IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO.

4, OCTOBER 2010 2205

Fast Distance Relay Scheme for Detecting

Symmetrical Fault During Power Swing
Chengzong Pang, Student Member, IEEE, and Mladen Kezunovic, Fellow, IEEE

Abstract—The power swing caused by various disturbances will swing takes place, the apparent impedance measured by a dis-
affect distance relay behavior and may result in relay misoper- tance relay may move away from the normal load area and into
ation. This paper provides a fast detection scheme for symmet-
one or more of the distance relay operating characteristics. This
rical fault during power swing for distance relay, which is based on
extracting the high-frequency component energy of forward and may cause unintended trips [6]–[8]. For example, the Northeast
backward traveling waves induced by faults. The multiresolution Blackout in 2003 was caused by distance relays operation in
analysis based on wavelet transform has the ability to decompose zone 3 under the overload and power swing condition, which
the analyzed signals into different frequency bands. The selection stressed the system and made the system collapse at the end [9].
of mother wavelet and the number of levels of wavelet transform
are carefully studied. The fault can be identified by feature ex- To ensure the security of operation, most modern distance re-
tracting from the d1 component of Daubechies-8 (Db8) wavelet lays detect and block the operation during the power swing [10].
transform. The proposed approach is verified by using the IEEE If a fault occurs during the power swing, the distance relay should
reference model implemented by using the Alternate Transients be able to detect the fault and operate correctly. In that case it is
Program and the test results have been presented in this paper.
This proposed method can be used for distance relay operation necessary to unblock the relay during power swing. The proce-
blocking or monitoring. dure is easy to implement for unsymmetrical faults, since the neg-
ative and zero sequence components do not exist during power
Index Terms—Fault detection, power swing, relay misoperation,
traveling wave, wavelet transform.
swing, which can be used as fault detection criterion. However, it
is much more difficult to identify symmetrical fault during stable
power swing, which may delay the operation of relay [11].
I. INTRODUCTION To solve this problem, many schemes using different methods
have been proposed. Mechraoui and Thomas present fault detec-
tion method based on load angle differences identification [4].
P OWER SYSTEM security and stability are becoming even
more challenging and important characteristics due to the
increasing complexity of power system operations. Since trans-
They did not consider the symmetrical fault in their case studies.
Benmouyal et al. present a fault detector based on tracking the
power swing center voltage (SCV) [12]. Choosing the appro-
mission lines are the vital links that enable delivering electrical
priate thresholds is still very difficult to implement. Su et al.
power to the end users, improved dependability and security of
introduced an improved method for a fast detector [13]. Their
transmission-line relays is required. According to the historical
scheme still needs two cycles to finish the fault detection, which
data [1], relay misoperation contributes to 70% of the major dis-
is not good enough when applied in an extremely high voltage
turbances in the U.S. Finding effective means to monitor and
(EHV) system. Brahma introduced the use of wavelet transform
improve distance relay operations is very important for under-
(WT) to detect the symmetrical fault quickly and reliably [14],
standing and mitigating relay misoperations on high voltage
but the sampling rate of 40.96 kHz is needed to satisfy all of the
transmission lines.
studied cases.
Power swing is a phenomenon of large fluctuations of power
Ultra-high-speed (UHS) protection schemes have been
between two areas of a power system. It is referred as the vari-
introduced based on the traveling wave detection techniques
ation of power flow, which often occurs with the instability of
[15]–[18]. Some are limited by their weakness in reliability and
synchronous generators. It is often caused by transmission line
feasibility. With the development of signal processing tools, the
faults, loss of generator units, or switching heavy loaded trans-
improved UHS protection schemes, such as the mathematical
mission lines. The occurrence of power swings is very difficult
morphology (MM)-based method [19]and WT-based method
to predict since they are quite unexpected [2]–[5]. When power
[20], [21] have been proposed. Although the traveling wave
protection scheme may not substitute the traditional protection
Manuscript received June 07, 2009; revised October 21, 2009. Date of pub-
methods right now, it provides a feasible method for fault
lication September 07, 2010; date of current version September 22, 2010. This
work was supported in part by the Power System Engineering Research Center detection with fast response and immunity to power swing and
(PSerc) under Pproject S-29 tilted “Detection, Prevention and Mitigation of Cas- other influences.
cading Events – Prototype Implementations” and in part by Texas A&M Uni-
Pang and Kezunovic [22] have shown that appropriately se-
versity. Paper no. TPWRD-00435-2009.
The authors are with the Department of Electrical and Computer En- lected wavelet-based method could detect and classify trans-
gineering, Texas A&M University, College Station, TX 77843-3128 USA mission-line faults during the power swing, which is aimed at
(e-mail: [email protected]; [email protected]). avoiding possible relay misoperations. This paper extends this
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. study by introducing a high-speed symmetrical fault detection
Digital Object Identifier 10.1109/TPWRD.2010.2050341 method for transmission lines under power swing conditions. By
0885-8977/$26.00 © 2010 IEEE
2206 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 4, OCTOBER 2010

