Computers and Electrical Engineering: M. Meenakshi Devi, M. Geethanjali, A. Rama Devi

JID: CAEE
ARTICLE IN PRESS [m3Gsc;February 7, 2018;8:48]

Computers and Electrical Engineering 0 0 0 (2018) 1–16
Contents lists available at ScienceDirect
Computers and Electrical Engineering

journal homepage: www.elsevier.com/locate/compeleceng
Fault localization for transmission lines with optimal Phasor

Measurement UnitsR
M. Meenakshi Devi∗, M. Geethanjali, A. Rama Devi
Department of EEE, Thiagarajar College of Engineering, Tamilnadu, India
a r t i c l e i n f o a b s t r a c t
Article history: This paper discusses a novel fault localization algorithm for multi-terminal transmission
Received 1 April 2017 lines. The technique is tested for different fault types and fault locations. The exact fault
Revised 31 January 2018
location is determined along with the latitude and longitude by involving geo-referenced
Accepted 31 January 2018
data of the power system. In addition, Phasor Measurement Units (PMUs) are effectively
Available online xxx
used for complete system observability. The number of PMUs is optimised using Genetic
Keywords: Algorithm. Thus, the objective is to find the distance of the fault occurrence and to locate
Smart grid the fault point along with the latitude and longitude, using optimal PMUs. For evaluation
Wide area monitoring, protection and purpose, Indian Utility (TamilNadu Electricity Board) 49 bus system is used and validated,
control using software such as MATLAB, Power World Simulator and Google Earth View. The re-
Phasor Measurement Units sults show the effectiveness of the proposed technique to spot the exact fault point in the
Fault location transmission lines with a minimum number of PMUs, to make the concept economical.
Genetic algorithm © 2018 Elsevier Ltd. All rights reserved.
Transmission lines
1. Introduction
Developing a model, maintenance, control aspects of electric power systems and communication networks are becoming
a challenging aspect. It is still more complicated for power engineers, to maintain the system reliability and to increase
the effectiveness in power system protection [1]. This is due to the increase in demand of power, diffusion of renewable
energy sources, enlarging interconnections and deregulated energy markets. This has raised the need for synchrophasor
measurement technology in the field of Wide Area Monitoring, Protection and Control (WAMPAC). Synchrophasor is said to
be a phasor value measured at an instant and therefore it consists of time-tagged data [2].
The time-tagged data are to be measured simultaneously and synchronised, to enhance the Phasor Measurements
recorded at a particular instant, across a huge geographical area. The time intervals are the multiples of a period of the
nominal power system frequency. This synchronisation is provided by Global Positioning System (GPS) receiver. In this mea-
surement technology, basic parameters such as voltage, current, phase angles, and frequency can be measured by using
Phasor Measurement Units (PMU). Also, the data provided by the PMUs are accurate and useful in the determination of the
exact sequence of events, which improves the monitoring. Likewise, it increases precision and speed of state estimation [2].
Due to these distinguishing features, PMUs are adopted for huge power system applications [3]. In the proposed algorithm,
the PMUs are used in fault localization for multi-terminal transmission lines.
The concept of fault location has attracted many researches to work on it to avoid the consequences due to the fault
occurrence. Numerous works are carried out on fault location in one-terminal and two-terminal transmission lines [4–6].
R
Reviews processed and recommended for publication to the Editor-in-Chief by Guest Editor Dr. J. Yogapriya.
∗
Corresponding author.
E-mail addresses: mdeee@tce.edu (M.M. Devi), mgeee@tce.edu (M. Geethanjali).
https://doi.org/10.1016/j.compeleceng.2018.01.043
0045-7906/© 2018 Elsevier Ltd. All rights reserved.
Please cite this article as: M.M. Devi et al., Fault localization for transmission lines with optimal Phasor Measurement
Units, Computers and Electrical Engineering (2018), https://doi.org/10.1016/j.compeleceng.2018.01.043
JID: CAEE
2 M.M. Devi et al. / Computers and Electrical Engineering 000 (2018) 1–16
Though the fault location algorithms using single and two terminals are uncomplicated and economical, it has some flaws
regarding accuracy and selection of the faulted section. In addition, real-time power system has numerous nodes making the
system complex [4]. To face this complexity and improvise the demerits of the older techniques multi-terminal fault location
algorithms are used. Here fault location for multi-terminal transmission lines is chosen because it does not completely rely
on the values of un-transposed lines and line-loading, and so it improves the accuracy too. Hence, instead of modifying
the lower level algorithms, multi-terminal fault location is taken into account. From the literature review, it is understood
that Smart Grids employing distributed and renewable energy resources are made secured by utilising the PMUs effectively.
There are several methods such as decision tree method [5], and least square estimation technique [6,7] which elaborate
the phasor estimation and classification of fault types. The methods based on synchronized sampling do not require line
parameters and settings free. The basic concepts could be understood using three-terminal transmission line [8,9]. A circuit
approach for three-terminal transmission lines is examined and learnt that it uses voltage and current phasor values [8]. In
this reference paper fault location for three-terminal transmission lines is done with time-time transformation to report the
arrival time of fault generated travelling waves in [9]. The paper [10] comes up with a novel method of fault location using
binary differential evolution optimization algorithm with ‘weight factor’ and ‘reduction factor’ that improves convergence
speed. Also, references [11,12] explain the fault location technique with optimal placement of PMUs. This reduces the cost of
installation and maintenance of PMUs. A thorough work on the fault detection and localization of smart power grids using
synchronous phasor angle measurements is achieved from the literatures [13,14]. References [15,16] are the inspirational
papers that deliver a clear derivation for fault location for multi-terminal transmission lines.
In real-time, the inspection of exact location of the fault is done manually and it is a tedious process which delays the
further rescue actions. In some places, even helicopters are used in search of fault points without knowing them exactly [17].
This accomplishes an expensive service of restoration. Additionally, the real-time power system includes various electrical
components and parameters within it and is located through a wide area in the landscape. In this aspect, with fault distance,
it will be hard for the experts to locate the exact fault point, particularly for long lines in uneven terrains [17]. In practical,
the assumption made is that the PMUs are placed at all the terminals [15,16]. But, in the economical point of view, the
installation cost of a PMU depends on numerous factors, such as the number of measuring channels, current transformer
connections, potential transformer connections, power connection, station ground connection and GPS antenna connection.
Here, according to the number of channels, the cost of PMU differs from one another [18]. The remaining connections are
similar to all the PMU installations. From the literature study [19], it is understood that the approximate cost of each PMU is
30,0 0 0–40,0 0 0 of US dollars. Thus, to make this an optimal technique, this paper proposes that the PMU counts optimised
through Genetic algorithm. Further, the references [15,16] focus on identifying the fault distance and fault point from the
sending end of the line. In order to deliver the fault point with latitude and longitude along with the fault type to the
control system operator, the fault location identification algorithm is proposed in this paper. To experiment on this aspect,
the goal of this paper is to capture the features of a real-time bus system in a simulation diagram, which could be used
to run fault analysis and locate the fault point with latitude and longitude. Regarding the quantities of voltage and current
phasors it is considered that they are completely observed and synchronously measured by the PMUs located at optimal
terminals. The simulation is created and the fault analysis is carried out for the considered test system. The fault region
is contoured and the exploration of the latitude and longitude of the fault point is done using real-time Indian utility -
TamilNadu Electricity Board (TNEB) 49 bus system. Finding the fault point with geo-referenced data will definitely replace
the operating time to restore the damaged section. To make this a winning concept, the proposed algorithm is developed in
MATLAB for identifying the fault location and the spotting of latitude and longitude of the fault point is extended in Power
World Simulator (PWS) along with Google Earth View. Geographic Information System (GIS) tools are the computer-based
tools that deliver content with the geographic location and satellite imagery [20].
The paper is organized as follows. Section 2 depicts the concepts involved in optimal PMU placement technique.
Section 3 describes the proposed algorithm for fault identification to establish the faulted section in the power system
which is interpreted for both bus and line fault. Section 4 elaborates the results and discussion obtained from the optimal
placement problem and the proposed method for the fault location. It is followed by the conclusion in Section 5.
2. Optimal PMU placement technique
In the field of Power system protection, the PMUs are known for their data acquisition technology for wide area monitor-
ing and protection. This regulation of PMUs with the wide area monitoring systems is done by data synchronisation using
Global Positioning Systems (GPS) receivers for 1 microsecond. In particular, the PMU is a device that is capable of mea-
suring the entire parameters of a bus in which the PMU is placed and those of the adjacent buses connected to that bus
[21]. Hence, the placement of PMUs at strategic buses makes the system completely observable. The target is to minimise
the number of PMUs with complete observability to observe the fault locations. For this purpose, Genetic algorithm (GA) is
considered for the optimal placement problem. GA is a popular one in meta-heuristic methods. In the field of Engineering
and Technology, for a particular scenario, GA is used to explore a prominent result.
GA is an iterative process and at first, it starts with the randomly generated chromosomes. These chromosomes contain
the information of the PMUs placed. This binary set has a length equal to the number of buses in the system. For example,
if a bit denotes a value of ‘1 , the PMU is placed in that location or bus. A bit having a value of ‘0 implies the absence
of PMU at that particular bus. The chromosome information is encoded by GA, to solve the optimization problem and
JID: CAEE
M.M. Devi et al. / Computers and Electrical Engineering 000 (2018) 1–16 3
Fig. 1. Fault on busi .
comes up with numerous solutions. The optimal placement technique involves the count and positioning of the PMU for
the entire system observability on the power network. Thus the constraint of the problem is complete observability towards
the objective of reducing the count of PMUs. Accordingly, the fitness function depends on two parameters, (i) the number
of PMUs positioned in a bus system and (ii) the number of buses unobserved.
For N-bus power system, the objective function of the optimal placement technique is given as follows,
Fitness function,
min f = aNPMU + b(Nh + NPMU ) (1)
For this Optimal PMU Placement (OPP) technique, the condition satisfies at min (f)

