Received: 17 August 2021 Revised: 27 October 2021 Accepted: 17 November 2021 IET Generation, Transmission & Distribution

DOI: 10.1049/gtd2.12367

ORIGINAL RESEARCH PAPER

Accurate fault location method for distribution network in

presence of DG using distributed time-domain line model

Faeze Mohajer Javad Sadeh

Electrical Engineering Department, Faculty of Abstract

Engineering, Ferdowsi University of Mashhad,
Mashhad, Iran
In this paper a novel technique is presented for fault location in distribution networks
in the presence of distributed generators (DGs). The proposed approach uses recorded
Correspondence voltage and current at the beginning of feeder and DG terminals. This method is based
Javad sadeh, Electrical Engineering Department, on transient state data, and due to presence of high frequencies in this state, to achieve
Faculty of Engineering, Ferdowsi University of
high accuracy it utilizes the distributed time-domain (DTD) line model. The fault location
Mashhad, Mashhad, Iran.
Email: [email protected] problem is formulated as an optimization problem and is solved using Genetic Algorithm
(GA). Through modal transformation, in addition to locate the symmetrical three-phase
to ground (3LG) faults, the location of asymmetrical faults including single line to ground
(SLG) and double line (LL) faults can also be determined. The effects of fault distance, fault
resistance and fault inception angle on the accuracy of the proposed algorithm have also
been investigated. A solution is also proposed to overcome the multi-response problem.
The simulations are conducted and the results are obtained via PSCAD and MATLAB
softwares, respectively. The effectiveness of the proposed approach has been verified using
the modified IEEE 34-node test feeder under different fault conditions. The simulation
results demonstrate the high efficiency and accuracy of the proposed method.

1 INTRODUCTION tion network. This multi-response nature is reported as the main

demerit of the impedance-based fault location methods as well.
Recently, distributed generation (DG) has been developed Mora-Florez, et al. compared ten different impedance-based
marked due to the economic and environmental concerns and methods which utilize the data at the beginning of the feeder
popularity of new resources. The distribution network has been [2]. The fundamental frequency component of the voltage and
thereby changed from a passive state to an active one with dif- current measured at the line terminal as well as the static load
ferent power supply locations, making the power flow bidirec- model and the lumped line model for network have been applied
tional. in [3]. Fifth-order polynomial equations have been derived in
Accurate fault location is one of the important issues for [4] by using a distributed-parameter line model, to locate the
improvement of the network availability and reliability. In the fault in DN. The proposed impedance-based algorithm does
active distribution network with bidirectional power flow, accu- not specify the faulted section, however. In the traveling-based
rate fault location is a challenging topic. In this regard, various approach with complex structure, the fault location has been
methods which can be categorized into three categories have determined using the time interval between forward and back-
been presented; (i) impedance-based, (ii) traveling-based and (iii) ward high frequency traveling waves generated by the fault. For
other methods. Impedance-based approaches are based on the instance, a novel algorithm has been adopted by Liang, et al.
fundamental frequency component of the voltage and current. exploiting the information of traveling wave recorders (TWRs)
Although these algorithms are realized simply, they are very sen- and network topology [5]. This algorithm collects fault signals
sitive to the fault resistance and phases mutual interaction [1]. from TWRs located in network, and then use the technology of
Moreover, different locations may be spotted as the fault loca- the global positioning system (GPS) to synchronize the voltages
tions regarding the scattered loads and laterals in the distribu- obtained from the wide-area measurement system (WAMS). In

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is
properly cited.
© 2021 The Authors. IET Generation, Transmission & Distribution published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology

IET Gener. Transm. Distrib. 2022;16:1305–1318. wileyonlinelibrary.com/iet-gtd 1305

 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1306 MOHAJER AND SADEH

the presented scheme, a distributed line model has been used accuracy of the fault location in distribution networks. An
and the percentage of fault location error has been reduced improved impedance-based scheme without considering DG
to 0.13 accordingly. The distance of the permanent fault has has been provided in [20], requiring solely the substation mea-
been obtained in [6] according to the reclosing instant and the surements. Aboshady, et al. modelled the fault as a step volt-
arrival instant of the traveling wave reflected from the fault age source and the error of the found fault location has been
point. Afterwards, the modified dynamic time warping has been reported less than 50 m. The zero-sequence voltage and cur-
employed to detect the exact fault section among the candidates. rent have been adopted in [21] to define a novel SLG fault loca-
The detection speed and accuracy have been raised using tion scheme for radial distribution network with single source.
heuristic and intelligent methods. These techniques often use Dashti and Sadeh exploited a dynamic model for the loads, con-
the high frequency component of the voltage and current and nected to the distribution network [22]. A genetic algorithm
require large databases or smart devices [7–11]. Jiang establishes has been used initially to solve the optimization problem, and
a mixed integer linear programming (MILP) method for esti- then the current pattern of the protection devices such as fuses,
mating the faulty line section using the incomplete and incorrect reclosers, and fault indicators has been analysed to find the
smart fault indicators (SFIs) statuses [7]. In the presented algo- fault section. A technique for locating faults in DNs with DGs
rithm, several simultaneous scenarios, including multiple faults, through the information of the voltage and current at begin-
failures, and malfunctions of SFIs can also be handled in the ning of the feeder has been reported in [23]. In the two-level
presence of DGs. Markov decision process and Bayes proba- algorithm, the fault distance has been determined in the first
bility theory has been suggested by Janjic and Velimirovic for stage and then solves the multi-response problem using the
the determination of the faulted feeder section and its isolation voltage frequency component analysis. Daisy and Dashti pro-
from the network [8]. This method has been investigated for vided a hybrid methodology to find the SLG faults location [24].
different operation conditions including the presence of DGs. First, the probable locations of the fault have been determined
Silos-Sanchez, et al. provides an algorithm for active distribution through the impedance-based on the algorithm presented in
network which has the capability of combination with internet [22]. Then a new algorithm has been subsequently employed
of things (IoT) and is independent of the network topology [9]. for determining the faulty section using a voltage sag matching
Buzo, et al. developed a fault location methodology for both technique.
active and passive distribution networks [10]. The presented Sadeh, et al. presented a fault location method for series com-
method requires impedance matrix and the information of volt- pensated transmission lines using DTD line model and the sam-
age sag measured by a large number of intelligent electronic ples of voltages and currents at both ends of the line are uti-
devices, installed at the substation and some of the network’s lized to increase the accuracy of the results [25]. Several simula-
end nodes. A fault location method based on Fibonacci search tions under different fault conditions including different fault
algorithm and minimum entropy theory that can be applied to inception angles, different fault resistances, symmetrical and
DNs with any DG types has been adopted in [11]. This scheme unsymmetrical faults highlighted the accuracy of the presented
uses the information collected by the phasor measurement units method. Like all two-terminal methods, the need for communi-
(PMUs) and can be only applied to recognize the location of the cation links and synchronization of the data are of the deficien-
single to the ground (SLG) faults. cies of this method. In this paper, a new and accurate fault loca-
In the fault location study, the accurate modelling of the tion method is proposed based on the transient states data in
network components including lines, loads, and resources has active DNs. On contrary of the presented works regarding the
vital effect on the results. Different models can be adopted to line models, a distributed time-domain (DTD) line model has
model the distribution network lines based on length, mutual been used in the proposed algorithm. The DTD model sim-
impedances, and capacitive effects. In the long length line ulates the lines characteristics precisely regarding the presence
model, all resistance, inductance, capacitance, and leakage con- of high frequency components in the transient states. The pre-
ductance parameters are considered while the leakage conduc- sented fault location scheme uses genetic algorithm to calcu-
tances and capacitances are neglected in the short and medium late the fault distance and fault resistance simultaneously. In the
line length models. In the most of existing researches, the short proposed technique, the location of asymmetrical LL and SLG
and lumped line models have been exploited [1–4, 12–19]. In faults has been calculated as well by applying modal transforma-
[18, 19] the π line model has been used for the lines model in tion. The current research also paves the way to determine accu-
DN. A new quadratic equation has been derived in [18] to find rately the faulty section. The proposed method provides several
the fault location in DNs using the recorded voltage and current advantages such as:
at the beginning of the feeder and DG terminals. The fault loca-
tion method in [19] employed the voltage magnitude, retrieved ∙ This algorithm is applicable to distribution networks with one
at the end of lateral. The outputs of the presented tests under or multiple DGs.
SLG and line-line to ground (LLG) faults highlighted the good ∙ Proposed scheme, has an outstanding performance irrespec-
performance of the algorithm disregarding the DG types and tive of fault inception angle and fault resistance.
dynamic loads. ∙ The presented scheme uses only two cycles of recorded
Further, several scholars have used the distributed-parameter voltage and current at the beginning of feeder and DG
line model in the frequency domain [20–23] to enhance the terminals.
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOHAJER AND SADEH 1307

