 17 – 22, 117.46 km long) is replaced with an equivalent
underground XLPE cable. The studies have been carried out using power flow in a commercial software package.
The assumed circuit parameters are shown in Table 2-5. For the purpose of monitoring voltage profile along the
cable circuit, the branch has been divided into four equal-length sections (Figure 2-3).
Full shunt compensation of this cable would require 440 Mvar in shunt reactors. However, the cable is so long that
some reactor arrangements require more than 100% compensation in order to achieve acceptable voltage profiles
along the cable. Higher shunt compensation levels are also required to avoid under-excitation of a nearby
generator. This example shows a workable solution, but intermediate switching stations along the cable might also
be considered. Although more expensive, the intermediate switching stations would enable a more uniform
distribution of compensation along the cable.

Table 2-5: Circuit 17-22 Parameters on 100 MVA Base

R [pu] X [pu] B [pu]
Underground Cable 0.0035 0.0351 4.4198
Overhead Line 0.0135 0.1053 0.2212

Figure 2-3: Power Flow Representation of Cable for Shunt Reactive Compensation Analysis

Step 1. Establish acceptance criteria for voltage rise and voltage step.

The assumed Planning Criteria, which may vary among different system operators, are:
 Normal voltage limits: 0.95 pu to 1.1 pu

 Emergency post-disturbance voltage limits: 0.9 pu to 1.1 pu

 Voltage step when switching reactive compensation devices : 3%

 For single contingencies the maximum step change between pre- and post- contingency steady-state
voltages shall not be more than 10%.

Step 2. Analyse minimum demand scenario in power flow.

a. Analyse voltage profile using intact network topology.

Results are plotted in Figure 2-4. Cable connection results in a significant voltage increase when
compared with the equivalent overhead line circuit. In particular, the line voltage along the cable

Page 15
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

circuit is well above the normal limit defined in Step 1 (1.1 pu). This result clearly indicates that
reactive compensation is required for the cable connection.

V [pu] UGC OHL

1.5

1.4

1.3

1.2

1.1

1
#17 25% 50% 75% #22

Figure 2-4: Voltage Profile for Intact Network Condition during Minimum Demand

b. Analyse voltage profile along cable length assuming open circuit-breaker at one end.

 A fault level analysis is carried out at buses 17 and 22 to identify the weakest source. Results in
Table 2-6 indicate that bus 22 has the lowest short circuit power. Therefore, the highest voltage
rise will occur when the cable circuit is energized from bus 22 with the circuit breaker open at bus
17.

Table 2-6: Short Circuit Power at Circuit Sending and Receiving End
Bus 17 Bus 22
Cable 1963 MVA 1686 MVA
Overhead line 1901 MVA 1475 MVA

 Analyse voltage profile assuming circuit energized from weakest source (bus 22) and circuit
breaker open at strongest side (bus 17). Simulate the opening of the circuit breaker at the
strongest cable end (bus 17) and solve load-flow for the cable energized from the weakest
source (bus 22). Results are plotted in Figure 2-5. In this particular example (for the UGC option
only) power flow could not converge due to the nearby generators reaching their under-excitation
limits and being unable to control voltage. A fictitious solution was found in this case, for
illustration purposes only, by increasing the reactive power limits of the generation units. This is a
severe condition and should be avoided. (In Figure 2-4, bus 17 absorbed the excess UGC
reactive power without exceeding any generator under-excitation limits, so that power flow
solution did not have to be modified.) In order to reduce voltages below 1.1 pu and ensure that
generators are operated within the reactive power capabilities, a shunt compensation scheme
must be implemented. The results of the open-circuit analysis indicate that the required shunt

Page 16
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

compensation must be directly connected to the cable, ruling out the bus or transformer location
options.

# 17 25% of 50% of 75% of # 22

Length Length Length
Strong Weak
R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4
Source Source

Open CB Closed CB

V [pu] UGC OHL

1.200
1.175
1.150
1.125
1.100
1.075
1.050
1.025
1.000
#17 25% 50% 75% #22

Figure 2-5: Voltage Profile Results with Circuit Energized from Weak Source (Bus 22) and
Open-end at Strong Source (Bus 17)

Assess various shunt compensation schemes.

 Scheme 1: Shunt reactors installed at the strongest source cable end (bus 17).

The results of the open-circuit test with cable energized from the weaker source are plotted in
Figure 2-6. The minimum size of shunt reactor that guarantees compliance with the voltage limit
criteria is 700 Mvar.
A second open-circuit test is carried out with Scheme 1 (700 Mvar at bus 17) by energizing the
cable from the strong source and leaving the circuit breaker open at the weak source end. The
results of this test are plotted in Figure 2-7. It can be seen that line voltages along the cable are
below the normal and emergency voltage limits defined in Step 1, therefore Scheme 1 is rejected.

Page 17
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

V [pu] 400 Mvar 500 Mvar 600 Mvar 700 Mvar

1.150

1.125

1.100

1.075

1.050

1.025

1.000
#17 25% 50% 75% #22

Figure 2-6: Scheme 1 (Reactors in Cable-end Bus 17) Voltage Profile with Cable Energized
from Weak Source (Bus 22) and Open-end at Strong Source (Bus 17)

# 17 25% of 50% of 75% of # 22

Length Length Length
Strong Weak
R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4
Source Source
700 Mvar
Closed CB Open CB

V [pu] 700 Mvar

1.000

0.975

0.950

0.925

0.900

0.875

0.850

0.825

0.800
#17 25% 50% 75% #22

Figure 2-7: Scheme 1 (Reactors in Cable-end Bus 17) Voltage Profile with Cable Energized
from Strong Source (Bus 17) and Open-end at Weak Source (Bus 22)

Page 18
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

 Scheme 2: Shunt reactors installed at the weaker source cable end (bus 22).

The results of the open-circuit test with cable energized from the weaker source are plotted in
Figure 2-8. It can be seen that the minimum size of shunt reactor that guarantees compliance
with the voltage limit criteria is 600 Mvar.
A second open-circuit test is carried out with Scheme 2 (600 Mvar at bus 22) by energizing the
cable from the strong source and leaving the circuit breaker open at the weak source end. The
results of this test are plotted in Figure 2-9. It can be seen that line voltages along the cable are
below the normal voltage limit defined in Step 1, therefore Scheme 2 is rejected.

# 17 25% of 50% of 75% of # 22

Length Length Length
Strong Weak
R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4
Source Source
Varies
Open CB Closed CB

V [pu] 400 Mvar 500 Mvar 550 Mvar 600 Mvar

1.200
1.175
1.150
1.125
1.100
1.075
1.050
1.025
1.000
0.975
#17 25% 50% 75% #22

Figure 2-8: Scheme 2 (Reactors in Cable-end Bus 22) Voltage Profile with Cable Energized
from Weak Source (Bus 22) and Open-end at Strong Source (Bus 17)

Page 19
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

# 17 25% of 50% of 75% of # 22

Length Length Length
Strong Weak
R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4
Source Source
600 MVar
Closed CB Open CB

V [pu] 600 Mvar

1.050

1.025

1.000

0.975

0.950

0.925

0.900
#17 25% 50% 75% #22

Figure 2-9: Scheme 2 (Reactors in Cable-end Bus 22) Voltage Profile with Cable Energized
from Strong Source (Bus 17) and Open-end at Weak Source (Bus 22)

 Scheme 3: Shunt reactors installed at both cable ends.

The results from Scheme 1 and Scheme 2 assessments indicate that the shunt compensation
needs to be split between the two cable ends in order to avoid excessive voltage rise or drop.
This is the objective of Scheme 3. The results of the open-circuit test with cable energized from
the weaker source are plotted in Figure 2-10. In this test the reactor size is increased at either
cable end until a solution that meets the Planning Criteria is found. This is initially achieved with
a 200 Mvar + 220 Mvar combination; however detailed inspection of the operating condition for
the generators connected to bus 22 revealed that they are running at their under-excitation limits,
which is an undesirable condition for stability. An increase of size in the reactor connected to the
bus 22 end results in a higher operating power factor in the nearby generators, providing an
adequate stability margin. Further reductions in the reactor size connected to bus 17 give rise to
overvoltages at the receiving cable end or result in the generators reaching their under-excitation
limits again. The selected shunt reactor combination from this test is 200 Mvar + 250 Mvar.
A second open-circuit test is carried out with Scheme 3 (200 Mvar at bus 17 + 250 Mvar at bus
22) by energizing the cable from the strong source and leaving the circuit breaker open at the
weaker source end. The results of this test are plotted in Figure 2-11. It can be seen that line
voltages along the cable are within the normal voltage limits defined in Step 1, therefore Scheme
3 is technically acceptable under open circuit breaker conditions. Scheme 3 becomes the
preferred option.

Page 20
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

# 17 25% of 50% of 75% of # 22

Length Length Length
Strong Weak
R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4 R/4, X/4, B/4
Source Source
Varies Varies
Open CB Closed CB

V [pu] 150-150 150-200 200-200 200-220

200-250 150-250 180-250
1.225
1.200
1.175
1.150
1.125
1.100
1.075
1.050
1.025
#17 25% 50% 75% #22

Figure 2-10: Scheme 3 (Reactors in Both Cable-ends) Voltage Profile with Cable Energized
from Weak Source (bus 22) and Open-end at Strong Source (Bus 17)

V [pu] 200-250 Mvar

1.100

1.075

1.050

1.025

1.000
#17 25% 50% 75% #22

Figure 2-11: Scheme 3 (Reactors in Both Cable-ends) Voltage Profile with Cable Energized
from Strong Source (Bus 17) and Open-end at Weak Source (Bus 22)

Page 21
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

c. Assess the preferred compensation scheme option in intact network topology

Preferred option is Scheme 3 with 200 Mvar at cable-end bus 17 and 250 Mvar at cable-end bus 22.
Power flow analysis for minimum demand scenario indicates that circuit thermal loading and system
voltages are within acceptable limits.
d. Assess the preferred compensation scheme option in single contingency

Contingency power flow analysis for minimum demand scenario indicates that circuit thermal loading
and system voltages are within acceptable limits.

Step 3. Analyse maximum demand scenario in power flow.

a. Assess the preferred compensation scheme option in intact network topology

Power flow analysis for maximum demand scenario indicates that circuit thermal loading and system
voltages are within acceptable limits.
b. Assess the preferred compensation scheme option in single contingency

Contingency power flow analysis for maximum demand scenario indicates that circuit thermal loading
and system voltages are within acceptable limits.

c. Assess voltage profile

The cable voltage profile under normal and open-circuit conditions is plotted in Figure 2-12. These
results indicate that the preferred scheme provides adequate performance under maximum demand
conditions.

V [pu] intact system cable open at #22 cable open at #17

1.10

1.09

1.08

1.07

1.06

1.05
#17 25% 50% 75% #22

Figure 2-12: Scheme 3 (Reactors in Both Cable-ends) Voltage Profile in Maximum Demand
Conditions

Page 22
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Step 4. Assess the maximum unit size of shunt reactor that can be switched individually.

Calculate the largest reactor size that can be switched individually to comply with the 3% voltage deviation
limit defined in the Planning Criteria. A minimum system loading scenario is used as it leads to the highest
voltage deviations. The results are:
 175 Mvar maximum at the bus 17 cable end

 150 Mvar maximum at the bus 22 cable end

Table 2-7 shows the selected compensation scheme out of three considered, accounting for all steady-state
constraints.

Table 2-7: Selected Compensation Scheme out of Three Considered

Cable-end bus 17 Cable-end bus 22 Degree of Compensation
2 x 100 Mvar 1 x 100 + 1 x 150 Mvar 101.81%

2.5 System Impact Issues

2.5.1 Harmonics and network resonances

Inductances and capacitances of power system components create series and parallel resonant frequency points.
The number and frequency of these resonant points depend on the number, size and lumped or distributed nature
of components, their electrical parameter values, and placement with respect to each other and the point of interest.
Switching operations and nonlinear elements such as iron cores or power electronics can generate a wide
spectrum of high-frequency voltage and current components, which may then trigger oscillations at system
resonant frequencies. Excessive harmonic distortion and overvoltages may result. System resistances tend to
dampen the oscillations, especially at higher frequencies and greater distances from the disturbance. Due to higher
capacitance values associated with cables, the resonant frequencies tend to be lower than without cables, and at
lower frequencies the damping also tends to be lower.
The first step in evaluation is to determine the resonant frequency points in a power system, and their series or
parallel nature. In a frequency scan analysis, one per-unit (or one ampere) current is injected at each bus and
frequency of interest; the resulting bus voltage response is equal to the driving point impedance in per-unit (or in Ω)
in frequency domain. The results efficiently identify sensitive locations and frequencies for harmonic sources, and
the impact of different system and parameter variations.
The second step is to determine the system response to excitation of those resonant frequency points. A harmonic
power flow, performed in the frequency domain, calculates the steady-state response to multiple harmonic sources,
which are described with theoretical or measured spectra. The results establish whether voltage distortion limits
and current ratings are met, and also verify performance of harmonic filters. EMT simulation also plays a role when
circuit breaker switching operations, faults, power electronic switching, control dynamics, or non-harmonic sources
excite system resonant points. With EMT simulation in the time domain, it is often necessary to post-process
waveforms using the Fast Fourier Transform (FFT) in order to evaluate them.
As discussed above, frequency scans are used to identify resonance points and resonant frequencies. A system
with cable will tend to have lower resonant frequencies, due to higher capacitance or line charging. The first
resonant harmonic order can be estimated by

MVA s
h , (2.2)
MVA c

where MVAs is the system short circuit strength at the cable and MVAc is the cable’s reactive power at nominal
voltage. Local shunt compensation can be added to MVAs. It must be pointed out that this formula neglects the
capacitance of the existing network and therefore overestimates the frequency of the first parallel resonance.

Page 23
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Table 2-8 shows the positive-sequence line parameters used to illustrate frequency scan analysis on the test
system in Figure 2-1. The 230 kV underground parameters match those used in the previous shunt compensation
example. The 138 kV underground parameters have been modified to yield realistic values for series impedance,
surge impedance and propagation velocity. In [18], the line charging value had been increased without a
corresponding reduction in the series impedance for 138 kV cables. The frequency scan cases cover gradual
conversion of all overhead lines to underground, in the following sequence:
 Base case with two existing underground 138 kV lines

 Convert the longest 230 kV overhead line to underground

 Convert all remaining 138 kV lines to underground

 Convert all remaining 230 kV lines to underground (no overhead lines left in the system)

At each step, shunt compensation of approximately 90% is added for each cable, split equally between the two
ends.

Table 2-8: Many-cable Example Test System Line Parameters for Frequency Scan Example
Line Type R1’[pu/km] X1’[pu/km] B1’[pu/km]
230 UGC 2.980E-05 2.988E-04 3.763E-02
230 OHL 1.149E-04 8.965E-04 1.883E-03
138 UGC 5.385E-04 4.158E-04 1.520E-02
138 OHL 6.168E-04 2.489E-03 6.775E-04

Figure 2-13 shows the positive sequence driving point impedance from one of the 138 kV transformer terminals on
Bus 10 (“H pu” on the horizontal scale refers to harmonic order, which is also the per-unit frequency). These results
include long-line PI adjustments at each frequency, but no other frequency-dependent line or cable effects. Loads
provide some damping. There are two short cables in the base case, including one that terminates on Bus 10.
These produce a parallel resonance around the 8th harmonic, and the addition of one long 230 kV cable does not
affect this resonance. Conversion of all 138 kV lines to underground will produce parallel resonance near the 2.5,
7.5, and 14th harmonics. Removing all 138 kV shunt reactors has little effect on the frequency scan results at higher
frequencies than the fundamental. Their effects could be more noticeable if the loads were not modelled. When all
overhead lines have been converted to underground, three additional resonant frequencies appear, and two of
them are tuned to the 2nd and 3rd harmonics.
Figure 2-14 shows the positive sequence driving point impedance from the 230 kV side of the same transformer on
Bus 12. The base case impedance is primarily inductive at this bus, even when all 138 kV lines are underground.
The long cable from Bus 17 to 22 also has little impact at Bus 12. However, when all of the 230 kV lines are
underground, at least nine parallel/series resonance points appear in the result.
Many systems have seen growth in background harmonic distortion levels, even at EHV transmission buses. After
converting overhead lines to underground, new parallel resonances at lower frequencies will probably exacerbate
the harmonic distortion. A complete evaluation should consider variable load level because that influences the
damping. Higher damping is indicated with a reduction in magnitude or increase in phase angle of the impedance
at resonance. The study should also consider line outages, transformer outages, and generator outages, because
they can shift the resonances to lower frequencies and increase the peak impedances.

Page 24
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure 2-13: Many-cable Example Test System Frequency Scans from Bus 10

Figure 2-14: Many-cable Example Test System Frequency Scans from Bus 12

Page 25
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

2.5.2 TOV - Resonance

Temporary overvoltages (TOV) have several causes, including ground fault back feed, generator self-excitation,
and resonance. The last cause is a special concern with cables, because they lower the system resonant
frequencies. Under normal switching conditions, the transient voltages may be damped. However, if there is an
excitation current at the resonant frequency, overvoltages may be sustained for several seconds, damaging surge
arresters or other equipment.

2.5.2.1 Parallel Resonance

Consider the long-cable test system (Figure 2-2) viewed from Sub A, energizing one cable with the other circuit
already in service. If the three Sub A transformers (22% impedance on 500 MVA) contribute most of the short-
circuit strength, then with shunt reactors on both cables, MVAs = 8980 in equation 2.2. The reactive power due to
the capacitance of both cables together amounts to 2340 MVA, and the per-unit resonance frequency can be
estimated as 1.96; very close to the 2nd harmonic.

On a system with a long HVAC cable the inductance of the shunt reactors and the cable's distributed capacitance
form a parallel resonance circuit. Parallel resonance is characterized by large impedance at the resonant
frequencies, with relatively low current drawn from the external circuit (current source). Larger currents may
circulate through the resonant L and C and produce higher transient overvoltages [26].
An example situation is the energization of a transformer through a cable. A simple model and its harmonic
equivalent circuit are shown in Figure 2-15. Harmonic current due to the transformer energization is injected to the
parallel resonance circuit composed of a long HVAC cable, its shunt reactor and equivalent source impedance.

Figure 2-15: Parallel Resonance due to a Transformer Energization and its Harmonic
Equivalent Circuit

The parallel resonance frequency of the equivalent circuit in Figure 2-15 can be found as:

1 1 1
f   (2.3)
2 LC L0 C

Page 26
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Severe TOV is expected when the parallel resonance frequency found by (2.3) is a multiple of the fundamental
frequency, especially 2nd harmonic. Considering typical compensation level of a long cable around 100 %, it
requires a weak source, namely large L0, to have the parallel resonance frequency at 2nd harmonic.
The parallel resonance caused by a transformer energization is simulated with the long-cable example test system
in Figure 2-2. The 400/230 kV transformer at Sub B is energized from Sub A and Sub C through long cables. As
explained in Section 2.3, a frequency scan from Sub B 400 kV bus shows a resonant frequency near 100 Hz, with
one cable in each corridor and simple equivalent source inductances of 30 mH at Sub A and 25 mH at Sub C. The
result of the frequency scan is shown in Figure 2-16.

Figure 2-16: Impedance of the Long-cable Example Test System Seen from Sub B 400 kV
Bus

Figure 2-17 shows the inrush current (bottom) and the parallel resonance overvoltage caused by the inrush current
(top), reaching a peak of 1.27 pu. This overvoltage has a strong second harmonic component. In a similar case
using overhead line with 80% shunt compensation, parallel resonance overvoltage also occurs, with a strong
seventh harmonic component. The relative voltage peaks depend on the instants of switch pole closings but in the
worst case, peak overvoltages were higher with cable. The overvoltages also decay faster with overhead line.
These observations hold true when the system model includes frequency-dependent loads and/or frequency-
dependent losses in lines and transformers.

Page 27
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

500 415.3 kV (1.27 pu)

[kV]
375

250

125

-125

-250

-375

-500
0.0 0.1 0.2 0.3 0.4 [s] 0.5
(file SimpleModel.pl4; x-var t) v:SUB_BA v:SUB_BB v:SUB_BC

300

[A]

200

100

-100

-200

-300
0.0 0.1 0.2 0.3 0.4 [s] 0.5
(file SimpleModel.pl4; x-var t) c:SUB_BA-X0001A c:SUB_BB-X0001B c:SUB_BC-X0001C

Figure 2-17: Parallel Resonance Overvoltage (Top) and Inrush Current (Bottom)

2.5.2.2 Series Resonance

The series (leakage) inductance of a transformer can create a series resonance circuit with the capacitance of a
cable. This situation is characterized by low harmonic impedance at the resonance frequency, which may result in
large harmonic current and high TOV at the connection between the transformer and the cable.
An example of this situation is the energization of a cable near a series resonance circuit. A simple model and its
harmonic equivalent circuit are shown in Figure 2-18. In this example, harmonics generated by the cable
energization excites the series resonance, causing high TOV at the location Vs.
The series resonance frequency of the equivalent circuit in Figure 2-18 can be found from the inductance of the
400/230 kV transformer (LS) and the total capacitance of 230 kV cable (CS).

Page 28
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

1
f  (2.4)
2 L S C S

Severe TOV is expected when the frequency of a harmonic component in the cable energization current matches
the series resonance frequency found by (2.4).

ZS = 0

Figure 2-18: Series Resonance due to a Cable Energization and its Harmonic Equivalent
Circuit

The series resonance caused by a cable energization is simulated with the long-cable example test system in
Figure 2-2 with some modifications. First, the 400 kV cable between Sub A and Sub B is shortened to 9 km. Also,
the 400 kV cable between Sub B and Sub C is shortened to half of the original length and is energized from Sub B.
After these modifications, the cable energization overvoltage contains a 140 Hz component.
In order to set the series resonance overvoltage at 140 Hz, the total capacitance of 230 kV cable (CS) is set as
287.9 MVA (CS = 17.33 μF), connected at Sub B. Due to this large charging capacity, 230 kV shunt reactors are
required to compensate the charging capacity and thus maintain reasonable steady-state voltage. In this example,
90% compensation is assumed with total 260 MVA shunt reactors.
1 1
f    140 Hz (2.5)
2 LS C S 2 0.07459  (17.33  106 )

Figure 2-19 shows the series resonance overvoltage on the LV side of the 400/230 kV transformer in Sub B.
Severe TOV occurs due to the series resonance.

Page 29
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

360

[kV]

240

120

-120

-240

329.6 kV (1.76 pu)

-360
0.0 0.1 0.2 0.3 0.4 [s] 0.5
(file SeriesResonance_rev3.pl4; x-var t) v:B2A v:B2B v:B2C

Figure 2-19: Series Resonance Overvoltage in the LV Side Bus of the 400/230kV
Transformer

TOV are usually evaluated against surge arresters TOV and energy absorption capabilities, because surge
arresters are the most likely to fail due to TOV. These resonance TOVs are weakly damped, because they occur at
low frequencies. These TOV have characteristics of both power frequency TOV and high-energy switching surges,
which are tested separately for surge arresters. A higher line discharge class or multiple columns would help with
energy absorption, but not with sustained TOV. A higher arrester voltage rating could help with both.
Other countermeasures to the parallel and series resonance overvoltage include synchronized (also known as
point-on-wave) switching for the transformer energization and the application of operational restrictions.

