Received: 2 June 2022 Revised: 8 March 2023 Accepted: 2 April 2023 IET Renewable Power Generation

DOI: 10.1049/rpg2.12734

ORIGINAL RESEARCH

Representative modelling of very long HVDC cables

Josh Schipper1 Saijie Sim2 Quy Dang2 Radnya Mukhedkar1

1
Electric Power Engineering Centre (EPECentre), Abstract
University of Canterbury, Christchurch, New
A representative frequency dependent model (FDM) of an HVDC cable is important in
Zealand
2
ensuring HVDC systems are designed harmoniously. A model of the proposed and very
Sun Cable Pty. Ltd., Sydney, Australia
long, 4200 km, AAPowerLink cables is implemented in PSCAD/EMTDC. The suitability
of approximations required for efficient simulation are evaluated according to the antic-
Correspondence
Josh Schipper, Electric Power Engineering Centre ipated cable specifications. In practice, a submarine cable of this length has between 40
(EPECentre), University of Canterbury, to 80 field joints (FJ), which can be onerous to individually express and computationally
Christchurch, New Zealand.
expensive to model. Frequency domain analysis shows a minimum of 8 FJs is sufficient to
Email: [email protected]
represent the behaviour of the core voltage. The sensitivity of the FDM to cable parameters
and operating conditions is assessed. The FDM is applied to establishing single ended fault
location techniques. For a 4200 km cable these techniques can determine fault location to
within 200 km (≈5%). Additional methods are explored to refine fault location estimates.

1 INTRODUCTION the existing longest undersea power cables do not exceed

720 km [3], and are using the higher 500 to 525 kV voltage
Renewable energy sources, including solar and wind, are con- technology.
sidered key in displacing fossil fuel energy sources in the Australia has excellent solar resources, averaging at 4.71 kWh
electricity sector. High yield renewable energy resources tend per installed kW per day [4]. The availability of undeveloped
to be regional and typically remote from the demand centres. and vast expanses of flat land in Australia make it an ideal loca-
Wind and solar energy are also predisposed to be variable as tion for giga-scale solar farms. Sun Cable is developing projects
the power output of these resources is dependent on geophys- in the Northern Territory. The first tranche, known as the
ical and weather conditions. Such variability can be overcome Australia-Asia Power Link (AAPowerLink), is looking to trans-
through energy storage or an appropriate energy mix. HVDC port the solar energy as electricity, to space constrained, high
has established its position as an enabling technology [1, 2] in energy density demand centres in South East Asia, supporting
integrating renewable energy sources into established AC power the regions decarbonisation initiatives.
systems. Pioneering projects of this magnitude push the boundaries
There is significant untapped solar resource around the globe, of the established knowledge base and amongst other things
some of which is thousands of kilometres from demand cen- necessitate assessment and further development of validated
tres. Giga-scale projects (>10 GW solar farms), long connection system models. This paper addresses HVDC cable models of
lengths (>3000 km) over land and water with bulk power length > 3000 km as required for the AAPowerLink project.
transmission (>2 GW), are now being considered seriously. Such a model is key in ensuring representative simulation of
The maturity of power electronics and cross-linked polyethy- the system behaviour in steady state (including sub-synchronous
lene (XLPE) cables have made such ambitious projects both and super-synchronous frequencies), dynamic and transient
technically and economically possible. Sun Cable and Xlinks conditions, as well as for evaluating DC cable fault location
are developing significant projects exporting solar energy from techniques.
Australia and the Sahara to the ASEAN region and UK respec- The cable model is implemented using the PSCAD/EMTDC
tively. Giga-scale projects are a significant step up to projects Phase Domain FDM, which approximates electromagnetic phe-
in operation now. The longest overhead HVDC transmis- nomena and material properties. Approximations are checked
sion lines approach lengths of 2500 km, particularly in Brazil for the AAPowerLink cable through independent mathematical
and China where they transmit hydro generation. However, modelling and review of published results.

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is
properly cited.
© 2023 The Authors. IET Renewable Power Generation published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.

IET Renew. Power Gener. 2023;17:2471–2488. wileyonlinelibrary.com/iet-rpg 2471

 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2472 SCHIPPER ET AL.

