Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO.

2, APRIL 2009

569

Hybrid Traveling Wave/Boundary Protection


for Monopolar HVDC Line
Xiaolei Liu, Student Member, IEEE, A. H. Osman, Member, IEEE, and O. P. Malik, Life Fellow, IEEE

AbstractA novel hybrid protection algorithm, based on traveling wave protection principle and boundary protection principle
for a monopolar HVDC line is proposed. Stationary wavelet
transform (SWT) is adopted in the traveling wave protection
to process the dc signal and then wavelet modulus maxima are
used to further represent the useful traveling wave signal. The
boundary protection principle based on SWT is used jointly
with traveling wave protection to distinguish internal faults from
external faults. The effect of border distortion, noise, high ground
fault resistance, close-up faults, transients caused by lightning
strokes and different dc line terminations are considered in the
paper.
Index TermsBoundary protection, HVDC line protection,
lightning strokes, relay, stationary wavelet transform (SWT),
traveling waves.

I. INTRODUCTION

HE main line protection system for HVDC transmission


line uses the voltage level and its rate of change to detect
the ground fault on the dc line. Backup line protection includes
dc minimum voltage protection, dc overvoltage protection, dc
line differential protection, etc. [1]. The major shortcoming of the
main protection is the influence of fault impedance. With the rapid
advance of microelectronics technology, traveling wave theorybased protection has been implemented [2] and adopted in real ac
and dc systems [3]. Traveling wave-based methods are more advantageous for HVDC line protection than for ac line protection.
In the ac line, the fault point cannot produce a transient traveling
wave for the fault inception angle at zero crossing while the dc line
does not have this problem. Besides, the structure of the ac bus has
a great influence on the traveling wave. The traveling wave has
to be distinguished from refracted wave of the bus. The dc line,
on the other hand, has a simpler structure and converters mainly
reflect waves instead of refracting waves [4].
Although successfully applied to line protection, traveling
wave-based methods still have problems that limit their application, such as lacking mathematical tools to represent traveling
wave and not being capable of detecting close-up faults. They
usually require the knowledge of characteristic impedance.
Also, pure time domain traveling wave methods are easily
influenced by noise [5].
Wavelet transform is found to be a useful mathematical tool to
represent traveling waves as a time-frequency method. The application of the wavelet transform to represent traveling waves
Manuscript received November 26, 2007. First published February 13, 2009;
current version published March 25, 2009. Paper no. TPWRD-00733-2007.
The authors are with the University of Calgary, Calgary, AB T2N 1N4,
Canada (e-mail: liux@ucalgary.ca; aosman@ucalgary.ca; maliko@ucalgary.ca).
Digital Object Identifier 10.1109/TPWRD.2008.2002687

is found in the literature. It has been applied in ac line directional relay to extract traveling wave information [6], [7]. Reverse voltage traveling wave is used as an input to a wavelet
modulus maxima (WMM)-based edge detector for the HVDC
line [5]. Other applications in HVDC line protection can also
be found in [8][10]. The HVDC model for these papers is
single-wire ground return system with a Bergeron line model
and the termination is inductive.
A hybrid algorithm consisting of an impedance relay and
traveling wave relay, is shown to be effective in solving
close-up faults in ac transmission systems [2]. However, power
frequency-based distance protection cannot meet the requirement of HVDC transmission, because the steady state voltage
and current are largely influenced by the HVDC control strategies. So the hybrid algorithm might not work for HVDC line
protection. A new boundary protection concept is proposed and
successfully applied to ac line protection [11]. It is believed
that current transient signals, particularly in the high frequency
range, will be shunted to earth by busbar capacitance. By
applying this theory, boundary protection can distinguish internal faults from external faults based only on one-end high
frequency signals. A similar technique has also been applied
in HVDC line protection [10], [12]. The smoothing reactor in
the HVDC system can attenuate the sudden change caused by
external faults and, thus, forms a boundary.
In the literature, most studies regarding lightning strokes
are for insulation or overvoltage studies; few studies are about
the effect of lightning on transient-based protection algorithms
[13][15]. In these studies, a current source of positive polarity
is used. However, it is suggested that a current source of negative polarity should be used for lightning study [16][18] since
90% of strokes are negative [19]. Current magnitude and tower
structure are not adequately considered in the existing lightning
studies for protection purposes. It is also found that when
lightning strikes the tower without causing back-flashover,
transients can still be induced on the main line and will possibly
cause confusion to the protection algorithm too. It is thus
necessary to further study the influence of lightning strokes on
transient-based transmission line protection algorithms.
In this paper, a hybrid protection algorithm consisting of traveling wave distance protection and boundary protection is proposed for mono-polar metallic return HVDC line main protection. The method to distinguish lightning strokes, brief introduction of stationary wavelet transform (SWT), and discrete
wavelet transform (DWT), the method to deal with border distortion, the basics of WMM, the de-noising method, and the
protection principle used in this paper are given in Section II.
The simulation model is given in Section III. The detailed procedure of the algorithm is presented in Section IV. Simulation

0885-8977/$25.00 2008 IEEE

570

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 2, APRIL 2009

Fig. 1. DC line fault.


