energies

Article
A Current Frequency Component-Based
Fault-Location Method for Voltage-Source
Converter-Based High-Voltage Direct Current
(VSC-HVDC) Cables Using the S Transform
Pu Zhao ID
, Qing Chen *, Kongming Sun and Chuanxin Xi
Key Laboratory of Power System Intelligent Dispatch and Control, Shandong University, Ministry of Education,
Jinan 17923, China; zhaopu106@163.com (P.Z.); skming.2008@163.com (K.S.); xichuanxin@126.com (C.X.)
* Correspondence: qchen@sdu.edu.cn; Tel.: +86-132-5616-9067

Academic Editor: Fengshou Gu

Received: 5 July 2017; Accepted: 28 July 2017; Published: 31 July 2017

Abstract: This paper proposes a fault-location method for voltage-source converter (VSC)-based
high-voltage direct current (VSC-HVDC) systems. This method relies on the current frequency
components generated by faults in the cable, and requires the arrival time of the frequency
components at two terminals. The S transform is a time–frequency analysis tool that is superior to the
wavelet transform in some respects. Therefore, the S transform was employed to determine the arrival
time in this paper. To obtain a reliable criterion, a novel phase-mode transform method for bipolar
cables was developed, and the propagation characteristics of the current frequency components
through out the cable were analyzed. A two-terminal VSC-HVDC system was modeled in power
system computer aided design/electromagnetic transients including DC (PSCAD/EMTDC). Various
faults under different conditions were simulated on the basis of this model, and the simulation
results verified a high accuracy, robustness against fault-resistance, and noise immunity of the
proposed method.

Keywords: current frequency component; DC cable; fault location; phase-mode transform;

the S transform; VSC-HVDC

1. Introduction
Voltage-source converter (VSC)-based high-voltage direct current (VSC-HVDC) is a new
transmission technology based on voltage-source converter topologies and fully-controlled
semiconductors. It can address not only conventional network issues, but also niche markets, such as
the integration of large-scale renewable energy sources with the grid and large onshore/offshore wind
farms [1,2]. Additionally, VSC-HVDC has the benefit of flowing a bi-directional power flow with ease,
and possesses the ability to create an electrical network [3]. The faults in DC cables may cause great
damage to the VSC-HVDC system. To ensure a safe DC fault extinguishment for VSC-HVDC systems,
a precise determination of the fault location is required.
The current fault-location techniques are mostly based on traveling wave theory [4,5].
The accuracy of the traveling wave-based fault-location methods is high. However, the feasibility
of these methods depends on the detection of the wave heads at the terminals of the transmission
line. If the measurements at the terminals of the transmission line fail to identify the wave heads,
the methods will be ineffective. To identify the wave head successfully, a high sampling frequency
of data and experienced technicians are required. Additionally, the traveling wave, after a DC fault
with a large fault resistance, is weak, and sometimes cannot be captured successfully. The wavelet

Energies 2017, 10, 1115; doi:10.3390/en10081115 www.mdpi.com/journal/energies

Energies 2017, 10, 1115 2 of 15

transform has the capability of time location and frequency location simultaneously during the signal
analysis [6,7]; thus, it is widely employed to detect the wave head. Many fault-location methods based
on the wavelet transform are proposed [8–10].
Some fault-location methods based on time-domain analysis have also been proposed.
A fault-location method for transmission lines using synchronized voltage measurements at both
endings of the line is described in [11]. Reference [12] proposes a fault-location method for VSC-HVDC
system transmission lines using an active impedance estimation technique. A fault-location method
is proposed in [13], which estimates the fault distance on the basis of the similarity of the post-fault
voltage signal to existing patterns. The fault-location algorithms proposed in [14] are based on the
distributed parameter model of the transmission line. An electrical circuit parameter evaluation-based
fault-location method for VSC-HVDC cables is proposed in [15]. The time-domain fault-location
methods can satisfy the system’s demand of a high reliability, but the accuracy of these methods is not
as high as the traveling wave-based fault-location methods.
The wavelet transform has made a great contribution to traveling wave-based fault-location
techniques. However, different types of signals require different wavelet bases, and it is not easy to
select a proper wavelet basis. The S transform was proposed by R. G. Stockwell in 1996 [16]. It is a
time–frequency analysis tool that is an extension of the wavelet and Fourier transforms. The S transform
can overcome the shortcomings of the wavelet transform, and additionally, its transform result for the
transient signal is more intuitive. In recent years, the S transform has been applied in many references
for power quality detection [17–20]. A fault-location method for AC transmission lines using the S
transform is proposed in [21]. However, the fault-location method for VSC-HVDC cables employing
the S transform has not been mentioned before. In this paper, a current frequency component-based
fault-location method for VSC-HVDC cables using the S transform is proposed. The feasibility and
applicability of the proposed fault-location method are researched deeply. The characteristics of the
traveling-wave in the DC cable are greatly different to those in the AC transmission line. To obtain a
reliable criterion for the proposed fault-location method, a phase-mode transform method for bipolar
DC cables was developed, and the propagation characteristics of the current frequency components in
the cable were analyzed in detail.
In this paper, a brief introduction of the S transform is given in Section 2, and the fault-location
principle is described. In Section 3, a phase-mode transform method for bipolar DC cables is proposed,
and the propagation characteristics of the current frequency components are represented. On the basis
of the analysis, a fault-location method employing the S transform for VSC-HVDC cables is proposed
in Section 4. In Section 5, the numerical simulation results and discussion based on a test system are
presented. Finally, a summary of the main conclusion and contributions of the paper are presented in
Section 6.

2. Fault-Location Principle

2.1. Brief Introduction of the S Transform

The S transform of a signal x(t) is expressed as [16]

 S(τ, f ) = +∞ x (t)ω (τ − t, f )e−i2π f τ dt

 R
−∞
f 2 ( τ − t )2 (1)
 ω (τ − t, f ) = √| f | e− 2
2π

where ω(τ −t, f) is the Gaussian window function with a time-shift factor of τ and a frequency of f. As
shown in Equation (1), the S transform is a time–frequency analysis tool, which can be derived from
the continuous wavelet transform (CWT) and the short-time Fourier transform (STFT).
Energies 2017, 10, 1115 3 of 15

The S transform of a discrete signal x[jT] is realized by the fast Fourier transform (FFT), and is
expressed as [16]
N −1 −2π 2 m2

 S[kT, n ] = ∑ X [ m+n ]e n2 e i2πmk
 N , n 6= 0
 NT NT
m =0 (2)
N −1
 S[kT, 0] = N1 ∑ X [ NT m
], n = 0


m =0

where k, m, n ∈ (0, N − 1). N is the sampling number, and T is the sampling interval. X[n/NT]
2 2 2
is the discrete Fourier transform of x[jT], and e−2π m /n is the Fourier spectrum of the Gaussian
window function.
As seen from Equation (2), the S transform result of a discrete signal is a complex matrix.
A row vector of the matrix represents the frequency-magnitude characteristic of the signal at a
specific moment, and the frequency interval is 1/NT, denoted by f 0 . A column vector represents
the time-magnitude characteristic of a frequency component with a specific frequency, and the time
interval is T. For example, we denote the S transform result by S, and S[k + 1, n + 1] represents the
frequency component with a frequency of nf 0 at the time kT.
The wavelet transform provides a cogent math tool for transient signal analysis. However,
the wavelet transform results for a signal may be different under different wavelet bases and scales;
thus, it is imperative to select the proper wavelet basis and scale for a specific signal. The Fourier
transform has a window with a fixed height and width; thus, it cannot adjust the time–frequency
resolution with variations of the time and frequency. The S transform inherits the advantages of the
wavelet transform and the Fourier transform, and can overcome both of their shortcomings. The S
transform is perfectly reversible. The phase of the S transform referenced to the time origin provides
useful and supplementary information regarding spectra that is not available from locally referenced
phase information in the wavelet transform [16]. By means of the S transform, all the frequency
components with a frequency interval of f 0 can be extracted, and the proper frequency components
can be selected conveniently. Additionally, the S transform results are more intuitive than the wavelet
transform results, which will be discussed in detail in Section 5.7. The height and width of the Gaussian
window can change with a variation of the frequency; thus, the S transform can adjust its time and
frequency resolution ability with respect to the frequency.
In this paper, the S transform was realized by means of matrix laboratory (MATLAB). We take a
series of signals that contained fault data at 0 ms as an example. The frequency components of the
signal were extracted by the S transform, and the time-magnitude curves of some typical frequency
components are shown in Figure 1. As can be seen, when the frequency is lower than 500 Hz,
the time-magnitude curves of the frequency components are smooth, but they are different from each
other. This indicates that, in the low-frequency range, the frequency resolution ability of the S transform
is remarkable, whereas the time resolution ability is not very good. When the frequency is greater
than 2.5 kHz, the time-magnitude curves are concentrated at the fault time; thus, the transient fault
signal can be represented intuitively. This indicates that the time resolution ability of the S transform is
improved in the high-frequency range. For fault identification and location, a good time resolution
ability is more critical. Therefore, due to the satisfactory time resolution ability of the S transform in the
high-frequency range, the high-frequency components were employed for fault location in this paper.