In time domain simulations, it is difficult to monitor signals TABLE 1 Radial dimensions of the AAPowerLink HVDC cable
along the length of the cable unless the points of interest are Thickness Outer
modelled as nodes between cable sections. A frequency domain Layer (mm) radius (m)
simulation is developed in MATLAB to observe behaviour of
Conduction layer 30 0.030
key variables along the length of the cable. The MATLAB simu-
Conductor screen 2 0.032
lation accepts cable characteristics from the PSCAD/EMTDC
line constants programme. The two tools together are utilised Insulation 26 0.058
to understand the behaviour of discontinuities, such as FJs, on Insulation screen and water blocking tapes 2 0.060
voltages and currents throughout the length of the cable. Metallic sheath 3.3 0.0633
AAPowerLink model is implemented with converter sta- Inner jacket and bedding 5.3 0.0686
tions. Modular multi-level converters (MMC) are connected in
Armour layer 6 0.0746
a bipole configuration with a metallic return (MR) cable. The
Serving 5.5 0.0801
Darwin and Singapore AC networks are represented by voltage
sources behind an impedance.
A sensitivity analysis of cable parameter is performed
to establish the potential variation in the cable’s fre-
quency dependent behaviour through DC fault simulation in current, is based on the superposition of individual frequencies:
PSCAD/EMTDC. A DC fault is analogous to a step change,
as the cable transitions from a healthy state to a faulted state. A (𝜔 ) e−𝛾 (𝜔 )z+ j 𝜔t = A (𝜔 ) e−𝛼 (𝜔 )z e j (−𝛽 (𝜔 )z+𝜔t ) (1)
A step response provides a means of understanding the cable’s
natural response. where 𝛾(𝜔) is the frequency dependent propagation, which can
An unsynchronised fault location technique is developed for be separated into attenuation (real component) and wave num-
the AAPowerLink by identifying features in terminal transients ber (imaginary component). The strength of each component
unique to fault location. A wide range of fault locations and is expressed in A(𝜔) and is determined by the boundary condi-
fault resistances are simulated to demonstrate the ability of the tions at the sending end (z = 0) and the receiving end (z = l ) at
proposed method to locate any fault. a distance l from the sending end. Equation (1) represents the
This paper presents a phase domain frequency dependent forward travelling wave: a similar equation exists for the back-
cable model for the AAPowerLink project. The assumptions wards travelling wave with an opposite sign on the propagation
in its development have been analysed and robustly assessed function. The characteristic impedance, Zc (𝜔), completes the
through a sensitivity analysis. In conjunction with a converter solution by determining the ratio between voltage and current.
station model, a new unsynchronised fault location technique is By definition, the length of an HVDC cable does not feature
developed. in the calculation of per unit length series impedance and shunt
The paper is organised as follows. Section 2 explains the admittance. The cable constitution is influenced by the require-
development of the cable model. The modelling of FJs is ment to efficiently transmit power over the longer distance.
assessed in Section 3. Section 4 explains the converter, the sys- Such as the properties of the new 525 kV insulation technology
tem configuration, and the sequence of events that occur during [5]. Or requiring multiple cable sections to be joined by FJs due
a DC fault. Section 5 performs the sensitivity analysis. Section 6 to transport and manufacturing constraints (analysed in the next
develops the method of fault location, and Section 7 explores section). This section analyses the assumptions made in imple-
methods of refining fault location estimates. menting the FDM for the cable specified by the parameters in
Tables 1 and 2.
The per unit length series impedance and shunt admittance of
2 HVDC CABLE MODEL a cable consisting of a concentric core, sheath, and armour has
been systemised by Ametani [6] and is based on earlier work by
A transient model of a cable has to accurately represent the Schelkunoff [7]. Ref. [7] made three approximations in relating
propagation of electromagnetic waves in and around the con- Maxwell’s macroscopic equations to TL equations (1) in calcu-
ductors. A model based on a complete solution to Maxwell’s lating the azimuthal magnetic field strength in the insulation,
equations in three dimensions is intractable for complex cable the axial displacement currents and dielectric conduction are
geometries, and a finite element method requires a significant insignificant in comparison to current in the conductors and
computational burden not suitable for the proposed studies. effectively zero. (2) The dielectric properties of conductors are
This impasse is broken by approximating Maxwell’s equations negligible, because metals have a large conductance in compar-
and linking it to the transmission line (TL) equations, which ison to the permittivity. (3) In determining azimuthal magnetic
distributes series impedances and shunt admittances along the field strength in the conductor, the propagation constant, 𝛾, is
direction of propagation. It is assumed that the material prop- effectively zero in comparison
√ to the solid conductor’s intrinsic
erties and cable profile do not change in the direction of propagation constant, j 𝜔𝜎𝜇, where 𝜎 is the metal’s conduc-
propagation. The solution to the TL equations describes two tance and 𝜇 is its permeability. This is satisfactory because the
travelling waves: one in the forward direction and the other negated components in the three approximations are several
in the backward. Each travelling wave, for both voltage and orders of magnitude lower than their respective dominant com-
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHIPPER ET AL. 2473

TABLE 2 Electrical parameters of the AAPowerLink HVDC cable conducting layer does not show an appreciable effect on the
Parameter Quantity cable characteristics in the 0.1 Hz to 1.0 MHz frequency
range. For a similar cable, ref. [12] agrees with this outcome
Core conductor resistivity (Ω m) 2.0358 × 10−8
and indicates that the semi-conducting layer results in higher
Core conductor permeability 1 attenuations for frequencies above 1.0 MHz.
Insulation permittivity 2.3 A wide range of water depths are traversed by the AAPower-
Insulation permeability 1 Link cable, especially as it passes deep water from the Australian
Sheath resistivity (Ω m) 2.14 × 10−7 continental plate to the Eurasian plate. The cable depth is
determined by the bathymetry along the cable route. The earth
Sheath permeability 1
medium is primarily seawater at 0.2 Ω m resistivity [13].
Inner jacket and bedding permittivity 2.4
Dielectrics consist of electric dipoles that oscillate in the
Inner jacket and bedding permeability 1 direction of applied electric field. If the electric field is oscillating
Armour resistivity (Ω m) 2.29 × 10−7 at high frequencies, the dipoles are dragged out of phase, the rel-
Armour permeability 10 ative permittivity drops, and power losses increase in a process
Serving permittivity 2.4 of relaxation. Electric dipoles form by multiple mechanisms,
which have relaxation frequencies well above the frequency
Serving permeability 1
range of interest for transient studies, that is, 1 MHz. The ori-
entational polarisation of molecules within XLPE appears to
have a relaxation frequency above 100 kHz [14, 15]. It is diffi-
ponent for frequencies up to 1 MHz. This is confirmed in [8] cult to ascertain the exact relaxation frequency for XLPE from
by comparing solutions to TL equations with solutions from the literature. Given the already high attenuation of frequen-
Maxwell’s equations, showing a close agreement in the range of cies above 100 kHz over the length of the AAPowerLink cable,
frequencies required for transient studies. uncertainty in the exact orientational polarisation relaxation fre-
The major assumption made in [7] is that electric and quency is not likely to make an appreciable change to transient
magnetic field strength are independent of azimuth, that is, behaviour. Therefore, the representation of the complex dielec-
rotationally invariant around the z-axis. This assumption was tric properties of XLPE by a single constant relative permittivity
required to simplify Maxwell’s equations, which is invalidated in PSCAD/EMTDC is sufficient.
when asymmetry is introduced by the electric and magnetic The formulae that describe the series impedance and shunt
fields from another cable in close proximity. Within one cable, resistance often include Bessel functions or require numerical
skin and proximity effects are modelled between the core, integration. These equations are difficult to accurately compute,
sheath, and armour conductors in PSCAD/EMTDC. For two and instead approximate formulae are used. According to the
nearby cables, the proximity effect between cables is not EMTDC manual [16], the approximations of Bessel functions
modelled [9]. In practice, the proximity effect influences the derived from [17] should not exceed a 0.5% error.
resistance of cables in proportion to their distance of separa-
tion [10]. Alternative modelling techniques have accounted for
the proximity effect, such as the method of moments with a sur- 3 EFFECT OF FIELD JOINT
face admittance operator [9]. For the AAPowerLink, its cables EARTHING
(primarily one HV and MR) will be placed together for a propor-
tion of the length, and the model will not include the proximity Field joints (FJ) are a discontinuity to the characteristic
effect on the high frequency resistance. Cable proximity has no impedance of cables and can generate reflections and refrac-
effect on DC operation. In the sensitivity analysis (Section 5), tions of electrical waves during transients. The earthing of the
the proximity effect is indirectly assessed by considering a range sheath and armour is the most discontinuous, where reflected
of cable resistivities. waves in the sheath and armour can induce waves in the core
Due to conducting and insulating properties, the semi- conductor by inductive and capacitive coupling. Most electro-
conducting layers on either side of the main insulation can magnetic energy is directed through the insulation between the
modify cable behaviour. The semi-conducting layer has a higher core and sheath, where a small amount of energy is reflected
capacitance than the XLPE main insulation, because of its as it passes around the deflectors. Accurately modelling this
higher relative permittivity and smaller thickness. The overall effect requires finite element modelling of the constructed joint,
capacitance is calculated to include all three insulating lay- which is beyond the scope of this paper [18]. Because the deflec-
ers in series. The capacitance of the main insulation is lower tion is gradual, it is assumed that the reflection is minimal. This
than the semi-conducting layer by at least two orders of mag- section assesses the impact of field joints on cable voltage pro-
nitude, and thus determines the overall capacitance. At high files in the frequency domain, and determines an appropriate
frequencies, skin effect may push current into the inner semi- number of cable sections to represent the 4400 km cable in
conducting layer, thereby increasing the effective resistance. The PSCAD/EMTDC for fault analysis at multiple locations.
effect of the semiconducting layer is calculated according to the Factory joints can connect sheath and armour at regular
methodology of [11] in MATLAB using the proposed dimen- intervals determined by the production facilities (≈1–20 km).
sions and electrical properties in Tables 1 and 2. The semi- FJs connect armour and sheath conductors of two separately
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2474 SCHIPPER ET AL.