2.5.2.3 Ferroresonance
Ferroresonance consists of interactions between capacitors and saturable iron-core inductors [27][28]. It can occur
when the transformer and cable become isolated, and the cable's capacitance is in series with the transformer
magnetizing characteristic. This may lead to sustained temporary overvoltages that damage surge arresters and
other equipment.
If one phase of a circuit breaker opens while the other two remain closed the open phase of the transformer is still
energized through the mutual inductance to the other phases. Overvoltage occurs on all phases, with highest
magnitude occurring on the open phase. This type of ferroresonance is more common on distribution systems,
where fuses may be employed. More information on particular cases can be found in WG C4.307 TB (under
development).
Ferroresonant voltages may also occur even without circuit breaker malfunction, if a transformer and cable are
energized together. To defer switchgear costs, high-voltage systems have sometimes used low-side switching or
transformer-terminated switching, in the early stages of development. In low-side switching, a transformer is
energized from the low side with an unloaded line already connected to the high side. This is often safe with
overhead lines, but should be avoided with long cables. In transformer-terminated switching, a line is energized
with an unloaded transformer already connected to the remote end, as shown in Figure 2-20.
An important factor in a ferroresonance analysis is the point-on-wave switching angle, which affects the voltage
distortion and amplitude, similar to inrush current effect in a transformer energization [29]. Its influence can be seen
in Figure 2-20, which shows the voltage in the transformer secondary during the energization of a transformer-
terminated 108 km cable for two different switching angles (simulation based on the long cable test system).

Page 30
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure 2-20: Energizing Transformer-terminated Line, Transformer Secondary Voltages.

a) Switching Angle: 90º; b) Switching Angle: 0º

2.5.3 TOV – Ground faults and clearing

Figure 2-21 shows an equivalent circuit with source and cable capacitance, during a nearby fault. After clearing the
fault, both cable capacitance and any nearby transformers are effectively re-energized as the system voltage
recovers. Because many transformers experience voltage recovery, this can produce a higher TOV than normal
energization of a single transformer.
It is important to model the transformer core’s non-linearity, loss, and ability to hold residual flux with hysteresis
characteristic as shown in Figure 2-22. Figure 2-23 shows the transformer current during a terminal fault at the
instant of voltage zero, which is near 11 ms in this case. The magnetizing current is at a peak when the fault occurs,
and holds that level with little decay during the fault. In Figure 2-21, this current circulates in the equivalent
magnetizing inductance and it supports a residual flux. When the fault is cleared at about 76 ms, residual flux
increases the inrush current in Figure 2-23. Only transformers near the fault can hold residual flux this way;
otherwise the current circulates through system component resistances to decay faster, and also the transformer
terminal voltage would not be zero during the fault.

Figure 2-21: Equivalent Circuit for TOV after Ground Fault Clearing

Flux vs. Current Lower Loop

Anhysteretic
300.00

200.00
Volt-Seconds Peak

100.00

0.00

-100.00

-200.00

-300.00
-20.00 0.00 20.00 40.00 60.00 80.00
Amperes Peak

Figure 2-22: Typical Transformer Hysteresis Characteristic

Page 31
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure 2-23: Transformer Current during the Fault

Figure 2-24 shows the ground-fault-clearing TOV from a case based on the long-cable test system in Figure 2-2. All
four cables are in service, with a parallel resonance at 80.5 Hz as determined by frequency scan. The fault occurs
at the Sub A end of a 108 km cable, and is cleared after 3 cycles. With that cable out of service, the first parallel
resonance frequency increases to 92.5 Hz. The shunt reactors have a Q factor of 500, and a 200 MVA load with
0.98 power factor was included at Sub B.
The surge arresters are usually most vulnerable to TOV; they are not intended to mitigate TOV and may fail
thermally during the event. Arresters are tested for TOV capability [30], but the standard test employs a stiff power-
frequency voltage source with constant output. The TOV in Figure 2-24 occurs at higher frequency and variable
magnitude, so both the energy discharge and TOV capabilities of the arrester must be evaluated.
The TOV peaks in Figure 2-24 do not follow a constant or steadily decaying profile, which makes the evaluation
difficult. Arrester switching surge energy and lightning discharge capabilities do not apply to the longer TOV time
frame, because the arrester failure mechanisms are different. In addition to thermal discharge capability, the
dielectric strength also plays a role in determining TOV capability. The arrester TOV capability should not be used
as a withstand envelope. The arrester won’t necessarily withstand a voltage waveform that lies under the TOV
curve, which describes a series of independent tests at different times and constant voltage magnitudes. It has
been suggested that comparing simulated energy discharge to the lightning impulse discharge energy capability,
as defined in Annex N of [31], may give conservative results for harmonic TOV, pending the possible adoption of a
new thermal rating system proposed for [31]. Another option is to define a TOV level and time that the voltage
peaks exceed for 90 % of the TOV time, and convert that to a 10-second equivalent TOV [32]. However, the
waveforms in Figure 2-24 are not steady enough to use that procedure. The arrester vendor should be consulted in
such cases; this was done for a recent study of TOVs in Southwest Connecticut [33].
If an arrester fails during this event, it creates a second fault right after the initial fault. The consequences to
reliability may be severe because at least one more component outage occurs. System loads provide significant
TOV damping; however, on the other hand higher pre-fault voltages exacerbate the TOV by increasing transformer
inrush currents. It’s difficult to control this TOV. To date, ground fault TOV has been mitigated by:
 Limiting the cable lengths to avoid resonances at or below the 3rd harmonic

 Increasing the arrester voltage ratings, which adversely affects insulation coordination

Page 32
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

 Reducing system voltage, especially during light load periods. This reduces transformer inrush current
magnitudes.

Other methods have been proposed, including C filters and reactor switching [34][35][36].
A TOV study should include various system contingencies, fault locations, shunt compensation levels, and loading
levels as these affect the resonant frequency and damping rate of TOVs.

Figure 2-24: Evaluation of a Typical TOV Envelope

2.5.4 TOV – System islanding

When one end of a long cable is opened, a part of a network can be separated from the main grid together with the
long cable. The overvoltage caused by the system islanding can be expressed by the following equations [37]:

v(t )  Vm sin t  Vm sin  0 t (2.6)
0

Em L 1 1
Vm  , 0   , (2.7)
L0 (1   2 CL)  L CL0 CL

where L0 is source impedance of the weaker islanded system and Em is the source voltage behind L0. Charging
capacity of the long cable and inductance of the shunt reactors directly connected to the cable are expressed by C
and L respectively.

Page 33
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure 2-25: Equivalent Circuit of the System Islanding

Equation (2.6) shows that the overvoltage contains two frequency components, the nominal frequency ω and the
resonance frequency ω0. Since the overvoltage is caused by the superposition of two frequency components, the
resulting overvoltage level is often difficult to estimate before the simulation. The result of a simulation performed
for the 500 kV Shin-Toyosu line is shown in Figure 2-26 [38].
The overvoltage level is sensitive to L0, which expresses the short circuit level in the islanded system. In order to
find the most severe overvoltage, it is necessary to study different network conditions.

2.23 pu

1.0s

Without Surge Arresters

1.69 pu

1.0s

With Surge Arresters

Figure 2-26: Overvoltage Caused by System Islanding [38]

In addition, excitation control of generators in the islanded system can affect the sustained overvoltage in the
islanded system, which affects the evaluation of insulation strength of the long cable and other equipment in the
system island. In this case, excitation control such as automatic voltage regulator (AVR), power system stabilizer
(PSS), and under-excitation limiter (UEL) needs to be modelled in the simulation.

Page 34
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

After forced disconnection at the land end of a long cable connecting a large offshore wind farm, the cable system
and the wind farm will be in island operation. Such forced events are rare, but do happen, introducing TOV in the
cable system. In the forced island operation of the 160-MW Horns Rev 1 offshore wind farm during 2006, the
temporary overvoltage was approximately 2 pu of the nominal cable voltage with total duration of a few seconds
[39]. The Horns Rev 1 cable system has no arresters; therefore, this overvoltage stressed the cable. In general, the
overvoltage magnitude and duration depend on the cable capacitance, compensating reactor inductance,
transformer, wind farm operating conditions, and wind turbine technology.

2.5.5 Temporary overvoltages versus cable tests recommended by IEC

The examples in Sections 2.5.2 to 2.5.4 revealed severe temporary overvoltages related to long or many cable
systems. In order to test the insulation performance of cables, lightning impulse voltages, switching impulse
voltages and power frequency test voltages are applied to cables according to international and national standards.
IEC 62067 defines the test methods and requirements for XLPE cables and accessories for rated voltages above
150 kV up to 500 kV. According to IEC 62067, the test voltage 440 kV is applied to 400 kV cables for 15 minutes,
as power frequency voltage test, after the lightning impulse voltage test.
Considering the voltage level that can be attained by TOVs, the level of applied power frequency voltage appears
to be somehow low even though the duration is sufficient. This mismatch is not an immediate issue as the
overvoltage is normally evaluated against the insulation performance and the energy absorption capability of surge
arresters. However, if one wishes to evaluate the insulation performance of cables, a theoretical conversion of
insulation performance to a shorter duration and a higher voltage level is possible using a voltage-time relationship
of insulation performance. With the voltage-time relationship, a severe temporary overvoltage can be evaluated to
be within the insulation performance of cables and accessories when the following inequality is satisfied.

V n  t  Vtest n  ttest , (2.8)

where V and t are the level and the duration of temporary overvoltages, and Vtest and ttest are the level and the
duration of applied power frequency test voltages. n is a parameter that can generally be obtained from a cable
manufacturer with typical values ranging from 9 up to 15.

2.5.6 Generator self-excitation

Excessive reactive power generated by a long transmission circuit connecting a power plant to the transmission
grid can cause self-excitation of a synchronous generator. This phenomenon can arise in the event of load
rejection and system separation, or during a black-start restoration procedure [40][41]. This condition can also arise
due to the presence of capacitor banks and filters associated with HVDC [42]. The self-excitation phenomenon can
cause an uncontrolled voltage rise in the generator terminals, only limited by the saturation of the step-up
transformer. Incidents of catastrophic equipment failure due to self-excitation of synchronous generators have been
reported [43].
The selection of a cable instead of an overhead line makes the event more likely to occur for shorter circuit lengths.
Figure 2-27 shows the minimum generator size to avoid self-excitation as a function of the line length l km, with
total reactance of 2 pu for the generator and step-up transformer [41][44][45]. The line and generator configuration
is radial to a strong source.
The physical phenomenon can be explained with Figure 2-28, which represents a round rotor synchronous
generator connected to a long transmission line at no-load. The internal no-load characteristic of the machine is
represented by “Gen Sat”. The external characteristic “UGC – No Rct” is determined by the transformer and
transmission line reactance and capacitance, without shunt compensation. The generator finds an equilibrium
operating point at VT1; this voltage can be well above the rated insulation of the machine.
A condition for self-excitation is that the generator internal synchronous reactance is higher than the external
equivalent reactance. Long transmission lines or cables connected in series with low impedance step-up
transformers are prime candidates to cause this phenomenon. Self-excitation can occur both in hydro (salient pole)
and thermal (round rotor) units, although it is more likely in thermal units due to their larger synchronous reactance.
Possible mitigation measures to avoid self-excitation include the connection of several generators in parallel (i.e.
reducing generator reactance) or the installation of shunt reactors in the transmission circuit (i.e. increasing
external reactance). The effect of the shunt reactor installation has been illustrated in Figure 2-28 with “UGC-w/Rct“,
resulting in a new operating point VT2 with lower voltage at the generator terminals.

Page 35
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

1500
reactance x*=2 p.u.
Cable line
Overhead line

2(31600 mm2)
1000
2(3630 mm2)
Size
[MVA] 31600 mm2)
3630 mm2)

500

OHL
0
0 10 20 30 40 50
[km]
Figure 2-27: Minimum Generator MVA to Avoid Self-excitation on an Unloaded Line
1.6

1.4

1.2
VT2 VT1
Voltage [p.u.]

1.0

0.8
Gen Sat
0.6
UGC-w/Rct
0.4 UGC-No Rct
0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Current [p.u.]

Figure 2-28: Self-excitation Phenomenon in a Synchronous Generator

2.5.7 Transient stability

To illustrate the effect of cables on transient stability, example simulations are conducted on a simplified system,
derived from the Sub A to Sub B portion of the long-cable test system in Figure 2-2. A 1000 MVA synchronous
generator at Sub B, generating 800 MW, is connected to a 10-GVA external grid at Sub A through two circuits. For
the overhead line system no shunt compensation is used. The study consists of applying a three-phase short circuit
on one of the lines and evaluating the critical clearing time of the fault. The short circuit is applied at different
locations, namely 10 %, 25 %, 50 %, 75 % and 90 % along the line length. Other parameters vary as discussed
below.

Page 36
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

The critical clearing time is mainly dependent on the X/R-ratio and the ratio between line and generator impedance.
Because these parameters are different for cables and overhead lines, the critical clearing time changes when an
overhead line is replaced with a cable. Figure 2-29 shows the critical clearing time for overhead line and
underground cable, based on the 108 km line length. The critical clearing time is slightly higher for the underground
cable, especially for fault locations close to the substations.
These factors influence the critical clearing time in a similar way for both overhead line and underground cable:
 Decreasing line length causes the critical clearing time to increase.

 Increasing the series inductance decreases the critical clearing time and vice versa. However, for very low
values of series inductance the generator reactance dominates, and further decreasing series inductance
has little effect.

 Increase or decrease in the shunt capacitance has no effect.

 Increasing the series resistance provides more damping and increases the critical clearing time.

OHL UGC
0.24
0.23
Critical Clearing Time [s]

0.22
0.21
0.20
0.19
0.18
0.17
0.16
0.15
0 20 40 60 80 100

Distance from Sub A [% of line]

Figure 2-29: Critical Clearing Times

A generator’s critical clearing time also depends on its operating power factor. For example, a typical 1300-MVA
machine might have critical clearing times of 290 ms at pf = 0.9, 190 ms at pf = 1, and 150 ms at pf = –0.9. In a
system with many cables, the generator operating points may shift to unity power factor or even into absorbing
reactive power. This will reduce critical clearing times and adversely affect transient stability.
Shunt compensation of the cables will mitigate this effect on transient stability. However, a fault in the
compensation equipment can still shift nearby generators into absorbing reactive power and reduce critical clearing
times. Also, clearing faults on compensated cables may alter the overall system compensation level and may
reduce local bus voltages. Either of these can reduce critical clearing times and adversely affect transient stability.

2.6 Equipment Design Issues

2.6.1 Cable energization

The energization of cables generates transient overvoltages with large time derivatives. The energization transient
of a cable is similar to the energization transient of a capacitor bank, with the difference that in a cable the
capacitance is distributed. Surge propagation occurs along the cable, damping the transients.
Energization transient studies can be local, depending on the strength of the bus from which the energization is
carried out. Energization transients are simulated to verify two issues:

Page 37
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

1. Is the cable insulation adequate, and if not, is it necessary to use remedial measures such as pre-insertion
resistor, synchronized switching or line surge arresters.
2. If the cable connects two systems with different strengths (i.e. fault levels), one may want an operating
restriction on the end from which the energization will take place (i.e. best-end switching).

Unless there is absolute certainty that there cannot be re-energization with trapped charge, a statistical simulation
of at least 200 cases should be run to assess the maximum overvoltages with trapped charge.
A simulation of overvoltage and overcurrent during cable energization was carried out using the long-cable test
system in Figure 2-2; specifically the 108 km cable connected directly to a 400 kV ideal voltage source. CB1 is
closed to energize the cable, with CB2, CB3, and CB9 open. Transformer models are not included in this test.
Figure 2-30 shows the voltage in the receiving end of the 108 km long cable and the current in the sending end.
The maximum value of overvoltage occurs in phase A (blue line). The dominant transient frequency is
approximately 260 Hz, which is lower than for typical overhead lines. An overvoltage of 2.1 pu is reached. The
sustained large current as a result of the transient might cause overcurrent relays to mis-operate.

Figure 2-30: Receiving End Voltages and Sending End Currents during Energization

The overvoltage caused by cable energization is also affected by cable type and sheath bonding. The next
theoretical example compares a 400 kV single-core cross-bonded cable, a 400 kV single-core two-point bonded
cable, and a 132 kV pipe-type cable. Figure 2-31 shows the cross section diagrams of the single-core cable and
the pipe-type cable. The single-core cable is assumed to be laid in a flat formation with 0.35 m phase separation.

Figure 2-31: Cross Section Diagram of Single Core Cable (Left) and Pipe Type Cable (Right)

The cable length is set to 1.5 km and the length of the minor section for the cross-bonded cable is 0.5 km. One
phase of the cable is energized through 1 Ω resistance. The cable sheath is grounded at both ends through 10 Ω
grounding resistance.
Figure 2-32 compares the energization overvoltages for the three types of cables. The dominant frequency of the
energization overvoltage in the cross-bonded cable is much lower than the other two types of cables, which leads
to lower damping of the overvoltage. In addition, in the cross-bonded cable, the overvoltage propagates to the
other phases (b and c) through sheath cross-bonding.

Page 38
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

600
[kV] 1.67 pu
500

400
Phase a
300

200

100
Phases b and c
0

-100

-200
0.0 0.4 0.8 1.2 1.6 [ms] 2.0

500

[kV]
1.47 pu
400

phase a
300

200

100

phases b and c
0

-100
0.0 0.4 0.8 1.2 1.6 [ms] 2.0

200

[kV]
1.75 pu
160

120 phase a

80

40

phases b and c
0

-40
0.0 0.4 0.8 1.2 1.6 [ms] 2.0

Figure 2-32: Comparison of Energization Overvoltage – Single Core Cross-bonded Cable

(Top), Single Core Two-point Bonded Cable (Centre), Pipe Type Cable (Bottom)

Page 39
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Energization of one cable from a bus requires particular attention. The event is similar to back-to-back switching of
capacitor banks, which can produce high inrush currents. Figure 2-33 shows the CB1 current upon energizing the
first 108-km cable in Figure 2-2. CB2, CB3, CB4, and CB9 are all initially open. This cable has over 90% shunt
compensation, so that relatively large currents flow into the shunt reactors. Each pole of CB1 closes at an instant of
voltage peak to minimize DC offset, but the currents will still take a long time to decay.
Figure 2-34 shows the CB1 currents upon closing CB2, after the currents in Figure 2-33 have reached steady state.
Again, each pole of CB2 closes at the instant of voltage peak, which minimizes the DC offset current into the
second cable but maximizes the capacitive back-to-back current. The transient currents in CB2 are symmetrical to
the transient currents in CB1, due to their back-to-back configuration. The back-to-back currents reach higher peak
values and oscillate at a higher frequency than in Figure 2-33. The first half cycle of phase A current in Figure 2-34
shows the traveling wave nature of back-to-back cable energization currents, but the reactive energization currents
and source impedance effects are still significant. Circuit breakers may have peak inrush current, frequency, and/or
current-frequency product ratings to check. Both single energizations and back-to-back energizations should be
considered.

Figure 2-33: CB1 Currents During First Cable Energization in the Long-Cable Test System

Figure 2-34: CB1 Currents During Second Cable Energization in the Long-Cable Test
System

Page 40
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

2.6.2 Auto-reclosing overvoltages

Auto-reclosing for mixed overhead / cable lines requires a careful risk assessment. Cable faults are typically
permanent; thus a reclose won’t succeed and may cause further damage or safety hazards. Auto-reclosing has
been applied to some mixed lines in the following ways:
 The line can be segmented with switchgear to allow auto-reclosing on part of it. This effectively converts a
mixed line to separate lines.

 Use fault detectors that block auto-reclosing in case of cable faults.

 Attempt auto-reclosing only during storms.

 Attempt auto-reclosing only if the underground cable section is much shorter than the overhead line section.

 Attempt auto-reclosing only if the fault current is relatively low.

When single phase auto-reclosing is applied to mixed overhead / cable lines, high degrees of shunt compensation
can cause open-phase resonances due to capacitive coupling between a disconnected phase and the energized
phases in the overhead line section [46].

2.6.3 Cable discharge

In a system with an AC cable compensated by a shunt reactor, when the cable is disconnected from the rest of the
system there is an energy exchange between the cable capacitance and the inductance of the shunt reactor. The
natural frequency of this exchange is approximately found by Equation (2.2). For example, on a 50-Hz system the
natural frequency is 38.73 Hz for 60 % compensation that is typical for overhead line, or 48.48 Hz for the long-
cable test system that is compensated to 94 %.

This can be verified in Figure 2-35 where the voltage during the cable de-energization is shown. The shunt reactor
compensates 46 % of the reactive power of the cable, producing a resonance frequency of approximately 34 Hz. In
Figure 2-35 b) it takes a long time to completely damp the voltage in the cable due to low resistance. Typical cable
discharge times are several minutes with shunt compensation since Q factor of typical shunt reactor ranges around
500. Without shunt compensation, inductive VTs will quickly saturate and cable discharge will be completed in
several hundred milliseconds. A three-phase three-legged shunt reactor will have unequal mutual inductance
between phases that should be included in the model. These mutual inductances may produce different modulating
beat frequencies in the discharge voltage on each phase.

Figure 2-35: Receiving End Cable Voltage during Disconnection: a) Zoom and b) First 5 s

Page 41
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

2.6.4 Zero-miss effect

Zero-miss phenomenon can happen when compensating more than 50 % of a cable’s reactive power using shunt
reactors [47][48][49]. It is characterized by a current not crossing zero for several cycles after energizing the cable
and reactor together. This would be a problem in case of a fault (or grounding chains) on the cable during its
energization. It may be impossible to open healthy-phase poles of a circuit breaker without damaging the breaker.
Shunt reactor current contains both an AC and a decaying DC component during energization. DC component
decay can take several seconds depending on the cable and shunt reactor resistances. Its initial value depends on
the terminal voltage value at the instant of energization, reaching its peak if energized at voltage zero crossing.
The AC components of the cable capacitive current and shunt reactor current are in phase opposition. These
opposing AC current components partially cancel. The current in the line breaker is the sum of the AC and DC
components, and may become dominated by the decaying DC component.
Using the circuit in Figure 2-36 to explain the phenomenon, the AC current in the shunt reactor is given by Equation
(2.9) and that in the line breaker is given by Equation (2.10).

Figure 2-36: Single Line Diagram of a System with a Cable, Shunt Reactor, and Circuit
Breaker

I SR   x  IUC (2.9)

I CB  I SR  IUC  1  x   IUC , (2.10)

where ISR is the AC current component in the shunt reactor, ICB is the AC current component in the circuit breaker,
IUC is the AC current component in the cable and x is the reactive power compensation ratio.
If the reactive power compensation ratio is higher than 50 %, the DC component in the circuit breaker may be
higher than the AC component. In this situation the current does not cross zero during several cycles.
Figure 2-37 shows an example of zero-miss phenomenon for a 150 kV 50 km cable, with 70 % of the reactive
power being compensated by a shunt reactor installed at cable's end. The system is energized when the voltage is
zero and therefore the DC component is at its maximum value. The DC component may saturate the current
transformer effectively blinding the protection devices.
Figure 2-37 a) shows the currents in the circuit breaker (blue), in the shunt reactor (green) and in the cable (red).
The current in the circuit breaker does not cross zero. Figure 2-37 b) shows the current in the circuit breaker for a
period of 5 s, and it takes several seconds before the current crosses zero for the first time. The X/R ratio of a
shunt reactor is typically between 400 and 700, but may be up to 1350 [50] leading to a decay time constant of
4.3 s, neglecting source resistance. The time needed for the current in the circuit breaker to cross zero is
calculated by Equation (2.11), assuming X/R = 1350 on a 50-Hz system.

Zoom

Figure 2-37: Zero-miss Currents for 70 % Shunt Compensation

Page 42
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Ls  x  1   x 1 
t ln    t   4.3  ln   (2.11)
Rs  x   x 

Table 2-9 summarises options to mitigate zero-miss phenomenon.