Fig. 2. Lattice diagram for the dc line fault.

results are given in Section V, followed by the conclusions in


Section VI.
II. BASIC THEORY AND PROTECTION PRINCIPLE

are the rectifier-side line attenuation coefficient and terminal reis expressed as [22]
flection coefficient.
(2)

A. Distinguish Lightning Strokes


Lightning strokes can sometimes cause similar transient
behaviors as line faults. Three types of lightning strokes
are considered here (i.e., lightning strikes the tower causing
back-flashover [20], lightning strikes the tower without
back-flashover, and lightning strikes directly on the line
(shielding failure). The lightning stroke is represented by a
current source of negative polarity [16]. A standard 1.2/50
waveform [20] is used in this paper, where 1.2 s and 50
s represent the rise time and fall time of the waveform,
respectively. Since the boundary protection is based on high
frequency transients, it is very important to distinguish a no
fault category (e.g., transients caused by shielding failure and
a lightning strike on a tower without back-flashover, from a
fault-induced category (transients caused by internal line fault
and back-flashover). The difference can be analyzed using
traveling wave equations.
First, for a 500 kV system, a ground fault 100 km away from
rectifier side on a single wire ground return dc line [9] is used
here to illustrate the general trend of traveling wave behavior
under an internal ground fault. The line voltage on the rectifier
side and the lattice diagram are shown in Figs. 1 and 2, respectively. The dc line fault is basically adding a step input at the
fault location. Similar to the traveling wave equation deduced
in [21], the traveling wave equation for this case is

(1)
where
is the voltage change at distance 0 and time
is the initial wave from the fault point,
is a step input.
is the reflection coefficient of the fault location.
and

is the characteristic
where is the smoothing reactor and
impedance of line. At instant , the voltage change consists of
and
,
as shown near 0.5004 s in Fig. 1. Then, the voltage decays
according to the constant of the reflection coefficient . The
voltage remains much lower than the steady-state level until
instant (near 0.5013 s) when the reflected wave correthe
sponding to the third and fourth item in (1) causes a voltage
jump. For back-flashover, the behavior is similar to a high
impedance ground fault.
Next, a direct lightning hit on the dc line is shown in Fig. 3
and its lattice diagram is shown in Fig. 4. Since the width of
the lightning stroke is very narrow, it can be regarded as adding
an impulse input at the lightning location. The traveling wave
equation can be written as

(3)
are
where is the initial wave at the lightning point. and
the inverter-side terminal reflection coefficient and line attenuis the impulse input. Ideally, there should
ation coefficient.
be two spikes only. The lightning source, however, is not a pure
impulse and its width, together with the constant of reflection
coefficients, causes the decay. It is found in Fig. 3 that other
(close to 0.5004 s),
(close to 0.5022 s)
than at instant
and their corresponding decays, the voltage remains near steady
state. The lighting on tower with no back-flashover is similar to
the direct hit. An impulse is induced onto the line through the
insulation capacitor.
This difference in the behavior of the traveling wave can be
used to distinguish the fault from no fault conditions. The energy

LIU et al.: HYBRID TRAVELING WAVE/BOUNDARY PROTECTION FOR MONOPOLAR HVDC LINE

571

Fig. 5. Stationary wavelet transform.

Fig. 3. Direct lightning strike on the line.


Fig. 6. Discrete wavelet transform.

and
are highpass and lowpass filters at the th
signal.
level.
Due to its translation-invariance property, SWT has been
successfully applied in singularity detection in a noisy environment. The application of the wavelet transform in HVDC line
traveling wave protection is also a signal singularity detection
problem. Besides, the high sampling frequency of traveling
wave based algorithms will inevitably bring in noise. So the
SWT is used in the traveling wave protection in this paper.
The Haar wavelet is chosen as the mother wavelet since it
can detect sharp transients effectively. It only has two filter
coefficients and, thus, requires much less computation than any
other mother wavelet.
C. Wavelet Modulus Maxima and Denoising

Fig. 4. Lattice diagram for the direct lightning stroke.

