You are on page 1of 17

Measurement 220 (2023) 113403

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Fault location in power grids using substation voltage magnitude


differences: A comprehensive technique for transmission lines, distribution
networks, and AC/DC microgrids
Mohammad Daisy a, b, c, Rahman Dashti b, *, Hamid Reza Shaker d, *, Shahram Javadi a, c,
Mahmood Hosseini Aliabadi a
a
Department of Electrical Engineering, Central Tehran Branch, Islamic Azad University, Tehran, Iran
b
Clinical-Laboratory Center of Power System & Protection, Engineering Faculty, Persian Gulf University, Bushehr 7516913817, Iran
c
Intelligent Power System Research Center, Central Tehran Branch, Islamic Azad University, Tehran, Iran
d
SDU Center for Energy Informatics, The Maersk Mc-Kinney Moller Institute, University of Southern Denmark, Odense, Denmark

A R T I C L E I N F O A B S T R A C T

Keywords: Power grids are highly susceptible to various types of faults and their associated consequences. In recent years,
Fault Location numerous fault location methods have been proposed for different types of power networks. Generally, these
Transmission Line methods determine the location of a fault by measuring current and voltage data on one or both sides of the line.
Distribution Network
However, the use of current data can result in calculation errors due to the saturation state of the current
Hybrid Microgrid
Voltage Magnitude Difference
transformer and the bidirectional fault current. Moreover, the use of measuring devices in different nodes can
D-STATCOM lead to increased costs and the need for advanced telecommunication systems and data synchronization. In this
paper, we propose a comprehensive technique for fault location in power networks that incorporates the pres­
ence of D-STATCOM and considers the effect of line capacitors. Our method estimates the distance and faulty
branch by measuring the difference in fault voltage magnitude at the substation and comparing it with simulated
faults in other branches. Unlike other methods that rely on current data, our proposed technique is independent
of current data, resulting in higher accuracy and faster fault detection. Furthermore, our method offers signifi­
cant cost savings compared to other fault location methods. To evaluate the performance of our technique, we
conducted simulations on a 32-node power network in MATLAB/SIMULINK and an 8-node network in a power
system simulator. We tested the sensitivity of our method to various fault locations, resistances, and DG pene­
tration levels. The results of our simulations demonstrate the high accuracy and speed of our proposed technique,
making it a promising alternative to other fault location methods in the field.

the distribution networks. Moreover, the integration of renewable en­


ergy sources in order to reduce greenhouse gas production and the use of
1. Introduction D-FACTS devices to improve reliability, power quality, and stability in
distribution networks are among the challenges of fault location
Power networks contain different parts such as transmission, distri­ methods. Renewable sources can lead to bidirectional current as well as
bution, smart, and microgrids in the form of DC or AC. These networks different levels of fault current in islanded or grid-connected modes. The
are exposed to a variety of transient and sustained faults caused by many fault currents of microgrids are different in islanded and grid-connected
factors such as weather conditions, broken lines, and human factors. modes. The fault current component is also variable depending on the
Fault location at the right time with high accuracy leads to reducing type of source and several times the nominal current. These issues lead
blackouts, increasing reliability, reducing damage to network equip­ to the misclassification of faults or their non-identification. In addition,
ment, and increasing customer satisfaction [1]. if the MG is in islanded mode, the component of the fault current caused
Each power grid has unique characteristics. Distribution networks by the energy sources may not reach the pickup level of the protection
have many branches, imbalance, asymmetry, and different loads. While device for fault detection.
the transmission networks do not have branches and other features of

* Corresponding authors.
E-mail addresses: r.dashti@pgu.ac.ir (R. Dashti), hrsh@mmmi.sdu.dk (H.R. Shaker).

https://doi.org/10.1016/j.measurement.2023.113403
Received 27 March 2023; Received in revised form 6 June 2023; Accepted 30 July 2023
Available online 1 August 2023
0263-2241/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
M. Daisy et al. Measurement 220 (2023) 113403

Nomenclature DVM Difference Voltage Magnitude


ESS Energy Storage System
DG Distributed Generation VSC Voltage Source Control
FL Fault Location Cp Variation of the power coefficient
AC Alternative Current AFD Actual Fault Distance
DC Direct Current DFB Determined Fault Branch
DPM Distributed Parameters Model GC Grid-connected mode
FR Fault Resistance IS Islanded mode
MG Microgrid FT Fault Type
DN Distribution Network Lambda Tip–speed ratio
PCC Point of Common Coupling D-FACTS Distributed Flexible AC Transmission System
CB Circuit Breaker D-STATCOM Distribution Static Synchronous Compensator
V2G Vehicle to Grid Branch-ij The line between nodes i and j
EV Electric Vehicle PMU Phasor Measurement Unit
PV Photovoltaic VMD Voltage Magnitude Difference
WT Wind Turbine GS Synchronous Generator
H Henry K Polarization constant, (V/Ah), or polarization resistance
T Transformer (Ω)
S Substation A Exponential voltage (V)
GPS Global Positioning System R Resistance (Ω)
E0 Constant voltage, (V) B Exponential capacity, (Ah− 1)

On the other hand, the presence of D-STATCOM for reasons such as scheme is presented in AC MGs using advanced μPMUs. This reference
injecting active and reactive power into the network in the face of has used the voltage magnitude and phase data in two terminals and
various disturbances to compensate for the load, reduce voltage flicker compensation theorem and backward-forward voltage. In this algo­
and regulate voltage and frequency, causes changes in the data used in rithm, the short-line model is used and only the faulty branch is calcu­
fault location methods. Therefore, a fault location method should be lated. In [5] a differential protection scheme based on the sensitivity of
able to be implemented in the presence of different types of DGs, the relay is presented for FL in a DC MG. In this reference, the FL is
different network equipment, AC/DC systems, different fault conditions, determined by measuring current and voltage at both sides of the faulty
and operation using the least measuring devices and data [1]. branch. Similar to the previous references, this reference is not eco­
In recent years, various methods have been proposed to locate faults nomic. Besides, the use of current data reduces the accuracy of calcu­
in power networks. According to the characteristics of each network, lations in different conditions. In [6], an estimation fault distance
these methods are used to measure data on one or both sides of the line method is presented using time domain equations in AC smart grids. In
to locate the fault. Current transformer saturation and bidirectional this reference, using the measured data of voltage and current in all
current can miscalculate fault location algorithms that use current data. nodes and calculating the difference of these data at the time of the fault,
On the other hand, methods that use both sides of the line are not the fault distance is obtained. Line parameters have been estimated
economical due to the increase in the number of measuring equipment, using the POS algorithm. In this reference, the distributed parameters
the need for advanced telecommunication systems, and data synchro­ line model is used and the measurement error is considered. The
nization. In general, FL techniques in power networks can be divided maximum error of this method is reported as 1.160 %. In [7], using time
into five groups: impedance-based, traveling waves, artificial intelli­ domain and differential equations, the fault distance for symmetrical
gence, differential equations, and state estimators. These methods can and asymmetrical faults in AC smart grids has been calculated using data
be implemented in all networks. However, due to the difference in the from both sides of the line. In this reference, the effects of the presence of
network topology and the presence of different equipment and re­ D-STATCOM, distributed parameters line model, and high impedance
sources, these methods must be specially optimized. For Example, faults are considered. Among the disadvantages of this reference, it can
impedance-based, differential, and traveling wave methods are able to mention the absence of DGs and the neglect to calculate the faulty
estimate only the fault distance. While in distribution networks and branch. Furthermore, the sensitivity of this method to the sampling rate
microgrids, the faulty branch must be calculated. Besides, with the and high costs are other disadvantages of this method. The maximum
addition of DGs and different equipment to the network, these methods error of this method is reported as 0.31 %. In [8], a method for esti­
need to be improved. In addition, the implementation of these methods mating the fault location in HVDC lines using DC components of current
in AC or DC systems is completely different due to the inherent differ­ and voltage in the time domain is presented. In this reference, the pa­
ences between these two systems. The disadvantages of these methods rameters of the line are estimated using the genetic algorithm. This
are reviewed in [2]. In the following, a review of the papers that method has been calculated by using the measured data of voltage and
implemented the mentioned methods in different power networks current in two terminals and calculating the difference of voltage and
(transmission, distribution, and AC/DC microgrids) has been investi­ current in the times before and after the fault. Among the disadvantages
gated [2]. of this method, it can mention the dependence of the accuracy of the
In [3] a ground FL technique for partially coupled transmission lines algorithm on the sampling rate, telecommunication systems, and data
is presented. This reference utilizes the measured current and voltage synchronizer. The maximum error of this reference is reported as 0.12
phasors at both sides of the line. This reference needs data from two %. In [9], a fault location method in AC smart grids with the presence of
PMUs. The presented method is based on Kirchhoff’s voltage and current DGs is presented. In this reference, the fault distance is calculated using
laws in the zero sequence circuit. It is assumed that the data of PMUs are the improved impedance based method and the faulty branch is calcu­
synchronized by the GPS network. This method is not economical due to lated using least squares. The studied network is assumed observable
the use of data from both sides of the line, telecommunication systems, and μPMU is used in all nodes. The maximum error of this method is
and data synchronization. In [4], an effective data-driven stochastic FL reported as 0.28 %. Using μPMUs in all nodes increases costs.

