IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO.

3, JULY 2006 1231

New One Terminal Digital Algorithm for Adaptive

Reclosing and Fault Distance Calculation
on Transmission Lines
Zoran M. Radojević, Senior Member, IEEE, and Joong-Rin Shin, Member, IEEE

Abstract—This paper presents a new numerical spectral do-

main algorithm devoted to blocking unsuccessful automatic
reclosing onto permanent faults and fault distance calculation.
Arc voltage amplitude and fault distance are calculated from the
fundamental and third harmonics of the terminal voltages and
currents phasors. The electric arc is modeled with its voltage
waveshape defined numerically on the basis of a number of arc
voltage records obtained in the high-voltage laboratory. From the
Fig. 1. Laboratory test circuit.
calculated arc voltage amplitude, it can be concluded whether
the fault is a transient arcing fault or a permanent arcless fault.
If the fault is permanent, automatic reclosure should be blocked. New numerical protection algorithms gave an opportunity to
The algorithm can be applied for adaptive autoreclosure, distance control the automating reclosing from a substation computer.
protection, and fault location. The results of algorithm testing The main problem to be solved is to make a distinction between
through computer simulation and real field record are given.
permanent and transient faults. Some interesting concepts were
Index Terms—Arc resistance, electric power systems, fault loca- proposed in the past. Most of them were based on the analysis
tion, numerical algorithms, protection, spectral analysis, transmis- of voltage on the opened phase conductor during the reclosing
sion lines.
dead time [4], [5]. The newly developed algorithms [6], [7]
were derived by processing the line terminal voltage and cur-
I. INTRODUCTION rent during the fault, in other words, before the breaker was
opened.
V ARIOUS studies have shown that somewhere between
70% to 90% of faults on most overhead lines are transient.
Lightning is the most common cause of transient faults. Other
In order to improve the quality of power delivery, digital de-
vices with the fault location algorithms are developed and pro-
posed in the open literature. Distance relays calculate the fault
possible causes are swinging wires and temporary contact with
distance in real time, while the fault location programs are exe-
foreign objects. Transient faults can be cleared by momen-
cuted after the fault using stored fault data. The rapid progress in
tarily de-energizing the overhead line, in order to allow the
microprocessor technology gives us hope that some numerical
fault to clear. For such faults, the service can be restored by
algorithms devoted to fault location will be used as algorithms
automatically reclosing the power circuit breaker (CB). This
for distance protection, particularly the so-called “one terminal
can improve power system transient stability and reliability,
fault location algorithms,” which use only data measured at the
providing much higher service continuity to the customers
relay place [8].
[1], [2].
In this paper, a new numerical spectral domain algorithm for
The remaining 10%–30% of faults are permanent faults that
arcing faults recognition and fault distance calculation using the
will not clear upon tripping and reclosing. An example of a per-
discrete Fourier technique will be given.
manent fault is a broken wire causing a phase to open, or a
broken pole causing the phases to short together. Reclosure onto
II. BASIC CHARACTERISTICS OF A LONG ELECTRIC ARC
a permanent fault may aggravate the potential damage to the
system and equipment. For some extra-high-voltage lines, espe- The long electric arc in free air is a plasma discharge, where
cially near generating plants, the classical automatic reclosing the plasma allows a current to flow through a gaseous medium.
of breakers cannot be used [3]. Because of those reasons, the The current can range from a few amps up to thousands of
problem of blocking automatic reclosing on permanent faults amps. In order to investigate the nature and the behavior of the
was investigated in the last decade. long arc in free air, data obtained in the FGH-Mannheim high-
power test laboratory are used [9], [10]. Voltage , current
, and arc voltage are digitized from the laboratory test
Manuscript received June 6, 2004; revised May 12, 2005. This work was
supported by the Korea Ministry of Commerce, Industry and Energy (MOCIE) circuit depicted in Fig. 1. The arc between the arcing horns of