NPMU < N
s.t.
Nh = 0
where, NPMU = Number of PMUs in the system, Nh = Number of buses unobserved, and a and b are the weighted factors
considered. The problem formulation of the GA explains that the number of PMUs should be less than the total number of
buses. Also, all buses should be observable, denoted as, Nh = 0.
The operators included in GA are as follows. Selection operator is to elect the suitable individuals from the ongoing set
of populations which is resolved by the fitness function. The method uses a proportionate scheme called roulette wheel
selection. By substituting the values of fitness function, the estimation is calculated to enumerate the fitness value of each
selected individual. Reproduction process is defined as generating heir from the parent. This is performed by crossover and
mutation [21]. A crossover operator is the blending of parent chromosomes which can be applied for reproducing their
offsprings. Mutation is a bit change process done to the individuals that are formed after the crossover. The Crossover
and mutation depends on the crossover and mutation probability constants. Here, binary crossover and binary mutation
are handled. The recently produced offspring is placed for the population. The algorithm is executed until it satisfies the
stopping criteria and it turns up with a better solution for the optimal placement. The resulting values are the optimal
places where the PMUs should be placed for complete observability.
3. Proposed algorithm for fault identification
The multi-terminal transmission line is considered with 1,2,…,i,.j....n terminals. By using the optimal placement of PMUs
at strategic locations, the synchronized voltages and current phasors at all buses can be made available. These data are
implemented for an application such as, fault identification process. Now, the PMUs are ready to record the values of volt-
ages and currents at both steady state and fault condition. At the steady state condition, the power system network with n
terminals can be represented with the node admittance matrix [22],
(0 ) (0 )
[YnXn ] Un = In (2)
where, [YnXn ] = [GnXn ] + j[BnXn ] is the node admittance matrix. [Un(0 ) ] is the pre-fault node voltage matrix and [In(0 ) ] is the
pre-fault node injection current matrix.
At the instant of fault occurrence, the line currents between the nodes of the faulted line are calculated based on the
fault node voltages that are measured by the PMUs. From the line currents, node injection currents at two terminals of the
faulted line are formed. The values of the node injection current are effective in the deduction of fault node. As a result,
the calculation of the node injection currents is essential for improving the accuracy of the fault localization process. The
computation of node injection currents and the interpretation of the faulted section are explained in the following sections.
3.1. Fault on a bus
Let us consider fault at bus-i, as shown in the Fig. 1.