′
−Rl ∕4 [ ′′
]
I j (t − 𝜏l ) = V j (t − 𝜏l ) + Zcl i j (t − 𝜏l )
′ 2
Z cl
Z [ ′′
]
− c 2 Vi (t − 𝜏l ) + Zcl ii (t − 𝜏l ) (4)
′
Z cl
FIGURE 1 Line with distributed parameter
Finally, the voltage of point j can be expressed in the terms
of the voltage and current at the beginning of the line as follows
by manipulating of Equations (1)–(4) [25]:
(
[ ]
V j (t ) = Z ′2 ′ (
cl Vi t + 𝜏l − Zcl ii t + 𝜏l
( ) )

′′2 [ ]
+ Z cl Vi (t − 𝜏l ) + Zcl′′ ii (t − 𝜏l )
FIGURE 2 Distributed line model for section i − j )
Zcl′ Rl′ [( ′ ′′
) ′′
]
− Rl ∕2Zcl Vi (t ) + 2Zcl ii (t ) ∕2Zc 2 (5)
4
2 PROPOSED METHOD
DESCRIPTION FOR FAULT LOCATION IN
Similarly, the current of point j can be given by:
SINGLE SECTION
2
Z cl
′ ⎛ −Z ′′ 2 [ ]
In this paper, new expressions for calculating fault location are ⎜ cl ′′
i j (t ) = − Vi (t − 𝜏l ) + Zcl ii (t − 𝜏l )
derived based on the time-domain equations of the distributed ′′
⎜
Zc Zcl ⎝ 2Z Z ′ 2
parameters line model. Since DNs have several sections and lat- c cl

erals, the equations of the fault distance and resistance estima- ′ [( ′ ) ( ) ]

Rl Rl Vi (t ) ′′ 1 1
tion in a single-section are derived at first. The equations are + ′ −1 ′
+ Zcl − ′ ii (t )
then developed to the whole DN in the next section. 4Zcl 4Zc Zcl Zc Zcl
A single-phase line diagram with distributed parameters ( ) ( )
model is shown in Figure 1. In this figure, Vi and V j are the
′
⎞
−1 1 Zcl
voltages at the beginning (i) and end points (j) of the line. The + + ′ Vi (t + 𝜏l ) + − 1 ii (t + 𝜏l )⎟
2Zc Zcl 2Zc ⎟
distributed-parameter line model in time-domain for section i–j ⎠
is illustrated in Figure 2 [26]. (6)
According to Figure 2, the current at the beginning (i) and
end points (j) of the line can be given by:
With the presence of DG in DN, fault current in some lines
feeds from two side, thereby, based on whether the fault current
1
ii (t ) = ′
Vi (t ) + Ii (t − 𝜏l ) (1) is fed from one or both sides of the line, the network lines are
Zcl divided into two groups, as follows:
1
i j (t ) = ′
V j (t ) + I j (t − 𝜏l ) (2) ∙ First group (FG): In the FG lines, the fault current is injected
Zcl only from the upstream side of the fault. Thus, the voltage
and current can be calculated just at the beginning of the line
where 𝜏l is the elapsed time of the wave propagation from after fault initiation. The FG includes lines and branches of
beginning to the end of line. In addition, Zc and Rl′ are the DG downstream (e.g. L5 in Figure 3) as well as DG upstream
characteristic impedance and total resistance of the line, respec- branches (e.g. L2 and L3 in Figure 3).
′ ′′
tively. These quantities can be used to define Zcl and Zcl , that is,
′ ′′
Zcl = Zc + Rl′ ∕4 and Zcl = Zc − Rl′ ∕4.
The dependent current sources Ii and I j can be then defined
as follows:
′
−Rl ∕4 [ ′′
]
Ii (t − 𝜏l ) = Vi (t − 𝜏l ) + Zcl ii (t − 𝜏l )
′ 2
Z cl
Z [ ′′
]
− c 2 V j (t − 𝜏l ) + Zcl i j (t − 𝜏l ) (3) FIGURE 3 Single line diagram of the sample network, categorize the
′
Z cl location of faults (FG & NG)
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1308 MOHAJER AND SADEH

FIGURE 4 Single line diagram of the symmetrical three-phase fault in a

single-section

∙ Next group (NG): In the NG lines, the current is injected

from both sides of the fault; hence, if a fault occurs, the volt-
age and current can be calculated at both beginning and end
of the line using voltage and current information of main sub-
station and DG. Some of the lines of DG upstream can be
considered in this category, e.g. L1 and L4 in Figure 3.