extracting the forward and backward traveling wave at the relay

point, wavelet analysis is performed to get the spectral energy.
The criteria function and implementation framework are also
discussed. This paper is organized as follows: Section II briefly
discusses the fundamentals of power swing and relay behavior
evaluation under power swing from [22]. Section III introduces
the principle of the proposed detection method, which includes
traveling wave theory and wavelet transform. The implementa-
tion of the proposed scheme is presented in Section IV. Sec-
tion V presents the test cases and test results. Conclusions are
given in Section VI.

II. DISTANCE RELAY BEHAVIOR DURING POWER SWING

The responses of the power system to different disturbances
depend on both the initial operating state of the system and
the severity of the disturbances. The steady state power system
operates at an equilibrium, which maintains the balance between Fig. 1. Z trajectory in the R–X phase.
the generated and consumed power. When system disturbances
occur, such as various faults, transmission-line switching,
sudden loss of load, loss of generators, loss of excitation, etc., the swing, will also vary accordingly. The plots of trajecto-
mechanical power input to the generators remains constant for ries in the plane with respect to voltage magnitude ratios
a short time under those sudden changes in power system. This and angle differences is shown in Fig. 1, under the condition of
will cause the oscillations in machine rotor angles and result in .
power flow swings [2], [23]. Power swing is a variation in power During the power swing, if certain values of the magnitude
flow which occurs when generator rotor angles are advancing or ratio and angle difference are satisfied, the impedance seen by
retracting relative to each other. It is possible for one generator, the relay will reach the zone settings and relay misoperation will
or group of generators that terminal voltage angles (or phases) occur. The traditional method for power swing blocking is to
go past 180 with respect to the rest of the connected power measure the rate of change of impedance through the zones of
system, which is known as pole slipping. The power swing is relay [26]. The speed of impedance moving during the power
considered stable if pole slipping does not occur and the system swing is slower than during the fault condition. This is the basic
remains stable and returns to a new equilibrium state. [24]. How- theory of how a relay may be able to distinguish the power
ever, large power swings, regardless of whether they are stable or swing from a fault. However, as Brahma mentioned in [14],
unstable, will cause large fluctuations of voltages and currents, if a symmetrical fault occurs during a power swing, it is not
which may lead to relay misoperations and finally result in loss possible to detect it based on the mentioned principle because
of synchronism between groups of generators. both power swing and symmetrical fault are both balanced phe-
Distance relays play an important role in assuring stability nomena, which may result in the relay not being able to “see”
of power systems by eliminating faults on transmission lines the fault and clear it.
leading to instability. The distance relays are proven to be influ-
enced by power swing [2], [11]–[13]. When and only when the III. PRINCIPLE OF DETECTION METHOD
faults occur within the desired zone, distance relay should iso- The system frequency during power swing only varies over
late the faults. It should not trip the line during the power swing the range around the nominal frequency, which can be as high as
caused by the disturbances outside the protected line. That is the 4–7 Hz [12]. The occurrence of fault, on the contrary, will gen-
reason why the power swing blocking function is intergrated erate transient signals in the waveforms of currents and voltages.
in most of modern distance relays, so that the relays shall be The type and degree of existences of transient signals are largely
blocked while power swing without faults occur. determined by the fault location, fault duration, and system pre-
Either a stable or unstable power swing will have impacts on fault conditions. Based on the difference in frequency behavior,
distance relay judgment. The detailed reasoning and an example it is feasible to detect the symmetrical fault during power swing
of the two machine system are given in [25]. by extracting the high frequency components from the voltage
If there is no fault on the considered transmission line, the and current waveform.
impedance seen by distance relay at bus is Wavelets are one of the relatively new mathematical tools for
signal processing [27]. Wavelet-based signal processing tech-
(1) nique is an effective tool for power system transient analysis and
power system relaying. The applications of wavelet transform
in power system have been reported for fault detection, fault
From (1), the apparent impedance seen by relay is deter- classification, power system disturbance modeling and identifi-
mined by two variables: 1) the magnitude ratio ( ) and 2) cation, power quality analysis, etc. [25], [27]–[30]. This paper
the angle difference ( ) of the bus voltages at presents a fast detection method for symmetrical faults by using
the two ends. Since the bus voltages will oscillate during power wavelet analysis to extract the high frequency components from
PANG AND KEZUNOVIC: FAST DISTANCE RELAY SCHEME FOR DETECTING SYMMETRICAL FAULT 2207