Table 2-9: Mitigation of Zero-miss Phenomenon

Technology Conditions

 requires higher leading current interruption capability

Sequential switching  requires single phase circuit breaker and current
differential relay

 causes higher switching overvoltage

Synchronized switching
 requires single-phase circuit breaker

CB with pre-insertion  may be necessary to develop a new circuit breaker

resistor (expensive)
Additional series resistance  requires special control to disconnect series resistance
in shunt reactor for
after energization
energization

Energize reactor and cable  causes steady state overvoltage or undervoltage and
in sequence voltage step

2.6.5 Couplings to other conductive infrastructures

The voltages and currents induced in the infrastructure close to a long cable (both during normal operation and
during phase(s)-to-ground fault on a cable) can cause danger to people, damage to the structure or may disturb the
operation of another system, e.g. telecommunication circuits [51], [52]. For example, the induced voltage due to
inductive or conductive coupling can cause touch voltage to the affected pipe-lines or contribute to corrosion [53].
Capacitive coupling from a cable is not significant because of its sheath.
The most common type of coupling is the inductive coupling, when the current or a part of it returns through the
earth. The fraction of the zero sequence current which returns through the earth is the inducing current. The
coupling is quantified by the mutual impedance between the inducing and induced structures with common earth
return. The result of the inductive coupling is the longitudinal electromotive force acting along the induced structure.
Due to the large skin depth of the return current in the earth, the inductive coupling can be significant even at a
distance of several hundred meters between the coupled systems. On the other hand, with balanced current (i.e.
positive sequence current) the inductive coupling is limited to practically five to ten times the phase conductor
separation. This is a very small influence distance in the case of a cable under normal operation.
Conductive coupling can occur when current is flowing from the structure of a cable to the earth directly or via an
earthing electrode, thus resulting in earth potential rise (EPR). Figure 2-38 shows an example with 300 Ω.m (left
side picture) and 3000 Ω.m (right side picture) soil resistivities. The area of the EPR is referred to as the zone of
EPR. The coupling is quantified by the mutual conductance between the two structures. The conductive coupling
exists along the cable trench only when there is a current exchange through the continuous leakage between the
sheath and the earth. The current exchange zones are typically at the fault location and at the feeding location(s).
High EPR can occur at the place of phase-to-sheath-and-earth fault. This causes risk to the affected structures
when the fault location coincides with a crossing of the cable and structures. The EPR of the sheath with insulating
cover could also cause risk of conductive coupling when the sheath potential is high enough to break through the
insulating jacket of the cable.

Page 43
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure 2-38: Substation Zone of Influence with and without Lines

The zone of EPR could appear near an earthing electrode used for local grounding of the sheath when the earth
fault location coincides with the earthing location of the sheath. The sheath of a cable connected to the ground grid
of a substation can transfer the ground potential from an earth fault in the station to remote locations.
The countermeasure can preferably be applied at the source cable. For the protection of the environment an
effective solution is the two-point grounding of the cable sheath. This can also be strengthened with conductor laid
along the cable and connected to the sheath at the grounding locations. Local countermeasures such as improved
grounding or filters may also be applied at the influenced system. However, the most efficient countermeasure is to
increase the cross-section of the sheath of the inducing cable, because the inducing effect is then decreased due
to the following reasons:
 The sheath - as a current return conductor - leads back an increased part of the return current, which
decreases the portion of the earth current in the zones close to the cable terminations (end effect zones)
and thereby decreases the EPR itself
 The sheath - as a compensating conductor - leads an increased part of the return current also along the
stationary section (between the end zones) of the cable line. This is also applicable to the sheath of the
telecom- and signal cables.
Therefore, with a given amount of material the most effective compensation can be achieved when this additional
material is built into the cable as e.g. metallic screen composed of copper wires. This is always more effective than
a separate compensating wire.

Page 44
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

2.7 Modelling of Other Equipment

This section discusses the modelling of equipment other than the cable under study. This includes:
 shunt reactors

 transformers

 loads

 overhead lines and other cables

 network equivalents

 surge arresters

 circuit breakers

Various models of cables are discussed in Chapters 3 and 4.

The important characteristics of shunt reactors and transformers are listed in Table 2-10 and Table 2-11,
summarizing their effects on the results of system impact studies and equipment design studies discussed in this
chapter. The importance of these models is well-documented [9].

Table 2-10: Effects of Shunt Reactor Characteristics

System impact study Equipment design study
Cable
Ground faults System Cable discharge
Resonance energization &
and clearing islanding and zero-miss
autoreclose
affects affects power- affects
affects
affects pre- frequency of frequency compensation
Linear inductance resonance
fault voltage transient component of level and initial
frequency
components overvoltages DC offset

Core non-linearity affects

causes
(knee-point and frequency of
ferro-
air-core transient
resonance
inductance) components

affects damping
Loss (magnetic small effect affects damping of oscillations
and copper) on damping of oscillations and time to
current zero
affects zero affects zero affects zero
Mutual (inductive) sequence sequence sequence
coupling impedance*1 impedance*1 impedance*1
*1: three-phase shunt reactor with a three-leg iron core

Page 45
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Table 2-11: Effect of Transformer Characteristics

Equipment design
System impact study
study
Ground faults Cable energization &
Resonance System islanding
and clearing autoreclose
affects affects frequency
Short circuit affects feeding affects feeding
resonance of transient
impedance impedance impedance
frequency components

Magnetizing
impedance and affect amplitude
affect frequency
core non-linearity cause and frequency of
of transient
(knee-point and ferroresonance transient
components
air-core components
inductance)
Losses (magnetic affect damping affect damping
and copper) of overvoltages of overvoltages
affects zero affects zero
Mutual (inductive) sequence sequence
coupling impedance*1 impedance*1
*1: Three-legged three-phase transformer

Shunt reactors, transformers and generators are often modelled without frequency dependence due to
unavailability of precise models. If the precise model is available, the frequency dependent model is useful to yield
reasonable damping in a high frequency region. For a simplified model of transformer damping, an equivalent
network of parallel RL sections can be used to realise a constant X/R ratio at power frequency and above.
In system impact studies discussed in this chapter, assumed load levels have a significant impact on the
magnitude of overvoltages. As the low load conditions generally cause more severe overvoltages, special attention
is required if the cable line is used for the black start restorations. The common practice among utilities is to vary
load levels to find the impact, assuming constant impedance loads. The assumed load power factor and
representation (e.g. series RL vs. parallel RL) also affects the damping provided for transients. For the same real
power load, higher power factors will provide more damping. At minimum load, which is a conservative assumption,
the power factor is likely to be 0.98 or higher. A series RL model is conservative because it provides less damping
than other equivalent circuit load models.

2.7.1 Extent of cable network to be modelled in detail

The peak voltage after a cable energization/re-energization does not occur at the energization/re-energization
instant, but some hundreds of micro-seconds later. This instant corresponds normally to the moment in which the
voltage wave generated at the CB switch-on reaches the cable receiving end for the second time. To observe any
effect of wave reflections from nearby buses before the time of peak overvoltage, the round trip wave travel time to
those buses must be less than the time of peak overvoltage. The wave travelling time in each line can be estimated
from the respective physical parameters, or typical values [54] can be used. This concept supports a zone of
interest for travelling wave line models.
The number of cross-bonded sections, when modelled, can also be minimised applying the same method and
considering the inter-sheath mode wave speed. In this case, only the lines closer to the cable being energized
contain a detailed model of the cross-bonded sections.
In summary, one should divide the simulation model into three zones. A first zone, closest to the switched cable,
uses full modelling detail, i.e., frequency-dependent (FD) models are used and all cross-bonded sections are
included in the model. A second zone, further out, uses FD models, but the cross-bonding is simplified to one major

Page 46
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

cross-bonded section. A third zone, with one busbar depth, uses lumped-parameter models. This reduces the
likelihood of erroneous high overvoltage at the expense of a relatively small increase in the computational effort. A
2-port or N-port boundary equivalent network represents the system outside these three zones.

Pi-models

FD-model

Detail Cross-bonded
models

Energised Cable

Figure 2-39: Three Zones of System Model Complexity

Page 47
Chapter 3. CABLE MODELLING FOR POWER FREQUENCY STUDIES

3.1 Introduction
In general, modelling of underground cables depends on the type of phenomenon to be considered. Several cable
models have been developed for system simulations. All available models can be broadly divided into two main
classes: lumped parameter models (alternatively referred to as PI-section models) and distributed parameter
models (alternatively referred to as travelling wave models which are covered in Chapter 4). A list of available cable
models for power frequency studies along with their characteristics is shown in Table 3-1.
Models with lumped parameters do not account for an infinite number of incremental cable segments but they
consider the total length of the transmission cable as a single section. Thus the parameters of the PI circuit result
from the multiplication of the cable constants (R’, L’ and C’ in ohms per unit length) by the total length of the
transmission cable. This model is usually referred in the literature as nominal PI model or short cable model and it
provides good steady state accuracy at power frequency. For steady state analysis of power systems, i.e. at
planning stages referred to in Chapter 2, the only frequency component of interest is the nominal power frequency
and it is not necessary to account for full frequency dependence of cable parameters. Therefore lumped parameter
models can be used with sufficient accuracy for studies such as power flow, short circuit, dynamics etc. The term
Nominal PI-model refers to a PI circuit with R, L, C elements calculated at power frequency (50/60 Hz) or at any
other frequency respectively. For relatively long transmission lines, a modification is made to R, L, C lumped
parameters to reduce the error in approximating distributed elements to lumped elements. If the parameters are
modified to compensate long line effects (distributed nature), the PI circuit is called Equivalent PI-Model or Long
Line Corrected PI-Model. An example of corrected pi model application is frequency scans which will be discussed
fully in Chapter 4.
A simplified way of compensating long line effects is to cascade a number of Nominal PI-Models longitudinally. In
such a case total values of line RLC values should be distributed proportionally between each Nominal PI-Model.
Cascading of Nominal PI-Models is always a less accurate line representation in comparison to a single Equivalent
PI-Model and it increases the simulation time depending on the number of cascaded Nominal PI-Models.
This chapter discusses the preparation and use of data for the calculation of parameters for lumped cable models
to be used for system planning studies. It leads the reader through available data and their treatment in the
calculation of long cable parameters for an accurate power frequency computer models based on IEC 60287-1-1
[55] and other references. For the purposes of model validation, it compares IEC 60287-1-1 calculations with
parameters from line constants functions found in commercially available software. It should be noted that a
reverse procedure may also be used where parameters are calculated with a software tool first and then checked
against the same parameters manually calculated with the standard, or other verification method such as
comparison against a test sheet. The process of parameter verification is very important in order to have
confidence in cable model parameters.
The techniques described in this brochure can be applied to all power cable types including solid dielectric type
(XLPE), mass impregnated type (MI), self-contained fluid filled type (SCFF), or high-pressure fluid filled (HPFF)
pipe type cables. However it should be noted that this technical brochure deals largely with long cable installations
with most of installations falling into the XLPE or MI type installations.

Table 3-1: Available Transmission Line Models for Power Frequency Studies
LUMPED PARAMETER MODEL

Implementation Name Characteristics Frequency

Nominal PI-model R,L,C at a fixed nominal frequency 50/60

Equivalent PI-model
Aliases include Exact
R,L,C calculated at a fixed frequency, but
Equivalent PI-model and DC-kHz
corrected for long transmission lines [86][87]
Long Line Corrected PI-
model

Page 48
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

3.2 Cable Data

At system planning stage of any cable project the exact details of the cable type, configuration, and installation are
not generally known. Hence it is normal to utilise data from cable catalogues published by various manufacturers
for the purposes of preparing cable parameters for planning stage power system models.
Table 3-2 shows an excerpt from a number of manufacturer’s catalogues providing physical information on cable
construction for different cross sectional area (sizes) of a 400 kV cable. The core diameter, insulation thickness,
current ratings, inductance and capacitance are all presented. Empty entries indicate that the information was not
available in the catalogue for that particular manufacturer.

Table 3-2: Indicative Cable Data Collected from Three Manufacturer Catalogues
for a 400 kV Cable
Cross- Diameter Inductance Outside Cable Current
Insulation Capacitance Surge weight Rating
section of the diameter
Manufacturer thickness [mH/km] impedance
of core core [μF/km] of cable
[mm] [Ω] [kg/m] [trefoil,A]
[mm2] [mm] trefoil [mm]
A 30.2 32 0.13 0.42 42.2 111 15.7 780
630 B 0.13 0.44 53 109 17 853
C 30 33 0.119 0.46 118 17 594
A 33.7 30 0.14 0.40 38.7 112 17.1 870
800 B 0.15 0.41 48 110 18 954
C 34.2 32 0.134 0.44 118 18 636
A 37.9 29 0.16 0.39 34.5 114 19 960
1000 B 0.17 0.38 45 112 21 1120
C 38.1 29 0.150 0.41 118 20 671
A 44 27 0.18 0.38 30.6 119 20.8 1115
1200 B 0.19 0.38 43 116 24 1209
C 42 27 0.171 0.55 121 22 1001

Details of screen or sheath thicknesses are not available in the catalogue but are important quantities to have for
accurate development of zero sequence impedances of cable systems. The zero sequence impedance is less
important in the planning stages. If necessary, one can use zero sequence data from similar projects and accept
some uncertainty until cable construction and installation details are known.
It should be noted that these are nominal values in dimensions. Cables are engineered for a particular application,
delivery method, installation method, warranty requirements, and manufacturer process. All of these elements may
have an impact on the final delivered product. The core DC resistance should be taken directly as a constant from
IEC 60228 [55] for conductors complying with this standard or otherwise, from the manufacture data sheet to be
provided. It is important to note that the data presented in Table 3-2 are for illustrative purposes only and does not
match the Example model cable developed in this chapter.
As the cable project moves to a more advanced stage a specification is developed to match the requirements of
the application. When more detailed information is available, complex computer models can be developed that can
be used for a broader range of study phenomena. The data from Table 2-2 is derived from various specification
and data sheets for a 400 kV cable, similar to that in Section 3.3. The data provided can be very detailed and is
typically supplied by a cable manufacturer as part of an order.

Page 49
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

3.3 Example Model

For the purposes of developing and validating computer models for studies, an example model for a 36 km section
of the 400 kV cable with data shown in Table 2-2 of Chapter 2 is to be considered to demonstrate the calculation of
nominal frequency parameters R, L (or reactance X) and C (or susceptance B). It is assumed that the cable is
uniform without explicit cross-bonding sections. The continuously cross-bonded methods are assumed neglecting
the grounding resistance of 5 Ω and 0.02 mH series inductances. Section 3.4 discusses some of the practical
considerations in modelling typical HVAC cable system. In Section 3.5, the resistance, inductance and capacitance
calculations are demonstrated for the 400 kV example cable system. Finally, the results were compared against a
commercially available line constant solver program with detailed modelling together with a discussion of the
possible errors due to approximations in RXB calculations.
When setting up a model for system studies, the user should be aware that cable spacing has a direct impact on
circuit inductance while the cable depth impacts the cable thermal properties. Length of cable sections and sheath
bonding details are also needed to prepare accurate data for the calculation of cable parameters.
Figure 3-1 below shows the cable buried in a flat configuration, with Figure 3-2 providing details on the cable
bonding

Figure 3-1: Example Cable Burial Details

Figure 3-2: Typical Example of Cable Bonding Details

Page 50
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

3.4 Practical Considerations

It is important to recognise that there are some assumptions that need to be given due consideration when
attempting to calculate cable parameters for power frequency studies and cable impedance sequence modelling.
These include the application of semi-conductor layers in cables and how they influence parameters, the effects of
earth on the zero sequence path, and the effect of stranded versus segmental core construction on resistance.

3.4.1 Semi-conducting layers

In a typical high-voltage cable, two layers of semiconductor material surround the metallic core: a semiconducting
layer on the core’s outer surface and a semiconducting layer on the insulation’s outer surface. The function of the
semiconducting material is to equalise the electric fields around the conductor or core reducing the electrical stress
on the insulation material and is to prevent electrolytic corrosion of metallic armour layers. Due to higher resistivity
of semiconductor material, current tends to flow in the core but charge passes through the semiconductor. Thus the
semiconductor layer acts as a part of insulation system for the magnetic and current calculations whereas for the
calculation of capacitance, the semiconductor acts as a part of the core. To achieve this, the semiconductor layer
is treated as part of insulation with a modified permittivity value (Equation 3.1), so that the capacitance is correct
[52].

R 
ln b 
 Ra 
 mod_ins   inu (3.1)
r 
ln b 
 ra 
where rb, ra are inner and outer radii of insulation. Ra is the outer radius of the core and Rb is the sum of the outer
radius of the insulator and thickness of semi-conducting layers.

3.4.2 Screen or sheath considerations

Electrically, screens or sheaths form one terminal of a capacitor across the insulation layer from the conductor.
From the [8], it is a usual practice to convert wire screens into solid sheath so that appropriate parameters can be
entered into the line constants functions of commercially available software.
Taking into account the helical nature of the wires, the actual resistance of the bundle of wires is:

1
Rs  (3.2)
d2
g s .N . . w . cos 
4

 is the laying angle, N is the number of wires, gs is the wires’ conductivity and dw the wires’ diameter.
The inner and outer radii, respectively (r2) and (r3) of the equivalent sheath being those of the bundle of wires, the
cross-sectional area of the equivalent sheath is:


As'   . r32  r22  with r3  r2  d w (3.3)

To get the same resistance, a corrected value of the conductivity has to be used:

d2
N . . w . cos 
g s'  g s . 4 (3.4)
As'

Page 51
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

The DC resistance per unit length of the conductor at its nominal operating temperature θ is given by:

Rdc = Ro [1 + α20 (θ - 20)] (3.5)

where
Ro is the DC resistance of the screen at 20°C, Ω/m
α20 is the constant mass temperature coefficient at 20 °C per Kelvin (see Table 3-3 for standard values)
θ is the nominal operating temperature in degrees Celsius.

The following table summarises the typical resistivity and temperature coefficients for different conductor materials
according to IEC 60028, IEC 60889 and IEC 60287-1-1.

Table 3-3: Electrical Resistivity and Temperature Coefficients of Metals

Resistivity () Temp. coefficient
Material
at 20 °C [Ωm] (α20) at 20 °C [1/K]
Copper 1.7241e-8 3.93e-3
Conductors
Aluminium 2.8264e-8 4.03e-3

Copper 1.7241e-8 3.93e-3

Lead and lead alloy 21.4e-8 4.0e-3

Steel 13.8e-8 4.5e-3

Sheaths and armour
Bronze 3.50e-8 3.0e-3

Stainless steel 70.0e-8 Negligible

Aluminium 2.84e-8 4.03e-3

When the sheath consists also a metal (but not ferromagnetic) tape wounded around the cable core then the
current in the tape follows the tape. This means that the resistance of the tape shall be calculated according to the
actual length of the tape (Ltape),not according to the length of the cable (L), using the following relationship.

Ltape D
 1 ( )2 (3.6)
L L
where D is the diameter of the cable core around which the metal tape is wound.

This correction is not needed for laminated foil, which is not wound.

Ferromagnetic armours for 3-core cables:

In typical studies, the ferromagnetic effects of the armours are neglected. The magnetic properties of the armouring
metal may depend on the nature or composition, the thermal treatment and mechanical stresses to which it was
subjected during the manufacture and processing of the metal tape and during its application to the cable. The
calculations may be difficult due to the complicated magnetic circuit of the armour [60]. Alternatively, a field test

Page 52
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

(measurement) can be used to obtain the internal impedances of the armouring as function of the frequency and
the pre-magnetization current of the fundamental frequency. The present EMT type program cable models neglect
the ferromagnetic effects.

3.4.3 Effects of earth

The characteristics of earth mainly impact the zero sequence impedance of transmission lines and underground
cables, since the earth forms part of the zero sequence path. As the zero sequence impedance is important for
network short circuit calculations, and hence various system planning studies utilizing network equivalents, it is
important to have accurate earth effects in the modelling procedure.

The earth return resistance increases when the cable is near the surface of earth, and decreases as cable burial
depth increases, and finally at large depths the earth resistance reaches a constant value. The earth return
reactance decreases with increasing depth and eventually reaches a constant value. At relatively low values of
earth resistivity, the earth return resistance increases and then decreases with high values of earth resistivity. The
earth return reactance increases with earth resistivity.
For a simple three-phase cable system (having only three-conductors and no sheaths), the zero sequence
impedance increases with increasing earth resistivity as expected. However, in the presence of sheath, the
magnitude of zero sequence impedance may be decreased with earth resistivity due to the effect of parallel sheath
impedance in the zero sequence circuit. This can be explained using a per phase equivalent circuit of the 3-phase
cable system as shown in Figure 3-3.

Figure 3-3: Equivalent Circuit of a Three-phase System

The impedances Zp, Zq and Zr can be derived based on basic impedance components in a cable and it can be
shown that Zp and Zq are independent of earth resistivity. However, Zr consists of both self and mutual impedances
between cables. The resulting zero sequence impedance is
ZqZr
Z0  Z p  (3.7)
Zq  Zr
Since reactance of Zr increases significantly (compared to resistance) with increasing earth resistivity, the
magnitude of Z0 will reduce with increasing earth resistivity.

3.4.4 AC resistance of conductors

High voltage AC cables with large cross sections have special construction characteristics aimed at minimizing the
increase of the AC resistance due to skin and proximity effects. These construction characteristics include
conductor stranding and segmental conductors. In particular in the case of segmental conductors, commonly
referred as Milliken conductors, the conductor has slightly insulated segments to reduce skin effect and hence
allow for larger cross-sections.

Page 53
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

If the core is composed of segmented conductors, the resistivity of the core material is increased to take care of the
isolated segments [57]. Hence the effective resistivity is,

r2
 eff   (3.8)
Ac
where, ρeff is the effective resistivity of the conductor with radius r and Ac is the nominal cross sectional area of the
core. However if the manufacturer directly provides the resistance per unit length (Rdc), then the resistivity is
calculated as,

 eff  Rdc   r 2 (3.9)

The conductor AC resistance is essential in cable impedance sequence modelling for system studies, and is
normally not given either in data sheets or manufacturer catalogues. The AC resistance is calculated from the
conductor DC resistance, utilizing IEC 60287-1-1, which gives empirical correction factors to deal with these special
conductor types.

Table 3-4 shows the Rac/Rdc ratio of cores with large cross-sections and different construction types as calculated by
the cable constants [61] and IEC 60287-1-1. While the influence of core stranding does not seem to be pronounced,
the results for round stranded, hollow and helical stranded conductors are in good agreement. The reader can see
that the error in the AC resistance of segmental conductors is excessively high. At higher frequencies the error
becomes even higher. Therefore, it can be concluded that calculations based on cable constants overestimate the
AC resistance of the conductor resulting in an overestimation of the losses and the damping of the transmission
cable. In the case of segmental (Milliken) conductors, manufacturers normally rely on IEC 60287-1-1 for the
calculation of the AC conductor resistance.

Table 3-4: Rac/Rdc Ratio of Conductors of Different Materials and Construction Types under
Consideration of Skin and Proximity Effect
Cross- Rac/Rdc @ 50 Hz Rac/Rdc @ 50 Hz
Material, Type of core
Section IEC 60287-1 Cable Constants
630 mm2 AL, Round stranded 1.051 1.052
1200 mm2 AL, Hollow, helical stranded 1.103 1.104
1200 mm2 AL, Round, 4 segments 1.012 1.144
630 mm2 CU, Round stranded 1.127 1.131
1200 mm2 CU, Hollow, helical stranded 1.236 1.250
1200 mm2 CU, Round, segmental 1.076 1.327

Cable constants programs that rely on a general formulation with solid conductors, typically allow the user to adjust
conductor resistivity in the conducting mediums to approximate this affect.