In the following subsections, the method for calculating

the fault distance and resistance for both groups is explained
thoroughly. It is worth to mention that operation mode of
DG (island or grid-connected) has no effect on the proposed
method and what matters is the recorded information in DG
location. FIGURE 5 Flowchart of GA for optimization problems

2.1 Calculating the fault distance and can be used to define Zcx′ and Zcx′′ , that is, Zcx′ = Zc + Rx′ ∕4 and
resistance in FG lines Zcx′′ = Zc − Rx′ ∕4.
Because of the continuity of the voltage along the line, Equa-
In Figure 4, line i − j is assumed to be a FG line and a fault tion (9) is obtained by combining Equations (7) and (8):
has been depicted on it. To calculate the location of fault in line ( )
i − j , the relationship between voltage and current in the point F Vi ,ii ,R f ,i f ,t ,𝜏x
of fault for symmetrical three-phase fault can be written as fol- (
[ ]
lows, where, Rf and if represent the fault resistance and current, = Z ′2 ′
cx Vi (t + 𝜏x ) − Zcx ii (t + 𝜏x )
respectively.
Vx (t ) = R f i f (t ) (7) [ ′′
]
+ Z ′′2
cx Vi (t − 𝜏x ) + Zcx ii (t − 𝜏x )
As mentioned earlier, voltage and current samples at bus i )
Zcx Rx [( ′ ) ]
′ ′
in FG lines can be obtained, therefore by applying DTD line ′ ′′ ( )
− Rx ∕2Zcx Vi (t ) + 2Zcx ii (t ) ∕2Zc 2 − R f i f
model of section i − F , similar to Equation (5) the voltage of 4
point F can be derived in terms of the voltage and current at the
beginning of the line as follows (Point F in Figure 4 indicates a =0 (9)
three-phase fault):
′
( The fault distance x appears in 𝜏x and Rx . It is worth to men-
[ ]
Vx (t ) = Z ′2 ′
cx Vi (t + 𝜏x ) − Zcx ii (t + 𝜏x )
tion that Vi and ii , are obtained using voltage and current sam-
ples at the main substation and i f is calculated by an iterative
[ ] algorithm in the current work.
+ Z ′′2 ′′
cx Vi (t − 𝜏x ) + Zcx ii (t − 𝜏x )
In order to find the fault distance and resistance, Equation (9)
)
Zcx′ Rx′ [( ′ ′
) ′′
] is initially discretized and afterwards, the following optimization
− Rx ∕2Zcx Vi (t ) + 2Zcx ii (t ) ∕2Zc 2 problem is solved by genetic algorithm in the MATLAB plat-
4
form (Flowchart of GA is shown in Figure 5):
(8) ( ) ∑ ( )
min J m,R f = min F 2 Vi ,ii ,R f ,i f ,k,m
m,R f m,R f
where 𝜏x is the elapsed time of the wave propagation from i k
′
to F . In addition, Rx is line resistance from i to F. Zc and Rx′ m × Δt = 𝜏x , k × Δt = t (10)
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOHAJER AND SADEH 1309

FIGURE 7 Single line diagram of fault at DG upstream

According to Figure 6b, Equations (12)–(14) can be rewritten

for LL fault as Equation (15)–(17):

Vxa − Vxb = R f i f (15)

FIGURE 6 Asymmetrical faults: (a) SLG fault (b) LL fault 3Vx1 − R f i f = 0 (16)

( ) ∑ ∑( )2
min J m,R f = min F 2 = min 3Vx1 − R f i f (17)
m,R f m,R f m,R f
where, Δt is a sampling step and m and k are arbitrary k k

integers.
Positive, negative, and zero sequences of voltage at the fault
The recent expressions can be applied to the asymmetrical
point in the above equations can be calculated as follows:
faults by modal transformation. The measured voltage and cur-
rent can be converted from the phase domain to the modal
⎛
domain, using the following equations [26]: ′2 [ ( ) ′ ( )]
Vx(n) (t ) = ⎜Zcx(n) Vi(n) t + 𝜏x(n) − Zcx(n) ii(n) t + 𝜏x(n)
⎜
⎝
⎡1 1 1⎤ ⎡ga ⎤ ⎡g0 ⎤ ′′ 2 [
+ Zcx(n)
( ) ′′ 2
Vi(n) t − 𝜏x(n) + Zcx(n)
(
ii(n) t − 𝜏x(n)
)]
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
M = ⎢1 −2 1⎥ ⎢gb ⎥ = M ⎢g1 ⎥ (11) ′ ′ /
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ Zcx(n) Rx(n) [( ′ ′ ) ′′ ]⎞
⎣1 1 −2⎦ ⎣ gc ⎦ ⎣g2 ⎦ − Rx(n) ∕2Zcx(n) Vi(n) (t ) + 2Zcx(n) ii(n) (t ) ⎟ 2Zc(n)
2
4 ⎟
⎠
where, g is the voltage or current quantities. The zero, positive, (18)
and negative sequences are also represented by “0”, “1”, and
“2” indices, respectively. where, n is sequence number and it can take 0, 1 or 2. 𝜏x(n) is
According to Figure 6a, Equation (12) can be derived for a the elapsed time of the wave propagation from i to F for nth
SLG fault. This expression can be expressed as Equation (13) ′
sequence. In addition Rx(n) and Zc(n) are line resistance from i to
by applying modal transformation. F and the characteristic impedance for sequence n, respectively.
′ ′′ ′
These quantities can be used to define Zcx(n) and Zcx(n) , i.e. Zcx(n) =
Vxa = R fa i fa (12) ′ ′′ ′
Zc(n) + Rx(n) ∕4 and Zcx(n) = Zc(n) − Rx(n) ∕4.

Vx0 + Vx1 + Vx2 = R fa i fa (13)

2.2 Calculating the fault distance and
Similar to Equation (9), the following term should be opti- resistance in NG lines
mized to calculate the distance and resistance in SLG faults by
knowing the i fa : The DTD model of the line and the data of both terminals are
used in [25] to locate the fault in the transmission lines. In the
( ) ∑ line i − j with NG, when a fault occurs, voltage and current
min J m,R fa = min F2
m,R fa m,R fa samples at the beginning and end of the line can be obtained
k
(Figure 7). Therefore, the same methodology as [25] is exploited
∑( )2
= Vx0 + Vx1 + Vx2 − R fa i fa (14) in the current work. Referring to Equation (7), the voltage of the
k fault point can be defined in the terms of the voltage and current
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1310 MOHAJER AND SADEH

at the end of the line as follows: ′ 2

⎛ −Z ′′ [
2
Zcj ]
iFdg (t ) = − ⎜ cj ′′
V j (t − 𝜏 j ) + Zc j i j (t − 𝜏 j )
⎛ [
′′
⎜
Zc Zc j 2Zc Z ′ 2

′ ] ⎝ cj
Vx (t ) = ⎜Z ′2 V j (t + 𝜏 j ) − Zc j i j (t + 𝜏 j )
⎜ c j
[( ) ( ) ]
⎝ Rj
′ ′
Rj V j (t ) ′′ 1 1
+ −1 + Zc j − ′ i j (t )
[ ′′ ] 4Zc j
′
4Zc ′
Zc j Zc Zc j
+ Z ′′2
c j V j (t − 𝜏 j ) + Zc j i j (t − 𝜏 j )