2.2. Estimation of the Fault Distance

After the occurrence of a fault in the DC cable, the fault current frequency components at the
fault point start traveling toward both terminals. Assuming a fault occurs in the cable at a time t0 ,
shown in Figure 2, and the fault point is at a distance l1 from the rectifier-terminal and a distance l2
from the inverter-terminal, for a fault frequency component with a specific frequency, if the frequency
components can be selected conveniently. Additionally, the S transform results are more intuitive
than the wavelet transform results, which will be discussed in detail in Subsection 5.7. The height
and width of the Gaussian window can change with a variation of the frequency; thus, the S
transform can adjust its time and frequency resolution ability with respect to the frequency.
Energies 2017, 10, 1115
In this paper, the S transform was realized by means of matrix laboratory (MATLAB). We4 take of 15

a series of signals that contained fault data at 0 ms as an example. The frequency components of the
Energies 2017, 10, 1115 4 of 15
signal were
Energies
component extracted
2017, 10, 1115
arrives byrectifier-terminal
at the the S transform, at and the time-magnitude
a time curves of some
t1 and at the inverter-terminal at atypical
time t2frequency
, the4fault
of 15
components
time canFigure are
be calculatedshown asin Figure 1. As can be seen, when the frequency is lower than 500 Hz, the
Figure 1.
1. The
The magnitude
magnitude of
of the
the frequency
frequency components
components L
of
of aa signal
signal extracted
extracted by
by the
the S
S transform.
transform.
time-magnitude curves of the frequency components (t1 + t2 −are v)
smooth, but they are different from each
t0 = (3)
other.
2.2. This indicates that, in the low-frequency range,
2 the frequency resolution ability of the S
2.2. Estimation
Estimation of of the
the Fault
Fault Distance
Distance
transform
where is remarkable,
v is the propagationwhereasvelocitythe time
of the resolution
frequency ability is not
component, andvery good.
L is the When
total length theoffrequency
the cable.
is After
greater
After the
than
the occurrence
2.5 kHz,
occurrence of
the
of aatime-magnitude
fault
fault in
in the
the DC
DC cable,
curves
cable, the
are
the fault current
concentrated
fault current frequency
at the
frequency components
fault time;
components at
at the
thus, the
Thus, the distance between the fault point and the rectifier-terminal can be calculated as
fault
fault point
transient
point start
fault
start traveling
signal can be
traveling toward both
represented
toward terminals.
intuitively.
both terminals. Assuming
This indicates
Assuming a fault
a fault occurs
that in
the time
occurs the
in the cable
resolution at a time
time ttof
cable at aability 0,
0,
shown
shown in
the S transformFigure
in Figure is 2, and the
2, improved fault
and the fault point
in the
point is at a
high-frequencydistance l 1from the rectifier-terminal and a distance l2
v(t1 −range.
is atl1a=distance tl01)fromFor thefault identificationand
rectifier-terminal anda location,
distance(4) la2
from
from the
the inverter-terminal,
good time resolution ability isfor
inverter-terminal, more
for aa fault frequency
critical.
fault Therefore,
frequency component
due to thewith
component aa specific
satisfactory
with timefrequency,
specific if
if the
resolution ability
frequency, the
frequency
of theAsS seen
frequency component
from Equations
transform
component in thearrives(3)at the
atand
the(4),
high-frequency
arrives rectifier-terminal
torange,
estimate thethe
rectifier-terminal at
at aa distance,
fault time
time tt11 and
high-frequency the at the
the inverter-terminal
arrival
components
and at time
were andemployed
propagation
inverter-terminal at
atforaa
time t
velocity2, the
of
fault tlocation
time fault
the
2, the fault
time
frequency
in this can
timepaper. be calculated
component
can be calculated as as
should be obtained.

L
(( tt 1 +
+ tt 22 −
L
− v ))
tt 0 = 1 (3)
(3)
= v
0 2
2

where
where v v is
is the
the propagation
propagation velocity
velocity of
of the
the frequency
frequency component,
component, and and L
L is
is the
the total
total length
length of
of the
the
cable. Thus, the distance between the fault point and the rectifier-terminal can be calculated
cable. Thus, the distance between the fault point and the rectifier-terminal can be calculated asas

ll1 =
=vv((tt11 −
−tt00 )) (4)
(4)
1

As
As seen
seen from
from Equations
Equations (3)
(3) and
and (4),
(4), to
to estimate
estimate the
the fault
fault distance,
distance, the
the arrival
arrival time
time and
and propagation
propagation
velocityFigure
of the1. The magnitude
frequency of the
component frequency
should components
be obtained. of a signal extracted by the S transform.
velocity of the frequency component should be obtained.

Figure 2.
2. Propagation
Figure 2.
Figure Propagation of
of the
the fault components.
fault components.

3. Analysis
3. Analysis of
of the
the Current
Current Frequency Components
Frequency Components
As can
As can be
be seen
seen from
from Equations
Equations (3)
(3) and
and (4),
(4), the
the proposed
proposed fault-location
fault-location method
method is based on
is based on the
the
current frequency
current frequency components
components in the DC
in the DC cable.
cable. To obtain
obtain aaa convincing
To obtain convincing criterion,
convincing criterion, the characteristics
criterion, the characteristics
characteristics
of the current frequency components should be analyzed first. We take a
of the current frequency components should be analyzed first. We take a VSC-HVDC systemVSC-HVDC system as
as an
an
example, whose schematic is shown in Figure 3. The two VSC stations are connected
example, whose schematic is shown in Figure 3. The two VSC stations are connected with a bipolar with a bipolar
cable. The
cable. The AC
AC systems
systems connected
connected to
to VSC1
VSC1 andand VSC2
VSC2 are
are modeled
modeled by two 420
by two 420 kV
kV generators.
generators. On the
On the
DC side, p1, p2, n1 and n2 are four relays taken
relays taken
DC side, p1, p2, n1 and n2 are four relays as
taken as protection
as protection devices
protection devices for
devices for the
for the cables.
the cables.
cables.

Figure
Figure 3.
3.3.Schematic
Schematic of
of the voltage-source
the of
voltage-source converter-based
converter-based high-voltage
high-voltage direct
direct current (VSC-HVDC)
currentdirect
(VSC-HVDC)
Figure Schematic the voltage-source converter-based high-voltage current
system.
system.
(VSC-HVDC) system.

3.1.
3.1. Phase-Mode
Phase-Mode Transform Method for Bipolar Cables
3.1. Phase-Mode Transform Method for
Transform Method for Bipolar
Bipolar Cables
Cables
The cross
The cross section
cross section of
of aaa bipolar
section of bipolar DC
bipolar DC cable
DC cable is
cable is shown
is shown in
shown in Figure
in Figure 4.
4. A
Figure 4. A single cable is composed of three
The A single
single cable
cable is
is composed
composed of of three
three
layers
layers from
from the
the inner
inner to
to the
the outside:
outside: a
a conductor
conductor layer,
layer, a
a sheath
sheath layer
layer and
and an
an armor
armor layer.
layer. Therefore,
Therefore,
layers from the inner to the outside: a conductor layer, a sheath layer and an armor layer. Therefore,
for
for aa bipolar
a bipolar cable,
bipolar cable, there
cable, there
there areare six
are six sets
six sets of
sets of parameters,
of parameters, which
which are
parameters, which are coupled
coupled with
with each other;
each other; thus,
thus, it
other; thus, it is
is
for are coupled with each it is
necessary to decouple the parameters of the cable. The decoupling methods for
necessary to decouple the parameters of the cable. The decoupling methods for the three-phase AC the three-phase AC
cable
cable are
are represented
represented in in [22,23].
[22,23]. Similarly,
Similarly, the
the decoupling
decoupling method
method for
for the
the bipolar
bipolar DC
DC cable
cable can
can bebe
derived; it is described in this subsection.
derived; it is described in this subsection.
Energies 2017, 10, 1115 5 of 15

necessary to decouple the parameters of the cable. The decoupling methods for the three-phase AC
cable are represented in [22,23]. Similarly, the decoupling method for the bipolar DC cable can be
derived; it is10,described
Energies 2017, 1115 in this subsection. 5 of 15
Energies 2017, 10, 1115 5 of 15

Figure 4. The cross section of a DC cable.

