www.nature.

com/scientificreports

OPEN Transmission line faults detection

and classification using new
tripping characteristics based on
statistical coherence for current
measurements
R. A. Mahmoud
Power transmission lines are critical components of a power system that connect power stations to
consumers. To maintain reliability and stability of the system, faults should be correctly classified and
cleared as soon as possible. In this article, a coherence-based protection scheme for faults detection
and classification on transmission lines (TLs) is proposed. Besides, the scheme introduces a new model
of tripping characteristics based on six coherence coefficients that are computed only for current
waves measured at the TL sending end. The power network under test is simulated using the ATP
software, and signals analysis and the performance evaluation of the technique are performed in the
MATLAB environment. The protection performance is investigated under different fault conditions,
such as fault type, fault location, fault resistance, fault inception angle and power flow angle. The
extensive simulation cases have demonstrated that the suggested technique is successful in detecting
and classifying all ten shunt faults in the transmission line within a half-cycle time. Moreover, the
protection security, sensitivity, and response speed are amended by changing the numerical values of
the coherence setting and data window. Furthermore, it is applicable in both conventional and smart
grids, and it is independent of the specifications of the system equipment and current transformers.

Keywords Power transmission line, Shunt faults, Series faults, Fault detection, Fault classification, Coherence
method

Abbreviations
ia(n), ib(n) and ic(n) The three-phase current waves at sample ‘n’ measured at the TL sending end of phases
A, B and C respectively
is(n) The current signal for every sample n of phase ’S’
ix(n) The current signal for every sample n of phase ’X’
is(n-Ns) The instantaneous phase current signal (is) at sample one-cycle prior to n of ’S phase’
n The n’th sample (in the time domain)
S and X The phase designation A, B or C; but they are not the same phase
Cis(k) The auto-coherence factor, on a given frequency (k), calculated between each two suc-
cessive data windows shifted from each other by one-cycle interval of the current signal
(is(n) and is(n - Ns)) of the ‘S’ phase
Cisx(k) he cross-coherence factor, on a given frequency (k), calculated between the two current
waves (is(n) and ix(n)) for the two phases ‘S’ and ‘X’, respectively
DFT Discrete fourier transform
RMS The root mean square
Δx The coherence setting deviation; it lies within the range of 0.0 and + 0.25
K The k’th frequency component
Ns The number of samples per cycle used in the simulation, (Ns = 100 samples/cycle)
N The number of samples per data window used in the technique (N ≤ Ns); it is select-

Department of Electrical Power and Machines Engineering (PME), College of Engineering Science and
Technology, Misr University for Science and Technology (MUST), 6th of October City, Giza, Egypt. email:
mohandes_ragab@yahoo.com; ragab.mahmoud@must.edu.eg

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 1

www.nature.com/scientificreports/

ed = Ns/2
Tc The periodic cycle, (Tc = 20 mSec)
Fc The fundamental frequency of one periodic cycle, (Fc = 50 Hz)
Ts The sampling time interval, (Ts = 0.2 mSec)
Fs The sampling frequency rate, (Fs = 5 kHz)
ω The angular velocity of the power system (ω = 2 πf)
Rf The fault resistance imposed from the faulted point on SG stator winding to the ground
point in the case of the ground fault, or inserted between the two faulted phases in case
of the phase fault.
tf The fault inception time
Nf The sample index at which the fault occurs
tc The fault clearing time
Nc The sample index at which the fault clears
Nsim The full number of samples per the simulation time.
Ө The fault inception angle
α The power system angle
SLGF Single line-to-ground fault
DLGF Double line-to-ground fault
DLF Double line fault
3LGF Three line-to-ground fault
ATP Alternative transient program
SG Synchronous generator
TL Transmission line
Rn Generator grounding impedance through the neutral point
Vn The nominal voltage of the power transmission line
In The nominal current of the power transmission line
SLD Single line diagram
PF Power factor
FD Fault detection
FL Fault location
BB Busbar
CB Circuit breaker
CT Current transformer
CTR Current transformer ratio
Rb The current transformer burden (it is 1 Ω)
RCT The current transformer secondary winding resistance
Rlead The lead resistance connected between the current transformer terminals and the
burden
V1Max The peak phase voltage of synchronous generator
V2Max The peak phase voltage of the electrical network
F1op The operating frequency of synchronous generator
F2op The operating frequency of the electrical network
δ1 The operating power angle of synchronous generator
δ2 The operating power angle of the electrical network

Power transmission lines (TLs) are vital elements of power systems because they link generating stations to
consumers in order to achieve a power supply continuity1. Moreover, it connects between the systems for
bidirectional power flow. Large blocks of electrical power can be economically transferred between these systems
with high-voltage TLs2. It is important to have a protection system against TL faults in the electrical networks
because most faults occur on TL3. TL faults can be divided into three types: shunt unsymmetrical, shunt
symmetrical and series faults4. There are three forms of the unsymmetrical shunt faults (i.e., SLG, DL, and DLG),
and one form of the symmetrical shunt faults (i.e., 3 L or 3LG)5. Such transient faults can cause overvoltage and/
or overcurrent, which can damage insulation materials and conductors of the power system depending upon
their severity6. Furthermore, they contain useful information that can be used for analyzing TL disturbances7. It
is essential to detect and locate the faults quickly and to classify them correctly in order to maintain the system
stability and reliability, and to provide high energy quality8. TL protective relays always utilize three-phase
currents, voltages or both to perform the task of fault detection, faulty phase(s) identification, classification,
and location9. The protective relays send a tripping control signal in order to TL CB(s) to isolate the faulted TL
from the rest of the system10. Traditional distance relays are the primary protection of TLs, which estimate the
impedance between the relay and the fault locations, and respond to the distance to the TL fault11. The relay speed,
reliability and accuracy of the fault detection and classification are important protection requirements for the
lines. With the applications of signals processing and modern techniques, a TL protection development for fault
diagnosis has been progressed12. Various techniques of fault detection, fault classification, fault location, faulty
phase(s) selection, and fault direction discrimination were approached in several articles. Some techniques used
synchro-phasors1, Travelling Waves (TW)2,3, least error squares4, sequential components5,6, a combination of
neural network and PMU measurements7, a combination of Wavelet and alienation8, genetic algorithm9, signal
processing10, and statistical methods such as alienation11, correlation12, and phasor measurement unit (PMU)13.
Moreover, other methods were based on the S-transform14, Discrete Fourier Transform (DFT)15, voltage
Travelling Waves (TW)16, Wavelet Packet Transform (WPT)17, Support Vector Machine (SVM)18, machine

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 2

www.nature.com/scientificreports/

learning19,20, fuzzy logic21, Decision Tree (DT)22, differential protection based on an incremental complex power
alpha plane23, and differential protection based on Continuous Wavelet Transform (CWT)24.
Several techniques for TL fault identification were also presented, including single-ended travelling wave25,
travelling wave theory26, decision tree aided travelling wave27, voltage response28, transmission line topology,
non-unit transient29, source impedance30, and the pilot impedance supported by synchronized data31. The
technique in32 suggested an integrated method for discriminating internal and external faults and locating the
faults on the TL compensated with a unified power flow controller. The paper33 introduced a digital protection
based on entropy principle, fast discrete orthogonal S-transform, support vector machine, and support vector
regression for faults detection, classification, and location on hybrid TL consisting of Overhead Transmission
Line (OHTL) and Underground Cable (UGC) sections. The method reported in34 was dependent on a Gaussian
Naïve Bayes approach for TL fault classification using phasor measurements. A modification of state estimation
formulation was used in35 for the fault location, selection of faulty line, identification of faulted phase(s), and
fault classification in the power network over a wide-area. TL fault determination, classification, and location
were also verified using travelling wave frequencies and extreme learning machine36, wide area measurement37,
Gabor transform38, and polarity and arrival time of asynchronously recorded travelling wave39.
Additionally, several transmission line protection solutions were recently published, including those based on
distances between current samples using the Hausdorff Distance40, Euclidean Distance41, Canberra Distance42,
and Dynamic Time Warping43. In addition, other protections were contingent on statistical techniques using
the Pearson Correlation44, Spearman Correlation45, Kendall Tau Correlation46, Biweight Mid-correlation47, Dice
Coefficient48, Bayesian Inference49, Structural Similarity50, Cosine Similarity51,52, and an impedance estimation–
based protection using Lissajous Fig53.
In this paper, a fault detection, classification, and faulty phase(s) identification algorithm is provided as a
secondary protection for TL primary protections using a coherence function. This algorithm relies on three
auto-coherence and three cross-coherence coefficients to identify the fault occurrence, discriminate the faulty
phase(s), and classify the ten types of shunt faults. Three-phase current waves are used at only one end to
process the proposed protection algorithm. Extensive simulation cases are conducted on a typical power system
to verify the algorithm’s performance. All fault types, locations, inception angles and fault impedance will be
included in this study. The results demonstrate that the proposed algorithm provides efficient discrimination
between all fault types within a sub-cycle time, increases protection requirements such as robustness, security
and dependability, and makes an accurate decision when faced with the faults. In the present manuscript, the
contributions to knowledge are given as follows:

(1) A new protection scheme is proposed to detect, select the faulty phase(s), and classify the different types of
faults using the auto- and cross-coherence functions computed for only current waves measured at the TL
sending end,
(2) The digital protection scheme has used two integrated modules to detect and classify the different TL faults,
as well as to select the faulty phase(s),
(3) Only two types of mathematical equations are applied for employing the protection algorithm. For the cur-
rent signals, three auto-coherence factors and three cross-coherence factors can be calculated using these
equations.
(4) The method introduces new models of tripping curves based on the auto- and cross-coherence coefficients
quantified for three-phase current signals. It can be used to isolate the faulty system, while keeping the
power grid safe during the normal operation,
(5) The approach can control the operating time required to identify the fault quickly by changing the selected
data set. One-cycle or sub-cycle time can be chosen, and.
(6) Only one coherence setting can be selected to discriminate between the normal and abnormal states, as well
as to achieve a compromise between the relay sensitivity and security levels.