2
M. Daisy et al. Measurement 220 (2023) 113403

Table 1
Investigation of several approaches in the field of FL in power networks.
Reference [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] Proposed Method
5 9 12 18 20 21
Method GPR IMB SVM - Differential- IMB- Local Analytic- MCE MCS, TW WT Voltage
FFT13 RLS14 LS15 measurement based MLP, Magnitude
NN19 Differences
Network MG DN DN MG DN MG MG MG MG TL TL All
System AC AC AC AC AC DC AC DC DC AC HVAC/ AC/DC
HVDC
Section √ – √ √ √ √ √ – √ – – √
Detection
10
Line model π D D Short D π π π π D D D
Model of Load C6 S11 S C S C C C S – – C
Laterals √ √ √ √ √ √ √ √ √ – – √
Load taps √ √ – √ √ √ √ √ √ – – √
Heterogeneous – √ √ – √ – √ √ – – – √
Unbalanced – √ √ √ √ – √ – √ – – √
Network
Type of Fault SP7/ SP/ SP/M SP/M SP/M PP16-PG17 SP/M PG PP-PG SP/ All All
M8 M M
Presence of DGs √ – √ √ √ √ √ √ √ – – √
KBl √ – √ – – – – – √ – √ –
TD2 √ – – √ – √ – √ √ √ √ √
PD3 – √ – – √ – √ – – – – –
SD4 – – – – √ – √ – – √ – –
l
Knowledge-based.
2
Time Domain.
3
Phasor Domain.
4
Sequence Domain.
5
Gaussian process regression.
6
Constant.
7
Single Phase to Ground.
8
Multi Phase to Ground.
9
Impedace Based.
10
Distributed Parameters Model.
11
Static.
12
Support Vector Machine.
13
Fast Fourier Transform.
14
Recursive Least-Square.
15
Least-Square.
16
Pole to Pole.
17
Pole to Ground.
18
Multiple Capacitive Earthing.
19
Multi-Criterion System and Multi- Layer Perceptron Neural Network.
20
Traveling Wave.
21
Wavelet Technique.

Furthermore, the need for current data reduces the accuracy of the al­ 64,512 data. In addition, changes in load level, penetration of DGs, FR,
gorithm in different conditions. In [10], an FL method in DNs is pre­ and types of fault are considered. Disadvantages of this method include
sented using characteristics of traveling waves, network topology, and the use of a big and offline data bank, the lack of considering different
two-terminal data. In this reference, variational mode decomposition FIA, and variation in line parameters. The maximum error of this
and teager energy operator are used to extract initial fault traveling reference is 4.198 %.
wave fronts, and fault distance difference and fault section difference According to the reviewed papers, it can be seen that the presented
matrices are used. The limitations of this method are the high sampling methods are designed for implementation in a specific network and
rate (2 MHz), dependence on the number of traveling wave recorders in cannot be implemented in all power networks and AC/DC systems.
the nodes, and the absence of DGs. The error of the mentioned method is Table 1 shows the details of several FL approaches in power systems.
reported to be 4600 m. In [11], an FL method in radial DNs using According to Table 1, it can be realized that the presented methods
simultaneous traveling wave detectors at the beginning node and end cannot calculate the fault location in all power networks. Furthermore,
nodes of the network is presented. This method relies on the recording these methods are designed for AC or DC systems. While the proposed
and arrival time of high-frequency traveling waves caused by the fault. method has the ability to calculate the fault location in all power net­
The delay time and the effect of power transformers have been inves­ works and both AC and DC systems. It can also be understood what
tigated. The disadvantages of this method include the absence of DGs, characteristics a fault location method should have. For instance, in
high sampling rate, the need for a large number of traveling wave de­ distribution networks, both the fault section and the fault distance must
tectors, and advanced telecommunications. The accuracy of this method be calculated. Moreover, conditions such as imbalance, heterogeneity,
is reported to be 1.118 %. In [12], an offline FL technique is presented in choosing the appropriate line model, type of load and type of line should
the IEEE 15-bus network with the presence of the photovoltaic system in be considered. In addition, a fault location method in the microgrid
islanded and grid-connected modes and using the data of phase voltage, should be independent of the type of DGs, different operation modes,
current, and cyclic frequency in the nodes. This reference is designed and islanded and grid-connected modes.
using a fault locator module and Gaussian process regression technique. In this paper, by measuring the voltage magnitude difference in the
The training of this module has been done through 3584 simulations and times before and after the fault at the beginning of the faulty network,

3
M. Daisy et al. Measurement 220 (2023) 113403

of the protective devices. The proposed method consists of two main


stages as follows:

Stage 1) fault branch estimation


Stage 2) fault distance estimation

This method has the same formulation in all power networks. In the
following, the equations of the suggested technique are analyzed.

2.1. Fault branch estimation in distribution and AC/DC microgrids

This section is implemented in distribution networks and hybrid


microgrids. Transmission lines do not have branches and do not need to
calculate the faulty branch.
Fig. 1. The voltage magnitude measured at the substation at the time of According to Fig. 1, when a fault occurs, the voltage magnitude
the fault.
changes. The difference between the pre-fault and post-fault values can
be used to indicate these changes. As a result, each fault has a voltage
the distance and branch of the fault in the transmission line, distribution magnitude difference (Δvm). When a fault occurs, the voltage magnitude
network, and AC/DC microgrids with the presence of D-STATCOM and difference is recorded by the measurement at the beginning of the
considering the effect of the line capacitor has been calculated. This network and calculated.
method requires the measured voltage data only at the beginning of the Then, an online data bank is generated by simulating the actual fault
faulty network. As a result, costs are reduced and it does not require (identical type and FR) in all branches and calculating Δvij for each
advanced telecommunication systems and data synchronization. More­ branch. Then, the difference between Δvm and Δvij in all branches is
over, due to not using current data, the saturation state of current determined using Eq. (1) and an index (δ) is calculated for each branch.
transformers and the direction of the fault current do not affect the The numerical difference between the voltage magnitude difference for
performance of this method. On the other hand, the use of voltage the actual fault (Δvm) and the fault voltage magnitude difference
magnitude data at the time of fault occurrence leads to the independence resulting from the simulation in the branch under consideration (Δvij) is
of the performance of the suggested technique against disturbances represented by this difference.
created at the moments of connection of converters, different fault √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
̅