of the original voltage signal (the summation of square of each


) near steady state
sample) above a certain level (denoted as
voltage is divided by the energy of the voltage signal below this
. If the
level; the ratio is defined as the lightning energy ratio
is lower than a threshold
, it indicates the existence
ratio
of faults; otherwise, it indicates no fault has occurred.
B. Stationary Wavelet Transform and Discrete Wavelet
Transform
The SWT has been studied by many researchers and mainly
applied in denoising. SWT does not downsample the signal but
upsamples the filters at each level by padding them with zeros
in a way defined in [23]. It thus avoids the translational variance
problem caused by decimation [24]. The SWT and traditional
DWT are shown in Figs. 5 and 6, respectively, for the first three
and ). x[n] is the original
decomposition levels (scale

The wavelet transform is defined as


. If there is an
such that
, for arbitrary belonging to
is one WMM
the right or the left neighborhood of , then
point. The WMM method is a more compact representation of
the signal than the wavelet representation and the edge of a
signal can be represented well by WMM since the wavelet is the
first derivative of a smooth function. The polarity of the WMM
reflects the direction of the edge and the magnitude reflects the
severity of change. Different from other sampled signals, the
WMM of white noise increases as the scale of wavelet transform decreases [24]. Many denoising methods are thus based
on this conclusion. A revised denoising method [9], proposed
by the authors, is used for the traveling wave protection in this
paper. The boundary protection is found to be much less influenced by noise [10] and, thus, no denoising process is required
is defor the boundary protection part. A universal threshold
fined as
(4)
where is the number of coefficients and is the estimation of
noise standard deviation. The method assumes that any wavelet

572

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 2, APRIL 2009

coefficient that is smaller than


is regarded as noise. Since
the signal at the finest scale is dominated by noise [24], the first
level wavelet detail coefficients are used for calculating . Soft
thresholding is used to threshold the coarsest level WMM (level
in this paper), after which those maxima whose
3 or scale
magnitude decreases with decreasing scale are further removed
since they are caused by noise. All three levels of WMMs (level
1 to 3) are used in later algorithms to locate the fault. By using
a threshold first, many WMMs caused by noise will be eliminated directly and will not need to be compared with their corresponding maxima in finer scales. It thus decreases the computational burden.
D. Traveling Wave Protection Principle
In this paper, a positive WMM spike corresponds to a signal
drop and a negative WMM spike corresponds to a signal rise.
is set to find the first spike
After denoising, a threshold
corresponding to the initial wave. For the HVDC line with inductive termination [25], the change of voltage signal is found
to be steep enough to represent traveling waves [10]. The initial
backward and forward voltage traveling wave has the same sign
at the fault location. The inductive termination initially looks
like an open circuit and, thus, the reflection coefficient [25] for
voltage at the termination is positive. The fault is approximately
a short circuit and its reflection coefficient for voltage is negative. The refraction coefficient at the fault point is always positive [26]. Therefore, the first spike (positive) corresponds to the
initial wave from the fault location. If the next spike is negative, then it corresponds to the first reflected wave from the fault
point and the fault location is less than half of the line length.
Or if the next spike is positive, then the fault location is more
than half the line length and the spike corresponds to the wave
reflected at the other termination and then refracted through the
for this terminal configuration is set
fault location. The
at 50. The first spike should be higher than this value.
Assuming that the first spike is found at sample , it is then
removed temporarily to find the maximum and minimum spikes
and , respectively).
among the rest of the data (at samples
Distances
and
can be calculated based on (5) and (6).
is greater than , it indicates that the fault occurs at a
If
is the
distance less than half of the line length and distance
correct distance. Otherwise, the fault location is more than half
is the chosen
of the line length and
(5)
(6)
where LL represents the length of the line.
is the sampling frequency, and is the velocity of the traveling wave. Information from all three decomposition levels (leves 1 to 3) is
used and, therefore, three distances can be obtained. The result
is regarded as the final distance if any two of them are close
enough. By using information from all three levels, the influence of noise can be further reduced.
Capacitive termination [25] of the HVDC line can cause very
steep changes in current signals when the traveling wave reaches
the termination [26]. Therefore, the current signal is used for

traveling wave protection for the HVDC line with capacitive


termination. The initial backward and forward current traveling
waves have opposite signs. The capacitive termination initially
looks like a short circuit [26] and, thus, the reflection coefficients
for current at the termination and at the fault point are both pos, after the first spike (negative)
itive. For a fault at less than
corresponding to the initial wave from the fault point, the next
spike should be negative. This corresponds to the first reflected
, the
wave from the fault point. For a fault at further than
next spike should be positive, corresponding to the wave reflected at the other termination and then refracted through the
and
are, respectively, chosen for
fault point. Distances
for this terminal configuration is
these two cases. The
. The first spike should be lower than this value.
set at
E. Boundary Protection Principle
Since smoothing reactors at the terminals of the HVDC line
attenuate the high frequency transient caused by external faults,
the high frequency transient can be used to distinguish internal
faults from external faults. In [10] and [12], the DWT is used to
process voltage transient signals, and the summation (denoted
here as ) of energy of detail coefficients from all DWT decomposition levels is used to compare with a threshold (in this
). In this paper, SWT is used instead of DWT for two
case,
reasons. First, SWT has better performance under a noisy environment. Second, for the inductive termination case, SWT has
already been used on the voltage signal in the traveling wave
protection algorithm part, so it is not necessary to start a new
DWT process. Four-level SWT decomposition is found to be effective for boundary protection. The trip criterion is expressed
as
(7)
where

