You are on page 1of 9

1

A New Impedance-Based Fault Location


Method for Radial Distribution Systems
K. Ramar, Senior Member, IEEE, and E. E. Ngu

Abstract-- A new impedance-based fault location method


suitable for radial distribution systems is presented in this paper.
The method uses the fundamental phasor components of voltage
and current signals available at the distribution substation end
only. Considering the unbalanced nature of the distribution
network with single-phase and two-phase laterals, and
unbalanced loads the fault location algorithm is derived using
phase-component analysis. The multiple-estimation problem
which is one of the main drawbacks of the one-end impedance
based algorithms is solved in the present method using the
estimated current information in the healthy phases. The
methodology is based on the during-fault and pure-fault values of
current phasors at the measuring end of the feeder and the
distribution matrix derived using the pure-fault equivalent circuit.
The efficacy of the method is demonstrated by simulating
different distribution system configurations with different source
impedances and fault types. The limitation of the proposed
method is also discussed.
Keywords: Radial distribution system, fault location,
impedance-based method, multiple-estimation problem.

I. INTRODUCTION

CCURATE fault location in power transmission and


distribution systems is very important for the
maintenance crews to reach the faulted section and undertake
the repair work quickly. In addition, the faults affect the
power quality in terms of service discontinuity and
propagation of disturbances created by faults. The fault
location can be determined by either monitoring the traveling
waves (TWs) generated by the faults (TW-based methods) or
processing the steady-state fundamental frequency
components of the voltage and current signals (Impedancebased methods). Several TW-based and impedance-based
fault location methods suitable for transmission systems have
been widely reported [1] [3]. These methods are not suitable
for distribution systems due to their specific characteristics
such as (i) radial topology of the system, (ii) lack of
measuring devices with very high sampling rate, (iii) presence
of single-phase and two-phase laterals along with
heterogeneous three-phase feeders, etc.
Several methods of fault location applicable to distribution
This work is partly supported by MOSTI E-Science Fund for project No.
of 03-02-01-SF0145.
K. Ramar is with the Faculty of Engineering, Multimedia University
Malaysia (e-mail: ramar@mmu.edu.my).
E. E. Ngu is with the Faculty of Engineering, Multimedia University
Malaysia (e-mail: eengu@mmu.edu.my).

978-1-4244-6551-4/10/$26.00 2010 IEEE

systems have been reported in the literature. The TW-based


methods [4] [5] mainly use the wavelet transform to detect
the high frequency disturbance caused by the fault in order to
estimate the fault location. Due to the existence of short and
heterogeneous feeders with many taps the TW generated by
the fault will undergo many reflections and refractions before
reaching the measuring end of the distribution system. Also,
because of the very short distance between the fault location
and the measuring end of the distribution system the time
difference between the incident and reflected TWs which is
required to estimate the fault location will be extremely short.
Therefore, measuring the time difference between the incident
and the reflected waves accurately even with very high data
sampling rate will be very difficult. Hence the TW-based
methods which are very popular for long transmission systems
are difficult to apply to distribution systems. Therefore,
researchers have concentrated on impedance-based methods
for fault location in distribution systems [6] [13].
The impedance based methods use the fundamental phasors
of voltages and currents measured at the substation end. The
equivalent distance to the fault from the substation end
(measuring end) is estimated based on the impedance
calculated using these phasor values. The pure-fault
(superimposed) and during-fault voltage and current values
are used in [1], [6] for computing the fault distance. An
iterative approach of single-ended fault location in distribution
system based on the concept of superimposed voltages and
currents is presented in [6]. A comparative study of ten
different fault location methods applicable to radial
distribution systems is presented in [11]. A new fault location
algorithm using direct circuit analysis specifically suitable for
unbalanced distribution system is presented in [9]. It uses the
current distribution factor calculated based on negative
sequence network which is suitable for balanced network and
the direct current analysis and matrix lemma for unbalanced
network. But the paper considers only phase-to-ground faults
and hence cannot be used for a general case.
The main drawback of the impedance-based methods
discussed above is the multiple estimations due to the
existence of multiple possible fault points at the same
electrical distance from the measuring end [13]. A solution to
solve this problem is presented in [13] using the load current
information in the unfaulted (healthy) phases. Though this
method identifies the faulted feeder in most cases the
calculated fault distance becomes inaccurate as the load
current in the faulted phases are neglected. Inaccuracy