where is the scaling (dilation) factor and is the shifting (trans-

lation) factor.
The application of the continuous wavelet transform in engi-
neering requires the feasibility evaluation. Similar to the discrete
Fourier transform, a discrete wavelet transform (DWT) is pro-
posed by adapting the discrete forms of , , and in (6), which
Fig. 2. Diagram for single phase transmission line. can be written as

(7)
the fault-induced voltage and current traveling waves propa-
gating along the transmission line.
where , , and are integer variables related to the sample
A. Traveling Wave Theory During Fault numbers in the input signal. The scaling and shifting factors
When a fault occurs in the power system, the voltage and cur- changed to the functions of , , and .
rent signals could be decomposed into two parts: 1) the prefault The performance of the wavelet transform highly depends on
steady-state component and 2) the fault injected component, or the selection of the mother wavelet. All mother wavelets have
often called superimposed component. The superimposed com- the common characteristics: the mother wavelet should be atten-
ponent can be expressed in terms of traveling waves, including uating and oscillating [32]. To perform wavelet transform, many
forward travelling wave and backward travelling wave. Fig. 2 approaches can be selected, such as Daubechies (Db), Symlets,
shows a single line diagram of a transmission system. When a Coiflets, Biorthogonas, etc. [35]. The different mother wavelets
fault occurs, the traveling wave will propagate along the line. will affect the performance of wavelet-based methods. Selecting
The wave propagation can be obtained by solving the partial the appropriate mother wavelet is very important to implement
differential equations, which are expressed as [21] the wavelet analysis.
Power swing is mostly the phenomena of low frequency os-
(2) cillation. The fault voltage or current contains high frequency
transient signals. The multiresolution analysis (MRA) will be a
(3)
best tool for decomposing the signal at the expected levels [34]
by which the faulted-derived signals can be represented in terms
where is the surge velocity, , is line surge of wavelets and scaling functions. Thus, we can easily extract
impedance, , and are the inductance and the desired information from the input signals into different fre-
capacitance per unit length, represents the distance that a surge quency bands related to the same time period.
travels away from the fault point. and are the forward and Considering an acquired digitized time signal , the ap-
backward travelling wave, respectively, which can be derived proximation coefficient (scaling coefficients) and wavelet
from (2) and (3) as coefficient (detail coefficients) after the decomposition at
(4) scales can be computed as [21]

(8)
(5)

where , are the fault injected voltage and current, re- (9)
spectively. They can be obtained by subtracting the steady-state
components from the postfault signals [31], and the steady-state where . is the total number of resolution levels.
components are those voltage and current waveforms one cycle (The maximum value of is determined by the number of sam-
before the fault. Thus, the forward and backward traveling pling points.) and are the low-pass and high-pass
waves can be calculated easily and fast. filter, respectively. Fig. 3 shows the procedure of two-scale de-
B. Wavelet Transform Analysis composition used by the MRA.
In order to represent the high frequency component in quan-
Wavelet transform (WT) is a relatively new and efficient tity, the wavelet energy spectrum is used to calculate the tran-
signal processing tool, which was introduced first at the be- sient energy in a different frequency band. From Parseval’s The-
ginning of the 1980s [32]. The application of wavelet-based orem, the energy of the analyzed signal can be represented by
techniques has been widely spread in the field of mathematics, the energy in each expansion components and their wavelet co-
physics, and engineering because of its capability of time and efficients if the used scaling function and wavelets form an or-
frequency domain analysis, which is its unique characteristic. thogonal basis, which can be shown as [27]
The fundamental theory and mathematics of the wavelet trans-
form was extensively studied and can be found in [32]–[36].
The definition of continuous wavelet transform (CWT) for a (10)
given signal with respect to a mother wavelet is
where is the norm value or the energy of
(6)
the signal component at the level after wavelet transform.
2208 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 4, OCTOBER 2010

Fig. 3. Diagram of the MRA decomposition into two scales.

