A simple faulted phase-based fault distance estimation algorithm for a loop distribution system

Indonesian Journal of Electrical Engineering and Computer Science
Vol. 25, No. 1, January 2022, pp. 14~24
ISSN: 2502-4752, DOI: 10.11591/ijeecs.v25.i1.pp14-24  14
Journal homepage: http://ijeecs.iaescore.com
A simple faulted phase-based fault distance estimation
algorithm for a loop distribution system
Shwe Myint, Warit Wichakool
Department of Electrical Engineering, Faculty of Engineering, Prince of Songkla University, Songkhla, Thailand
Article Info ABSTRACT
Article history:
Received Jan 25, 2021
Revised Nov 19, 2021
Accepted Nov 23, 2021
This paper presents a single ended faulted phase-based traveling wave fault
localization algorithm for loop distribution grids which is that the sensor can
get many reflected signals from the fault point to face the complexity of
localization. This localization algorithm uses a band pass filter to remove
noise from the corrupted signal. The arriving times of the faulted phase-
based filtered signals can be obtained by using phase-modal and discrete
wavelet transformations. The estimated fault distance can be calculated
using the traveling wave method. The proposed algorithm presents detail
level analysis using three detail levels coefficients. The proposed algorithm
is tested with MATLAB simulation single line to ground fault in a 10 kV
grounded loop distribution system. The simulation result shows that the
faulted phase time delay can give better accuracy than using conventional
time delays. The proposed algorithm can give fault distance estimation
accuracy up to 99.7% with 30 dB contaminated signal-to-noise ratio (SNR)
for the nearest lines from the measured terminal.
Keywords:
Band pass filter
Discrete wavelets transform
Fault location
Loop distribution system
Noise
Traveling wave
This is an open access article under the CC BY-SA license.
Corresponding Author:
Warit Wichakool
Department of Electrical Engineering, Faculty of Engineering, Prince of Songkla University
Hat Yai, Songkhla 90112, Thailand
Email: warit.wi@psu.ac.th
1. INTRODUCTION
A fault in a power system interrupts a modern way of living. An ability to locate a fault in a power
system loop distribution network is very important to improve reliability for customer satisfaction and
utilities. The outage location is needed to know accurately to remove the outage line quickly from the
remaining parts of the system. Therefore, different conventional fault distance estimation methods such as
impedance-based method, frequency domain method, and traveling wave (TW) method are used to solve this
problem.
Some researchers utilized impedance-based algorithms with the help of system fault signals [1], [2].
Next, the knowledge-based method has been used to detect the fault location using neural networks, decision
trees and support vector machine techniques [3]-[7]. Some previous researcher used voltage sag matching
algorithm in [8], comparing faulty phase voltage with measure value approach was used in [9] and weighted
least square approach in [10] respectively. All the above algorithms require training data sets with pre-
defined different fault situations.
Traveling wave-based algorithms can give the best accuracy with the help of discrete wavelet
transforms (DWT) filters. Most of fault localization algorithms use the measured fault signals from single
end of the system by identifying the first sudden changing point in the waveform [11]-[13]. This method was
analyzing in time domain that cannot get accurate arriving time for high impedance faults. A simple fault
localization algorithm based on characteristics of network topology and recloser-generating traveling wave

Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 
A simple faulted phase-based fault distance estimation algorithm for a loop distribution system (Shwe Myint)
15
signals was introduced in [14]. However, the feeders without recloser cannot be protected by this algorithm.
The combination of impedance and traveling wave-based algorithms are proposed [15]-[17]. The arriving
times of zero and aerial components are extracted by DWT filter (db4, level1) and utilized to estimate fault
distance [15]. Dijkstra shortest path algorithm was proposed based on arrival time of initial voltage traveling
wave fronts at the two terminals [16]. However, these algorithms did not consider about fault section to trip
correctly the faulty part. Moreover, these methods utilized the general aerial mode arriving time of modal
components which cannot give the accurate fault information for all fault types.
A decision tree aided traveling wave fault location algorithm was introduced in [18] and this
algorithm used the two adjacent arrival times of the faulted line. This method could not get highest accuracy
because it used the arriving time of general aerial mode components of Clark’s transform which cannot give
the accurate fault information for all fault types. In addition, some single line to ground TW based methods
used DWT filter and support vector regression (SVR) model [19], polarities of initial voltage and current
signals [20], the difference of velocities by separating the initial arriving times of zero and aerial mode
components [21] and the last one based on zero-sequence components distribution characteristics [22],
respectively. These methods also used arriving time of conventional aerial mode that cannot give correct fault
information for all faulted phases and fault types. The next method is asynchronous voltage-based TW data
fusion method [23]. This method requires the data training. Moreover, a single phase to ground fault
localization method [24] used the fault characteristics include in zero sequence voltages and currents.
However, the method is suitable for radial distribution systems.
According to literature reviews, the best accuracy can be achieved based on the arrival time
difference between the modal components of the current waveform monitored at the substation bus [19]. All
the traveling wave methods applied in the above literature used the arriving time of conventional aerial mode
components and only for transmission line and radial network topologies. However, the arriving time of these
components cannot give the accurate and correct fault information and significantly one of the difficulties in
loop distribution network is that the sensor can get many reflected signals from the fault point to face the
complexity of detection of actual faulted length. Therefore, the proposed algorithm can choose the correct
faulted section in [25], the accurate faulted path and uses the arriving time of faulted phase-based modal
component [26] that can give higher accuracy than that of conventional methods with the consideration of
noise effects, such as white noise. The paper is organized as section: 2 presents traveling wave method and
modal transformation, research methodology is explained in section 3. Section 4 focus on simulation results
and discussions for loop distribution test systems. The last section is conclusion.
2. THEORY BACKGROUDS
In this paper, the traveling wave reflected signal is mainly used to detect the arriving time of fault
signal. These reflected signals are necessary to be significant and accurate. Therefore, the reflected signals
are pass through the discrete wavelet filter via the phase signals are needed to be transformed using phase to
modal transformation matrix to avoid mutual effects. The following subsections are theories and detail
information of the component’s background.
2.1. Traveling wave theory
In the proposed algorithm, the recorded fault signals from the measure bus are transformed to modal
components. Different modes have different wave velocity depending on the system characteristic
impedance. Hence, the arrival time of difference mode will result in a time delay. The location of single-
ended faults is given as (1) [24].
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(1)
The time delay between zero mode current and aerial mode currents of Park’s transformation is
shown in (2). The time delay between zero mode current and modal difference currents is shown in (3) and
the time delay between the zero mode and aerial mode currents of the Karenbauer’s transformation in (4).
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(2)
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(3)
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(4)

 ISSN: 2502-4752
Indonesian J Elec Eng & Comp Sci, Vol. 25, No. 1, January 2022: 14-24
16
The variable to is the arriving time of the zero mode, which can be obtained using the residual current. The
value can be calculated from (5).
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(5)
The zero-mode wave velocity can be estimated based on the system characteristic impedance,
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(6)
where 𝐿0 is the zero-sequence inductance and 𝐶0is zero sequence capacitance of the power line. The arriving
time of the aerial mode using different transformation can be calculated using the same in (7).
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(7)
The wave velocity of the aerial mode is given by (8),
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(8)
where 𝐿1 and 𝐶1are positive sequence inductance and capacitance of the power line respectively. The arriving
time of the aerial mode can be different depending on the transformation used to process the time domain
signal. Using these different arriving times of aerial components can result different accuracies; therefore, the
modal difference component and faulted phase-based arriving time are used and compared with that of the
conventional aerial mode components.
2.2. Phase to modal transform
The phase components are transformed to modal components using Karenbauer’s transform to avoid
mutual effects. The transformation matrix is shown in (9). The 𝛼, 𝛽 and 𝛾 modal represent the difference
between two current phases. On the other hand, the zero-modal (0) is similar to the zero component in the
sequence network.
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(9)
In this article, the most accurate arriving time of aerial mode is chosen from the aerial components related
with Park’s, Clark’s and Karenbauer’s transform, and aerial different component.
2.3. Wavelet transforms
The DWT splits into two components. They are approximation coefficients (cA) (low frequency)
and detail coefficients (cD) (high frequency). Given an input signal x(k), its DWT can be calculated as (10),
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(10)
where a0 is the scale factor, b0 is the translation factor. The variable k represents an input sample number.
Variables m and n represent mother wavelet and decomposed level, respectively.
In this paper, detail coefficients of db6 mother wavelet for level1, 2 and 3 are extracted and utilized
to reconstruct the signals. The arriving times are detected from the component that is combining the three
reconstructed signals using combinations of three detail levels, D1, D2 and D3 to obtain accurate arriving
time of the reflected fault signal. The accuracy comparison of db4 and db6 was described in [20]. To estimate
fault distance using traveling wave method, in this paper, faulted phase-based time delay is utilized. It is
necessary to get correct faulted phase from the fault information accuracy and faulted line from previous
work [20]. Figure 1 shows the three levels detail coefficients decomposition diagram.