3.4.5 Influence of temperature on skin effect

Skin effect influences the AC resistance of the conductor and should therefore be correctly accounted for in the
cable modelling. A skin effect factor is defined by the ratio Rac/Rdc.
The skin effect is the consequence of the depth of penetration of the electrical field in the conductive material. This
penetration depth is given by the following equation:

2
De  (3.10)


Page 54
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

In this equation  is the angular frequency,  is the permeability and  is the resistivity of the material. Lower
penetration depth is said to increase the skin effect by restricting current flow to a smaller cross-section. Since the
conductor metal’s conductivity varies with temperature the skin effect is temperature dependent. Higher
temperature increases  and the penetration depth. The formulas in IEC standard 60287-1-1 chapter 2.1.2 show
this dependence in another manner.
The temperature dependence can be illustrated with the plot (Figure 3-4) of the skin effect factor of segmented
conductors versus temperature. It can be observed that the skin effect decreases when temperature increases.

1.50

1.45

1.40

1.35

1.30
Rac / Rdc

1.25 STANDARD CONDUCTOR

1.20 ENAMELLED CONDUCTOR

1.15

1.10

1.05

1.00
20 30 40 50 60 70 80 90

CONDUCTOR TEMPERATURE ( °C )

Figure 3-4: Skin Effect of Segmented Conductor vs. Temperature

3.4.6 Influence of bonding and transposition

Long cables usually have cross-bonded sheaths to reduce induced currents on sheath and avoid the sheath
overvoltage at the same time. In such a grounding method the sheaths are transposed and grounded at points in
between. To control possible overvoltages, surge arresters are usually connected to the cross-bonding points.
Core transpositions can be performed to balance mutual coupling in order to reduce sheath currents. It will also
balance the series impedance of the phases and hence avoid an unbalanced system caused by the cable
installation. Untransposed transmission lines may require special attention regarding the protection settings.
A comprehensive description of various bonding arrangements, and their influence on the calculation of sequence
impedances are discussed in [8]. Proper modelling of bonding and transposition is important in the determination of
the cable current rating, as the bonding method influences the sheath losses.

3.4.7 Long transmission line correction

In some cases, for example harmonic analysis, it is possible to approximate the lumped parameters to distributed
parameters using long transmission line model. These formulas can be derived by combining the equations for
phase domain transmission lines (equations A1.5 in Annex 1) and nominal PI model with some assumptions and it
includes the different resonance points of the cable, due to its parameter distribution.

Page 55
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

z sinh  l
Z   Z C sinh  l  sinh  l  zl (3.11)
y l zy

l 
tanh 
Y 1 cosh  l  1 1 l  y   l  yl  2
  tanh   tanh   (3.12)
2 Z C sinh  l ZC  2 z  2  2  l  zy
 
2
where z and y are the per unit length impedance and admittance of the line respectively and l is the length of the
line.

z
ZC  Z (3.13)
y

is defined as the characteristic impedance of the line and

  zy (3.14)

is called the propagation constant.

It should be noted that for short lines where the value of l is small,
sinh  l   l sinh γl γl

and

  l  yl
tanh    (3.15)
 2  2

and therefore Z   z l Z ' zl and Y   y l Y ' yl suggests that the PI representation with lumped parameters are
accurate enough. Table 3-5 illustrates cable parameters as calculated in Section 3.5 with modified valued based on
the long line transmission correction for a 36 km cable at nominal frequency. These values will be used for the
basis of validation in Section 3.6.

Table 3-5: Cable Parameters Corrected for Long Line Model

Calculated Parameter Without correction With correction

Rpos [Ω] 0.5055 0.5024

Xpos [Ω] 7.5932 7.5700

B [S] 0.00243 0.0024

Rzero [Ω] 2.257 2.2496

Xzero [Ω] 2.2936 2.2870

3.5 Sequence Impedance Calculations according to Published Standards

When developing model parameters for a cable model for nominal frequency, practitioners can either use a
software tool that is readily available in commercial power system simulation packages, or develop the parameters
from calculations based on published technical material; in this case the Standard was the IEC Standards 60287-1-

Page 56
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

1. In all cases, it is a best practice to compare and verify results of one method against the other, or from a
measured data sheet. For this chapter, one has calculated the expected results from the IEC standard, and then
repeated the calculation in a software tool with the same parameters to verify the results.
All calculations in this section of the document conform to IEC 60287-1-1, IEC Technical Report 60909-2, CIGRE
Technical Brochures 283 and 347, and the underground transmission system Reference Book by EPRI and are all
referenced where appropriate.
Electrical parameters of cables published in manufacturers’ datasheet are typically calculated according to the
formulation given in IEC 60287-1-1 [55] applicable to steady-state operation of AC cables, where the calculated
parameters are valid at power frequency only. The standard assumes a continuous, constant current operation of
the cable (100% loading) to produce the maximum conductor temperature under stable thermal conditions. The
Technical Brochure 531 “Cable Systems Electrical Characteristics” [8] fully describes these formulas, use, and
application for different types of cable systems and is an excellent reference for the calculation of cable parameters.
IEC 60287-1-1 takes the following into account:
 Data related to the construction characteristics of the cable: layout (parallel single core cables and pipe-
type cables), geometry and properties of the conducting and insulating materials (conducting and insulating
layers).
 Data related to the surrounding conditions (in air, buried directly underground, in ducts, troughs or in steel
pipes).
 Skin and proximity effects for the calculation of AC resistance of conductors.
 Dielectric losses, screen and armour losses, losses in steel pipes, etc.

The following sections provide a detailed example in calculating cable parameters based on available cable data
from manufacturer’s specification. The example utilises the 400 kV cable data given in Table 2-2 in flat formation at
a burial depth of 1.5 m and a 500 mm cable spacing between phases, and goes on to calculate the lumped
parameters shown in Figure 3-5 according to IEC 60287-1-1.
All the equations are fully explained in [8].

Figure 3-5: Lumped Parameter Model

3.5.1 Calculation of AC resistance (R)

The DC resistance of the conductor of 400 kV cable can be calculated as,
l
R dc  (3.16)
A
where the parameters are:

ρ Resistivity of the solid copper conductor at 20 °C [Ωm] 1.724e-8

Cross-sectional area of the conductor
A 0.002516
(πr2, r = 28.3e-3 m) [m2]

Rdc = 6.896×10-6 [Ω/m] at 20 °C. As the highest loss occurs at operating temperature 90 °C, the DC resistance used
is

Page 57
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

R dc 90  R dc 20 (1   (  20 )) (3.17)

where the parameters are:

Rdc20 DC resistance of the conductor at 20 °C [Ω/m] 6.896 ×10-6

3.93×10-3
α Temperature coefficient of copper at 20 °C per kelvin
(from table 1 in [55])
θ Maximum operating temperature in degrees Celsius 90

Rdc = 8.737 ×10-6 [Ω/m] at 90°C

The AC resistance at 90 °C can be computed as [55],

Rac 90  Rdc 90 (1  y s  y p ) R _ R _ 1 Y Y Ω/M [Ω/m] (3.18)

It should be noted that the example shown above is based on 90°C but depending on the manufacturer
specifications, utility practices and geographical location other temperatures may be used.
The formulas for ys and yp are given in reference [55]. In calculating the AC resistance, it is assumed that the
conductor has a solid copper core. Hence the kp and ks factors are both unity (to calculate ys and yp) [62]. ys and yp
are described below and earth resistivity is assumed 100 Ω.m.

ys Skin effect factor 0.6066

Proximity effect factor
yp (In this example, the proximity factor is very small 0.00067
compared to the skin effect factor)

Rac = 14.04 ×10-6 [Ω/m] at 90°C

3.5.2 Calculation of positive sequence impedance

The equivalent per unit resistance of the earth can be found as


Re  (3.19)
8

ω Angular Frequency (=2πf, f=50 Hz) [rad/s] 314.1593

μ0 Permeability of non-magnetic material, or vacuum. 4π×10-7

Re = 4.9348×10-5 [Ω/m]

Depth of penetration for earth is

 earth
De  659 [m] (3.20)
f

Page 58
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

earth Resistivity of earth [Ω.m] 100

f Frequency [Hz] 50

De = 931.9667 [m]

The reactance of the cable conductor per unit length is,

  De 
Xa  ln   (3.21)
2  GMR a 

The geometric mean radius for a solid conductor can be shown to be equal to re-1/4 where r is the radius of the
conductor.

GMRa Geometric mean radius of core (= 0.779×r) [m] 0.0220

r Radius of the core [m] 0.0283

Permeability of the conductor (assume relative

μ 4π×10-7
permeability is unity)

Xa = 6.6928×10-4 [Ω/m]

The conductor self-impedance per unit length is given by,

Z a  R e  R ac  jX a (3.22)

Za = 6.3391×10-5 + j6.6928×10-4 [Ω/m]

The mutual impedance between two adjacent conductors in the flat configuration is given by

  D e 
Z c  Re  j ln   (3.23)
2  S 

Spacing between two adjacent cables in flat formulation

S 0.5
[m]

Zc = 4.9348×10-5 + j4.7315×10-4 [Ω/m]

The mutual impedance between outer conductors in flat configuration is given by

Page 59
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

  D e 
Z l  Re  j ln   (3.24)
2  2 S 

Zl = 4.9348×10-5 + j4.2960×10-4 [Ω/m]

The average mutual impedance (Zm,ave) between cables for transposed flat formulation,

2Z c  Z l
Z m ,ave  (3.25)
3

Zm,ave = 4.9348×10-5 + j4.5863×10-4 [Ω/m]

For cross-bonded cable, the positive sequence impedance is given by

Z pos  Z a  Z m,ave (3.26)

Zpos = 1.404296×10-5 + j2.10646×10-4 [Ω/m]

Since the length of the cable is 36 km, the total positive sequence impedance is,

Zpos = 0.5055 + j7.5832 [Ω]

3.5.3 Calculation of zero sequence impedance

The reactance (Xs) and the mutual impedance (Zm) between phase conductor and sheath are,

  De 
Xs  ln   (3.27)
2  rs 

Z m  R e  jX s (3.28)

Zm = 4.9348×10-5+j6.0711×10-4 [Ω/m]

rs Mean radius of the sheath [m] 0.0593

Self-resistance of sheath is given by,

Page 60
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

1
Rs  (3.29)
g sc  Ash

It is assumed that sheath consists of aluminium tape and the sheath resistivity is corrected using
sc90  sc20 (1   (  20)) (where   4.03  10 3 (Table 1 in [55]) and   90 o C ). The maximum operating
temperature for sheath and core may be different depending on various factors [55]. However in planning stage,
usually the maximum operating temperature of both conductor and sheath is assumed with a typical value such as
90 °C.

ρsc-20 Resistivity of the aluminium tape sheath at 90 °C [Ωm] 2.84x10-8

gsc Conductivity of the sheath at 90 °C [S/m] 2.7463×107

Ash Cross-sectional area of the sheath [m2] 7.4519×10-4

Rs = 4.8862×10-5 [Ω/m]
Self impedance of the sheath is given by,

Z e  R e  R s  jX s (3.30)

Ze = 9.8210×10-5+ j6.0710×10-4 [Ω/m]

In the average cross-bonded system, the sheaths of the three cables are transposed and connected to ground
continuously through the total length of the cable system. Assuming average cross-bonded cable, the zero-
sequence impedance is given by,

Z m  2 Z m,ave
Z zero  Z a  2 Z m, ave  ( Z m  2 Z m ,ave ) (3.31)
Z e  2 Z m,ave

Zzero = 6.2706×10-5 + j6.3712×10-5 [Ω/m]

Hence, the total zero-sequence impedance is,

Zzero = 2.257+ j2.2936 [Ω]

Typically zero sequence impedance is greater than the positive sequence impedance on normal overhead
transmission lines. However for underground or submarine cables, as shown in this example, positive sequence
impedance is often greater than zero sequence impedance. It has to be noted that the grounding resistances are
neglected here.

3.5.4 Calculation of capacitance (C)

The capacitance of the cable is calculated by,

Page 61
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

2 r  0
C [F/m] (3.32)
D
ln 
d

External diameter of the main insulation (including outer

D 116.6
semi-conducting layer) in mm
Diameter of the inner conductor (excluding first semi-
d 56.6
conducting layer)in mm
Relative permittivity of the insulation (assuming effective
εr 2.8
permittivity as 2.8)
ε0 Vacuum permittivity in F/m 8.8542×10-12

The capacitance and the corresponding total susceptance (B = ωC) are,

C = 2.152×10-4 [μF/m] or 7.7472 [μF] for a 36 km cable.

B = 0.00243 [S]

3.6 Verification of the Nominal Frequency Cable Model

The following table is a comparison of manual calculations (expected) against a software line constants solver tool.
With this software tool, provision for average cross-bonding was available which more closely resembles how the
IEC standard deals with the issue of cross-bonding.

Table 3-6: Comparison of Calculated Power Frequency Cable Parameters (Total Length)
with Average Cross-bonding
Manual Manual Software A Software A
Calculation Calculation (detailed/explicit (average cross- Software Software
(expected) (long line cross-bonding) bonding) B C
corrected)
Rpos [Ω] 0.5055 0.5024 0.4973 0.4995 0.5130 0.5036
Xpos [Ω] 7.5832 7.5700 7.311 7.4353 7.4795 7.4250
B [mS] 2.43 2.40 2.40 2.437 2.45 2.437
Rzero [Ω] 2.257 2.2496 2.2454 2.2513 2.2731 2.2561
Xzero [Ω] 2.2936 2.2870 2.2146 2.1377 2.1354 2.1225

The differences between parameters from software solver and the calculations can be explained. In calculation,
simplified formulas with approximations are used. For example, the depth of cable and actual cross-bonded
configuration are not considered in the manual calculation. However in simulation, the cable is modelled with actual
details such as 1.5 km cable segments with grounding resistances (5 Ω) at each major section. The sheath has an
additional 0.02 mH series inductance.
A check against the anticipated result of the manual calculation provides a form of verification and thus greater
confidence of study results. The results above show that the developed cable model, which includes input
parameters, has been verified at power frequency with reasonable agreement between manual calculations and a
software line constants solver.

Page 62
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

3.7 Construction of the Nominal Frequency Cable Model

0.5055 Ω 7.5832j Ω

1.215 mS 1.215 mS

Figure 3-6: Realization of Positive Sequence PI Model from Parameters

2.257 Ω 2.2936j Ω

1.215 mS 1.215 mS

Figure 3-7: Realization of Zero Sequence PI Model from Parameters

Most simulation software will take these parameters as presented, or they may require modification or data
conditioning depending on the input requirements of your simulation environment.

Page 63
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Chapter 4. CABLE MODELLING FOR TRANSIENT STUDIES

4.1 Introduction
Simulation studies of transients in cables are of great importance when analysing the behaviour of cables in power
systems. Transients in power systems can cause steep voltage wave fronts (travelling waves) that contain high
frequency components resulting in overvoltages at bus connections or reflections at open ends. For modelling a
cable for transient simulations, the frequency dependence of cable parameters must be taken into account. Precise
simulations of transients require a detailed cable model.
The major difficulty in modelling underground cables in power system electrical simulation software is that they are
frequency dependent, mainly due to the frequency dependency of conductors (skin effect), as well as the ground or
earth return path. In order to accurately represent a frequency-dependent transmission line when simulating with
time domain EMT-type solvers, the frequency dependent transfer function must be convolved into their equivalent
time-domain characteristics. The techniques required for this convolution are quite complex, and are not a part of
this technical brochure. This chapter is to guide the reader in appropriate selection and use of various cable models
for transient phenomena, and only provides references to cable models that have been developed by various
researchers. No derivations of cable model implementations are described.
Frequency dependent parameters need to be calculated with consideration to the construction of the cable, the
surrounding conditions, the skin effects, the dielectric losses, etc. It is therefore necessary to consider the
parameters in the frequency domain.
Many computer models utilise simplifications in their calculations, for example, assuming constant permeability,
neglecting dielectric losses and proximity effects, and assuming coaxial arrangements of the conducting and
insulation layers. Furthermore, an accurate model for the earth impedance is quite complicated and requires
specific data typically not available. Note that parameter values provided on data sheets by manufacturers are
usually only measured at power frequency and are useful for model benchmarks.
Checking of output parameters calculated by software over a wide frequency range is an important consideration
when determining the suitability of a model for the intended application domain. A table of suggested computer
models for specific applications is provided. Details on the mathematical formulations are provided in Annex 1.
A fundamental description of impedances and admittances has been documented in [61]. These formulations have
been widely used for calculating cable parameters for transient simulations and the model has been verified for
frequencies up to 1 MHz. The frequency range of interest for transient analysis is usually below this frequency.

4.2 Available Computer Models

Computer models form the basis for system technical performance studies. Computer models are comprised of two
basic areas:
1. data describing the cable installation: length, size, depth, ground material, conductor material, layer
information etc, and
2. a mathematical implementation of the physics based behaviour of the cable when exposed to some form of
electrical stimulus.
Computer models have evolved and became more sophisticated over time as the following table captures the
historical model name, aliases, and their characteristics.

Page 64
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Table 4-1: Available Transmission Line Cable Models for Transient Studies

DISTRIBUTED PARAMETER MODELS

Implementation Name Characteristics Frequency
One fixed frequency at
Dommel model which surge
Surge impedance and propagation velocity at a
Aliases include the impedance and
fixed frequency in modal domain [63]
Bergeron and K.C.Lee propagation velocity
are evaluated
Frequency dependent except for
Semlyen model DC-kHz
transformation matrix in modal domain [64]
Frequency dependent except for
J. Marti model DC-kHz
transformation matrix in modal domain [65]
Noda model Frequency dependent in phase domain [66] DC-kHz
Frequency dependent including transformation
L. Marti model DC-kHz
matrix in modal domain [67]
Gustavsen model
Alias: Universal Line Frequency dependent in phase domain [68] DC-kHz
Model

A number of terms are used that require some explanation to fully understand the implementation details of the
various models:
- Modal domain. The modal domain is a set of equations used to describe key features of a phenomena
based on physical characteristics of transmission paths. This is very similar in treatment to sequence
components commonly used in power systems theory. For a three phase overhead transmission line,
there are only three modes and they are equivalent to the sequence component definitions. For
underground or submarine cable systems, the modes are defined by the number of conducting elements
including cores and sheaths or screens.

- Curve fitting is another technology used in transmission model construction, and is applied to define an
approximated transfer function to represent a frequency response, or a time domain response.

- Convolution: the theoretical definition is the folding or blending of one function with another. However, in
transmission line construction, convolution is used to transfer a frequency domain response into an
equivalent time domain response.

- Phase domain. In the context of transmission line modelling, phase domain refers to the mathematical
treatment of defining frequency dependent transmission line transfer functions without the use of a modal
transformation, i.e. the transfer function equivalents are realised directly on a per conductor basis or phase
domain basis.

4.2.1 Bergeron transmission line model

The Bergeron transmission line model is a Norton equivalent model based on traveling wave solution for lossless
lines in time domain. This model takes account the distributed nature of transmission line parameters, such as
inductance and capacitance by its lossless Norton equivalent. The series resistance is treated as lumped element
and represents the losses of the line. The transmission line parameters are evaluated at a single frequency. The
model is suitable for modelling long cables and is usually recommended as the minimum model for a range of
studies since, as discussed in Section 4.3, the frequency of interest is confined due to the large charging capacity
and the long propagation time.
One benefit of using Bergeron models is that the computational burden (computer time and memory) is low when
compared to the frequency dependent models. The computational burden of the fully detailed Frequency
Dependent model has a quadratic relationship to the complexity of the cable model (number of conductors), and

Page 65
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

the order of the curve-fitted transfer function. The Bergeron model computational burden has a linear relationship to
the number of conductors.

Figure 4-1: Bergeron Model for a Transmission Cable [A4]

4.2.2 Modal domain models

Semlyen Model
Another model, developed by Semlyen, constructs a model by approximating some transmission line parameters
(e.g. the characteristic admittance and the propagation constant) of each mode through two exponentials [64]. A
limitation of this model is that a constant transformation matrix is used to convert phase matrices into modes. In
case of underground cables and vertically asymmetrical overhead lines, the transformation matrix is frequency
dependent.

J. Marti transmission line model

Both the frequency dependence and distributed nature of the parameters are taken into account in this frequency
dependent model [65]. This model is more accurate than the Semlyen model as the frequency domain transfer
function is curve-fitted directly in the frequency domain as opposed to Semlyen method of curve fitting in the time
domain. The J. Marti model approximates the transmission line parameters by rational functions which is simply a
ratio of two polynomial functions as shown in equation 4.1.

⋯
y (4.1)
⋯

The application of rational functions has many advantages. For example, rational functions offer a generalised way
to represent curves plus the flexibility to fit various shapes. One of the limitations of this model is that it uses a
constant transformation matrix to convert from modal domain to phase domain.

L. Marti transmission line model

This model takes into account both the frequency dependence of the cable parameters, distributed nature and the
frequency dependence of the modal transformation matrix [67]. This model can accurately simulate low and high
frequency transients. A limitation in this model is that for some cable or overhead line configurations, there can be
difficulties associated with curve fitting the transformation matrix. The L. Marti model also approximates the
transmission line parameters by rational functions.

4.2.3 Phase domain models

Noda model
This model is called ARMA model (Auto-Regressive Moving Average model) and the transmission line parameters
are approximated by Z domain functions based on Z-transformation (hence it is called a discrete model) [66]. The
model is time step dependent and if the time step of the simulation changes, the model must be reconstructed. This
model is formulated directly in phase domain.

Page 66
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Gustavsen Line model

This frequency dependent model (also known as the universal line model) was developed by B. Gustavsen et al.
[68]. It is directly formulated in the phase domain and can be used to model both underground cables and
overhead lines. The frequency dependent behaviour and distributed nature of parameters are taken into account.
This model can also be used in situations where the transformation matrix is highly frequency dependent as in the
case of underground cables and unsymmetrical overhead lines. The transmission line parameters are
approximated by rational functions.

4.3 Selection of Cable Models fo r Various Types of Studies

As stated in the brochure, there are three main types of studies: planning, system impact and equipment design
studies. The planning studies all focus on power frequency while both system impact and equipment design studies
have a larger variety of frequencies. This section discusses the selection of cable models applicable for various
types of studies and considerations in their usage.

4.3.1 Bergeron model

As Bergeron models produce conservative results in transient studies, many practitioners start with Bergeron
models due to its numerical stability and lower computational burden. When the results exceed the withstand
voltages or other equipment ratings, detailed studies can be performed with frequency dependent models due to
higher accuracy.
In the analysis of many or long cable systems, transient components of the overvoltages do not include high
frequency components. The frequency of interest is normally limited to around 1000 Hz due to large charging
capacity of cables. Frequency dependencies of cable impedances are not significant within this range, for typical
physical and electrical parameters of cables and burial layouts, supporting the selection of Bergeron models as a
starting point.
When using Bergeron models, the choice of the target frequency, at which cable impedances are calculated, is of
critical importance. It is one common practice to use a power frequency as the target frequency since the lower
target frequency leads to conservative results due to the weaker damping. However, different target frequencies
may be more appropriate and the following analysis introduces a simple method in choosing the target frequency.
Figures 4-1 and 4-2 show frequency dependencies of modal propagation velocities and modal characteristic
impedances (real part) of the example long-cable introduced in Section 2.3.2. The modal propagation velocities
and modal characteristic impedances at different frequencies were found with the use of a Cable Constants
program. The frequency dependencies in the figures demonstrate that the analysis with Bergeron models will
contain minor errors when cable impedances are calculated at 200 Hz. The modal propagation velocities and
modal characteristic impedances at 200 Hz and 1000 Hz are different only by around 5 %, but it is recommended
to calculate cable impedances at a lower frequency in order to produce conservative results with lower resistance.
200
Modal propagation velocity [m/μs]

150

Mode 1
Mode 2
100
Mode 3
Mode 4-6

50

0
0 500 1000 1500 2000
Frequency [Hz]

Figure 4-2: Frequency Dependence of Modal Propagation Velocities for the Long-cable
Example

Page 67
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

25

Modal characteristic impedance [Ω] 20

Mode 1
15 Mode 2
Mode 3
Mode 4
10 Mode 5
Mode 6

0
0 500 1000 1500 2000
Frequency [Hz]

Figure 4-3: Frequency Dependence of Modal Characteristic Impedances (Real Part) for the
Long-cable Example

The magnetic field produced by the core current cannot be shielded by the metallic sheath if the transient
frequencies contained in the overvoltages are lower than the characteristic frequency fs calculated by (4.2) [69]. On
the other hand, the magnetic field produced by the core current is completely shielded by the metallic sheath
(coaxial model) if the transient frequencies contained in the overvoltages are higher than the critical frequency fc
calculated by (4.3) [70].

 s3/ 4
fs  (4.2)
10 s ds
s
fc  (4.3)
 s d 2
where  s : resistivity of metallic sheath [Ωm]

 s : permeability of metallic sheath [H/m]

d : thickness of metallic sheath [m]
s : separation between cables [m]

The characteristic frequency and critical frequency are calculated as 55 Hz and 1798 Hz, respectively, for the
example long-cable. It is recommended to select the target frequency at which cable impedances are calculated,
with the following considerations:
- Since the frequency of interest is between the characteristic frequency and the critical frequency in most
analyses of many or long cable systems, the target frequency should be selected between the two
frequencies.