is defined as
(8)

and
is the th detail coefficients of the th decomposition
level ( starts from 2 because the first level is dominated by
noise). is the maximum level of SWT decomposition and is
the total number of detail coefficients at each level. Four-level
SWT decomposition is found to be necessary to make a clear
distinction between internal and external faults.
F. Border Distortion
Due to the limited number of data in wavelet applications,
the border distortion can arise during the wavelet transform, but
users usually ignore it. For the fault location application, the position of spikes caused by traveling wave reflection is critical to
the fault distance calculation. For the boundary protection algorithm, border distortion can create extra energy and, thus, cause
wrong trip decision. The border extension [10] is a simple way
of dealing with border distortion; however, the default border
extension in the MATLAB wavelet toolbox uses too few extension samples and sometimes cannot solve the problem mentioned in this paper. Therefore, this problem is addressed here.

LIU et al.: HYBRID TRAVELING WAVE/BOUNDARY PROTECTION FOR MONOPOLAR HVDC LINE

573

Fig. 7. Block diagram of the system.

Fig. 9. Tower model used for the lightning study.

Fig. 8. Inductive and capacitive termination at the rectifier side.

III. SIMULATION MODEL


The model used in this paper is built in PSCAD. It is a revised version of the first HVDC benchmark model proposed in
[27]. The system frequency is changed from 50 Hz to 60 Hz
to meet the North American standard and frequency dependent
parameters are changed correspondingly. The dc transmission
model to a 300 km
line model is changed from a simple
long Bergeron line model in [10]. However, it is found that the
Bergeron model is not accurate enough to represent the line;
especially the effect of refracted wave cannot be properly represented. Therefore, a frequency dependent line model [28] with
the same length is used in this paper. The HVDC transmission
system also has been changed from a single-wire ground return
system to a metallic return system. The block diagram of the
system is shown in Fig. 7. The terminations of the HVDC system
are different from case to case and mainly can be inductive or
capacitive, as shown in Fig. 8. The parameters are chosen based
on [25]. The parameters of line arresters are shown in [29]. For
a specific system, the termination at the inverter side is usually
the mirror of that at the rectifier side. The relay location is set at
the rectifier side of the line. The system is a 500 kV, 1000 MW
monopolar system and other details are described in [27].
Since the PSCAD frequency dependent model can only provide the impedance matrix and admittance matrix information
to users, the velocity of the traveling wave has to be calculated
based on this information. Since the line mode velocity does
not change above 100 Hz and the transients created by fault are
mainly preserved by the line mode signal [26], the line mode velocity is calculated by using eigenvalue analysis as 295975.133

km/s for simulation purposes [30]. In real cases, this velocity


can be calculated based on previous field data.
The model for lightning is a current source with negative polarity. The standard 1.2/50 lightning waveform is obtained using
the double exponential expression [31]. For a direct lightning
hit on the line, the magnitude of current is generally less than 20
kA [16]. Lightning hit on tower with a magnitude from 20 kA to
200 kA can cause back-flashover [16]. In order to better evaluate
the back-flashover, a tower model is also included, as shown in
and are 10 m
Fig. 9. The tower geometry is based on [32].
and 57.5 m, respectively, representing the distance between the
shield wire and point A and between the shield wire and ground.
Other tower model parameters are chosen according to [16]
and [33]. The capacitance for the insulator is 10 F, the surge
impedance of the tower is 200 ; and the footing impedance
is 20 . For back-flashover, an arc is formed across the insulation capacitor. The resistance of the arc is chosen at 1 [34].
Transmission line and substation insulation design, whether
a back-flashover will occur and when it occurs, are further decided by models, such as the leader development model [20] and
usually the effect of the adjacent tower is also taken into account.
In this paper, these processes are simplified. Since the time to
flashover for a standard 1.2/50 lightning stroke is usually 0.5 to
16 s and all versions of the FLASH program evaluate stress
at 6 s after the stroke to decide whether back-flashover occurs
[20], the breakdown time for back-flashover cases is chosen at
6 s in this paper. It is also proven that the effect of reflection
wave from the adjacent tower can reduce tower top potential and
reduce the line flashover rate [33]. However, it is found that the
adjacent tower does not have much of an effect on the proposed
protection algorithm, so it is neglected. Besides, for the study
described in this paper, the PSCAD step time is 2.604 s, the
sampling frequency is 192 kHz and the data window is at the
millisecond level, so it is not necessary to include the same details as insulation studies do.
The shield wire can slightly influence the magnitude of lightning-induced voltage. It is included in the model because it provides a shunt to ground for the lightning stroke. However, the
coupling effect between the shield wire and phase conductors

