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

2, APRIL 2009 579

Fault-Location Algorithms Without Utilizing

Line Parameters Based on the Distributed
Parameter Line Model
Yuan Liao, Senior Member, IEEE, and Ning Kang, Student Member, IEEE

Abstract—This paper presents novel power system transmis- the Newton–Raphson method, may be harnessed for solving
sion-line fault-location algorithms without requiring transmis- nonlinear equations, if any, during the solution process [8].
sion-line parameters. The voltages and currents from both ends of Intelligent techniques have also been drawn on for fault lo-
a line are taken as inputs and no synchronization is required. Both
prefault and fault data may be utilized. The proposed methods cation. A genetic-algorithm (GA)-based method is discussed in
are based on a distributed parameter line model and, hence, fully [9], which first formulates the fault location as an optimization
consider the impacts of shunt capacitance of the line. Positive-se- problem and then searches the optimal estimate through the net-
quence line parameters may also be estimated as a byproduct. No work.
assumption on the source impedance and fault resistance is made. One characteristic of these algorithms is that they require
Evaluation studies based on Electromagnetic Transients Program
(EMTP) simulation data demonstrate that the new methods are line parameters to be available. In some cases, the values of line
able to achieve quite accurate estimates. parameters may not be available. The authors of [10] scheme
a method for estimating line parameters, but this method
Index Terms—Distributed parameter line model, fault location,
needs continuously monitoring of the line during normal
transmission-line parameters.
operations and demands phasor measurement units for data
synchronization.
I. INTRODUCTION To locate the fault in absence of line parameters, an itera-
tive algorithm is proposed in [11] that employs unsynchronized
voltages and currents from two ends of the line. However, this
method is based on the lumped parameter line model and does

S PEEDY and precise location of faults in a transmission

system can help maintenance engineers quickly pinpoint
the faulted component and perform repair work and, hence, sig-
not consider the shunt capacitance of the line and, hence, may
yield estimation errors for long lines.
To more accurately estimate the fault location in absence of
nificantly reduce outage time [1], [2].
line parameters, this article puts new algorithms forward based
Recognizing the importance as well as the challenges of fault
on a distributed parameter line model that can fully consider the
location, a number of researchers have worked in this area and
impacts of line shunt capacitances. It is assumed that the consid-
developed a valuable set of algorithms. Based on the available
ered line is transposed and, thus, its positive- and negative-se-
data, one-terminal, two-terminal, or multiterminal algorithms
quence parameters are equal.
have been proposed in the past. In one-terminal algorithms, only
Depending on the availability of measurements preceding the
local data are entailed, thus obviating the need of transferring
fault, we will propose two fault-location algorithms, one taking
data from the remote end. However, the accuracy of this type of
advantage of both prefault and fault measurements, and the other
algorithm is normally adversely affected by the fault resistance,
utilizing only the fault data. It will be evident that the proposed
although certain compensation techniques may be adopted to
methods are also capable of yielding the estimates of the posi-
alleviate this effect [1]. The authors of [2] make use of the
tive-sequence line parameters.
remote source impedance to improve the accuracy. Two-ter-
In the rest of this paper, Sections II and III present the fault-
minal fault-location techniques employing measurements at
location algorithms. Section IV describes the evaluation studies,
both ends of the line have been proposed in [3]–[8]. References
followed by the conclusion.
[3]–[5] present methods that employ unsynchronized measure-
ments for estimating the fault location and synchronization
II. FAULT-LOCATION METHOD UTILIZING
angle between measurements from different terminals of the
PREFAULT AND FAULT DATA
line. Algorithms based on synchronized measurements have
been discussed in [6]–[8]. A computational technique, such as Consider the line, assumed to be a transposed line, between
terminals P and Q as shown in Fig. 1, where and repre-
sent the Thevenin equivalent sources.
Manuscript received December 12, 2007. First published February 27, 2009; The following discussion assumes that unsynchronized
current version published March 25, 2009. Paper no. TPWRD-00806-2007. voltage and current measurements at P and Q preceding the
The authors are with the Department of Electrical and Computer Engineering, fault and during fault are available. Fig. 2 depicts the equiva-
University of Kentucky, Lexington, KY 40506 USA (e-mail: [email protected].
edu). lent circuit of the positive-sequence network of the system
Digital Object Identifier 10.1109/TPWRD.2008.2002698 preceding the fault [12]. Note that the developed algorithms are
0885-8977/$25.00 © 2009 IEEE

Authorized licensed use limited to: ExxonMobil. Downloaded on March 28,2022 at 06:05:02 UTC from IEEE Xplore. Restrictions apply.
580 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 2, APRIL 2009

Fig. 1. Transmission line considered for analysis.