Then node-admittance matrix in (2) is updated as [23],
(0 )
Un( ) + [Un ] = + [In ]
0
[YnXn ] In (3)
JID: CAEE
Fig. 2. Fault on linei − j .
(f) (f)
where, [Un ] = [Un ] − [Un(0 ) ] are fault components of node voltages, [In ] = [In ] − [In(0 ) ] are fault components of node
(f) (f)
injection currents, and [Un ] is node voltage matrix at post fault conditions. [In ] is node injection current matrix at post
fault conditions. If is the fault current at fault point.
Hence, the node equation becomes, [YnXn ] [Un ] = [In ] by solving (3) in (2). The fault node injection current turns into,
⎡ ⎤ ⎡ ⎤
0 0 1
⎢ .. ⎥ ⎢ .. ⎥ ..
⎢ . ⎥ ⎢.⎥ .
⎢ ⎥ ⎢ ⎥
[In ] = ⎢Ii ⎥ = ⎢I f ⎥ i (4)
⎢ . ⎥ ⎢.⎥ ..
⎣ .. ⎦ ⎣ .. ⎦ .
0 0 n
Thus, it is confessed that if [In ] has single non-zero element, the fault is said to be occurred at busi . The location of
index i in matrix [In ] depicts the bus number.
3.2. Fault on a line

Let us consider that fault occurs at linei − j between busi and busj as shown in the Fig. 2.
During the pre-fault condition, the node admittance matrix is [23],
⎡ ⎤⎡U (0) ⎤ ⎡ (0) ⎤
Y11 ··· Y1i ··· Y1 j ··· Y1n 0 1 I1
⎢.. .. .. .. .. .. .. ⎥⎢ .. ⎥ ⎢ . ⎥
⎢. . . . . . . 0 ⎥⎢ ⎥
⎢ (.0) ⎥ ⎢
.. ⎥
⎢Y ··· Yii ··· Yij ··· ⎥ ⎢
Yi f ⎥⎢Ui ⎥ ⎢I 0 ) ⎥ ( ⎥
⎢ i1 Yin
⎢ ⎥⎢ ⎥ ⎢i ⎥
(0) ⎢... .. .. .. .. .. .. ⎥⎢ ... ⎥ ⎥ ⎢ .. ⎥
Y(n+1)X (n+1) Un+1 = ⎢ . . . . . . 0 ⎥⎢ ⎢ ⎥ =⎢ . ⎥ (5)
⎢Y ··· Y ji ··· Y jj ··· Y j f ⎥⎢U j( ) ⎥ ⎢I (0 ) ⎥
⎥ 0 ⎢
⎢ j1 Y jn ⎥
⎢. .. .. .. ⎥⎢ ⎥
⎢ .. ⎥ ⎢
j
⎥
⎢.. .. .. .. ⎥
0 ⎥⎢ . ⎥ ⎢ .. ⎥
⎢ . ⎥
⎢ . . . . . .
⎢ (0 ) ⎥ ⎣ ⎦
⎣Yn1 ··· Yni ··· Yn j ··· Ynn Y ⎦ ⎣U ⎦ (0 )
ff n In
0 0 Y fi 0 Y f j 0 0 0 U f( )
0
0
where, Yii and Y j j are the self admittance of busi and busj respectively, Y f f is the self admittance of busn + 1 , Yij = Y ji are the
mutual admittance between busi and busj , Yif = Y f i are the mutual admittance between busi and busn + 1 and Y j f = Y f j are
the mutual admittance between busj and busn + 1 . From (5), the following equations are obtained,
The expansion of ith row of Eq. (5) denotes,
Yi1U1( ) + . . . + Y iiUi( ) + . . . + Y i jU j( ) + . . . + YinUn( ) + Y i f U f( ) = Ii( )

0 0 0 0 0 0
(6)
The expansion of jth row of Eq. (5) means,
Y j1U1( ) + . . . + Y ji Ui( ) + . . . + Y jjU j( ) + . . . + Y jnUn( ) + Y j f U f( ) = I (j )

0 0 0 0 0 0
(7)
The expansion of n+1th row of Eq. (5) results in,
Y f iUi( ) + Y f jU j( ) + Y f f U f( ) = 0
0 0 0
(8)
JID: CAEE
From (8), voltage at fault point
−Y f iUi( ) − Y f jU j( )
0 0
U f( ) =
0
is obtained. (9)
Y f f
By solving (9) in (6),

(0 )
Y i f Y f i (0 )
Y i f Y f j
Y ii − Y i j − U j( ) + . . . + YinUn( ) = Ii( ) is derived.
0 0 0
Yi1U1 + ...+ Ui + ...+ (10)
Y f f Y f f
Similarly, by solving (9) in (7),

(0 )
Y j f Y f i (0 )
Y j f Y f j
Y ji − Y j j − U j( ) + . . . + Y jnUn( ) = I (j ) is obtained.
0 0 0
Y j1U1 + ...+ Ui + ...+ (11)
Y f f Y f f
Here, self admittances of node i, node j and node f are denoted by,
1 1 1 1 1
Y ii = Yii − + , Y i j = Y ji = Yi j + , Y j j = Y j j − + (12)
zi j xzi j zi j zi j ( 1 − x )z i j
1 1 1 1
Y f f = + , Y = Y f i = − , Y j f = Y f j = − (13)
xzi j (1 − x )zi j i f xzi j (1 − x )zi j
By solving (12) and (13) in (10) and (11), the expression attained under the pre-fault condition is,
Yi1U1( ) + . . . + YiiUi( ) + . . . + Yi jU j( ) + . . . + YnnUn( ) = Ii( )

0 0 0 0 0
Y j1U1( ) + . . . + Y jiUi( ) + . . . + Y j jU j( ) + . . . + YnnUn( ) = I (j )

