Fault Location in Radial Distribution Lines Using
Travelling Waves and Network Theory
Ajendra Dwivedi, Xinghuo Yu
School of Electrical and Computer Engineering, RMIT University, Melbourne, Victoria, Australia
ajendra.dwivedi@student.rmit.edu.au, x.yu@rmit.edu.au
Abstract— This paper proposes a method for locating a fault in a
radial distribution network using travelling waves together with
network theory. The single-ended travelling wave measurement
scheme is used to calculate the distance of the fault from the
terminal node. However, due to the radial structure of the
distribution network, there could be more than one possible fault
candidates at the same distance. The spanning tree structure and
unsymmetrical nature of such networks are then taken into
consideration to isolate the actual faulty link from all possible
fault candidates. The methodology is explained on a simple 12
node example and further verified on the IEEE 34 node test
feeder network using Matlab and Alternate Transients
Program/Electromagnetic Transients Program (ATP/EMTP).
Index Terms – Distribution system, fault location, network
theory, radial network, spanning tree, travelling waves.
I. INTRODUCTION
The size of distribution systems have grown immensely
with increasing power demand which has resulted in a
complex structure with large number of transmission lines.
This increased size with variety of loads adds to the
complexity of the network and poses new challenges in the
maintenance of reliable power supply. Such systems can
experience faults due to physical or electrical breakdowns
which could cause interruption in supply of power. To
maintain high quality of customer service, it is essential to fix
these faults and restore power supply to minimize
interruptions. Accurate location of these faults can help the
utilities in quick repair and maintenance of the problem and
hence reducing the downtime.
Various fault location techniques have been studied and
developed to identify and locate faults in power lines and
networks. References [1]-[4] give comprehensive reviews of
the existing fault location methods for power networks and
distribution networks. Fault location techniques usually
involve measurement and calculation of impedance, current,
voltage and other fault transients and the changes are used
along with other system parameters to predict the fault point.
Mora-Florez et al [5] compared 10 different fault location
methods based on impedances. The fault location methods can
be either single-ended where measurements are taken only at
one end of the faulty line or double-ended where these
parameters are measured at both ends. Han et al [6] proposed
a fault location algorithm for radial distribution systems with
single-ended measurements using a sinusoidal steady state
analysis method. One of the advantages of such approaches is
that they do not require the line parameters which make them
more robust. This has been rightly demonstrated by Radojevic
et al [7]. Djuric et al [8] proposed an algorithm which allows
978-1-4244-9312-8/11/$26.00 ©2011 IEEE 1051
calculating the fault distance irrespective of the fault
resistance and load.
Another method which has been a subject of interest for
many researchers is based on travelling waves [9]-[16].
Unlike the impedance based methods, travelling wave
technique can be used to measure distance to faults in all
kinds of power lines such as AC transmission lines, HVDC
transmission lines, lines with cable sections and overhead
lines, lines with T-branches etc. They are based on high
frequency signals where the fault location is provided by the
arrival time of the transients at the line terminals [1]. Such
fault location techniques can be broadly classified in to two
types, namely, single-ended measurement and double-ended
measurements. The important factor in this type of fault
location are detection of transient current or voltage signals
and signal processing using wavelet transform, correlation etc.
to calculate the arrival time at the measurement points [13]-
[16]. In case of double ended measurements, the faulty signals
are recorded at the two ends of the transmission line and the
arrival times help in calculating the fault distance. With the
developments in the global positioning systems (GPS), it is
easy to synchronize the recording devices at different nodes to
accurately calculate these time stamps. In case of single ended
method [12]-[14], fault distance is determined by analyzing
the fault travelling waves recorded at one end of the line.
Time difference between the initial detected fault surge and
the corresponding reflected surge is used for such
calculations. ATP/EMTP is a powerful simulation tool for
analysis of transients and electromagnetic waves. It has been
used quite often to compare and validate field measurement
results [17].
Travelling wave theory has been mostly used to calculate
fault distance in transmission lines between points of
measurements. However, it does address the problem of
locating a fault if there are branches in the transmission lines
such as in radial distribution systems. In such cases, there
could be multiple fault candidates at the same distance from
the measurement point due to the radial structure of the
network. Yu et al [18] reported the potential use of network
theory in power system research and in locating faults. This
paper utilizes the knowledge of tree-like structure and
unsymmetrical nature of the distribution systems to resolve the
above mentioned problem and identify the actual faulty line.
The paper is organized as follows: Section I gives the
introduction and rationale of this research. Section II lists
some preliminaries and basics of travelling waves and network
theory. Section III describes the modeling of a distribution
system as a spanning tree network and explains the
methodology used to calculate the fault distance and identify
the faulty link using travelling waves and network theory.
Section IV validates the methodology on the IEEE 34 node
test feeder system and demonstrates the identification of faulty
link and distance. Section V concludes the paper with
discussions and future recommendations.
II. PRELIMINARIES
This section describes some basics of the travelling waves and
network theory. It discusses the ideas behind travelling wave
theory in locating faults in transmission lines. It also includes
definitions of few basic network elements and other
topological parameters. It briefly explains the tree structures
and the way information can be searched through such
networks.
A. Travelling Wave Theory [1]
Fig. 1 Travelling Voltage and Current Waves due to Fault
The basic idea of travelling wave fault location theory lies
in the fact that when there is a fault in any power system it
generates electromagnetic current and voltage transients
which travel forward and backwards along the transmission
lines to all the terminals or nodes until a post-fault steady state
is reached. A lattice diagram illustrating the travelling of
waves due to fault transients is shown in Fig. 1. Forward (with
subscripts f) and reverse (with subscript r) voltage and current
waves leave the fault point and travel towards the nodes.
These nodes represent discontinuity or impedance change so
part of the energy is reflected back and the remaining continue
to travel towards other nodes of the network. kA and kB are
reflection coefficients which are determined by the
characteristic impedance ratios and represent wave
amplitudes. Similarly, tA and tB represent the travel times from
the fault point to the nodes A and B. Construction of lattice
diagrams becomes difficult if attenuation and distortion of
signals are taken into account, so time-frequency resolution of
the transients can also be used to determine the travel time of
these fault signals to the line terminals.
The voltage and current at any point x obey a set of partial
differential equations given by:
∂e ∂i ∂i ∂e
= −L and = −C (1)
∂x ∂t ∂x ∂t
Where, L and C are the inductances and capacitances of the
line per unit length and e and i are the electromagnetic voltage
1052
and current transients. The resistance is assumed to be
negligible so the solution to the above wave equations for a
lossless line in terms of forward and backward travelling
waves can be expressed as:
e( x, t ) = e f ( x − vt ) + er ( x + vt ) 