- The target frequency should be selected at the lower end of the two frequencies in order to produce
conservative results with lower damping.

- The cable impedances and other related values are in the transition near the two frequencies. Hence the
target frequency should have some offset from the two frequencies.

Page 68
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

The target frequency 200 Hz, as proposed above, can be selected as a result of these considerations. When
selecting a target frequency which is different from the power frequency, one should confirm that it does not lead to
an unrealistic steady-state condition.

4.3.2 Recommended cable models

Table 4-2 recommends the level of cable modelling requirement as a minimum for various types of studies. The
types of studies correspond to the issues listed in Table 2-1. In most cases, a Bergeron model will produce reliably
conservative results. In some cases, a frequency-dependent (FD) or segmented model is recommended for
improved accuracy. Segmented model in this context refers to explicit modelling of cross-bonded sections. In Table
4-2, FD refers to frequency dependence in the cable resistance, inductance, and modal transformation matrix (if
used), not just frequency dependence of the reactance and susceptance. In cases where the cable capacitance
dominates its response, the modelling effort is often better spent on representing the frequency dependence of
other system components. On the other hand, in a forensic study, or any other study in which it is important to
match cable transient measurement data, use of frequency-dependent and segmented cable models is
recommended to the extent where possible.
To briefly summarise Table 4-2, Bergeron and PI models are adequate for most issues, except those involving
sheath overvoltages and lightning protection. Some issues mention that FD or segmented models are “moderately
important” for that issue. This means that more detailed modelling of other system components should be
considered first for those issues since the detailed modelling of cables is not likely to have a major impact on
system design or cost.

Table 4-2: Recommended Cable Models for Various Types of Studies

Importance of Detailed Models
Minimum Recommended Cable (Segments, Bonding, Grounding
Issue
Model
Frequency-Dependent (FD))
Phase unbalance and Bergeron (Unbalanced) Segments are important.
transposition Equivalent PI FD is not important.
Segments are not important.
Bergeron FD is moderately important; cable
Harmonic frequency scans
Equivalent PI parameters may be re-calculated for each
frequency of interest if necessary.
TOV – Ground faults and Minimal; damping from other system
Bergeron
clearing components is more important.
Minimal; damping from other system
TOV – Resonance Bergeron
components is more important.
TOV – System islanding and Minimal; damping from other system
Bergeron
load rejection components is more important.
Minimal; damping from other system
Ferroresonance Bergeron
components is more important.
Segments are not important.
Bergeron FD is moderately important; cable
Harmonic sources and flows
Equivalent PI parameters may be re-calculated for each
frequency of interest.
Segments are not important.
Cable energization Bergeron (at Target Frequency) FD is moderately important; this is usually
the first case in which to apply an FD model.
Segments are not important.
FD is moderately important, but power-
Auto-reclosing overvoltages Bergeron
frequency solution should not be
compromised.
Discharge of the cable Bergeron Minimal.
Sheath overvoltages and Segments are important.
Bergeron (Unbalanced)
protection FD is moderately important.
DC offset / zero-miss effect Bergeron Minimal.
Current transformer (CT)
Bergeron Minimal.
saturation

Page 69
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Importance of Detailed Models

Minimum Recommended Cable (Segments, Bonding, Grounding
Issue
Model
Frequency-Dependent (FD))
Capacitive current
Bergeron Minimal.
interruption
Bergeron (Unbalanced, at Segments are important.
Lightning overvoltage
Target Frequency) FD is moderately important.
Segments are not important.
Shunt reactor restrike Bergeron
FD is moderately important.
Inductive coordination and Bergeron (Unbalanced) Segments are important.
EMI Equivalent PI FD is not important.

Figure 4-4 shows the results of an energization overvoltage analysis with different cable models in different EMT-
type simulation software packages. The cable from Sub B to Sub C in the long-cable example of Chapter 2 is
energized from a voltage source behind the source impedance of 1 Ω and 30 mH. The energization occurs at the
voltage peak of phase a, and the voltage (phase a) at the open terminal is monitored.
From the results, the Bergeron model evaluated at 50 Hz produces a similar overvoltage to the Gustavsen model in
Softwares B and C during the first 20 ms after the energization. In finding the maximum overvoltage, the choice of
cable models between the Bergeron model and FD models does not have a noticeable impact. On the other hand,
the overvoltages after the first 20 ms following energization have notable differences, which require the use of FD
models when matching simulation results to field measurements.

600

400

200
Voltage [kV]

-200

-400

-600

-800
0 0.01 0.02 0.03 0.04 0.05 0.06
Time [s]

Bergeron 50Hz (Software A) Gustavsen Model (Software B) Gustavsen Model (Software C)

Figure 4-4: Comparison of Energization Overvoltages with Different Cable Models

4.3.3 Guideline for setting up phase-domain FD models

During transient studies one may encounter unstable simulations which may give erroneous or nonsensical results.
When frequency-dependent phase-domain models are used, unstable simulations are often caused by passivity
violations of the cable model which is common to any phase-domain model developed in literature.
The following describes the suggested steps to resolve unstable simulations due to cable modelling errors in order
of importance:

Page 70
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

1. Verify the input fitting parameters (see Table 4‐3)

2. Check fitting results
3. Reduce the curve fitting frequency – lower limit (e.g. 0.1, 0.01,0.001 Hz)
4. Increase the Maximum Order of fitting

Table 4-3 shows guideline to select proper fitting parameters for frequency-dependent models.

Table 4-3: Guideline to Select Proper Fitting Parameters

Possible consequences as a results of significant deviation from
Typical the typical values
Parameter
Values
An increase A decrease
(e.g. 0.01 or 0.001 Hz)
- May improve stability of the
(e.g. 10 Hz) simulation
1 Hz - Fitting may be only (A very low value, e.g. 1e-4 Hz)
0.5 Hz accurate between 10 Hz to
Curve fitting frequency - May result in unstable simulation
the upper limit
– lower limit 0.1 Hz (due to over-fitting)
- Possibility of unstable
0.01 Hz simulation (due to under- - Simulation may be slow (require a
fitting) higher order function for accurate
fitting at low frequency)

Curve fitting frequency (e.g. 10MHz)

(e.g. 10kHz)
– upper limit - Transmission line models
- inaccurate simulation at high
(it is recommended to are only accurate and tested
1 MHz frequencies
keep the default value, up to 1 MHz
1 MHz, see Section - possibility of unstable simulation (due
- Possibility of unstable
4.8…) to inaccuracy outside the fitting range)
simulation
Total number of (e.g. 500)
frequency increments
- increase accuracy
(total number of 100
samples for curve- (Usually do not need to
fitting) change this value)
Maximum order of
fitting for propagation When plotted as a function of frequency, the angle of the propagation
function (Ap and Q) function shows an oscillatory behaviour (due to the transportation
(It is sometimes 20-30 delay of the transmission line). Hence a higher order approximated
required to increase function may be required to obtain an accurate fitting (compared with
this value to obtain Ysurge).
accurate solution)
Maximum order of
fitting for Ysurge Compared with propagation function, curve-fitting of Yc is relatively
easy and hence accurate fitting may be achieved less than 20 poles
(In many instances, the 20-30 for many transmission line configurations.
maximum order, 20 is
sufficient for accurate These values may be increased based on fitting results.
results)

Page 71
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

4.4 Calculation of Electrical Parameters of a Cable

4.4.1 General formulation

Reference [61] describes the general formulation of impedances and admittances of single-core coaxial cables
consisting of a core, sheath and armour, pipe-type cables with finite pipe thickness and overhead cables. These
fundamental generalizations form the basis of all cable models in use today. An in-depth discussion of cable model
formulations is beyond the scope of this Technical Brochure and important references to prior work are noted.
The calculation focuses on the transient behaviour of the cable system, a challenging modelling domain, and hence
the determination of its frequency-dependent electrical parameters defining the electromagnetic coupling among
the conducting layers of the cable system. Results have been verified to be correct up to approximately 1 MHz,
while in most of the cases, the frequency of interest in the analysis of transients on a cable system is less than
1 MHz.
This calculation method should therefore be the preferable approach for the calculation of the electrical parameters
of a cable for electromagnetic transient simulations, where a wide range of frequencies other than the power
frequency may be involved. Most EMT-type simulation software supports the calculation of the parameters based
on this formulation.
The calculation method assumes:
 Coaxial arrangement of the conducting and insulating layers of the single-core cable. This is normally
the case in HV cables but may not be the case in MV or LV cables (e.g. compact cables).
 Dielectric losses are not considered, as they are assumed to be small in comparison with the losses in
the conducting layers of the cable and the earth.
 Each conducting layer has a constant permeability; hence no saturation effects are taken into account.
 Finite pipe thickness (pipe type cable) with earth return, so that the calculation does not assume a
complete shielding of the pipe.
 No correction factor for temperature dependency of the AC and DC resistance.
 No proximity effects are considered. Work done by C. H. Chien and R. Bucknall [71] shows that
proximity effect is significant only for three-core cables.

4.4.2 How to deal with non-standard configurations

Non-standard configurations are those that employ arbitrary geometries such as sectored 4 quadrant cables,
umbilical or elliptical conductor designs. Work by Gustavsen et. al [72] used a 2-D finite-element method for
calculating electrical parameters of cables. This approach can be applied to non-standard configurations. The
proximity, skin effects and twisted elements and layers are accurately taken in to account with these tools and the
parameters have been validated with measurements up to 50 kHz.
Work by Ametani and Fuse [73] develops impedance formula of a conductor with an arbitrary cross section. This
reference discusses calculation of the cable parameters for arbitrary shaped cables (e.g. sector-shaped cables) by
approximating the arbitrary shaped conductor using an equivalent cylindrical conductor. Once approximated,
conventional formulas can be directly used to calculate cable parameters. The proposed method is accurate
enough for many power system studies including surge analysis and compared with finite element based
approaches, this method requires less computational effort and memory. The cable parameters from the
approximate method for several sector–shaped cable examples have been compared with field measurements and
finite element method. These methods discussed here usually do not form the part of typical commercial EMT-type
software tools and requires expert knowledge to apply.

4.5 Installation and Environmental Effects on Long Cable Models

The physical environment and installation methods all have an impact on the model of a long land or submarine
cable. This section describes some common considerations to be factored into cable models in order to achieve
the desired accuracy.

4.5.1 Bonding of cable screens

Page 72
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Since cross-bonding affects cable impedance and propagation velocity, it is highly recommended to model sheath
cross-bonding for switching studies. As the computational burden required by Bergeron models is relatively small,
all cross-bonding points can be modelled without a noticeable effect on computational efficiency. With some
frequency dependent models, however, one may wish to reduce cross-bonding points in order to reduce
computational burden and possible numerical stability issues. The reduction, to some extent, can be made without
a significant impact on the simulation results [74][75].
Bonding of cable sheath is an important factor for the cable models, also when assessing the sheath overvoltage in
transient studies. The sheath overvoltage is generally higher in single point bonded cables.

4.5.2 Installation and layout of cables

Cable installation and layout refer to where and how individual cables are position with respect to one another, for
example trefoil or flat configurations. Cable installed underground, in water, in a tunnel or pipe also has an
influence. These are important input parameters for modelling purposes. It should be noted that this information is
only available after installation, or as part of an installation specification.
Some specific considerations need to be taken into account when modelling tunnel installed cables as described in
reference [76]. The distance between the cable and the tunnel wall as well as physical and electrical parameters of
the tunnel are important factors for modelling purposes. Similar modelling considerations apply to pipe type cables
as well. Some commercially available software tools have pipe type cable models. Simplifications provided in [76]
could be utilised to model tunnel-installed cables with an equivalent pipe type model. Modelling of pipes or tunnels
is important when studying the surface voltage of gas insulated transmission line or bus which is above ground [56].

4.5.3 Effects of temperature

Effects of conductor or sheath temperature are often ignored in the analysis even though they can affect the
damping of an overvoltage. Generally, transient analysis is performed assuming low temperature (often ambient
temperature), which should yield a result with less damping. This is in contrast to the calculations performed in
Chapter 3 which uses 90 °C for planning studies. Losses and the damping are, in many cases, accurately modelled
only to reproduce measured results in simulations.

4.5.4 Effects of earth resistivity

The contributions of the earth path to underground cable parameters are significant. The earth impedances can
comprise up to about 90% of the total self-impedance values of individual cables. In contrast to overhead
transmission lines where the mutual impedances between phase conductors depends on both the air paths and the
earth paths, the mutual impedances of underground cables are contributed purely by the earth paths [77].
Resistance of the self and mutual earth impedances increases with frequency, whereas inductance decreases with
frequency. The mutual earth impedance depends on the configuration of cable layout (e.g. trefoil or flat) and
decreases with increasing distance between cables and depth of the cable.
Effects of stratified earth can be considered in Cable Constants but are typically ignored due to unavailability of
data. They can have a significant effect on high (or very high) frequency earth-return impedance [78][79]. The high
frequency effect of earth on models is shown in [80].
Modelling of ground resistance at normal joints and substations has an impact on transient studies since it adds to
earth-return impedance. This ground resistance needs to be explicitly modelled outside of the cable and earth
model.

4.6 Example of Sensitivity Analysis of Input Parameters on Study Results

Input parameters of a cable model are resistivity of the conductors (including sheath), permittivity of the insulations,
thickness of semi-conducting layers, geometric data such as conductor radius, thickness of layers. These
parameters can be obtained from the manufacturer data sheets and more importantly, the supplied data from the
manufacturers may differ from the actual values. The specification sheet lists contractual limits (minimum or

Page 73
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

maximum tolerances) and the delivered cable may be different. This will have an effect on the accuracy of the
cable models and care must be taken in the use of cable parameter data.
A sensitivity analysis in this section highlights which parameters of the cable are more significant or have more
impact to the accuracy of the cable model for different frequency ranges. This helps simulation software users to
focus more attention to the accuracy of those parameters for various power system studies.
The cable data is same as that of example cable discussed in chapter 2 (Table 2-2) and the cable configuration
used for this analysis is shown in Figure 4-5. The trefoil formation is selected, instead of the flat formation, since
sequence impedances are defined on a balanced three-phase circuit.

Figure 4-5: Cable Configuration for Sensitivity Analysis

In order to determine the effect of cable input parameters for different studies, current and voltage waveforms from
time domain simulations involving typical power system studies can be used. However, it is difficult to conclude the
sensitivity of a particular parameter for the entire frequency range for each type of transients. For example, the
switching transients typically contain frequencies up to 10 kHz, but a particular time domain switching study may
generate transients only up to 1 kHz. A better approach therefore will be to determine the sensitivity of the input
cable parameter in frequency domain using frequency domain transmission line functions.
Many power system studies such as short circuit and fault analysis, dynamic simulations with fault/unbalanced load
shedding etc. involve the analysis using sequence components (positive and zero sequence impedances etc.). So
it is useful for the simulation engineer to know the impact of cable input parameters on sequence components. The
switching studies and lightning studies may not involve direct sequence components analysis and the simulation
depends on phase domain quantities such as phase impedances (relates to phase voltages and currents). Since
sequence impedances are components of phase impedances, a high impact of a parameter on either positive or
zero sequence can cause significant impact on such a study.
In this sensitivity analysis, the sensitivity of a parameter was determined in frequency domain using sequence
impedances and admittances of the cable. This requires a derivation of the relationship between such functions
and the parameter itself. When one parameter is changed (e.g. conductor radius), other parameters are assumed
constant (e.g. thickness of the other layers such as main insulation, sheath, semi-conducting layer etc.)
For a 10% change in the parameter (except for earth resistivity and semi-conducting layer where a 50% change
was applied), the change in the cable model sequence impedances was observed for different frequency ranges,
and checked for any significant impact. It is assumed that the sensitivity of input parameters is classified as high if it
is above 10% for a particular frequency range. This means that a small error in the input parameter may cause a
relatively high voltage or current magnitude change, when exciting the cable with a frequency belonging to that
frequency range.
The cable model sensitivity analysis has identified a set of most sensitive parameters that include conductor radius,
permittivity of inner insulator layer and thickness of inner insulator among others. The following tables summarise
the results with three sensitivity ranges, below 5%, 5% - 10% and above 10%.

Page 74
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

-3
x 10 Resistance of Positive Sequence Impedance [Ohms]
2.5

Default
2 Conductor radius
Conductor Resistivity
1.5 Sheath Resistivity
Earth resistivity
Thickness of the inner insulator layer
1
Thickness of the sheath

0.5

0
-3 -2 -1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10

Difference as a percentage
20

15

10

0
-3 -2 -1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10
Frequency [Hz]

Figure 4-6: Positive Sequence Resistance as a Function of Frequency

Table 4-4: Impact on Positive Sequence Resistance by Frequency Range

IMPACT ON POSITIVE SEQUENCE RESISTANCE

PARAMETER DC/low frequency Power frequency Switching studies Lightning
studies studies (50/60Hz to studies
(0 to 50/60Hz) (50/60Hz) 10kHz) (up to 1MHz)
Conductor radius
above 10% 5 to 10% 5 to 10% 5 to 10%
(+10%)
Conductor Resistivity
10% below 5% below 5% below 5%
(+10%)
Sheath Resistivity
below 5% below 5% 5 to 10% below 5%
(+10%)
Earth resistivity
below 5% below 5% below 5% below 5%
(+50%)
Thickness of the inner
below 5% below 5% below 5% below 5%
insulator layer (+10%)
Thickness of the sheath
below 5% below 5% 5 to 10% below 5%
(+10%)

Page 75
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Reactance of Positive Sequence Impedance [Ohms]

1
Default
0.8 Conductor radius
Conductor Resistivity
0.6 Sheath Resistivity
Earth resistivity
0.4 Thickness of the inner insulator layer
Thickness of the sheath

0.2

0
-3 -2 -1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10

Difference as a percentage
7

0
-3 -2 -1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10
Frequency [Hz]

Figure 4-7: Positive Sequence Reactance as a Function of Frequency

Table 4-5: Impact on Positive Sequence Reactance by Frequency Range

IMPACT ON POSITIVE SEQUENCE REACTANCE

PARAMETER DC/low frequency Power frequency Switching studies Lightning
studies studies (50/60Hz to studies
(0 to 50/60Hz) (50/60Hz) 10kHz) (up to 1MHz)
Conductor radius
below 5% 5 to 10% 5 to 10% 5 to 10%
(10%)
Conductor Resistivity
below 5% below 5% below 5% below 5%
(10%)
Sheath Resistivity
below 5% below 5% below 5% below 5%
(10%)
Earth resistivity
below 5% below 5% below 5% below 5%
(50%)
Thickness of the inner
below 5% below 5% 5 to 10% 5 to 10%
insulator layer (10%)
Thickness of the sheath
below 5% below 5% below 5% below 5%
(10%)

Page 76
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

-3
x 10 Resistance of Zero Sequence Impedance [Ohms]
2.5

Default
2
Conductor radius
Conductor Resistivity
1.5 Sheath Resistivity
Earth resistivity
1 Thickness of the inner insulator layer
Thickness of the sheath

0.5

0
-3 -2 -1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10

Difference as a percentage
20

15

10

0
-3 -2 -1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10
Frequency [Hz]

Figure 4-8: Zero Sequence Resistance as a Function of Frequency

Table 4-6: Impact on Zero Sequence Resistance by Frequency Range

IMPACT ON ZERO SEQUENCE RESISTANCE

PARAMETER DC/low frequency Power frequency Switching studies Lightning
studies studies (50/60Hz to studies
(0 to 50/60Hz) (50/60Hz) 10kHz) (up to 1MHz)
Conductor radius
above 10% 5 to 10% 5 to 10% 5 to 10%
(10%)
Conductor Resistivity
10% below 5% below 5% below 5%
(10%)
Sheath Resistivity
below 5% 5 to 10% 5 to 10% below 5%
(10%)
Earth resistivity
below 5% below 5% below 5% below 5%
(50%)
Thickness of the inner
below 5% below 5% below 5% below 5%
insulator layer (10%)
Thickness of the sheath
below 5% 5 to 10% 5 to 10% below 5%
(10%)

Page 77
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Reactance of Zero Sequence Impedance [Ohms]

1

Default
0.8 Conductor radius
Conductor Resistivity
0.6 Sheath Resistivity
Earth resistivity
0.4 Thickness of the inner insulator layer
Thickness of the sheath

0.2

0
-3 -2 -1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10

Difference as a percentage
14

12

10

0
-3 -2 -1 0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10
Frequency [Hz]

Figure 4-9: Zero Sequence Reactance as a Function of Frequency

Table 4-7: Impact on Zero Sequence Reactance by Frequency Range

IMPACT ON ZERO SEQUENCE REACTANCE

PARAMETER
DC/low frequency Power frequency Switching studies Lightning
studies studies (50/60Hz to studies
(0 to 50/60Hz) (50/60Hz) 10kHz) (up to 1MHz)
Conductor radius
below 5% 5 to 10% 5 to 10% 5 to 10%
(10%)
Conductor Resistivity
below 5% below 5% below 5% below 5%
(10%)
Sheath Resistivity
below 5% below 5% below 5% below 5%
(10%)
Earth resistivity
below 5% below 5% below 5% below 5%
(50%)
Thickness of the inner
below 5% 5 to 10% 5 to 10% 5 to 10%
insulator layer (10%)
Thickness of the sheath
below 5% below 5% below 5% below 5%
(10%)

Page 78
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Table 4-8: Impact on Capacitance

IMPACT ON CABLE CAPACITANCE

Parameter Sensitivity

Conductor radius (10%) 5 to 10%

Permittivity of inner insulator (+10%) 10%

Thickness of the inner insulator layer (10%) 5 to 10%

Thickness of inner semi-conducting layer

below 5%
(+50%)

The sensitivity study results presented above suggest that accuracy of the conductor radius is relatively important
for DC or low frequency studies and less important for other studies. The accuracy of conductor resistivity may
have relatively large effect on DC or low frequency studies. The sheath resistivity may have medium effect on
power and switching frequency studies. The accuracy of the inner insulation layer thickness has a medium effect
on all studies except DC or low frequency studies, while the thickness of the sheath has a medium effect on power
and switching frequency studies. The earth resistivity and thickness of inner semi-conducting layer may not cause
considerable impact on all studies (see Section 3.43. for further discussion on effect of earth). The permittivity
value shows a linear relationship with capacitance (as expected from the theory) and the effect is relatively high.
The conductor radius and thickness of inner insulation layer have a medium effect on the capacitance.
It must be underlined that the above observations are based on the relative importance of input parameters on
sequence impedances and admittances for this particular case, and actual effects on voltage and current values
obtained when performing simulations can vary form study to study. For example, Figure 4-10 shows the sensitivity
of the same parameters on cable energization overvoltages. The cable from Sub B to Sub C in the long-cable
example of Chapter 2 is energized from a voltage source behind the source impedance of 1 Ω and 30 mH. The
energization occurs at the voltage peak of phase a, and the voltage (phase a) at the open terminal is monitored.
From the results, it is clear that 10 % or 50 % deviations in conductor, sheath, or earth resistivity do not have
noticeable impact on the energization overvoltages. The deviations in resistivity can affect the damping of the
overvoltages in theory, but the effect cannot be observed in the simulation time 60 ms. The other two parameters
have rather clear impact on the energization overvoltages. The increase of conductor radius leads to lower
reactance and higher capacitance. As the effect on capacitance is more significant than that on reactance, the
results show lower propagation velocity due to the increase of conductor radius. In contrast, the increase in the
thickness of the inner insulator layer leads to lower capacitance, which results in higher propagation velocity as
shown in Figure 4-10. The simulation results demonstrate the importance of modelling the capacitance accurately
for cable energization studies.