Figure
Figure 4. The cross
4. The cross section
section of
of aa DC
DC cable.
cable.
The schematic diagram of the bipolar DC cable is shown in Figure 5. The output terminal of the
The station
rectifier schematic diagram by
is denoted of the
1 andbipolarthe DC output cableterminalis shownofinthe Figure 5. The
inverter output
station is terminal
denotedofbythe 2.
rectifier
rectifier station
station isis denoted
denoted byby1 1
and andthe the
output output terminal terminalof
Taking terminal 1 as an example, the DC voltages of the conductor layer, shielding layer the of the inverter
inverter station station
is is
denoted denoted
byand by 2.
2. Taking
armor
Taking
terminal
layer forterminal
1the 1 as an
as positive
an example, example,
cablethe
areDC the DC
voltages
denoted byvoltages
of
up1cthe , up1s of the
conductor
and up1aconductor layer,
layer, shielding
, respectively. Theshielding
layer andlayer
DC voltagesarmor and
of thearmor
layer for
three
layer
the for
positivethe positive
cable are cable are
denoted denoted
by
layers for the negative cable are denoted u , u by uand
p1c p1sby un1c, p1ap1c , u u p1s and
, u , respectively.
respectively.
p1a The DC The DC
voltages voltages
of
un1s and un1a, respectively. The DC currents of the sixthe of the
three three
layers
layers
for the for
layers at the negative
negative
terminalcable cable
are
1 are are denoted
denoted
denoted by un1c
byip1c ,by
, ip1s u, n1sun1c
ip1a i,n1cu,n1s
,and u and
n1a
in1s and un1ain1a, ,respectively.
, respectively. The DC
respectively. The DC currents
currents of the
of the six six
layers
layers
at at terminal
terminal 1 are denoted
1 are denoted by ip1c ,by
ip1sip1c
, i,p1a
ip1s, ,in1c
ip1a, in1sn1c, iand
n1s and
in1ai,n1a , respectively.
respectively.

Figure 5. The schematic diagram of the bipolar DC cable.

Figure 5. The schematic diagram of the bipolar DC cable.
Figure 5. The schematic diagram of the bipolar DC cable.
Defining R, L, G and C as the matrices of the resistance, inductance, conductance and
Definingper
capacitance R, meter
L, G and of C as
the DC the matrices
cable, of the resistance,
respectively, then inductance,
the wave conductance
equations with and
mutual
Defining
capacitance R, L, G
perbemeter and C as the
of theasDC matrices
cable, of the resistance, inductance, conductance
respectively, then the wave equations with mutual and capacitance
inductance
per meter of can
thebeDCrepresented follows
cable, respectively, [22]:
then the wave equations with mutual inductance can be
inductance can represented as follows [22]:
represented as follows [22]: ( ∂U ∂I
= RI
−RI1 1−−L
∂U1 1 ∂I
∂ U=1 − L∂∂tI1 1
1 ∂=x − ∂t1 1
∂I
∂x
=GU
−RI1 − − LC ∂U (5)
∂ x 1 ∂ tU (5)
∂x ∂ I
 1 = −GU − C 1 ∂∂t
(5)
where U1 = [up1c , un1c , up1s , un1s , up1a , un1a ∂I1 I = [i1 , i∂U1,i , i , i , i ]. The key to realizing
], ∂and
x = −1GU1 p1c − C n1c ∂t p1s n1s p1a n1a
the decoupling is to transform R, L, G and  ∂C x to diagonal ∂t
matrices. According to this principle, the
phase-mode
where U1 = [u transform matrix
p1c, un1c, u p1s, un1sof
, uDC
p1a, uvoltages
n1a], andcan I1 =be derived
[ip1c , in1c,ip1sand
, in1srepresented as follows:
, ip1a, in1a]. The key to realizing the
where U 1 = [up1c, un1c, up1s, un1s, up1a, un1a], and I1 = [ip1c, in1c,ip1s, in1s, ip1a, in1a]. The key to realizing the
decoupling is to transform R, L, G and C to diagonal matrices. According to this principle, the
decoupling
phase-modeistransform
to transformmatrix R,ofL,DC
G voltages
and C tocan diagonal
be derived matrices. According as
and represented to follows:
this principle, the
phase-mode transform matrix of DC voltages can be derived and represented as follows:
Energies 2017, 10, 1115 6 of 15

 
I I I
P = k⊗ I I 0  (6)
 
I 0 0

where 0 and I are the two-dimensional zero matrix and unit matrix, respectively, while k is the
Karenbauer transform matrix, expressed as
" #
1 1
k= (7)
1 −1

Introducing Q = P−T as the phase-mode transform matrix of DC currents, then Equation(5) can
be transformed to (
∂U1m −1 −1 ∂I1m
∂x = −P RQI1m − P LQ ∂t (8)
∂I1m −1 −1 ∂U1m
∂x = −Q GPU1m − Q CP ∂t

where I1m is expressed as

 
i p1c + in1c + i p1s + in1s + i p1a + in1a

 i p1c − in1c + i p1s − in1s + i p1a − in1a 

−1
 i p1c + in1c + i p1s + in1s 
I1m = Q I1 =  (9)
 
i p1c − in1c + i p1s − in1s

 
 
 i p1c + in1c 
i p1c − in1c

Through the operation above, the currents in the bipolar cable can be transformed to six
independent mode currents. In the following, the current I1m [p] is denoted as the pth mode current,
where p = 1, 2, . . . , 6.

3.2. Propagation Characteristics of the Current Frequency Components

The attenuation constant matrix of the mode currents through the cable can be obtained as [22]
√
A = real ( Zm Ym ) (10)

where Zm and Ym are the impedance matrix and admittance matrix after decoupling, and real() is the
function to extract the real part of a complex number. It is easy to see that A[p, p] is the attenuation
constant of the pth mode current.
As seen from Equation (8), the resistance, inductance, conductance and capacitance matrices
after decoupling are expressed as Rm = QT RQ, Lm = QT LQ, Gm = PT GP and Cm = PT CP, respectively.
Denoting Rm [p, p], Lm [p, p], Gm [p, p] and Cm [p, p] by Rp , Lp , Gp and Cp , respectively, then the
propagation velocity of the pth mode current can be calculated as [22]
ω
vp = s  r (11)
 
1 2 2 2 2 2 2 2
2 ω L p Cp − R p Gp + R p + ω L p Gp + ω Cp

where ω is the angular velocity.

Figure 6a,b shows the attenuation constants and propagation velocities of the mode currents with
respect to the frequency, respectively. In Figure 4, the line marked as “p” represents the attenuation
constant or the propagation velocity of the pth mode current, where p = 1, 2, . . . , 6. As seen from
Figure 6a, the attenuation constants of the fifth and sixth mode currents stay low with the increase of
the frequency, while the attenuation constants of the other mode currents increase markedly. As seen
from Figure 6b, the velocities of the fifth and sixth mode currents are close to that of the traveling wave
 Figure 6a,b shows the attenuation constants and propagation velocities of the mode currents
with respect to the frequency, respectively. In Figure 4, the line marked as “p” represents the
attenuation constant or the propagation velocity of the pth mode current, where p = 1, 2, …, 6. As
seen from
Energies 2017, Figure
10, 1115 6a, the attenuation constants of the fifth and sixth mode currents stay low with 7 ofthe
15
increase of the frequency, while the attenuation constants of the other mode currents increase
markedly. As seen from Figure 6b, the velocities of the fifth and sixth mode currents are close to that
through the DC wave
of the traveling cable when
through thethe
frequency
DC cableiswhenhigherthethan 5 kHz, whereas
frequency is higherthe velocities
than of the other
5 kHz, whereas the
mode currents are significantly lower. According to Equations (3) and (4), although
velocities of the other mode currents are significantly lower. According to Equations (3) and (4), influenced by the
high propagation
although influenced velocity,
by thethe fault-location
high propagationaccuracy
velocity,can
thebe improved by
fault-location increasing
accuracy can the sampling
be improved
rate
by increasing the sampling rate of the current signal. However, if the attenuation constant signal
of the current signal. However, if the attenuation constant is too large, the fault current is too
will bethe
large, toofault
weakcurrent
to be detected, andbe
signal will thustoothe fault-location
weak method
to be detected, andwill
thus fail.
theTherefore, the fifth
fault-location and
method
sixth mode
will fail. currentsthe
Therefore, arefifth
moreandsuitable for fault
sixth mode location.
currents are According to the
more suitable forreference direction
fault location. of the
According
currents,
to the reference direction of the currents, as in Figure 5, the sixth mode current was selected asthis
as in Figure 5, the sixth mode current was selected as the fault-location parameter in the
paper. To obtain
fault-location the sixth mode
parameter current,
in this paper.theTocurrents
obtain ofthe
thesixth
conductor
modelayers in both
current, thepoles are needed,
currents of the
as in Equation
conductor (9).in both poles are needed, as in Equation (9).
layers

Figure 6.
Figure 6. Attenuation
Attenuation constants
constants and
and propagation
propagation velocities
velocities of
of the
the mode
mode currents.
currents. (a)
(a) Attenuation
Attenuation
constants. (b) Propagation velocities.
constants. (b) Propagation velocities.