and Korea Electric Power Research Institute (KEPRI) under the IERC Program. a vertical insulator chain is initiated by means of a fuse wire,
Paper no. TPWRD-00675-2003. when switch S in Fig. 1 is closed. The distance between elec-
The authors are with the Department of Electrical Engineering, Konkuk Uni-
versity, Seoul 143-701, Korea. trodes is selected to be 60 cm. Data are digitized using a tran-
Digital Object Identifier 10.1109/TPWRD.2005.860285 sient recorder with the sampling frequency of 10.417 kHz. In
0885-8977/$20.00 © 2006 IEEE
1232
Fig. 2. Real arc voltage and current waveforms.
Fig. 3. Typical accepted arc voltage waveshape.
Fig. 1, resistance and inductance are parameters of a short,
approximately 10-m-long, single-phase line. The recorded cur-
rent and the arc voltage are depicted in Fig. 2. The
nonlinear variations of the arc causes the arc voltage waveform
distortion, distorting it into a near square wave with arc voltage
amplitude , what is given in Fig. 2. From Fig. 2, it is obvious
that the sign of the arc voltage wave is the same as the sign
of the arc current .
In order to check how general these recordings are, compared
to variations in outdoor conditions and variations in arc current
levels, the arc voltage waveform given in Fig. 2 is compared with
experimental data obtained from the well-designed high-power
arc test, reported in [11], with arc voltage initiated in a 3.4-m-
long gap and with the arc current varying from 5 to 50 kA, and
with available field records collected from transient recorders
located at different points of the transmission network. Based on
these comparisons, one can conclude that arc waveform given in
Fig. 2 could serve as a representative form of high-current fault
arc in long gaps.
The arc voltage could be modeled by a nonlinear arc resis-
tance [12], or by the piecewise arc voltage–current character-
istics [13]. In this paper, a new approach is chosen. The arc
voltage waveshape is defined numerically on the basis of a great
number of arc voltage records obtained in the high-voltage lab-
oratory [10]. The arc voltage model presented in Fig. 3 is ac-
cepted. Fig. 3 gives a normalized cycle of the arc voltage model
with arc voltage amplitude p.u. It is assumed that the
sign of these values is determined by the sign of the arc current.
IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 3, JULY 2006
TABLE I
COEFFICIENTS OF THE hTH HARMONICS OF THE ARC VOLTAGE
The arc model given in Fig. 3 has an important feature: it can
be represented by a Fourier series containing odd sine compo-
nents only, as follows:
(1)
where is the harmonic order, is the
th harmonic of the arc voltage, is the fundamental radian
frequency, and is the coefficient of the th harmonic.
Using the discrete Fourier transform (DFT) algorithm, it is
easy to obtain coefficients for the accepted arc voltage model.
These coefficients are given in Table I.
In comparison to the other models, the advantage of the arc
voltage presentation through the sequence of numerical values
is its flexibility. One can create various waveform shapes and
calculate the corresponding coefficients , depending on the
modeling application.
III. FAULT MODEL
Let us consider the most frequent single-phase-to-ground
fault on a transmission line fed from both line terminals. The
fault model will be given by the relation between faulted phase
voltage, fault current, and earth return path parameters.
Fault voltage may be selected to be zero for a pure metallic
fault. Such cases occur, but relatively seldom. The former model
may be improved by including it into the fault resistance. The
fault resistance originates from the resistance of the arc and re-
sistance of tower footing and ground. In this paper, a new accu-
rate model with all important features of arc taken into account
is developed. The current path for ground faults adopted in this
model includes the arc voltage and the fault resistance.
Fault phase voltage can be given in time domain, what is
depicted in Fig. 4(a), by the next relation
(2)
where is arc voltage, is fault resistance, and is fault
(arc) current.
The new spectral domain fault model, developed in this paper,
is depicted in Fig. 4(b). From this picture, the th harmonic of
the fault voltage can be expressed by the next relation
(3)
where is the th harmonic of the arc voltage and is th
harmonic of the fault (arc) current.
In this paper, only the fundamental and third harmonic fault
model will be used for algorithm developing.
It is possible to express the amplitude of the th harmonic of
the arc voltage by the next relation
(4)
RADOJEVIĆ AND SHIN: NEW ONE TERMINAL DIGITAL ALGORITHM 1233