Page 79
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

800

600

400

200 Base
Voltage [kV]

Conductor radius
0
Conductor resistivity
-200 Sheath resistivity

-400

-600

-800
0 0.01 0.02 0.03 0.04 0.05 0.06
Time [s]

800

600

400 Base

200
Voltage [kV]

Earth resistivity
0
Thickness of the
-200 inner insulator layer
Thickness of the
-400 sheath
-600

-800
0 0.01 0.02 0.03 0.04 0.05 0.06
Time [s]

Figure 4-10: Comparison of Energization Overvoltages with Different Cable Parameters

4.7 Verification of Line Constant Solver Results against Manual Calculations

Frequency dependent cable models built for transient studies can be verified at power frequency as a spot check of
input parameters. For verification, it is recommended to go through this procedure before performing transient
cable model validation in Section 4.8.
R, X and B results generated from the line constants solver [61] available in commercial software packages can be
compared with parameter values calculated in Chapter 3. Table 3-6 compares the calculated parameter values
against several commercially available line and cable constant solver programs. In general R, X and B values
calculated for flat formation with the line constant show very good agreement against the expected hand
calculations at power frequency. Thus, users will have a high level of confidence in suitability of the parameters and
computer models.
Planning model parameters from the left-hand column of Table 3-6 may be used to derive Bergeron model
parameters for transient studies. First, one divides Rpos and Rzero by the cable length, 36 km, to derive resistances

Page 80
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

per meter. Second, the positive sequence (or line mode) surge impedance and wave velocity are determined from
(4.4) and (4.5).

(4.4)

1 1
2 (4.5)

Where f is the system power frequency, Xpos is in Ω, and B is in siemens. Xpos and B can both be for the total cable
length, as in Table 3-6, in which case the units of Vpos are cable lengths per second. If Xpos and B are both per meter,
the units of Vpos are meters per second. For Xpos = 7.5832 and B = 0.00243, we have Zpos = 55.86 Ω and Vpos =
2314.31 * 36e3 = 8.332e7 m/s. The surge impedance is about one order of magnitude lower than for an OHL, and
the wave velocity is about fourth the speed of light, both within the range of typical values for UGC. Equations (4.4)
and (4.5) may be used for zero sequence (or ground mode) Bergeron parameters, and to cross-check steady-state
and Bergeron model parameters for consistency. For example, if the derived Bergeron parameters are outside of
typical ranges, it could mean that the planning-model parameters included a series current-limiting reactor, or
described two cables in parallel.

4.8 Transient Cable Model Validation Procedure

The normal procedure when validating transient cable models for system transient studies, such as insulation
coordination on cable systems, is to verify the EMT based simulation model with power frequency data only as
described in Chapter 3. This method is limited, and will not give perfect validation of all aspects of transient
modelling, but can be assumed adequate in the absence of real transient measurement data. This is suggested as
adequate, as the program developer validates each individual detailed EMT model by measurements or
calculations. The model user must rely on the model developer regarding accuracy of each modelled component.
However, when transient measurement data is available, it is possible to validate transient cable models by
comparing simulation results with the results of field measurements. Reference [76] and Annex 3 give good
examples of field measurements of the cable core / sheath overvoltage.
Both low voltage or system voltages are used for the field measurements. When low voltage is applied, disturbance
recorders or oscilloscopes can be used for the measurements. When system voltage is applied, capacitive dividers
are required to transform the actual line voltage to signals suitable for measurement by disturbance recorders or
oscilloscopes. When VTs or CTs are used for the measurement, their frequency responses must to be known.
Cable model validation should be conducted in the frequency range that is important to the analysis. It is not
recommended to validate the cable model against very high frequency (greater than 1 MHz) since Pollaczek’s
formula, on which all Cable Constants are based, is not accurate at very high frequency range and can lead to
numerical instability [63].

Since it is often difficult to obtain transient measurement data, future challenges should include the development of
a method to validate transient cable models without using transient measurement data. This section introduces one
promising method that can be applied to any time-domain model described in Section 4.2.
The typical frequency domain program is described in this paragraph. The current and voltage relationship at a
terminal of a transmission line can be expressed in terms of transmission line (frequency dependent) parameters
such as series impedance and shunt admittance. If a simple linear network (possibly with current or voltage
sources) is connected to transmission line terminals, (frequency domain) transfer functions for voltages and
currents at terminals can be derived by combining equations for the transmission lines and the external network.
Using techniques such as Numerical Laplace Transform, it is possible to convert the frequency domain transfer
function into a time domain waveform (e.g. for sending-end current).
Indeed, most cable constant programs that generate data for use in time domain simulation use frequency domain
equations to calculate the cable parameters. Typically, the cable being simulated in the time domain would be
cross-bonded and may have additional resistive terminations to ground along its length. Depending on the

Page 81
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

sophistication of the frequency dependent program, it may be possible to validate this entire cable. However, if
such a program is not available, it is still possible to write a simple program (using commercial software) to
calculate the various frequency dependent parameters such as series impedance, shunt admittance as functions of
the cable geometry and material properties such as resistivity of the conductors and ground, relative permittivity etc.
(see Annex 1, frequency dependent phase domain models). Some electromagnetic transients programs may also
directly provide this information.
In the simplest validation approach, a smaller cable segment (in between cross-bonding points) can be considered.
It can be excited on the sending end by a simple voltage or current source, and terminated at the receiving end
with suitable impedance (including open or short circuit). Using the frequency dependent program mentioned in the
previous paragraph, the phase impedances of this cable as seen by the source at any given frequency can be
calculated in a straightforward manner. The magnitudes and phases of these impedances can be plotted or
tabulated as a function of frequency. The same cable segment can then be setup in the EMT program with the
identical terminations as in the frequency dependent analysis, and excited with a sending end source with a given
frequency. Simulation can be conducted till steady state is reached after which the voltages and currents in each
phase are measured from which the impedances can be calculated. The process is repeated with a different
source frequency to yield the impedances at several different frequencies. The impedances are then compared
with the frequency domain solution with a view to validate the EMT model. In order to save validation effort, the
comparisons may be conducted at a handful of frequencies where more accuracy is desired.
A comprehensive validation is possible if a more sophisticated frequency domain formulation is developed, that
allows cable segments to be cascaded. In such cases the cross-bonding and other intermediate terminations can
be represented in the frequency dependent program. The process discussed in the above paragraph can then be
applied to entire cable system rather than to a smaller segment, thereby increasing the confidence in the EMT
model. In order to assess the accuracy of the time domain simulation directly, a time-domain response can be
calculated in the frequency dependent program by using Numerical Laplace Transform techniques [83]. This
resultant waveform can then be used to validate the corresponding EMT simulation response of the full cable.
As an example, consider a 36 km cable (24 segments with 1.5 km each). Each segment is cross-bonded with the
next and there is a 5 Ω resistance to ground after every three segments. The full model is simulated on an EMT
program and also for validation on a frequency domain program with FFT capability. The cable is energized with a
DC voltage step as shown in Figure 4-11. Terminating and source impedances are 1 Ω resistors as shown in the
figure. Figure 4-12 shows the waveforms indicating a good agreement, thereby validating the EMT model.

Figure 4-11: Cable Model under Validation (1 Ω terminations)

Page 82
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Main : Graphs
Sending Current (EMT) Sending Current ( FreqDomain Calc)
400
350
300
250
200
150
100
50
0
-50
Sending Voltage (EMT) Sending Voltage ( FreqDomain Calc)
1.2k

1.0k

0.8k

0.6k

0.4k

0.2k

0.0

x 0.080 0.100 0.120 0.140 0.160 0.180 0.200

Figure 4-12: Comparison of Sending End (phase a) Current and Voltage Waveforms for
Step Excitation Obtained from EMT Simulation and FD Model

Figure 4-13 shows the arrangement used for energization with a fundamental frequency (50 Hz) ac source. The
terminations are with 1 Ω resistors at the sending end and essentially open circuit (100 kΩ) on all phase and
sheath conductors at the receiving end. The cable breaker (BRK) is switched on at the peak of the ac voltage
waveform. The corresponding sending end current and voltage waveforms obtained from EMT as well as
frequency domain simulation are shown in Figure 4-14. The agreement is very good, once again confirming the
validity of the EMT model of the cable.

Figure 4-13: Cable Model under Validation (near-open circuit termination on receiving
end)

Page 83
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Main : Graphs
Sending Current (EMT) Sending Current ( FreqDomain Calc)
40

20

-20

-40

Sending Voltage (EMT) Sending Voltage (FreqDomain Calc)

2.0k
1.5k
1.0k
0.5k
0.0
-0.5k
-1.0k
-1.5k
x 0.090 0.100 0.110 0.120 0.130 0.140

Figure 4-14: Comparison of Sending End (phase a) Current and Voltage Waveforms for
Sinusoidal Excitation with Receiving End Open Circuit Obtained from EMT Simulation and
FD Model

Page 84
Chapter 5. Conclusions and Recommendations

This Technical Brochure has looked at the system technical performance issues associated with the installation of
long AC cables. One of the first tasks has been to identify a definition for a long cable that would be acceptable for
the many stakeholders; thus, a long cable has been defined as the one that normally requires shunt reactor
compensation and / or other components installed with the cable.
Studies associated with any cable installation project have been divided into four separate stages; system planning
studies, system impact studies, equipment design studies and forensic studies. Likely issues that may be looked at
for each study stage have been summarised to help identify system technical performance related phenomena.
A weighting factor has been attached to each study type in order to emphasise the importance of the study due to
likely consequences. It is possible that different weighting factors may be more applicable for different power
networks due to some unique peculiarities and characteristics and therefore the given factors should only be taken
as indicative but nevertheless applicable in vast majority of cases. Relevant frequency range and phase along with
appropriate analysis tools have also been drawn for each particular study type.
Two example systems have been presented; one to demonstrate technical performance issues associated with
long cables and one to show issues associated with the presence of many cables. Many cable example test
system has been based on the IEEE 24-bus reliability test system by replacing overhead lines with cables in order
to illustrate issues associated with planning and system impact studies. A new test system has been developed to
show issues associated with long cables.
Although similar in principle to overhead lines, particular attention was given to reactive compensation studies with
illustrative calculations shown in a step by step manner in order to emphasise the differences between overhead
lines and cables.
It has been shown that replacing existing overhead lines with cables will shift parallel resonances to lower
frequencies. This has the potential to exacerbate the existing harmonic voltage distortion on power systems.
Considering that many systems have seen an upward trend in background harmonic distortion levels, it has been
suggested that a complete evaluation at different system loading levels with many system contingencies should be
considered.
Synchronous generator self-excitation with possible uncontrolled voltage rise at generator terminals based on
reported incidents has been highlighted to be another possible issue due to excessive reactive power generated by
cables connecting a power plant to the transmission grid. Connection of generators in parallel, thus reducing their
equivalent reactance or introduction of shunt reactors to the transmission circuit and thereby increasing the
external reactance seen by the generator have been suggested as possible mitigation measures to avoid generator
self-excitation.
Various temporary overvoltage (TOV) phenomena such as those due to ground fault back feed, generator self-
excitation, system islanding and resonance have been shown. Since temporary overvoltages are experienced in a
relatively wide area, the evaluation is often performed against the insulation performance of the weakest equipment
in the area. Therefore particular attention has been drawn to surge arresters due to their vulnerability to TOV.
Although tested for TOV capability under a stiff power frequency voltage source, this capability should not be used
as a withstand envelope. This is because TOVs shown in example cases occur at higher frequencies and with
variable magnitudes. As arresters may not necessarily withstand a voltage waveform that is under their TOV
capability curve it is particularly important to consult equipment vendors/manufacturers in such cases in order or
ascertain capability against calculated values. It is also important that the calculation of TOVs consider various
system contingencies, loading levels and other parameter sensitivities as these affect the resonant frequency and
the damping rate of TOVs.
Example mitigation measures to control TOV especially in the case of ground fault have been drawn. These
measures include limiting cable lengths to avoid resonances at or below 3rd harmonic, increasing arrester voltage
ratings and reducing system voltage under light loading conditions.
For design stage studies, energization of cables has been classified as particularly important as severe TOVs could
be established and therefore assessment to ascertain adequate insulation level is critical. It has been suggested
that unless absolute certainty exists that there can be no re-energization, studies should be run on a statistical
simulation basis to assess the maximum overvoltage with trapped charge. The insulation performance of cables is
normally tested by application of lightning and switching impulse voltages in combination with a power frequency
voltage. IEC 62067 defines the test methods and requirements for XLPE cables and accessories for rated voltages
of 150 kV and up to 500 kV. For example, for a 400 kV cable, a 440 kV power frequency voltage is applied for 15
minutes following a lightning impulse test. Considering the voltage level that can be attained by TOVs, the level of

Page 85
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

applied power frequency voltage appears to be somehow low even though the duration is sufficient. Therefore if a
match between possible TOVs and the applied test voltage is preferred one possibility that may be considered by
the relevant Study Committee on cables could be to shorten the duration of the test while raising the applied
voltage level. In the meantime, it is possible to utilise a theoretical conversion of insulation performance to a shorter
duration but with a higher voltage level using voltage-time relationship of insulation characteristics.
It has been shown that transient frequencies involved in cable energization are lower than for typical overhead lines.
This is more so in cross-bonded cables which leads to lower damping of overvoltages. Cross-bonding has been
shown to be the factor for overvoltages to propagate to other phases through the sheath. It has been concluded
that in the case of cable energization through a cable fed busbar particular attention should be given to high inrush
currents and checking capacitive switching capabilities of breaker as it is analogous to back to back capacitor
energization.
Zero-miss phenomenon characterized by a current not crossing zero for several cycles following energization of a
cable has been identified as another potentially dangerous issue in that if there was a fault on the cable it may be
impossible to open healthy phase poles of a circuit breaker without due damage. Various mitigation measures have
been identified and presented with some explanation.
Inductive coupling to other metallic structures has been suggested as another area of study due to being a possible
safety hazard by introduction of touch voltages as well as contributing to corrosion.
The effect of other peripheral equipment on system impact and equipment design studies has been tabled along
with appropriate size of network model for these studies. In most cases, peripheral power system component
models often lack frequency dependency due to unavailability of data and precise models. However, if available,
inclusion of frequency dependent models is useful as it usually provides reasonable damping in the frequency
region of interest.
Cable modelling has been covered extensively under two main headings: lumped parameter models for use in
steady-state power system analysis and distributed parameter models for use in electromagnetic transient analysis.
Manual calculation of sequence parameters for a lumped model has been illustrated extensively. Results have
been compared against a number of line constant solvers of commercially available software packages with very
good agreement. This in itself forms a good verification tools in terms of the developed cable model that can be
used in studies requiring sequence models.
Reader’s attention has been drawn to some practical consideration while attempting to calculate cable parameters.
It is recommended that semi-conducting layer is treated as part of the insulation by appropriate adjustment of the
permittivity of the insulation. Likewise, due to difficulties in modelling wires that make up the screening of a cable it
has been recommended that screen wires are treated as sheath with the application of approximation equations.
Other practical considerations such as the decrease of skin effect factor with increasing conductor temperature,
and the effect of earth on zero sequence impedance have also been illustrated.
The importance of AC resistance in cable impedance sequence modelling for system studies cannot be ignored. It
has been shown that use of line constant solvers does overestimate AC resistance especially in the case of
segmented or as more commonly referred to Milliken conductors. It has been concluded that the introduced error is
not so pronounced in the case of round stranded, hollow and helical stranded conductors.
As high frequency is intrinsic to most power system transient phenomena, relevant power system component
modelling includes some form of frequency dependence. Cables are no exception to this and the difficulty lies in
the convolution of the frequency dependent transfer function from frequency domain into the equivalent time
domain. Various models that have been developed to account for this are given in a short form without going into
the complexity of their derivation.
A correlation has been drawn between the various issues identified for power system technical performance related
phenomena and the minimum cable model requirement. Bergeron model has been recommended as a minimum in
most cases owing to the fact that it produces relatively conservative results with solid numerical stability and lower
computational burden and hence being the first choice of most engineers.
An alternative method in the selection of target frequency when employing Bergeron model has been presented in
order to minimise the conservative results produced by the use of power frequency for Bergeron models. Choice of
target frequency selection criteria has been drawn based on the calculated characteristic and critical frequencies. It
has been shown that Bergeron model with target frequency set according to the given methodology produces
sufficiently accurate results in the frequency range of interest with errors in the region of 5% on the modal
propagation velocities and modal characteristic impedances. Hence, the use of Bergeron models with this type of
approach is suggested for many transient modelling studies identified in this brochure. A comparative study to
calculate the maximum overvoltage during a cable energization has been included to illustrate the validity of this

Page 86
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

recommendation. However, it must be emphasised that although the usage of Bergeron model is sufficient for
establishing the maximum overvoltage, it has its limitations in producing accurate representation of subsequent
overvoltages when inspected over a range of timeline in a simulation study. Therefore, in cases where higher
accuracy is required, such as in forensic studies where there is usually a need to match measured data with
simulation, frequency dependent models should be used whenever possible.
Cable models in all existing commercial power system analysis tools are based on the formulation of impedance
and admittance matrices for coaxial cables and this trend has been followed in this Technical Brochure. However,
important references were included for those more willing practitioners to attempt modelling of non-standard
configurations by use of either finite element analysis or by employing some form of approximation to fit an arbitrary
shaped conductor into an equivalent cylindrical form.
Various installation and environmental factors that should be taken into account during modelling has been drawn.
For switching studies, modelling of cross-bonding arrangements is highly recommended due to the effect on the
cable impedance and propagation velocity. Same applies during the assessment of sheath overvoltages in
transient studies. Taking into account the environment where cables are installed is equally important. This ranges
from modelling the electrical parameters of a tunnel where the cables are laid to simplifications to reduce a tunnel
into a pipe model. The importance of earth return path was highlighted due to its significant effect on cable
parameters.
A comprehensive sensitivity analysis with respect to the cable model input parameters was performed to establish
their effect on the positive and zero sequence resistance and reactance as well as admittance. This analysis
identified a set of most sensitive parameters that include conductor radius, permittivity of inner insulator layer and
thickness of inner insulator among others. Tables and graphs showing the effects in detail have been included in
Chapter 4 along with the relevant percentage change they introduce.
The best way to verify models is to compare calculated values with field measurements. For verification, a
measurement set up has to be made. For validation of high frequency models, a step function should be supplied
at the cable termination. These measurements should be carried out when the cable is out of service. Measuring
when the cable connection is still under construction is by far the best opportunity for this. Model validation for
higher frequencies is more cumbersome in practice than for power frequency.

Page 87
Bibliography/References
[1] CIGRE Technical Brochure 379, Update of Service Experience of HV Underground and Cable Systems, WG
B1.10, April 2009.
[2] CIGRE Technical Brochure 110, Comparison of High Voltage Overhead Lines and Underground Cables,
JWG 21/22.01, December 1996.
[3] CIGRE Technical Brochure 189, Insulation Coordination for HVAC Underground Cable System, JWG 21/33,
June 2001.
[4] CIGRE Technical Brochure 250, General Guidelines for the Integration of a New Underground Cable System
in the Network, WG B1. 19, August 2004.
[5] CIGRE Technical Brochure 268, Transient Voltages Affecting Long Cables, WG B1.05, April 2005.
[6] CIGRE Technical Brochure 283, Special Bonding of High Voltage Power Cables, WG B1.18, October 2005.
[7] CIGRE Technical Brochure 347, Earth Potential Rises in Specially Bonded Screen Systems, TF B1.26, June
2008.
[8] CIGRE Technical Brochure 531, Cable Systems Electrical Characteristics, WG B1.30, April 2013.
[9] IEC Technical Report 60071-4, Computational Guide to Insulation Co-ordination and Modelling of Electrical
Networks, June 2004.
[10] R. Benato, A. Paolucci: EHV AC Undergrounding Electrical Power. Performance and Planning, SPRINGER,
Series: Power Systems, ISBN: 978-1-84882-866-7, 2010.
[11] R. Benato, A. Paolucci: Operating Capability of Long AC EHV Transmission Cables, Electric Power Systems
Research, Vol. 75/1, luglio 2005, pp. 17-27.
[12] R. Benato, A. Paolucci: "Operating capability of ac EHV mixed lines with overhead and cables links", Electric
Power System Research, Vol. 78/4, aprile 2008, pp. 584-594.
[13] IEEE Working Group Report, “Protective Relaying Considerations for Transmission Lines with High Voltage
AC Cables”, IEEE Trans. on Power Delivery, vol. 12, no. 1, pp. 83-96, January 1997.
[14] J. B. Lee, C. W. Ha, C. H. Jung, “Development of Digital Distance Relaying Algorithm in Combined
Transmission Lines with Underground Cables”, Proc. IEEE PES Summer Meeting, pp. 611-616, 15-19 July
2001.
[15] B. Kasztenny, I. Voloh, J. G. Hubertus, “Applying Distance Protection to Cable Circuits”, 57th Annual
Conference for Protective Relay Engineers, pp. 46-69, 30 March – 1 April 2004.
[16] A. Ametani, C. T. Wan, “Sheath Overvoltages due to Faults on an EHV Cable”, Presented at IPST 1995 in
Lisbon, Portugal.
[17] R. Benato: "Multiconductor Analysis of Underground Power Transmission Systems: EHV AC Cables",
Electric Power System Research, Vol. 79, Issue 1, gennaio 2009, pp. 27-38.
[18] IEEE Task Force Report, "IEEE Reliability Test System", IEEE Trans. on Power Apparatus and Systems, vol.
98, no. 6, pp. 2047-2054, Nov/Dec 1979.
[19] IEEE Task Force Report, “The IEEE Reliability Test System – 1996”, IEEE Trans. on Power Systems, vol. 14,
no. 3, pp. 1010-1020, August 1999.
[20] CIGRE Technical Brochure 30, Reactive Compensation Analyses and Planning Procedure, TF 03.01 of
Committee 38 (Power System Analysis and Technique). 1989.
[21] R. Benato, S. Lauria, F.M. Gatta, L. Colla, M. Rebolini, F. Renaud: "Steady-State And Transient Ehv Ac
Cable Shunt Reactive Compensation Assessment", Proc. of CIGRÉ 2010, Paper C4-109, 2010.
[22] Transmission Line Reference Book 200kV and Above, third edition, EPRI. Palo Alto, USA.2006, 1011974.
[23] M. Val Escudero, M. Redfern. “Effects of Transmission Line Construction on Resonance in Shunt
Compensated EHV Lines”, Presented at IPST 2005 in Montreal, Canada, Paper No IPST05-109.
[24] F. Iliceto, E. Cinieri, A. Di Vita, “Overvoltages due to Open-Phase Occurrence in Reactor Compensated EHV
Lines”, IEEE Trans. On PAS., Vol PAS-103, no. 3, March 1984. pp. 474-482.
[25] S.Lauria, F.M.Gatta, L.Colla, “Shunt compensation of EHV cables and mixed overhead-cable lines” , Proc.
2007 IEEE Lausanne Power Tech Conf., paper n. 562.
[26] R. A. Turner, K. S. Smith, "Resonance Excited by Transformer Inrush Current in Inter-connected Offshore
Power Systems", Industry Applications Society Annual Meeting, 2008.
[27] Surya Santoso, Roger C. Dugan, Thomas E. Grebe, Peter Nedwick, "Modeling Ferroresonance Phenomena
in an Underground Distribution System", International Conference on Power System Transients (IPST), 2001.