transported cable sections through a finite resistance to earth. equations, (2) is rearranged as:
This earth resistance is dependent on the size of the FJ, as the
exterior housing forms the earth electrode, and how the cable ⎡ Vi ⎤
and FJ are laid or buried in the seafloor. If earth resistance is low, [ ]⎢ ⎥
E −Zs Zm 0 ⎢ Ii,in ⎥
then it is possible to remove the sheath and armour conductors
⎢ ⎥=0 (4)
from the cable model via Kron reduction. Especially, if the 0 −Zm Zs E ⎢Ii+1,out ⎥
outer cover is permeable to sea water, so that the armour is con- ⎢ ⎥
⎣ Vi+1 ⎦
tinuously connected to earth. A comparison is made between
models that do and do not include Kron Reduction to assess the
appropriateness of the latter option in modelling the full cable. where Zs = TV Zc coth(𝜸l )TI −1 , Zm = TV Zc csch(𝜸l )TI −1 and
The 4400 km AAPowerLink cable will have somewhere E is the identity matrix.
between 44 and 74 cables sections, each of 100 to 60 km in Each FJ is modelled by a solid connection between the sheath
length. Note, an additional 200 km is added to the commonly and armour, which is then connected to an earth impedance, Ze ,
stated 4200 km for the AAPowerLink to accommodate for any of nominally 1.0 Ω. This forms the matrix equation:
variances. A full representation of 44 or more cable sections and
FJs in PSCAD/EMTDC is impractical. Compact models of the ⎡Ii,out ⎤
[ ]⎢ ⎥
full cable that reduce the number of FJs are simulated and com- ZFJ EFJ −ZFJ ⎢ Vi ⎥ = 0
pared to the full representation. Frequency domain analysis in ⎢ ⎥
this section shows a model with 9 sections and 8 FJ is sufficient ⎣ Ii,in ⎦
to represent a cable with 74 sections.
A linear system of equations for the voltages and currents ⎡1 0 0⎤ ⎡0 0 0⎤
at the nodes along the length of the cable is formed in MAT- ⎢ ⎥ ⎢ ⎥
ZFJ = ⎢0 0 0 ⎥ EFJ = ⎢0 1 −1⎥
LAB. The solution is used to construct the voltage and current ⎢ ⎥ ⎢ ⎥
profiles between two nodes by applying the differential equa- ⎣0 Ze Ze ⎦ ⎣0 0 −1⎦
tions of travelling wave propagation. The linear equations for
the two-port model of each cable section are: Similarly, the sheath and armour conductors at the terminals
are connected to an earth impedance, Ze . The Singapore termi-
V1 = TV Zc coth (𝜸l ) TI −1 I1,in − TV Zc csch (𝜸l ) TI −1 I2,out nal core conductor is connected to a resistive load, ZL , of 230.82
Ω. This is to model full load of 2000 A at a Darwin terminal
(2a) voltage of 525 kV. The core conductor at the Darwin termi-
nal is connected to a voltage source, V0 , behind an impedance,
Z0 , of 1.0 Ω. Accordingly, the termination conditions
are:
V2 = TV Zc csch (𝜸l ) TI −1 I1,in − TV Zc coth (𝜸l ) TI −1 I2,out [ ] [ ]
[ ] V1 V0
(2b) Darwin∶ ED −ZD =
I1,in 0
Equations (2a) and (2b) are matrix equations, where the volt- [ ]
ages (Vi ) and currents (Ii,in , Ii+1,out ) are three by one complex [ ] IN ,out
vectors, where each component refers to either the core, sheath, Singapore∶ ZR ER = 0
VN
or armour conductors. Each vector has a subscript referring to
a group of core, sheath and armour nodes, where the Darwin
where N is the number of nodes and the Singapore terminal
terminal is the first index. The transformation matrices, TV and
index number,
TI , are composed of the eigenvectors of ZY and YZ , respec-
tively, where Z is the per unit length series impedance and Y
is the per unit length shunt √ admittance of the cable. The prop-
⎡V0 ⎤ ⎡1 0 0⎤ ⎡−Z0 0 0⎤
agation values, 𝜸, are 𝛾 j = 𝜆 j , where 𝜆 j is an eigenvalue of ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
V0 = ⎢ 0 ⎥ ED = ⎢0 1 −1⎥ ZD = ⎢ 0 0 0⎥
both ZY and YZ . The hyperbolic functions are interpreted as ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
the diagonal matrix: ⎣0⎦ ⎣0 0 −1⎦ ⎣ 0 Ze Ze ⎦

coth (𝜸l ) = diag ([coth (𝛾1 l ) , coth (𝛾2 l ) , coth (𝛾3 l )]) (3)
⎡−1 0 0⎤ ⎡ZL 0 0⎤
⎢ ⎥ ⎢ ⎥
where l is the length of a cable section. The characteristic ER = ⎢ 0 1 −1⎥ ZR = ⎢ 0 0 0⎥
impedance, Zc , is expressed in its modal form as a diagonal ⎢ ⎥ ⎢ ⎥
⎣0 0 −1⎦ ⎣0 Ze Ze ⎦
matrix: Zc = diag(𝜸 )−1 TV−1 ZTI . To form the linear system of
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHIPPER ET AL. 2475