0 0 0 0 0
Under the fault condition, the fault point is regarded as a new bus and located at n + 1 location. Here, the node admit-
tance matrix becomes,
⎡ ⎤⎡ ⎤ ⎡ ⎤
Y11 ··· Y1i ··· Y1 j ··· Y1n 0 U1 0
⎢.. .. .. .. .. .. .. ⎥⎢ .. ⎥ ⎢ .. ⎥
⎢. . . . . . . 0 ⎥⎢ . ⎥ ⎢ . ⎥
⎢Y Yif ⎥⎢
⎥
⎢ i1 ··· Yii ··· Yij ··· Yin Ui ⎥ ⎢ ⎥
⎢. ⎥⎢ ⎥ ⎢0⎥
⎢. .. .. .. .. .. .. ⎥ .. ⎥ ⎢
⎢ ⎥ .. ⎥
Y(n+1)X (n+1) [Un+1 ] = ⎢ . . . . . . . 0 ⎥⎢ ⎢ . ⎥=⎢ ⎢ .⎥⎥ (14)
⎢Y ··· Y ji ··· Y jj ···
⎥
Y j f ⎥⎢ U j ⎥ ⎢ 0 ⎥
⎢ j1 Y jn
⎢. ⎥⎢ . ⎥ ⎢ . ⎥
⎢ ⎥ ⎢
⎢. .. .. .. .. .. .. ⎥ .. ⎥ ⎢ .. ⎥
⎢. . . . . . . 0 ⎥⎢ ⎣ ⎦ ⎣ ⎦
⎥
⎣Yn1 ··· Yni ··· Yn j ··· Ynn 0 ⎦ Un 0
0 0 Y fi 0 Y f j 0 0 Y f f U f If
From the matrix, fth line represents Y fi Ui + Y fj Uj + Y ff Uf = If = If .
This could be modified as
I f − Y f i Ui − Y f j U j
U f = (15)
Y f f
while solving (15) in ith line of node admittance matrix, it results in
Yi1 U1 + . . . + Y ii Ui + . . . + Y i j U j + . . . + Yin Un + Y i f U f = 0

1/xzi j
Yi1 U1 + . . . + Y ii Ui + . . . + Y i j U j + . . . + Yin Un − If = 0 (16)
(1/xzi j )+1/(1−x)zi j
Yi1 U1 + . . . + Yii Ui + . . . + Yi j U j + . . . + Yin Un = (1 − x )I f
Similarly jth line results in,
Y j1 U1 + . . . + Y ji Ui + . . . + Y j j U j + . . . + Y jn Un = xI f (17)
JID: CAEE
From this,
⎡ ⎤ ⎡ ⎤1
0 0
.
⎢ . ⎥ ⎢
. .
.. ⎥ ..
⎢ ⎥ ⎢
. ⎥
⎢ I ⎥ ⎢ ( 1 − x ) I ⎥ i
⎢ i⎥ ⎢ f⎥
⎢ . ⎥ ⎢ .. ⎥ ..
⎢ . ⎥ ⎢ ⎥.
[YnXn ] [Un ] = [In ] = ⎢ . ⎥ = ⎢ . ⎥ is derived. (18)
⎢I j ⎥ ⎢ xI f ⎥ j
⎢ ⎥ ⎢ ⎥.
⎢ .. ⎥ ⎢ .. ⎥ ..
⎢ . ⎥ ⎢ . ⎥
⎣ ⎦ ⎣ ⎦.
0 0 ..
0 0 n
In the aforementioned equation, Ii and Ij are the fault injection currents at busi and busj respectively. These fault
injection currents are proportional to the fault current If .
The variable x may be defined as the percentage of the fault distance from busi to the fault point at busn + 1 of the line
length.
The variable is denoted by
I j
x= . (19)
Ii + I j
Fault location is found by an unknown variable x that varies from 0 to 1. Fault current at fault point is the sum of node
injection currents of both ith and jth node.
I f = Ii + I j (20)
The fault current at fault point will be placed at n + 1 location [23].

To examine the fault type, the zero, positive and negative sequence values are noted and solved as follows,
⎡ ⎤ ⎡ ⎤1
0 0
.
⎢ .. ⎥ ⎢ .. ⎥ ..
⎢ . ⎥ ⎢ . ⎥
⎢ I ⎥ ⎢(1 − x )I ⎥ i
⎢ i (0 ) ⎥ ⎢ f (0 ) ⎥
⎢⎢ ... ⎥ ⎢
⎥ ⎢ .. ⎥ ..
⎥.
Y nXn(0 ) Un(0 ) = ⎢ ⎥ ⎢
= . ⎥ (21)
⎢I j (0) ⎥ ⎢ xI f (0) ⎥ j
⎢ ⎥ ⎢ ⎥.
⎢ .. ⎥ ⎢ .. ⎥ ..
⎢ . ⎥ ⎢ . ⎥
⎣ ⎦ ⎣ ⎦.
0 0 ..
0 0 n
⎡ ⎤ ⎡ ⎤1
0 0
.
⎢ .. ⎥ ⎢ .. ⎥ ..
⎢ . ⎥ ⎢ . ⎥
⎢ I ⎥ ⎢(1 − x )I ⎥ i
⎢ i (1 ) ⎥ ⎢ f (1 ) ⎥
⎢⎢ .. ⎥ ⎢
⎥ ⎢ .. ⎥ ..
⎥.
Y nXn(1 ) Un(1 ) = ⎢ . ⎥ = ⎢ . ⎥ (22)
⎢I j (1) ⎥ ⎢ xI f (1) ⎥ j
⎢ ⎥ ⎢ ⎥.
⎢ .. ⎥ ⎢ .. ⎥ ..
⎢ . ⎥ ⎢ . ⎥
⎣ ⎦ ⎣ ⎦.
0 0 ..
0 0 n
JID: CAEE
⎡ ⎤ ⎡ ⎤1
0 0
.
⎢ .. ⎥ ⎢ .. ⎥ ..
⎢ . ⎥ ⎢ . ⎥
⎢ I ⎥ ⎢(1 − x )I ⎥ i
⎢ i (2 ) ⎥ ⎢ f (2 ) ⎥
⎢⎢ ... ⎥ ⎢
⎥ ⎢ .. ⎥ ..
⎥.
Y nXn(2 ) Un(2 ) = ⎢ ⎥=⎢ . ⎥ (23)
⎢I j (2) ⎥ ⎢ xI f (2) ⎥ j
⎢ ⎥ ⎢ ⎥.
⎢ .. ⎥ ⎢ .. ⎥ ..
⎢ . ⎥ ⎢ . ⎥
⎣ ⎦ ⎣ ⎦.
0 0 ..
0 0 n
where, subscript 0, 1, 2 denotes the zero, positive and negative sequence components transformed from three-phase system.
Equation with zero sequence components i.e. (21) is used for all earth faults such as line to ground (LG) and Double line
to ground (LLG). Similarly, rare three phase faults and all symmetrical faults can be found using (22) and (23) respectively.
Thus, the derivation for the fault condition has been obtained.
The fault identification error is said to be the difference between actual and estimated location and is expressed as
follows,
|est imated locat ion − actual location|
Fault location error = ∗ 100 (24)
total l ine l ength
The fault location error brings out the difference between the estimated location and actual location.
3.2.1. Interpretation of the fault section. Based on the zero and non-zero elements, there are three different cases present in
the node fault injection current [In ]. This node fault injection current [In ] has dissimilar features for each and every case.
This notable element is used to resolve the faulted section.
Case 1) If all elements are zero in [In ] the condition is denoted as [In ] = [0, 0, …0, …, 0]T . From this, it is inferred
that the system has no fault.
Case 2) If [In ] has one non-zero element the condition is denoted as [In ] = [0, 0, …Ii ,…, 0]T . From this, it is under-
stood that the fault has occurred on a bus. The node i that has the non-zero element is the faulted bus.
Case 3) If [In ] have two non-zero elements the condition is denoted as [In ] = [0, 0, …Ii ,…, 0, …, Ij , 0]T . From this
it is concluded that the fault has occurred on the line. The node i and j that have the non-zero element is the faulted line
section.
3.3. Effect of fault resistance