Page 88
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

[28] J. R. Marti, A. C: Soudack, "Ferroresonance in power systems: Fundamental solutions", IEE Proceeding-C.,
Vol. 138, No. 4, July 1991.
[29] V. Valverde, A. J. Mazón, I. Zamora, G. Buigues, "Ferroresonance in Voltage Transformers: Analysis and
Simulations", International Conference on renewable Energies and Power Quality, Sevilla, 2007.
[30] IEEE Std. C62.11, IEEE Standard for Metal-Oxide Surge Arresters for AC Power Circuits (> 1 kV).
[31] IEC Std. 60099-4, Surge arresters – Part 4: Metal-oxide surge arresters without gaps for a.c. systems
[32] IEC Std. 60099-5, Surge arresters – Part 5: Selection and application recommendations.
[33] Reliability and Operating Committee (ROC) Final Report, 20 December 2004, available at:
http://www.ct.gov/csc/cwp/view.asp?a=3&q=287674
[34] J. H. R. Enslin, J, Knijp, C. P. J. Jansen, J. H. Schuld, “Impact of Reactive Power Compensation Equipment
on the Harmonic Impedance of High Voltage Networks”, IEEE Bologna Power Tech Conference, June 23-26,
2003, Bologna, Italy.
[35] J. H. R. Enslin, R. A. Wakefield, Y. Hu, S. Eric, “Harmonic Impedance Study for Southwest Connecticut
Phase II Alternatives”, October 18, 2004, available at
http://www.ct.gov/csc/lib/csc/docket_272/kema_harmonic_analysis_report.pdf
[36] A. Ketabi, A. M. Ranjbar, R. Feuillet, “Analysis and Control of Temporary Overvoltages for Automated
Restoration Planning”, IEEE Trans. on Power Delivery, vol. 17, no. 4, pp. 1121-1127, October 2004.
[37] T. Karasaki, T. Goto, A. Ametani, ”An Abnormal Overvoltage due to Load Rejection on EHV Underground
Transmission Lines”, International Conference on Power System Transients (IPST), September 1995.
[38] N. Momose, H. Suzuki, S. Tsuchiya, T. Watanabe, “Planning and Development of 500 kV Underground
Transmission System in Tokyo Metropolitan Area”, CIGRE Session 37-202, 1998.
[39] V. Akhmatov, B. C. Gellert, T. E. McDermott, W. Wiechowski, “Risk of Temporary Over-Voltage and High-
Voltage Fault-Ride-Through of Large Wind Power Plants”, 9th International Workshop on Large-Scale
Integration of Wind Power into Power Systems as well as on Transmission Networks for Offshore Wind
Farms, October 18-19, 2010, Quebec City, PQ..
[40] T. Petersson, “Reactive Power Compensation”. ABB Power Systems. Information NR 500-028E. December
1993.
[41] W. Diesendorf, “Insulation Co-Ordination in High Voltage Electric Power Systems”. Butterworths, 1974. ISBN
0 408 70464 0.
[42] CIGRE Technical Brochure 119, Interaction between HVDC Convertors and nearby Synchronous Machines,
WG 14.05. October 1997.
[43] X. Youfang, “The Practical Criterion of Generator Self Excitation and its Field Example Analysis”,
TENCON’93 Proceedings. Conference on Computer, Communication, Control and Power Engineering, IEEE
Region 10 Conference. DOI 10.1109/TENCON.1993.320572.
[44] C. Concordia, Synchronous Machines, J. Wiley, New York, 1951.
[45] E. W. Kimbark, Power System Stability, J. Wiley, New York, 1948.
[46] F. Iliceto, E, Cinieri, A. Di Vita, “Overvoltages Due to Open-Phase Occurrence in Reactor-Compensated EHV
Lines”, IEEE Trans. on Power Apparatus & Systems, vol. 103, no. 3, pp. 474-482, March 1984.
[47] B. Kulicke, H. H. Schramm, “Clearance of Short-circuits with Delayed Current Zeros in the Itaipu 550 kV-
Substation”, IEEE Trans. on Power Apparatus and Systems, vol. PAS-99, no. 4, July/Aug 1980.
[48] T. Michigami, S. Imai, O. Takahashi, “Theoretical Background for Zero-miss Phenomenon in the Cable
Network and Field Measurements”, IEEJ General Meeting 1459, 1997 (in Japanese).
[49] H. Hamada, Y. Nakata, T. Maekawa, “Measurement of Delayed Current Zeros Phenomena in 500 kV Cable
System”, IEEJ General Meeting 2001 (in Japanese).
[50] K. Søgaard, C. L. Bak, "Analyse og modellering af dynamiske forhold for 400kV linjen FERSLEV-TRIGE",
Institute of Energy Technology, Aalborg University, June 2005 (In Danish???).
[51] ITU-T (International Telecommunication Union, Telecommunication Standardization Sector), Directives
Concerning the Protection of Telecommunication Lines against Harmful Effects from Electric Power and
Electrified Railway Lines; Volumes II and III.
[52] CIGRE Joint Working Group C4.207, Guide for Assessment of Transferred EPR on Telecommunication
Systems due to Faults in A.C. Power Systems: Propagation of EPR (Earth Potential Rise) by metallic
conduction, characteristics of used system configuration and precautions for telecommunication equipment.
Under preparation.

Page 89
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

[53] CIGRE/CIRED Technical Brochure 291, AC Corrosion on Metallic Pipelines due to Interference from AC
Power Lines, Phenomenon, Modelling and Countermeasures, JWG C4.202, April 2006.
[54] A. Ametani, "Wave Propagation Characteristics of Cables", IEEE Transactions on Power Apparatus and
Systems, Vol. PAS-99, No. 2, March/April, 1980.
[55] IEC 60228, “Conductors of insulated cables”, International Standard, Third Edition 2004-11.
[56] IEC 60287-1-1, “Electrical cables- Calculation of the current rating. Part 1: current rating equations and
calculation of losses”, International Standard, First Edition 1994-12.
[57] Bjørn Gustavsen “Panel Session on Data for Modeling System Transients. Insulated Cables”, Proc. IEEE.
Power Engineering Soc. Winter Meeting , 2001.
[58] P. Graneau, “Underground power transmission”, New York : John Wiley & sons, 1979, ISBN 0471057576
[59] M.M. Kashani,”Frequency Dependent Impedance of Stranded Conductors using Subdivision method”, PhD
Thesis, University of British Colombia, Canada, 1990.
[60] CIGRE Technical Brochure 250 Technical and environmental issues regarding the integration of a new HV
underground cable system in the network, 2004.
[61] A. Ametani, “A general formulation of impedance and admittance of cables”, IEEE Transactions on Power
Apparatus and Systems, vol. PAS-99, no. 3, pp. 902-910, May/June 1980.
[62] CIGRE Technical Brochure 272, Large cross-sections and composite screens design, WG B1.03, June 2005.
[63] H. W. Dommel, EMTP Theory Book, 2nd ed. Vancouver, British Columbia Canada: Microtran Power System
Analysis Corporation, Vancouver, British Columbia, 1996.
[64] A. Semlyen and A. Dabuleanu, "Fast and accurate switching transient calculations on transmission lines with
ground return using recursive convolutions", IEEE Transactions on Power Apparatus and Systems, Vol. PAS-
94(2), p. 561-571, 1975.
[65] J. Marti, Accurate Modeling of Frequency Dependent Transmission Lines in Electromagnetic Transients
Simulation, IEEE Transactions on Power Apparatus and Systems, PAS-101, #1, pp. 147-155, Jan. 1982.
[66] T. Noda, N. Nagaoka, A. Ametani, “Phase Domain Modelling of Frequency Dependent Transmission lines by
means of ARMA model”, IEEE transactions on power delivery, Vol. 11, No. 1, January 1996.
[67] L. Marti, “Simulation of transients in underground cables with frequency dependent modal transformation
matrices”, IEEE transactions on Power delivery, Vol. 03, No 03, July 1998.
[68] A. Morched, B. Gustavsen, M. Tartibi, A Universal Line Model for Accurate Calculation of Electromagnetic
Transients on Overhead Lines and Cables, Paper PE-112-PWRD-0-11-1997.
[69] A. Ametani, Distributed-Parameter Circuit Theory, Corona Pub. Co., 1990 (in Japanese).
[70] A. Ametani, “A study of cable transient calculations, Part II Transient calculations”, Science and Engineering
Review of Doshisha University, vol. 24, no. 2, pp. 119-127, July 1983.
[71] C. H. Chien, R. Bucknall, “Harmonic Calculations of Proximity Effect on Impedance Characteristics in Subsea
Power Transmission Cables”, IEEE Trans. on Power Delivery, vol. 24, no. 4, October 2009.
[72] B. Gustavsen, A. Bruaset, J. Bremnes, A. Hassel, ”A Finite-Element Approach for Calculating Electrical
Parameters of Umbilical Cables”, IEEE Trans. On Power Delivery, vol. 24, no. 4, October 2009.
[73] A. Ametani, Ikuko Fuse, “Approximate Method for Calculating the Impedance of Multicounductors with Cross
Sections of Arbitrary Shapes”, Electrical Engineering in Japan, vol. 112, no. 2, 1992.
[74] N. Nagaoka, A. Ametani, “Transient Calculations on Crossbonded Cables”, IEEE Trans. on Power Apparatus
and Systems, vol. PAS-102, no. 4, April 1983.
[75] A. Ametani, Y. Miyamoto, N. Nagaoka, “An Investigation of a Wave Propagation Characteristic on a
Crossbonded Cable”, IEEJ Transaction B, vol. 123, no. 3, 2003 (in Japanese).
[76] N. Nagaoka, K. Yoshida, T. Karasaki, “Transient Sheath Voltage Characteristics of a Crossbonded Cable
Installed within a Tunnel”, IEEJ vol. 112, no. 5, 1992.
[77] L.Wedepohl and D.Wilcox, “Transient analysis of underground power-transmission systems,” IEE
Proceedings, vol. 120, no. 2, pp. 253–260, February 1973.
[78] A. Ametani, “Stratified Effects on Wave Propagation”, IEEE Trans. PAS, col. 93 (5), pp. 1233-1239, 1974.
[79] A. Ametani, “On the Impedance and the Admittance in the EMTP Cable Constants/Parameters”, European
EMTP User Group Meeting, October 2009.
[80] A. Ametani, T. Yoneda, Y. Baba, N. Nagaoka, “An Investigation of Earth-Return Impedance between
Overhead and Underground Conductors and Its Approximation”, IEEE Trans. on Electromagnetic
Compatibility, vol. 51, no. 3, August 2009.

Page 90
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

[81] B. Gustavsen, “A study of overvoltages in high voltage cables with emphasis on sheath overvoltages”,
Trondheim, Norway: PhD. Thesis, NTH, 1986.
[82] A. Ametani, C. T. Wan, “Sheath Overvoltages due to Faults on an EHV Cable”, International Conference on
Power System Transients (IPST), September 1995.
[83] Y. Goto, A. Ametani, T. Kubo, N. Nagaoka, T. Baba, “A Surge Analysis of a Cable System Composed of
Submarine and Underground / Overhead Cables”, UPEC P58-7, September 2009.
[84] A. Ametani, “Problems and Countermeasures of Cable Transient Simulations”, EMTP Journal, vol. 5, April
2000 (in Japanese).
[85] S. Nishiwaki, T. Koshizuka, O. Hosokawa, K. Yamamoto, “Comparison between Measurement and ATP
Computation on Traveling Wave Propagation Characteristics of Long Power Cable”, International Conference
on Power Systems Transients (IPST 2009), available on the web
(http://www.ipst.org/techpapers/2009/papers/25.pdf).
[86] W. I. Bowman and J. M. McNamee, “Development of Equivalent Pi and T Matrix Circuits for Long
Untransposed Transmission Lines”, IEEE Trans. On Power Apparatus and Systems, Vol. 83, Issue 6, pp.
625-632, June 1964.
[87] William D Stevenson Jr., “Elements of power system analysis”, fourth edition, McGraw-Hill, 1982.
[88] L. Colla, M. Rebolini, F. Iliceto, “400 kV AC new submarine cable links between Sicily and the Italian
mainland. Outline of project and special electrical studies”, CIGRE Session C4-116, 2008.
[89] IEEE PES Task Force on Data for Modeling System Transients of IEEE PES Working Group on Modeling
and Analysis of System Transients Using Digital Simulation (General Systems Subcommittee), “Parameter
Determination for Modeling System Transients—Part II: Insulated Cables”, IEEE Trans. on Power Delivery,
vol. 20, no. 3, JULY 2005.
[90] Juan A. Martinez-Velasco, ed., “Power System Transients: Parameter Determination”, CRC Press, 2009.
[91] L. Colla, S. Lauria, F. Palone “Finite Sections Modeling of Power Cable Systems”,International Conference on
Power Systems Transients (IPST2011), paper 139, Delft, Netherlands, June 14-17, 2011

Page 91
Annex 1. CABLE MODEL IMPLEMENTATIONS
Annex 1 describes available computer modelling employed in the typical power system engineering analysis
software:
 PI models
 Method of finite sections
 Bergeron travelling wave models
 Frequency dependent mode models
 Frequency dependent phase domain models

PI models

Figure A1-1: Representation of a PI-section Cable Model

Cable models, using PI-sections, are used for short cables or where the travel time does not exceed the solution
time-step for the simulation [A1]. There are two PI-section models; the nominal PI and the exact PI. The nominal
model is based on lumped R, L and C parameters for the transmission cable where coupling to ground and
between phases is considered. The series branch in the nominal PI-section model is calculated by multiplying the
series impedance of the cable line per unit distance by the cable length. The shunt branch is the shunt admittance
and treated in the same fashion, for example per km multiplied by the cable length. This type of model can only be
used for single phase steady state simulations while the exact PI-model represents the correct representation for
the cable line at a given frequency, see equation (A1.3) [A2].
 sinh( l ) 
Z  ( R  j L)  l   
 l 
 l 
tanh 
(G  j L)  l  2
Y   (A1.1)
2   l 
 2 
  ( R  j L)(G  jC )

where ( R  j L ) is the series impedance of the cable line,

(G  j C ) is the shunt admittance of the cable line,

 is the wave propagation constant and

l is the cable length.
The exact model can be used both for steady-state analysis and for frequency scan, for a given frequency. Neither
of the two PI-section models can be used for long cables or transient studies where more than one frequency
component is considered.

Method of finite sections

A stable and accurate method of modelling electromagnetic propagation in conductive materials, termed the
method of finite sections was introduced by Meredith [A3]. The method addresses the issues of modelling
frequency-dependent impedances and frequency-dependent coupling of conductors. Representing the state of the

Page 92
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

art of the time, the method used passive elements to construct a complex network representing a number of
coaxial cables. Finite sections modelling represents the integration of accurate conductor modelling utilizing
transmission pi-section modelling. Figure A1-2 below shows such an example. The single frequency branches of a
pi-section model are extended to synthesised multi-frequency equivalents.

Rcn Rsn Rcn Rsn

Lcn Lsn Lcn Lsn

Rc1 Rs1 Rc1 Rs1

Lc1 Ls1 Lc1 Ls1

Rc0 Rs0 Rc0 Rs0

Ld Ld
G/2 C/2 G C C/2 G/2

Figure A1-2: Two Transmission Line PI-sections Modelling Conductor-guided Propagation

in a Dielectric with Multi-frequency Conductor Loss

The addition of earth return complicates the nodal matrix representation of the coaxial cables. For effective use, a
method to transform cable parameters is required to generate the nodal matrix. This method may be the only fully
passive and accurate EMT means of modelling wave propagation in non-linear resistive and/or inductive
conductors. This method has been implemented in an EMT-type software [88].
The disadvantage of this modelling technique is that the nodal matrix expands dramatically impacting simulation
numerical efficiency due to the N2 solution factor of nodal methods [42].

Bergeron travelling wave models

The distributed parameter model is based on travelling wave theory and explains the cable as a series connection
of many line elements of a differential length dx as shown in Figure A1-3.

Figure A1-3: Equivalent Circuit of a Differential Length dx Cable Elements for One Cable,
without Mutual Coupling to Other Cables

The Bergeron model is a Norton equivalent model based on travelling wave theory [A1]. This model is often used
when modelling longer power transmission networks and represents the cables inductance and capacitance by its

Page 93
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Norton equivalent with lumped resistance representing the cables distributed series resistance, where the Norton
equivalent is lossless and the resistance represents losses in the transmission line.
The model is usually set up so that the lumped resistance is represented by dividing the line length into only two
sections, where each end has 1/4 part of the resistance and the centre contains 1/2 part. This is shown in Figure 4-
1 in Chapter 4.

The sending end currents in Figure A1-3 can be calculated by using voltage division as shown in equation (1.4)
[A4]. Receiving end currents are obtained similarly.
v1 (t )
i12 (t )   I1 (t   / 2)
ZC  R / 4
(A1.2)
v (t   / 2) Z C  R / 4
I1 (t   / 2)  2  I 2 (t   / 2)
ZC  R / 4 ZC  R / 4

This type of cable modelling can be found in most EMT simulation software. In multi-conductor case, there are
several assumptions (e.g. using constant transformation matrix) in transforming phase quantities in to modal
quantities and hence the model approximately represents the fundamental frequency.

Frequency dependent modal domain models

Based on the formulations of L. Marti [67], the Frequency Dependent Modal domain model utilises a constant (or
frequency-independent) modal transformation matrix to de-couple multiple-phase systems into separate, mutually
exclusive modes. Each mode is thereafter treated as a single-phase circuit. Although classified as 'frequency-
dependent', this model is exact in its frequency dependence only for geometrically balanced transmission systems,
such as ideally transposed circuits or any other systems where a naturally occurring, constant modal transformation
matrix occurs.

These models as outlined in references [65], [67], [A6] are based on modal theory in which the coupled
transmission line equations are decoupled using transformation matrices. As an example, consider the following
system of equations:
Y=GX (A1.3)
Using suitable transformation matrices, the above can be decoupled as shown below. Each decoupled mode can
be treated as a single phase circuit in the solution.

YM =GM XM (A1.4)

In general the transformation matrix from direct phase domain to modal domain is frequency dependent and
complex. However for most overhead transmission lines (e.g. horizontally symmetrical overhead lines) or simple
cable systems this transformation matrix is almost constant and also the imaginary part of the transformation matrix
is negligible compared with the real part. For such cases, the transformation matrices can be considered to be real
and constant with little error in the time domain solution. The assumption of the real constant transformation
matrices leads to efficient time domain modelling, since less number of convolution are required to solve for a given
multi-conductor transmission line.

Frequency dependent phase domain models

The frequency dependent model is a model with all parameters distributed (RLC), based on travelling wave theory
and without any lumped parameters. These models are frequency dependent in all parameters and they overcome
problems associated with transforming from modal domain to time domain.
By not using the modal domain, the transmission line equations remain coupled. The frequency dependent line
equations are given in equation (A1.5).

Page 94
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

I send  YCVsend  H (YCVrec  I rec )

(A1.5)
I rec  YCVrec  H (YCVsend  I send )

where all parameters are a function of frequency (ω), Isend/Vsend is the sending end current/voltage, Irec/Vrec is
the receiving end current/voltage,

(A1.6)

is the cable’s characteristic admittance,

Y ( ) Z ( )l
H ( )  e 
(A1.7)
is the cable’s propagation matrix and l is the cables length.
The frequency dependent phase domain models all calculated YC and H matrices by eigen decomposition methods
and then use fitting methods or convolution in order to convert the frequency dependent line equations of (A1.5) to
time domain equations.
The frequency dependent phase domain models are of more interest than modal domain models, because phase
domain models avoid problems with frequency dependent transformation matrices. Various phase domain models
have been developed. To name a few: The ARMA model (Auto-Regressive Moving Average) [66], the zcable
model [A7] and the Gustavsen model [A8].
Frequency dependent phase models can be found in various implementations in EMT simulation software. The
model accurately represents the frequency dependence of the cables parameters and is therefore recommended
for studies with various frequencies, such as transient studies and harmonic studies.

References
[A1] Watson, E.F. & Arrillaga, J, 2003. Power systems electromagnetic transient simulation. London: The
Institution of Electrical Engineers. ISBN: 0 85296 106 5.
[A2] Tleis, N.D, 2008. Power Systems Modelling and Fault Analysis Theory and Practice. Oxford: Newnes. ISBN:
978-0-7506-8074-5.
[A3] R. Meredith, “EMTP Modeling of Electromagnetic Transients in Multi-Mode Coaxial Cables by Finite
Sections”, IEEE Transactions on Power Delivery, Vol. 12, No. 1, January 1997.
[A4] Dommel, H.W: Digital Computer Solution of Electromagnetic Transients in Single- and Multiphase Networks,
IEEE Transactions on Power Apparaturs and systems, Vol. PAS-88, No. 4, April 1969, pp. 388-399.
[A5] Nguyen, T.T, "Earth-return path impedances of underground cables. I. Numerical integration of infinite
integrals", IEEE Proceedings on Generation, Transmission and Distribution, Vol. 145, Issue: 6, pp 621 – 626,
November 1998.
[A6] A. Ametani, “A highly efficient method for calculating transmission line transients”, IEEE transactions on PAS,
PAS-95, No 05, Sept/Oct 1976.
[A7] T.-C. Yu and J. R. Martí, “A robust phase-coordinates frequency-dependent underground cable model
(zcable) for the emtp,” IEEE Transactions on Power Delivery, vol. 18, no. 1, pp. 189–194, January 2003.
[A8] B. G. A. Morched and M. Tartibi, “A universal model for accurate calculation of electromagnetic transients on
overhead lines and underground cables,” IEEE Transactions on Power Delivery, vol. 14, no. 3, pp. 1032–
1038, July 1999.

Page 95
Annex 2. MEASUREMENT OF CABLE ELECTRICAL PARAMETERS
This Annex summarises how the cable vendors determine the electrical parameters given on data sheets. It is
important to understand this process when verifying cable model parameters against the data sheets. In some
cases, it’s helpful to duplicate the vendor test method in an EMT program when verifying consistent results.
Both the DC resistance and cable capacitance can be measured directly from the cable either at the factory or
once installed in the field. Each manufacturer may have a different procedure and it is important to understand
their method when reviewing the data as no standard measurement method exists. For a DC resistive or capacitor
bridge measurement, test sets are used to measure the cable parameters directly.
The conductor resistances of circuit A-B, B-C and C-A are measured, then the resistance must be corrected to
20°C. By averaging the measurements, the reported resistance per phase per unit distance is calculated .
Positive, negative, and zero sequence parameters for the above test report are calculated as follows:
For positive sequence values, balanced three phase voltage is applied to all three phase cables with the far end
bonded. For every phase conductor, voltage (phase to ground), current (I), active (P) and reactive power (Q),
power factor and frequency are measured.

Figure A2-1: Positive Sequence Impedance (Zpos) Measurement Test Circuit

Using the following formulas,

Im Z pos  Q I 2
(A2.1)
Re Z pos  P I 2
The imaginary component of Zpos is modified to the desired frequency. The impedance Zpos is then adjusted for the
maximum operating temperature 90 °C utilizing known formulas [55]. The per-phase impedance Zpos (Ω/m) is then
calculated by averaging the values for all three phases.