Fig. 4. Daubechies-8 (Db8) wavelet.

Multiresolution analysis is a hierarchical and fast solution. It
can be implemented by a set of successive filter banks as shown
above. An import issue remains for the wavelet analysis: choice
of a suitable wavelet. A particular type of wavelet should be se-
lected depending on the application purpose of wavelet analysis.
The main concerns when selecting the appropriate wavelet are:
1) it should be one of the orthogonal wavelets;
2) it should be easy to implement and with acceptable perfor-
mance.
When wavelet analysis is used to detect transient distur-
bances, the mother wavelet shape should be close to the shape
of the detected disturbance in order to reach the higher ef-
ficiency. However, the proposed method in this paper wants
to extract the energy distribution at each frequency band in
the detected signals. Thus, an orthogonal wavelet should be
adopted to satisfy energy conservation of the Parseval’s The-
orem. Among those mother wavelets, the Daubechies wavelet
family is one of the most suitable orthogonal wavelets in
multiresolution analysis due to their powerful performance,
which has been widely used in different fields [36]. Other
orthogonal wavelets may also have the ability to function as
well as Daubechies wavelets. But Daubechies wavelets are very
easy to implement by using the fast wavelet transform with
remarkable performance [35].
Different Daubechies wavelets have different filter lengths,
which determined the performances of the Daubechies wavelet
Fig. 5. Implementation diagram of symmetrical fault detection during the
family. The longer the length of the wavelet, the higher the power swing.
computation burden of the filter. The smoother the wavelet
waveform in the time domain, the better the localization ca-
pability in the frequency domain. Literature [37] discussed IV. IMPLEMENTATION OF THE DETECTION METHOD
the comparison results in fault diagnosis using Daubechies-4
(Db4), Daubechies-8 (Db8), and Daubechies-20 (Db20) The diagram of the proposed symmetrical fault detection
wavelets, which shows that the wavelet with longer filter length method based on the travelling-wave technique and wavelet
is superior to the shorter ones. However, the computation transform is conceptually shown in Fig. 5.
burden is also an important factor to be considered when
A. Data Acquisition
choosing the wavelets to implement the fast detecting scheme
in this paper. The proposed symmetrical fault detection during power
Based on these considerations, multiresolution analysis based swing requires data acquisition for all three phase voltages and
on the Daubechies-8 (Db8) wavelet shown in Fig. 4 is selected currents. Those data can be obtained directly from measurement
for the investigations in this paper. Db8 wavelet is compactly units of advanced digital relays. Most wavelet transform based
represented in time, and this is good for the short and fast tran- methods require high frequency sampling rate. The detection
sient analysis due to its better localization performance in fre- method proposed in this paper can be used in a wide range
quency [32]. It is relatively easy to localize and detect the fault of sampling rates. Considering advanced digital relays using
part under the power swing by extracting features of transients sampling rate of 10 kHz, the same sampling rate is selected in
in the wavelet domain. this study, which can satisfy the requirements of the wavelet
PANG AND KEZUNOVIC: FAST DISTANCE RELAY SCHEME FOR DETECTING SYMMETRICAL FAULT 2209

transform proposed in this scheme with good results. In order

to avoid aliasing due to the fault transients and low sampling
rate, an analog antialiasing filter needs to be employed before
the sampling of the input waveforms coming from instrument
transformers. There are many known solutions available for im-
plementing the antialiasing filters; hence, no further discussion
is given.

B. Modal Transformation
In three phase power transmission line, the electromagnetic
coupling exists among three phases. Thus when a fault occurs in
one phase, transient currents will be induced in the other phases
due to the mutual coupling. The induced currents may distort
the travelling waves on each phase, and it’s difficult to solve
the coupled equations describing wave propagation. Therefore,
the modal transformation is adopted in the proposed scheme to Fig. 6. Border distortion effect for wavelet analysis.
uncouple the dependent phase components into three indepen-
dent propagation modes. Clarke transformation is selected in
this paper for three phase voltage and current, which is shown
as

(11)
(12)

where is called the Clarke transformation matrix Fig. 7. Periodic-padding for wavelet analysis.