1 1  (2)
i ( x, t ) = e f ( x − vt ) − er ( x + vt )
Zc Zc 
Zc is the characteristic impedance of the transmission line
and v is the velocity of propagation of wave which is close to
the speed of light. They can be calculated in terms of line
parameters as follows:
L 1
Zc = and v= (3)
C LC
These travelling waves are recorded at the terminals or
nodes with Travelling Wave Recorders (TWRs). Further, they
can be processed to get the time taken by them to travel to the
nodes. These recordings can be done from one terminal of the
transmission line called single-ended measurement or from
both terminals of the line called double-ended measurements.
B. Network Theory and Spanning Tree [19][20]
A network can be defined as a set of nodes or vertices with
connections called links or edges. We can see in Fig. 2, that
vertex is the fundamental unit of a network, also called as site
(in physics) node (in computer science) or actor (in
sociology). They are connected together by lines called edges,
also known as bond (in physics), link (in computer science) or
a tie in (sociology).
A spanning tree or tree is part of a graph or a network G
such that it contains all vertices and some (or all) edges of the
original network in such a way that there is a unique path
between every pair of vertices and there are no loops or cycles
formed. Further, there can be weights associated with the links
of the network. These weights incorporate some of additional
features of the system under study apart from the topological
characteristics.
Travelling through a network or the flow of information
usually happens with an alternate sequence of nodes and edges
which is defined as path. Depth first search [20] is one of the
algorithms for traversing or searching through a tree structure.
We select and start at the root node and search to the farthest
point in a branch before selecting another branch.
Fig. 2 Spanning Tree and Depth First Search
 Fig. 2 shows an example of a tree structure and the method
of traversing through the network using depth first search. The
numbers within the circles give the order in which the nodes
are searched.