inception angles, and AC/DC systems. Unlike other fault location δij = |Δvm − Δvij |2 (1)
methods, the proposed algorithm has the ability to locate faults in all aεp
power networks. The prominent innovations of this paper are as follows:
Where:
• Ability to locate faults in transmission lines, distribution networks, Δvm: voltage magnitude difference for the actual fault.
and microgrids. ij: number of nodes at the beginning and end of each branch.
• Calculation of both fault distance and faulty section. a: faulty phase.
• Applicable in AC and DC systems. p: faulty phases.
• Independent of current data and transformer saturation condition. δ: index of the faulty branch.
• Use the voltage-measuring device only at the beginning of the If a single-phase fault has occurred, the difference between the
network. voltage magnitude difference for the actual fault and the simulated fault
• No need for telecommunication systems and data synchronization. voltage magnitude difference (Δvm − Δvij), calculates the index (δ). But
• No need for an enormous and offline data bank. in the case of two-phase and three-phase faults, the voltage magnitude
• Insensitivity to the fault type and different topologies of the network. difference for each phase should be calculated separately and then
averaged to get the index (δ) using Eq. (1).
The evaluation of the proposed method has been investigated using Based on Eq. (2), the minimum value of δij is for the faulty branch.
2000 simulations in MATLAB/SIMULINK and a power systems simu­
lator. In these tests, various types of DGs have been used, including WT, CalculatedFaultyBranch = δij minimum (2)
PVs, ESS, and EVs. The obtained results indicate the acceptable accuracy
The value of Δvm and Δvij depends on the fault resistance. The higher the
and speed of this technique in different conditions, including resistances
fault resistance, the smaller the values of Δvm and Δvij. If these values
and different fault locations, and different DG penetration levels.
approach zero, the performance of the algorithm will be disturbed. In
The rest of the paper is as follows: In the second part, the suggested
addition, according to Eq. (1), it can be seen that the fault inception
fault distance and faulty branch methods in power networks are
angle has no effect on AC systems.
described. Simulation results, case studies, and experimental results are
reviewed in the third part. In the fourth part, the comparison of the
suggested technique with other fault location approaches in power 2.2. Fault distance estimation in AC/DC microgrids, transmission and
networks is discussed. Finally, the conclusion is presented in the fifth distribution networks
part.
The fault distance is estimated in this part using the calculated faulty
2. The proposed fault location procedure branch acquired in the previous part. To do so, first, calculate Δvm
related to the actual fault, and then simulate faults similar to the actual
In this technique, it is assumed that all the protection devices are fault with 0.1 km steps from the faulty branch’s beginning. The smaller
coordinated with each other and in case of a fault, only the faulty the selected steps, the higher the accuracy of the algorithm. However, it
network is separated from the other parts of the network. Therefore, the increases the size of the data bank. After various simulations, steps of
fault location algorithm can be implemented in the faulty network. In 0.1 km are chosen due to acceptable accuracy in this method. The
addition, the diagnosis of the type of fault has been done using the data voltage magnitude differences for these simulated faults (Δv1:n) are
calculated and saved. As a result, an online data bank is generated. The

4
M. Daisy et al. Measurement 220 (2023) 113403

Table 2 the proposed method in the transmission lines, this curve has been used
The data bank Structure. as a cubic. The behavior of the voltage magnitude difference in pro­
Distance of fault 0 0.1 0.2 … l portion to the fault distance is nonlinear. In addition, in limited cases,
the cubic curve cannot be fitted on the curve obtained from the simu­
voltage magnitude difference Δv1 Δv2 Δv3 … Δvn
lation of faults. In these cases, higher-degree curves (4th or 5th degree
polynomial) should be used. However, the use of equations of higher
degrees will increase the calculations. As a result, due to the rarity of
these situations, the cubic curve is used in this paper. This behavior can
be described as Eq. (3):

y = a3 x3 a2 x2 + a1 x + a0 (3)

Where:
x: distance of fault from the beginning of the branch.
i: a part of the employed equation.
a0, a1, a2, a3: coefficients of the cubic equation.
When a fault occurs, the coefficients of cubic equation are calculated
using the basic fitting tool (in MATLAB/Software) and simulated faults
in the faulty branch. The voltage magnitude differential related to the
fault distance in a faulty branch is shown in Fig. 2 as a cubic curve.
The voltage magnitude difference equation for the fault distance is
shown in this figure as fΔvm(Δvm,x). This equation includes a voltage
Fig. 2. The cubic curve is based on the voltage magnitude difference relative to
drop for the faulty branch along with two adjacent buses. Eq. (4) is used
the faulty branch.
to obtain the distance of single-phase and PG faults in AC and DC sys­
tems. Likewise Eq. (5) is used to obtain two and three-phase faults in AC
structure of this data bank is shown in Table 2.
systems. According to Eqs. (4) and (5), three answers are obtained by
The first row in Table 2 indicates the various distances with 0.1 km
equating the Δvm with the cubic equation.
steps, and the second row gives the voltage magnitude difference. Be­
( )
sides, l represents the branch length, and n represents the voltage a3 x3 + a2 x2 + a1 x + a0 − Δvm = 0 (4)
magnitude difference for l. This data bank is used to generate a cubic
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
polynomial based on the Δvm relative to the fault distance from the ∑( )
xΔvm = (a3 x3 + a2 x2 + a1 x + a0 − Δvm)p = 0 (5)
beginning of branch (x). The shorter the length of the line, the closer the aε p
curve is to a linear curve. In the same way, the longer the line length, the
curve leaves the linear and goes towards the curves of higher degrees The correct answer should be a positive, real number and less than
(quadratic, cubic, or 4th degree polynomial). Due to the investigation of the length of the branch. Therefore, the fault distance (xΔvm) is obtained,

Fig. 3. The flowchart of the proposed technique.

5
M. Daisy et al. Measurement 220 (2023) 113403

Fig. 4. The studied power network.

which represents the fault distance from the beginning of the branch.
In this method, the location of the fault is calculated from the
measured voltages of the fault at the beginning of the network and the
calculation of the voltage difference in the time before and after the
fault. As a result, the type and operation condition of DGs, the control
modes of converters, the connection type of transformers, and the
islanded and grid-connected modes in microgrids do not have an effect
on the performance of the proposed method.

2.3. The flowchart of the proposed technique

The steps of the proposed technique are as follows:

1. In this technique, network topology, pick-up signals from protection


devices, and data obtained from voltage measurements at the
beginning of each network (transmission, distribution, AC microgrid, Fig. 5. Characteristics of the studied wind turbine.
and DC microgrid) are collected as input to the algorithm.
2. After that, if a fault occurs, voltage magnitude difference for actual 8. The obtained answer indicates the location of the fault determined
fault (Δvm) is calculated. by the proposed technique (xΔvm) and is shown as the output of the
3. Then, the data bank is created online by simulating the similar actual algorithm.
fault in all branches and determining the voltage magnitude differ­
ence (Δvij) for each fault (For distribution network and microgrids). The flowchart of the suggested technique is displayed in Fig. 3.
4. In the next step, for distribution network and microgrids, an index (δ) In the next section, by simulating a power network at the time of
for each branch is calculated using Eq. (1). The minimum value of faults in different modes and conditions, the proposed technique is
this index indicates the faulty branch (The results obtained in this evaluated.
step are shown as the output of the algorithm).
5. To compute the fault distance, faults similar to the actual fault are 3. Simulation verification
simulated in 0.1 km steps from the beginning of the faulty branch
and saved in a data bank. (The transmission line is considered as a The performance of the suggested technique has been evaluated
branch). using various simulations in MATLAB and a power system simulator. In
6. Then, a cubic equation is drawn using this data bank. this part, various tests have been performed to survey the effect of
7. Afterward, three answers (roots) are achieved by equating this cubic different fault locations and resistances and different DG penetration
equation with the Δvm. The root that has the following constraints is levels. In the following, the studied network and case studies are
chosen as fault distance: presented.
• Positive number.
• Real number. 3.1. The power network under study
• Less than branch length.
A 32-node power network with a length of 282.5 km consists of
transmission networks (120 km) [21], distribution (Modified IEEE 15-

6
M. Daisy et al. Measurement 220 (2023) 113403

node) 140 km, an AC 7-node microgrid with a length of 12.4 km, and an
8-node DC microgrid (10.1 km) has been simulated in MATLAB/SIMU­
LINK to evaluate the proposed method. Fig. 4 shows the studied
network.
This network has equipment and resources including a wind turbine
at node 19, photovoltaic systems at nodes 23 and 27, lithium-ion battery
at node 26, electric vehicles at node 22, synchronous generator at node
5, and D-STATCOM at node 6. The resource models are selected from the
MATLAB library. The characteristics and uncertainty of sources are
shown in Figs. 5-8.
Besides, in Fig. 9, the topology of D-STATCOM (voltage source
converter device) under study (±3 Mvar) which consists of two power
switches, a filter, a DC source, and the capacitor is shown. In this paper,
the method [7], which is based on the control theory of the simultaneous
Fig. 6. Active and reactive powers of the studied electric vehicles. reference frame, is applied to D-STATCOM. The frequency of the AC and
DC system is 50 and 60 Hz, respectively, the network voltage in the

Fig. 7. I-V and P-V specifications of PV.