The article is organized as follows: In Sect. 2, the coherence-based protection algorithm is discussed. In Sect. 3,
a typical power system model with a list of real data parameters is demonstrated. Section 4 presents the results
of various case studies, as well as an analysis of the technique performance for fault detection and classification.
Advanced technique properties and a comparison between the proposed and published protection techniques
are demonstrated in Sect. 5. Ultimately, the conclusions are drawn in Sect. 6.

Protection methodology
Coherence application justification
The proposed TL protection is based on recent advances in signal processing and numerical analysis for discrete
values of the three-phase currents measured at only one end of the protected transmission line. It is capable of
detecting the fault occurrence, selecting the faulty phase(s) and classifying the different types of the TL faults
using the coherence function. There are two types of coherence: auto-coherence and cross-coherence54–56.
The coherence factors in this study are divided into three auto-coherence factors and three cross-coherence
factors55–57.
The coherence function possesses the following attributes:

(1) Creating new, square-shaped relay tripping curves that are bounded between the two real values of 0.0 and
+ 1.0,
(2) Distinguishing between the healthy and faulty states for the power TLs using both auto-coherence and
cross-coherence functions. The auto-coherence factors obtained for the three-phase currents are constant
and equal to approximately + 1.0 in the normal operating conditions, while they are unstable and unequal

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 3

www.nature.com/scientificreports/

in fault states. Also, the cross-coherence factor calculated for each two phase currents is fixed and close to
+ 0.25 in the healthy conditions, while the three cross-coherence factors are upset and unlike in fault events,
(3) Selecting the faulty phase(s), and classifying the ten types of shunt faults using the auto-coherence function.
The relationship is strong and stable when the auto-coherence factor computed for each phase current is
nearly + 1.0, while it is weak and unstable when the auto-coherence factor is otherwise less than + 1.0, and.
(4) Differentiating between the DL and DLG faults using the cross-coherence function. The cross-coherence
factor calculated for the two currents of the faulty phases is equal to about + 1.0 in the case of the DL fault,
while its value is disturbed and unequal to unity in the case of the DLG fault.

Coherence mathematical formulas

Auto-coherence function
The auto-coherence coefficient (Cis(k)) can be found between two data sets that are shifted by one-cycle time of
the current wave for ‘S’ phase. This coefficient can be quantified using the following mathematical formula for
each phase current58,59:
[( )2 (N −1 )2 ]
∑
N −1 ∑
Is1 (k) × Is3 (k) + Is2 (k) × Is4 (k) + Is1 (k) × Is4 (k) − Is2 (k) × Is3 (k)
n=0 n=0
Cis (k) = (1)
∑ [
N −1
2 2
] ∑ [
N −1
2 2
]
(Is1 (k)) + (Is2 (k)) × (Is3 (k)) + (Is4 (k))
n=0 n=0

Where,

∑[
N −1 (
2πkn
)]
Is1 (k) = is (n) · cos
N
n=0

∑[
N −1 (
2πkn
)]
Is2 (k) = is (n) · sin
N
n=0

∑[
N −1 (
2πkn
)]
Is3 (k) = is (n − Ns ) · cos
N
n=0

∑[
N −1 (
2πkn
)]
Is4 (k) = is (n − Ns ) · sin
N
n=0

Is1(k) Cosine component of the DFT for wave is(n),

Is2(k) Sine component of the DFT for wave is(n),

Is3(k) Cosine component of the DFT for wave is(n-Ns), and.

Is4(k) Sine component of the DFT for wave is(n-Ns).

Equation (1) can be used to get the three auto-coherence coefficients: Cia(k), Cib(k), and Cic(k).

Cross-coherence function
The cross-coherence coefficient (Cisx(k)) can be determined between each two corresponding data sets of the
two current waves (is(n) and ix(n)) for the ‘S’ and ‘X’ phases. The following equation can be used to calculate the
coefficient for each two phase currents57–59:
[( )2 (N −1 )2 ]
∑
N −1 ∑
Is1 (k) × Ix1 (k) + Is2 (k) × Ix2 (k) + Is1 (k) × Ix2 (k) − Is2 (k) × Ix1 (k)
n=0 n=0
Cisx (k) = (2)
∑ [
N −1
2 2
] ∑ [
N −1
2 2
]
(Is1 (k)) + (Is2 (k)) × (Ix1 (k)) + (Ix2 (k))
n=0 n=0

Where,

∑[
N −1 (
2πkn
)]
Ix1 (k) = ix (n) · cos
N
n=0

∑[
N −1 (
2πkn
)]
Ix2 (k) = ix (n) · sin
N
n=0

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 4

www.nature.com/scientificreports/

Fig. 1. (a) Flow chart of the proposed method for fault detection. (b) Flow chart of the proposed method for
fault classification. (c) Flow chart of the proposed method for fault classification (Continued).

Ix1(k) Cosine component of the DFT for wave ix(n), and.

Ix2(k) Sine component of the DFT for wave ix(n).

Equation (2) can be used to obtain the three cross-coherence coefficients: Ciab(k), Cibc(k), and Cica.(k). The
concept of changing the amount of data window can be applied to the present algorithm. This study uses a pre-
determined fixed value for the data window, which is selected to be a half-cycle time of the fundamental power
frequency in the system.

Proposed algorithm strategy

Two modules of the proposed protection methodology are sequentially processed as follows:

(1) Module 1: It can be used to detect the fault occurrence, discriminate the faulty phase(s), and classify the
fault types using the auto-coherence coefficients of the current waves, and.
(2) Module 2: It can be used to ensure the fault presence, and distinguish between the DL and DLG faults using
the cross-coherence factors of the current waves.

Figure 1a presents the first flow chart of the proposed method for detecting fault events. Whereas, Fig. 1b, c show
the second flow chart of the proposed method for classifying the ten shunt faults on the power transmission line.
The proposed algorithm employs both auto-coherence and cross-coherence functions.

Fault detection
There are several variables that change when the TL fault occurs, including electrical voltage, current, power
factor, impedance, and frequency. In addition, the parameters of power quality, such as magnitude, frequency,
symmetry, and shape of the measured waveforms may be disturbed. The proposed technique can be used to
determine the fault occurrence in three scenarios:

(I) The estimation of the auto-coherence factor for each phase current,
(II) The post-fault auto-coherence value of the phase current is compared with the pre-fault value for the same
phase current, and.
(III) The values of the three auto-coherence factors are compared over the same time span.

In this algorithm, a fault transition can be detected if 0 ≤ Cia˂ (1- Δx), 0 ≤ Cib˂ (1- Δx), or 0 ≤ Cic˂ (1- Δx);
where, the selected value of the coherence setting margin (Δx) is + 0.15.

Fault confirmation
To confirm that the fault is present, the following conditions should be satisfied:

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 5

www.nature.com/scientificreports/

Figure 1. (continued)

(0.25 + Δx) < Ciab< (0.25- Δx), (0.25 + Δx) < Cibc< (0.25- Δx), or (0.25 + Δx) < Cica.< (0.25- Δx).

Fault classification
After detecting the fault, fault classification and the faulty phase(s) discrimination can be accomplished according
to the rules tabulated in (Table 1).

Algorithm procedure
The sequence of the algorithm is as follows:

(1) Read the instantaneous values of the three-phase currents (is(n)) measured at the TL sending end,
(2) Choose the numerical values for the data window area (N) and the coherence margin (Δx),
(3) Quantify the three auto-coherence factors (Cia, Cib, and Cic) using Eq. (1).
(4) Differentiate between the healthy and the faulty states according to the conditions given in Table 1, and the
flow chart depicted in (Fig. 1a). When the values of the three auto-coherence factors are within the range of
+ 1.0 and 1.0 - Δx, then the system status is healthy, while if the values of at least one factor are outside this
range, then the system status is faulty.
(5) Calculate the three cross-coherence factors (Ciab, Cibc, and Cica.) using Eq. (2).
(6) Check the power system status is whether normal or abnormal according to the conditions in Table 1 and
the flow chart shown in (Fig. 1a). If the values of the three cross-coherence factors are within the range of
0.25 ± Δx, then the system status is healthy, while if the values of at least one factor are outside this range,
then the system status is faulty.
(7) Discriminate the phases that are faulty, and classify the TL fault type according to the conditions given in
(Table 1) and the flow chart depicted in (Fig. 1b).
(8) The protection action is based on the conditions noted in (Table 1).