Fig. 3. Positive-sequence network of the system during the fault.

Fig. 3 delineates the positive-sequence equivalent circuit of

the line during the fault [8]. R indicates the fault point. Utilizing
Fig. 2. Positive-sequence network of the system preceding the fault. positive-sequence voltage and current, we obtain

based on fundamental frequency phasors, to which all of the

phasors throughout this paper refer to. (5)
In the figure, the following notations are adopted:

, prefault positive-sequence voltage and where

current at P;
, equivalent series impedance of the line
, prefault positive-sequence voltage and
segment PR and QR;
current at Q;
, equivalent shunt admittance of the line
positive-sequence characteristic impedance
segment PR and QR;
of the line;
, positive-sequence voltage and current
positive-sequence propagation constant of
during the fault at P;
the line;
, positive-sequence voltage and current
length of the line in miles or kilometers.
during the fault at Q;
Based on Fig. 2, we can derive the following equations [8],
, positive-sequence voltage and current
[12]:
during the fault at R.

(1) The equivalent line parameters are expressed based on the

distributed parameter line model as follows:
(2)
(6)
(3)
(7)
(4)
where (8)
synchronization angle between
measurements at P and Q, representing any (9)
possible synchronization error;
, , positive-sequence series resistance, where is the fault distance from P to R.
reactance, and shunt susceptance per mile Define as a function vector composed of the following
or kilometer. These line parameters are six functions:
treated as unknown variables in this paper.
(10)
And is defined as the unknown variable vector
(11)

where Re(.) and Im(.) yield the real and imaginary part of the
where symbolizes the vector and matrix transpose operator. argument, respectively.

Authorized licensed use limited to: ExxonMobil. Downloaded on March 28,2022 at 06:05:02 UTC from IEEE Xplore. Restrictions apply.
LIAO AND KANG: FAULT-LOCATION ALGORITHMS WITHOUT UTILIZING LINE PARAMETERS 581

Then, the five unknown variables can be obtained by fol- Referring to Fig. 3, we further obtain
lowing the well-established least squares approach [8]:

(12)
(13)

(14)
where
, variable vector at and after the th iteration;
iteration number starting from 1; (20)

variable update vector; and similarly using negative-sequence components, we can

write
Jacobian matrix.
The iterative process can be terminated when the variable update
is less than the specified tolerance.
The derivatives of the with respect to the unknown
variables can be determined as follows:

(15) (21)
The following sections present the algorithms for unsymmet-
and rical faults.
(16) A. For Single-Line-to-Ground (LG) Fault

(17) Let us take a phase-A-to ground fault as an instance and we

have
Then, we acquire (22)

(18) It can be seen that (5), (19), and (22) hold true based on the
voltage and current measurements during the fault, from which
The derivatives of the other functions in with respect six real equations can be formulated by separating each equation
to the unknown variables can be obtained similarly and are not into real and imaginary parts. The unknowns can be obtained
shown here. similarly by following (12)–(14).

B. For Line-to-Line (LL) Faults

III. FAULT-LOCATION METHOD UTILIZING ONLY FAULT DATA
Let us take the phase B to C fault as an example
If prefault data are not available, this section shows that it is
still possible to pinpoint the fault location by utilizing only the (23)
fault data. Different methods are proposed for unsymmetrical
faults, to which the following discussions apply. Thus, (5), (19), and (23) are valid for phase B to C faults. Sim-
Employing negative-sequence voltage and current leads to ilarly, the unknowns can be calculated based on the iterative
method.