′ ′ / ( ) ( )
]⎞ ⎞
′
Zc j R j [ −1 1 Zc j
−
′ ′
(R j ∕2Zc j )V j t + 2Zc j i j t
( ) ( ) ⎟ ′′
2Zc2 + + ′ V j (t + 𝜏 j ) + − 1 i j (t + 𝜏 j )⎟ .
4 ⎟ 2Zc Zc j 2Zc ⎟
⎠
⎠
(24)
(19)

′ ′ In addition, the fault resistance in Equation (22) can be cal-

The relationships between the parameters Rx , R j and culated using the following equations in which vectors A and
′ ′ ′
𝜏x , 𝜏 j can be defined as Rx + R j =Rl and 𝜏x + 𝜏 j = 𝜏l B are the functions of voltage and current at the beginning and
respectively. end of the line.
Due to the voltage continuity along the line, Equations (7)
and (19) are combined to determine the fault location as follows: A = iFs + iFdg and B = Vx (t ) (25)
( )
F Vi ,ii ,V j ,i j ,t ,𝜏x ,𝜏 j = 0 (20) ( )−1
A = R f B → R f = A T A AT B (26)
Similar to Equation (9), Equation (20) is discretized and
solved by the following optimization problem through genetic In this expression, T represents the vector transpose.
algorithm to identify the fault distance:
∑ ( ) 3 APPLYING THE PROPOSED
min J (m ) = min F 2 Vi ,ii ,V j ,i j ,k,m (21) METHOD TO THE DISTRIBUTION
m m
k
NETWORK
As mentioned earlier, by modal transformation, the above
In DNs, only the voltage and current in the substation and DG
relationships can be applied to asymmetrical faults. Further-
resources are measured. Subsequently, to determine the fault
more, the fault resistance can be determined by calculat-
distance in DN, first:
ing the fault distance and voltage at the point of fault. For
example, for three-phase fault, the following equation can be
∙ calculation of voltage and current samples at the beginning
written:
of each line
( ) ∙ calculation of the equivalent load at the end of each line (Rload
Vx (t ) = R f iFs + iFdg (22)
and Lload )
In this expression, iFs and iFdg are the injected currents to the
fault from the upstream and downstream of the fault, respec- Thus, using an iterative algorithm, the fault location is calcu-
tively. Similar to Equation (6), iFs and iFdg can be determined as lated precisely.
a function of Vi and ii (Equation (23)) and V j and i j (Equation
(24)), respectively.
3.1 Determination of voltage and current at
′ 2 ( ′′ 2 the beginning of each line
Z cx −Z cx [ ′′ ]
iFs (t ) = − ′′
Vi (t − 𝜏x ) + Zcx ii (t − 𝜏x )
Zc Zcx ′ 2
2Zc Z cx Voltage and current at the beginning of the feeder are used to
calculate voltage and current at the beginning of other sections.
′ [( ′ ) ( ) ]
Rx Rx Vi (t ) ′′ 1 1 First, one main path (MP) should be selected, preferably the
+ ′
−1 ′
+ Z cx − ′
ii (t )
4Zcx 4Zc Zcx Zc Zcx longest path to which most branches are connected, for instance
from node S to node 6 in Figure 8. So depending to the lines
( ) ( ′ ) ) position, voltage and current at the beginning of each line are
−1 1 Zcx calculated by combining three different ways. Figure 9 illustrates
+ + ′ Vi (t + 𝜏x ) + − 1 ii (t + 𝜏x )
2Zc Zcx 2Zc the flowchart of voltage and current calculation at different sit-
(23) uations and in the following is explained in detail.
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOHAJER AND SADEH 1311

Equation (28) can be used for calculating current derivative,

also, Equation (29) can be obtained by merging Equations (27)
and (28). In the following equations, k is the sample number.

d i2 (tk ) i (t ) − i2 (tk−1 )
= 2 k (28)
dt Δt
( )
Lload 1 L
V2 (tk ) = Rload 1 + i2 (tk ) − load 1 i2 (tk−1 ) (29)
FIGURE 8 Single-line diagram of sample network with DG Δt Δt

Using DTD model of line between nodes 1 and 2, V2 and

i2 can be rewritten similar Equations (5) and (6) respectively as
a function of voltage and current at the beginning of the line
′
(V1 ,i1 ). Then, by merging the relations of V2 and i2 by Equation
′
(29), i1 is calculated in term of V1 using Equation (30).

⎛−Z ′ 3 ( )⎛ ′ 2 ⎞( ) −1
Zcl ⎞
′
′
⎜ Lload 1 ⎜ Z cl ⎟ ⎟
i1 (tk + 𝜏 ) = cl
+ Rload 1 + −1 +
⎜ 2Zc2 Δt ⎜ Zc Z ′′ ⎟ 2Zc ⎟
⎝ ⎝ cl ⎠ ⎠

⎛ ( )⎛ ′ 2 ⎞ ⎛ ′′ 2

⎜ Lload 1 ⎜ Z cl ⎟ ⎜ −Z cl [ ′′ ′ ]
× − Rload 1 + V (t −𝜏 ) +Zcl i1 (tk − 𝜏l )
⎜ Δt ⎜Zc Z ′′ ⎟ ⎜ 2Z Z ′ 2 1 k l
FIGURE 9 Flowchart of calculating voltage and current at different ⎝ ⎝ cl ⎠ ⎝ c cl
situations ′
[( ′ ) ( ) ]
R Rl V1 (tk ) ′′ 1 1 ′
+ l′ −1 + Z − i (t )
4Z 4Zc Z ′ cl cl Zc Z
′ 1 k
cl cl

( )
3.1.1 Determination of voltage and current at ⎞
−1 1
the end of the lines in the main path + + ′ V1 (tk + 𝜏l ) ⎟
2Zc Zcl ⎟
⎠
Main substation (node s) is the first node of the MP, therefore ′
Z cl ⎛ −Z ′′ 2 [
2
]
using the Vs and is , the voltage and current at the end of the ⎜ ′′ ′
+ cl
V1 (tk−1 − 𝜏l ) + Zcl i1 (tk−1 − 𝜏l )
first line can be calculate. Generally, the voltage and current at ′′
⎜
Zc Zcl 2Zc Z ′ 2
⎝ cl
the end of the MP lines are calculated by employing the voltage
and current at the beginning of the line, hierarchically to the ′
[( ′
) ( ) ]
Rl Rl V1 (tk−1 ) ′′ 1 1 ′
last node of the MP. For example, using the DTD model of line + −1 + Zcl − ′ i1 (tk−1 )
4Zcl
′
4Zc Z ′ cl Zc Zcl
between nodes 1 and S in Figure 8 and the voltage and current at
the beginning of line (Vs ,is ), the input current to node 1 and its ( ) ( )
voltage (i1 ,V1 ) can be calculated. Moreover, if the fault current
′ ⎞
−1 1 Zcl
− 1 i1 (tk−1 + 𝜏l )⎟
′
+ + ′ V1 (tk−1 + 𝜏l ) +
is fed from two directions (such as F1 in Figure 8), the voltage 2Zc Zcl 2Zc ⎟
⎠
and current of DG bus (VDG ,iDG ) is exploited to calculate the
′′
voltage and current of the fault downstream buses (V4 ,i4 ). ⎛
′′ [ ]
− ⎜Zcl [V1 (tk + 𝜏l )] + Zcl V1 (tk − 𝜏l ) + Zcl i1 (tk − 𝜏l )
′ ′′ ′