Fig. 8. Characteristics of the studied energy storage system.

7
M. Daisy et al. Measurement 220 (2023) 113403

Fig. 9. D-STATCOM topology under study.

Table 3
Characteristics of the studied PVs.
PV Powers and Voltage Array Data Module Data
Num.
Active Reactive voltage Series-connected Parallel Maximum Number of series- Open-circuit Short-circuit
Power power modules per string strings Power connected cells voltage current

PV1 4 × 100 0 VAR 260 V 5 64 315.072 W 96 64.6 V 6.14 A


kW
PV2 100 kW 0 VAR 500 V 5 66 305.226 W 96 64.2 V 5.96 A

Table 4 Table 7
Characteristics of the studied WT. Line lengths.
Active Reactive voltage Stator Rotor Magnetizing Network Node A-Node B Length (km)
Power power [Rs,Lls] [Rr’, Llr’] inductance Lm
Transmission line 1–2 120
1.5 MW 0var 575 V [ 0.023 [ 0.016 2.9p.u. Distribution network Each branch 10
0.18] p.u. 0.16] p.u. AC microgrid 18–19 2.4
18–20 2.1
20–21 2.7
21–22 1.5
Table 5 20–23 2.1
Characteristics of the studied EVs. 21–24 1.6
DC microgrid 25–26 1
Active Reactive voltage Rated Mode Efficiency
25–27 2
Power power capacity
27–28 1.7
40 × 20 kW 0 var 600 V 85 kWh V2G 90 % 25–29 1.5
29–30 1.2
29–32 1.7
30–31 1
Table 6
Characteristics of the studied ESS.
Battery type Nominal voltage Nominal capacity Primary SOC Table 8
Lithium-ion 120 V 800 A/H 80 % Characteristics of transformers.
Transformer Number VA kV-high V-low X%

transmission line is 220 kV, the distribution network is 20 kV, the AC T1 2 × 108 220 1.575 × 104 12
T2 3 × 106 220 2 × 104 12
microgrid is 11 kV, and the DC microgrid is 600 V. In the distribution
T3 3 × 106 20 11 × 103 12
network, the length of each branch is assumed to be 10 km [12]. In this T4 3 × 106 11 575 12
simulation, the sampling rate is 1e-6 and the π line model is used in all T5 1 × 106 11 260 6
branches. The details of the studied network are shown in Tables 3-9. T6 2 × 106 11 600 6
T7 2 × 106 11 600 12

8
M. Daisy et al. Measurement 220 (2023) 113403

Table 9 obtained using this databank and plotting the obtained points on the x-y
Load characteristics. axes. This curve is converted into a cubic curve using the basic fitting
Network Node No. Active Power (kW) Reactive power (kvar) tool (In MATLAB/Software). This curve is shown in Fig. 12.
Then by equating the value of Δvm with this cubic equation and
AC microgrid 20 135 kW, 1 φ (phase A) 0
21 110 kW, 1 φ (phase C) 55 obtaining the roots of this equation, three answers are obtained. These
24 82 kW, 3 φ 43 answers are 338.0127, 196.0888, and 60.7994. The answer that is
DC microgrid 26 113 kW 0 shorter than the length of the desired line (transmission line), a real and
28 208 kW 0 positive number, indicates the fault distance calculated by the proposed
31 208 kW 0
32 167 kW 0
algorithm. As a result, the answer 60.7994 km is chosen as the fault
distance. The error of the suggested technique in this test is calculated to
be 0.6661 %. To further investigate, several different faults in the
3.2. Performance evaluation simulation transmission network and the results of these tests are shown
in Table 10. (In Eq. (6), the total length of the line is 120 km.) According
This part presents two case studies to demonstrate the performance to the obtained results, the maximum error of the suggested technique is
of the suggested scheme under various fault conditions for the power 1.1 %, which indicates the acceptable performance of this technique in
network in Fig. 4. the transmission line.
In this paper, the error of the suggested technique is calculated using In the following, several P-G faults are simulated in different loca­
Eq. (6). tions and resistances. P-P faults are not investigated in this paper.
⃒ ⃒
⃒Actualx − Determinedx ⃒
Error% = ⃒⃒ ⃒ × 100
⃒ (6)
Totallinelength

Where:
Actualx = Actual fault distance from the beginning of the faulty
branch.
Determinedx = Determined fault distance from the beginning of the
faulty branch using the proposed method.
Total line length = The total length of the desired network (trans­
mission line (120 km), distribution network (140 km), AC (12.4 km), or
DC microgrids (5.3 km)).

3.2.1. Case study 1: Several fault locations and resistances


Different locations and resistances of the fault are important factors
affecting the performance of the fault location methods. For instance,
Fig. 10 shows the cubic curve obtained in branch 1–2 in case of a B-G Fig. 11. The voltage magnitude difference of node-1 at the time of the fault
fault with 0, 25 and 50 Ω fault resistance. According to these figures, it (Transmission line).
can be realized that the obtained cubic curves change with the fault
resistance. In this part, in order to investigate the effect of resistances
and different fault locations on the accuracy and performance of the
suggested technique, several tests are performed on the studied network
and the results are shown in the rest of the paper.
In the first test, an A-G fault is simulated in the transmission line
(branch 1–2) at a distance of 60 km from the beginning of node-1 with a
resistance of 10 Ω. In this step, using the data of the protection device,
the fault type is determined and the fault location algorithm is imple­
mented in the transmission line. As mentioned in the previous parts,
when a fault occurs in the transmission line, only fault distance calcu­
lations are used. Therefore, after the occurrence of the fault, using the
data measured in node-1 and Eq. (4), the voltage difference for the fault
is calculated. The value of this difference is calculated to be 431.7378 V.
Fig. 11 shows this voltage difference.
Then, by simulating faults similar to the actual fault with steps of 0.1
km in the transmission line and calculating the voltage difference for Fig. 12. Cubic curve obtained from the simulation of A-G faults with 10 Ω
these faults, an online data bank is generated. In the next step, a curve is resistance in the transmission line.

Fig. 10. Cubic curves determined in branch 1–2 (a) 0-Ω FR (b) 25-Ω FR (c) 50-Ω FR.

9
M. Daisy et al. Measurement 220 (2023) 113403

Table 10
Simulation results of different fault locations and resistances in the transmission line.
FT1 FR2(Ω) AFD3(km) Δvm Coefficients of the cubic equation Roots Error%

× × √

BG 0.1 30 960.3290 − 0.009377 2.098 − 151.7 3942 96.2055 + 29.9089i 96.2055–29.9089i 31.3280 1.10

AG 10 60 431.7378 − 0.0001765 0.104 − 17.3 1141 338.0127 196.0888 60.7994 0.66


BC 10 80 515.0279 0.001829 − 0.3699 17.18 564.6 125.9632 − 2.7236 79.0021 0.83
ABCG 50 40 167.5800 1.436 × 10^− 05 − 0.01344 2.496 88.25 698.3725 197.5103 40.0503 0.04
1
Fault Type.
2
Fault Resistance.
3
Actual Fault Distance.

The voltage magnitude difference of the simulated faults in all


branches of the DC microgrid using the data obtained from the mea­
surement installed in node-25 are shown in Fig. 14 and Table 11.
According to Table 11, it can be seen that the minimum calculated
index is related to branch 25–27. Consequently, this branch is consid­
ered a faulty branch. Then by simulating the real fault in the faulty
branch with steps of 0.1 km and drawing a cubic curve, a cubic poly­
nomial has been obtained. The coefficients of the cubic polynomial and
the curve obtained from it are shown in Table 12 and Fig. 15, respec­
tively. Then by subtracting this equation from Δvm and setting them
equal to zero, three answers are obtained. These answers are shown in
Table 8. It can be seen that only the answer 1.19 has a real and positive
number and is less than the desired branch length. Consequently, this
answer is considered for the calculated fault distance. The error of the
suggested technique in this test is 10 m.
Fig. 13. Voltage waveform of node-25 at the time of actual fault occurrence To further evaluation, several faults are simulated at different loca­
(DC microgrid). tions with different resistances in the DC microgrid. Tables 11 and 12
display the results of these tests. In addition, Fig. 15 illustrates the curves
Moreover, based on [23–25], the maximum fault resistance in the DC obtained from the 3th degree polynomials obtained in the faulty
system is considered less than 2 Ω. In this part, a P-G fault in branch branches. According to this figure, it can be seen that different fault
27–25 with a resistance of 0.01 Ω and a distance of 1.2 km from node-25 locations and resistances cause changes in the cubic curves. However,
in the DC microgrid is simulated. The voltage difference value of this the performance of the algorithm is not disrupted.
fault is calculated to be 0367.96 V. Fig. 13 shows the voltage waveform According to the results obtained from Tables 7 and 8, the maximum
of node-25 at the time of the fault. error of the suggested technique in the DC microgrid is 0.63 %.