The conventional techniques handled for accurate fault location on transmission lines have shortcomings as they are
influenced by the fault resistance. This demerit has been overcome in the proposed method. The effect of fault resistance
on fault identification algorithm can be explained with the expression given in the following equation,
Uf = If ∗ Rf (25)
where, Uf is the voltage of fault point. If and Rf are the fault current and fault resistance values respectively.
From this relation, it is evident that the fault current value is indirectly proportional to Rf . Thus, the fault resistance value
will resist the fault current value that occurs at the fault point [23]. Consequently, it forces the fault voltage to a lower value.
Therefore, this will affect the fault identification process. Hence, it is necessary to devise a method in which the value of
the Rf should not have an impact on If and Uf values. In other words, to retain the fault injection current value, the fault
identification algorithm should not be affected by the fault resistance values. Hence, the proposed technique is based on
node fault injection current alone.
3.4. Geographic data view

The entire power system is drawn as a geographic representation and the fault is located in the Google Earth View by
spotting its latitude and longitude. The variable x gives the percentage of the fault distance from busi . This data is taken to
the simulation and the fault is localised using PWS and Google Earth View. The steps involved in this process are clearly
given in the Fig. 3 and explained below.
(1) Steps involved

(a) Creation of one line: The one-line diagram is a pictorial representation of a power system [24]. The PWS has the
facility to draw the one-line of a power system easily with all the basic details. PWS has two modes such as, edit
mode and run mode. The one-line can be drawn in edit mode and the parameters of the system can be fed within
it. The one-line can be run in the run mode and the necessary power flow can be obtained with the tools in built.
JID: CAEE
Fig. 3. Flowchart explaining the steps involved in proposed methodology.
(b) Fault analysis and Contouring: Fault analysis dialog is used to execute fault analysis for the currently loaded power
system. Fault analysis dialog includes fault location, fault distance, fault type, and fault current. The fault analysis
dialog makes the user to define the problem clearly and the results are obtained for various operating conditions
[24]. Contouring provides the display with an animated contour map for the system quantities including voltage
magnitudes, angle, real and reactive loads, MW transactions and transmission loading.
(c) Geographic data view: The geographic data view portrays the geographical representations of buses, generators, loads,
transmission lines, substations, areas, zones, super areas and injection groups. The latitude and longitude coordinates
should be compulsorily given for using this option. Object attributes make the visibility obvious. The simulator admits
the conversion of one-line to ESRI shapefile (∗ .shp/∗ .dbf pairs) and vice-versa. The one-line can also be imported from
KML (Keyhole Markup Language) and exported to KML file format. Also KML file format is supported by geographical
software such as Google earth.
(d) Google Earth View: Google Earth exhibits satellite images including 3D representation of the Earth, GIS data, and
aerial photography features. Google Earth View relinquishes about location based information with fast access and
high resolution [20]. The exported one-line of the power system from PWS can be imported from Google Earth View.
Hence, the places of the buses included in the power system can be placed on the geographic map. The graphic
appearance depicts colourful images with the locations of the buses of a power system which makes the readers
understand easily.
JID: CAEE
Table 1
Results for optimal PMU placement technique.
Bus system Number of PMUs required Location of PMUs (Bus numbers)
TNEB 49 bus system 9 1, 4, 14, 32, 34, 37, 41, 42, 46.
Table 2
Comparison Between proposed and reference papers for number of PMUs.
Total number of PMUs for Number of PMUs required Installation cost estimation
N bus power system for TNEB 49 bus system (K US Dollars)
Ref [22] = N 49 1470