4. Fault-Location Method
4. Fault-Location Method

4.1. Extraction
4.1. Extraction and
and Processing
Processing of
of Frequency
Frequency Components
Components
It has
It hasbeenbeen mentioned
mentioned in Section
in Section 2 that2the that theresolution
time time resolution
ability of ability of the Ssignificantly
the S transform transform
significantly
improves withimproves
an increasewith an frequency.
of the increase ofAccording
the frequency. to FigureAccording
1, when to theFigure
frequency1, when the
is higher
frequency is higher than 10 kHz, the time resolution ability of the S transform
than 10 kHz, the time resolution ability of the S transform is satisfactory. Additionally, as seen from is satisfactory.
Additionally,
Figure as seen from
6b, the velocities Figure
of the 6b, thecomponents
frequency velocities of(10–20the frequency
kHz) of thecomponents
sixth mode (10–20
currentkHz)areof the
close
sixth mode current are close to each other; thus, it can be considered that they
to each other; thus, it can be considered that they arrive at one of the cable terminals at the same time. arrive at one of the
cabletoterminals
Due at the
the analysis same the
above, time. Due to the
frequency analysis above,
components (10–20the frequency
kHz) components
of the sixth (10–20 were
mode current kHz)
of the sixth
selected modelocation
for fault currentinwere
this selected
paper. Byfor fault of
means location in this paper.
the S transform, By means of thecharacteristic
the time-magnitude S transform,
the time-magnitude characteristic of the frequency components
of the frequency components (10–20 kHz) of the sixth mode current at the cable terminals (10–20 kHz) of the sixthcan
mode be
current at the cable terminals can be extracted. Denoting the summation
extracted. Denoting the summation of their magnitudes at the time kT as M, then M is expressed as of their magnitudes at the
time kT as M, then M is expressed as
M = ∑ abs(S6 (k + 1, n + 1)) (12)
= n≤
Mn1 ≤ n2 ab s ( S 6 ( k + 1 , n + 1))
n ≤n≤n
1 2
(12)
where the matrix S6 is the S transform result of the sixth mode current signals, and n1 f 0 = 10 kHz and
nwhere
2 f 0 = 20 matrix S6 is the S transform result of the sixth mode current signals, and n1f0 = 10 kHz and
thekHz.
n2f0 = 20 kHz.
4.2. Determination of the Arrival Time
4.2. Determination of the Arrival Time
When the fault frequency components (10–20 kHz) first arrive at the cable terminals, they will
causeWhen the firstthesudden change points
fault frequency of M. Therefore,
components according
(10–20 kHz) totheat
first arrive first
thesudden change point
cable terminals, M,
theyofwill
the
cause firstthe
arrival time of the
first sudden frequency
change pointscomponents can beaccording
of M. Therefore, determined. The
tothe sudden
first suddenchange point
change of M
point of
can be extracted through the following equation:
(
dM
dt > 0, t < tm
dM , and M(tm ) > Mh (13)
dt < 0, t > tm
of M can be extracted through the following equation:

 dM
 dt > 0, t < tm
 , and M(tm ) > Mh (13)
Energies 2017, 10, 1115  dM < 0, t > t 8 of 15
 dt m

where ttmm is the

where the sudden
sudden change
changetimetimeofofM, andMM
M,and h is
h isthe
thethreshold
thresholdvalue MM
ofof
value to to
avoid thethe
avoid influence
influenceof
disturbances. The first sudden change point of M can be determined easily because
of disturbances. The first sudden change point of M can be determined easily because its magnitude its magnitude is
usually
is usuallymuch higher
much higherthan
thanthose of of
those thethe
subsequent
subsequent change
changepoints.
points.
We
We denote
denotethethesixth
sixthmode
mode currents
currentsat the
at therectifier-terminal
rectifier-terminalandandinverter-terminal
inverter-terminalby i1m6 i1m6 iand
byand 2m6 ,
respectively. The value of M calculated from
i2m6, respectively. The value of M calculated from i is denoted by
1m6 i1m6 is denoted by M , and the value of M calculated
1 M1, and the value of M calculated from
ifrom
2m6 isi2denoted by M
m6 is denoted 2. M
by The arrival
2. The timest
arrival 1 and
timest 1 andt2 tcan
2 canbebeobtained
obtainedthrough
throughthethefirst
firstsudden
sudden change
points of M11 and M22,, respectively.
respectively.

4.3. Flow Chart

4.3. Flow Chart of
of the
the Fault-Location
Fault-Location Method
Method
After
After determining
determining the the arrival
arrival timest and tt22,, the
timest11and the fault
fault distance
distance can
can bebe estimated
estimated through
through
Equations (3) and (4). The propagation velocities of the frequency
Equations (3) and (4). The propagation velocities of the frequency components (10–20components (10–20 kHz)
kHz) cancan
be
be obtained
obtained from
from Figure
Figure 6, 6,
and and their
their averagevalue
average valueisisconsidered
consideredasasthe velocityvvin
thevelocity in Equations
Equations (3)
(3) and
and
(4). Theflow
(4).The flowchart
chart of
of the
the fault-location
fault-location method
method is is shown
shown in in Figure
Figure 7.
7. The
The fault-location
fault-location method
method is
is
simple but efficient.
simple but efficient.

Read the currents

at both terminals of
the cable.
Determine t1 and t2
through the first
sudden change points
Calculate sixth of M1 and M2 .
mode currents i1m6
and i2m6.

Extract M1 and M2 Calculate the fault

by means of the S distance through
transform. Equation (10) and(11).

Figure
Figure 7. Flowchart of
7. Flowchart of the
the proposed
proposed fault-location
fault-location method.
method.

The fault-location error is

is defined
defined as
as
|l − l |
Error = 1 l1 −1act
l1 act × 100% (14)
Error = L × 100% (14)
L
where l1act is the actual distance between the fault point and rectifier-terminal. The fault-location error
where
is l1actinfluenced
mainly is the actual distance
by the between
sampling thecalculation
rate and fault pointerror.
and rectifier-terminal. The fault-location
error is mainly influenced by the sampling rate and calculation error.
5. Test System and Simulation Results
5. Test System and Simulation Results
5.1. Test System
The VSC-HVDC system in Figure 3 was established using a power system computer-aided
design/electromagnetic transients including DC (PSCAD/EMTDC). The DC cable was modeled by a
frequency-dependent model, and its total length L was 200 km. Some typical faults were simulated in
this model to verify the proposed fault-location method. The internal positive-cable-to-ground (PG)
5.1. Test System
The VSC-HVDC system in Figure 3 was established using a power system computer-aided
design/electromagnetic transients including DC (PSCAD/EMTDC). The DC cable was modeled by a
frequency-dependent model, and its total length L was 200 km. Some typical faults were simulated
Energies 2017, 10, 1115 9 of 15
in this model to verify the proposed fault-location method. The internal positive-cable-to-ground
(PG) faults and positive-cable-to-negative-cable (PN) faults were simulated in the system. The
faults and positive-cable-to-negative-cable
occurrence (PN) faults
time for all the faults in the following waswere simulated
set to in the
0 ms. The unitsystem. The occurrence
of the current was per
time for
unit (p.u).all the faults in the following was set to 0 ms. The unit of the current was per unit (p.u).