Fig. 14. The voltage magnitude difference of the simulated faults in the DC microgrid.

10
M. Daisy et al. Measurement 220 (2023) 113403

In the following, Tables 13 and 14 of the performance of this tech­

67.0548&28.9819
34.6314&18.8012
nique in the AC-MG and the DN have been examined by simulating

13.1469&1.9595
7.8118&0.6649

2.4324&0.2339
5.4496&0.408
different fault locations with different FR.

Branch 30–31
In this part, to calculate the error of the suggested technique, the
total length of the AC-MG is considered 12.4 km. According to Table 10,
the maximum error of this method in the AC-MG is 1.6370 % and the
minimum error of this method is 0.02 %. It can be seen that the distance
of the faults to the measurement location (PCC1) with different fault
resistances (5, 10, 15, 25, and 50 Ω) had no effect on the performance of

73.1260&22.9107
39.6412&13.7914
13.8148&1.2916
this method.

7.9268&0.5499
5.5299&0.3277
2.6544&0.0119
Branch 29–32
In the following, the evaluation this method in the DN has been
investigated using the simulation of several types of faults in different
locations and resistances of 5, 10, 25, and 50 Ω. Tables 15 and 16 show
the faulty branches and calculated fault distance in these tests, respec­
tively. According to the results of Table 15, it can be seen that with the
increase of FR, the value of Δvij has decreased. In practice, the resistance
of short-circuit faults in the DN is approximately 20 Ω. As a result, ac­
79.3713&16.6654
41.9479&11.4847
14.1628&0.9436
8.0036&0.4731
5.8252&0.0324

cording to Table 16, the performance of this method is acceptable in


2.5513&0.115
Branch 29–30

different fault resistances (0–50 Ω).


This case study investigated the effect of different fault resistances
(0, 10, 25, and 50 Ω in the AC system and 0.01 to 2 Ω in the DC system).
Moreover, the performance of the proposed algorithm has been analyzed
in different fault locations. According to the results and Tables 10, 12,
14, and 16, it can be seen that the error of the proposed method in the
64.2038&31.8329
33.5170&19.9156

transmission line is determined between 0.04 % and 1.1 %, in the DC


13.3374&1.769
8.4255&0.0512
5.4932&0.3644
2.5369&0.1294
Branch 27–28

microgrid is determined between 0.009 % and 0.6 %, in the AC micro­


grid is determined between 0.02 % and 1.6 % and in the distribution
network is determined between 0.006 % and 0.02 %. In addition, the
average error of the proposed technique in the studied power grid is
0.81 %. According to these results, it can be seen that despite the change
in the cubic curves with the change in fault resistance and different fault
locations, the performance and accuracy of the proposed method are
114.1510&18.1143

acceptable. In the following, the performance of the proposed technique


49.7715&3.6611
15.3863&0.2799
9.1024&0.6257
6.4881&0.6305
2.9587&0.2924
Branch 25–29

has been evaluated at different levels of DG penetration.

3.2.2. Case study 2: Different DG penetration levels


Renewable sources have variable production capacity for reasons
Simulation results for different fault branch determination by proposed method (DC microgrid).

such as changes in wind speed, changes in sunlight, etc. These changes


can affect the performance of fault location techniques. In order to
clarify these effects on the performance of the proposed method, the
115.4048&19.3681
54.2234&0.7908
16.3870&1.2806

cubic curves in the case of a C-G fault in the branch 20–21 have been
6.5687&0.7111
3.0040&0.3377
9.5317&1.055
Branch 25–26

compared in the presence and absence of DGs. These simulations are


done in an AC microgrid part. Fig. 16 shows the voltage magnitude in
two cases. In the first case, only PV-1 is present. In the second case, the
WT, EVs, and PV-1 are present. According to this figure, it can be seen
that the presence or absence of DGs can change the magnitude of voltage
difference pre and post-fault.
Fig. 17(a) shows the cubic curve obtained in the presence of EVs, WT
102.4029&6.3662
47.5201&5.9125
16.5086&1.4022
9.5835&1.1068
6.5937&0.7361
3.0096&0.3433

and the PV-1. Fig. 17(b) displays the cubic curve obtained in the pres­
Branch 25–27

ence of the WT and EV. Fig. 17(c) depicts the cubic curve obtained in the
⃒ΔVij ⃒& δ ij

presence of PV-1.

Fig. 17 shows the changes in the cubic curves to the presence or


absence of DGs. Now the effect of these changes on the performance of


the method should be analyzed.
Table 17 illustrates three scenarios considered for DG penetration
96.0367
53.4326
15.1064
8.4767
5.8576
2.6663

levels. In these scenarios, the power produced in DGs is considered


|ΔVm|
Grid-connected and islanded modes

variable between + 50 and − 50 %. The tests of this part are performed


only in AC and DC microgrids. The results of these tests are shown in
Table 18. According to the results obtained in Table 18, the maximum
error of the suggested scheme in this part is calculated to be 0.52 %,
FR Ω

0.01

which corresponds to branch 25–27 with a resistance of 1.5 Ω in the DC-


0.1
0.4
0.7
1
2

MG. Furthermore, the minimum error of the proposed technique in this


part is determined at 0.004 %, which is related to the PG (DC) fault in
branch 29–32 with 2 Ω FR. According to the results obtained in this case
Table 11

25–27
25–26
25–29
27–28
29–30
29–32

study, it can be realized that the DG penetration levels have no effect on


AFB

the performance of this method. Although the presence or absence of DG

11
M. Daisy et al. Measurement 220 (2023) 113403

Table 12
Simulation results for different fault distance determination by the proposed method (DC microgrid).
DFB1 AFD2(km) Coefficients of the cubic equation Roots Error%

× × √

25–27 1.2 –23.53 103.6 − 170.1 191.4 1.6064 + 0.9083i 1.6064–0.9083i 1.1900 0.099
25–26 0.6 − 0.9761 1.104 − 8.091 58.1 0.2656 + 2.8107i 0.2656–2.8107i 0.5999 0.0009
25–29 1.2 − 3.346 7.528 − 5.627 16.95 0.4927 + 0.4393i 0.4927–0.4393i 1.2645 0.6386
27–28 0.3 0.0006423 0.2598 − 0.9381 8.701 − 408.0651 3.3238 0.2575 0.4207
29–30 0.2 − 0.0009932 0.0107 − 0.3298 5.923 5.2868 + 17.378i 5.2868––17.378i 0.1996 0.0039
29–32 0.3 − 8.918 × 10-05 0.002323 − 0.1204 2.702 12.8751 + 34.3021i 12.8751–34.3021i 0.2982 0.0178
1
Determined Fault Branch.
2
Actual Fault Distance.

Fig. 15. Cubic curves obtained in faulty branches (DC microgrid).