There are six coherence factors that can recognize fault conditions in the system, discriminate the faulty phase(s),
classify the fault kind, and operate in response to it. The proposed algorithm monitors fault events continuously

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 6

www.nature.com/scientificreports/

Figure 1. (continued)

and sends a trip control signal to the TL CB(s) when there is a short-circuit current, while it will be held under
normal operating conditions.
This scheme should take into account the following concerns:

(I) It is recommended that the setting margin (Δx) should be in between the two values of 0.0 and + 0.25,
(II) The number of samples per data set (N) should be lower than or equal to the number of samples per cycle
(Ns). This allows the proposed method to respond rapidly to detect short-circuit currents, and
(III) Changes to the protection security, sensitivity, and response speed can be made via the data set (N) and the
coherence margin (Δx).

In this work, the auto-coherence function has the following roles:

(I) Detecting the fault situation,

(II) Selecting the faulty phase(s), and
(III) Classifying the fault type,

Whereas, the cross-coherence function has the following tasks:

(I) Confirming the fault event,

(II) Identifying and estimating the three-phase current imbalance, and
(III) Discriminating between DL and DLG faults.

When the auto-coherence factors assure that two phases are faulted, and the cross-coherence factor estimated
between the two current waves of the two faulty phases is nearly + 1.0, the fault type is a DL fault. Otherwise, it
is DLG.

Square tripping characteristics

The proposed technique can actually complete its operation and trip TL circuit breaker(s) in the event of
fault detection, based on a tripping zone of the coherence characteristic involved. Furthermore, the length of
time it takes for the fault locus to traverse the tripping zone. The coherence factors range from 0.0 to + 1.0, as
mentioned before. In Fig. 2a–c, new closed-tripping curves with square forms are presented. All tripping curves
are restricted and known because they depend on the coherence coefficients. As shown in (Fig. 2a–c), each

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 7

www.nature.com/scientificreports/

TL fault detection and Technique action

Required signals Coherence coefficients range classification (tripping/blocking)
Module type The selected numerical value of the coherence margin (Δx) is 0.15
Module (1) for detecting the
fault occurrence, selecting the
1 ≥ Cia≥ 1- Δx,
faulty phase(s), and classifying
ia(n) and ia(n - Ns) 1 ≥ Cib≥ 1- Δx and Normal operation Blocking
the fault types using the auto-
1 ≥ Cic≥ 1- Δx
coherence coefficients for the
current waves
ib(n) and ib(n - Ns)
ic(n) and ic(n - Ns)
0 ≤ Cia <1- Δx,
ia(n) and ia(n - Ns) 0 ≤ Cib <1- Δx or Fault detection Tripping
0 ≤ Cic <1- Δx
ib(n) and ib(n - Ns)
ic(n) and ic(n - Ns)
0 ≤ Cia <1- Δx,
0 ≤ Cib <1- Δx, 3LG fault (A-B-C-G) or 3 L fault
ia(n) and ia(n - Ns) Tripping
0 ≤ Cic <1- Δx and (A-B-C)
Cia ≈ Cib ≈ Cic
ib(n) and ib(n - Ns)
ic(n) and ic(n - Ns)
0 ≤ Cia <1- Δx,
ia(n) and ia(n - Ns) 1 ≥ Cib≥ 1- Δx and SLG fault (A-G) Tripping
1 ≥ Cic≥ 1- Δx
ib(n) and ib(n - Ns)
ic(n) and ic(n - Ns)
1 ≥ Cia≥ 1- Δx,
ia(n) and ia(n - Ns) 0 ≤ Cib <1- Δx and SLG fault (B-G) Tripping
1 ≥ Cic≥ 1- Δx
ib(n) and ib(n - Ns)
ic(n) and ic(n - Ns)
1 ≥ Cia≥ 1- Δx,
ia(n) and ia(n - Ns) 1 ≥ Cib≥ 1- Δx and SLG fault (C-G) Tripping
0 ≤ Cic <1- Δx
ib(n) and ib(n - Ns)
ic(n) and ic(n - Ns)
0 ≤ Cia <1- Δx, Cia ≈ Cib and
ia(n) and ia(n - Ns) 0 ≤ Cib <1- Δx and Ciab = 1 Tripping
1 ≥ Cic≥ 1- Δx DL (A-B) fault
Cia ≠ Cib and
ib(n) and ib(n - Ns) Ciab≠ 1
DLG (A-B-G) fault
ic(n) and ic(n - Ns)
1 ≥ Cia≥ 1- Δx, Cib ≈ Cic and
ia(n) and ia(n - Ns) 0 ≤ Cib <1- Δx and Cibc = 1 Tripping
0 ≤ Cic <1- Δx DL (B-C) fault
Cib ≠ Cic and
ib(n) and ib(n - Ns) Cibc≠ 1
DLG (B-C-G) fault
ic(n) and ic(n - Ns)
0 ≤ Cia <1- Δx, Cia ≈ Cic and
ia(n) and ia(n - Ns) 1 ≥ Cib≥ 1- Δx and Cica. = 1 Tripping
0 ≤ Cic <1- Δx DL (A-C) fault
Cia ≠ Cic and
ib(n) and ib(n - Ns) Cica.≠ 1
DLG (A-C-G) fault
ic(n) and ic(n - Ns)
Module (2) for confirming
the fault occurrence, and
0.25 + Δx ≥ Ciab≥ 0.25- Δx,
differentiating between the
ia(n) and ib(n) 0.25 + Δx ≥ Cibc≥ 0.25- Δx and Normal operation Blocking
DL and DLG faults using the
0.25 + Δx ≥ Cica.≥ 0.25- Δx
cross-coherence factors for the
current waves
ib(n) and ic(n)
ic(n) and ia(n)
0.25 + Δx ˂ Ciab < 0.25- Δx,
ia(n) and ib(n) 0.25 + Δx ˂ Cibc < 0.25- Δx or Fault detection Tripping
0.25 + Δx ˂ Cica. < 0.25- Δx
ib(n) and ic(n)
Continued

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 8

www.nature.com/scientificreports/

TL fault detection and Technique action

Required signals Coherence coefficients range classification (tripping/blocking)
Module type The selected numerical value of the coherence margin (Δx) is 0.15
ic(n) and ia(n)
ia(n) and ib(n) Ciab = 1 DL (A-B) fault Tripping
ib(n) and ic(n) Cibc = 1 DL (B-C) fault
ic(n) and ia(n) Cica. = 1 DL (C-A) fault
ia(n) and ib(n) Ciab≠ 1 DLG (A-B-G) fault Tripping
ib(n) and ic(n) Cibc≠ 1 DLG (B-C-G) fault
ic(n) and ia(n) Cica.≠ 1 DLG (C-A-G) fault

Table 1. TL shunt faults detection and classification based on the six coherence factors computed for the
current waves and the technique action.

Fig. 2. (a–c) Tripping characteristic curves based on the coherence coefficients. (a) Tripping characteristic
curve based on auto-coherence coefficients. (b) Tripping characteristic curve based on both auto-coherence
and cross-coherence coefficients. (c) Tripping characteristic curves based on cross-coherence coefficients.

characteristic has two areas: a tripping area used to isolate TL CBs in the fault situations, and a blocking area
used to prevent tripping TL CBs during the normal operating conditions.
Figure 2a depicts the relay operating characteristic curve based on the pair of auto-coherence factors, (Cis and
Cix). As illustrated in Fig. 2a, if the operating points of one of the three auto-coherence coefficients fall within at
least one tripping zone of the auto-coherence-based characteristic curves, then a fault situation is present. From
Fig. 2a, the proposed characteristic curve is able to carry out the following tasks: (1) monitoring and detecting
the fault conditions, (2) selecting the faulty phase(s), and (3) classifying the shunt fault type.

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 9

www.nature.com/scientificreports/

Figure 2b shows the relay operating characteristic curve based on the pair of coherence factors, (Cis and Cisx).
As demonstrated in Fig. 2b, the power transmission system is subject to both fault and unbalanced currents
when the operating points of both auto-coherence and cross-coherence coefficients are inside the tripping zone
of this curve. From Fig. 2b, the proposed characteristic curve has the ability to perform the following roles: (1)
confirming the presence of the power quality disturbances (PQDs), and (2) differentiating between DL and DLG
faults.
Figure 2c presents the relay operating characteristic curves based on the three pairs of cross-coherence
factors: (Ciab and Cibc), (Cibc and Cica.), and (Cica. and Ciab). As shown in Fig. 2c, if the operating points of one of
the three cross-coherence coefficients are situated within at least one tripping zone of the cross-coherence-based
characteristic curves, then a current imbalance instance is figured out. From Fig. 2c, the proposed characteristic
curves can execute the following functions: (1) assuring the abnormal conditions, and (2) finding and evaluating
the three-phase current unbalance.
The TL CBs receives a tripping signal once the protection scheme senses the fault/imbalance situation. If
all six coherence coefficients are present within the blocking zones of the relay characteristics, there is no fault
nor unbalanced currents condition. As a consequence, the algorithm tripping signal will be active if the fault/
imbalance instance is present. Otherwise, it is inactive.