The complete linear system is generalised for the number of For high relative accuracy in sheath (and armour) voltages it is
cable sections in the following matrix equation: necessary to model the actual number of joints, compare results
against the 74 cable section models (Figure 2). The sheath (and
⎡ED −ZD ⎤ armour) voltage magnitude is insignificant and a minimum of 9
⎢E −Zs Zm ⎥ cable sections is sufficient for representing accurate behaviour
⎢ ⎥ of the core cable voltage.
⎢ −Zm Zs E ⎥ The magnitude and phase of the combined impedance of the
⎢ ⎥
⎢ ZFJ EFJ −ZFJ ⎥ cable and load as seen from the Darwin terminal over a fre-
⎢ ⎥ quency range from 0.1–100 Hz is shown in Figures 3 and 4,
⎢ E −Zs Zm ⎥ respectively. It is observed that the conclusions from the analy-
⎢ ⎥
⎢ ⋱ ⎥ sis of 10.72 Hz signals can be generalised across the frequency
⎢ ⎥ range. Counter-intuitively with a fixed resistive load (230.82 Ω)
⎣ ZR ER ⎦
at the Singapore terminal, the impedance magnitude drops at
higher frequencies as more power is lost in the cable.
⎡ V1 ⎤
⎢ ⎥ The results presented so far have allowed for all states of
⎢ I1,in ⎥ Kron reduction. Each state gives a sufficiently accurate repre-
⎢I ⎥ sentation of the core conductor voltage. Selection of a particular
⎢ 2,out ⎥ ⎡V ⎤
⎢ V ⎥ ⎢ 0⎥ state relies on secondary considerations, such as complexity,
⎢ 2 ⎥ ⎢0⎥ computational speed, and study scenarios. Sheath and armour
× ⎢ I2,in ⎥ = ⎢ ⎥ electrical connections are retained within PSCAD/EMTDC
⎢ ⎥ ⋮
⎢ I3,out ⎥ ⎢⎢ ⎥⎥ model to enable access to the various conductor layers for
⎢ ⎥ ⎣0⎦ simulation of events such as cable faults within a cable section.
⎢ ⋮ ⎥
⎢ ⎥
⎢IN ,out ⎥
⎢ ⎥ 4 CONVERTER STATION MODEL
⎣ 6VN ⎦
This section describes the modelling of the converter sta-
Three different numbers of cable sections are simulated: a tions to accurately simulate the overall system behaviour in
single section, 9 sections, and 74 sections. In addition, three PSCAD/EMTDC. The AAPowerLink (Figure 5) consists of
different states of Kron reduction are analysed: no conductors three cables: positive and negative HVDC cables, and a metal-
reduced (3C), only the armour reduced (2C), and both sheath lic return (MR) cable with smaller main insulation thickness.
and armour reduced (1C). A range of frequencies between 0.1 These cables form an MR bipole configuration, where the Dar-
and 100 Hz are analysed. For frequencies below 0.1 Hz, the win terminal MR is earthed. Each converter pole is a modular
induced current into the sheath and armour is minor and the multilevel converter (MMC) and is modelled with the PSCAD
number of FJs has no impact. For frequencies above 100 Hz example model [19, 20]. MMC model includes a converter trans-
the attenuation in all three conductors is too high for standing former, six arm reactances, half-bridge sub-modules, controller
waves to be produced in the sheath or armour. For 10.72 Hz, and modulator. The Darwin and Singapore AC networks are
the voltage profiles across the core and sheath are shown in represented by a voltage source behind an RRL impedance. On
Figures 1 and 2, respectively. 10.72 Hz is a sampled frequency the DC side of the converter, a DC reactor is included on all
from the PSCAD/EMTDC line constants program. It was cho- three cables and at both terminals. It is primarily included to
sen because it is one of many frequencies where the results differ analyse its impact on the shape of electrical wave reflections
between the three states of Kron reduction and the number of from faults.
modelled FJs. The voltage and current transients recorded at the cable ter-
The configuration of 74 sections (3C) is the expected voltage minals is a result of the natural response of the cable and the
profile for the cable, where the single and 9 section configura- operation of the converter. When analysing DC faults to per-
tions are approximations to it. The single cable section models form a sensitivity analysis of cable parameters and identifying
(3C 1 Sec. FD and 2C 1 Sec. FD) show the greatest deviation fault location techniques, it is necessary to differentiate the
from the expected (3C 74 Sec. FD) in Figures 1 and 2. While the cable’s natural response from converter side influences. There-
9 section (3C 9 Sec. FD) model shows closer agreement to the fore, the converter control model is briefly described under
expected core voltage (Figure 1, 3C 74 Sec. FD). Including Kron normal and recovery conditions. Then the sequence of protec-
reduction of sheath and armour further improves agreement to tion events is explained to show when the interactions between
the expectation. the converter and cable change.
The close agreement between the time domain results Converter control is separated into differential (AC side) and
of 9 section cable models (in the core, Figure 1) and fre- common (DC side) mode control, which are added together in
quency domain simulations in MATLAB shows that the the modulator to determine the switching state of each submod-
PSCAD/EMTDC conversion from the frequency domain to ule. The output current controller (differential mode current)
the time domain has a minimal build-up of numerical errors. is decoupled into d-axis and q-axis components. For both the
17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHIPPER ET AL.

Sheath voltage magnitude across cable length for a 10.72 Hz input. Voltage is normalised by the core voltage at the sending end source voltage.
Core conductor voltage magnitude across cable length for a 10.72 Hz input. Voltage is normalised by the source voltage.
FIGURE 1

FIGURE 2
2476
17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2477

Input impedance magnitude of an AAPowerLink HVDC cable from the sending with the receiving end terminated with a 230.82 Ω resistance.

Input impedance phase of an AAPowerLink HVDC cable.

SCHIPPER ET AL.

FIGURE 3

FIGURE 4
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2478 SCHIPPER ET AL.

FIGURE 5 AAPowerLink circuit configuration.