Table 13
Simulation results for different fault branch determination by proposed method (AC microgrid).
Grid-connected and islanded modes
⃒ ⃒
FT AFB FRΩ |ΔVm| ⃒ΔVij ⃒& δ ij

Branch 18–19 Branch 18–20 Branch 20–23 Branch 20–21 Branch 21–22 Branch 21–24

BG 18–20 5 1.3203e + 04 1.3249e + 04&46 1.3223e + 04&20 1.3121e + 04&82 1.3023e + 1.2916e + 04&287 1.2927e + 04&276
04&180
CG 20–23 10 1.2792e + 04 1.2860e + 04&68 1.2820e + 04&28 1.2789e + 04&3 1.2738e + 04&54 1.2680e + 04&112 1.2685e + 04&107
AB 20–21 15 6.8633e + 03 6.9155e + 6.8923e + 03&29 6.8692e + 03&5.9 6.8645e + 03&1.2 6.8486e + 6.8492e +
03&52.2 03&14.7 03&14.1
ACG 21–22 25 9.1692e + 03 9.1117e + 9.1370e + 9.1603e + 03&8.9 9.1603e + 03&8.9 9.1699e + 03&0.7 9.1676e + 03&1.6
03&57.5 03&32.2
ABC 21–24 50 5.9725e + 03 5.8841e + 5.9216e + 5.9569e + 5.9643e + 03&8.2 5.9900e + 5.9741e + 03&1.6
03&88.4 03&50.9 03&15.6 03&17.5

Table 14
Simulation results for different fault distance determination by the proposed method (AC microgrid).
DFB AFD (km) Coefficients of the cubic equation Roots Error%

× × √

18–20 1.1 0.7253 − 3.97 –32.13 1.325 × 1004 9.4393 − 5.2687 1.3030 1.6370
20–23 0.9 0.2404 − 3.786 − 25.17 1.282 × 1004 20.5646 − 5.7935 0.9776 0.6258
20–21 2.4 − 0.03482 0.05097 − 11.48 6891 − 0.4664 + 18.2130i − 0.4664––18.2130i 2.3966 0.0274
21–22 1 0.1309 − 1.176 8.413 9162 4.0049 + 6.3584i 4.0049–––6.3584i 0.9741 0.2088
21–24 0.3 0.955 − 2.184 16.72 5968 1.0045 + 3.9925i 1.0045–––3.9925i 0.2780 0.1774

and different penetration levels of DG cause changes in the difference 3.3. Practical validation of the proposed technique
voltage magnitude before and after the fault, the performance of the
suggested algorithm are not disturbed. For practical testing of the proposed technique, an AC power system
simulator has been used. This test was conducted in the power systems

12
M. Daisy et al. Measurement 220 (2023) 113403

Table 15
Simulation results for different fault branch determination by proposed method (DN).
⃒ ⃒ ⃒ ⃒ ⃒ ⃒ ⃒ ⃒
B1 Actual fault ⃒ΔVij ⃒& δ ij Actual fault ⃒ΔVij ⃒& δ ij Actual fault ⃒ΔVij ⃒& δ ij Actual fault ⃒ΔVij ⃒& δ ij

3–4 Fault type = BG 7.2146e + 03&48.3 Fault type = ABG 6.0410e + Fault type = ABC 3.6661e + Fault type = 1.9872e +
Branch = 3–4 Branch = 8–10 03&439 Branch = 5–6 03&155.4 ABCG 03&255.4
4–5 |ΔVm| = 7.1663e 6.7252e + 03&441.1 |ΔVm| = 5.6020e 5.8715e + |ΔVm| = 3.8215e 3.7556e + Branch = 14–15 2.0725e +
+ 03 + 03 03&269.5 + 03 03&65.9 |ΔVm| = 2.2426e 03&170.1
5–6 FR = 5 Ω 5.9518e + FR = 10 Ω 5.5557e + FR = 25 Ω 3.8204e + + 03 2.1687e +
Distance = 5 km 03&1.2145e + 03 Distance = 1 km 03&46.3 Distance = 9 km 03&1.1 FR = 50 Ω 03&73.9
6–7 5.1345e + 5.1424e + 3.8354e + Distance = 3 km 2.2031e +
03&2.0318e + 03 03&459.6 03&13.9 03&39.5
4–8 6.5724e + 03&593.9 5.8224e + 3.7598e + 2.0755e +
03&220.4 03&61.7 03&167.1
8–9 5.6803e + 03&1486 5.4449e + 3.8201e + 2.1506e +
03&157.1 03&1.4 03&92
8–10 5.6739e + 5.5640e + 3.8008e + 2.1227e +
03&1.4924e + 03 03&38 03&20.7 03&119.9
4–11 6.5627e + 03&603.6 5.8353e + 3.7698e + 2.0785e +
03&233.3 03&51.7 03&164.1
11–12 5.6523e + 03&1514 5.4663e + 3.8466e + 2.1606e +
03&135.7 03&25.1 03&82
5–13 5.9469e + 5.5610e + 3.8203e + 2.1452e +
03&1.2194e + 03 03&41 03&1.2 03&97.4
13–14 5.1419e + 5.1480e + 3.8410e + 2.2053e +
03&2.0244e + 03 03&454 03&19.5 03&37.3
14–15 4.5031e + 4.7631e + 3.8036e + 2.2467e +
03&2.6632e + 03 03&838.9 03&17.9 03&4.1
6–16 5.0106e + 5.1419e + 3.8340e + 2.2024e +
03&2.1557e + 03 03&460.1 03&12.5 03&40.2
6–17 5.0179e + 5.1406e + 3.8300e + 2.2005e +
03&2.1484e + 03 03&461.4 03&8.5 03&42.1
1
Branch i-j.

Table 16
Simulation results for different fault distance determination by the proposed method (DN).
DFB AFD (km) Coefficients of the cubic equation Roots Error%

× × √

3–4 5 0.186 − 4.626 − 18.01 7349 27.1063 − 7.2403 5.0050 0.0142


8–10 1 0.01608 − 0.3691 − 37.04 5639 60.4375 − 38.4731 0.9896 0.0074
5–6 9 0.00259 − 0.3494 6.434 3790 114.0586 11.8354 9.0094 0.0067
14–15 3 0.008989 − 0.1702 4.917 2229 7.9504 + 20.8701i 7.9504––20.8701i 3.0334 0.0239

loads. The length of each branch is 15 km. In addition, the conductors of


branches 5–6 and 6–7 are of type ACSR-226 and other of the branches
are of type ACSR-118.5. Fig. 19 displays the implementation of the
studied network on this simulator. The details of the loads and DG are
shown in Table 19.

3.3.2. Experimental testing and results


In this test, a two-phase fault to ground with FR 30 Ω occurred at a
distance of 7.5 km in branch 2–3. In the first step, the voltage data are
collected at node-1 using METREL-MI 2892. Its sampling rate is set to
48000 Hz. Then, the measuring device sends the data to MATLAB and
their curves are plotted. Afterward, the difference voltage magnitude is
calculated for the actual fault (5.5021e + 03 V) and the online databank
is generated by simulating a similar fault in all branches. Subsequently,
seven indexes (δ) are calculated for all branches. Table 20 indicates
Fig. 16. Voltage magnitude in branch 20–21. these indexes.
According to Table 20, the minimum value of δ is 3.7, which is
laboratory at Persian Gulf University. related to branch 2–3. As a result, branch 2–3 is selected as faulty
branch. In the next step, faults with actual fault characteristics (identical
3.3.1. Experimental setup type and resistance) in branch 2–3 with steps of 0.1 km are simulated,
In this part, an 8-bus network with a total line length of 105 km is and the voltage magnitude difference are calculated for all faults. Then,
simulated on an AC power system simulator. Fig. 18 illustrated the the points obtained using the basic fitting tool and a cubic equation are
single-line model of this network. plotted on the x–y axis. Fig. 20 depicts this curve and cubic equation.
The system’s frequency and voltage are 50 kHz and 20 kV respec­ The obtained equation is set equal to the Δvm and the value of x is
tively. This simulator is made of five parts consist the grid cabinet, DG, calculated. The answers obtained are 64.8796, 29.1961, and 7.5087. An
substation, distribution and transmission lines, and constant and static answer that is less than the length of the branch, real and positive is the

13
M. Daisy et al. Measurement 220 (2023) 113403

Fig. 17. Changes in the cube curve in the presence and absence of DGs in the branch 20–21.

fault branches in the transmission line, distribution network, and hybrid


Table 17
microgrids using the difference voltage magnitude at the beginning of
Different scenarios of DG Penetration Levels.
the grids. Recently presented fault location methods are designed for
Scenario Wind Photovoltaic Photovoltaic Electric ESS specific networks. Therefore, the proposed method should be compared
Turbine 1 2 Vehicle
separately with the methods presented in different networks. Most of the
First − 10% +15% +5% − 20% +6% FL techniques in power networks have used the measurement of voltage
Second − 30% − 20% +10% +3% − 10% and current data in the first node and buses with DG. While the proposed
Third +8% − 50% − 40% +50% − 25%
technique is calculated the location of the fault using measuring only the
voltage magnitude at the beginning of the grids. The proposed technique
correct answer. As a result, the answer 7.5087 is chosen as the calculated only requires voltage data. Whereas other similar methods require
fault distance in km. The error of the proposed technique in this test is voltage and current data. As a result, changes in the value and direction
8.7 m, which indicates the high accuracy of this technique. of the fault current and the saturation condition of the transformer have
no effect on the performance of this technique. In addition, other typical
4. Comparison with other fault location methods methods compared in Table 21 are designed for AC or DC systems. While
this approach is applicable in both AC and DC systems. The advantages
The focus of this technique is on the calculation of both distance and of this method compared to other methods include high speed,