Power model under test

To study a power system model under different operating and fault conditions, the ATP platform is used. In
Fig. 3, a single line diagram of the system is shown. The system consists of one synchronous generator (SG)
with a step-up transformer (SUT) linked to busbar 1 (BB1) and two transmission lines (TL1 and TL2) that are
connected to busbar 2 (BB2) of a unified power network. Additionally, there are load 1 linked at the SG terminals,
load 2 located at busbar 2 (BB2), and eight circuit breakers (CBs) distributed, as shown in (Fig. 3). BB1 is the TL1
sending end, and BB2 is the TL1 receiving end. Appendix 160 contains the real data of parameters for the different
system components. This study will concentrate on the first transmission line (TL1) in order to test the protection
based on the coherence function. Therefore, three-phase currents should be measured using protection current
transformers installed at the TL1 sending end.

Results analysis
To investigate the accuracy and reliability of the algorithm, the power network under test is simulated using the
ATP platform, and signals processing and analysis are employed in MATLAB/M-file environment. The current
waveforms, which are measured at the TL1 sending end, are used in a discrete form for signal processing and
numerical analysis. The sampling frequency of the signals is 5 kHz, which yields 100 samples per cycle of the
signal. In this study, the full simulation time is 0.2 s. (i.e., the total number of samples, Nsim = 1000), the fault
inception instant (ti) is 0.05 s. (i.e., the fault occurs at the sample order of 250), the fault clearing instant (tc) is
0.2 s, and the shunt faults are existing at 50% of TL1 long. To validate the technique performance, the system
and fault conditions are changed. This will be accomplished by utilizing different scenarios such as fault type,
fault location, fault resistance, fault inception angle and power flow angle. In this work, it is supposed that the
electrical power network is loaded before the fault occurrence, and that the pre-fault operating conditions of the
simulated power system are tabulated in (Table 2). In Table 3, the fault conditions for each case are given.
Below, there are a few assumptions that are taken into consideration in order to improve the algorithm
performance.

(1) The same type of three-phase current transformers installed to protect the TL,
(2) The same CTRs for all current transformers in the three-phase circuit, and
(3) All secondary currents are transformed without core saturation for at least a quarter-cycle starting from the
fault inception.

This paper demonstrates only five case studies. The pre-fault results of the proposed scheme are involved in
(Table 4).

Fig. 3. One line diagram of the power system model.

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 10

www.nature.com/scientificreports/

Electrical component (operating condition) Data

Operating peak phase voltage (V1Max) of synchronous generator 16.063 kV
Operating peak phase voltage (V2Max) of electrical power network 408.24 kV
F1op of synchronous generator 50 Hz
F2op of electrical power network 50 Hz
Synchronous generator operating power angle (δ1) 10o
Electrical power network operating power angle (δ2) 0o
Generator grounding impedance 0.77 Ω
Electrical load 1 10.85 + j 6.72 Ω
Electrical load 2 8.5 + j 5.26 Ω
CB11, CB12, CB13, CB21, CB22 CB23, CB24 and CB33 Status Close

Table 2. Pre-fault operating conditions of the simulated power system.

CT
burden,
fault inception time, tf Fault Resistance, Rf Current transformer Rb (in
Case number Fault type Fault location (in Sec) (in Ω) ratio (CTR) Ω)
Case 1 Series fault F1 (located at 50% of TL1) 0.05 Not applicable 1000/1 0.5 + j0.0
SLG shunt fault )A-G) through Rf
Case 2 500
= 500 Ω
Case 3 DLG shunt fault )A-B-G) 0.0
Case 4 DL fault (A-B) fault 0.0
Case 5 3LG shunt fault )A-B-C-G) 0.0

Table 3. Fault conditions of the case studies.

Parameter Ia(Amp) Ib(Amp) Ic(Amp)

Peak phase primary value 117 117 117
Peak phase secondary value 0.1165 0.1165 0.1165
Auto-coherence factor Cia Cib Cic
+ 1.0 + 1.0 + 1.0
Cross-coherence factor Ciab Cibc Cica.
+ 0.25 + 0.25 + 0.25

Table 4. Pre-fault results of the proposed protection.

Case 1: series fault

In this case, there is a series fault due to a single pole open condition of the three-phase circuit breaker (CB13).
In the case of an open “A” pole for CB13, unbalanced currents flow in the three-phase system. The series fault
initiates at a time (ti) of 0.05 s. from the beginning of the simulation time. Figure 4a illustrates three-phase
current waves in the event of the series fault. The figure reveals that the magnitudes of the three-phase current
before the fault are normal and balanced for the three-phase system. During the series fault time, the current
values of the open “A” phase are zero, while the currents of the two healthy phases (“B” and “C”) are unbalanced,
and their amplitudes are lower than the pre-fault current. Figure 4b presents the three auto-coherence factors
(Cia, Cib and Cic) estimated for the three-phase current waves. It is observed that their values are stable, equal,
and close to unity (i.e., Cia = Cib = Cic = + 1.0) before the series fault presence. But the three coefficients are
uneven and unstable during the first two data windows of the fault time, as shown in (Fig. 4b). During the first
two data sets, it is obvious that the values of Cib ≈ Cic ≈ 0.95, while the Cia values are less than + 0.3. This indicates
that there is a fault on the phase “A” of TL1. Three cross-coherence factors (Ciab, Cibc and Cica.) are shown in
(Fig. 4c). In this case, it is noticed that their values are balanced and close to + 0.25 before the event of series
fault, as shown in (Fig. 4c), while their post-fault values become unsymmetrical and unequal. Three tripping
characteristics with square shapes are illustrated in Fig. 4d for case 1. Each characteristic consists of one auto-
coherence factor and another cross-coherence factor. In this case, it can be evident that the operating points are
located within the tripping zone of each characteristic. It is therefore possible to conclude that the values of the
six coherence factors are capable of detecting the series fault event, and selecting the faulty phase.

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 11

www.nature.com/scientificreports/

Fig. 4. (a–d) Simulation results for case 1. (a) Three-phase secondary currents taken at the TL1 sending end.
(b) Coherence factors (Cia, Cib and Cic). (c) Coherence factors (Ciab, Cibc and Cica.). (d) Tripping characteristics
based on the coherence factors.

Fig. 5. (a–d) Simulation results for case 2. (a) Three-phase secondary currents taken at the TL1 sending end.
(b) Coherence factors (Cia, Cib and Cic). (c) Coherence factors (Ciab, Cibc and Cica.). (d) Tripping characteristics
based on the coherence factors.

Case 2: SLG shunt fault with Rf = 500 Ω

The shunt fault is located at the midpoint of TL1 in this case. The fault type is SLG (A-G) with a fault resistance
(Rf ) of 500 Ω, and the fault starts at a time (ti) of 0.05 s. The curves of three-phase secondary currents taken at
the TL1 sending end for case 2 are presented in (Fig. 5a). In the figure, the magnitudes of the pre-fault secondary
currents are almost constant. The SLG fault increases the “A” phase current, which is approximately 2.75 times
higher than the current before the fault. Figure 5b depicts the three auto-coherence factors (Cia, Cib and Cic)
computed for the three-phase currents (ia, ib and ic), respectively. In this figure, it is clear that the three factors
(Cia, Cib and Cic) are equal and close to unity before the fault happens, while Cia≈ 0.84, Cib ≈ 0.87 and Cic≈ 0.86
during the first two data windows of the fault time, as shown in (Fig. 5b). This denotes the fault occurrence on
“A” phase of TL1 because Cia is lower than the coherence setting of 1.0 – Δx = 0.85. Figure 5c illustrates the three

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 12

www.nature.com/scientificreports/

cross-coherence factors (Ciab, Cibc, and Cica.). It is seen that their values are fixed and close to + 0.25 before the
fault presence, while their values are asymmetrical, unequal and outside the domain of 0.25 ± Δx during the fault
time, as depicted in (Fig. 5c). Figure 5d offers the three tripping characteristics with quadratic forms for case 2.
One auto-coherence factor and another cross-coherence factor are used to determine each characteristic. It is
noted that the operating points of the three coherence factors (Cia, Ciab, and Cica.) are placed within the tripping
zones of the developed characteristics. This is due to the fault location on the “A” phase of the TL1. This event
shows that the suggested technique is sensitive because of its ability to recognize the short-circuit current state
through the high fault resistance of 500 Ω. Besides, it can choose the faulty phase that is the “A” phase of the TL1,
and classify the fault type that is SLG (A-G).