After Clarke transformation, three phase variables will be

converted into earth mode 0, and two aerial modes and .
0 mode is prone to frequency dispersion and is not appropriate
to be used since our goal is to extract the high frequency com- Fig. 8. Symmetric-padding for wavelet analysis.
ponents. Since a symmetrical fault is considered, choosing be-
tween the mode or is not a critical decision. Here, the
aerial mode is selected to be used with the voltage and current kHz, so higher sampling rate will be helpful to improve the de-
signals. tection reliability.
When implementing the wavelet analysis, one cycle window
C. Wavelet Transform Implementation
data is calculated by wavelet transform. The moving speed of
As mentioned before, the frequency of the system varies over shifting data window can be set based on the requirements of
a range around the nominal frequency during power swing. The system protection scheme. The proposed fault detection method
fault will result in transient components of the voltages and cur- could be running at point-by-point shifting, which is feasible
rents, which may typically be dc or higher harmonics. In order to with the aids of high-speed DSPs.
implement the proposed method, the levels of wavelet transform Since the data for wavelet analysis windowed the dis-
and the choice of key level for analysis are carefully studied. For crete signal, the results after wavelet transform will have
a given signal, multi-resolution analysis based on Daubechies-8 border distortion, which is shown in Fig. 6. Although many
(Db8) wavelet is performed. Different level components are the extended signal methods have been used in FFT, WT, etc.,
analysis results for different frequency bands, which can offer such as zero-padding, symmetric-padding, smooth padding,
different information about given signals. Selection of sampling periodic-padding, none of them could eliminate the border dis-
rate affects the frequency band for the wavelet transform. Based tortion effects completely. For example, Figs. 7 and 8 show the
on the sampling rate of 10 kHz and Nyquist theorem, the wavelet results after WT with periodic-padding and symmetric-padding,
level d1 will cover 2.5 kHz –5 kHz. Our study shows that the respectively, in which border distortion still exists although it
d1 component from the wavelet transform is able to capture the has been improved by those signal extending methods.
energy of the transients for all kinds of faults, which includes The detection scheme proposed in this paper adopts a com-
symmetrical fault, and it is sensitive to the fault occurrence irre- promised method: discarding the first and last 10 coefficients in
spective of the occurrence of power swing condition. However, the d1 component after the wavelet transform. It will cause 1 ms
the high frequency transient component will disperse up to 100 time delay for the fault detection based on the sampling rate of
2210 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 4, OCTOBER 2010

10 kHz. The proposed method is still fast enough to detect the

symmetrical fault. The only factor to consider when introducing
a detection lag is the time needed for calculations. The calcula-
tion burden is mainly coming from performing modal transform,
wavelet transform, and wavelet energy spectrum. This is less
than the calculation burden of newly transient-based ultra-high-
speed directional protection relays that use the same wavelet
transforms. It is estimated that they use less than 6000 multipli-
cation and 5500 addition operations, which can be completed
within 1.5 ms [21]. The method proposed in this paper is able to
finish the fault detection within 2.5 ms after a fault occurs based
Fig. 9. One-line diagram of IEEE EMTP reference model.
on the sampling rate of 10 kHz. This detection scheme may be
more rapid with faster DSPs and a higher sampling rate. For ex-
ample, the time delay for discarding the 10 points will be 0.1 ms
based on the sampling rate of 100 kHz.

D. Fault Detection Criterion

The criterion for the symmetrical fault detection is defined as

(13)