574

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 2, APRIL 2009

is not considered [26], because it does not affect the proposed


algorithm. So the shield wire is directly connected to ground for
the studies in this paper.
IV. PROPOSED METHOD
The combination of traveling wave protection and boundary
protection is suggested as the HVDC line main protection here.
The feasibility of traveling wave protection has already been
proved experimentally and in industry. Besides, the proposed
traveling wave protection can generate the fault location information as well. However, the closeup faults are difficult to detect with traveling wave protection. The boundary protection is
thus combined mainly to solve this problem. Although boundary
protection has shown satisfactory results, it has not been tested
experimentally and is widely accepted by the industry. Besides,
it cannot generate fault location information. So each protection
algorithm has its own shortcomings and would have better performance by combing them as the HVDC line main protection.
Lightning strokes can also be distinguished by the proposed algorithm.
A sampling frequency at 192 kHz is found to be high enough
to provide satisfactory distance resolution and is theoretically
feasible considering the relay arrangement and sampling frequency at 400 kHz as suggested in [7]. The analog-to-digital
converter samples at 192 kHz (3200 samples per cycle for a
60-Hz system) and stores the dc signal into the DATA memory.
For the dc line with inductive termination, only dc voltage is
required; for the dc line with capacitive termination, dc current is required for traveling wave protection and dc voltage
is required for boundary protection and to distinguish lighting
strokes. Ideally, DSP should also read the memory at a high frequency of 192 kHz; however, many digital signal processing
(DSP) chips use the same registers for input and output [10].
Therefore, the calculation must be done within 0.005 ms, which
puts too much restriction on the DSP selection. It is suggested
that the DSP read the DATA memory every 0.5 ms (96 samples),
which should be enough time for DSP to finish the calculation.
A flowchart for the proposed method for system with inducis
tive termination is shown in Fig. 10. The starter threshold
to divide the
set at 80% of the normal voltage. The level
calculation is selected
signal for the lightning energy ratio
at 300 and its threshold
is 3.0. The boundary protection
is set at
to compare with energy
to
threshold
distinguish internal from external faults. All of these thresholds
are selected based on extensive simulation results. If any sample
, DSP will check the DATA
in the th window is lower than
memory every 0.5 ms to find whether enough data have been
accumulated to cover from the fault inception to the time point
where the wave can travel twice the length of the transmission
line. Assume the fault start sample is found in the th window
th
and enough data are found when it comes to the
th
window. The DSP will then read the data from the
th window, and this
window (prefault window) to the
series of data is used in the following calculations. The data used
start from the first data
to calculate the lightning energy ratio
to the last data of the
th window. For
that are below
the traveling wave protection, distances are calculated and the
final distance is selected according to the protection principle

Fig. 10. Proposed line protection algorithm.

mentioned in Section II. If it is a positive number within the


length of the transmission line LL, the trip signal is generated.
Only if the traveling wave protection does not detect the fault,
the boundary protection algorithm will be triggered and the ento
ergy , calculated according to (8), is compared with
determine whether it is an internal fault or not.
For systems with capacitive termination, the process is very
similar except that when enough data are accumulated, voltage
and current data are input to DSP. SWT is carried out on voltage
and current data. The distance is calculated based on wavelet
detail coefficients from the current signal instead of the voltage
signal.
V. SIMULATION RESULT
Disturbances on a transmission line usually include transients
caused by switching, lightning strokes, line faults and startup,

LIU et al.: HYBRID TRAVELING WAVE/BOUNDARY PROTECTION FOR MONOPOLAR HVDC LINE

575

TABLE I
PERFORMANCE UNDER DC FAULTS

and, at the same time, the other end of the line is open. Since it
is a highly improbable coincidence, the main switching process
is only a ground fault on the transmission line [35]. For the dc
startup process, because of the control system, the voltage and
current increase mildly without much overshoot. So the disturbances for the dc line are mainly caused by line faults, lightning
stroke, and external faults.
A. Effect of Noise, Fault Resistance, and External Faults
(Inductive Termination)

Fig. 11. Removal of noise with SWT. (a) DC voltage with 30 dB noise. (b)
WMM before denoising. (c) WMM after denoising.