Case 3: DLG shunt fault with Rf = 0.0 Ω

In this case, the fault type is changed to bolted DLG fault (A-B-G). The three-phase currents measured at the TL1
sending end are shown in (Fig. 6a). The two faulty phases (A and B) have current magnitudes that are nearly
40 times larger than the pre-fault current, while the current value of the healthy phase (C) is nearly the same
as the pre-fault current. Figure 6b depicts the three auto-coherence factors (Cia, Cib, and Cic). Their values are
equilibrium and equal to almost unity before the fault starts. When starting the fault, the values of Cia and Cib
are lower than 0.2, while the values of Cic are greater than 0.85. These values indicate that the A and B phases
are faulty, but the C phase is healthy. Figure 6c plots the three cross-coherence factors (Ciab, Cibc, and Cica.).
Before the fault inception, their values are identical and equal to nearly + 0.25, while their values are out of the
range restricted from 0.1 to 0.4 during the fault period. Therefore, this confirms that the TL1 state is defective.
Figure 6d depicts the three tripping characteristics for case 3. It is evident that the values of the five coherence
factors (Cia, Cib, Ciab, Cibc, and Cica.) are located within the tripping zones of the developed characteristics. This
means that the fault classification is DLG, and it is present in the ‘A and B’ phases of the TL1. It is obvious from
the obtained results, the three auto-coherence factors (Cia, Cib, and Cic) can define the DLG fault. Furthermore,
the three cross-coherence factors (Ciab, Cibc, and Cica.) can be used to assure the fault occurrence, and the fault
type is DLG. It is not a DL fault since the post-fault values of the factor Ciab are below 0.8 (i.e., the factor Ciab
computed between the two faulty phases is not equal to unity). It is observed from the results, the proposed
algorithm (based on the six coherence factors calculated for the current waves) is capable of finding out the fault
situation, discriminating the faulty phases that are “A” and “B” phases of TL1, and classifying the fault type that
is DLG (A-B-G).

Case 4: DL fault with Rf = 0.0 Ω

The fault type in this case is a bolted DL fault (A-B) free of grounding. Figure 7a presents the curves for the
three-phase currents transformed at the TL1 sending end. During the fault period, the current magnitudes of
the two faulty phases are identical, but their directions are different. The post-fault currents of the two faulty
phases (A and B) are nearly 38 times the pre-fault current, while the current magnitude of the healthy phase (C)
is nearly the same as the pre-fault current. Figure 7b introduces the three auto-coherence coefficients (Cia, Cib
and Cic). Before the fault inception, their values are stable, equal and close to unity. When the fault occurs, the
values of Cia and Cib are less than 0.15, while the values of Cic are close to unity. Thus, these values affirm that the
A and B phases are faulty, but the C phase is healthy. Figure 7c illustrates the three cross-coherence coefficients
(Ciab, Cibc, and Cica.). Their pre-fault values are symmetrical and equal to about + 0.25. In this case, it is observed

Fig. 6. (a–d) Simulation results for case 3. (a) Three-phase secondary currents taken at the TL1 sending end.
(b) Coherence factors (Cia, Cib and Cic). (c) Coherence factors (Ciab, Cibc and Cica.). (d) Tripping characteristics
based on the coherence factors.

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 13

www.nature.com/scientificreports/

Fig. 7. (a–d) Simulation results for case 4. (a) Three-phase secondary currents taken at the TL1 sending end.
(b) Coherence factors (Cia, Cib and Cic). (c) Coherence factors (Ciab, Cibc and Cica.). (d) Tripping characteristics
based on the coherence factors.

that Cia ≈ Cib, Cic ≈ +1.0, Cibc ≈ Cica., Ciab ≈ +1.0 and ia(n)= −−ib(n) during the fault time. From the results, it is
obvious that the present algorithm can find the fault condition, choose the faulty phases of TL1 that are A and B,
and classify the fault kind that is a DL (A-B) fault free of grounding. In this study, the technique can determine
the fault type whether DLG or DL by calculating the cross-coherence factor computed between the two faulty
phases (A and B), where if Ciab ≈ +1.0, then the fault type is DL. Otherwise, it is DLG. Figure 7d plots the three
tripping characteristics based on the six coherence factors of the current waveforms for case 4. It is noticed that
the values of the five coherence factors (Cia, Cib, Ciab, Cibc, and Cica.) are located within the tripping zones of the
developed characteristics.
The simulation results reveal that the proposed technique is accurate, secure, dependable and reliable to
determine the faulted phases, and distinguish between DL isolated and grounded faults without adding any extra
algorithm to carry out this task. In addition, the algorithm is fast because the selected data set is within one cycle.

Case 5: 3LG fault with Rf = 0.0 Ω

In the present case, the fault type is a bolted 3LG fault (A-B-C-G). Figure 8a presents the three-phase currents
transformed at the TL1 sending end. In this case, it is noticed that the three-phase currents are higher than the
pre-fault currents; the post-fault currents are nearly 40.5 times the pre-fault current. As depicted in (Fig. 8b),
the pre-fault values of the three factors (Cia, Cib, and Cic) are balanced, similar and close to + 1.0. But, these
factors are less than + 0.5 during the first data set of the fault time. Figure 8c shows the three cross-coherence
factors (Ciab, Cibc, and Cica.). Their pre-fault values are symmetrical and equal to roughly + 0.25, while their
post-fault values are out of the scope bounded between 0.1 and 0.4. Therefore, the TL1 situation is abnormal.
Figure 8d depicts the three tripping characteristics with quadruple forms for case 5. It is shown that the values
of the six coherence factors (Cia, Cib, Cic, Ciab, Cibc, and Cica.) are settled inside the tripping areas of the proposed
characteristics. As a result, the fault classification is 3LG (A-B-C-G).
From the results of the extensive simulation cases, the suggested scheme is smart, effective, and quick for
detecting the situations of the series and ten shunt faults, selecting the faulty phase(s), and classifying the
fault type. This demonstrates that the coherence function is considered a proper fault detector, classifier, and
discriminator of faulty phase(s) for power TL.
The quantified findings are presented in Table 5 for the first two data windows of the fault time.
The coherence algorithm has been examined in situations involving noise in the current signals, stable power
swings, unstable power swings, CT saturation, and currents through high fault resistances. In addition, cases in
which the fault occurs at sending and receiving ends of the transmission line. Table 6 summarizes the results of
other case studies under the effect of these situations for the simulated power system. As presented in Appendix
(2), case studies for stable and unstable power swings have been discriminated using the coherence algorithm.
As illustrated in Table 6, the algorithm is blocked in the instance of stable power swings, while it responds in the
event of unstable power swings. This is one of the advantages of the proposed algorithm based on the coherence
statistic.

Salient features of the proposed technique

Advanced technique features
The present technique has several distinctive characteristics as follows:

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 14

www.nature.com/scientificreports/

Fig. 8. (a–d) Simulation results for case 5. (a) Three-phase secondary currents taken at the TL1 sending end.
(b) Coherence factors (Cia, Cib and Cic). (c) Coherence factors (Ciab, Cibc and Cica.). (d) Tripping characteristics
based on the coherence factors.

The obtained results during the first two windows of the fault time
Auto-coherence Cross-coherence
Case number Fault type Cia Cib Cic Ciab Cibc Cica
Pre-fault 1.0 1.0 1.0 0.25 0.25 0.25
Series fault (open A
Case 1 Cia < 0.3 Cib ≈ Cic ≈ 0.95 Cib ≈ Cic ≈ 0.95 Out of the range Out of the range Inside the range
phase)
SLG shunt fault )A-G)
Case 2 Cia < 0.84 Cib ≈ 0.87 Cic≈ 0.86 Out of the range Inside the range Out of the range
through Rf = 500 Ω
Out of the range
Case 3 DLG shunt fault )A-B-G) Cia < 0.2 Cib < 0.2 Cic > 0.85 Out of the range Out of the range
Ciab ≠ + 1.0
Out of the range Ciab Out of the range Cibc Out of the range
Case 4 DL fault (A-B) fault Cia ≈ Cib < 0.5 Cia ≈ Cib < 0.5 Cic ≈ 1.0
≈ +1.0 ≈ Cica., Cibc ≈ Cica.,
3LG shunt fault )A-B-
Case 5 Cia < 0.1 Cib < 0.5 Cic < 0.1 Out of the range Out of the range Out of the range
C-G)

Table 5. Post-fault results of the proposed protection scheme. The selected Δx is 0.15, the setting of the auto-
coherence factors is 1.0 − Δx, and the setting of the cross-coherence factors is 0.25 ± Δx.