17
0.625MHz
-1.25MHz
High Pass
Low Pass
H
L
H
L
Original
Signal 2.5MHz-5MHz
0-2.5MHz
10MHz
H
L
1.25MHz-2.5MHz
0-1.25MHz
0-.625MHz
Figure 1. DWT decomposition
3. PROPOSED ALGORITHM
In this process, the recorded fault signals will firstly pass through the noise filter to remove noise
from the input signal and then the filtered signals will be transformed to modal components to avoid mutual
effects. In traveling wave fault location methods, the arriving times of fault signals are very important
parameters to estimate fault distance, they can be extracted in time domain, sequence domain and wavelet
domain as well. Therefore, the arriving times are extracted by using detail coefficients of DWT in this paper.
After that, the correct faulted phase is detected by using fault information accuracy content in the three-phase
current signals. Finally, the estimated fault distance can be estimated by using time delays of detail
coefficients of modal current components. The process is generally illustrated in Figure 2. The arriving times
of absolute maxima of detail coefficient of zero and aerial mode current traveling wave signals can be
estimated by using db6 mother wavelet because db6 mother wavelet can give more accurate arriving time
than db4 mother wavelet that is used by most of previous fault location researchers for that purpose.
Recording three-phase
fault current signals
Noise Filtering
Modal Transformation
Discrete Wavelet
Transformation
Faulty Phase
Identification
Fault Distance
Estimation
Figure 2. Fault estimation process
4. SIMULATION RESULTS AND DISCUSSIONS
This paper uses a loop overhead distribution (modified IEEE 14-bus) test system that consists of
three power supplies of 33 kV/10 kV, 50Hz as shown in Figure 3. The faults are simulated in one kilometer
step on all branches of the test system. This paper uses the distributed line parameter model in the simulation.
Before locating fault, the system firstly identifies the faulted line with the help of the fault detectors at all
buses. Therefore, the fault detectors are installed at all buses of the test system. In this paper, the faulted
section identification algorithm [25] is applied to get accurate faulted feeder. After obtaining faulty feeder,
the system continues to estimate fault distance using only the recording signals from bus 1 (B1) for all lines

 ISSN: 2502-4752
18
of the test system. Various fault distances with fault resistance 0.001 𝛺 to 50 𝛺 considering noise effects with
30 to 70 signal-to-noise ratio (SNR) dB are simulated and analyzed as well. Total 1650 cases for 55 points
with 100 iterations for every case are studied. System parameters are generator source resistance 13.96 𝛺 and
source inductance of 0.35H.
4.1. Accuracy calculation
In the loop distribution system, there are many possible predefined fault distance values because
there are many paths from the fault location to the measure terminal. For example, when a fault occurs at the
middle of line (L7) which is 4 km long, there are five possible ways to get different five arriving signals from
the fault point (F) to the measure terminal (B1). They are shown in the Figure 4.
B-1
B-2
G
G
G
L2
L3
L4
L5
L6
L7
L8
L14
L10
L11
L12
L13
L9
L15
L16
B-3
B-4
B-8
B-7
B-9
B-5
B-6
B-12
B-13
B-11 B-10
B-14
2km
4 km
4 km
6 km
10 km
4 km
2 km
4 km
5 km
4 km
4 km
5 km
4 km
6 km
4 km
Figure 3. IEEE 14-Bus modified test system [25]
Fault
Fault
Fault
Fault
Fault
B5 B1
2km 6km
8km
B5 B2 B1 16km
2km 10km 4km
B4 B2 B1 10km
4km 4km
2km
B4 B3 B2 B1
4km 4km
2km
B4 B2 B5 B1
4km 10km
2km
12km
22km
2km
6km
Loop (1)
Loop (2)
Loop (3)
Loop (4)
Loop (5)
Figure 4. Possible actual fault distances when fault occurs at L7
According to these loops, possible fault distances from the measuring point are 8, 16, 10, 12 and
22km that are possible actual fault distances for that case. Hence the range of possible fault distance (PFD) in
this case is between 8 and 22 km. If the estimated fault distance (EFD) resides within the possible fault
distance, the estimated fault distance is feasible. Otherwise, the algorithm assumes that the estimated distance
is erroneous and should be discarded. This scenario can occur under noisy environment. distance helps
eliminate erroneous estimation. If the estimated fault distance is within the valid range, the algorithm can
continue. Then we can access the performance of the algorithm by calculating the fault distance estimation
accuracy (FDEAC) which can be defined by (11).
𝑑 =
𝑣0𝑣1(𝑡2−𝑡1)
(𝑣1−𝑣0)
(11)