Ref [23] = N 49 1470
Proposed fault Localization algorithm ≤ N3 9 270
4. Results and discussion
To judge the fault identification and localization methodology for multi-terminal transmission lines, the Indian utility
(TNEB) 49 bus system has been considered and the one -line diagram is shown in the Fig. 4.
PMUs are placed in optimal locations using GA to make the network completely observable. Hence, the optimal PMUs
monitor the basic parameters of the buses in pre-fault and post-fault conditions. The proposed technique is evaluated with
Pentium® Dual Core CPU @ 2.00 GHz, 2GB RAM system configuration, using MATLAB and the localization of the fault with
latitude and longitude is carried out in PWS version 16.
4.1. Optimal PMU placement technique using GA
The PMUs are installed in the buses to monitor the system. Although the PMU installation maximises the fault location
performance in wide area, it also leads to the increase in the cost of installation and maintenance. Hence, the aim of this
work is to optimise the PMU count as well as the cost. To accomplish this aspect, GA is used here to optimise the PMU
locations by solving the fitness function. The advantages of GA are explained by the authors in their previous work [25].
Significant results are acquired for weighted factor values of a = 1 and b = 2. GA is tested on Indian utility (TNEB) 49 bus
system and the results are obtained as shown in Table 1.
Generally, the optimal placement of PMUs for achieving complete observability requires only N3 PMUs for N bus power
systems [25]. This satisfies the complete observability of a power grid by placing optimal PMUs with 25% of the total number
of buses. The result obtained reveals that it is optimal to place the PMUs at only 9 locations, instead of placing the PMUs
at all the 49 buses. Thus, the condition NPMU ≤ N3 is satisfied. The installation cost is also reduced accordingly. Thus, the
proposed approach is significant in placing the PMUs at lower number of buses and the faulted section will be diagnosed
in a better way. The comparative study with other literatures is given in the Table 2. In the literatures [22,23], it is assumed
that the PMUs are placed at all the buses. In the proposed algorithm, it is convinced that PMUs at 9 locations are sufficient to
attain the complete observability. This reduces the installation cost of the PMUs as the numbers are reduced considerably.
Since, the PMUs are approximately 30,0 0 0 – 40,0 0 0 of US dollars. For example, let us consider the cost of one PMU is
around 30,0 0 0 US dollars, the optimal placement of PMUs will reduce the installation cost drastically. This Installation cost
estimation is compared with the published literatures [22,23] and furnished in Table 2. Thus, according to this approach, for
the TNEB practical bus system taken for this study, the total cost of PMUs is remarkably reduced. In particular, nearly 75%
of the cost reduction is achieved.
4.2. Fault identification algorithm
To evaluate the fault identification algorithm, the coding based on the algorithm is developed and the data of the Indian
utility (TNEB) 49 bus test system is fed. The algorithm is run for two fault locations such as, bus fault and line fault. The
proposed technique is applicable for all the transmission lines in the power system with different fault types and fault
distances. For simplification and understanding purpose, detailed results for the line 2–3 (Ennore Thermal Power Station
(ETPS)-MANALI)) are given. The valuation of the method depends on the number of non-zero elements of index InX1 as
described in the Section 1. Figs. 5 and 6 show the calculated index of fault node injection current. Since Fig. 5 denotes that
InX1 has single non-zero element at bus 2 located at Ennore Thermal Power Station (ETPS) it is confessed that the fault has
occurred at bus 2 (ETPS) with the amount of fault current value plotted. Fig. 6 explains that the fault has occurred at line
2–3 (ETPS-MANALI) along with the current at faulted point at the n + 1 terminal. In addition, the line 2–3 (ETPS-MANALI)
spans for 7.13 km. The distance of the fault point is varied and the estimated location and the error are given in Table 3 for
three phase balanced fault. Likewise, the algorithm is executed for unsymmetrical faults also. For unsymmetrical faults, the
fault resistance values are considered to show that the algorithm is not affected by the different fault resistance values. Any
suitable resistance value shall be assumed for fault resistance.
JID: CAEE
Fig. 4. One-line diagram of the Indian utility (TNEB) 49 bus system.
JID: CAEE
Fig. 5. Node injection current value for Bus fault occurred at bus 2 (ETPS).
Fig. 6. Node injection current value for Line fault occurred at line 2–3 (ETPS-MANALI).
Table 3
Results for symmetrical fault location estimation.
Fault type Actual location (% of Estimated location (% of Latitude and longitude details Error (%)
the line) the line) of the estimated location

3 phase balanced fault 20 21.02 130 11 43.59 N 0.14

800 18 12.20 E

40 41.31 130 11 10.33 N 0.18

800 17 35.75 E

60 60.79 130 10 43.40 N 0.11

800 17 01.11 E

80 81.41 130 10 18.57 N 0.19

800 16 22.20 E
In the literature, for illustration purpose, 10 and 50 fault resistance values are considered [23]. From Table 4, it is
inferred that the fault identification results acquired with 10 fault resistance are eventually closer to the results obtained
for 50 fault resistance. In other words, the algorithm is not influenced for higher and lower values of fault resistance.
Based on this, it is concurred that the effectiveness of the proposed algorithm will persist insensitive for different fault
resistance values.
The varied distances and the fault resistances for fault types namely single line to ground (SLG), line to line (LL) and
double line to ground (LLG) at line 2–3 (ETPS-MANALI) are furnished in Table 4. The error between the actual and estimated
location is noted to be pretty low.
The simulation of the Indian utility (TNEB) 49 bus system is started with the one-line diagram. The simulation is executed
while getting the fault location value from x variable. Then the load flow analysis and fault analysis is carried out. The fault
voltage values of the system are calculated with the variations in fault location, fault distance and fault type. Here, three
phase balanced line fault between line 2–3 (ETPS-MANALI) is considered at 40% of the line. After this, a contouring process
is done to visualise the faulted point in a colourful pattern based on the voltage magnitude profile after fault condition. The
one-line diagram of the test case with the contour diagram is given in the Fig. 7. The darker region marked red indicates
the faulted region. This is done to localise the faulted area. From the fault analysis, the geodataform simulation has been
created. This creation is done by plotting the buses at their geographical locations by inserting GIS data from shape-file. The
transmission lines are connected and the fault point is located after the fault analysis as shown in Fig. 8.
Using the closest facilities to point in the GIS tools, the fault point is described with latitude and longitude as given in
the Fig. 9. This is extended to Google Earth View by exporting the one-line as KML file. The buses presented in the test case
JID: CAEE
Fig. 7. Contouring for test case system in PWS.
Fig. 8. Geodataform-simulation result.
Fig. 9. Result with the closest facility point showing Latitude and Longitude of fault point.
JID: CAEE
Table 4
Results for unsymmetrical fault location estimation.
Fault type Actual location Fault resistance (Ω) Estimated location Latitude and longitude details Error (%)
(% of the line) (% of the line) of the estimated location