(1) The coherence function is able to detect different fault events, select the faulty phase(s), and classify all ten
types of shunt faults. Furthermore, it can recognize series faults,
(2) It is highly sensitive because fault events through large fault resistances can be identified,
(3) It is capable of distinguishing between DL and DLG faults without any extra algorithm,
(4) This approach has a quick response, since the operating time taken by this approach is about half-cycle,
(5) Its security, sensitivity, and response speed can be set by adjusting the data set amount or the coher-
ence-setting margin,
(6) Three-phase current measurements are sufficient for the proposed scheme to be processed,
(7) New tripping curves are proposed, which have square shapes and are restricted and determined because
they are based on the coherence factors,
(8) The performance of the technique is independent on the specification of the current transformers and
power system components,
(9) The coherence method works like a digital low-pass filter to reduce the effect of DC contents and transient
harmonics on measured currents,
(10) The technique can be implemented in practice,
(11) It is advantageous with light data transmission burden because it only requires three-phase current meas-
urements at the TL sending or receiving end,
(12) Data synchronization systems are not required, and

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 15

www.nature.com/scientificreports/

Correct/
Ser. incorrect
no. Case type coherence algorithm action (blocking/tripping) action
Correct
1 Stable power swings Blocking
action
Unstable power Correct
2 Tripping
swings action
Blocking
Noise in the current Correct
3 (The reasonable coherence settings and the data window size can be used to avoid the effect of the acceptable harmonics content in the
signals action
current signals.)
The same author presented another paper57 that explains the methodology for CT saturation detection using the coherence algorithm. Correct
4 CT saturation
In57, case studies were described for busbar protection with CT saturation extent. action
Weak infeed and
outfeed (in which Tripping
Correct
5 the fault currents are (To make the algorithm more sensitive for fault detection, the size of the data window and coherence setting deviations should be
action
reduced and resemble reduced, which can be selected a sub-cycle period.)
load currents)
Tripping
Faults located The proposed algorithm installed at the sending end can monitor and protect the entire transmission line. In other sense, the proposed
at sending and algorithm constructed at the sending end can monitor and detect the faults at the receiving end of the power system under test. Correct
6
receiving ends of the Hints: The relay action depends on the impedance specification of the transmission line required to be protected and the fault action
transmission line resistance.
Additionally, the developed algorithm placed at the receiving end can monitor and protect the same transmission line.

Table 6. The relay responses for other situations.

(13) It is called high-reliability because it is immune to issues like fault time, fault inception angle, fault loca-
tion, fault resistance or fault type.

Comparison
Table 7 illustrates a comparison between the proposed protection technique based on the coherence criterion
and other recently published techniques used to protect transmission line systems.

Conclusions
A numerical analysis of the coherence function for detecting, selecting phase(s), and classifying TL faults
has been presented in this paper. The method has used six coherence factors that are calculated only for the
current measurements taken at the TL sending end. In order to evaluate the technique’s performance, numerous
simulation studies have been conducted on a segment of a power grid with real data parameters. The system
model has been simulated under different operating and fault conditions, such as fault type, fault location,
fault resistance, fault inception angle and power flow angle. The simulation results have confirmed the high
capabilities of the proposed protection for detection, classification, and faulty phase(s) identification of various
faults. In addition, it is simple, smart, exact, robust, stable, and reliable, and it is not affected by the specification
of the TL parameters or current transformers. Moreover, its response is fast as the relay operating time is within
a half-cycle time, and it is sensitive because the faults through high fault resistance are detectable. Furthermore,
the relay response speed, sensitivity, and security levels can be set by adjusting the data set amount and/or the
coherence setting margin. In conclusion, a new method has been introduced that uses square forms of tripping
curves based on the coherence values (which are similar to the per-unit values). Each curve includes two
restricted areas: tripping and restraining. The protection can be active within the tripping area for the abnormal
states of the system, while it can be inactive within the restraining area for the normal operating states. In the
following items, the proposed method based on the coherence function is superior to the existing methods based
on the impedance calculation:

(1) Both the tripping and restraining zones are restricted because their boundary ranges between 0.0 and 1.0,
and the normal operating points are neatly predetermined,
(2) The coherence coefficients are similar to the per-unit values,
(3) There is no need to calculate the setting parameters and the specification of the transmission lines and in-
strument transformers are not required,
(4) A low-pass filter and a data synchronization system are unnecessary,
(5) The algorithm demands only instantaneous measurements of the current signals acquired at one of the two
terminals of the transmission line,
(6) The same coherence algorithm is able to identify, classify, and discriminate the faulty phases. Besides, it can
find out and measure the current unbalance level,
(7) No additional algorithm is needed to differentiate between DL and DLG faults, and.
(8) The coherence technique can distinguish between stable and unstable power swings, and perform an appro-
priate decision-making process.

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 16

www.nature.com/scientificreports/

Ser.
no. Item The proposed protection technique The published protection techniques
Several existing methods require three-phase voltages and currents
The three-phase currents measured at one end of the transmission line are required
Required at one or two ends of the transmission line1,4,6,7,13,34,37,38. Thus, these
1 for estimating the six coherence coefficients. This results in swift data transfer and
measurements existing methods have a greater computational burden than the
processing.
proposal.
It is contingent on the mean-squared coherence (MSC) estimators, which are Numerous methods are employed to estimate the impedance required
2 Main basic
applied to only three-phase currents. for processing the protection of the transmission lines4,6,49.
3 Its function It has an online function. Certain protection methods have an offline function.
In5,26,32,36,39, the techniques used a higher sampling rate to protect the
4 Sampling size The sampling rate of the developed method is 100 samples per cycle. transmission line. The sampling rate exceeded 100 samples per cycle,
leading to a delay in the relay operation5,26,32,36,39.
The RMS values of voltage and current measurements are used to
In the developed algorithm, the coherence coefficients are computed using the
compute the impedance needed to operate the existing impedance
Data window/ moving data set concept. The data set quantity can be modified in accordance with
5 relays4,6,7,13,37,49. Other techniques used the data set concept, which
RMS concept the protection characteristics required and the prevailing conditions of the power
could be modified in line with the protection characteristics required
network.
and the prevailing operating conditions of the power system12.
The approach significantly outperforms other RMS quantities-based techniques
that require several cycles to compute the relay operating time. This fast response
reduces electrical and mechanical stresses during fault conditions, providing
improved protection for power transmission systems and ensuring power network
stability.
The tripping and blocking zones within the proposed quadratic characteristic The characteristic curves of the impedance relays had different
Tripping
curves derived from the coherence coefficients can be used to distinguish between shapes depending on the impedance estimation These curves were
6 characteristic
a transmission line fault and none, respectively. These curves are restricted from not limited because of the variation in the parameters of different
curves
0.0 to + 1.0. transmission system sizes4,6,30,31,49,53.
The tripping characteristics curves of the impedance relays depend on
the parameters of the power transmission lines and the specification
of the voltage and current transformers4,6,30,31,49,53.
The numerical values of the data set size and the specified coherence deviations
The settings of the impedance relays should be neatly set to avoid any
7 Relay settings can be used to alter the requirements of the protection attributes, such as speed,
protection malfunction4,30,31,53.
reliability, and sensitivity.
There is no need to estimate the coherence settings for changing the operating and
restraining regions of the coherence characteristic curves, and it is unnecessary to Numerous conventional relays of transmission lines protection need
modify the relay settings in response to changes in transmission system topologies/ extensive offline studies to prepare their settings1.
dimensions.
In3,7,8,10,11,13,]14,18,27, a variety of mathematical models were applied. As
a result, the algorithms needed diverse setting values to be set.
Algorithm
Several techniques are significantly complicated, making them
8 complexity It is easy to use in practice due to its simple mathematical formulas.
impractical18,19,22,36,41.
degree
Multiple Many protection functions can be performed for transmission systems using the In several protection techniques, only one or two protection functions
9
functions coherence coefficients quantified for the three-phase currents as follows: were implemented13,31.
(1) Detection of the series and shunt faults,
(2) Determination of the severity of shunt faults,
(3) Classification of TL shunt faults,
(4) Discrimination between DL and DLG faults,
(5) Differentiation between symmetrical and unsymmetrical faults,
(6) Identification of three-phase current imbalance, and
(7) Assessment of three-phase current imbalance.
The data window concept, which is used to estimate the coherence coefficients, Some protection systems may need additional low-pass filters to
10 Digital filter
serves as a low-pass filter. remove or alleviate ripples in voltage or current waves28,29.
Response to
Some protection strategies were incorrectly employed during stable
11 stable power The methodology does not operate during stable power swings.
power swings4,6,30,31,49.
swing
The variations in the topology and parameters of the power
transmission system affect the operation of several protection
strategies1,28.
Redundancy The utilization of auto-coherence and cross-coherence algorithms increases Numerous conventional methods lack the attributes of protection
12
protection protection reliability and redundancy. reliability and redundancy4,13,23,31.
Protection The present algorithm is capable of integrating with other conventional protection The absence of protection integration in numerous techniques
13
integration techniques to improve the protection reliability. resulted in a reduction in the protection reliability18,19.
In8,11, the approaches were dependent on the alienation estimator
The protection system is contingent upon the direct coherence models calculated
derived from the coherence/correlation estimator obtained for the
Protection for the transmission line current measurements. As a consequence, the operating
14 transmission line current measurements. Hence, the operational
speed speed of the coherence-based technique is superior to that of the alienation-based
speed of the alienation-based model is lower than that of the
technique.
coherence-based model.
In7,13,37, the methods used Phasor Measurement Units (PMUs) that
made a significant delay in fault detection.
The protection accuracy of several impedance relays is low owing to
Protection The simulation findings reveal that the suggested scheme has the ability to
15 the elevated fault resistance, which reduces the RMS magnitude of the
precision accurately identify fault events.
fault current13,17.
Continued