Darwin and Singapore terminals, the q-axis controls AC side model. The dimensions and electrical parameters of the cable
voltage magnitude. At Darwin, the d-axis controls real power may vary due to variances in material properties and tolerances
into the converter to maintain DC voltage, while at Singapore, in the manufacturing process and installation. Furthermore, the
it controls real power into the AC grid to match the reference. operational conditions, such as cable loading and ambient con-
Common mode (internal) control is a variation of direct volt- ditions affect the cable temperature and associated parameters
age control [21] to maintain submodule capacitor voltage, which temporarily. Some of these effects may become permanent over
includes the suppression of 2nd and 4th harmonic circulating time through the cable aging process.
currents. In addition to the PSCAD example model [19], the Although, a sensitivity analysis cannot describe all phenom-
internal controller is modified to aid capacitor voltage recovery ena suppressed in the approximation process, it can provide
in the healthy pole during and following clearance of a DC cable an insight into potential variations in the frequency dependent
fault on the opposite HVDC cable. characteristics, and define an envelope of operation upon which
The sequence of events that occur during a DC fault are the design of the HVDC controller, fault location methods etc.
as follows: the apparent resistance of the faulted cable from can be tested.
the perspective of the converter decreases, thereby resulting in Four parameters of the cable model are varied: the core
increased current in the cable, converter, and AC network. The conductor resistivity, main insulation thickness, main insulation
rated steady state current of the cable is 2.0 kA, but if one of the relative permittivity, and state of Kron reduction for the sheath
six arm currents exceeds 2.7 kA then protection is triggered. and armour. Results are presented for a DC fault 1015 km away
Firstly, the converter is blocked and the switches are protected. from the Darwin terminal, which will establish the need for
A 150 µs delay is modelled to measure current and send the calibration of the FDM.
blocking signal. At this time fault current still flows through the Variations in core conductor resistivity is dependent on core
converter via the free-wheeling diodes. The converter acts as an temperature, which is correlated with cable loading, and to
uncontrolled rectifier, which is identified by a 300 Hz ripple in a lesser extent seawater and seafloor temperature. Consider-
the transients. The fault current is cleared by the opening of the ing variance in core resistivities, also allows for the analysis of
AC circuit breaker. A further 20 ms is given for the AC circuit potential impacts caused by proximity effect between cables.
breaker to open, and current is extinguished in each phase at the Resistivity is varied in the range from 20◦ C (Table 2) to 70◦ C
current zero crossing. A current remains in the HVDC cable, as (2.4316 × 10−8 Ω m) core temperature. Higher core resistivity
the stored energy in the cable dissipates through the cable and increases attenuation and is expected to reduce the size of the
diode resistance. incoming travelling wave from the fault. Simulations show that
this effect is relatively minor (Figure 6).
The electrical parameters and radial dimensions of the cable
5 CABLE MODEL SENSITIVITY is based on published values and an initial proposal for the
ANALYSIS AAPowerLink as there are no 525 kV XLPE submarine cables
in operation by 2022. The insulation thickness is expected to
A sensitivity analysis of the AAPowerLink cable is conducted be between 26 and 30 mm [22], and will encompasses smaller
to understand the potential variation in frequency dependent variations from the manufacturing process. Based on the ideal
characteristics due to uncertainty in cable parameters. A num- theory of coaxial cables, a larger insulation thickness results in a
ber of aspects can contribute to the need to calibrate the cable higher characteristic impedance, but propagation speed remains
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHIPPER ET AL. 2479

FIGURE 6 The impact of different cable temperatures (core resistivity) on sending end terminal voltage in response to a DC fault on the positive HVDC
cable. Fault is 1015 km from the sending end terminal.

FIGURE 7 The impact of insulation thickness on the sending end terminal voltage in response to a DC fault on the positive HVDC cable.

constant. The transient simulation results of Figure 7 agree with occurs during the cable manufacturing process, which gives the
this expectation. polyethylene its final properties before operation. Small vari-
The relative permittivity of XLPE insulation is obtained from ations in manufacturing, variations in operating temperature
the manufacturer datasheets [23]. The cross-linking process [22], and accumulated age can result in variation of insulation
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2480 SCHIPPER ET AL.

FIGURE 8 The impact of the relative permittivity of the main XLPE insulation on the sending end terminal voltage in response to a DC fault on the positive
HVDC cable.

properties. A range of relatively permittivity from 2.0 to 2.6, when de-energised, such as time domain reflectometry (TDR),
centred around the nominal value of 2.3, is considered. Figure 8 the optical variant of TDR for embedded fibre optical cables,
shows the resulting difference in the propagation speed, as per and resonance approaches.
TL theory. An established online overhead line fault location method
Section 3 analysed different representations of FJ in the [24] has been tested for the 2375 km Rio Madeira HVDC,
AAPowerLink cable. Figure 9 compares the different states of where the time of arrival of the traveling-waves produced by a
Kron reduction for the 15-section cable model, and as expected fault [25] is recorded. This double-ended fault location method
from Section 3 there is minor variation in the transients. synchronises data acquisition with GPS at each terminal, and
Variations in parameters assessed in this section have shown estimates fault location by the difference in arrival times. Cal-
a small effect on the frequency dependent characteristics and ibration of propagation velocity and effective cable length is
fault transient recordings. Calibration can further refine the required, which can be performed by applying test faults at
FDM to improve the accuracy of fault location methods. This the terminals. Further details about this method are found
could be achieved through system identification of terminal in [26].
measurements during the energisation process and fast power If synchronisation between terminals is unavailable, then
ramps. single-ended fault location methods are possible. One method
records the times of arrival of multiple traveling-wave reflec-
tions at one terminal [27]. A Bewley lattice diagram is
6 FAULT LOCATION ANALYSIS constructed to understand the path of travel for each wave,
which accounts for all the reflections and refractions at the
The AAPowerLink cable model is used to develop fault fault and terminals. This method is successfully applied to
pre-location techniques based on terminal measurements. Pre- AC transmission line faults from current measurements [28],
location is the first stage in locating a fault, and provides a and is commercially available (SEL-411L [29] and the TWS
general vicinity for repair crews to begin pinpointing the fault. FL-8 [30]).
For an undersea cable, the accuracy of pre-location is crucial, as This section analyses unsynchronised online fault location
pinpointing requires a significant amount of time and resource. methods as they are most challenging for cables as long as the
Fault pre-location methods fall into two categories: online AAPowerLink project. Simulations will show that single-ended
and offline. Online methods are here defined by recording sig- method described above is ineffective for most fault locations,
nals during a fault, for example, the relative timing of signal as a second reflection is unidentifiable. Alternative single ended
wave fronts arriving at each of the terminals. An offline method methods that correlate the shape of the first travelling wave
analyses the response from a signal injected into a faulted cable front with fault location are studied [31, 32].
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHIPPER ET AL. 2481

FIGURE 9 The impact of the model configuration, where the outer layers of the cable can be removed by Kron reduction, on the sending end terminal voltage
in response to a DC fault on the positive HVDC cable.