SLG (LINE 2–3) 20 10 20.0220 130 11 43.03 N 0.003

800 18 08.67 E

50 20.0276 130 11 43.03 N 0.003

800 18 08.67 E

40 10 40.0741 130 11 12.58 N 0.010

800 17 31.27 E
0
50 40.0855 13 11 12.59 N 0.011

800 17 27.69 E

60 10 60.0571 130 11 43.00 N 0.008

800 16 10.73 E

50 60.0922 130 11 43.08 N 0.012

800 16 48.49 E

80 10 80.0197 130 10 14.72 N 0.002

800 16 10.06 E

50 80.0294 130 10 14.34 N 0.004

800 16 10.73 E

LL (LINE 2–3) 20 10 20.0159 130 11 43.00 N 0.002

800 18 16.67 E

50 20.0249 130 10 43.03 N 0.003

800 18 08.67 E

40 10 40.0478 130 11 13.12 N 0.006

800 17 24.16 E
0
50 40.0910 13 11 12.64 N 0.012

800 17 23.76 E

60 10 60.0747 130 10 43.03 N 0.010

800 16 48.35 E

50 60.0576 130 10 43.00 N 0.008

800 16 53.05 E

80 10 80.0281 130 10 14.34 N 0.004

800 16 10.73 E

50 80.0208 130 10 14.34 N 0.002

800 16 10.73 E

LLG (LINE 2–3) 20 10 20.0235 130 11 43.03 N 0.003

800 18 08.67 E

50 20.0287 130 11 43.03 N 0.004

800 18 08.67 E

40 10 40.0187 130 11 13.16 N 0.002

800 17 22.40 E

50 40.0142 130 11 13.16 N 0.002

800 17 22.40 E

60 10 60.0246 130 10 42.16 N 0.003

800 16 51.69 E

50 60.0245 130 10 42.16 N 0.003

800 16 51.69 E

80 10 80.0831 130 10 17.33 N 0.011

800 16 20.91 E

50 80.0793 130 10 17.32 N 0.011

800 16 21.54 E
are plotted in the satellite view as presented in the Fig. 10. The faulted point is placed at n + 1 location, i.e., 50th bus and
marked with red marker is clearly shown in the Fig. 10. Hence, the exact point of the fault is given in the satellite image
of Google Earth View. In this aspect, the fault localization methodology is visualised easily so that the untrained users can
also read the result effortlessly. This enables the experts in the control room to visualise and spot the fault point easily. The
execution time taken for locating the fault using the proposed algorithm is 15.39 seconds. After the indication of the fault
occurrence, the relay will be operated at that instant. Thus, it reduces the operation time and supports it to recover fast.
4.3. Merits of the proposed method
The advantageous aspects of spotting latitude and longitude are as follows,

• Instead of inspecting the faulted zone or whole line, the location of the fault can be found precisely.
• This reduces the cost and computational burden of assessment and repair.
• The faster restoration and accuracy of fault location increases the customer service.
• The faster operation of isolating the faulted section helps to save the devices from destruction.
JID: CAEE
Fig. 10. Results in Google Earth View with n + 1buses.
• Hence, it elevates the quality and reliability in power system protection and prevents severe blackouts.
• Also, the total installation cost of PMUs is reduced by optimising the number of PMU for complete observability.
5. Conclusion
In recent decades, the advancements in the power system have enhanced and increased the complexity of the trans-
mission lines. The major difficulty is to find the faulted section and to locate the place at which the fault had occurred. To
solve this problem, a new method to locate the fault for multi-terminal transmission lines with optimal number of Phasor
Measurement Units is proposed. For optimization, the powerful meta-heuristic method namely, Genetic algorithm is used.
The pre-fault and post-fault data (voltage and current values) recorded by the Phasor Measurement Units are used to esti-
mate the exact location of the fault. After locating the fault, the type of fault such as bus fault and line fault are classified.
If line fault is identified, the distance between the sending bus and the faulted point is predicted. Then, the error between
the actual and estimated locations is calculated. Moreover, the type of fault (symmetrical and asymmetrical fault) is also
predicted by using the novel method. Then the scheme is extended to find the latitude and longitude of the fault location
using Geographic Information System mapping. The proposed technique has been demonstrated for a typical Indian utility-
TamilNadu Electricity Board 49 bus system. The theoretical analysis and simulation work are carried out to evaluate the
exact location of the fault. The test results indicate the accuracy and the enhancement of spotting the exact location of the
fault with latitude and longitude. Hence, the proposed technique is a promising way to reduce the burden in searching the
fault point and to help the fast recovery from the outages. This, in turn, elevates the power system reliability.
In the future research work, the authors will investigate the fault localization algorithm for simultaneous fault occurrence
and extend it to incorporate the concepts of fault clearing mechanisms.
Acknowledgement
The authors are thankful to the authorities of Thiagarajar College of Engineering, Madurai-625015, India, for providing all
the facilities to do the research work. This research was supported by Ministry of Human Resource Development, Govern-
ment of India through Technical Education Quality Improvement Programme (TEQIP) Scholarships under Ref: TCE/Dean-R&D/
TEQIP/ T&R Assistantship (2016–2017).
References
[1] Acharjee P. Strategy and implementation of smart grids in India. Energy Strategy Rev 2013;1:193–204.
[2] Korres GN, Manousakis NM. A state estimator including conventional and synchronized phasor measurements. Comput Electr Eng 2012;38(2):294–305.
[3] Li Weilin, Zhang Xiaobin. Simulation of the smart grid communications: challenges, techniques, and future trends. Comput Electr Eng 2014;40(1):
270–288.
[4] Mahamdedi B, Sanaye-Pasand M, Azizi S, Zhu JG. Unsynchronised fault-location technique for three-terminal lines. IET Gener Transm Distrib
2015;9(15):2099–107.
[5] Dehkordi PZ, Dobakhshari AS, Ranjbar AM. A decision tree-based method for power system fault diagnosis by synchronized phasor measurements. In:
3rd IEEE PES innovative smart grid technologies Europe (ISGT Europe); 2012. p. 1–9.
[6] Oubrahim Z, Choqueuse V, Amirat Y, Benbouzid M. An improved algorithm for power system fault type classification based on least square phasor
estimation. In: IECON2015-Yokohama, 9; 2015. p. 2735–40. (12).
JID: CAEE
[7] Dobakhshari AS, Ranjbar AM. Robust fault location of transmission lines by synchronised and unsynchronised wide-area current measurements. IET
Gener Transm Distrib 2014;8(9):1561–71.
[8] Dobakhshari AS, Ranjbar AM. A circuit approach to fault diagnosis in power systems by wide area measurement system. Int Trans Electr Energy Syst
2013;23:1272–88.
[9] Ahmadimanesh A, Shahrtash SM. Time-time-transform-based fault location algorithm for three-terminal transmission lines. IET Gener Transm Distrib
2013;7(5):464–73.
[10] Ghaffarzadeh N, Parpaei M, Tayar MZ. Fast fault location method based on detecting the minimum number of phasor measurement units using a novel
adaptive binary differential evolution optimization algorithm. Int Trans Electr Energy Syst 2015;25(11):2933–47.
[11] Jegarluei MR, Dobakhshari AS, Ranjbar AM, Tayebi A. A new algorithm for fault location on transmission lines by optimal PMU placement. Int Trans
Electr Energy Syst 2015;25(10):2071–86.
[12] Mortazavi SH, Sadeh J. An analytical fault location method based on minimum number of installed PMUs. Int Trans Electr Energy Syst
2016;26(2):253–73.
[13] Gopakumar P, Bharata Reddy MJ, Mohanta DK. Adaptive fault identification and classification methodology for smart power grids using synchronous
phasor angle measurements. IET Gener Transm Distrib 2015;9(2):133–45.
[14] Gopakumar P, Bharata Reddy MJ, Mohanta DK. Transmission line fault detection and localisation methodology using PMU measurements. IET Gener
Transm Distrib 2015;9(11):1033–42.
[15] Carlos Eduardo Morais dePereira, Luiz Cera Zanetta Jr. Fault location in multitapped transmission lines using unsynchronized data and superposition
theorem. IEEE Trans Power Delivery 2011;26(4):2081–9.
[16] Jiang G, Li X, Wang B, Wang H. PMU-based fault location using voltage measurements in large transmission networks. IEEE Trans Power Delivery
2012;27(3):1644–52.
[17] Saha MM, Izykowski J, Rosolowski E. Fault location on power networks. Springer Press; 2010.
[18] Shahraeini Mohammad, Ghazizadeh MohammadS, Javidi MohammadHossein. Co-optimal placement of measurement devices and their related com-
munication infrastructure in wide area measurement systems. IEEE Trans Smart Grid 2012;3(2):684–91.
[19] Abbasy NabilH, Ismail HanafyMahmoud. A unified approach for the optimal PMU location for power system state estimation. IEEE Trans Power Syst
2009;24(2):806–13.
[20] Aathinarayanan T, Shanmugam M. Estimation of fault distance in transmission line using DWT algorithm and GIS. Int J Eng Res Technol
2014;3(5):1120–3.
[21] Müller HeloisaH, Castro CarlosA. Genetic algorithm-based phasor measurement unit placement method considering observability and security criteria.
IET Gener Transm Distrib 2015;10(1):270–80.
[22] Jiang G, Wang B, Li X. An efficient PMU-based fault-location technique for multi-terminal transmission lines. IEEE Trans Power Delivery
2014;29(4):1675–82.
[23] Wang C, Dou CX, Li XB, Jia QQ. A WAMS/PMU-based fault location technique. Electr Power Syst Res 2007;77:936–45.
[24] Ramachandran Rajeswari, Karthick S. Optimal PMU placement for Tamil Nadu grid under controlled islanding environment. Procedia Technol
2015;21:240–7.
[25] Meenakshi Devi M, Geethanjali M. Optimal placement of phasor measurement units for observability using heuristic and meta-heuristic techniques.
Int J Appl Eng Res 2014;9(24):8457–63.
JID: CAEE
M. Meenakshi Devi received B.E. in Electrical and Electronics Engineering from Thiagarajar College of Engineering, India, and M.E. in Power Systems Engi-
neering from Anna University, Coimbatore, India. Currently, she is working towards the Ph.D. degree in Electrical and Electronics Department at Thiagarajar
College of Engineering, India. Her research interests include applications of Wide-area Monitoring, Protection and Control, and Smart Grid.
M. Geethanjali received her Ph.D degree in Electrical Engineering from Madurai Kamaraj University, India. Currently, she is working as Assistant Professor
in Department of Electrical and Electronics Engineering in Thiagarajar College of Engineering, India. She has authored around 50 papers published in
International journals and conferences. Her research interests include Power system protection, Smart Grid and Wide-area monitoring and Control.
A. Rama Devi received B.E. in Electrical and Electronics Engineering from University College of Engineering, Dindigul, India and M.E. in Power Systems
Engineering from Thiagarajar College of Engineering, Madurai, India. Her research interests include applications of Power System protection, Power System
analysis and control.