C. For Line-to-Line-to-Ground (LLG) Faults

(19) Let us take the phase B to C to ground fault as an example. As
illustrated in [12], the positive- and negative-sequence voltages
where at the fault point are equal. Hence
, negative-sequence voltage and current
during the fault at P;
, negative-sequence voltage and current
during the fault at Q. (24)

Authorized licensed use limited to: ExxonMobil. Downloaded on March 28,2022 at 06:05:02 UTC from IEEE Xplore. Restrictions apply.
582 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 2, APRIL 2009

So, we have (5), (19), and (24) for phase B to C to ground faults. TABLE I
Similarly, the unknowns can be calculated based on the iterative FAULT-LOCATION ESTIMATES OBTAINED BY BOTH METHODS
method.
As will be illustrated in Section IV, for LLG faults, the
proposed method will fail to provide reliable line parameter es-
timation if the fault location is 0.5 p.u. This is because for a fault
location of 0.5 p.u., the matrix will become ill-condi-
tioned due to structural characteristics of (5), (19), and (24),
when the fault-location and synchronization angle approach
the correct values during the iteration process. Nonetheless,
the fault-location and synchronization angle estimates are still
highly precise.
In absence of line parameters, it is not possible to derive a
fault-location method for three-phase faults (LLL) based only
on the fault data, which can take line shunt capacitances into
account.

IV. EVALUATION STUDIES

This section presents evaluation results based on Electromag-
netic Transients Program (EMTP) simulation studies [13]. We
will present the estimates for both fault location and line param-
eters obtained by the proposed methods.
The 500-kV and 200-mile transmission-line system as uti-
lized in [11] is adopted here. The line is modeled based on the
distributed parameter line model. The per unit system is em-
ployed with a voltage base of 500 kV and a voltampere base
of 1000 MVA. The actual values of the line parameters in per
unit are 0.00099667 for , 0.0023566 for , and 0.0018349 for for the fault location, zero for , 1E-3 for , 2E-3 for , and 1E-3
, respectively. Diverse types of faults with different fault lo- for , respectively. The algorithm reaches convergence within
cations and fault resistances have been simulated to generate nine iterations for all cases.
voltage and current waveforms. Then, the Fourier transform is The fault-location estimates obtained by both methods are
applied to extract the voltage and current phasors to evaluate the presented in Table I. The fault type, fault resistance, and actual
performance of the proposed fault-location algorithms. To sim- fault location are given in the first, second, and third column, re-
ulate the synchronization error, we have rotated the phasors at spectively. The estimation errors resulting from the two methods
terminal Q by 10 in all cases. are displayed in the 4th and 5th column, respectively. Note that
The proposed algorithms have been implemented in Matlab. the approach based only on fault data is not applicable (N/A) to
The following technique has been adopted to speed up conver- LLL faults.
gence. During each iteration, the fault location is kept between It has been observed that the estimates are highly accurate
zero and one by setting it to zero when it becomes negative and and are virtually independent of fault resistance.
setting it to one when it exceeds one. Variables , , and are To examine the effects of starting values on the convergence
constrained as positive values by setting them to initial values behavior, different starting values have been attempted. As an
when they become negative. Evaluation studies show that this example, the total number of iterations entailed to reach con-
accelerating method works well and will not cause uncontrol- vergence with different starting values for an LG fault with a
lable oscillations in the convergence process. fault location of 0.2 p.u. and a fault resistance of 50 by uti-
In this study, the fault-location accuracy is measured by the lizing only fault data are shown in Table II. It is manifested that
percentage error calculated as speedy convergence has been achieved.
Actual location Estimated location As illustrated in Sections II and III, as a byproduct, the
(25) proposed algorithms also yield the line parameter estimates.
Total line length
Table III shows the line parameter estimation error by utilizing
The line parameter estimation accuracy is measured by the only fault data. The errors in percentage for estimated resis-
percentage error evaluated as tance, reactance, and susceptance are listed in the 4th, 5th, and
Actual parameter value Estimated parameter value 6th columns, respectively. It is evident that the estimates are
Actual parameter value quite accurate, except for the last case, which is discussed as
(26) follows.
The last case in Table III is for an LLG fault with a fault loca-
For all of the cases, the iteration process is terminated when tion of 0.5 p.u. As pointed out in Section III-C, the line param-
the update of the fault location becomes less than 1E-6. Unless eter estimates obtained by the method based only on fault data
otherwise specified, the initial values for adopted are: 0.5 p.u. may not be reliable for LLG faults. The convergence process