III. LOCATING FAULTS USING TRAVELLING WAVES AND

NETWORK THEORY
The following section discusses the methodology to calculate
the fault distance and identify the faulty line in a distribution
system based on network theory. It briefs the use of travelling
wave theory to record and process the fault transients. Further, Fig. 4 Sample Radial Network
this section defines the modeling of a radial distribution The above network can be modeled mathematically using
system with proper weights on edges to incorporate the the adjacency list. If there is a connection between node i and
topological property. Finally, the fault locating algorithm is j then it can be represented as an element (i, j) along with the
explained using a sample network. weight wij of the connection. All other nodes which are not
A. Single-Ended Travelling Wave Recording and Processing connected have a weight 0. In the current example, the weight
of the connection is the distance between the nodes. The
Travelling wave recorders (TWRs) are installed at the two
complete connection of the sample network has been listed in
ends of the radial distribution systems. Whenever there is a
Table I.
fault, current and voltage transients in the form of
electromagnetic waves are originated at the fault point and
Table I. Network Connection
travel to the terminals. These signals are recorded and
Node Node
processed individually at both ends of system using single- Weights Weights
Connections Connections
ended fault location technique.
(1,2) 11 (6,8) 14
The recorded transients are processed using wavelet
(2,3) 7 (8,9) 20
transform to extract the time information at each node [13]-
[16]. The wave velocity v, which is close to the speed of light, (3,4) 12 (8,10) 8
can be calculated using the line constant program of (3,5) 30 (10,11) 19
ATP/EMTP [21][22]. (2,6) 12 (10,12) 15
As shown in Fig. 3, in case of single-ended measurement, (6,7) 22 All Others 0
the time difference ∆t = (t 3 − t 2 ) between the initial fault
transient and the corresponding reflected surge from fault C. Fault Location in Radial Networks
point is calculated and then the fault distance from the Fig. 5 illustrates the block diagram of the fault locator. It
measurement terminal can be given as: consists of two travelling wave data acquisition units at the
∆t.v two ends of the radial network which record the
x= (4)
2 electromagnetic current and voltage transients generated due
to the fault. The wave data is analyzed using wavelet
transform to extract the time stamps and calculate the fault
distance. Next, different fault candidates are isolated using

Travelling Wave Travelling Wave

Data Acquisition Data Acquisition
Unit 1 Unit 2

Calculate Fault Calculate Fault

Distance X1 From Distance X2 From
TWR 1 TWR 2

Fig. 3 Lattice Diagram for Single-Ended Fault Measurement Depth First Search Depth First Search
From TWR 1 to From TWR 2 to
Find Fault Find Fault
B. Network Model of a Radial System Candidates Candidates
Fig. 4 shows a sample radial network with 12 node and 11
edges. The numbers within the circles are the node identities Output Faulty Link
and Fault Distance
and the number on the edges show the length of the
connecting links.
Fig. 5 Block Diagram of Fault Locator

1053
depth first search from both ends. The candidate which is
common in both the searches is output and the faulty link.
Although the fault distance can be calculated form any one
end but due to the radial structure of the network, there could
be multiple fault candidates at the same distance from the
measurement node. If x is the calculated fault distance then the
flowchart in Fig. 6 explains the algorithm to identify and
isolate all possible fault candidates.
Calculated Fault
Distance = x
Sum = 0
Advance to a new adjacent node
Sum = Sum + Lf
(Lf is length of link traversed
forward)
The identification of the faulty link can be summarized in
the following steps:
(a) Model the radial system as a network and store the
topology and distance between nodes in an
adjacency list as shown in Table I.
(b) Time stamps are recorded individually at the two
ends of the radial system using single-ended
travelling wave measurement.
(c) Calculate the distance of the fault from one end of
the network using Equation (4).
(d) Use depth first search to isolate all the possible fault
candidates from the measurement node of the
network.
(e) Calculate the fault distance from the other end of the
network as in step (c).
(f) Use the depth first search to isolate all possible
candidates from the other end of the network.
(g) Due to the unsymmetrical structure of the network,
there will be only one common fault point among
the list of fault candidates resulting from the two
searches in step (d) and (f). This common candidate
will be the actual faulty link.