The AAPowerLink cable model considered for this study and when the AC circuit breaker is opened is marked by a tri-
has 15 cable sections and can test 14 possible fault locations. angle and square, respectively. The effect of the fault appears at
A range of possible fault resistances and cable loadings are the terminal after a finite time depending on the distance from
simulated to test the robustness of the proposed method. DC the fault to the terminal. For faults far from the terminal, deter-
Reactors are varied in size to determine its influence on the mining the exact instance when the fault appears at the terminal
reflection of wave fronts. is a challenge, since voltage and currents gradually change, for
Instrumentation has a finite bandwidth and can limit the example, Figure 10 for a 4300 km fault.
capabilities of fault location. A recent review [33] has con- In practise, the time when a fault occurs is unknown, and
sidered different measurement technologies in the context of the relative appearance of a fault at both terminals is required
the demands of fault location methods. Travelling wave meth- for synchronised two-ended traveling-wave based fault location.
ods require a significant bandwidth, as high sampling rates are The standard method is to filter the transient with a wavelet [28]
required to precisely determine the time of arrival of a travelling and determine the appearance of the peak component.
wave. The unsynchronised fault location techniques proposed A single-ended method of fault location based on travelling
in this section record voltage. A commercially available resistive- wave records the times when wave fronts return back to the ter-
capacitive voltage divider [34] has a time constant of 33 µs. The minal upon reflection from the fault. These reflections are most
simulation time step is set to 50 µs to recognise this limitation. noticeable in the transient voltage oscillations of Figures 12
and 13 for faults within 600 km of the terminal. The time
between swings in voltage is the time the wave takes to travel
6.1 Base case simulations from the terminal to the fault and back again. When wave
velocity is known from calibration, the fault location can be esti-
The base case system configuration considers a load causing a mated. To a lesser extent, these wave reflections are visible in the
2 kA DC current in each pole, 20◦ C core conductor tempera- current transients of Figures 10 and 11.
ture, 26 mm insulation thickness, and a fault resistance of 0.01 Careful consideration of converter operation is required to
Ω between the core and sheath. The DC Reactor inductance is identify and discriminate wave reflections. When the converter
200 mH and the arm reactor inductance is 60 mH. is blocked a small notch occurs in the voltage waveform, which
The base case transient voltage and current at the Darwin may be miss identified as a reflection. Images of the blocking
(sending end) and Singapore (receiving end) terminals for DC action are also visible in the reflected waves.
faults are shown in Figures 10–13. Simulated faults occur at time The uncontrolled rectification before the AC circuit breaker
zero. The time when the corresponding converter is blocked opens creates a 300 Hz oscillation, which interferes with the
17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHIPPER ET AL.

Receiving end positive HVDC cable current (out of the cable) for faults with 0.01 Ω resistance at selected locations.
Sending end positive HVDC cable current (into the cable) for faults with 0.01 Ω resistance at selected locations.
FIGURE 10

FIGURE 11
2482
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHIPPER ET AL. 2483

FIGURE 12 Sending end positive HVDC cable voltage for faults with 0.01 Ω resistance at selected locations.

FIGURE 13 Receiving end positive HVDC cable voltage for faults with 0.01 Ω resistance at selected locations.

ability to identify smaller reflections, especially in the current additional characteristic to refine the fault location technique.
transients. Faults close to the terminal experience sharp initial changes
For faults located beyond 600 km from the terminal, a in voltage, while the faults further away demonstrate shallow
different method is required for fault location. changes.
The wave front created by a fault can be approximated by
a step response. In the frequency domain, a step comprises of
6.2 Impact of fault resistance a broad spectrum. When the wave travels along the cable, the
high frequency components of the step are attenuated faster
Correlation between the sharpness of the first wave reflection than lower frequency components. Dispersion also has a role
and the distance to the fault (Figures 12 and 13) suggests an in the process. A fault location method can be established when
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2484 SCHIPPER ET AL.

FIGURE 14 Sending end positive HVDC cable voltage for faults of varying fault resistance and located 1015 km from the sending end terminal. A high
voltage remains after the 100 Ω fault has cleared. This is because the Singapore positive pole converter did not trigger protection as arm current did not exceed
threshold. Consequently, current continues to flow through the fault from the receiving end, creating a potential at the sending end.

a relationship between fault distance and the rate of change of difference is called the time to inflection, which is plotted in
voltage is found. Figure 16 as a function of fault location and fault resistance
The proposed method must account for a range of con- from the simulations. The horizontal contours of Figure 16
ditions for reliable fault location. Differences in the pre-fault show that the time to inflection is invariant to fault resistance for
voltage can be removed by normalising the voltage transient faults located within 2500 km of the terminal of measurement.
to the corresponding pre-fault value. Identifying characteristics A similar plot to Figure 16 can be produced for the receiving
of transients invariant to fault resistance and obtaining a better end measurements to cover fault locations closer to the receiv-
definition of reflection sharpness requires further assessment. ing end. The vertical range of each contour determines the
A range of fault resistances from 0.01 to 100 Ω are simulated range of possible fault locations for a recorded time to inflec-
at all 14 fault locations. The impact of fault resistance on the tion, which is about 200 km for faults within 2200 km of the
transient voltage is shown in Figure 14. A higher fault resistance terminal.
results in a smaller electrical wave at the fault radiating out to the Fast times to inflection close to the voltage transformer time
terminal, which has an appreciable reduction in the magnitude constant of 33 µs are seen in faults close to the terminals, mak-
of the rate of change of voltage. ing the time to inflection point an inappropriate measure for
In theory for an electrical wave generated by a fault, the faults located within 600 km of the terminal.
attenuation of each constituent frequency is independent of
the fault resistance once the wave has departed from the fault.
Therefore, the shape of the transient voltage should be con- 6.3 Cable loading conditions
sistent across fault resistances except for scaling and vertical
translation. The single ended fault location method of [31] is DC faults are simulated for different pre-fault loading of the
based on this principle and fits an exponential function to the HVDC link to assess if the fault location method of Section 6.2
initial transient, where the decay rate is correlated with fault is affected. For 10, 200 and 2000 A DC current loading, simula-
location. tion results in Figure 17 show voltage transients are unaffected
The method proposed in this paper characterises the shape until the converter is blocked. Therefore, the time to inflection
of the initial voltage transient by the time between when the point method can be considered independent of cable loading.
fault first appears and the time of the maximum absolute rate The difference in blocking time is reflective of the additional
of change of voltage, as shown by ti in Figure 15. The time of time required for at least one converter arm current to reach the
largest rate change of voltage is an inflection point, and the time protection threshold (2.7 kA) from a lower initial current.
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHIPPER ET AL. 2485

FIGURE 15 Properties of initial voltage transient (green).

FIGURE 16 Time to inflection for transients recorded at the sending end terminal.

6.4 The effect of the DC reactor The converter appears more as a constant current source with
a larger DC reactor. Hence, the terminal appears as an open
The impact of the DC reactor in conjunction with the arm reac- circuit from the perspective of a reflecting wave, and a near per-
tors is to reduce the rate of change of current to within the fect image of the wave front is recorded at the terminal. The
equipment rating. In Figure 18, where DC reactor sizes of 2.0, smaller DC reactors in Figure 18 result in larger voltage dis-
20.0 and 200.0 mH are simulated, the slower rate of change of tortions in the incoming wave front from the fault before the
current for largest reactor further delays the blocking of the con- converter is blocked. A larger DC reactor is generally beneficial
verter. Thereby providing additional time separating the fault for fault location. However, the size of arm and DC reactors
wave (cable response to fault) and the blocking impulse wave is weighed against other engineering and economic consider-
(converter response to fault). ations that seek to minimise inductance. The effectiveness of
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2486 SCHIPPER ET AL.