5.2.
5.2. Testof
Testof the
the Determination Method of
Determination Method of theArrival
theArrival Time
Time
Many
Many simulations
simulationswerewereperformed
performedtotoverify
verifythe
theaccuracy
accuracy ofof
the determination
the determination t1 and
of of t1 andt2 .tFigure 8
2. Figure
shows the curves of the sixth mode currents at the two terminals of the cable and the
8 shows the curves of the sixth mode currents at the two terminals of the cable and the corresponding corresponding
M and M
M11 and during the
M22 during the PG
PG fault
fault at
at the
the middle
middle point
point of
of the
the cable
cable with
with aa fault
fault resistance
resistance of 100 Ω.
of 100 Ω. As
As
shown in the figure, the curves of M and M almost overlap, and they both reached
shown in the figure, the curves of M11 and M22 almost overlap, and they both reached the first sudden the first sudden
change
change points
points 0.53
0.53 ms
msafter
afterthe
theoccurrence
occurrenceofofthe
thefault;
fault;this
thistime equaltotoL/2v.
timeisisequal L/2v.This
This indicates
indicates thatthat the
the
arrival timest 1 and t
arrival timest1 and t2 can be determined accurately through the sudden change points of M1 and M2. 2 .
2 can be determined accurately through the sudden change points of M 1 and M

Figure 8. Curves of sixth mode currents and the corresponding M1 and M2during the
Figure 8. Curves of sixth mode currents and the corresponding M1 and M2 during the
positive-cable-to-ground (PG) fault with a fault resistance of 100 Ω at the middle point of the cable. (a)
positive-cable-to-ground (PG) fault with a fault resistance of 100 Ω at the middle point of the cable.
The sixth mode currents. (b) M1 and M2.
(a) The sixth mode currents. (b) M1 and M2 .

5.3. Fault-Location Results

5.3. Fault-Location Results
A number of PG faults 50 km from the rectifier-terminal were simulated. The fault resistance
A number of PG faults 50 km from the rectifier-terminal were simulated. The fault resistance took
took the values of 0, 25, 50, …, 500 Ω. The sampling rate was set to 100 kHz. To provide the details,
the values of 0, 25, 50, . . . , 500 Ω. The sampling rate was set to 100 kHz. To provide the details, the
the currents of the conductor layers at p1, n1, p2 and n2 during the faults are displayed in Figure 9.
currents of the conductor layers at p1, n1, p2 and n2 during the faults are displayed in Figure 9. The
The sixth mode DC currents at the two terminals of the cable were calculated according to the
sixth mode DC currents at the two terminals of the cable were calculated according to the currents
currents in Figure 9, and the corresponding M1 and M2 during the faults were plotted and are
in Figure 9, and the corresponding M1 and M2 during the faults were plotted and are displayed in
displayed in Figure 10a,b. The numbers of the lines in the figures represent the fault resistances (Ω)
Figure 10a,b. The numbers of the lines in the figures represent the fault resistances (Ω) of the faults. As
of the faults. As seen from Figure 9, there were oscillations in the currents when the fault resistance
seen from Figure 9, there were oscillations in the currents when the fault resistance was small, and with
was small, and with the increase of the fault resistance, the change rates and change values of the
the increase of the fault resistance, the change rates and change values of the currents decreased
currents decreased markedly. As seen from Figure 10, for each of the faults, the first sudden change
markedly. As seen from Figure 10, for each of the faults, the first sudden change points of M1 and
points of M1 andM2wereprominent and sharp, while their sudden change times were not influenced
M2 wereprominent and sharp, while their sudden change times were not influenced by the oscillations
by the oscillations of the currents or the fault resistance. When the fault resistance was0 Ω, the peak
of the currents or the fault resistance. When the fault resistance was0 Ω, the peak values of the sudden
values of the sudden change points of M1 and M2werenearly 190 p.u., and when the fault resistance
change points of M1 and M2 were nearly 190 p.u., and when the fault resistance was 500 Ω, the peak
was 500 Ω, the peak values of the sudden change points of M1 and M2 were about 4.4 p.u. In general,
values of the sudden change points of M1 and M2 were about 4.4 p.u. In general, the peak values of
the peak values of the sudden change points of M1 and M2 decreased gradually with the increase of
the sudden change points of M1 and M2 decreased gradually with the increase of the fault resistance.
the fault resistance. However, they were large enough to be observed even when the fault resistance
However, they were large enough to be observed even when the fault resistance was very large.
was very large.
The simulation results for some PG faults are displayed in Table 1.The fault distance between
the fault point and the rectifier-terminal took the values of 0, 50, 110 and 180 km, whereas the fault
resistance took the values of 0, 100 and 500 Ω. The obtained terminal current signals were sampled
with a sampling rate of 100 kHz. To provide the details, the detected arrival time, estimated distance
between the fault point and rectifier-terminal and the fault-location error are displayed in Table 1.
As shown in Table 1, the minimum fault-location error was 0.1857% and the maximum fault-location
error was 0.8948%.
Energies 2017, 10, 1115 10 of 15
Energies 2017, 10, 1115 10 of 15
Energies 2017, 10, 1115 10 of 15

Figure 9. The DC currents of the conductor layers during PG faults 50 km from the rectifier-terminal
Figure
under 9. The
Figure 9. The DC
different
currents
DCfault
currents of
of the
the conductor
resistance conductor layers during
conditions.layers
(a) DCduring PG
PG faults
currents faults 50 km
at p1.50(b)
from
kmDCfrom the
therectifier-terminal
rectifier-terminal
currents at n1. (c) DC
under
under different
different fault resistance
fault conditions.
resistance (a)
conditions. DC
(a) currents
DC at p1.
currents (b)
at DC
p1. currents
(b) DC at n1. (c)atDC
currents n1.currents
(c) DC
currents at p2. (d) DC currents at n2.
at p2. (d)at
currents DCp2.currents
(d) DC at n2.
currents at n2.