Is Sum > x In order to further illustrate the above listed fault location
No methodology, a fault is introduced between node 8 and node 9
in the sample network of Fig. 4. The travelling wave recorders
Yes
at the two ends of the network record the fault transients.
Record Last link added as a fault These transients are analyzed to calculate the time stamps at
candidate the two ends, which are further used to calculate the fault
distance. The fault distance from node 1 is 43 units and from
node 12 is 29 units.
Back track to previous nodes
until adjacent node found
Although the fault distance from node 1 is calculated but
due to the radial structure of the network, there are more than
one fault candidates (links) at the calculated distance. In order
Sum = Sum – Lb to search the fault candidates we take advantage of the
(Lb is length traversed back) unsymmetrical tree structure of the network and use depth first
search. The process has been illustrated in Fig. 7, where the
numbers within circles are node identities and the numbers
outside give the search order. There are four fault candidates,
No
All adjacent nodes namely on link (3, 5), (6, 7), (8, 9) and (8, 10), searched from
traversed node 1.

Yes
Output fault
candidates

Fig. 6 Flowchart to Search Fault Candidates

Fault location in a radial network can be broadly classified

into two parts:

Calculating the fault distance which is done using
travelling wave theory. Fig. 7 Search for Fault Candidates from Node 1

Identifying the faulty link which is done using
network theory. In a similar manner, as shown in Fig. 8, we calculate the
distance of the fault point from other end of the network and
search for possible fault candidates. Three fault candidates are

1054
 Table II (b). Identification of Faulty Link (Case II)
Case II: A fault is introduced at 12390 feet from TWR2 on the
line connecting node 858 and node 864 i.e. link (858, 864)
Fault distance calculated Fault distance calculated
using single-ended travelling using single-ended travelling
wave measurement from wave measurement from
TWR1 X1 = 178589 feet TWR2 X2 = 12383 feet
Fault candidates identified Fault candidates identified
using depth first search at a using depth first search at a
distance of X1 from TWR1 distance of X2 from TWR2
are on links (858, 834), (858, are on links (858, 832) and
Fig. 8 Search for Fault Candidates from Node 12 864) and (888, 890) (858, 864)
Since, the fault point on line connecting node 858 and node
found from node 12, namely on link (10, 11), (8, 9) and (8, 6). 864 is common among the fault candidates searched from
We can clearly see that fault on line (8, 9) is the common TWR1 and TWR2 respectively, so the output faulty link is
candidate in both the searches and hence the actual fault point. (858, 864).