FIGURE 17 The impact of cable loading on the sending end terminal voltage in response to a DC fault on the positive HVDC cable. The DC fault is located
1015 km from the sending end terminal.

FIGURE 18 The impact of DC reactor size on the sending end terminal voltage in response to a DC fault on the positive HVDC cable.

the time to inflection point method, for a fault 1015 km from 7 ALTERNATIVE CONCEPTS FOR
the terminal, is minimally affected by a small DC reactor, as FAULT LOCATION
the inflection point occurs before the converter is blocked. The
slight difference in the times to inflection can be accounted for As this paper has shown, there are at least two methods for
calibration. single-ended fault location based on electrical measurements at
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SCHIPPER ET AL. 2487

the terminals. These methods supply the fault location within a Validation; Visualization; Writing - original draft; Writing -
range of about 200 km (Section 6.2). The fault location estimate review & editing. Saijie Sim: Conceptualization; Funding acqui-
can be refined by two-ended methods based upon synchronising sition; Writing - review & editing. Quy Dang: Resources;
terminal measurements, and offline methods. Signals injected Writing - review & editing. Radnya Mukhedkar: Conceptualiza-
by offline methods are attenuated by the long length of the tion; Funding acquisition; Investigation; Methodology; Project
cable making reflections undetectable. This difficulty could be administration; Supervision; Validation; Writing - original draft;
overcome by placing offline fault location equipment along the Writing - review & editing.
cable length at regular intervals. The challenge is powering such
equipment over long distances under water. Ref. [35] suggests ACKNOWLEDGEMENTS
energy harvesting by superimposing a low frequency voltage The authors wish to acknowledge funding support from
onto the core conductor to induce a voltage between the sheath Ministry of Business, Innovation & Employment’s Advanced
and armour, where the equipment is connected. Alternatively, a Energy Transformation Programme. Grant Number: CONT-
medium voltage DC cable similar to communication cables [36], 69662-SSIFAETP-UOC. The authors also like to thank David
battery, or local power generation at the ocean surface can offer Griffin from Sun Cable for his input.
an external power source.
Offline fault location equipment placed along the cable CONFLICT OF INTEREST STATEMENT
requires fibre optic cable (FOC) to be embedded within the The authors declare no conflicts of interest.
HVDC cables so that optical time domain reflectometry can be
performed. Each section of FOC is terminated to the nearest FJ,
where a distributed device, such as a signal generator, receiver or DATA AVAILABILITY STATEMENT
repeater can be located. The FOC can monitor parameters such Data sharing not applicable—no new data generated, or the
as cable temperature in addition to fault location. article describes entirely theoretical research
Controlling and communicating with the distributed devices
is a challenge. A connection could be achieved through a spare ORCID
FOC core, or by transmission (acoustic, electric, or optical) up Josh Schipper https://orcid.org/0000-0003-3371-0917
to the ocean surface and then via satellite. This system will
REFERENCES
present several challenges, the foremost is the design of the FJ
1. Deane, P., Brinkerink, M.: Connecting the continents-A global power grid.
to pass FOC through the sheath.
IEEE Power Energy Mag. 18(2), 121–127 (2020)
2. Gellings, C.W.: A globe spanning super grid. IEEE Spectr. 52(8), 48–54
(2015)
8 CONCLUSION 3. National Grid: National Grid powers up world’s longest sub-
sea interconnector between the UK and Norway. (2022).
https://www.nationalgrid.com/national-grid-powers-worlds-longest-
The AAPowerLink is a proposed HVDC bipole link transport-
subsea-interconnector-between-uk-and-norway. Accessed 25 May 2022
ing solar energy from the Northern Territory in Australia to 4. Suri, M., Betak, J., Rosina, K., et al.: Global photovoltaic power potential by
Singapore. Electricity is to be transmitted through 525 kV cables country. Energy Sector Management Assistance Program (ESMAP), The
over a 4200 km distance, more than five times the length of World Bank Washington DC (2020)
any HVDC cable in operation. Approximations inherent to an 5. Gustafsson, A., Saltzer, M., Farkas, A., Ghorbani, H., Quist, T., Jeroense,
M.: The new 525 kV extruded HVDC cable system – World’s most pow-
implementation of a cable’s FDM in PSCAD/EMTDC was
erful extruded cable system. ABB Grid Systems. 2014 (Technical Paper).
assessed. The effect of FJ on the core conductor voltage pro- https://resources.news.e.abb.com/attachments/published/12792/en-
file, current profile, and input impedance, was simulated in the US/D6D3A232A090/The-new-525-kV-extruded-HVDC-cable-system-
frequency domain. The results determined that if electrical con- White-PaperFinal.pdf. Accessed 14 April 2023
nections of the sheath are explicit, then it is necessary to split 6. Ametani, A.: A general formulation of impedance and admittance of
cables. IEEE Trans. Power Appar. Syst. PAS-99(3), 902–910 (1980)
the cable into a minimum of 9 sections. Otherwise, Kron reduc-
7. Schelkunoff, S.A.: The electromagnetic theory of coaxial transmission lines
tion is sufficient to model the regular factory and field joints. and cylindrical shields. Bell Syst. Tech. J. 13(4), 532–579 (1934)
An MMC model at each terminal was included to understand 8. Theethayi, N., Baba, Y., Rachidi, F., Thottappillil, R.: On the choice
the cable interactions in a practical system. A sensitivity analy- between transmission line equations and full-wave maxwell’s equations
sis of the cable FDM to variations in cable parameters showed for transient analysis of buried wires. IEEE Trans. Electromagn. Compat.
50(2), 347–357 (2008)
the importance of calibrating the model through onsite tests.
9. Patel, U.R., Triverio, P.: MoM-SO: A complete method for computing the
The fault location method, time to inflection, was developed impedance of cable systems including skin, proximity, and ground return
for locating faults between 600 and 2200 km of the terminal to effects. IEEE Trans. Power Delivery 30(5), 2110–2118 (2015)
an accuracy of 200 km. In conjunction with established single- 10. Asada, T., Baba, Y., Nagaoka, N., Ametani, A., Mahseredjian, J., Yamamoto,
ended travelling wave fault location methods, which covers fault K.: A study on basic characteristics of the proximity effect on conductors.
IEEE Trans. Power Delivery 32(4), 1790–1799 (2017)
location within 600 km of the terminal, it is possible to locate
11. Ametani, A., Miyamoto, Y., Nagaoka, N.: Semiconducting layer impedance
faults along the entire cable length. and its effect on cable wave-propagation and transient characteristics.
IEEE Trans. Power Delivery 19(4), 1523–1531 (2004)
AUTHOR CONTRIBUTIONS 12. Ghosh, S., Ghosh, M., Das, S.: Effects of multiple semiconducting screens
Josh Nathanael Schipper: Conceptualization; Data cura- on line parameters and wave properties of underground cable. IEEE
Access 7, 169371–169384 (2019)
tion; Formal analysis; Investigation; Methodology; Software;
 17521424, 2023, 10, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rpg2.12734 by Test, Wiley Online Library on [25/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2488 SCHIPPER ET AL.