where and are the energy of the d1 wavelet component

for the forward and backward traveling waves and , re-
spectively. According to the principles discussed before, and Fig. 10. IEEE EMTP reference model in ATP.
only exist after fault occurs, not only limited to symmetrical
faults. Due to the reflection effects at the bus boundary, is
bigger than after the reflection at the boundary. Thus, the is obvious that the wavelet d1 component is “quiet” during the
fault detection criteria will be defined as: if , the sym- power swing.
metrical fault occurs. When it is used in practice, in order to Fig. 12 shows the three phase voltage of a symmetrical fault
avoid the possible situation of dividing by zero, the values of during the power swing. The values of criteria factor around
and are being monitored. If any of them is close to zero the fault point are shown in Fig. 13. In this case, there is at least
(for example, less than ), the value of is set to zero. The one point that , which means a fault occurred during
threshold value of is determined by the bus reflection coef- the power swing.
ficient. Here, is set to be 1.15 after a large number of simu- In order to validate that the proposed scheme is transparent to
lation trials. More test cases are discussed in Section V. fault locations, a variety of scenarios under different locations
has been conducted. Table I shows some typical simulation re-
V. CASE STUDY sults for the three phase symmetrical fault under power swing.
The proposed scheme is also immune to the variety of fault
A. Simulation of Power Swing types and locations, under both power swing and normal condi-
In this paper, the power system model for the case study is tions. In order to validate the effectiveness of this scheme, the
based on the EMTP reference model for transmission-line relay cases of one single-phase-to-ground fault (Fault AG) and one
testing, which is introduced by the IEEE PES Power System double-phase fault (Fault AB) are studied and results are equally
Relaying Committee (PSRC) WG D10 [38]. This model is de- good without loss of generality. Table II shows the case of sim-
scribed as a “standard” system model, which can be used to ulated results for different fault locations of Fault AG and Fault
generate uniform relay test scenarios. In order to generate the AB when the power swing is absent, and Table III shows the
needed conditions, the Alternative Transient Program (ATP) is results for different locations when the power swing is present.
used to simulate the power swing [39]. The one-line diagram of All of the results show the proposed scheme is effective under
the studied system and its ATP model are shown in Figs. 9 and these conditions as well.
10, respectively.
VI. CONCLUSION
B. Example Cases A novel and fast symmetrical fault detection scheme for the
In this part, some case results are illustrated to test the per- distance relay during power swing is presented in this paper
formance of the proposed fault detection method during power aiming at avoiding possible relay misoperations during power
swing. swing conditions. It extracts the traveling waves from transient
Fig. 11 shows the typical power swing voltage waveform gen- signals induced by faults and calculates the energy of high fre-
erated by the model discussed before with the 5-level multires- quency components extracted by using the wavelet transform.
olution wavelet analysis based on Db8. d1-d5 are the high fre- Based on the discussions presented in this paper, conclusions
quency components at different wavelet level, respectively. It can be drawn as follows.
PANG AND KEZUNOVIC: FAST DISTANCE RELAY SCHEME FOR DETECTING SYMMETRICAL FAULT 2211

Fig. 13. Values of criteria factor k around the fault point.

TABLE I
SIMULATION CASE FOR DIFFERENT FAULT LOCATIONS FOR
THE SYMMETRICAL FAULT UNDER POWER SWING

Fig. 11. Five-level Db8 wavelet transform results.

TABLE II
SIMULATION CASE FOR DIFFERENT FAULT LOCATIONS
UNDER NORMAL CONDITIONS

TABLE III
SIMULATION CASE FOR DIFFERENT FAULT LOCATIONS
UNDER THE POWER SWING

Fig. 12. Three phase voltage waveforms for a symmetrical fault during the
power swing.

• The proposed method is novel since applying the ultra-

• The proposed scheme is very fast. It could detect the fault high-speed protection method based on traveling waves to
within 3 ms after the symmetrical fault occurs during the fault detection during the power swing is considered for the
power swing, which could be beneficial for system protec- first time.
tion, especially in the EHV system.
• The sampling rate for data acquisition can be as low as 10 REFERENCES
kHz, which fits the requirements of most modern micropro- [1] NERC Disturbance Reports North American Electric Reliability
Council, NJ, 1996–2001.
cessor-based distance relays. A higher sampling rate could [2] D. Tziouvaras and D. Hou, “Out-of-step protection fundamentals and
offer better results. advancements,” presented at the 30th Annu. Western Protective Relay
• The implementation of the proposed scheme is straightfor- Conf., Spokane, WA, Oct. 21–23, 2003.
[3] L. Wang and A. Girgis, “A new method for power system transient
ward since the theoretical research results may easily be instability detection,” IEEE Trans. Power Del., vol. 12, no. 3, pp.
transferred into practical application. 1082–1089, Jul. 1997.
2212 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 4, OCTOBER 2010