IV. CASE STUDY

This section describes the application of the methodology
explained in the previous section. The proposed method has
been applied on IEEE 34 node test feeder [23] which is
modeled and simulated in ATP/EMTP and calculations are
done using Matlab.
Fig. 9 shows the one line diagram of the IEEE 34 node test
feeder. The two TWRs as shown in the figure, record the fault
transient waves reaching the nodes, which are further
processed using wavelet transform to calculate the wave
arrival time at each end.
The wave arrival time is used to calculate the fault distance
from the two ends respectively. Due to the radial structure of
the network, there are more than one fault candidates at the
calculated distance. This ambiguity is resolved by using the
network topology and doing a search from the two ends as
described in the previous section.
In three different cases, faults are introduced in different
parts of the network and the proposed methodology is applied
to calculate the fault distance and identify the faulty link. The
results have been tabulated in Table II. X1 is the fault distance
calculated from TWR 1 and X2 is the fault distance calculated
form TWR 2.
Table II (a). Identification of Faulty Link (Case I)
Case I: A fault is introduced at 116500 feet from TWR1 on
the line connecting node 828 and node 830 i.e. link (828, 830)
Fault distance calculated Fault distance calculated
using single-ended travelling using single-ended travelling
wave measurement from wave measurement from
TWR1 X1 = 116372 feet TWR2 X2 = 72677 feet
Fault candidates identified Fault candidates identified
using depth first search at a using depth first search at a
distance of X1 from TWR1 distance of X2 from TWR2
are on links (828, 830), (824, are on links (830, 828) and
826) and (818, 820) (854, 856)
Since, the fault point on line connecting node 828 and node
830 is common among the fault candidates searched from
TWR1 and TWR2 respectively, so the output faulty link is
(828, 830).
Table II (c). Identification of Faulty Link (Case III)
Case III: A fault is introduced at 23290 feet from TWR2 on
the line connecting node 888 and node 890 i.e. link (888, 890)
Fault distance calculated Fault distance calculated
using single-ended travelling using single-ended travelling
wave measurement from wave measurement from
TWR1 X1 = 179670 feet TWR2 X2 = 23276 feet
Fault candidates identified Fault candidates identified
using depth first search at a using depth first search at a
distance of X1 from TWR1 distance of X2 from TWR2
are on links (858, 834) and are on links (852, 854) and
(888, 890) (888, 890)
Since, the fault point on line connecting node 888 and node
890 is common among the fault candidates searched from
TWR1 and TWR2 respectively, so the output faulty link is
(888, 890).
The distance measurements are in feet and the calculated
fault distance has been rounded to nearest integer. Although,
the proposed methodology can identify the faulty link
accurately but the errors are higher in single-ended travelling
wave measurement scheme as compared to double-ended
measurements. So, if the fault is very close to the node or on a
link with very small length then the output result might be
erroneous.
V. CONCLUSION AND FUTURE SCOPE
This paper has utilized some of the existing principles from
network theory to extend its application to locating faults in
radial distribution systems. Single-ended measurements are
done individually from the two ends of the network and fault
distances are calculated using travelling wave theory. Further,
network theory is used to isolate the actual faulty link from a
number of possible fault candidates at the calculated distance.
The proposed methodology has been demonstrated and
validated on the IEEE 34 node test feeder. The modeling,
simulations and calculations are done using Matlab and
ATP/EMTP.