etc., For a dc system, the main switching process of a dc link is


when there is a fault in the automatic rectifier startup procedure

Noise is usually unavoidable especially when the sampling


frequency is high. White noise with 30 dB SNR is added to the
dc voltage in this case. The fault is simulated 100 km away from
the relay location with 1 ground resistance. The dc voltage
is shown in Fig. 11(a); WMM before and after denoising are
shown in Fig. 11(b) and (c), respectively. It can be found that
most noise is successfully removed and the distance calculated
is 100.1999 km for this case.
Cases with 30-dB noise and without noise, including different
fault resistance and location, are shown in Table I. For all of
is less than
,
these cases, the lightning energy ratio
which indicates the existence of a fault. The distances are correct
under all cases. So the trip signal can be generated by the travis listed just to show
eling wave algorithm alone. The energy
that even if the traveling wave algorithm fails, the trip signal will
are above
.
still be generated because all
When an external fault occurs, the protection algorithm
should not generate the trip signal. Since the HVDC system
always has dc reactors installed at both ends of the HVDC line,
the waves caused by external faults have less magnitude and
less steep wavefront when they go through the dc reactor to the
relay position. So by setting proper thresholds, external faults
can be distinguished from dc faults. External faults studied
in this paper include rectifier ac bus faults, inverter side ac
bus faults [36] with ground fault resistances up to 100 and
commutation failure on the inverter side system as shown in
Table II. For external fault cases, the lightning energy ratio
is not considered because even if
indicates there is no fault,

576

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 2, APRIL 2009

TABLE II
PERFORMANCE UNDER EXTERNAL FAULTS

TABLE IV
ENERGY RATIO UNDER DIFFERENT SITUATIONS

TABLE III
CLOSEUP FAULTS UNDER 30-dB NOISE (INDUCTIVE TERMINATION)

TABLE V
PERFORMANCE UNDER DC FAULTS

it does not affect the algorithm since external faults should not
trigger dc line protection anyway.
B. Performance Under Closeup Faults (Inductive Termination)
When the fault occurs extremely near the relay location, for
example 0.5 km away, the traveling wave algorithm may not
detect the fault [6]. Some closeup faults (with 30 dB noise) 0.5
km away from the relay location are shown in Table III. It can be
seen that the hybrid method can effectively detect these closeup
faults.
C. Effect of Lightning Strokes (Inductive Termination)
In order to distinguish transients caused by lightning strokes
from fault cases, the lightning energy ratio
is adopted. It can
be found that the ratio is below
for fault cases and above
the threshold for no fault cases as shown in Table IV. The ratio
is not influenced much by noise.
D. Performance Under Capacitive Termination
For a system with capacitive termination, the current signal
is used for the traveling wave protection part. All other settings
are basically the same. Some dc faults are shown in Table V. It
can be seen that all of these faults will generate the trip signal
are
including closeup faults. External faults and the energy
shown in Tables VI and VII, respectively. It is obvious that the
and, thus,
energy of all external faults is below the threshold

will not generate the trip signal. The lightning energy ratio
can distinguish fault cases from no fault cases.
VI. CONCLUSION
A novel hybrid HVDC line main protection algorithm consisting of traveling wave protection and boundary protection
is proposed. A revised denoising method is used to deal with
noise. Many factors that affect the performance of protection algorithms are considered, such as transients caused by lightning
strokes, border distortion, high ground fault resistance, external
faults, closeup faults, and line terminal configuration. The proposed algorithm shows satisfactory performance under various
conditions and does not require the knowledge of characteristic
impedance to calculate the traveling wave signal.

LIU et al.: HYBRID TRAVELING WAVE/BOUNDARY PROTECTION FOR MONOPOLAR HVDC LINE

TABLE VI
PERFORMANCE UNDER EXTERNAL FAULTS

TABLE VII
ENERGY RATIO UNDER DIFFERENT SITUATIONS

In some rare cases with high fault resistance and strong noise,
the influence of noise cannot be completely removed. Sometimes, noise can change the magnitude of the real spikes caused
by traveling wave reflection. Under these rare cases, the computed distance can be incorrect and/or the traveling wave protection may fail. However, the boundary protection can catch
the faults. There are two distance sets calculated from (5) and
(6). If the distance is wrong, the distance set that is not chosen
might be correct since it is based on a different principle and,
thus, is worth being considered.
The threshold of boundary protection is set assuming that
30 dB noise exists. However, in cases where the noise level is
known to be lower, the threshold can be set higher so that higher
fault resistance can be covered by the boundary protection.
REFERENCES
[1] P. M. Anderson, Power System Protection. New York: McGraw-Hill,
1999, pp. 915955.
[2] V. Pathirana, E. Dirks, and P. G. McLaren, Hardware and software
implementation of a travelling wave based protection relay, in Proc.
IEEE Power Eng. Soc. General Meeting, Jun. 2005, vol. 1, pp.
701706.