⎜
⎝
3.1.2 Determination of current at the beginning [ ] /
′ ′ ′
Rl ∕2 ⎞ ⎞
of the branch Z R
V1 (tk ) + 2Zcl i1 (tk ) ⎟ 2Zc ⎟
′′ ′
− cl l (30)
4 Zcl
′
⎟ ⎟
⎠ ⎠
In some situations, it is necessary to calculate the current at the
′′
beginning of the branch. For example to calculate i1 and also in
the case of fault occurrence in branch 1–2 of Figure 8, it is need 3.1.3 Determination of the load currents
′ ′′ ′ ′
to calculate i1 ( i1 = i1 − i1 ). To calculate i1 , assuming that Rload 1 connected to the main path
and Lload 1 are respectively the resistance and inductance of the
load at the end of the line, the relationship between voltage and Load 2 in Figure 8 is connected directly to the main path. For
current of load can be given by: ′ ′′
calculating i3 , it is necessary to obtain load current (i3 ). At first,
for Load 2, the relationship between voltage and current can be
d i2 (t ) ′
V2 (t ) = Rload 1 i2 (t ) + Lload 1 (27) written similar to Equation (29). Then the samples of i3 can be
dt
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1312 MOHAJER AND SADEH

calculated by Equation (31) as a function of load voltage and lowed; otherwise, the method jumps to step 11.
previous samples of load current.
|xnew − xold | < 𝜀 (33)
( )−1 ( )
′ L L ′
i3 (tk ) = Rload 2 + load 2 V3 (tk ) + load 3 i3 (tk−1 ) where, xold is the calculated fault distance in the previous
Δt Δt
iteration. Moreover, xnew is the calculated fault distance in
(31) the step 9 whilst 𝜀 is the predefined tolerance.
10. (a) If xnew > Lline : add one to the line number and go back
It is worth mentioning that the first sample of current in the to step 4. (b) If xnew < Lline : print values of x, R f , and line
above equations is considered equal to the last sample of current number, add one to the line number and then go back to
before fault. step 4.
11. Calculate the voltage of the fault point using Equation (7).
3.2 Determination of the equivalent load 12. Update iu and id according to the following equations, then
impedance at the end of each section go back to step 7. Similar to Equation (6), the following
equation can be written for updating id by applying the
The voltage and current data at the end of the line are obtained equations of DTD line model to section i − F :
(
by applying the relations that are expressed in Section 3.1. Fol- ′
Z cx
2 ′′
−Z cx [
2
]
′′
lowing this, the equivalent load (Rload and Lload ) at the end of id (t ) = − ′′
Vi (t − 𝜏x ) + Zcx ii (t − 𝜏x )
′ 2
Zc Zcx 2Zc Z cx
each line is estimated through dividing the fundamental fre-
quency component of voltage and current. ′ [( ′ ) ( ) ]
R Rx Vi (t ) ′′ 1 1
+ x′ −1 ′
+ Zcx − ′ ii (t )
4Zcx 4Zc Zcx Zc Zcx
3.3 Flowchart of the proposed method ( ) ( ′ ) )
−1 1 Zcx
+ + ′ Vi (t + 𝜏x ) + − 1 ii t + 𝜏x
( )
The flowchart of the proposed fault location method in pres- 2Zc Zcx 2Zc
ence of DG is shown in Figure 10 and its steps are explained
(34)
below:
−1
1. Fault detection and fault type determination: The presented ′ 3 ( )⎛ ′ 2 ⎞⎛ ′
( ) ⎛ −Z c j Lload ⎜ Z c j ⎟ ⎜ Zc j ⎞⎞
method in [27] or similar algorithms can be used. iu tk + 𝜏 j = ⎜ + Rload + −1 + ⎟⎟
⎜ 2Z 2 Δt ⎜ Zc Z ′′ ⎟ ⎜ 2Zc ⎟⎟
2. (a) Determine voltage and current at the beginning of the ⎝ cj ⎝ cj ⎠ ⎝ ⎠⎠
lines of FG and also equivalent load at their end. (b) Deter-
⎛ ( )⎛ ′ 2 ⎞
Z ⎛ −Z ′′ 2 [ ]
mine voltage and current at the beginning and end of the L ′′ ′
× ⎜− Rload 1 + load 1 ⎜ cl ′′ ⎟ ⎜ cl
V1 (tk − 𝜏l ) +Zcl i1 (tk − 𝜏l )
lines of NG as well as the equivalent load of DG down- ⎜ Δt ⎜ Zc Z ⎟ ⎜ ′ 2
⎝ ⎝ cl ⎠ ⎝ 2Zc Z cl
stream. It is worth to mention that in lines of NG, voltage ( )
′ ′
and current at the end of the line are calculated by using the R j ⎡⎛ R j ⎞ V (t )
′′ 1 1
⎤
+ ⎢⎜ − 1⎟
f k
+ Zc j − ′ iu (tk )⎥
DG voltage and current. ′
⎢⎜ 4Z
4Zc j ⎣⎝ c ⎟ Z′ Zc Zc j ⎥
⎠ cj ⎦
3. Line selection.
( )
4. If the line is belonged to NG, the distance and resistance of ⎞
−1 1
fault are obtained using the proposed method in Section 2.2 + + ′ V f (tk + 𝜏 j )⎟
2Zc Zc j ⎟
and step 9 should be followed; otherwise if line is belonged ⎠
to FG, step 5 should be followed. ′ 2
⎛ −Z ′′ [ (
2
Zcj ) ′′ ( )]
5. As shown in Figure 11, load current (iL ) before and after + ⎜ cj
V f tk−1 − 𝜏 j + Zc j iu tk−1 − 𝜏 j
′′
⎜
Zc Zc j ⎝ 2Z Z ′ 2
the fault inception is assumed to have the same values. In c cj
addition, the output current from the fault point to the end ′ ′ ( )
of line (iu ) is assumed to be equal to iL . R j ⎡⎛ R j ⎞ V (t )
′′ 1 1
⎤
+ ′ ⎢⎜ − 1⎟ iu (tk−1 )⎥
f k−1
+ Zc j − ′
6. Input current to the fault point from upstream of the line ⎢⎜ 4Z
4Zc j ⎣⎝ c ⎟ Z
′ Zc Zc j ⎥
⎠ c j ⎦
(id ) is assumed to be ii .
( ) ′
7. Calculate the fault current by the following equation: −1 1 ( ) ⎛ Zc j ⎞ ( )⎞
+ + ′ V f tk−1 + 𝜏 j + ⎜ − 1⎟ iu tk−1 + 𝜏 j ⎟
2Zc Zc j ⎜ 2Zc ⎟ ⎟
⎝ ⎠ ⎠
i f = id − iu (32)
⎛ [ ( )]
′ ′′ [ ( ) ′′ ( )]
8. Calculate the fault distance and resistance using one − ⎜Zc j V1 tk + 𝜏 j + Zc j V f tk − 𝜏 j + Zc j iu tk − 𝜏 j
⎜
of the Equations (10), (14) or (17) regarding the fault ⎝
type. ′ ′ ′
Zc j R j ⎡ R j ∕2 ⎤⎞ / )
′′
9. For the next iteration, check the convergence of x using − ⎢ V f (tk ) + 2Zc j iu (tk )⎥⎟ 2Zc (35)
4 ⎢ Z ′
⎥⎟
Equation (33). If this inequality is satisfied, step 10 is fol- ⎣ cj ⎦⎠
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOHAJER AND SADEH 1313