increases with the increase in fault resistance. Other proposed


solutions for multiple estimations of fault location are by
using (i) fault sensors which detect abnormal fault current and
voltage events [15], (ii) voltage measurements at all the nodes
to identify the faulted feeder [14] and (iii) knowledge based
techniques [12]. The additional cost and the computational
difficulties involved in the above methods make them not
attractive for distribution systems.
In this paper a new impedance-based fault location method
suitable for radial distribution systems is presented. The
proposed method uses the voltage and current signals
available at the distribution substation end only. The
fundamental phasor components of the voltage and current
signals required to compute the fault distance can be extracted
from the measured digital signals by using Digital Fourier
Analysis. Considering the unbalanced nature of the
distribution network with single-phase and two-phase laterals,
the fault location algorithm is derived using the phasecomponent analysis instead of the commonly used
symmetrical component analysis. To compute the fault
distance the prefault current in the faulted feeder is not
neglected as is done in [13] and hence the fault location
accuracy is considerably improved. The method uses the
current distribution matrix derived using the three-phase purefault (superimposed) network. The multiple-estimation
problem is resolved by examining the estimated fault currents
in the healthy phases. The performance of the proposed
method is demonstrated considering three typical distribution
networks thereby proving that the method accurately
computes the fault location by eliminating the problem of
multiple estimations in three-phase radial distribution systems.
The fundamental principle of the proposed fault location
algorithm is explained in Section II. Application of the
method to a feeder with multiple load taps is discussed in
Section III. How the multiple-estimation problem is solved is
explained in Section IV. The sensitivity of the new fault
location method to errors in load estimation and feeder
impedances is discussed in Section V. The discussion and
conclusions are given in section VI.

analysis of the network to the left of the node S. Let ZR be the


3-phase equivalent impedance of the RDS to the right of node
R. Assuming a lumped parameter model for feeder S R, the
series impedance of the feeder is represented by ZSR. The
shunt capacitance of the feeder is neglected. As the length of
the feeders in a typical distribution system is normally very
short the assumption of lumped parameter model without
shunt capacitance is justified. In the figure,

II. FAULT LOCATION METHOD

Let a shunt fault occur at F which is at a distance ml from


node S as shown in Fig. 1. The length of the feeder is denoted
by l and m is the per-unit fault distance based on the length of
the feeder l. Let RF be the fault resistance matrix.
During the fault, the currents in the feeder are shown in the
figure. IS is the (during-fault) current at node S, IR is the
current at node R, and IF is the fault current. Referring to the
figure,

The principle of fault location considered in this paper is


illustrated first with reference to the equivalent system shown
in Fig. 1. It is the equivalent of a general 3-phase radial
distribution system (RDS) between any two adjacent nodes S
and R. Taking into account the unbalanced nature of the
distribution system the analysis is done in phase components.
Let S be the substation side node of the feeder S R and R
be the node at the other end of the feeder. Let the equivalent
impedance matrix of the radial distribution system to the left
of the node S including the substation source impedance be ZS
and the equivalent source voltage be ES. It is assumed that the
voltage and current measurement data are available at the
substation. The during-fault voltage at node S and the prefault and during-fault currents from node S to node R can be
computed by performing simple three-phase load flow

Fig. 1. Equivalent system

Z Saa
Z S = Z Sba
Z Sca

Z Sab

Z Raa
Z R = Z Rba
Z Rca

Z Rab

Z SR

Z Sbb
Z Scb

Z Sac
Z Sbc
Z Scc

(1)

Z Rac
Z Rbc
Z Rcc

(2)

Z Rbb
Z Rcb

Z aa
= Z ba
Z ca

Z ab
Z bb
Z cb

Z ac
Z bc
Z cc

E Sa
E S = E Sb
E Sc

(3)

(4)

VS = mZ SR I S + RF I F

(5)

where

VS = [VSa VSb VSc ]

(during) fault condition,

I S = [I Sa

I Sb

is the voltage of node S at the

I Sc ] and
T

I F = [I Fa

I Fc ] .
T

I Fb

(6)

The superscript T denotes the matrix transpose.