577

[3] P. F. Gale, P. V. Taylor, P. Naidoo, C. Hitchin, and D. Clowes, Travelling wave fault locator experience on ESKOMs transmission network, in Proc. Inst. Elect. Eng. 7th Int. Conf. Developments in Power
System Protection, Apr. 2001, pp. 327330.
[4] L. Al and W.-H. Chen, Discussion on line protection of HVDC transmission line, (in Chinese) Relay, vol. 32, no. 4, pp. 6163, Feb. 2004,
(indexed by INSPEC).
[5] L. Shang, G. Herold, J. Jaeger, R. Krebs, and A. Kumar, High-speed
fault identification and protection for HVDC line using wavelet technique, in Proc. IEEE Porto Power Tech Conf., Porto, Portugal, 2001,
vol. 3, pp. 15, pt. 3.
[6] W. Chen, O. P. Malik, X. Yin, D. Chen, and Z. Zhang, Study of
wavelet-based ultra high speed directional transmission line protection, IEEE Trans. Power Del., vol. 18, no. 4, pp. 11341139, Oct.
2003.
[7] X. Dong, Y. Ge, and J. He, Surge impedance relay, IEEE Trans.
Power Del., vol. 20, no. 2, pt. 2, pp. 12471256, Apr. 2005.
[8] K. Rajesh and N. Yadaiah, Fault identification using wavelet transform, in Proc. IEEE/Power Eng. Soc. Transmission Distribution Conf.
Exhibit.: Asia Pacific, 2005, pp. 16.
[9] X. Liu, A. H. Osman, and O. P. Malik, Stationary wavelet transform
based HVDC line protection, presented at the NAPS 2007, Las Cruces,
TX, Sep. 30, 2007.
[10] X. Liu, A. H. Osman, and O. P. Malik, Hybrid traveling
wave/boundary protection for HVDC line, presented at the Int.
Conf. Power System, Bangalore, India, Dec. 12, 2007.
[11] Z. Q. Bo, A new non-communication protection technique for transmission lines, IEEE Trans. Power Del., vol. 13, no. 4, pp. 10731078,
Oct. 1998.
[12] G. Wang, M. Wu, H. Li, and C. Hong, Transient based protection for
HVDC lines using wavelet-multiresolution signal decomposition, in
Proc. IEEE/Power Eng. Soc. Transmission Distribution Conf. Exhibit.:
Asia Pacific, 2005, pp. 14.
[13] H. X. Ha and B. H. Zhao, Study on identification of fault and lightning
strokes in boundary protection for EHV transmission lines, in Proc.
Int. Conf. Power System Technology, Oct. 2002, vol. 1, pp. 270274.
[14] H. Li, G. Wang, and Z. Liao, Distinguish between lightning strikes and
faults using wavelet multi-resolution signal decomposition, in Proc.
8th Inst. Elect. Eng. Int. Conf. Developments in Power System Protection, Apr. 2004, vol. 1, pp. 8083.
[15] J. Ren, J. Duan, B. Zhang, and P. Li, Identification of lightning disturbance in ultra-high-speed transmission line protection, in Proc. IEEE/
Power Eng. Soc. Transmission Distribution Conf. Exhibit.: Asia Pacific, 2005, pp. 15.
[16] Fast Front Transients Task Force of the IEEE Modeling and Analysis
of System Transients Working Group, Modeling guidelines for fast
front transients, IEEE Trans. Power Del., vol. 11, no. 1, pp. 493506,
Jan. 1996.
[17] L. A. Kraft, Modelling lightning performance of transmission systems
using PSpice, IEEE Trans. Power Syst., vol. 6, no. 2, pp. 573549, May
1991.
[18] T. Narita, T. Yamada, A. Mochizuki, E. Zaima, and M. Ishii, Observation of current waveshapes of lightning strokes on transmission
towers, IEEE Trans. Power Del., vol. 15, no. 1, pp. 429435, Jan.
2000.
[19] J. D. Glover and M. S. Sarma, Power system analysis and design,
Thomson Learning, p. 591, 2002.
[20] IEEE Guide for Improving the Lightning Performance of Transmission
Lines, IEEE Std. 1243-1997, Dec. 1997.
[21] E. W. Kimbark, Transient overvoltages caused by monopolar ground
fault on bipolar DC line: Theory and simulation, IEEE Trans. Power
App. Syst., vol. PAS-89, no. 4, pp. 584592, Apr. 1970.
[22] H. Li, F. Lin, J. He, Y. Lu, H. Ye, and Z. Zhang, Analysis and simulation of monopolar grounding fault in bipolar HVDC transmission
system, in Proc. IEEE Power Eng. Soc. General Meeting, Jun. 2007,
pp. 15.
[23] G. P. Nason and B. W. Silverman, The stationary wavelet transform
and some statistical applications, in Lecture Notes in Statistics, A.
Antoniadis and G. Oppenheim, Eds. Bristol, U.K.: Univ. Bristol,
1995, pp. 281300. [Online]. Available: www.maths.bris.ac.uk/
~magpn/Eliospub/reports/Wavelets/swtsa.ps.gz
[24] S. Mallat and W. L. Wang, Singularity detection and processing,
IEEE Trans. Inf. Theory, vol. 38, no. 2, pt. 2, pp. 617643, Mar. 1992,
Special issue.
[25] N. G. Hingorani, Transient overvoltage on a bipolar HVDC overhead
line caused by DC line faults, IEEE Trans. Power App. Syst., vol.
PAS-89, no. 4, pp. 592610, Apr. 1970.