Authorized licensed use limited to: ExxonMobil. Downloaded on March 28,2022 at 06:05:02 UTC from IEEE Xplore. Restrictions apply.
LIAO AND KANG: FAULT-LOCATION ALGORITHMS WITHOUT UTILIZING LINE PARAMETERS 583

TABLE II TABLE VI
CONVERGENCE BEHAVIOR WITH DIFFERENT STARTING VALUES FOR IMPACTS OF VOLTAGE MEASUREMENT ERRORS ON FAULT-LOCATION
AN LG FAULT WITH A FAULT LOCATION OF 0.2 p.u. AND A FAULT ESTIMATES BY UTILIZING ONLY FAULT DATA FOR LG FAULTS
RESISTANCE OF 50 BY UTILIZING ONLY FAULT DATA (2% ERROR FOR VOLTAGE)

TABLE III
TABLE VII
LINE PARAMETER ESTIMATES BY UTILIZING ONLY FAULT DATA
IMPACTS OF VOLTAGE MEASUREMENT ERRORS ON LINE PARAMETER
ESTIMATES BY UTILIZING ONLY FAULT DATA FOR LG FAULTS
(2% ERROR FOR VOLTAGE)

TABLE IV
CONVERGENCE PROCESS FOR AN LLG FAULT WITH A FAULT RESISTANCE
OF 1 AND FAULT LOCATION OF 0.5 p.u. USING INITIAL VALUES OF
[0.5; 0; 1e-3; 2e-3; 1e-3] TABLE VIII
LINE PARAMETER ESTIMATION ERRORS DUE TO MEASUREMENT
INACCURACY FOR AN LG FAULT WITH A FAULT LOCATION
OF 0.2 p.u. AND FAULT RESISTANCE OF 1

TABLE V
CONVERGENCE PROCESS FOR AN LLG FAULT WITH A FAULT RESISTANCE
OF 1 AND FAULT LOCATION OF 0.5 p.u. USING INITIAL VALUES OF
[0.2; 0; 1e-3; 2e-3; 1e-3]

are given in Tables VI and VII. It can be seen from Table VI that
the fault-location estimates are relatively not sensitive to mea-
surement errors and are still within 2% accuracy. Performance
of the methods for other fault types is comparable. Errors in cur-
rent measurements have similar effects.
However, it is seen that the line parameter estimates are rather
sensitive to measurement errors. A closer examination of how
with different starting values is shown in Tables IV and V. The
the voltage error at terminal P affects line parameter estimation
first row represents the estimated values after the first iteration,
accuracy is presented in Table VIII for an LG fault. Generally,
etc. Note that the estimated synchronization angle is shown in
to achieve 5% accuracy in line parameter estimates, the mea-
degrees. It reveals that while the line parameter estimates are
surements need to be as accurate as 0.5%. Hence, if an accurate
not reliable, the fault-location estimate is still highly accurate
line parameter estimate is desired, the proposed methods will
regardless of the starting fault-location values.
demand relatively precise measurements.
In comparison, the method based on both prefault and fault
data is able to produce reliable line parameter estimation for any
type of fault and does not have the aforementioned problem. V. CONCLUSION
To study the impacts of measurement errors, we have carried This paper shows that it is possible to develop fault-loca-
out sensitivity studies. For example, suppose that the voltage tion algorithms by utilizing unsynchronized voltage and cur-
measurements at terminal P have 2% error in the magnitude, rent measurements from both ends of the line without requiring
the estimates of fault location and line parameters for LG faults line parameters based on the distributed parameter line model.