[4] A. Mechraoui and D. W. P. Thomas, “A new blocking principle with [28] M. Sanaye-Pasand and A. N. Jahromi, “Study, comparison and simula-
phase and earth fault detection during fast power swings for distance tion of power system swing detection and prediction method,” in Proc.
protection,” IEEE Trans. Power Del., vol. 10, no. 3, pp. 1242–1248, Power Eng. Soc. General Meeting, 2003, Jul. 2003, vol. 1, pp. 27–32.
Jul. 1995. [29] IEEE Power Syst. Relaying Comm. WG D6 Workgroup, “IEEE PSRC
[5] M. A. Redfern and M. J. Checksfield, “A study into new solution for Report, power swing and out-of-step considerations on transmission
the problems experienced with pole slipping protection,” IEEE Trans. lines,” 2005.
Power Del., vol. 13, no. 2, pp. 394–404, Apr. 1998. [30] N. Zhang and M. Kezunovic, “A study of synchronized sampling based
[6] Z. Xu, New Distance Relay for Transmission Line. Beijing, China: fault location algorithm performance under power swing and out-of-
China Water and Hydroelectricity Press, 2002. step conditions,” presented at the PowerTech, St. Petersburg, Russia,
[7] N. Zhang, H. Song, and M. Kezunovic, “Transient based relay testing: Jun. 2005.
A new scope and methodology,” presented at the 13th IEEE Mediter- [31] P. A. Crossley and P. G. McLaren, “Distance protection based on trav-
ranean Electrotechnical Conf., Torremolinos, Spain, May 2006. elling waves,” IEEE Trans. Power App. Syst., vol. PAS-102, no. 9, pp.
[8] N. Zhang and M. Kezunovic, “Verifying the protection system opera- 2971–2982, Sep. 1982.
tion using an advanced fault analysis tool combined with the event tree [32] I. Daubechies, Ten Lectures on Wavelets, ser. CBMS-NSF Regional
analysis,” in Proc. 36th Annu.North American Power Symp., Moscow, Conf. Ser. Appl. Math. Philadelphia, PA: Society for Industrial Ap-
ID, Aug. 2004, pp. 133–139. plied Mathematics (SIAM), 1992, vol. 61.
[9] “Final report on the August 14, 2003 blackout in the United States and [33] C. S. Burrus, R. A. Gopinath, and H. Guo, Introduction to Wavelets and
Canada: causes and recommendations,” U.S.–Canada Power System Wavelet Transforms: A Primer. Englewood Cliffs, NJ: Prentice-Hall,
Outage Task Force, 2004. 1998.
[10] NARI-Relays Electric Ltd. Co., RCS-901A/B EHV transmission line [34] S. Mallat, “A theory for multiresolution signal decomposition: The
distance protection. wavelet representation,” IEEE Trans. Pattern Anal. Mach. Intell., vol.
[11] X. Lin, Y. Gao, and P. Liu, “A novel scheme to identify symmetrical 11, no. 7, pp. 674–693, Jul. 1989.
faults occurring during power swings,” IEEE Trans. Power Del., vol. [35] D. B. Percival and A. T. Walden, Wavelet Methods for Time Series
23, no. 1, pp. 73–78, Jan. 2008. Analysis. Cambridge, U.K.: Cambridge Univ. Press, 2000.
[12] G. Benmouyal, D. Hou, and D. Tziouvaras, “Zero-setting power-swing [36] C. Kim and R. Aggarwal, “Wavelet transform in power system,” Proc.
blocking protection,” presented at the 31st Annu. Western Protective Inst. Elect. Eng. Power Eng., vol. 15, no. 4, pp. 193–202, Aug. 2001.
Relay Conf., Spokane, WA, Oct. 19–21, 2004. [37] J. Wu and C. Liu, “Investigation of engine fault diagnosis using discrete
[13] B. Su, X. Dong, Y. Sun, B. R. J. Caunce, D. Tholomier, and A. Apos- wavelet transform and neural network,” Expert Syst. Appl., vol. 35, no.
tolov, “Fast detctor of symmetrical fault during power swing for dis- 3, pp. 1200–1213, Oct. 2008.
tance relay,” in Proc. IEEE Power Eng. Soc. General Meeting, 2005, [38] Power Syst. Relaying Comm., EMTP reference models for transmis-
pp. 604–609. sion line relay testing report, Draft 10a, Tech. Rep. 2004. [Online].
[14] S. M. Brahma, “Distance relay with out-of-step blocking function Available: http://www.pserc.org/
using wavelet transform,” IEEE Trans. Power Del., vol. 22, no. 3, pp. [39] L. Priker and H. K. Hoidalen, ATPDraw Version 4.0 for Win-
1360–1366, Jul. 2007. dows9X/NT/2000/XP. SINTEF Energy Research AS, Norway.