An expression for m can be obtained by pre-multiplying (5)
by the complex conjugate transpose of IF, denoted by IF* and
equating the imaginary parts as

m=

imag I F V S

imag I F Z SR I S

(7)

It can be noted that in (7) m and IF are unknowns. Using


the pure fault equivalent circuit, IF can be written in terms of
the pure- fault (superimposed) current at the node S as

I F = [(1 m) Z SR + Z R ] 1 [ Z S + Z SR + Z R ] I Sp = D(m) I Sp (8)


where

D ( m ) = [(1 m ) Z SR + Z R ] 1 [ Z S + Z SR + Z R ]

load taps as shown in Fig. 2. Let the impedance matrices of


the feeders 0-1, 1-2, and 2-3 be Z01, Z12, and Z23, respectively.
Let the impedance matrices of the taped lines connected to tap
points 1, 2, and 3 be Z1, Z2, and Z3, respectively. The length of
the feeders 0-1, 1-2, and 2-3 are denoted by l1, l2, and l3,
respectively. From the given load data the equivalent load
admittance matrices can be easily computed using the prefault load voltages. The pre-fault voltages at all the nodes can
be obtained by performing three-phase load flow analysis. Let
the equivalent load admittance matrices of loads L1, L2 and
L3 be YL1, YL2 and YL3, respectively. The method of computing
the equivalent load admittance matrices is explained in [6].
When the system is subjected to fault, the fundamental phasor
quantities of during-fault currents and the during-fault voltage
at the substation node 0 can be computed from the measured
digital data by performing digital Fourier analysis.

(9)

ISp is the pure-fault (superimposed) current vector at node S,


D(m) is the current distribution matrix relating ISp and IF.
The current ISp can be computed from the pre-fault and
during-fault currents measured at node S as

I Sp = I S I S 0

(10)

where IS0 is the pre-fault current vector at node S. Substituting


(8) in (7),

m=

imag I Sp D(m)*VS

imag I Sp D(m) Z SR I S

(11)

* denotes the complex conjugate transpose of a matrix.


Equation (11) is a nonlinear equation in one unknown
variable, m which can be easily evaluated using any one of the
standard numerical algorithms. A possible simple way to solve
(11) is as follows:
Step 1: Assume m = 0.5
Step 2: Compute D(m) using (9) as all qualities are known.
Step 3: Evaluate the update value of m using (11).
Step 4: Repeat steps 2 and 3 till it converges.
It has been observed that, in most cases, the algorithm
converges in 3 to 4 iterations. After convergence, the fault
current can be computed using the current distribution matrix
D(m), if required. It can be observed that the matrix
[(1 m) Z SR + Z R ] in (9) is nonsingular even if ZR is
singular and hence the current distribution matrix D(m) can
always be computed.
III. FAULT LOCATION IN A SYSTEM WITH MULTIPLE LOAD
TAPS
In order to explain the procedure for a distribution system
with multiple load taps, consider a simple system with three

Fig. 2. Feeder with three load taps

Firstly, the fault is assumed to be in feeder 0-1. Compute


the equivalent impedance to the left of the node 0, Z0S. In this
case, Z0S = ZS. Also, compute the equivalent impedance of the
distribution system to the right of node 1 including load L1
and denote it as Z1R. The derivation of the equivalent
impedance Z1R is shown in Appendix I. From the measured
values of pre-fault and during-fault currents at node 0
compute the pure-fault (superimposed) current as

I 01 p = I 01 I 010

(12)

where
I01 is the during-fault current flowing from node 0 to node 1
and
I010 is the pre-fault current flowing from node 0 to node 1.
Note that, for brevity, in many places the impedance matrix
and the current/voltage vector are simply written as
impedance and current/voltage.
Next, compute the current distribution factor D(m) using (13)

D(m) = [(1 m) Z 01 + Z1R ]1[ Z 0 S + Z 01 + Z1R ]