Fig. 6. Single-phase-to-ground arcing fault on the three-phase power line.

Fig. 4. Fault model given (a) in time domain and (b) in spectral domain for the
hth harmonic.

Fig. 7. Positive- and negative-sequence line equivalent circuit.

IV. ARC VOLTAGE AMPLITUDE AND FAULT

DISTANCE CALCULATION

It is practically impossible to measure the arc voltage, since

the fault has an arbitrary location on the line. The idea of this
paper is the calculation of the arc voltage amplitude from the
line terminal voltage and current signals obtainable as uniformly
Fig. 5. Phases of the arc voltage and arc current harmonics. digitized quantities at one of the line terminals. The voltages and
currents at the line terminals contain harmonics. The distortion
of waves depends on the fault distance and the arc voltage ampli-
where is the amplitude of the th harmonic of the arc tude. It could be concluded through spectral analysis that the line
voltage. terminal voltage and the current contain the harmonics induced
From the relation (4), it follows: by arc voltage, particularly the odd components. These ampli-
tudes are not comparable to the fundamental harmonic ampli-
and (5) tude but, despite that, are recognizable and measurable.
Thus, one of the obvious ways in which arc voltage harmonics
The arc voltage wave is in phase with the fault arc current, could be calculated is to analyze the circuit depicted in Fig. 6. In
as shown in Fig. 2. That means that the phase of the first har- this circuit, all variables have radian frequency and all of the
monic of the arc voltage has to be the same as the phase of the line parameters are calculated in terms of . Index denotes
first arc current harmonic (i.e., fault current). The phase of the the order of the harmonic.
third harmonic of the arc voltage has to be three times greater Let us assume a single-phase-to-ground arcing fault occur-
than the phase of the first harmonic of arc current. The former ring on a three-phase overhead line depicted in Fig. 6. In Fig. 6,
observation could be expressed as is the th harmonic of the left line terminal phase voltage,
is the is the th harmonic of the left line terminal current, is
and (6) the th harmonic of the arc voltage, is fault resistance, and
is the th harmonic of the faulted phase voltage on the fault
where and are vectors of the first and the third har- place.
monics of the arc voltage, , and , The three-phase circuit from Fig. 6 can be presented by three
where is the phase of the first harmonic of the fault current single-phase equivalent circuits: positive- (p), negative- (n), and
. zero-sequence (0) equivalent circuits. Positive- and negative-se-
Fundamental harmonics of the arc voltage and arc current and quence equivalent circuits are equal and are depicted in Fig. 7.
third harmonic of the arc voltage, and the relation between their In Fig. 7, is a positive- or negative-sequence line impedance.
phases is depicted in Fig. 5. The zero-sequence equivalent line circuit is depicted in Fig. 8.
1234 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 3, JULY 2006

Fig. 8. Zero-sequence line equivalent circuit.

In Fig. 8, all variables and parameters are zero-sequence vari- Fig. 9. Test power system.
ables and parameters.
For the equivalent circuits depicted in Figs. 7 and 8, the fol- Complex equations (15) and (16) give the system of four
lowing equations can be written: scalar equations
(7)
(8) (18)
(9)

By adding (7)–(9), and using basic symmetrical components (19)

equations
(20)

and (21)
from which unknown arc voltage amplitude and fault distance
can be calculated.
one obtains Theoretically, the fifth or seventh harmonics phasors could
be used in (20) and (21) instead of the third harmonic phasors.
(10) But, because of a considerable decrease in the amplitudes of the
higher order harmonics, the accuracy of the algorithm output is
where is the zero-sequence compensation
the best if the third harmonic phasors are used. For example,
factor, which can be calculated in advance.
the amplitude of the fifth harmonic of the electrical arc is only
Substituting fault model equation (3) in (10), and using rela-
54% of the amplitude of the third harmonic of the electrical arc,
tions (6), the next faulted loop equations for fundamental and
and the amplitude of seventh harmonic of the electrical arc is
third harmonic are obtained
only 34% of the amplitude of the third harmonic of the electrical
(11) arc (Table I). The authors can strongly recommend using third
(12) harmonics phasors.
After calculating the arc voltage, the amplitude algorithm can
Because the zero-sequence network is passive, we can assume make the decision on whether the fault is with arc (transient
that zero-sequence currents supplied from the local and remote fault) or without arc (permanent fault). The fault is transient
systems are in phase. Then, fundamental and third harmonics of if the calculated value of the arc voltage amplitude is greater
fault current can be expressed as than the product of arc voltage gradient and the length of the arc
path, which is equal to or greater than the flashover length of
(13) a suspension insulator string. The average arc-voltage gradient
(14) lies between 12 and 15 V/cm [14].

where and are real proportional coefficients. In case in V. COMPUTER-SIMULATED TESTS

which only the arc voltage amplitude and fault distance are to
be calculated, the value of and are not necessary to be The tests have been done using the Electromagnetic Transient
known in advance. Program (EMTP) [15]. The schematic diagram of the 400-kV
Using the above assumption, (11) and (12) get the form power system on which the tests are based is shown in Fig. 9.
Here, and are digitized voltages and currents, and
(15) is the line length. The line parameters, calculated via the line
(16) constants program, were km, km,
km, km, and km.
where The data for network A were , ,
, and . Data for network B were
and (17) , , , and .
RADOJEVIĆ AND SHIN: NEW ONE TERMINAL DIGITAL ALGORITHM 1235

Fig. 10. Arc voltage shape used by EMTP. Fig. 13. Calculated fault distance (exact value used by EMTP was 60 km) and
arc voltage amplitude (exact value used by EMTP was 5.4 kV).