Table 18
Simulation results of different DG Penetration Levels.
Grid-connected and islanded modes

Scenario Fault Type FR Ω Actual Fault Branch Calculated Fault Branch Actual Fault Distance (Km) Calculated Fault Distance (Km) Error%

First CG 10 21–24 21–24 0.4 0.3467 0.4298


Second ABG 30 20–23 20–23 1.7 1.6971 0.0233
Third ABC 50 20–21 20–21 0.6 0.5935 0.0524
First PG 1 25–26 25–26 0.6 0.5989 0.0108
Second PG 1.5 25–27 25–27 1.3 1.3534 0.5287
Third PG 2 29–32 29–32 1.15 1.1505 0.0049

Fig. 18. Single line model of the studied network.

14
M. Daisy et al. Measurement 220 (2023) 113403

0.81 %, which indicates the acceptable accuracy of this method


compared to other approaches.

5. Conclusion

In this paper, we proposed a comprehensive method for accurately


determining fault distance and location in transmission lines, distribu­
tion networks, and AC/DC microgrids. An online data bank is generated
(at the time of the fault) by measuring the voltage magnitude difference
at the beginning of the network and simulating faults in other branches.
Consequently, this technique is not sensitive to different topologies, does
not require updating data in case of changes in the network and massive
databank. By comparing the actual fault voltage magnitude difference
with the simulated faults, the faulty branch is identified. Then the
minimum value between the actual and simulated fault voltage differ­
ences is used to calculate the fault distance using a cubic equation. This
approach takes into account the effect of line capacitors and the pres­
ence of D-STATCOM, and is independent of current data, which avoids
errors resulting from the saturation conditions of current transformers
Fig. 19. Implementation of the studied network on the power sys­
and bidirectional fault currents. The proposed technique only requires a
tem simulator.
voltage-measuring transformer at the beginning of the network, making
it simple and cost-effective to implement. The suggested method was
simplicity, and the use of π line model. In addition, insensitivity to the
evaluated using 2000 simulations in MATLAB and a power systems
presence of renewable resources, D-FACTS, electric vehicles, different
simulator. The performance of this method was investigated in different
FL, FR, and different DG Penetration Levels can be mentioned as other
fault resistances (0, 10, 25, and 50 Ω in the AC system and 0.01 to 2 Ω in
advantages of this technique. Compared to the impedance-based and
the DC system), different fault locations, different penetration levels of
traveling wave methods, this technique is independent of the fault
DGs (±50 percentage), and uncertainty in sources. The error percentage
inception angles. In addition, unlike differential equations and traveling
of the proposed method for different fault resistance in the transmission
wave methods that depend on measuring devices at both sides of the
line between 0.04 % and 1.1 %, in the DC microgrid between 0.009 %
line, advanced communication systems, and data synchronization, this
and 0.6 %, in the AC microgrid between 0.02 % and 1.6 %, and in the
method requires only one measuring device at the beginning of the
distribution network between 0.006 % and 0.02 % are determined.
network. This method is independent of building a big data bank, which
Besides, the maximum and minimum error of this algorithm has been
is one of the disadvantages of intelligent methods. On the other hand,
some of the main limitations are as follows:

• In the part of fault branch estimation, the cubic curve obtained from
the simulation of different faults does not fit on the main curve of the
simulation faults. This happens in some three-phase faults. In order
to solve this problem, a curve with higher degrees should be used.
• The algorithm’s accuracy decreases if a fault occurs with high re­
sistances (>100 Ω) and the fault voltage difference decreases (close
to zero).
• If the measurement device at the beginning of the network has an
error, the algorithm’s accuracy will decrease.

Table 21 displays the comparison results of the proposed technique


with other papers. In addition, Fig. 21 compares the accuracy of the
proposed method with other similar methods in power networks. Ac­
cording to the results obtained in the Table 21 and Fig. 21, the average
error of the proposed technique in the estimation of the fault distance is Fig. 20. Cubic equation and obtained curve in branch 2–3.

Table 19
Loads and DG characteristics.
Node 1 2 3 4 5 6 7 8

Load – – Static 400 kVA, 0.85 lag – – Constant, 500 kVA, 0.8 lag Constant, 350 kVA, 0.8 lag –
(1φ) (3φ) (1φ)
Asynchronous Machine – – – – 0.5 kW, 2.2 A, 380 V – – –
(DG)

Table 20
Calculated difference fault voltage magnitude and indexes for all branches.
Branch 1–2 2–3 2–4 4–5 5–6 6–7 7–8
⃒ ⃒
⃒Δvij ⃒ 5.3732e + 03 5.4984e + 03 5.4712e + 03 5.4308e + 03 5.2891e + 03 5.1159e + 03 4.9501e + 03
⃒ ⃒
⃒δij ⃒ 128.9 3.7 30.9 71.3 213 386.2 552

15
M. Daisy et al. Measurement 220 (2023) 113403

Table 21
Comparison of the proposed technique with other existing FL methods in power networks.
Reference [6] [12] [4] [26] [27] [18] [28] [7] [29] [30] [17] [19] Proposed
Method

Measurement All All S3 + last S+ S+ S+ S + DERs + All S+ All All FTEC4 S


Nodes buses DERs DERs DERs Loads DERs
Network Smart MG MG MG DN MG Active DN Smart MG MG MG MG Power
DN DN networks
System AC AC AC AC AC AC AC AC DC DC DC DC AC/DC
Required I,V I,V I,V I,V I,V I,V I,V I,V I,V I,V I,V I,V V
parameters
DG Penetration √ √ √ √ – √ – – √ – – – √
Levels
Presence of EVs – – – – – – – – – – – – √
D-FACTS – – – – – – – √ – – – – √
Modes – IS1- IS-GC GC – GC IS-GC – GC GC GC GC IS/GC
GC2
Fault section – √ √ √ √ √ √ – √ √ √ – √
estimation
Fault Distance 1.16 4.198 – 5.75 8.5 1.38 3 0.318 14.36 25 2.45 2 Avg. = 0.81
Error% m
1
Islanded mode.
2
Grid-connected mode.
3
Substation.
4
The Feeder Terminals and Earthing Capacitors.

Fig. 21. Comparison of the accuracy of the suggested scheme with other similar methods in power networks. (See above-mentioned references for further
information.)

obtained in different DG penetration levels in the DC microgrid 0.5287 penetration level, unbalanced network, fault type, heterogeneous, line
% and 0.0049 % and in the AC microgrid 0.4298 % and 0.0233 % and load model, and type of DGs. Furthermore, unlike other similar
respectively. After various tests, the maximum error of the proposed methods, current data is not required. In future research, this method
method in transmission lines, distribution networks, AC microgrids, and can be extended to investigate ring networks, different types of micro­
DC microgrids is 1.1 %, 0.023 %, 1.63 %, and 0.63 %, respectively, in grids (3-Net), and high impedance faults. Overall, the proposed method
islanded and grid-connected modes. Furthermore, the average error of represents a significant contribution to the field of fault location in
the proposed method in power grids is calculated as 0.81 % in simula­ power networks and microgrids and has the potential to enhance the
tion and 8.7 m in laboratory tests. The results demonstrate the high reliability of power systems and the security of energy supply.
accuracy and speed of the proposed method, as well as its economic
benefits compared to other existing methods. In addition, the suggested
approach is not sensitive to fault resistance, fault locations, different