(13)

where Z01 is the series impedance of the feeder 0-1. Once


D(m) and I01p are known the per unit fault distance m can be
computed following the procedure explained in Section II. If
the value of m is between 0 and 1 then the assumed faulted
feeder 0-1 is the correct faulted feeder and ml1 is the computed
fault distance from node 0. The value of m greater than 1
indicates that the fault is beyond node 1.
If the value of m computed in the previous step is greater
than 1 then proceed further by assuming that the fault is in
feeder 1-2. Compute the equivalent impedance Z1S to the left

of the node 1, and compute the equivalent impedance Z2R to


the right of node 2 including the load L2. Also, compute the
during fault voltage at node 1, and the prefault and during
fault currents flowing from node 1 and node 2 using the
voltage and current data at node 0. Then, following the
procedure given above compute the per unit fault distance m.
If the value of m lies between 0 and 1 then the feeder 1-2 is
the faulted feeder and ml2 is the fault distance from node 1.
The value of m greater than 1 indicates that the fault is beyond
node 2. Repeating this procedure for feeder 2-3 the fault
location can be obtained.
This procedure gives the correct faulted feeder and the
correct fault location in 3 or less than 3 attempts. Once the
faulted section is identified there is no further iteration to
compute the fault location. It is interesting to note that the
method presented in this paper requires not more than n
attempts for the distribution system with n load taps. On the
other hand the procedure given in [6] requires a large number
of iterations, generally much more than n to locate the fault. A
numerical example is presented below to demonstrate the
effectiveness of the proposed method.
A. Example 1
Consider the 11-kV distribution system shown in Fig. 3.

locations and various fault resistances to get the fault currents


and voltages at the substation node. Tests with all the four
types of faults are conducted using MATLAB. The faulted
feeder and the fault distance are computed using the proposed
method using the substation end voltage and current
information, and the results are tabulated for symmetrical and
phase-to-ground faults only in Table 1(a) and 1(a),
respectively. The fault location process is started by first
assuming the feeder closest to the substation end as the faulted
one and then repeating the procedure from one feeder to the
other in a sequence. Once the calculated m value lies between
0 and 1 the procedure is terminated as this value of m
indicates the correct fault distance. From the tables it can be
seen that for the actual faulted feeder 1-2 two estimations only
have been done. It is interesting to note that the proposed
method gives accurate fault location for all the types of faults.
It is not surprising to note that the error in fault distance
estimation is almost zero as the accurate steady-state values of
current and voltage data are used in the algorithm. In practice,
there may be some error in estimating the phasor values from
the recorded transient fault data and hence the fault location
may contain some errors. However, the inaccuracy will be due
to the inaccuracy in estimating the phasor values from the
transient digital data. As the aim of the test here is to
demonstrate the efficacy of the proposed method, the accurate
steady-state values have been used. It may be observed that
there is no possibility of multiple fault location in this example
for a given set of data.
IV. MULTIPLE ESTIMATION PROBLEMS

Fig. 3. System for Example 1

The system data are given in Appendix II. The feeder


impedances are taken from IEEE 4-node data [16]. It is
assumed that the loads are directly connected to the nodes
without tap feeders. Load flow analysis is performed to get the
pre-fault node voltages and pre-fault current at the substation
node 0. Fault simulations are performed for various fault

In the example system considered in the previous section there


is no possibility of multiple estimations. But, in most of the
practical radial distribution systems there may be more than
one point which will be at the same electrical distance from
the substation end. Hence the computation of fault location
will require some additional information. In this section a
method

TABLE 1(a)
THREE-PHASE SYMMETRICAL FAULT

5
TABLE 1(b)
PHASE-TO-GROUND FAULT IN PHASE a

is proposed to identify the correct faulted feeder and location


based on the estimated fault current(s) in the healthy phase(s).
The method can best be illustrated by a simple three-feeder
system shown in Fig. 4. Assume that the fault data at the
substation only are available. Using the data available at the
substation there is a possibility of multiple fault locations.

Fig. 4. Three-feeder distribution system

Following the procedure explained in Section II, compute


the fault distance by assuming sequentially the faulted feeder
as 0-1, 0-2 and 0-3. For each case, estimate the fault currents
in all the phases including the healthy phases by using the
relation (8) which is repeated below

I F = D(m) I Sp

(14)

where ISp is the pure-fault current at the substation end. Let IFb
and IFc be the estimated fault current in the health phases for a
phase-to-ground fault in phase a. Define an error index as
2

eindex = I Fb + I Fc

(15)

The assumed faulted feeder for which the eindex is the


minimum is the correct faulted feeder and the corresponding
value of m gives the correct fault distance.
In the case of phase-to-phase faults or phase-to-phase-to

ground faults involving phases b and c, the error index is


defined as

eindex = I Fa

(16)