Figure 10. The curves of M1 and M2 duringthe PG faults 50 km from the rectifier-terminal under
Figure 10.
different The curves ofconditions.
M1 and M2 duringthe ofPG
M1faults 50 kmoffrom
M2. the rectifier-terminal under
Figure 10.fault
Theresistance
curves of M1 and M2(a)during
Curvesthe PG . faults
(b) Curves
50 km from the rectifier-terminal under
different fault resistance conditions. (a) Curves of M1. (b) Curves of M2.
different fault resistance conditions. (a) Curves of M1 . (b) Curves of M2 .
The simulation results for some PG faults are displayed in Table 1.The fault distance between
The point
the fault simulation
and the results for some PGtook
rectifier-terminal faults
theare displayed
values in Table
of 0, 50, 110 1.The
and 180fault distance the
km, whereas between
fault
Table 1. Fault-location results for various PG faults.
the fault point and the rectifier-terminal took the values of 0, 50, 110 and 180 km,
resistance took the values of 0, 100and 500 Ω. The obtained terminal current signals were sampled whereas the fault
resistance
with aFault took the
ratevalues
Distance
sampling of
of 100Fault0, 100and
kHz. 500 Ω.
To provide theThe obtained
details, terminal
the detected current
Estimated
arrival signals
time, were sampled
Fault-Location
estimated distance
t1 (ms) t2 (ms)
with a sampling
(km) rate of 100 kHz.
Resistance To
(Ω) provide the details, the detected arrival
between the fault point and rectifier-terminal and the fault-location error are displayed(%)
Distance (km)time, estimated
Error in distance
Table 1.
between the fault point and 0rectifier-terminal
0.01 and the fault-location
1.03 error
1.7897 are
As shown in Table 1, the minimum fault-location error was 0.1857% and the maximum fault-locationdisplayed
0.8948in Table 1.
As shown in
0
error was0.8948%.Table 1, the minimum
100 fault-location
0.01 error was
1.04 0.1857% and the
0.8268 maximum fault-location
0.4134
error was0.8948%. 500 0.01 1.04 0.8268 0.4134
0
Table 0.26 results 0.78
1. Fault-location 49.932
for various PG faults. 0.3402
50 100 1. Fault-location
Table 0.26 results 0.78 49.932
for various PG faults. 0.3402
Fault Distance Fault Resistance
500 Estimated Distance
0.26 t2 (ms) 0.78 49.932 Fault-Location
0.3402 Error
Fault(km)
Distance Fault Resistance t1 (ms) Estimated Distance Fault-Location Error
(Ω) t1 (ms) (km) (%)
(km) (Ω) 0 0.57 t2 (ms) 0.47 (km) 109.63 0.1857
(%)
0 0.01 1.03 1.7897 0.8948
0
110 1000100 0.57 1.03
0.01
0.01 1.04
0.47 1.7897 109.63
0.8268
0.1857
0.8948
0.4134
0 100500 0.57
0.01 1.04 0.47 0.8268 109.63 0.1857
0.4134
500 0.01 1.04 0.8268 0.4134
500
00 0.01 1.04
0.93 0.78
0.26 0.11 0.8268 178.95
49.932 0.4134
0.5233
0.3402
50 180 1000100 0.26
0.93
0.26 0.78
0.78 0.11 49.932
49.932 178.95 0.3402
0.5233
0.3402
50 100
500 500 0.26
0.93
0.26 0.78
0.78 0.11 49.932
49.932 178.95 0.3402
0.5233
0.3402
500
0 0.26
0.57 0.78
0.47 49.932
109.63 0.3402
0.1857
110 1000 0.57
0.57 0.47
0.47 109.63
109.63 0.1857
0.1857
Various
110 PN faults with500100a fault resistance
0.57 0.47 Ω were simulated.
0.57 of 100
0.47 109.63
109.63 The distance between
0.1857 the fault
0.1857
point and the rectifier-terminal
500
0 varied from0.57 0 to 0.11
0.93 180 km. The obtained
0.47 109.63
178.95 terminal current signals were
0.1857
0.5233
sampled180with a sampling 100 0
rate of 100 kHz. 0.93
0.93 Figure0.11 178.95
0.11 11 shows the 0.5233
curves of the sixth mode
178.95 0.5233 currents
180 100
500 0.93 0.11 178.95 0.5233
i1m6 and i2m6 and the corresponding M1 0.93
and M2 .0.11The numbers178.95
of the lines in Figure 110.5233
represent the
500 0.93 0.11 178.95 0.5233
distance (km) between the fault point and rectifier-terminal. As can be seen, for each of the faults,
Various
the first PNchange
sudden faults with
pointsa fault
of M resistance of 100
and M2 can Ω were simulated.
be extracted easily and The
can distance between the
exactly represent the
fault Various
point PN
and faults
the with a fault 1resistance
rectifier-terminal varied of 0100
from to Ω were
180 km. simulated.
The obtainedThe distance
terminal between
current the
signals
sudden change of the currents. Table 2 shows the detected arrival time, the estimated distance between
fault point
were and with
sampled the rectifier-terminal
a sampling ratevaried
of 100from
kHz.0 Figure
to 180 km. The obtained
11 shows terminal
the curves current
of the sixthsignals
mode
currents i1m6 and i2m6 and the corresponding M1 and M2. The numbers of the lines insixth
were sampled with a sampling rate of 100 kHz. Figure 11 shows the curves of the mode
Figure 11
currents i1m6 and i2m6 and the corresponding M1 and M2. The numbers of the lines in Figure 11
 Energies 2017, 10, 1115 11 of 15

represent the distance (km) between the fault point and rectifier-terminal. As can be seen, for each
of the faults, the first sudden change points of M1 and M2 can be extracted easily and can exactly
Energies 2017, 10, 1115 11 of 15
represent the sudden change of the currents. Table 2 shows the detected arrival time, the estimated
distance between the fault point and rectifier-terminal and the fault-location error for the PN faults.
the
As fault pointinandTable
shown rectifier-terminal and the fault-location
2, the minimum fault-location error
errorforwas
the PN faults. As
0.1857% andshown
the in Table 2,
maximum
the minimum fault-location
fault-location error was 0.1857% and the maximum fault-location error was 0.5233%.
error was 0.5233%.

Figure 11. The curves of sixth mode currents and the corresponding M1 and M2 during the
Figure 11. The curves of sixth mode currents and the corresponding M1 and M2 during the
positive-cable-to-negative-cable (PN) faults with a fault resistance of 100 Ω at various locations. (a)
positive-cable-to-negative-cable (PN) faults with a fault resistance of 100 Ω at various locations.
Curves i1m6i . (b). (b)
of of Curves of M
of1.M
(c). Curves of i2m6. (d) Curves of M2.
(a) Curves 1m6 Curves 1 (c) Curves of i2m6 . (d) Curves of M2 .

Table 2. Fault-location results for PN faults with a fault resistance of 100 Ω at different locations.
Table 2. Fault-location results for PN faults with a fault resistance of 100 Ω at different locations.
Fault Distance t1 t2 Estimated Distance Fault-Location Error
Fault Distance (ms)
(km) (ms) Estimated
(km) Fault-Location
(%)
t1 (ms) t2 (ms)
0 (km) 0.01 1.04 Distance
0.8268 (km) Error 0.4134
(%)
50 0 0.26 0.01 0.78 1.04 49.932
0.8268 0.3402
0.4134
110 50 0.57 0.26 0.47 0.78 49.932
109.63 0.3402
0.1857
180 110 0.93 0.57 0.11 0.47 109.63
178.95 0.1857
0.5233
180 0.93 0.11 178.95 0.5233
The simulation results in Tables 1 and 2 show that with a sampling rate of 100 kHz, the fault
distance can be estimated
The simulation resultsaccurately
in Tables by the 2proposed
1 and show that fault-location method.
with a sampling rate of 100 kHz, the fault
distance can be estimated accurately by the proposed fault-location method.
5.4.Influence of the Fault Resistance
5.4. Influence of the Fault Resistance
It can be seen from Tables 1 and 2 that the accuracy of the proposed fault-location method is
It can be seen
not sensitive to thefrom
faultTables 1 and Figure
resistance. 2 that the accuracy
10 shows ofwith
that the proposed fault-location
an increase of the fault method is not
resistance, the
sensitive to theoffault
peak values the resistance. Figurepoints
sudden change 10 shows
of Mthat withMan
1 and increase of
2 decrease the fault resistance,
gradually. However, thethe peak
peak
values
valuesofofthe
thesudden
suddenchange
change points
points M1M
of of and M2M
1 and decrease gradually.
2 are both However,
significantly large,the peak
even values
when theoffault
the
sudden change
resistance points
is 500 of M1indicates
Ω. This and M2 are
thatboth
thesignificantly large, even when
proposed fault-location the fault
method resistance
is not is 500toΩ.a
vulnerable
This
largeindicates that the proposed fault-location method is not vulnerable to a large fault resistance.
fault resistance.

5.5.
5.5.Influence
InfluenceofofNoise
Noise
To
Toinvestigate
investigate thethe
influence of noise
influence on the
of noise onproposed fault-location
the proposed method,method,
fault-location the obtained current
the obtained
signals
current signals were contaminated with 0.005 p.u. of random noise (RN). For the PG fault 50 the
were contaminated with 0.005 p.u. of random noise (RN). For the PG fault 50 km from km
rectifier-terminal with a fault resistance
from the rectifier-terminal with a of 500 Ω,
fault the noiseof
resistance was added
500 Ω, to
thethenoise
rectifier-terminal
was addedcurrents.
to the
The curves of M1 are
rectifier-terminal displayed
currents. The in Figure
curves of 12.
M1 As
are seen from the
displayed figure,12.
in Figure theAsnoise
seencould
fromalso cause the
the figure, the
sudden change points of M , but the magnitudes of the sudden change points
noise could also cause the1sudden change points of M1, but the magnitudes of the sudden changecaused by noise were
much
pointslower than
caused bythose
noisecaused by faults.
were much lowerTherefore, a threshold
than those caused by Mhfaults.
was set to reject the
Therefore, sudden change
a threshold Mh was
points that were associated to noise, as shown in Figure 12. The threshold
set to reject the sudden change points that were associated to noise, as shown M h in Figure 12 is smaller
in Figure 12. The
than 1 p.u.,M
threshold and it is much smaller than the peak value of the sudden change point of M caused by
h in Figure 12 is smaller than 1 p.u., and it is much smaller than the peak 1value of the
faults. Therefore, the noise can be ignored effectively by setting a threshold Mh .
Energies 2017, 10, 1115 12 of 15

sudden
Energies change
2017, 10, 1115point of M1 caused by faults. Therefore, the noise can be ignored effectively
12 of by
15
setting a threshold Mh.

Figure 12.
Figure 12. The
The curve
curve of
of MM1 with
with 0.005p.u.
0.005p.u. of random noise
of random noise (RN)
(RN) for
for the
the PG
PG fault
fault 50
50 km
km from
from the
the
1
rectifier-terminalwith
rectifier-terminal withaaresistance
resistanceof
of500
500Ω.
Ω.