[15] M. Chamia and S. Liberman, “Ultra high speed relay for EHV/UHV [Online]. Available: http://www.eeug.org/files/secret/atpdraw
transmission lines-development, design and application,” IEEE Trans.
Power App. Syst., vol. PAS-97, no. 6, pp. 2104–2116, Nov. 1978.
[16] A. T. Johns, “New ultra-high-speed directional comparison technique
for the protection of EHV transmission lines,” in Proc. Inst. Elect. Eng.
Conf., 1980, vol. 127, pp. 228–239, 4.
[17] M. Vitins, “A fundamental concept for high speed relaying,” IEEE
Trans. Power App. Syst., vol. PAS-100, no. 1, pp. 163–168, Jan. 1981.
[18] A. T. Johns, R. K. Aggarwal, and Z. Q. Bo, “Non-unit protection tech- Chengzong Pang (S’07) received the B.S and M.S.
nique for EHV transmission systems based on fault-generated noise, degrees in electrical engineering from North China
Part 1: Signal measurement,” in Proc. Inst. Elect. Eng., Gen., Transm. Electric Power University, China, in 2000 and 2003,
Distrib., 1994, vol. 141, pp. 133–140. respectively, and is currently pursuing the Ph.D. de-
[19] Q. H. Wu, J. F. Zhang, and D. J. Zhang, “Ultra high-speed directional gree at Texas A&M University, College Station.
protection of transmission lines using mathematical morphology,” His research interests are power system analysis,
IEEE Trans. Power Del., vol. 18, no. 4, pp. 1127–1133, Oct. 2003. protection, and stability and control.
[20] W. Chen, O. P. Malik, X. Yin, D. Chen, and Z. Zhang, “Study of
wavelet-based ultra high speed directional transmission line protec-
tion,” IEEE Trans. Power Del., vol. 18, no. 4, pp. 1134–1139, Oct.
2003.
[21] J. Duan, B. Zhang, S. Luo, and Z. Yi, “Transient-based ultra-high-speed
directional protection using wavelet transforms for EHV transmission
lines,” in Proc. 2nd IEEE/Power Eng. Soc. Transmission and Distribu-
tion Conf., Aug. 14–18, 2005, pp. 1–6. Mladen Kezunovic (S’77–M’80–SM’85–F’99) re-
[22] C. Pang and M. Kezunovic, “Wavelet-based method for transmission ceived the Dipl.Ing. degree in electrical engineering
line fault detection and classification during power swing,” presented from the University of Sarajevo in 1974 and the M.S.
at the MedPower, Thessaloniki, Greece, Nov. 2008. and Ph.D. degrees in electrical engineering from the
[23] F. Jiang, Z. Bo, and M. A. Redfern, “A new generator fault detec- University of Kansas in 1977 and 1980, respectively.
tion scheme using wavelet transform,” in Proc. 33rd Univ. Power Eng. Currently, he is the Eugene E. Webb Professor
Conf., Edinburgh, U.K., Sep. 1998, pp. 360–363. and Site Director of the Power Engineering Research
[24] S. Huang, C. Hsieh, and C. Huang, “Application of Morlet wavelets to Center (PSerc), a National Science Foundation
supervise power system disturbances,” IEEE Trans. Power Del., vol. I/UCRC.at Texas A&M University He was with
14, no. 1, pp. 235–243, Jan. 1999. Westinghouse Electric Corp., Pittsburgh, PA, from
[25] J. L. Blackburn, Protective Relaying Principles and Applications, 2nd 1979 to 1980 and with the Energoinvest Co., Europe,
ed. New York: Marcel Dekker, 1998. from 1980 to 1986. He spent a sabbatical at EdF, Clamart, France, from 1999
[26] A. H. Osman and O. P. Malik, “Wavelet transform approach to dis- to 2000. He was also a Visiting Professor at Washington State University,
tance protection of transmission lines,” in Proc. IEEE Power Eng. Soc. Pullman, from 1986 to 1987 and The University of Hong Kong in 2007. His
Summer Meeting, 2001, vol. 1, pp. 115–120. main research interests are digital simulators and simulation methods for
[27] A. M. Gaouda, M. M. A. Salama, M. R. Sultan, and A. Y. Chikhani, relay testing as well as the application of intelligent methods to power system
“Power quality detection and classification using wavelet multi-resolu- monitoring, control, and protection.
tion signal decomposition,” IEEE Trans. Power Del., vol. 14, no. 4, pp. Dr. Kezunovic is a member of CIGRE and a Registered Professional Engineer
1469–1476, Oct. 1999. in Texas.