FIGURE 10 Flowchart of the proposed fault location method

For updating iu using only fault point voltage, the following

equation can be written (similar to Equation (30)). Moreover,
in the Equation (35), the load current samples before the fault
occurrence are exploited for iu of the time samples before 𝜏 j .

3.4 Determination of fault section

In DNs, multiple fault locations may be obtained due to the

presence of multiple laterals in fault location algorithms. In few
cases, the proposed algorithm identifies several possible fault
location candidates. This paper presents a simple and accu-
FIGURE 11 Single-line diagram of the lines of FG under fault incident
rate methodology to distinguish the correct answer. In this
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1314 MOHAJER AND SADEH

25
20
15
10

Vol ta ge (kV)
5
0
-5
-10
-15
-20
FIGURE 12 Single line diagram of a part of the 34-bus IEEE test system -25
in presence of DG in node 3 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30 32.5 35
me (ms)
(a)
TABLE 1 Setting of MATLAB GA toolbox

Population size 40 1.05

0.85
generation 100
0.65
Mutation Gaussian
0.45

Cur r ent (A)

Crossover Scattered
0.25
Crossover fraction 0.8 0.05
-0.15
-0.35

context, the probable fault locations and their resistances are -0.55

first simulated. The voltage samples at the beginning of the -0.75

0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30 32.5 35
(k)
feeder obtained from simulation for the kth answer (Vsimulation ) me (ms)
are then compared with the measured samples (Vreal ) by the (b)
index p as shown in Equation (36). The correct location of the
fault has the minimum index among the answers. FIGURE 13 Voltage and current recorded at the beginning of the feeder
for SLG fault: (a) Voltage (b) Current
n (
∑ )2
(k )
p(k ) = Vreal (i ) − Vsimulation (i ) (36)
i=1

4 SIMULATION RESULT

In order to evaluate the accuracy of the proposed method, a

part of the 34-bus IEEE test system is simulated in PSCAD
software. The total line length of the feeder is 68.5 km and has
7.1 MW load, 12 buses and 6 sub-laterals. Single line diagram of
the system is shown in Figure 12. A DG is installed in node 3
with the nominal capacities of 1.8 MW and is modelled by the
Norton equivalent circuit [28]. The basic parameters of GA in
MATLAB toolbox are presented in Table 1. The voltage and
current recorded at the beginning of the feeder for a SLG fault FIGURE 14 Proposed algorithm accuracy versus fault distance
at line 1 are shown in Figure 13. The proposed scheme is anal-
ysed under various scenarios such as different fault types, fault
distances, fault resistances, and fault inception angles. Moreover, 4.1 Effect of fault distance
the error percentage of the calculated fault distance is deter-
mined via Equation (37). In this part, the effects of fault distance on the accuracy of
the proposed algorithm has been assessed. The simulations are
|xcalculated − xreal | performed for 10 Ω fault resistance, zero fault inception angle,
%Error = × 100 (37)
ltotal and various distances from substation. As shown in Figure 14,
maximum error of the proposed method is for SLG fault and
where, xcalculated is the calculated fault distance, xreal is the real is equal to 0.66% (452 m at the distance of 32,629.092 m from
distance of fault from the beginning of the line, and ltotal is the the beginning of the feeder in L4). Furthermore, according to
total length of the feeder. Figure 14 the proposed algorithm has the desired accuracy for
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOHAJER AND SADEH 1315

TABLE 2 Effect of fault resistance and fault type on accuracy of the proposed method

Fault resistance (Ω) 0 10 30 50

Fault Calculated Calculated Calculated Calculated

type Fault distance(m) %Error resistance (Ω) %Error resistance (Ω) %Error resistance (Ω) %Error resistance (Ω)

3LG 6000 0.03 0.08 0.03 9.92 0.03 29.9 0.03 49.91
11,522.392 0.16 0.11 0.16 10.03 0.16 29.63 0.16 49.86
12,437. 392 0.3 0.00 0.10 9.87 0.10 29.75 0.10 49.58
29,137. 392 0.11 0.00 0.11 9.64 0.11 29.69 0.11 49.67
33,629.092 0.12 0.09 0.12 10.01 0.12 29.89 0.12 49.76
39,988.592 0.32 0.24 0.08 10.18 0.07 29.58 0.08 49.39
44,783. 232 0.02 0.00 0.02 9.88 0.02 29.15 0.02 50.02
48,863. 232 0.02 0.00 0.02 9.79 0.019 29.49 0.02 49.22
50,064. 064 0.18 0.12 0.18 9.95 0.18 29.59 0.18 49.23
50,164. 064 0.07 0.00 0.07 9.95 0.07 30.29 0.07 49.01
50,924. 064 0.008 0.00 0.01 9.81 0.008 29.33 0.41 50.40
52,417. 584 0.24 0.14 0.16 9.63 0.16 29.25 0.16 48.89
54,124. 564 0.14 0.05 0.26 9.47 0.26 28.36 0.26 48.70
SLG 6000 0.03 0.35 0.03 10.39 0.03 30.18 0.03 50.03
11,522.392 0.16 0.12 0.16 9.54 0.16 30.12 0.16 50.7
12,437. 392 0.3 0.00 0.10 9.85 0.10 29.54 0.10 49.26
29,137. 392 0.29 1.09 0.11 10.65 0.11 29.82 0.11 49.08
33,629.092 0.12 0.09 0.52 10.00 0.12 30.48 0.52 50.09
39,988.592 0.32 0.27 0.32 10.01 0.08 30.06 0.08 49.36
44,783. 232 0.02 0.28 0.02 9.679 0.02 29.26 0.02 48.90
48,863. 232 0.02 0.00 0.40 10.08 0.38 30.32 0.38 50.25
50,064. 064 0.18 0.12 0.21 9.49 0.18 29.83 0.18 49.80
50,164. 064 0.07 0.05 0.07 9.75 0.07 30.08 0.07 49.58
50,924. 064 0.39 0.32 0.39 10.02 0.39 29.52 0.39 49.05
52,417. 584 0.16 0.51 0.24 9.85 0.16 29.38 0.16 49.17
54,124. 564 0.14 0.47 0.26 9.78 0.26 29.84 0.26 50.17

the three-phase and LL faults at various fault distances with a

maximum error of 0.26%.