For more detailed study, various PG faults were simulated and the 0.005p.u. of RN was added
For more detailed study, various PG faults were simulated and the 0.005p.u. of RN was added to
to the current signals. The fault distance between the fault point and the rectifier-terminal took the
the current signals. The fault distance between the fault point and the rectifier-terminal took the values
values of 0, 50, 110 and 180 km, and the fault resistance was set to500Ω. The obtained terminal
of 0, 50, 110 and 180 km, and the fault resistance was set to500Ω. The obtained terminal current signals
current signals were sampled with a sampling rate of 100 kHz. The fault-location errors are
were sampled with a sampling rate of 100 kHz. The fault-location errors are displayed in Table 3.
displayed in Table 3. As can be observed from Tables 1 and 3, the fault-location error may increase
As can be observed from Tables 1 and 3, the fault-location error may increase under the condition of
under the condition of additive noise.
additive noise.
Table3.3. Fault-location
Table Fault-location errors
errors for
for PG
PG faults
faults at
at various
various locations
locations with
with aa fault
fault resistance
resistance of
of 500
500 Ω,
Ω,
considering the influence of noise.
considering the influence of noise.
Fault Distance (km) 0 50 110 180
Fault-Location Error (%) 0 0.5494 500.3402 110
Fault Distance (km)
0.2957 180
1.004
Fault-Location Error (%) 0.5494 0.3402 0.2957 1.004

5.6. Effect of Sampling Rate

5.6. Effect of Sampling Rate
The accuracy of the fault-location method is usually influenced by the sampling rate. In the
The
above, theaccuracy
samplingof the fault-location
rate was set to method
100 kHz. is To
usually influenced
investigate the by the of
effect sampling rate. Inrate
the sampling the above,
on the
the sampling rate was set to 100 kHz. To investigate the effect of the sampling
proposed method, the obtained current signals during various PG faults were also sampled rate on the proposed
under
method,
samplingthe obtained
rates current
of 200 and 500 signals
kHz. Theduring
fault various
distancePG faults the
between were alsopoint
fault sampled under sampling
and rectifier-terminal
rates
took of
the200 and of
values 5000,kHz. The
50, 110 fault
and 180distance
km, andbetween
the faultthe fault point
resistance wasand rectifier-terminal
set to took the
100 Ω. The fault-location
values 50, 110 and 180 km, and the fault resistance was set to 100 Ω.
errors for the faults are displayed in Table 4. As shown in the table, with the increase errors
of 0, The fault-location of the
for the faults are displayed in Table 4. As shown in
sampling rate, the fault-location error decreased, in general. the table, with the increase of the sampling rate,
the fault-location error decreased, in general.
Table 4. Fault-location errors for PG faults with a fault resistance of 100 Ω at different locations,
Table 4. Fault-location
considering errors
the influence of thefor PG faults
sampling with a fault resistance of 100 Ω at different locations,
rate.
considering the influence of the sampling rate.
Fault Distance (km)
Sampling Rate (kHz) 0 Fault50Distance (km)
110 180
Sampling Rate (kHz) 0 Maximum50 Fault-Location
110 Error (%) 180
100 0.4134
Maximum 0.3402 0.1857
Fault-Location 0.5233
Error (%)
200 0.2474 0. 4154 0.0549 0.2826
100 0.4134 0.3402 0.1857 0.5233
200500 0.0991
0.2474 0.0.3511
4154 0.0005
0.0549 0.0716
0.2826
500 0.0991 0.3511 0.0005 0.0716
5.7. Comparison to the Conventional Method
5.7. Comparison to the Conventional
In this subsection, Method method based on the discrete wavelet transform (DWT) is
the fault-location
compared
In thiswith the proposed
subsection, method in method
the fault-location this paper. It isonmentioned
based in wavelet
the discrete [24] thattransform
the db4 and db6
(DWT)
is compared with the proposed method in this paper. It is mentioned in [24] that the db4 and db6
Energies 2017, 10, 1115 13 of 15

wavelets are the most suitable for the detection of short-time and fast high-frequency transient
Energies 2017, 10, 1115 13 of 15
signals. In this case, the db4 and db6 wavelets were employed to detect the wave head.
Taking the PG fault50 km from the rectifier-terminal with a fault resistance of 100 Ω as an
example, thethe
wavelets are d1,mostd2 and d3 DWT
suitable results
for the for the
detection ofsixth mode and
short-time currents at the rectifier-terminal
fast high-frequency during
transient signals.
the fault
In this are
case, shown
the
Energies in1115
db410,and
2017, Figure 13. The fault
db6 wavelets were was locatedtoaccording
employed detect theto wavethe modulus
head. maximums
13 of 15 of the
wavelet
Taking the PG fault50 km from the rectifier-terminal with a fault resistance of 100 Ω as the
transform results, as shown in Figure 13. By comparing Figure
wavelets are the most suitable for the detection of short-time and fast high-frequency transient
10 with Figure 13, an
sudden
example,change
the d1, points
d2 and of the
d3 DWTS transform
results results
for the are
sixth more
mode intuitive
currents than
signals. In this case, the db4 and db6 wavelets were employed to detect the wave head.at the those of the DWT
rectifier-terminal results.
during
Additionally,
the fault are shown the wavelet
Taking in PG
the Figuretransform fromresults
13.kmThe
fault50 fault wasof located
the signals
the rectifier-terminal with aunder
accordingfault to different
the modulus
resistance wavelet
of 100 as anbases are
Ω maximums of
different; thus,
example, it is imperative
the d1, d2 and d3to choose
DWT the
results forproper
the sixth wavelet
mode basis
currents at for
the wavelet transform results, as shown in Figure 13. By comparing Figure 10 with Figure 13, thethe a specific signal.
rectifier-terminal As
during for the S
transform, the fault are
the Gaussian shown in Figure 13. The fault was located according to the modulus maximums of the
sudden change points of window functionresults
the S transform and theare FFT
moretransform
intuitive than can ensure
wavelet transform results, as shown in Figure 13. By comparing Figure 10 with Figure 13, the
those ofthe thequality of the
DWT results.
transform results
Additionally,
sudden
for
thechange different
wavelet points
signals.
transform results of
of the S transform the signals
results are moreunder
intuitivedifferent
than thosewavelet
of the DWT bases are different;
results.
thus, it is imperative
Additionally, tothechoose
waveletthe properresults
transform wavelet basis
of the for aunder
signals specific signal.
different As for
wavelet the are
bases S transform,
the Gaussian different;
window thus, function
it is imperative
and tothechoose
FFT the proper wavelet
transform basis forthe
can ensure a specific
quality signal. As transform
of the for the S results
transform, the Gaussian window function and the FFT transform can ensure the quality of the
for different signals.
transform results for different signals.

Figure 13.Figure
The 13. discrete wavelet
The discrete transform
wavelet transform (DWT) resultsof of
(DWT) results the the
sixthsixth
mode mode
currentscurrents
at the at the
Figure 13. The discrete
rectifier-terminal wavelet
during PG transform
faults 50 km from (DWT)
the results
rectifier-terminalof the
with a sixth
fault mode
resistance
rectifier-terminal during PG faults 50 km from the rectifier-terminal with a fault resistance of 100 of currents
100 Ω. at Ω.
the
rectifier-terminal during PG faults 50 km from the rectifier-terminal with a fault resistance of 100 Ω.
5.8. Comparison of Simulation Results to the Other Mode Currents
5.8. Comparison of Simulation Results to the Other Mode Currents
5.8. ComparisonInoftheSimulation
above, the sixth
Resultsmode to current was Mode
the Other employed for fault location. To provide more detail,
Currents
In thethe
above, the sixth
comparison of themode currentresults
fault-location was employed
using differentfor mode
fault currents
location. is To provide
discussed more detail,
in this
In thesubsection.
the comparisonabove,ofthe thesixth
Taking themode
PG fault
fault-location current
50 km wasthe
from
results employed for fault
rectifier-terminal
using different with
mode location.
a fault To provide
resistance
currents 100 Ωmore
is ofdiscussed at indetail,
this
the
the comparison middle
of thepoint of the
fault-locationcable as an example, the curves of M and M extracted from the third to
PG fault 50 results using
thedifferent mode currents with is discussed in thisof subsection.
1 2
subsection. Taking the km from rectifier-terminal a fault resistance 100 Ω at
Taking the sixth
the middle PG mode currents are shown in Figure 14. The numbers of the lines in the figure represent the
fault
point of50the
kmcable
from the as anrectifier-terminal
example, with a fault
of Mresistance of 100 Ω atfrom the middle point
corresponding mode currents. The values of the curves
M1 and M2 extracted 1 and
from M extracted
the2 first and second mode the third to
of the cable as
sixth modecurrentsan example,
currents
wereare the
too shown curves
small, and in theyof
FigureM and
1 14.
could M
notThe extracted
numbers
be2displayed from
in of
thethe the third
lines
figure. to
It isineasy sixth mode
thetofigure currents
see thatrepresent
the are
the
shown in Figure
frequency
corresponding 14.components
mode The numbers
currents. theof
of The thetolines
first
values ofinM
fourth the
mode figure
M2 represent
currents
1 and were too weak,
extracted the
from corresponding
andthe that it was
first and mode
hard to currents.
second mode
The values capture
of M the
and fault
M signal through them.the
Therefore, thesecond
first to fourth mode currents are not suitable
currents were too 1 small, 2 and they could not be displayed in the figure. It is easy to see and
extracted from first and mode currents were too small, thatthey
the
for the proposed fault-location method.
could not be
frequency displayed of
components in the
the first
figure. It is easy
to fourth modeto see that the
currents were frequency
too weak, components
and that itofwas thehard
first to
to
fourth mode currents were too weak, and that it was hard to capture
capture the fault signal through them. Therefore, the first to fourth mode currents are not suitable the fault signal through them.
Therefore, the firstfault-location
for the proposed to fourth mode currents are not suitable for the proposed fault-location method.
method.