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 17

www.nature.com/scientificreports/

Ser.
no. Item The proposed protection technique The published protection techniques
Many fault detection methods are unreliable in the instance of
Protection The proposed approach achieves the requirements of the protection reliability. The
16 symmetrical faults4–6,32. Other existing protective relays achieve the
reliability failure of the relay operation indicates the unreliability of the protection system.
requirements of the protection reliability.
Increasing both the size of the data window and the coherence setting deviations
The protection security of the conventional impedance relays can be
Protection can enhance the protection security of the proposed technique. The increase in
17 enhanced by reducing the impedance curve and increasing the relay
security the coherence setting deviations increases the blocking region and decreases the
operating time4,6,30,31.
tripping region, with each other, inside the setup of characteristic curves.
Reducing both the amount of the data window and the coherence setting
deviations can improve the protection dependability of the proposed technique. The protection dependability of the conventional impedance relays
Protection
18 The reduction in the coherence setting deviations decreases the blocking can be improved by increasing the impedance curve and reducing the
dependability
region and increases the tripping region, with each other, inside the established relay operating time4,6,30,31.
characteristic curves.
The protection sensitivity of the proposed technique can be boosted by reducing
both the data set size and the coherence setting deviations of the suggested tripping
protection characteristic curves based on the coherence coefficients. The coherence setting High threshold impedance can enhance the protection sensitivity of
19
sensitivity deviations are beneficial for modifying both blocking and tripping regions placed the conventional impedance relays1,4,6,30,31.
in the proposed characteristic curves. Therefore, these settings are capable of
adjusting the protection sensitivity.
The proposed quadratic curves are limited because they are based on the coherence The size of the impedance tripping curve relies on the parameters
coefficients that expand from 0.0 to + 1.0. Thus, the reduction in the blocking of the transmission lines and the specifications of the voltage and
region results in an increase in the tripping region within the proposed curves. current transformers1,4,6,30,31.
The entire area of the proposed characteristic curve is independent of the Only one tripping curve can be selected for the impedance-based
parameters of the transmission lines or the specifications of the current protection. Consequently, the area of tripping region of the relay
transformers. characteristic curve is fixed4,6,30,31,49,53.
Application Different scales of transmission lines can be protected with the coherence-based It could be applied to protect diverse sizes of transmission lines
20
fields protection system. against different fault scenarios40–53.
In addition, the scheme can be applied to identify series and shunt faults in both Additionally, some existing techniques were applicable in both
traditional and smart grids with different voltage and power ratings. traditional and smart grids with a variety of voltage standards26,36,39.
It can be applied to protect many of power system components, such as
Conventional impedance relays are used to protect only power
transmission lines, AC generators54,56,59, AC motors, power transformers and
transmission lines26,36,39.
busbars57.
Experimental The investigation of the coherence approach involves thorough computer The impedance approaches were verified through comprehensive
21
verifications simulations54 and experimental verifications59. simulation programs and practical experiments40–53.

Table 7. A comparison between the proposed and published protection techniques.

Data availability
All data generated or analysed during this study are included in this published article [and its supplementary
information files].

Received: 27 August 2024; Accepted: 20 January 2025

References
1. Vibhuti, N., Jena, M. K. & Panigrahi, B. K. Synchro-phasors assisted back-up protection of transmission line. IET Gener. Transmi.
Distrib. 12, (14), 3414–3420, (2018).
2. Suraj, S. & Ashok K. P. Travelling-wave-based protection of transmission line using single-end data. IET Gener. Transmi. Distrib.
13 (20), 4659–4666 (2019).
3. Saeid Hasheminejad, S. G., Seifossadat, M. R. & Joorabian, M. Traveling-wave-based protection of parallel transmission lines
using teager energy operator and fuzzy systems. IET Gener. Transmi. Distrib. 10 (4), 1067–1074 (2016).
4. Shenyi, L., Xingxing, J. & Ramakrishna, G. High-speed distance relaying using least error squares method and testing with FPGA.
IET Gener. Transmi. Distrib. 13 (16), 3591–3600 (2019).
5. Chen, S., Tai, N., Liu, C. F. J. & Shubin, H. Sequence-component-based current differential protection for transmission lines
connected with IIGs. IET Gener. Transmi. Distrib. 12 (12), 3086–3096 (2018).
6. Behnam, M. & Jian, G. Z. Double-ended technique for distinguishing series faults from shunt faults on transmission lines using
the sequentialcomponents of impedance. IET Gener. Transmi. Distrib. 11 1381–1388 (2017).
7. Shahabodin Afrasiabi, M., Afrasiabi, M. & Mohammadi, B. P. Fault localisation and diagnosis in transmission networks based
on robust deep gabor convolutional neural network and PMU measurements. IET Gener. Transmi. Distrib. 14 (26), 6484–6492
(2021).
8. Bhuvnesh, R. & Abdul G. S. Wavelet-alienation based transmission line protection scheme. IET Gener. Transmi. Distrib. 11 (4)
995–1003 (2017).
9. Javier, A. C., Martinez-Velasco, j. A. & Magnusson, J. Optimum design of hybrid HVDC circuit breakers using a parallel genetic
algorithm and a MATLAB-EMTP environment. IET Gener. Transmi. Distrib. 11 (12), 2974–2982 (2017).
10. Abhishek Gupta, R. K., Pachar, B. K. & Mahela, O. P. A multivariable transmission line protection scheme using signal processing
techniques. IET Gener. Transmi. Distrib. 15 (22), 3115–3137 (2021).
11. Ola, S. R., Saraswat, A., Goyal, S. K., Jhajharia, S. K., Rathore, B. & Mahela, O. P. Wigner distribution function and alienation
coefficient-based transmission line protection scheme. IET Gener. Transmi. Distrib. 14 (10), 1842–1853 (2020).
12. Abd Allah, R. Protection scheme for transmission lines based on correlation coefficients. Int. J. Energy Power Eng. 3 (2), 93–102
(2014).
13. Zhang, C., Wang, X., Liu, Y., Shi, D. & Li, Y. Fast faulted line identification method for wide-area backup protection with PMU
optimal placement strategy. IET Gener. Transmi. Distrib. 13 (9), 1599–1611 (2019).
14. Gil, M. & Abdoos, A. A. Intelligent busbar protection scheme based on combination of support vector machine and S-transform.
IET Gener. Transmi. Distrib. 11 (8), 2056–2064 (2017).