16
M. Daisy et al. Measurement 220 (2023) 113403

CRediT authorship contribution statement [13] R. Dashti, M. Daisy, H.R. Shaker, M. Tahavori, Impedance-based fault location
method for four-wire power distribution networks, IEEE Access 6 (2017)
1342–1349.
Mohammad Daisy: Conceptualization, Methodology, Data curation, [14] H. Mirshekali, R. Dashti, A. Keshavarz, H.R. Shaker, Machine learning-based fault
Formal analysis, Writing – original draft, Resources, Validation, Visu­ location for smart distribution networks equipped with micro-PMU, Sensors 22 (3)
alization. Rahman Dashti: Conceptualization, Methodology, Writing – (2022) 945.
[15] T. Gush, S.B.A. Bukhari, R. Haider, S. Admasie, Y.-S. Oh, G.-J. Cho, C.-H. Kim, Fault
original draft, Resources, Supervision, Validation, Project administra­ detection and location in a microgrid using mathematical morphology and
tion. Hamid Reza Shaker: Conceptualization, Methodology, Project recursive least square methods, Int. J. Electr. Power Energy Syst. 102 (2018)
administration, Supervision, Validation, Writing – review & editing. 324–331.
[16] H. Mirshekali, R. Dashti, A. Keshavarz, A.J. Torabi, H.R. Shaker, A novel fault
Shahram Javadi: Conceptualization, Methodology, Supervision, location methodology for smart distribution networks, IEEE Trans. Smart Grid 12
Writing – review & editing. Mahmood Hosseini Aliabadi: Conceptu­ (2) (2021) 1277–1288.
alization, Methodology, Supervision, Writing – review & editing. [17] Y. Yang, C. Huang, D. Zhou, Y. Li, Fault detection and location in multi-terminal
DC microgrid based on local measurement, Electr. Pow. Syst. Res. 194 (2021),
107047.
Declaration of Competing Interest [18] A. Bretas, C. Orozco-Henao, J. Marín-Quintero, O. Montoya, W. Gil-González,
N. Bretas, Microgrids physics model-based fault location formulation: Analytic-
based distributed energy resources effect compensation, Electr. Pow. Syst. Res. 195
The authors declare that they have no known competing financial (2021), 107178.
interests or personal relationships that could have appeared to influence [19] A. Makkieh, V. Psaras, R. Peña-Alzola, D. Tzelepis, A.A. Emhemed, G.M. Burt, Fault
the work reported in this paper. location in DC microgrids based on a multiple capacitive earthing scheme, IEEE J.
Emerg. Select. Topics Power Electron. 9 (3) (2020) 2550–2559.
[20] A. Abdali, K. Mazlumi, R. Noroozian, High-speed fault detection and location in DC
Data availability microgrids systems using multi-criterion system and neural network, Appl. Soft
Comput. 79 (2019) 341–353.
Data will be made available on request. [21] O. Naidu, A.K. Pradhan, A traveling wave-based fault location method using
unsynchronized current measurements, IEEE Trans. Power Delivery 34 (2) (2018)
505–513.
References [22] M. Fayazi, M. Joorabian, A. Saffarian, M. Monadi, A single-ended traveling wave
based fault location method using DWT in hybrid parallel HVAC/HVDC overhead
[1] S. Hossain-McKenzie, E.C. Piesciorovsky, M.J. Reno, J.C. Hambrick, Microgrid transmission lines on the same tower, Electr. Pow. Syst. Res. 220 (2023), 109302.
Fault Location: Challenges and Solutions, Sandia National Laboratories, [23] S. Dhar, R.K. Patnaik, P. Dash, Fault detection and location of photovoltaic based
Albuquerque, NM, USA, 2018. DC microgrid using differential protection strategy, IEEE Trans. Smart Grid 9 (5)
[2] R. Dashti, M. Daisy, H. Mirshekali, H.R. Shaker, M.H. Aliabadi, A survey of fault (2017) 4303–4312.
prediction and location methods in electrical energy distribution networks, [24] A. Meghwani, S.C. Srivastava, S. Chakrabarti, Local measurement-based technique
Measurement 184 (2021), 109947. for estimating fault location in multi-source DC microgrids, IET Gener. Transm.
[3] F. Abbasi, A.A. Abdoos, S.M. Hosseini, M. Sanaye-Pasand, New ground fault Distrib. 12 (13) (2018) 3305–3313.
location approach for partially coupled transmission lines, Electr. Pow. Syst. Res. [25] R. Mohanty, A.K. Pradhan, DC ring bus microgrid protection using the oscillation
216 (2023), 109054. frequency and transient power, IEEE Syst. J. 13 (1) (2018) 875–884.
[4] A. Kavousi-Fard, S. Nikkhah, M. Pourbehzadi, M. Dabbaghjamanesh, A. Farughian, [26] A. Srivastava, S. Parida, A robust fault detection and location prediction module
IoT-based data-driven fault allocation in microgrids using advanced µPMUs, Ad using support vector machine and gaussian process regression for AC microgrid,
Hoc Netw. 119 (2021), 102520. IEEE Trans. Ind. Appl. 58 (1) (2021) 930–939.
[5] P. Chauhan, C. Gupta, M. Tripathy, High speed fault detection and localization [27] M. Gholami, Detecting the location of short-circuit faults in active distribution
scheme for low voltage DC microgrid, Int. J. Electr. Power Energy Syst. 146 (2023), network using PMU-based state estimation, IEEE Trans. Smart Grid 11 (2) (2019)
108712. 1396–1406.
[6] H. Mirshekali, R. Dashti, H.R. Shaker, R. Samsami, A.J. Torabi, Linear and [28] P.K. Ganivada, P. Jena, A fault location identification technique for active
nonlinear fault location in smart distribution network under line parameter distribution system, IEEE Trans. Ind. Inf. 18 (5) (2021) 3000–3010.
uncertainty, IEEE Trans. Ind. Inf. 17 (12) (2021) 8308–8318. [29] R. Bhargav, B.R. Bhalja, C.P. Gupta, Algorithm for fault detection and localisation
[7] A. Fathy, R. Dashti, M. Najafi, H.R. Shaker, Transient and steady-state faults in a mesh-type bipolar DC microgrid network, IET Gener. Transm. Distrib. 13 (15)
location in intelligent distribution networks compensated with D-STATCOM using (2019) 3311–3322.
time-domain equations and distributed line model, Electr. Eng. 103 (2021) [30] K. Saleh, A. Hooshyar, E.F. El-Saadany, Fault detection and location in medium-
3033–3048. voltage DC microgrids using travelling-wave reflections, IET Renew. Power Gener.
[8] M. Daisy, R. Dashti, H.R. Shaker, A new fault-location method for HVDC 14 (4) (2020) 571–579.
transmission-line based on DC components of voltage and current under line [31] R.L. Reis, F.V. Lopes, Correlation-based single-ended traveling wave fault location
parameter uncertainty, Electr. Eng. 99 (2) (2017) 573–582. methods: A key settings parametric sensitivity analysis, Electr. Pow. Syst. Res. 213
[9] H. Mirshekali, R. Dashti, A. Keshavarz, A.J. Torabi, H.R. Shaker, A novel fault (2022), 108363.
location methodology for smart distribution networks, IEEE Trans. Smart Grid 12 [32] Y. Yang, Q. Zhang, M. Wang, X. Wang, E. Qi, Fault location method of multi-
(2) (2020) 1277–1288. terminal transmission line based on fault branch judgment matrix, Appl. Sci. 13 (2)
[10] L. Xie, L. Luo, Y. Li, Y. Zhang, Y. Cao, A traveling wave-based fault location method (2023) 1174.
employing VMD-TEO for distribution network, IEEE Trans. Power Delivery 35 (4) [33] T. Takagi, Y. Yamakoshi, M. Yamaura, R. Kondow, T. Matsushima, Development of
(2019) 1987–1998. a new type fault locator using the one-terminal voltage and current data, IEEE
[11] A. Tashakkori, P.J. Wolfs, S. Islam, A. Abu-Siada, Fault location on radial Trans. Power Syst. 8 (1982) 2892–2898.
distribution networks via distributed synchronized traveling wave detectors, IEEE [34] S. Das, S. Santoso, A. Gaikwad, M. Patel, Impedance-based fault location in
Trans. Power Deliv. 35 (3) (2019) 1553–1562. transmission networks: theory and application, IEEE Access 2 (2014) 537–557.
[12] A. Srivastava, S. Parida, Data driven approach for fault detection and Gaussian
process regression based location prognosis in smart AC microgrid, Electr. Pow.
Syst. Res. 208 (2022), 107889.

17

You might also like