Computers and Electrical Engineering: M. Meenakshi Devi, M. Geethanjali, A. Rama Devi

Uploaded by

Copyright:

Available Formats

You might also like

Computers and Electrical Engineering: M. Meenakshi Devi, M. Geethanjali, A. Rama Devi

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Computers and Electrical Engineering: M. Meenakshi Devi, M. Geethanjali, A. Rama Devi

Uploaded by

Copyright:

Available Formats

JID: CAEE

ARTICLE IN PRESS [m3Gsc;February 7, 2018;8:48]

Contents lists available at ScienceDirect

Computers and Electrical Engineering

Fault localization for transmission lines with optimal Phasor

2. Optimal PMU placement technique

Fig. 1. Fault on busi .

3. Proposed algorithm for fault identiﬁcation

3.1. Fault on a bus

Let us consider fault at bus-i, as shown in the Fig. 1.

Fig. 2. Fault on linei − j .

3.2. Fault on a line

Yi1U1( ) + . . . + Y iiUi( ) + . . . + Y i jU j( ) + . . . + YinUn( ) + Y i f U f( ) = Ii( )

The expansion of jth row of Eq. (5) means,

Y j1U1( ) + . . . + Y ji Ui( ) + . . . + Y jjU j( ) + . . . + Y jnUn( ) + Y j f U f( ) = I (j )

The expansion of n+1th row of Eq. (5) results in,

From (8), voltage at fault point

By solving (9) in (6),

Similarly, by solving (9) in (7),

Yi1U1( ) + . . . + YiiUi( ) + . . . + Yi jU j( ) + . . . + YnnUn( ) = Ii( )

Y j1U1( ) + . . . + Y jiUi( ) + . . . + Y j jU j( ) + . . . + YnnUn( ) = I (j )

while solving (15) in ith line of node admittance matrix, it results in

Yi1 U1 + . . . + Y ii Ui + . . . + Y i j U j + . . . + Yin Un + Y i f U f = 0

Similarly jth line results in,

Y j1 U1 + . . . + Y ji Ui + . . . + Y j j U j + . . . + Y jn Un = xI f (17)

The fault current at fault point will be placed at n + 1 location [23].

3.3. Effect of fault resistance

3.4. Geographic data view

(1) Steps involved

Fig. 3. Flowchart explaining the steps involved in proposed methodology.

Bus system Number of PMUs required Location of PMUs (Bus numbers)

Ref [22] = N 49 1470

4. Results and discussion

4.1. Optimal PMU placement technique using GA

4.2. Fault identiﬁcation algorithm

Fig. 4. One-line diagram of the Indian utility (TNEB) 49 bus system.

Fig. 7. Contouring for test case system in PWS.

Fig. 8. Geodataform-simulation result.

4.3. Merits of the proposed method

The advantageous aspects of spotting latitude and longitude are as follows,

Fig. 10. Results in Google Earth View with n + 1buses.

You might also like