Based on the procedure explained above a numerical


example is worked out in the next subsection.
A. Example 2
Consider the simple three-feeder system shown in Fig.4.
The system parameters are selected from the typical IEEE-34
node test feeder [13], [16]. The feeders 0-1 and 0-3 have the
feeder configuration 300 and feeder 0-2 has the feeder
configuration 301. The feeder lengths are assumed to be 10
km each and the feeder capacitances are neglected. The load
L1 at the end of feeder 0-1 has the value similar to the load at
node 860 and the loads L2 and L3 have the value similar to
the load at node 830. The source impedance is assumed to be
1.0+j8.0 per phase and the source voltage is taken as 11.0
kV. The short line connecting the substation to the node 0 is
assumed to have negligible impedance.
The test results for phase-to-ground fault in phase a are
presented in Table 2. Tests have been done for many fault
resistances, but the results are tabulated for the fault resistance
of 1 and 10 only. From the table it is clear that the
proposed method is able to correctly identify the faulted
feeder and the fault distance. For example consider the fault in
feeder 0-1. The per unit fault distance m for all the assumed
faulted feeders lies between 0 and 1 indicating possible
multiple fault locations. But the error index for the assumed
faulted feeder 0-1 is very small compared to the error index
obtained for other assumed faulted feeders and hence it is
clear that the fault is in feeder 0-1. Tests have also been
conducted with phase-to-phase faults and phase-to-phase-toground faults and the method correctly identifies the faulted
feeder and estimates the fault location with high accuracy.

6
TABLE 2
PHASE-TO-GROUND FAULT IN FEEDER 0-1 AT A DISTANCE OF ml FROM NODE 0

phase loads is considered.


Figure 5 gives the system configuration. Unlike the system
considered in Example 1, the impedances of load tap feeders
which are different from that of the main feeders are also
considered in this example. The system data are given in
Appendix III.

Fig. 5. Distribution system for Example 3


TABLE 3
COMPARISON WITH THE METHOD IN [12] FOR PHASE-TO-GROUND FAULTS

For comparison, the fault distance is also computed using


the method proposed in [13] for phase-to-ground fault and the
results are given in Table 3. It can be observed that the faulted
feeder is correctly identified by both the methods, but the
present method computes the fault distance more accurately
for the fault resistance of 10 . The error in fault distance
estimation using the method in [13] increases with increase in
the fault resistance. This is expected as the method in [13]
neglects the load impedance in the faulted phase and the
impedance of the portion of the faulted line beyond the fault
point.
B. Example 3
Example 2 is presented to prove the efficacy of the
proposed method to solve the multiple-estimation problem.
Here, another example with a typical distribution system
consisting of a number of load taps with two-phase and three-

To illustrate the fault location procedure, a phase-to-ground


fault in phase a is created in feeder 4-5 at a distance m = 0.2
from node 4. The fault resistance is 10 . As explained
earlier the fault, at first, is assumed in feeder 0-1 and the per
unit fault distance m is computed as 3.758 which indicates that
the fault is not in the assumed feeder and is well beyond node
1. The calculated fault distance is (ml =) 11.274 km from node
0 and hence the approximate fault location may lie between
nodes 4 and 5 or between node 4 and the load L4 thus
indicating two possibilities. Therefore, one can directly
assume that the fault is either between nodes 4 and 5 or
between node 4 and load L4 and proceed to find the fault
distance. But here it is suggested to proceed systematically by
assuming next that the fault is in feeder 1-2. If the calculated
value of m is greater than 1 then proceed to the next feeder.
The results of this systematic procedure are given in row 1 of
Table 4. It is found that the value of m for the assumed faulted
feeder 4-5 and feeder 4-4L is less than 1. Based on the error
index it is clearly identified that the actual faulted feeder is 4-5
and the fault distance is (0.2 x 2.5 =) 0.5 km from node 4.
The results for different fault location in feeder 4-5 and with
the fault resistances 1 and 10 are presented in the table.
From the table it may be observed that the proposed method
gives accurate fault location for all the cases. The error index
is not computed if the value of m is greater than 1 as it clearly
indicates that the assumed feeder is not the faulted one.