4.2 Effect of fault resistance

Fault resistance is one of the main parameter that can affect the
accuracy of the fault distance estimation. Hence, the accuracy
of the proposed method under different fault resistances has
been evaluated in this subsection. Based on the presented data
in Table 2, it can be inferred that the fault distance estimation
with the proposed scheme in the presence of DG has an out-
standing performance irrespective of the amount of fault resis-
tance. The maximum error for SLG fault is 0.52% for 10 and 50 FIGURE 15 Effect of different fault resistances on the proposed
Ω fault resistances. Furthermore, the maximum error percent- algorithm accuracy for LL fault
age for symmetrical three-phase fault is 0.32%.
The effect of different fault resistance on the proposed algo- imum error is 0.46% for zero fault resistance case, occurred at
rithm for LL fault has been depicted in Figure 15. The error the downstream of the DG and at a distance of 5150 m from
for 3 and 5 Ω fault resistances is quite consistent, and the max- the beginning of the feeder.
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1316 MOHAJER AND SADEH

TABLE 3 Effect of fault inception angle and fault type on accuracy of tion network, the accuracy of the proposed method is also
proposed method increased.
Inception angle (deg) 0 45 90 120

Fault Fault 4.5 E valuation of the proposed method for

type distance (m) %Error
section estimation
3LG 6000 0.03 0.03 0.03 0.03
29,137.392 0.11 0.11 0.11 0.11 In order to evaluate the performance of the proposed method
38,838.592 0.24 0.15 0.16 0.24 for determining the fault section, several case studies have been
simulated (Table 5). At first, all multiple responses of the pro-
52,417.584 0.16 0.16 0.16 0.16
posed algorithm are simulated. Then, the voltage samples in
LL 6000 0.03 0.03 0.03 0.03
the substation obtained by the simulation of the algorithm
29,137.392 0.11 0.11 0.29 0.29 responses are compared with the measured voltage samples
38,838.592 0.16 0.24 0.24 0.24 through index p. As seen in the Table 5, the index value at the
52,417.584 0.16 0.16 0.16 0.24 faulty line is less than those of other answers. This criterion can
SLG 6000 0.03 0.03 0.36 0.03 effectively find the fault section accordingly.
29,137.392 0.11 0.11 0.11 0.11
38,838.592 0.64 0.64 0.64 0.64
52,417.584 0.24 0.16 0.24 0.24 4.6 Comparison with a few existing
methods

In Table 6, the accuracy of the proposed method is compar-

ing with the recently presented methods. According to the pre-
sented results in this Table, it can be seen that the proposed
4.3 Effect of fault inception angle method has minimum error among others, except for one case
that used micro PMUs and a large number of smart meters.
Simulations have been extended for 10 Ω fault resistance, dif-
ferent fault inception angle and various fault types. Accord-
ing to the obtained results in Table 3, the performance of
the proposed algorithm does not rely on the fault inception 5 CONCLUSION
angle.
Although the maximum error of fault location for SLG fault In this paper, a new and accurate fault location method is pro-
is greater than other two fault types, its value is in the acceptable posed for distribution network in the presence of DG, based
range and its maximum is 0.64%. on the distributed-parameter time-domain line model. The pre-
sented scheme uses only two cycles of the post fault measured
samples of voltage and current at the beginning of the feeder
4.4 Effect of multiple DGs and DG terminal. During the first step, the fault distance and
resistance are calculated by applying genetic algorithm. After-
The performance of the proposed method for different num- wards, an index is presented to solve the multi-response prob-
bers of DGs in the distribution network is examined in this lem. It is worth noting that proposed algorithm is also appli-
subsection. As shown in Table 4, with increasing the num- cable for DN without DG and DG type has no effect on this
ber of DG and measuring information in the active distribu- algorithm. Using the modal transformation, in addition to sym-

TABLE 4 Effect of multiple DGs on the accuracy of the proposed method

Number Of DG 0 1 2 3

DG Bus number – 3 3-6 3-6-12

Real parameter

Fault Calculated Calculated Calculated Calculated

Distance (m) Resistance (Ω) %Error Resistance (Ω) %Error Resistance (Ω) %Error Resistance (Ω) %Error Resistance (Ω)

6000 0 0.03 0.00 0.03 0.08 0.03 0.00 0.03 0.05

44,783.323 10 0.32 9.07 0.02 10.00 0.02 10.3 0.02 9.89
54,124.560 50 0.26 50.07 0.26 48.70 0.14 49.55 0.14 50.01
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOHAJER AND SADEH 1317

TABLE 5 Accuracy evaluation of the proposed method for determining the fault section

The real location of fault Answers of proposed method

Correct
Fault type Line number x (m) Line number x (m) Fault resistance (Ω) Index value answer

3LG 1 4000 1 4107.52 9.88 25.3 Line 1

4 547.66 6.46 128.2
9 500 9 547.66 9.74 22.3 Line 9
8 547.66 9.86 37.6
LL 2 385 2 547.66 9.73 5.6 Line 2
4 273.83 9.53 217.7
3 1300 2 1369.15 9.91 97.9 Line 3
3 1369.15 9.80 89.5
SLG 6 3300 6 3285.96 9.65 13.9 Line 6
7 3012.13 9.03 62.8
8 400 8 547.66 9.49 2.3 Line 8
9 273.83 9.70 48.3

TABLE 6 Comparing the proposed method with some presented methods

Ref. Used information Tested network Fault type DG Section estimation Error (m)