The calculation of the zero sequence parameters on the above test report are based on the test setup shown in
figure A2-2. The voltage between phase and earth (V), current (IAB), active power (P), power factor, reactive
power (Q), and frequency are all measured.

Page 96
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure A2-2: Zero Sequence Impedance (Zzero) Measurement Test Circuit

Im Z pos  3Q I AB
2

(A2.2)
Re Z pos  3P I AB
2

Using the above formulas, the imaginary part of Zzero is modified to the desired frequency. The impedance Zzero is
adjusted for the maximum operating temperature 90 °C. Then, the per-phase impedance Zzero(Ω/m) is then
calculated as follows.

 Re Z zero    Im Z zero 
2 2
Z zero  (A2.3)

The theoretical calculations used by the manufacturers are based on the assumption that all current will return
through metallic screens. However, this is not true for installed cables and must be accounted for.

Page 97
Annex 3. MEASUREMENTS ON THE 100 km LONG 150 kV AC CABLE
CONNECTION OF THE HORNS REV 2 OFFSHORE WIND FARM
A new 100 km long 150 kV AC cable line was installed in Denmark in 2009. This cable line is a radial connection
from an offshore windfarm to an onshore substation in the Danish transmission network. There was a unique
opportunity to perform detailed impulse measurements during the installation process, for model verification
purposes, and to perform full scale switching measurements. Both measurements were performed in relation to on-
going PhD projects. The purpose, performance, outcome and use of the measurements are described in detail in
[A9] and [A10].
The measurement method described here, can be adapted for model verification of other systems. Furthermore,
the results obtained can be used for validation of new modelling methods.
A schematic diagram of the cable connection for the Horns Rev 2 wind farm is shown in Figure A3-1. The main
components installed directly at the onshore substation called Endrup are also illustrated.

Figure A3-1: Diagram of the Cable Connection between the Onshore Substation in the
Danish Transmission Network and the Horns Rev 2 Wind Farm

The cable line is divided into three sections; a long land cable section, a short land cable section and a submarine
cable. The land cable sections are of the same cable type. The submarine and the land cable are connected with a
direct buried joint, such that the whole cable line is operated as a single 100 km long cable line.

Land cable
The land cable line consists of three 150 kV single core 1200 mm2 XLPE cables. The cables are laid in a tight
trefoil with the bottom cables at 1.3 m depth, see Figure A3-2.

Page 98
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure A3-2: Cross Sectional Layout of the 150 kV Land Cable System

A picture of the cable is shown in Figure A3-3. This picture shows all cable layers: core, core screen, insulation,
insulation screen, metal screen strands, metalic screen laminate and outer covering.

Figure A3-3: Cross Sectional Layout of the 150 kV Landcable

The land cable data are given below:

Core inner radius………………………………….. 0 mm
Core outer radius………………………………….. 20.75 mm
Core material………………………………………… Compact stranded Al
Core screen thickness……………………………. 1.5 mm
Insulation radius……………………………………. 39.25 mm
Insulation material………………………………… XLPE
Insulation screen thickness……………………. 1 mm
Diameter of each metal screen strand…... 1.91 mm
Thickness of Metal screen laminate………. 0.8 mm
Outer covering outer radius…………………… 47.94 mm
Outer covering material…………………………. XLPE

Page 99
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Submarine cable
The submarine cable is a three core armoured cable. The cross section of the submarine cable is shown in Figure
A3-4.

Figure A3-4: Cross Sectional Layout of the 150 kV Submarine Cable

The submarine cable data are given below:

Core inner radius………………………………….. 0 mm
Core outer radius………………………………….. 15.25 mm
Core material………………………………………… Compact stranded Cu
Core screen thickness……………………………. 1.5 mm
Insulation radius……………………………………. 34.75 mm
Insulation material………………………………… XLPE
Insulation screen thickness……………………. 1 mm
Metal sheath radius…............................... 40.15 mm
Material of metal sheath Lead alloy
Inner cable outer covering radius…..……… 42.45 mm
Inner cable outer covering material………. XLPE
Galvanised steel armour wire thickness… 5.6 mm
Cable outer covering radius…………………… 104 mm
Cable outer covering material………………. Polypropylene yarn

When modelling these specific cables, corrections of core resistivity and insulation permittivity must be included,
according to Section 3.4. When using the obtained results for validation of new modelling methods, the below
parameter corrections should be included.

Core resistivity
The core conductor is compact stranded. It is therefore necessary to correct the conductor resistivity, as explained
in Section 3.4.4. For the Horns Rev 2 land cable, the conductor resistivity is 3.19 x 10-8 Ω.m (A3.1).
∙ 20.75
2.8264 10 3.19 10 (A3.1)
1200

Page 100
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Insulation permittivity
The semiconductive layers are considered as being part of the insulation and the permittivity is corrected as
explained in Section 3.4.1.
The corrected relative permittivity is given in (A3.2). The insulation thickness is increased from 17 mm to 20.1 mm.

40.85
2.5 20.75 2.95 (A3.2)
39.50
22.25

Screen
The cable screen has a first layer of copper wires with a 95 mm2 cross-section and a second 0.8 mm thick
aluminium foil used as a water barrier. The diameter of the copper wires was measured with a micro meter
measuring tool and it was equal to 1.11 mm.
The wires are spaced, see Figure A3-3. Consequently, the cable screen cannot be modelled as a solid layer.
Applying the same method as used to correct conductor resistivity. The correction of the copper wires' resistivity is
as done in (A3.3).

, ,

41.96 40.85 (A3.3)

1.724 10
95
5.240 10 Ωm

The simulation models used in A3.1-Measurements for Model Verification have both screen layers modelled
separately in order to improve the results of the intersheath mode.
Such high level of accuracy is not necessary in A3.2-Full Scale Measurements and the two layers are simplified
into one by using (A3.4).

(A3.4)

5.240 10 2.826 10 288.8 10 212.9 10

5.240 10 212.9 10 2.826 10 288.8 10

8.98 10 Ωm

Submarine Cable
The correction of the submarine cable layers is made using the same approach used for the land cable. The
corrected conductor resistivity is given by (A3.5), and the corrected insulation permittivity by (A3.6).

Page 101
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

15.25 (A3.5)
1.724 10
630
1.999 10 Ωm

37.75
2.5 15.25 3.105 (A3.6)
34.75
16.75

For model verification purposes, the field measurements were performed in 3 steps; measurements on a single
minor section, measurements on a single major section, measurements on multiple major sections (the entire land
part of the cable line).

A3.1 Measurements for Model Verification

A3.1.1 Measurements on a single minor section

The purpose of this set of measurements was to verify impedance calculations and EMT type simulations for the
cable as a component. Therefore the coaxial and intersheath modes of propagation are excited on a single 1780 m
long minor section of the land cable.
In order to evaluate the velocity and attenuation of the waves propagating in the cable, it is practical to have a look
at the modal currents. There are six modes of propagation which can be described as three coaxial modes
(conductor-screen loop), two intersheath modes (screen-screen loop) and one ground mode (screen-ground loop).
As described in [A12], above a few kHz, the loops between core conductor and sheath becomes decoupled from
the outer loop between sheath and earth return. This is due to the fact that depth of penetration on the inside of the
sheath becomes smaller than the sheath thickness.
The velocity of the ground mode is normally many times less than those of both the intersheath mode and the
coaxial mode. The modal currents for the measured system can be seen in Figure A3-5.

Figure A3-5: Simulated Modal Velocity for the Single Minor Section of the Measured 150
kV Cable System

Page 102
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Because of the low velocity of the ground mode, only the coaxial and intersheath modes are tested in the field.
For the coaxial mode, the current flows in the core conductor and fully returns in the screen of the same phase and
for one of the intersheath modes, the current flows in one screen and fully returns in the screen of an adjacent
cable.
Therefore, for high frequencies, the impedance matrix for coaxial and intersheath modes can be simplified as
shown in (A3.7) and (A3.8) respectively.

(A3.7)

(A3.8)

where denotes an impedance of cable 1, Zgmij denotes the mutual earth impedance between cables i and j. VCSi
and VSGi are cable i conductor-screen and screen-ground loop voltages respectively and ICS1, ISG2 and ISG3 are the
conductor-screen loop current of cable 1 and screen-ground loop currents of cables 2 and 3 respectively. It should
be noted, that ISG3 = −ISG2.

For the intersheath mode, the earth return impedance therefore has an impact on the results. And by exciting only
the coaxial and intersheath modes of propagation, it is possible to evaluate the accuracy for separate impedance
parts of the full cable impedance. The cables are identical and in a tight trefoil, and therefore the different cable
impedances for each phase are assumed identical.

The purpose of the measurements is to obtain a base for validation of simulation models. The measurements
should validate the models for fast transient behaviour. According to IEC 60071-2 an impulse of 1.2x50 μs can be
used to simulate lightning overvoltages. The field measurements therefore use an impulse test to validate the cable
performance and model calculations for the cable terminal conditions. Using such an impulse as an input to the
cable, and measuring voltage/current in both ends of the cable gives results, where each reflection can clearly be
seen. This way, it is possible to validate modelled traveling time and the simulated cable parameters.

When performing simulations, one should use correct time step. Furthermore, during measurements, the sample
time should be sufficient. The single section of the underground 150 kV cable for the offshore HR2 wind farm is
1780 m long. The travelling time, assuming the speed of light, for 1780 m is approximately 5.94 μs. The front time
of the impulse, 1.2 μs, is therefore the dominating shortest time, and the time step and sample time should be no
larger than 0.12 μs. There should at least be 10 samples during the front time, in order to have an acceptable
resolution. The sample time should therefore be no less than 120 ns, which implies sampling frequency no less
than 8.33 MHz. The sampling frequency used for the measurements was 10 MHz.

The power source used for the impulse test is a 1.2/50 µs surge tester with a maximum output voltage of 7 kV.
For excitation of the coaxial mode, a 1.4 kV impulse is used and for excitation of the intersheath mode a 2 kV
impulse between the two sheaths is used. These impulses require a sampling frequency range of approx. 8 MHz
for measurements. For this short cable length, with an open end (and no cross-bounding point) the sending end
voltage will approximately double at the receiving end. In order not to risk any damage of the screen or measuring
equipment, the sending end voltage is limited to only 1,4 kV and 2 kV (between the sheaths). Normally the screen
is protected with a 5 kV surge voltage limiter.

The impulse generator was connected to the HV cable by use of a 5m long 300 mm2 Cu wire. The inductance of
such a wire is estimated as 1 μH per meter. The wire is therefore represented by a 5 μH inductance connected in
series with the surge generator. The two measurement setups are shown in Figure A3-6 and Figure A3-7. In Figure

Page 103
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

A3-6, the sending end voltage and current as well as the receiving end voltage, is measured for the excited phase.
In Figure A3-7, the sending end voltage and current for the excited sheaths, all core voltages as well as the
receiving end voltage for the excited sheaths are measured.

Figure A3-6: Excitation of the Coaxial Mode for a Single Minor Section

Figure A3-7: Excitation of the Inter-sheath Mode for a Single Minor Section

The results from measurements on a single minor section are shown in Figure A3-8 – Figure A3-10.

Page 104
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure A3-8: Results for Single Minor Section Measurements, Excitation of the Coaxial
Mode

Page 105
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

(A) SENDING END VOLTAGE, SCREEN AND CORE CONDUCTORS (B) SENDING END VOLTAGE, SCREEN AND CORE ONDUCTORS
CABLE 2, VSH2 AND VC2 RESPECTIVELY CABLE 2, VSH3 AND VC3 RESPECTIVELY

(C) SENDING END VOLTAGE, INDUCED VOLTAGE ON (D) SENDING END CURRENS, SCREEN CONDUCTORS CABLE 2
CORE CONDUCTORS AND 3

Figure A3-9: Results for Single Minor Section Measurements, Excitation of the Inter-sheath
Mode, at Sending End of the Cables

Page 106
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure A3-10: Results for Single Minor Section Measurements, Excitation of the Inter-
sheath Mode

As shown in equations (A3.7) – (A3.8), these results can be used to validate the impedances in the cable
impedance matrix. The measurements can therefore be used in order to verify a specific cable model or cable
parameter calculations. This has already been done in [A9]. The results are shown in Figure A3-11 and Figure A3-
12.
The coaxial mode showed no difference between measurements and simulations, and therefore calculations of
core impedance parameters were verified. There was however difference for the intersheath setup, and therefore
the proximity effect between wires of the screen, and between cables was included, as well as correct modelling of
the two screen layers (Al laminate and Cu wires) in a new cable model. The validation of this model improvement
can be seen in Figure A3-12, whereas detailed description of the model improvements and simulations can be
found in [A9].

Page 107
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure A3-11: Comparison of the Sending End Current and Receiving End Voltage on
Energized Screen Conductor for the Coaxial Mode

Figure A3-12: Comparison of the Sending End Current and Receiving End Voltage on
Energized Screen Conductor for the Intersheath Mode

Page 108
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

A3.1.2 Measurements on a single major section

The purpose of this set of measurements was to validate EMT type modelling of cable with cross-bonding points
and grounding of the screen.
There are two cross-bonding points in the setup for measurements on a single major section. The three cable
segments lengths are 952 m, 768 m and 742 m.
The generator used is the same 1.2/50 µs surge tester as before. For this test, the impulse does not double at the
end, because of damping in the cross-bonding points. Furthermore, other measuring equipment is used, allowing
for higher impulses, with still a sampling frequency of 10 MHz. A 4.3 kV impulse is used for excitation. The impulse
is kept below 5 kV in order not to exceed the protection level of the surge limiter. The impulse generator was, as
before, connected to the HV cable by use of a 5m long 300 mm2 Cu wire. The wire is therefore represented by a 5
μH inductance connected in series with the surge generator. The measurement setup is shown in Figure A3-13. In
the figure, the sending end voltage and current of the excited phase is measured, as well as all receiving end core
voltages.

Figure A3-13: Field Test Setup for the Single Major Section Measurements

The results are shown in Figure A3-14.

Page 109
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure A3-14: Results for Single Major Section Measurements

As the cross-bonding points are included for this measurement setup, and since the exact value of grounding
resistance was not known, the results cannot be used for validation of impedance and parameter calculations. The
results can however, be used to verify modelling of a cable line with cross-bonding points. This has been done in
[A9], and the results can be seen in Figure A3-15.

Page 110
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure A3-15: Validation of a Cable Model including Cross-bonding Points

A3.1.3 Measurements on multiple major sections

The purpose of this set of measurements was to validate EMT type modelling of a long cable line without shunt
compensation.

There are several cross-bonding points in the setup for measurements on multiple major sections. The cable
segments lengths are given below.

Page 111
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

The generator used is the same 1.2/50 µs surge tester as before. For this test, one must ensure that the impulse is
sufficiently large so it will not be damped too much along the long cable line. The voltage should still be below the
protection level of the surge voltage limiter of 5 kV. If this is not taken into account, the SVL’s can influence the
results. A 4.3 kV impulse is used for excitation. The impulse generator was, as before, connected to the HV cable
by use of a 5 m long 300 mm2 Cu wire. The wire is therefore represented by a 5 μH inductance connected in series
with the surge generator. The field measurement setup is as shown in Figure A3-16.

Figure A3-16: Field Test Setup for the Multiple Major Section Measurements

The results from measurements on multiple major sections are shown in Figure A3-17.

Page 112
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure A3-17: Results for Multiple Major Section Measurements

As the cross-bonding points are included for this measurement setup, and since the exact value of grounding
resistance was not known, the results cannot be used for validation of impedance and parameter calculations.
There are 22 cross-bonding points in this long line. These results should therefore not be used for validation of
modelling cross-bonding points. The results can however, be used to verify modelling of an entire long cable line,
without shunt compensation. This has been done in [A9], and the results can be seen in Figure A3-18.

Page 113
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Figure A3-18: Validation of a Cable Model for a Long Line, without Shunt Compensation

The set of measurements shown above can be used for various type of model validation. The results are split into
verifying the cable model itself, model of a cable with cross-bonding points and a model of a long cable line.
Measurements of only the coaxial and intersheath modes are acceptable and can be used for cable parameter and
modelling software verification. As cross-bonding points cause reflections and influence the measurement results,
measurements on a major cable section cannot be used for validation of cable parameter calculations. Field
measurements on a major section and a longer cable line can however be used for verifying the modelling as a
whole.
The cables are excited with a fast impulse, and the measurements thereby include a very high frequency spectrum.
In order to verify the model used for system studies, a full scale field test has also been performed. The results
from this test are given in Section A3.2.

Page 114
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

A3.2 Full Scale Measurements

In the full scale measurements, the cable is open in the receiving end (the wind farm) and it is energized directly
from the 150 kV network. Measurements were taken on three different points along the cable. The points and
measured signals were located as follows:
 Onshore substation (Point A in Figure A3-19)
o Three-phase voltages in the cable
o Three-phase voltages in the busbar (before the circuit breaker)
o Three-phase currents in the cable
 Shunt reactor (Point B in Figure A3-19)
o Three-phase currents in the shunt reactor
 Wind Farm (Point C in Figure A3-19)
o Three-phase voltages in the cable

C A
B

Figure A3-19: Measured Points

The energization, steady-state and de-energization waveforms were recorded three times in all three measuring
points. The closing of the circuit breaker was made by synchronised switching at voltage zero crossing. The
voltage and current were transformed to values readable by the measurement equipment by inductive voltage
transformers and inductive current transformers.
The recorded waveforms were then compared with simulations performed using a commercial software in order to
evaluate the model accuracy.

SIMULATION SETUP
The cable was modelled as being connected to an equivalent network modelled by an ideal voltage source and a
2372ej87.91° MVA short-circuit power.
For the cross-bondings a ground resistance of 3 Ω and a wire inductance of 1 μH were considered. All the minor
sections of the cross-bondings were modelled. The shunt reactors parameters were obtained directly from the
respective test report.
The simulated and measured waveforms are compared in the next pages.

Page 115
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

Energization Transient
Current
The energization of Horns Rev 2 cable is made using synchronised switching. As the three phases are connected
at voltage zero crossing, the initial DC component in the shunt reactor installed in the middle of the cable is virtually
maximal.
Figure A3-20 shows the currents in all the three-phases of the shunt reactor in the moments after its energization.
The value of the initial DC component is of approximately 400 A, a value virtually equal to the magnitude of the AC
component. Figure A3-21 shows that this DC component can also be observed in the currents registered at the
sending end. During the first 50 ms there is a high-frequency component that is superimposed on the 50 Hz
component, increasing the current amplitude.
There is good agreement between the simulated and measured voltage and current at the sending end during the
first moments of the transient. Approximately 10 ms after the energization, the simulation accuracy starts to reduce,
as observed by comparing the current waveforms.
The divergence between simulations and measurements after this instant is most likely because the model does
not include the lines adjacent to the energized node. It is a common practice in the transient overvoltage studies as
it causes more severe results. As a result, however, the reflections in the adjacent lines and nodes are not
reproduced in the simulation (see Section 2.7), which can result in the divergence between simulations and
measurements.

a) Current-Shunt Reactor-Connection b) Current-Shunt Reactor-Connection (Simulation)

1000 1000

500 500
Current [A]

Current [A]

0 0

-500 -500

-1000 -1000
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
time [s] time [s ]

Figure A3-20: Current in the three-phases of the shunt reactor in first 0.5s after the
connection; a) Measured b) Simulated

a) Current-Cable Sending End-Connection b) Current-Cable Sending End-Connection (Simulation)

500 500

0 0
Current [A]
Current [A]

-500 -500

-1000 -1000

-1500 -1500
0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2
time [s] time [s]

Figure A3-21: Three-phases currents in the sending end in the first 0.2s after the
connection; a) Measured; b) Simulated

Voltage
Due to the synchronised switching the voltage transients are very small, which can be seen in Figure A3-22.
When comparing voltage and current it can be seen that current is much more distorted due to the synchronised
switching, at voltage zero crossing.

Page 116
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

a) Voltage-Cable Sending End-Connection b) Voltage-Cable Sending End-Connection (simulation)

200 200

100 100

Voltage [kV]
Voltage [kV]

0 0

-100 -100

-200 -200
0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
time [s] time [s]

Figure A3-22: Voltage in the cable's sending end during its energization; a) Measured; b)
Simulated

Steady-state
Table A3-1 compares the peak voltage and current in the sending end, receiving end and shunt reactor. The
comparison is made between the measured results and the simulations.

Table A3-1: Comparison between the measured results and the simulations: peak voltage
and current in the sending end, receiving end and shunt reactor.
Voltage in the Voltage in the Current in the Current in the
sending end [kV] receiving end [kV] sending end [A] shunt reactor [A]
Measured 133 142.6 510.6 372.8
Simulated 132.2 134.4 523.4 369.3

The steady-state results show that:

- The voltage in the sending end is equal in simulations and measurements;
- The current in the sending end has lower amplitude and is more distorted in the measurements. The difference
between the peak values of the simulation and the measured results is of approximately 16 A or 3 %;
- The voltage in the receiving end is lower in the simulation than in the measurements. The difference between
the peak values of the simulation and the measured results is of approximately 8.3 kV or 6 %;
- The current into the shunt reactor is practically equal in simulation and measurements;

The simulation results are in good agreement with the measurements with the exception of the voltage in the
receiving end, which is higher in the measurements.
The current into the shunt reactor is practically equal in the simulations and measurements. The shunt reactor was
modelled according to the information provided in the test report, and it was validated against the measurement
results shown in the same test report. Therefore, as the current into the shunt reactor is the same in simulations
and measurements, the voltage in that point, at a distance of 55.4 km from the sending end, is also the same.
Consequently, it is concluded that, for steady-state condition, the land cable model is accurate and the most likely
reason for the difference in the receiving end voltage is the submarine cable model.
The inaccuracy in the submarine cable model may be caused by several reasons: modelling of the return path,
inaccurate permeability and/or resistivity values of the seabed/water, electric connectivity between the screens
[A10], [A11]. All these factors are difficult to model as the necessary data is normally unavailable.

Page 117
 POWER SYSTEM TECHNICAL PERFORMANCE ISSUES RELATED TO THE APPLICATION OF LONG HVAC CABLES

A3.3 Conclusion
The field measurements performed on the cable line for the offshore wind farm Horns Rev 2 show how cable
parameter calculations, and an EMT type cable model can be validated, as well as the modelling of the connection
as a whole, for system studies.
The field test results have already been used for model improvements and network model validation. The methods
of measurements can therefore be used for validation on other cable lines, as well as the results shown here can
be used for validation of new cable models.
For the purpose of parameter calculations and model validation, it is recommended to use an impulse to excite the
cable, while for verifying the equivalent network modelling for system studies, a full scale network voltage should
be used during energization transients.
The full scale measurement allows the validation of not only a cable model, but also the models of other equipment
installed, such as shunt reactors and transformers.

References
[A9] U. S. Gudmundsdottir, Modelling of long High Voltage AC cables in Transmission Systems. Fredericia,
Denmark: PhD. Thesis, Aalborg University, 2010, ISBN: 978-87-90707-73-6.
[A10] F. M. F. da Silva, Analysis and simulation of electromagnetic transients in HVAC cable transmission grids.
Fredericia, Denmark: PhD. Thesis, Aalborg University, 2011, ISBN: 978-87-89179-99-5.
[A11] K. Yamabuki, K. Kubori, “A transient analysis of a scaled model for a submarine cable connected with an
offshore wind farm”, Electric Power Systems Research 85 (2012) 59-63.
[A12] H. W. Dommel, EMTP Theory Book. 2nd ed. Microtran Power System Analysis Corporation, Vancouver,
British Columbia, Canada, 1996.

Page 118
ISBN: 978-2-85873-251-7