TABLE 4
TEST RESULTS FOR THE PHASE-TO-GROUND FAULT IN FEEDER 4 5

7
TABLE 6

V. SENSITIVITY TO LOAD VARIATIONS


In practice the loads on the system are subjected to
variations with the times of day and hence it will be very
difficult to estimate the correct load data to be used in the fault
location algorithm. Therefore it is very important to know as
to what extend the accuracy of the fault location algorithm is
affected as a result of errors in load estimation. For this
purpose the voltage and current data for the system in
Example 3 are generated by randomly varying all the loads by
some percentage from the value used in the algorithm and the
error in the fault location is estimated. Table 5 summarizes
the performance of the algorithm for three types of faults
when the loads are subjected to 2%, 5% and 10%
variations randomly. The fault is assumed in feeder 4-5 at a
distance m = 0.6 from node 4. The fault resistance is 10 . In
all the cases the correct faulted feeder is identified and the
maximum error is found to be 1.81%. These results clearly
indicate the robustness of the fault location algorithm to errors
in the load estimation.
To know the effect of the error in feeder impedances on the
accuracy of fault location a sensitivity study is performed. For
this study the voltage and current data for the system in
Example 3 are generated by randomly varying all the seven
main feeders and seven tap feeders by some percentage from
the value used in the algorithm and the maximum error in the
fault location for each case is estimated. For this study the
fault is assumed in feeder 4-5 at a distance m = 0.6 from node
4. The fault resistance is 10 . Table 6 shows the worst case
fault location errors for three types of faults when the feeder
impedances are subjected to random variations of 1%, 2%
and 5%. In all the test cases the correct faulted feeder is
identified. From the table it is observed that the fault location
is sensitive to the values of the feeder impedances. This result
is expected as the method relays heavily on the feeder
impedances including the impedance of the faulted feeder to
compute the fault location. Therefore accurate values of the
feeder impedances should be supplied to the algorithm to
obtain accurate fault location. In most cases the details of the
feeder data are accurately known and hence computing the
feeder impedances accurately may not be a major problem.
TABLE 5
EFFECT OF LOAD ESTIMATION ERROR ON FAULT LOCATION ACCURACY

EFFECT OF ERROR IN FEEDER IMPEDANCES ON FAULT LOCATION ACCURACY

VI. DISCUSSION AND CONCLUSION


A new fault location method suitable for radial distribution
systems using one-end data is presented in this paper. The
efficacy of the method has been demonstrated with the aid of
three numerical examples. The information about the type of
fault is not required for the fault distance calculation.
However it is needed for resolving the multiple-estimation
problem. The type of fault can be determined easily by
observing the magnitude of the fault current recorded at the
substation or from the magnitudes of fault currents computed
using (14).
The data required for the fault distance computation has
been obtained by steady-state fault simulation using
MATLAB. The method accurately locates the fault as the
phasor values of the fault current and voltage data are
accurate. In practice the steady-state fault data have to be
extracted from the recorded transient data using Digital
Fourier analysis and hence there will be small error in the
location depending upon the accuracy of the digital Fourier
analysis. To verify the accuracy of the proposed method with
transient data, digital simulation of distribution systems with
different fault conditions are being conducted using the
software PSCAD/DigSilent and the results will be reported
shortly.
As the multiple-estimation problem has been resolved in
this method using the magnitudes of fault currents in the
healthy phases it cannot be used to find the actual faulted
feeder from the multiple possibilities for three-phase faults.
However, the method works well to locate the fault accurately
for this three-phase fault case also if the multiple possibilities
do not exist. This has been demonstrated in Example 1. Work
is in progress to resolve the multiple-estimation problem for
three-phase faults also.
In the proposed method it is assumed that the loads are of
constant impedance type during the fault analysis which is the
normal valid assumptions in most cases [6], [9], & [13].
However, the method can be extended to cases where the
loads are different from constant impedance type. Work in this
direction is also in progress. Also it may be observed that the
capacitances of the feeders are neglected in the fault location.
But, the underground cable feeders may have significant shunt
capacitances. In such cases, the effect of the shunt
capacitances of underground feeders may be taken into effect

by modeling the feeders, preferably, by equivalent -models.