[6] reclosure-generating traveling waves. Real DN All No Yes 600

[10] Micro PMUs and Smart Meters 134 bus All Yes Yes 201–230
[11] 14 PMUs 34 bus SLG Yes Yes 452
[24] V and I at the beginning pf the feeder 34 bus SLG No Yes 753
Proposed method V and I at the beginning of the feeder 34 bus All Yes Yes 412.336
and DG locations

metrical faults, the location of asymmetrical faults including LL ORCID

and SLG is also estimated. The suggested technique is assessed Javad Sadeh https://orcid.org/0000-0002-3331-2508
for a part of the 34-bus IEEE test system and the maximum
errors of 0.66%, 0.46% and 0.32% are observed for SLG, LL REFERENCES
and 3LG faults, respectively, that show the superior precision 1. Liao, Y., Kang, N.: Fault-location algorithms without utilizing line param-
of the fault location method. The accuracy of the fault location eters based on the distributed parameter line model. IEEE Trans. Power
doesn’t have notable dependency on fault distance, fault resis- Delivery 24(2), 579–584 (2009)
2. Mora-Florez, J., Melendez, J., Carrillo-Caicedo, G.: Comparison of
tance, and fault inception angle. Thus, the presented algorithm impedance based fault location methods for power distribution systems.
performance is reliable and accurate. Electr. Power Syst. Res. 78(4), 657–666 (2008)
3. Das, R., Sachdev, M., Sidhu, T.: A fault locator for radial subtransmission
and distribution lines. IEEE Power Eng. Soc. Summer Meet. 1, 443–448
FUNDING INFORMATION (2000)
4. Dashti, R., Sadeh, J.: Accuracy improvement of impedance-based fault
location method for power distribution network using distributed-
None parameter line model. Int. Trans. Electr. Energy Syst. 24(3), 318–334
(2014)
CONFLICT OF INTEREST 5. Liang, R., Wang, F., Fu, G., Xue, X., Zhou, R.: A general fault location
No conflict of interest method in complex power grid based on wide-area traveling wave data
acquisition. Int. J. Electr. Power Energy Syst. 83, 213–218 (2016)
6. Shi, S., Zhu, B., Lei, A., Dong, X.: Fault location for radial distribution net-
DATA AVAILABILITY STATEMENT work via topology and reclosure-generating traveling waves. IEEE Trans.
Smart Grid 10(6), 6404–6413 (2019)
Data available on request from the authors: The data that sup-
7. Jiang, Y.: Toward detection of distribution system faulted line sections in
port the findings of this study are available from corresponding real time: A mixed integer linear programming approach. IEEE Trans.
author upon the request. Power Delivery 34(3), 1039–1048 (2019)
 17518695, 2022, 7, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/gtd2.12367 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [29/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1318 MOHAJER AND SADEH

8. Janjic, A., Velimirovic, L.: Integrated fault location and isolation strategy in 20. Aboshady, F., Thomas, D., Sumner, M.: A new single end wideband
distribution networks using Markov decision process. Electr. Power Syst. impedance based fault location scheme for distribution systems. Electr.
Res. 180, 106172 (2020) Power Syst. Res. 173, 263–270 (2019)
9. Silos-Sanchez, A., Villafafila-Robles, R., Lloret-Gallego, P.: Novel fault 21. Rui, L., Nan, P., Zhi, Y., Zare, F.: A novel single-phase-to-earth fault loca-
location algorithm for meshed distribution networks with DERs. Electr. tion method for distribution network based on zero-sequence components
Power Syst. Res. 181, 106182 (2020) distribution characteristics. Int. J. Electr. Power Energy Syst. 102, 11–22
10. Buzo, R.F., Barradas, H.M., Leao, F.B.: A new method for fault location (2018)
in distribution networks based on voltage sag measurements. IEEE Trans. 22. Dashti, R., Sadeh, J.: Applying dynamic load estimation and distributed-
Power Delivery 36(2), 651–662 (2020) parameter line model to enhance the accuracy of impedance-based fault-
11. Zhang, T., Yu, H., Zeng, P., Sun, L., Song, C., Liu, J.: Single phase fault location methods for power distribution networks. Electr. Power Compon.
diagnosis and location in active distribution network using synchronized Syst. 41(14), 1334–1362 (2013)
voltage measurement. Int. J. Electr. Power Energy Syst. 117, 105572 (2020) 23. Gord, E., Dashti, R., Najafi, M., Santos, A.Q., Shaker, H.R.: Determining
12. Dashti, R., Sadeh, J.: Fault section estimation in power distribution network an accurate fault location in electrical energy distribution networks in the
using impedance-based fault distance calculation and frequency spectrum presence of DGs using transient analysis. Measurement 151, 107270 (2020)
analysis. IET Gener. Transm. Distrib. 8(8), 1406–1417 (2014) 24. Daisy, M., Dashti, R.: Single phase fault location in electrical distribution
13. Alamuti, M.M., Nouri, H., Ciric, R.M., Terzija, V.: Intermittent fault loca- feeder using hybrid method. Energy 103, 356–368 (2016)
tion in distribution feeders. IEEE Trans. Power Delivery 27(1), 96–103 25. Sadeh, J., Hadjsaid, N., Ranjbar, A., Feuillet, R.: Accurate fault location
(2012) algorithm for series compensated transmission lines. IEEE Trans. Power
14. Gabr, M.A., Ibrahim, D.K., Ahmed, E.S., Gilany, M.I.: A new impedance- Delivery 15(3), 1027–1033 (2000)
based fault location scheme for overhead unbalanced radial distribution 26. Dommel, H.W.: Digital computer solution of electromagnetic transients in
networks. Electr. Power Syst. Res. 142, 153–162 (2017) single and multiphase. IEEE Networks 88(4), 388–399 (1969)
15. Aggarwal, R., Aslan, Y., Johns, A.: New concept in fault location for over- 27. Salim, R.H., de Oliveira, K.R.C., Filomena, A.D., Resener, M., Bretas, A.S.:
head distribution systems using superimposed components. IEE Proc. Hybrid fault diagnosis scheme implementation for power distribution sys-
Gener. Transm. Distrib. 144(3), 309–316 (1997) tems automation. IEEE Trans. Power Delivery 23(4), 1846–1856 (2008)
16. Das, S., Karnik, N., Santoso, S.: Distribution fault-locating algorithms 28. Jamali, S., Talavat, V.: Dynamic fault location method for distribution net-
using current only. IEEE Trans. Power Delivery 27(3), 1144–1153 (2012) works with distributed generation. Electr. Eng. 92(3), 119–127 (2010)
17. Salim, R., Salim, K., Bretas, A.S.: Further improvements on impedance-
based fault location for power distribution systems. IET Gener. Transm.
Distrib. 5(4), 467–478 (2011)
18. Dashti, R., Ghasemi, M., Daisy, M.: Fault location in power distribu- How to cite this article: Mohajer, F., Sadeh, J.:
tion network with presence of distributed generation resources using Accurate fault location method for distribution network
impedance based method and applying π line model. Energy 159, 344–
in presence of DG using distributed time-domain line
360 (2018)
19. Zhang, S., Lin, S., He, Z., Lee, W.-j..: Ground fault location in radial distri- model. IET Gener. Transm. Distrib. 16, 1305–1318
bution networks involving distributed voltage measurement. IET Gener. (2022). https://doi.org/10.1049/gtd2.12367
Transm. Distrib. 12(4), 987–996 (2018)