578

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 2, APRIL 2009

[26] M. Ando, Fault location techniques for HVDC lines: analysis, development, simulation and field-data evaluation, Ph.D. dissertation,
Washington State University, Pullman, WA, 1984.
[27] M. Szechtman, T. Wess, and C. V. Thio, First benchmark model for
HVDC control studies, Electra, no. 135, pp. 5673, Apr. 1991.
[28] EMTDC Users Manual v3, Manitoba HVDC Research Center,
1988, pp. 2633, 239249.
[29] H. Elahi, R. W. Flugum, S. E. Wright, and D. R. Brown, Insulation coordination process for HVDC converter stations: Preliminary and final
designs, IEEE Trans. Power Del., vol. 4, no. 2, pp. 10371048, Apr.
1989.
[30] A. Maffucci and G. Miano, Irregular terms in the impulse response of
a multiconductor lossy transmission line, IEEE Trans. Circuits Syst.
I: Fundam. Theory Appl., vol. 46, no. 7, pp. 788805, Jul. 1999.
[31] J. Wang and X. Zhang, Double-exponential expression of lightning
current waveforms, in Proc. 4th Asia-Pacific Conf. Environmental
Electromagnetics, Aug. 2006, pp. 320323.
[32] S. J. Shelemy and D. R. Swatek, Monte Carlo simulation of lightning strikes to the Nelson river HVDC transmission lines, presented
at the Int. Conf. Power Systems Transients, Brazil, Jun. 2001. [Online].
Available: www.ipst.org/TechPapers/2001/IPST01Paper099.pdf.
[33] Electric Power Research Inc., Transmission Line Reference Book, 345
kV and Above, 2nd ed. Palo Alto, CA: Electric Power Research, 1987,
pp. 552575.
[34] J. L. Blackburn, Protective Relaying, Principles and Applications, 2nd
ed. New York: Marcel Dekker, 1998, p. 420.
[35] E. Uhlmann, Power Transmission by Direct Current. Berlin, Germany: Springer-Verlag, 1975, pp. 214215.
[36] K. G. Narendra, V. K. Sood, K. Khorasani, and R. Patel, Application
of a radial basis function (RBF) neural network for fault diagnosis in a
HVDC system, IEEE Trans. Power Syst., vol. 13, no. 1, pp. 177183,
Feb. 1998.

Xiaolei Liu (S07) received the B.Sc. degree from Qingdao University of Science and Technology, Qingdao, China, in 2005. He is currently pursuing the
Ph.D. degree at the University of Calgary, Calgary, AB, Canada.
His areas of interest include digital protection relaying, signal processing,
neural network, and HVDC systems.

A. H. Osman (S01M03) received the B.Sc. and M.Sc. degrees in electrical


engineering from Helwan University, Cairo, Egypt, in 1991 and 1996, respectively, and the Ph.D. degree in electrical engineering from the University of Calgary, Calgary, AB, Canada, in 2003.
Currently, he is an Assistant Professor in the Department of Electrical and
Computer Engineering at the University of Calgary. His areas of interest include
power system engineering, digital protection relaying, signal processing, and
power electronics.

O. P. Malik (M66SM69LF00) received the National Diploma in electrical


engineering from Delhi Polytechnic, Delhi, India, the M.E. degree in electrical
machine design from Roorkee University, Roorkee, India, in 1962, and the Ph.D.
degree from the University of London and D.I.C. from the Imperial College,
London, U.K., in 1965.
In 1974, he became a Professor in the Department of Electrical and Computer Engineering, University of Calgary, Calgary, AB, Canada, where he is
currently Professor Emeritus. He has performed research work in collaboration
with teams from Russia, Ukraine, China, Egypt, and India. In addition to his
research and teaching, he has served in many additional capacities at the University of Calgary, including Associate Dean of Academics/Student Affairs and
Acting Dean of the Faculty of Engineering.

You might also like