Figure 14. The curves of M1 and M2 extracted from different mode currents during the PG fault 50
km from the rectifier-terminal with a fault resistance of 100 Ω. (a) Curves of M1. (b) Curves of M2.

Figure 14. The curves of M1 and M2 extracted from different mode currents during the PG fault 50
Figure 14. The curves of M1 and M2 extracted from different mode currents during the PG fault 50 km
km from the rectifier-terminal with a fault resistance of 100 Ω. (a) Curves of M1. (b) Curves of M2.
from the rectifier-terminal with a fault resistance of 100 Ω. (a) Curves of M1 . (b) Curves of M2 .
Energies 2017, 10, 1115 14 of 15

6. Conclusions
In this paper, a fault-location method for VSC-HVDC cables employing the S transform is
proposed. With the proposed method, the fault distance can be determined using the current frequency
components caused by the fault and the first arrival time of the frequency components at the cable
terminals. By means of the S transform, the first arrival time of the current frequency components can
be determined accurately. To obtain a reliable criterion, a new phase-mode transform method for the
bipolar cable was developed. The characteristics of the mode currents were analyzed in detail and the
high-frequency components of the sixth mode currents were selected for the proposed method.
A two-terminal VSC-HVDC system was modeled using PSCAD/EMTDC, and various faults
under different conditions were simulated using this model to verify the proposed fault-location
method. The simulation results showed that this method has the advantage of a high accuracy, and it
is not sensitive to a large fault resistance and noise.

Acknowledgments: There is no source of support for our research work.

Author Contributions: This paper was a collaborative effort between the authors. All authors read and approved
the final manuscript. Pu Zhao designed and performed the experiments, and wrote the paper; Qing Chen
contributed basic theory; Kongming Sun analyzed the data; and Chuanxin Xi contributed analysis tools.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Flourentzou, N.; Agelidis, G. VSC-based HVDC power transmission systems: An overview. IEEE Trans.
Power Electron. 2009, 24, 592–599. [CrossRef]
2. Bresesti, P.; Kling, W.L.; Hendriks, R.L. HVDC connection of offshore wind farms to the transmission system.
IEEE Trans. Energy Convers. 2007, 1, 37–43. [CrossRef]
3. Muhammad, R.; Kevin, S.; Oriol, G.B. Droop control design of multi-VSC systems for offshore networks to
integrate wind energy. Energies 2016, 9, 826.
4. Xie, J.; Chen, P. A traveling wave based fault locating system for HVDC transmission lines. In Proceedings
of the 2006 International Conference on Power System Technology, Chongqing, China, 22–26 October 2006;
pp. 1–4.
5. Girgis, A.; Hart, D.G.; Peterson, W.L. A new fault location technique for two- and three-terminal lines.
IEEE Trans. Power Deliv. 1992, 7, 98–107. [CrossRef]
6. Magnago, F.H.; Abur, A. Fault location using wavelets. IEEE Trans. Power Deliv. 1998, 13, 1475–1480.
[CrossRef]
7. Rajesh, K.; Yadaiah, N. Fault identification using wavelet transform. In Proceedings of the 2005 IEEE PES
Transmission and Distribution, Dalian, China, 15–18 August 2005; pp. 1–6.
8. Nanayakkara, O.M.K.K.; Rajapakse, A.D.; Wachal, R. Traveling-wave-based line fault location in
star-connected multiterminal HVDC systems. IEEE Trans. Power Deliv. 2012, 27, 2286–2294. [CrossRef]
9. Azizi, S.; Sanaye-Pasand, M.; Hasani, A. A traveling-wave-based methodology for wide-area fault location
in multiterminal DC systems. IEEE Trans. Power Deliv. 2014, 29, 2552–2560. [CrossRef]
10. Nanayakkara, O.M.K.K.; Rajapakse, A.D.; Wachal, R. Location of DC line faults in conventional HVDC
systems with segments of cables and overhead lines using terminal measurements. IEEE Trans. Power Deliv.
2012, 27, 279–288. [CrossRef]
11. Brahma, S.M.; Girgis, A. Fault location on a transmission line using synchronized voltage measurements.
IEEE Trans. Power Deliv. 2004, 19, 1619–1622. [CrossRef]
12. Shukr, M.; Thomas, D.W.P.; Zanchetta, P. VSC-HVDC transmission line faults location using active line
impedance estimation. In Proceedings of the 2012 IEEE International Energy Conference and Exhibition,
Florence, Italy, 9–12 September 2012; pp. 244–248.
13. Farshad, M.; Sadeh, J. A novel fault-location method for HVDC transmission lines based on similarity
measure of voltage signals. IEEE Trans. Power Deliv. 2013, 28, 2483–2490. [CrossRef]
14. Suonan, J.; Gao, S.; Song, G. A novel fault-location method for HVDC transmission lines. IEEE Trans.
Power Deliv. 2010, 25, 1203–1209. [CrossRef]
Energies 2017, 10, 1115 15 of 15

15. Yang, J.; Fletcher, J.E.; O’Reilly, J. Short-circuit and ground fault analysis and location in VSC-based DC
network cables. IEEE Trans. Ind. Electron. 2011, 59, 3827–3837. [CrossRef]
16. Stockwell, R.G.; Mansinha, L.; Lowe, R.P. Localization of the complex spectrum: The S transform. IEEE Trans.
Signal. Process. 1996, 44, 998–1001. [CrossRef]
17. Lee, I.W.C.; Dash, P.K. S-transform-based intelligent system for classification of power quality disturbance
signals. IEEE Trans. Ind. Electron. 2003, 50, 800–805. [CrossRef]
18. Huang, N.T.; Peng, H.; Cai, G.W. Power quality disturbances feature selection and recognition using optimal
multi-resolution fast S-transform and CART algorithm. Energies 2016, 9, 927. [CrossRef]
19. Zhao, F.; Yang, R. Power-quality disturbance recognition using S-transform. IEEE Trans. Power Deliv. 2007,
22, 944–950. [CrossRef]
20. Wang, H.H.; Wang, P.; Liu, T. Power quality disturbance classification using the S-transform and probabilistic
neural network. Energies 2017, 10, 107. [CrossRef]
21. Ahmadimanesh, A.; Shahrtash, S.M. Transient-based fault-location method for multiterminal lines employing
S-transform. IEEE Trans. Power Deliv. 2013, 28, 1373–1380. [CrossRef]
22. Dommel, H.W. EMTP Theory Book; Bonneville Power Administration: Portland, OR, USA, 1987; pp. 150–161.
23. Chen, Y.L.; Gong, Q.W.; Chen, Y.P. Traveling wave fault location for crossbonded HV cables. High. Volt. Eng.
2008, 34, 799–803.
24. Heydt, G.T.F.; Galli, A.W. Transient power quality problems analyzed using wavelets. IEEE Trans. Power Deliv.
1997, 12, 908–915. [CrossRef]

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).