Authorized licensed use limited to: ExxonMobil. Downloaded on March 28,2022 at 06:05:02 UTC from IEEE Xplore. Restrictions apply.
584 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 2, APRIL 2009

Meanwhile, positive-sequence line parameters can also be ob- [8] Y. Liao and M. Kezunovic, “Optimal estimate of transmission line fault
tained as byproducts. The proposed algorithms are independent location considering measurement errors,” IEEE Trans. Power Del.,
vol. 22, no. 3, pp. 1335–1341, Jul. 2007.
of fault and source impedances [9] S. Luo, M. Kezunovic, and D. R. Sevcik, “Locating faults in the trans-
If prefault data are available, then a method is proposed that mission network using sparse field measurements, simulation data and
is applicable to all types of faults. If only fault data are em- genetic algorithms,” Electric Power Syst. Res., vol. 71, no. 2, Oct. 2004.
[10] C. S. Chen, C. W. Liu, and J. A. Jiang, “A new adaptive PMU based
ployed, we have suggested approaches for unsymmetrical faults. protection scheme for transposed/untransposed parallel transmission
Evaluation studies based on EMTP simulation data have demon- lines,” IEEE Trans. Power Del., vol. 17, no. 2, pp. 395–404, Apr. 2002.
strated quite promising results. It shows that for both methods, [11] Y. Liao and S. Elangovan, “Unsynchronized two-terminal transmis-
sion line fault location without using line parameters,” Proc. Inst. Elect.
the fault-location estimate is desirably not sensitive to measure- Eng., Gen., Transm. Distrib., vol. 153, no. 6, pp. 639–643, Nov. 2006.
ment errors while line parameter estimates are rather sensitive [12] J. Grainger and W. Stevenson, Power System Analysis. New York:
to measurement errors. Thus, relatively precise measurements McGraw-Hill, 1994.
[13] “Alternative Transient Program, User Manual and Rule Book,” Leuven
are required to obtain accurate line parameter estimates. EMTP Center, Leuven, Belgium, 1987.

REFERENCES
[1] L. Eriksson, M. M. Saha, and G. D. Rockefeller, “An accurate fault
locator with compensation for apparent reactance in the fault resistance Yuan Liao (S’98–M’00–SM’05) is an Assistant
resulting from remote-end infeed,” IEEE Trans. Power App. Syst., vol. Professor in the Department of Electrical and Com-
PAS-104, no. 2, pp. 424–436, Feb. 1985. puter Engineering at the University of Kentucky,
[2] A. T. Johns, P. J. Moore, and B. Whittard, “New technique for the Lexington. He was an R&D Consulting Engineer and
accurate location of earth faults on transmission systems,” Proc. Inst. then Principal R&D Consulting Engineer with the
Elect. Eng., Gen., Transm. Distrib., vol. 142, no. 2, pp. 119–127, Mar. ABB Corporate Research Center, Raleigh, NC. His
1995. research interests include protection, power-quality
[3] M. Sachdev and R. Agarwal, “A technique for estimating transmission analysis, large-scale resource scheduling optimiza-
line fault locations from digital impedance relay measurements,” IEEE tion, and EMS/SCADA design.
Trans. Power Del., vol. 3, no. 1, pp. 121–129, Jan. 1998.
[4] J. Izykowski, R. Molag, E. Rosolowski, and M. Saha, “Accurate loca-
tion of faults on power transmission lines with use of two-end unsyn-
chronized measurements,” IEEE Trans. Power Del., vol. 21, no. 2, pp.
627–633, Apr. 2006.
[5] A. A. Girgis, D. G. Hart, and W. L. Peterson, “A new fault location Ning Kang (S’07) received the bachelor’s degree in
technique for two-and three-terminal lines,” IEEE Trans. Power Del., electrical engineering from Xi’an Jiaotong Univer-
vol. 7, no. 1, pp. 98–107, Jan. 1992. sity, Xi’an, China, in 2004 and is currently pursuing
[6] S. M. Brahma, “Fault location on a transmission line using synchro- the Ph.D. degree in electrical and computer engi-
nized voltage measurements,” IEEE Trans. Power Del., vol. 19, no. 4, neering from the University of Kentucky, Lexington.
pp. 1619–1622, Oct. 2004. Her research interests include protection and
[7] D. J. Lawrence, L. Cabeza, and L. Hochberg, “Development of an ad- power-quality analysis.
vanced transmission line fault location system, Part II-Algorithm de-
velopment and simulation,” IEEE Trans. Power Del., vol. 7, no. 4, pp.
1972–1983, Oct. 1992.

Authorized licensed use limited to: ExxonMobil. Downloaded on March 28,2022 at 06:05:02 UTC from IEEE Xplore. Restrictions apply.