In the computation of equivalent impedances to the left and
right of nodes of the faulted feeder the equivalent -model of
the feeders can be used and the fault distance can be
calculated. For the faulted feeder the effect of the shunt
capacitance are neglected as the voltages of the faulted node
will be very small. Capacitor banks connected to certain nodes
can be modeled as fixed shunt admittances and the fault
location procedure can be continued as explained earlier.
Similar to other papers in this area, the fault location
algorithm derived in this paper is for 3-phase 3-wire
distribution systems. However, a 3-phase 4-wire system with
neutral grounding at all nodes can be converted to an
equivalent 3-phase 3-wire system using Kron reduction
technique. Finally, it may be observed that the expression
given in (13) to evaluate the current distribution matrix cannot
be used as given if the fault is in a two-phase or single-phase
lateral. A slight modification should be done in the expression
to take care of this problem.
In summary, the fault location method proposed in this
paper is novel and identifies the faulted feeder and the fault
location accurately for all types of faults except three-phase
faults. It has been shown that the method is robust for small
errors in load estimation. However it is sensitive to error in
feeder impedances. Therefore, accurate values of feeder
impedances should be provided to the algorithm to obtain
accurate fault location. Obtaining accurate feeder impedance
values may not be a difficult job as the details of the feeders
are known almost accurately. The extension of the method for
three-phase faults is in progress and the result will be reported
in near future.
VII. APPENDICES
APPENDIX I
Computation of Y3R, Y2R and Y1R
Referring to Fig. 2, the equivalent admittance of system to
the right of node 3 (including load L3) is given by

Y3 R = YL 3 (U + Z 3 YL 3 ) 1

Y0S = YS

Y1S = YL1 (U + Z1 YL1 ) + YS (U + Z 01 YS )

(A-4)

Y2S = YL2 (U + Z1 YL2 ) + Y1S (U + Z12 Y1S )

(A-5)

APPENDIX II
Data for Example 1
Substation data: 58 MVA SCL, 11 kV, Zs0/Zs+ = 1.0, X/R =
10.
Feeder data:
0.4013+ j1.4133 0.0953+ j0.8515 0.0953+ j0.7266
Z = 0.0953+ j0.8515 0.4013+ j1.4133 0.0953+ j0.7802/km
0.0953+ j0.7266 0.0953+ j0.7802 0.4013+ 1.4133i

Load data: L1 = 3-ph load of 100 kW, pf =0.9 lagging


L2 = 2-ph load of 50 kW/ph, pf = 0.9 lagging
L3 = 3-ph load of 80 kW, pf = 0.9 lagging
APPENDIX III
Data for Example 3
The substation data and main feeder data are the same as in
Example 1 (APPENDIX II).
The impedance of the load tap feeder,
0.8021+ j0.8006 0.1261+ j0.3467 0.1278+ j0.3009
Ztap = 0.1261+ j0.3467 0.8021+ j0.8006 0.1240+ j0.2755/km
0.1278+ j0.3009 0.1240+ j0.2755 0.8021+ j0.8006

Lengths of the main feeders and load tap feeders are given
in the figure.
Loads: L1 = 500 kVA, 3-ph
L2 = 25 kVA, b c
L3 = 50 kVA, a b
L4 = 150 kVA, 3-ph
L5 = 250 kVA, 3-ph
L6 = 375 kVA, 3-ph
L7 = 500 kVA, 3-ph
Power factor of all loads = 0.9 lagging

(A-1)
VIII. REFERENCES

and the admittance of system to the right of node 2 is given by

[1]

Y2R =Y3R (U + Z23Y3R)1 +YL2 (U + Z2 YL2)1

[2]

(A-2)

where
[3]

U is the unit matrix of dimensions 3


Similarly, the admittance of system to the right of node 1 is
given by

[4]

Y1R = Y2R (U + Z12 Y2R )1 + YL1 (U + Z1 YL1)1

[5]

(A-3)

From the admittance matrices the impedance matrices can


be easily computed by inverting the matrices.
Computation of Y0S, Y1S and Y2S

[6]