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 18

www.nature.com/scientificreports/

15. Liang, Y. et al. A robust and accurate discrete fourier transform-based phasor estimation method for frequency domain fault
location of power transmission lines. IET Gener. Transmi. Distrib. 16 (10), 1990–2002 (2022).
16. Nan, P., Menghan, C., Rui, L. & Zare, F. Asynchronous fault location scheme for half-wavelength transmission lines based on
propagation characteristics of voltage travelling waves. IET Gener. Transmi. Distrib. 13 (4), 502–510, (2019).
17. Patel, B. & Parthasarathi, B. Fast fault detection during power swing on a hybrid transmission line using WPT. IET Gener.
Transmi. Distrib. 13 (10), 1811–1820 (2019).
18. Koley, E., Shukla, S. K.,S. K., Ghosh, S., Mohanta, D. K. Protection scheme for power transmission lines based on SVM and ANN
considering the presence of non-linear loads. IET Gener. Transmi. Distrib. 11 (9), 2333–2341 (2017).
19. Ghashghaei, S. & Akhbari, M. Fault detection and classification of an HVDC transmission line using a heterogenous multi-
machine learning algorithm. IET Gener. Transmi. Distrib. 15 (16), 2319–2332 (2021).
20. Barbosa, D. C. P. et al. Machine learning approach to detect faults in anchor rods of power transmission lines. IEEE Antennas Wirel.
Propag. Lett. 18 (11), 2335–2339 (2019).
21. Hasmat, M., Rajneesh, S. Transmission line fault classification using modified fuzzy Q learning. IET Gener. Transmi. Distrib. 11
(16), 4041–4050 (2017).
22. Taheri, M. M., Seyedi, H. & Behnam, M. DT-based relaying scheme for fault classification in transmission lines using MODP. IET
Gener. Transmi. Distrib. 11 (11), 2796–2804 (2017).
23. Maria Leonor, S., Almeida, K. M. & Silva. Transmission lines differential protection based on an alternative incremental complex
power alpha plane. IET Gener. Transmi. Distrib. 11 (1), 10–17 (2017).
24. Saeed, A. G. & Heresh, S. Adaptive CWT-based transmission line differential protection scheme considering cross-country faults
and CT saturation. IET Gener. Transmi. Distrib. 10 (9), 2035–2041 (2016).
25. Zhang, G. et al. Single-ended travelling wave-based protection scheme for double-circuit transmission lines. Int. J. Electr. Power
Energy Syst. 97, 93–105 (2018).
26. Hasheminejad, S. et al. Ultra-high-speed protection of transmission lines using traveling wave theory. Electr. Power Syst. Res. 132,
94–103 (2016).
27. Chaitanya, B. & Yadav, A. Decision tree aided travelling wave based fault section identification and location scheme for multi-
terminal transmission lines. Measurement 135, 312–322 (2019).
28. Santiago, J. & Tavares, M. C. Analysis of half-wavelength transmission line under critical balanced faults: voltage response and
overvoltage mitigation procedure. Electr. Power Syst. Res. 166, 99–111 (2019).
29. Luo, S. et al. Non-unit transient based boundary protection for UHV transmission lines. Int. J. Electr. Power Energy Syst. 102,
349–363 (2018).
30. Spielböck, T. & Beláň, A. Design of a one-sided, impedance-based transmission line fault locator using line topology and source
impedances’, Electr. Power Syst. Res. 161, 123–138 (2018).
31. Tong, X. & Wen, H. A novel transmission line fault detection algorithm based on pilot impedance. Electr. Power Syst. Res. 179,
106062 (2020).
32. Chatterjee, B. & Debnath, S. Sequence component based approach for fault discrimination and fault location estimation in UPFC
compensated transmission line. Electr. Power Syst. Res. 180, 106155 (2020).
33. Patel, B. A new FDOST entropy based intelligent digital relaying for detection, classification and localization of faults on the hybrid
transmission line. Electr. Power Syst. Res. 157, 39–47 (2018).
34. Swain, K. B., Mahato, S. S. & Cherukuri, M. Expeditious situational awareness-based transmission line fault classification and
prediction using synchronized phasor measurements. IEEE Access 7, 168187–168200 (2019).
35. Ghaedi, A., Golshan, M. E. H. & Sanaye-Pasand, M. Transmission line fault location based on three-phase state estimation
framework considering measurement chain error model. Electr. Power Syst. Res. 178, 106048 (2020).
36. Akmaz, D. et al. Transmission line fault location using traveling wave frequencies and extreme learning machine. Electr. Power
Syst. Res. 155, 1–7 (2018).
37. Das, S., Singh, S. P. & Panigrahi, B. K. Transmission line fault detection and location using wide area measurements. Electr. Power
Syst. Res. 151, 96–105 (2017).
38. Moravej, Z., Movahhedneya, M. & Pazoki, M. Gabor transform-based fault location method for multi-terminal transmission lines.
Measurement 125, 667–679 (2018).
39. Peng, N. et al. Fault location of transmission lines connecting with short branches based on polarity and arrival time of
asynchronously recorded traveling waves. Electr. Power Syst. Res. 169, 184–194 (2019).
40. Lei, K. Differential protection of distribution lines based on Hausdorff algorithm. J. Phys. 3rd Int. Conf. Green Energy Power Syst.
Ser. 2788 (1), 012032 (2024).
41. Guilherme, T. A. Ricardo, C. D. S. & Aline, N. Euclidean distance-based method for fault detection and classification in
transmission lines. J. Control Autom. Electr. Syst. 33 (5), 1–11 (2022).
42. Ke, J., Wen, Z., Tianshu, B., Tonghua, W., Yang, Z., Fanzhe, K. & Lei, Y. Waveform differential based pilot protection for transmission
line connected to renewable energy sources. Int. J. Electr. Power Energy Syst. 149, 108993 (2023).
43. Manassero, G. & Rozenblit Tiferes, R. Dynamic time warping based pilot protection algorithm for AC and HVDC transmission
lines. IEEE Access 11, 56846–56857 (2023).
44. Jian, L. I. U. et al. Fault discrimination of two-level VSC-HVDC DC line based on Pearson correlation of transient currents. J.
Electro Technol. 32 (03), 74–85 (2017).
45. Ke, J., Zhe, Y. & Wei et al. Longitudinal protection of new energy transmission line based on Spearman’s rank correlation
coefficient. J. Power Syst. Automat. 44 (15), 103–111 (2020).
46. Deng, L. Wu, H. & Li, X. Pilot protection of new energy field station transmission line based on kendall correlation coefficient. J.
Res. Sq. (2024).
47. Tiferes, R. R., Manassero, G., Pellini, E. L. & Silvio G. D. S. Biweight midcorrelation based transmission line pilot protection
algorithm. IEEE Open. Access J. Power Energy 11, 68–82 (2024).
48. Zhichang L. & Yin, X. A dice similarity coefficient-based pilot protection method for 500 kV transmission lines of large-scale
integrated photovoltaic power supply. Int. J. Electr. Power Syst. Res. 226 (03), 109918 (2024).
49. Tiferes, R. R. & Giovanni, M. J. Transmission line distance protection algorithm based on bayesian inference. Int. Trans. Electr.
Energy Syst. (4), 1–15 (2022).
50. Zheng, L., Jia, K., Bi, T., Yang, Z. & Yu, F. Structural similarity based pilot protection for renewable power transmission line. Int.
J. IEEE Trans. Power Deliv. 35 (6), 2672–2681 (2020).
51. Ke, J. I. A. et al. Cosine similarity-based longitudinal protection of wind farm station transmission line. Chin. J. Electr. Eng. 39,
6263–6275 (2019).
52. Liming, Z., Jia, K., Bi, T., et al. Cosine similarity-based longitudinal protection of T-connected feeder lines for new energy field
stations. J. Power Syst. Automat. 43 (18), 111–119 (2019).
53. Patel, B., Bera, P. & Sunita H. D. A novel method to distinguish internal and external faults during power swing. Int. J. IEEE Trans.
Power Deliv. 36 (5), 2595–2605 (2021).
54. Mahmoud, R. A. Coherence-based integrated detection and assessment algorithm for voltages and currents unbalance/disturbance
in three-phase synchronous generators: A validation of simulation results. IET J. Eng. (5), e12272 (2023).
55. Mahmoud, R. A. & Emam, A. Coherence-based automatic power factor correction (APFC) algorithm for power grids. IET J. Eng.
(5), 512–527 (2022).

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 19

www.nature.com/scientificreports/

56. Mahmoud, R. A. Smart automatic synchronization system based on coherence algorithm for power grids. IET J. Eng. (1), e12214.
(2023).
57. Mahmoud, R. A. Integrated busbar protection scheme utilizing a numerical technique based on coherence method. IET J. Eng.
2022 (1), 94–119 (2022).
58. Mahmoud, R. A. & Elwakil, E. S. Experimental investigations using quadratic-tripping characteristics based on alienation/
coherence coefficients of voltage and current signals for synchronous generators protection. IET Gener. Transmi. Distrib. 15 (21),
2978–3000 (2021).
59. Mahmoud, R. A. Experimental performance verification of an intelligent detection and assessment scheme for disturbances and
imbalances of three-phase synchronous machine output using coherence estimators. Int. J. Sci. Rep. 14, 26278 (2024).
60. Mahmoud, R. A. Detection and assessment scheme of voltage and current unbalance for Three-phase synchronous generators
using dual numerical techniques based on correlation and deviation percentage concepts. IET J. Eng. 477–495 (2021).

Acknowledgements
Personal ProfileAssoc. Professor/ Ragab Abd Allah Was Born in Giza, Egypt, In 1976. He Received the B.Sc.
Degree in Electrical Machine & Power Engineering from Cairo University, Faculty of Engineering, Egypt, In
2001; and The M.Sc. and Ph.D. Degrees in Electrical Machine & Power from Helwan University, Faculty of Engi-
neering, Department of Electrical Machines& Power Engineering, Cairo, Egypt, In 2007 And 2012, Respectively.
I Addition to That I Possess Associate Professor Certificate. M.Sc. Thesis Under the Following Title: ‘’ A Digital
Busbar Protection Scheme Avoiding Current Transformer Saturation Effects’’; And Ph. D. Thesis Under the
Following Title: ‘’Multifunction Digital Relay for Large Synchronous Generators Protection’’. Assoc. Professor/
Ragab Abd Allah Is a Chairman of Electrical Power and Control Department in Pyramids Higher Institute for
Engineering and Technology-Electrical Power and Control Department- 6th of October City-Giza-Egypt. He
Is a Live Member of the Quality Committee of Electrical Power Engineering Department in the Institute. His
Current Research Interests Include Power System Protection and Automation, Power System Operation and
Control, Renewable Energy, Smart Grids (SG), And Power Quality.

Author contributions
R. A. Mahmoud: Conceptualization, Methodology, Software, Data curation, Writing-Original draft preparation,
Visualization, Investigation, Supervision, Validation, Writing-Reviewing and Editing.

Funding
Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooper-
ation with The Egyptian Knowledge Bank (EKB).

Declarations

Competing interests
The authors declare no competing interests.

Additional information
Supplementary Information The online version contains supplementary material available at ​h​tt​ ​p​s​:​/​/​d​o​i​.o
​ ​r​g​/​1​
0​.​1​0​38​ ​/​s​4​1​5​9​8​-​02​ ​5​-​8​7​5​7​7​-​5​​​​.​​
Correspondence and requests for materials should be addressed to R.A.M.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and
indicate if changes were made. The images or other third party material in this article are included in the article’s
Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included
in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or
exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy
of this licence, visit http://creativecommons.org/licenses/by/4.0/.

© The Author(s) 2025

Scientific Reports | (2025) 15:8487 | https://doi.org/10.1038/s41598-025-87577-5 20