1055
 Although, the proposed methodology identifies the faulty
candidate (link) accurately but single-ended travelling wave
scheme is less accurate compared to double-ended
measurements. Furthermore, it only takes into account one
fault at a time. However, this paper gives a new direction to
the use of network theory in power system research. Future
work will include optimizing the current methodology and
validating it on a real distribution networks.
Fig. 9 IEEE 34 Node Test Feeder
VII. REFERENCES
[1] M.M. Saha, J.J. Izykowski, E. Rosolowski, “Fault location – Basic
concept and characteristic of methods,” Fault Location on Power
Networks, Springer London, pp. 1-26, 2010.
[2] M.M. Saha, J.J. Izykowski, E. Rosolowski, “Fault location in
distribution networks,” Fault Location on Power Networks, Springer
London, pp. 333-360, 2010.
[3] M. Mirzaei, M.Z. A Ab Kadir, E. Moazami, H. Hizam, “Review of
fault location methods for distribution power network,” Australian
Journal of Basic and Applied Sciences, Vol. 3, No. 3, pp. 2670-2676,
2009.
[4] M.M. Saha, R. Das, P. Verho, D. Novosel, “Review of fault location
techniques for distribution systems,” Proceedings of Power Systems
and Communications Infrastructures for the Future, Beijing,
September 2002.
[5] J. Mora-Florez, J. Melendez, G. Carrillo-Caicedo, “Comparison of
impedance based fault location methods for power distribution
systems,” Electric Power Systems Research, Vol. 78, No. 4, pp. 657-
666, April 2008.
[6] F. Han, X. Yu, M. Al-Dabbagh, Y. Wang, “Locating phase-to-ground
short-circuit faults on radial distribution Lines,” IEEE Transactions on
Industrial Electronics, Vol. 54, No. 3, pp. 1581-1590, June 2007.
[7] Z. M. Radojevic, C.H. Kim, M. Popov, G. Preston, V. Terzija, “New
approach for fault location on transmission lines not requiring line
parameters,” Proceedings of International Conference on Power
Systems Transients, Kyoto, Japan, June 2009.
[8] M.B. Djuric, Z.M. Radojevic, V.V Terzija, “Distance protection and
fault location utilizing only phase current phasors,” IEEE Transactions
on Power Delivery, Vol. 13, No. 4, pp. 1020-1026, 1998.
[9] Q. Jian, C. Xiangxun, Z. Jianchao, “Travelling wave fault location of
transmission line using wavelet transform,” Proceedings of
International Conference on Power System Technology
(POWERCON), Vol. 1, pp. 533-537, 1998.
[10] A. Dwivedi, X, Yu, P. Sokolowski, P. Wong, F. Buratto, “Fault location
in power networks using graph theory,” Proceeding of 36th IEEE
Industrial Electronics Conference (IECON), Phoenix, Arizona, USA,
November 2010.
[11] A. Montaes, M. P. Comech, M. Gracia-Gracia, “Travelling wave
distance protection in a distribution network: A comparative study,”
1056
VI. ACKNOWLEDGEMENT
The authors would like to thank two power utilities
Powercor/Citipower and Jemena (previously named Alinta) in
Australia for their valuable feedback and support for this
project. This project has been funded by the Australian
Research Council (ARC) linkage scheme (project number
LP0562568).
Proceedings of 14th IEEE Electrotechnical Conference, pp. 768-773,
Ajaccio, September 2008.
[12] N. E. Eng, K. Ramar, “Single-ended travelling wave based fault
location on two terminal transmission lines,” Proceedings of IEEE
Region 10 Conference on TENCON, pp 1-4, Singapore, January 2009.
[13] D. W. P. Thomas, R. J. O. Carvalho, E. T. Pereira, “Fault location in
distribution systems based on travelling waves,” Proceeding of IEEE
Power Tech Conference, Bologna, Italy, June 2003.
[14] A. Elhaffer, M. Lehtonen, “Travelling wave based earth fault location
in 400 kV transmission network using single end measurement,”
Proceeding of IEEE Large Engineering Systems Conference Power
Engineering, pp. 53-56, July 2004.
[15] T. Hisakadu, K. Tanaka, K. Okumura, “A transmission line fault-
location system using the wavelet transform,” Electrical Engineering in
Japan, Vol. 40, No. 4, 2002.
[16] A. Borghetti, S. Corsi, C.A Nucci, M Paolone, L. Peretto, R. Tinarelli,
“On the use of continuous-wavelet transform for fault location in
distribution power systems,” International Journal of Electrical Power
and Energy Systems, Vol. 28, No. 9, pp. 608-617, November 2006.
[17] T.I.A.H. Mustafa, D.W.P. Thomas, C. Christopoulos, A. Raizer, “
Comparison of simulated and recorded transients for travelling wave
fault location,” Proceedings of IEEE Power Tech Conference, Vol. 3,
Bologna, Italy, June 2003
[18] X. Yu, A. Dwivedi, P. Sokolowski, “On complex network approach for
fault detection in power grids,” Proceedings of IEEE International
Conference on Control and Automation,” Christchurch, New Zealand,
December 2009.
[19] M. E. J. Newman, “The structure and function of complex networks,”
SIAM Review, Vol. 45, pp. 167-256, 2003.
[20] V. K. Balakrishnan, Schaum’s Outline of Theory and Problems of
Graph Theory, McGraw-Hill, 1997.
[21] ATP/EMTP Rule Book, Canadian-American EMTP Users Group.
[22] L. Prikler, H.K. Hoidalen, ATPDRAW version 3.5 for Windows
9x/NT/200/XP, October 2002.
[23] IEEE 34 Node Test Feeder [Online, September 2010].
Available:
http://ewh.ieee.org/soc/pes/dsacom/testfeeders/index.html