Fig. 11. Distorted input voltages generated by EMTP.

Fig. 14. Input voltages measured by the relay.

Fig. 12. Distorted input currents generated by EMTP.

Fig. 15. Input currents measured by the relay.

The equivalent electromotive forces of networks A and B were
and , respectively. The phase angle
The fault distance and arc voltage calculated by algorithm
between them was 20 .
are depicted in Fig. 13. The exact unknown model parameters
Single-phase-to-ground faults are simulated at different
( and ) are obtained fast, after 20 ms,
points on the transmission line. The prefault load was present
and accurate. From the algorithm speed and accuracy point of
on the line. The left line terminal voltages and currents are
view, the results obtained confirm that the algorithm is useful
sampled with the sampling frequency Hz. The
for the application to real overhead lines protection.
duration of the data window was ms.
The arc voltage used by EMTP is assumed to be of square wave-
shape with an amplitude of kV, corrupted by the random VI. FIELD TESTING
noise. Typical waveforms of arc voltage obtained through com- In order to check the validity of the algorithm presented,
puter simulation are depicted in Fig. 10, where the instant of the voltages and currents, recorded during faults on a 110-kV net-
fault inception was 23 ms. Fault resistance was . work, are processed. Here, a typical example of an arcing fault
Impute phase voltages and line currents, measurable at the will be demonstrated. In Figs. 14 and 15, voltages and currents
relay place and calculated by the EMTP for the selected study measured by the relay before and during a single-phase-line-to-
case, are plotted in Figs. 11 and 12, respectively. ground fault over arc are respectively presented. All signals
1236
Fig. 16. Calculated arc voltage amplitude and fault distance.
Fig. 17. Arc voltage amplitude calculated using the arc voltage model given
in Fig. 3, curve (a), and using the square-wave arc model, curve (b).
are sampled with the sampling frequency Hz. The
duration of data window was ms.
After data processing, the results of the application of the al-
gorithm presented are depicted in Fig. 16. The exact fault loca-
tion (Fig. 16) km was calculated. Additionally, it is
determined that the fault was over an arc with the calculated am-
plitude plotted in Fig. 16. As it can be observed, the arc voltage
amplitude was approximately 2 kV.
The accuracy of the arc voltage amplitude calculated in al-
gorithm is basically depending on the model of the arc voltage
shown in Fig. 3. But the actual arc voltage shape during the
arcing fault on a transmission line in real world could be dif-
ferent from the model. Because of that, it is interesting to inves-
tigate the dependency of the algorithm output on accuracy of
the arc voltage model. In Fig. 17, the algorithm output obtained
using arc voltage shape given in Fig. 3 ( , ),
and algorithm output calculated using square-wave shape arc
voltage ( , ) are presented. For square-
wave shape function, harmonic coefficients are calculated by the
theoretical expression
(22)
where is the harmonic order. One can conclude that from the
point of view of adaptive reclosing, both results are obtained
with sufficient accuracy.
Because the minimal length of the arc path determined by
the arcing horns distance was 0.85 m, the minimal arc voltage
amplitude expected was 1.1 kV, obtained as a product of the
IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 3, JULY 2006
minimal arc length and the electric field inside the arc, which is
practically constant along the arc and has the average value of
1.3 kV/m. The calculated arc voltage amplitude indicates that
the arc was prolonged with regard to its minimal length, equal
to the distance between arcing horns.
By inspection of the fault analyzed, it was concluded that the
estimated distance was the exact one.
VII. CONCLUSION
A new numerical algorithm for arcing faults recognition and
for fault distance calculation is developed. The algorithm is de-
rived by processing line terminal voltages and currents during
the period between the fault inception and fault clearance. It
is based on the spectral analysis of the input phase voltages
and line currents signals measured by numerical relay. Only the
fundamental and third harmonic phasors calculated by discrete
Fourier technique are needed for algorithm development.
The arc voltage amplitude calculated in the algorithm can be
used for blocking reclosing of transmissions lines with perma-
nent faults, whereas the fault distance calculated in algorithm
can be used for distance protection or for fault location.
A new spectral domain fault model with all significant arc
features included is given.
The algorithm was successfully tested with data obtained
through computer simulation and data recorded in the real
power system.
REFERENCES
[1] IEEE Power System Relaying Committee Working Group, “Single
phase tripping and auto reclosing of transmission lines IEEE committee
report,” IEEE Trans. Power Del., vol. 7, no. 1, pp. 182–192, Jan. 1992.
[2] E. W. Kimbark and E. W. Kimbark, Power System Stability. New York:
Wiley, 1995, p. 998.
[3] C. E. J. Bowler, P. G. Brown, and D. N. Walker, “Evaluation of the effect
of power circuit breaker reclosing practices on turbine-generator shafts,”
IEEE Trans. Power App. Syst., vol. PAS-99, pp. 1764–1779, 1980.
[4] R. K. Aggarwal, A. T. Johns, Y. H. Song, R. W. Dunn, and D. S. Fitton,
“Neural-network based adaptive single-pole autoreclosure technique for
EHV transmission systems,” Proc. Inst. Elect. Eng., Gen. Transm. Dis-
trib., vol. 141, pp. 155–160, Mar. 1994.
[5] Y. Ge, F. Sui, and Y. Xiao, “Prediction methods for preventing single-
phase reclosing on permanent fault,” IEEE Trans. Power Del., vol. 4, no.
1, pp. 114–121, Jan. 1989.
[6] Z. Radojević, V. Terzija, and M. Ðurić, “Spectral domain arcing faults
recognition and fault distance calculation in transmission systems,”
Elect. Power Syst. Res., vol. 37, pp. 105–113, 1996.
[7] Z. Radojević and M. Ðurić, “Arcing faults detection and fault distance
calculation on transmission lines using least square technique,” Int. J.
Power Energy Syst., vol. 18, no. 3, pp. 176–181, 1998.
[8] M. S. Sachdev, Advancements in microprocessor based protection
and communication, in IEEE Tutorial Course Text, Publication no.
97TP120-0, 1997.
[9] V. Terzija and H.-J. Koglin, “Long arc in free air: laboratory testing,
modeling, simulation and model-parameters estimation,” Proc. Inst.
Elect. Eng., Gen. Transm. Distrib., vol. 149, no. 3, pp. 319–325, May
2002.
[10] V. Terzija and D. Nelles, “Parametrische Modelle des Lichtbogens und
Parameterschatzung auf Grund der simulierten und echten Daten,” Univ.
Kaiserslautern, Germany, Jul. 1993.
[11] Y. Goda, M. Iwata, K. Ikeda, and S. Tanaka, “Arc voltage characteristics
of high current fault arcs in long gaps,” IEEE Trans. Power Del., vol. 15,
no. 2, pp. 791–795, Apr. 2000.
[12] T. Funabashi, H. Otoguro, Y. Mizuma, L. Dubé, M. Kizilcay, and A.
Ametani, “Influence of fault arc characteristics on the accuracy of digital
fault locators,” IEEE Trans. Power Del., vol. 16, no. 2, pp. 195–199, Apr.
2001.
RADOJEVIĆ AND SHIN: NEW ONE TERMINAL DIGITAL ALGORITHM 1237