K. Ramar and A. A. Eisa, New Fault Location Algorithms for Overhead


Transmission Lines, Int. J. Power and Energy Syst., vol. 28, no. 2, pp.
241 251, 2008.
M. Gilany, D. K. Ibrahim and E. S. Tag Eldin, Traveling-Wave Based
Fault-Location Scheme for Multiend-Aged Underground Cable
Systems, IEEE Trans. Power Delivery, vol. 22, no. 1, pp. 82 88,
2007.
X. Y. Zheng, X. M. Li, J. Y. Ding, Z. Y. Duan, Study on Impedancetraveling Wave Assembled Algorithm in One-terminal Fault Location
System for Transmission Lines, Third International Conference on
DRPT2008, Nanjing, China,6-9 April 2008.
F. H. Magnago, and A. Abur, Fault location using Wavelets, IEEE
Trans. Power Delivery, vol. 11, no. 3, pp.1475 1480, 1998.
X. A. Qin, X.J. Zeng, X.L. Zhang, Z.W. Li, Traveling Wave Based
Distribution Lines Fault Location Using Hilbert-Huang Transform,
IEEE Industrial Application Society, Annual Meeting, pp. 1 5, 2008.
R. K. Aggarwal, Y. Aslan, and A. T. Johns, New concept in fault
location for overhead distribution systems using superimposed
components, Proc. IEE Geer. Transm. Distrb. vol. 144, no. 3, pp. 309
316, May 1997.

9
[7]
[8]
[9]
[10]

[11]
[12]

[13]

[14]

M. Sachdev, R. Das, and T. S. Sidhu, Distribution line shunt fault


locations from digital relay measurements, Can. J. Elect. Comput. Eng.,
vol. 24, no. , pp. 41 47, Jan. 1999.
R. Das, M. Sachdev, and T. S. Sidhu, A fault locator for radial
subtransmission and distribution lines, Proc. Power Engineering
Society Summer Meeting, vol. 1, pp. 443 448, 2000.
M. Choi, S. Lee, D. Lee, and B. Jin, A new fault location algorithm
using direct circuit analysis for distribution systems, IEEE Trans.
Power Delivery, vol. 19, no. 1, pp. 35 41, Jan. 2004.
S. Lee, M. Choi, S. Kang, B. Jin, D. Lee, and B. Ahn, An intelligent
and effecient fault location and diagnosis scheme for radial distribution
systems, IEEE Trans. Power Delivery, vol. 19, no. 2, pp. 524-531, Apr.
2004.
J. Mora-Florez, J. Melendez, and G. Carrillo-Caicedo, Comparison of
impedance based fault location methods for power distribution systems,
Elect Power Syst Res. vol. 78, no. 4, pp. 657 666, Apr. 2008.
J. Mora-Florez, V. Barrera-Nunez and G. Carrillo-Caicedo, Fault
Location in Power Distribution systems Using a Learning Algorithm for
Multivariable Data Analysis, IEEE Trans. Power Del., vol. 22, no. 3,
pp. 1715-1721, Jul. 2007.
G. Morales-Espana, J. Mora-Florez and H. Vargas-Torres, Elimination
of Multiple Estimation for Fault Location in Radial Power Systems by
Using Fundamental Single-End Measurement, IEEE Trans. Power Del.,
vol. 24, no. 3, pp. 1382-1389, Jul. 2009.
L. Garcia-Santander, P. Bastard, M. Petit, I. Gal, E. Lopez and H. Opaza,
Down-conductor fault detection and location via a voltage based
method for radial distribution networks, IEE Proc. Gener. Transm.
Distrib., vol. 152, No.2, pp. 180 -184, Mar. 2005

[15] IEEE Guide for Determining Fault Location on AC Transmission and


Distribution Lines, IEEE Std. C 37.114 2004, (R2009).
[16] IEEE 34-Bus Test Feeder, Distribution System Analysis Subcommittee
http://ewh.ieee.org/soc/pes/dsacom/testfeeders.html

IX. BIOGRAPHIES
K. Ramar received the Bachelor of Engineering degree from University
of Madras in 1964, and Master of Technology and Ph.D. degrees from Indian
Institute of Technology Madras, India in 1968 and 1972 respectively. He
worked as a Professor in the Department of Electrical Engineering, Indian
Institute of Technology Madras until December 2002. He is currently a
Professor in the Faculty of Engineering, Multimedia University, Malaysia.
He has more than 40 years of teaching and research experience. He has
published more than 75 papers in leading international journals and
conferences. His fields of interest include Power System Operation and
Control, Computer Application to Power Systems, Power Systems Protection,
and Fault Location in Overhead Transmission Lines. He is a Senior Member
of IEEE.
E. E. Ngu was born in Malaysia. She received the Bachelor of Electrical
Engineering and Master of Engineering (Electrical) degrees from Universiti
Teknologi Malaysia in 2002 and 2004, respectively. She is currently working
as a Lecturer and pursuing her Ph.D. on part time basis at Multimedia
University, Malaysia. Her research interests include fault location on power
system and lightning protection.

You might also like