13. Mouhaidali, A., Chadebec, O., Silvant, S., Tromeur-Dervout, D., Guichon, lines-part 2: Algorithm and evaluation. IEEE Trans. Power Appar. Syst.
J.: Evaluation of HVDC cable impedance and admittance matrices by finite PAS-104(12), 3531–3537 (1985)
element method. In: 2018 IEEE International Conference on Industrial 28. Saha, M.M., Izykowski, J., Rosolowski, E.: Fault Location on Power
Technology (ICIT). Lyon, France (2018) Networks. pp. 1–422. Springer London (2010)
14. Das-Gupta, D.K., Scarpa, P.C.N.: Modeling of dielectric relaxation spectra 29. Schweitzer Engineering Laboratories, Inc. (SEL): SEL-411L Data
of polymers in the condensed phase. IEEE Electr. Insul. Mag. 15(2), 23–32 Sheet – Line Current Differential Protection Automation and Control
(1999) System, United States of America. (2022). https://cms-cdn.selinc.
15. Mugala, G.: High frequency characteristics of medium voltage XLPE com/assets/Literature/Product%20Literature/Data%20Sheets/
power cables. Ph.D. Thesis, KTH Royal Institute of Technology, Stock- 411L_DS_20220324.pdf ?v=20220413–225712. Accessed 25 May
holm, Sweden (2005) 2022
16. EMTDC: Transient Analysis for PSCAD Power System Simulation User 30. Qualitrol: TWS FL-8 and TWS FL-1 Traveling wave fault locators
Guide Version 4.6. https://www.pscad.com/knowledge-base/topic-94/ (Product Brochure) 2011. https://2hfybu1lrdue3%D7;9wnu1dvw7s-
v-. Accessed 14 April 2023 wpengine.netdna-ssl.com/wp-content/uploads/2016/09/IP-F21-02L-
17. Wedepohl, L.M., Wilcox, D.J.: Transient analysis of underground power- 01E_TWS_FL8_FL1.pdf (2011). Accessed 25 May 2022
transmission systems. System-model and wave-propagation characteristics. 31. Zhang, C., Song, G., Wang, T., Yang, L.: Single-ended traveling
Proc. Inst. Electr. Eng. 120(2), 253–60 (1973) wave fault location method in DC transmission line based on wave
18. Norouzi, Y., Braun, S., Frohne, C., Seifi, S., Werle, P.: Effect of cable joints front information. IEEE Trans. Power Delivery 34(5), 2028–2038
on frequency domain analysis. In: 2018 IEEE Conference on Electrical (2019)
Insulation and Dielectric Phenomena (CEIDP). Cancun, Mexico (2018) 32. Naidu, O., Pradhan, A.K.: A traveling wave-based fault location method
19. PSCAD: Modular multi-level converter (MMC). https://www.pscad.com/ using unsynchronized current measurements. IEEE Trans. Power Delivery
knowledge-base/article/234 (2022). Accessed 25 May 2022 34(2), 505–513 (2019)
20. Gnanarathna, U.N.: Efficient Modeling of Modular Multilevel HVDC 33. Tzelepis, D., Psaras, V., Tsotsopoulou, E., Mirsaeidi, S., Dyśko, A., Hong,
Converters (MMC) on Electromagnetic Transient Simulation Programs. Q., et al.: Voltage and current measuring technologies for high voltage
Ph.D. Thesis. University of Manitoba, Winnipeg, Canada (2014) direct current supergrids: A technology review identifying the options for
21. Sharifabadi, K., Harnefors, L., Nee, H.P., Norrga, S., Teodorescu, R.: protection, fault location and automation applications. IEEE Access 8,
Design, Control, and Application of Modular Multilevel Converters for 203398–203428 (2020)
HVDC Transmission Systems. 1st ed. IEEE Press, Wiley, Chichester, West 34. Trench: HVDC Divider – Compensated voltage divider for HVDC
Sussex, UK (2016) transmission systems. (2012). https://trench-group.com/wp-content/
22. Mazzanti, G., Marzinotto, M.: Extruded Cables for High-Voltage Direct- uploads/2019/03/DC-Voltage-Divider.pdf. Accessed 25 May 2022
Current Transmission: Advances in Research and Development. 1st ed. 35. Schipper, J., Sim, S., Dang, Q., Mukhedkar, R.: Representative modelling
John Wiley & Sons, Inc., Hoboken, NJ (2013) of very long HVDC cables. In: The 17th International Conference on AC
23. Borealis: Borlink LS4258DCE – Crosslinkable Insulation Com- and DC Power Transmission (ACDC 2021). Glasgow, United Kingodm
pound (Datasheet). https://www.borealisgroup.com/product/borlink- (2021)
ls4258dce/data-sheets (2022). Accessed 26 May 2022 36. Chesnoy, J.: Undersea Fiber Communication Systems. 2nd ed. Academic
24. Dewe, M.B., Sankar, S., Arrillaga, J.: The application of satellite time Press, London (2016)
references to HVDC fault location. IEEE Trans. Power Delivery 8(3),
1295–1302 (1993)
25. Soeth, A.B., de Souza, P.R.F., Custódio, D.T., Voloh, I.: Traveling wave fault
location on HVDC lines. In: 2018 71st Annual Conference for Protective
Relay Engineers (CPRE). College Station, TX (2018) How to cite this article: Schipper, J., Sim, S., Dang, Q.,
26. Bjorklund, H., Ni, J., Nordtorp, N.: Next Generation of Line and Cable Mukhedkar, R.: Representative modelling of very long
fault Locator for HVDC transmissions. B4-118, In: 2018 Cigre Paris
HVDC cables. IET Renew. Power Gener. 17, 2471–2488
Session. Paris, France. Cigre (2018)
27. Ando, M., Schweitzer, E.O., Baker, R.A.: Development and field-data eval- (2023). https://doi.org/10.1049/rpg2.12734
uation of single-end fault locator for two-terminal HVDV transmission