[13] A. T. Johns, R. K. Aggarwal, and Y. H. Song, “Improved technique for
modeling fault arc on faulted EHV transmission systems,” Proc. Inst.
Elect. Eng., Gen. Transm. Distrib., vol. 141, pp. 148–154, Mar. 1994.
[14] A. S. Maikapar, “Extinction of an open electric arc,” Elektrichestvo, vol.
4, pp. 64–69, 1960.
[15] D. Lönard, R. Simon, and V. Terzija, “Simulation von Netzmodellen
mit zwei seitiger Einspeisung zum Test von Netzschutz-einrichtungen,”
Univ. Kaiserslautern, Germany, 1992.
Joong-Rin Shin (M’87) received the B.S., M.S., and Ph.D. degrees from Seoul
National University, Seoul, Korea, in 1977, 1984, and 1989, respectively.
Currently, he is a Professor of Electrical Engineering with Konkuk Univer-
sity, Seoul. From 1977 to 1990, he was a Researcher Staff with KEPCO, Seoul,
Korea. His major research fields are in power system operation and planning.

Zoran M. Radojević (M’02–SM’03) received the B.S., M.S., and Ph.D. degrees
in electrical engineering from the University of Belgrade, Belgrade, Yugoslavia.
Currently, he is an Associate Professor in the Electrical Engineering Depart-
ment, Konkuk University, Seoul, Korea. He was a Professor with the Faculty
of Electrical Engineering, University of Belgrade. He was also with the Elec-
trotechnology R&D Center, LG Industrial Systems Co. Ltd., Cheongju, Korea.
His areas of research are power system protection, digital signal processing ap-
plications in power systems, short circuits, power system stability, high-voltage
engineering, power cables, and electric power distribution systems.