19
According to the example shown in Figure 4, the actual fault distance can have multiple values
depending on the path the traveling wave take. Therefore, the algorithm must choose the possible fault
distance from the list to calculate the accuracy value. In this paper, the value closest to the average estimated
fault distance is chosen. If there are two possible AFD values, the smallest one is chosen.
For example, in the case of signals mixed with noise 60 SNR dB, and the estimated fault distance is
7.9908 km. Therefore, the actual fault distance would be 8km because it is the closest one to the estimated
value in this case. The FDEAC is 99.89% by (11).
4.2. Arriving time based on faulted phase
This paper uses the arrival time of the mode related to the faulted phase to achieve better accuracy.
For example, the arrival times of the travelling wave when phase ‘c’ to ground (CG) fault occurs in the
middle of L7. Four modes of travelling wave are extracted from the simulation and shown in Table 1.
From this example, the arrival time of 𝛽 and 𝛾modal components are the same because they are
related to phase ‘c’, the faulted phase. If wrong information is used, the estimation could be erroneous.
Therefore, the proposed algorithm used the arriving time of faulted phase related modal component to get
higher accuracy.
Table 1. Arriving times of detail coefficients of zero and aerial mode components when a single line to
ground (CG) fault occurs at L7
Components Arriving time (ms)
𝛼modal 40.05
𝛽 modal 40.03
𝛾 modal 40.03
Zero modal 40.15
4.3. Detail level analysis without noise
In order to obtain the time delay, the wavelet transform is used to localize the arrival time among
different modes. This paper uses the db6 mother wavelet and three level of decomposition to extract the
arrival time information. To verify the best accuracy, different combinations of detail coefficients are used to
compared the fault estimation error. The cases are the arriving times taken from: i) the combination of
reconstructed signals by using detail coefficients level1 (D1), level2 (D2) and level3 (D3) shown here as
(D1+D2+D3); ii) the combination of reconstructed signals by using detail coefficients, D1 and D2, shown
here as (D1+D2); iii) the reconstructed signals by using detail coefficients, D1; iv) the reconstructed signals
by using detail coefficients, D2; and v) the reconstructed signals by using detail coefficients, D3,
respectively. In this study, the proposed algorithm is tested with single line to ground (SLG) fault for all lines
of test system. The result is shown in Table 2. Table 2 shows the performance of FDEAC for all lines of the
test system without noise in the simulation. This accuracy using the time delay with combination of the fault
information from the detail level 1, 2 and 3 (D1+D2+D3) give the best accuracy in five comparative cases,
which is expected because they cover widest frequency range of all combinations.
Table 2. Average % of FDEAC for all lines with different detail levels without noise effects
Faulted Lines
% FDEAC by
D1+D2+D3 D1+D2 D1 D2 D3
L1 99.68 99.68 99.90 0.00 98.19
L2 99.75 99.97 99.68 99.68 98.90
L3 99.91 93.53 93.62 93.81 94.66
L4 99.82 99.82 85.91 99.82 99.68
L5 99.68 99.81 99.98 99.64 66.27
L6 99.89 99.58 99.79 99.79 85.52
L6 99.89 99.58 99.79 99.79 85.52
L7 99.89 94.44 94.27 99.99 99.82
L8 99.91 99.83 99.91 63.17 99.19
L9 99.85 99.98 99.94 98.34 94.99
L10 99.85 99.98 99.93 99.98 99.65
L11 99.85 94.11 99.94 94.76 85.63
L12 99.92 95.63 95.63 98.48 99.92
L13 95.85 93.90 96.86 93.90 51.82
L14 99.85 88.52 99.94 88.52 80.95
L15 99.88 96.10 95.05 99.78 99.83
L16 99.92 97.50 99.99 97.50 89.40
Average 99.59 97.02 97.52 95.14 97.05

 ISSN: 2502-4752
20
4.4. Detail level analysis with noise
In order to verify the accuracy of the set of detail coefficients used to extract the arrival time, the
Gaussian white noise is added to simulate a practical situation where the input signal may be noisy. The
simulation was done with SNR from 30 to 70 dB. The result is shown in Figure 5.
According to Figure 5, although the estimation accuracies by D1 are the best in overall, the values
are not significant in single case for five parameters with different detail levels. However, the parameter with
combination of detail level 1, level 2, and level 3 can give better accuracy in high noise levels, such as 40dB
and 50dB and the fault distance estimation accuracy by using this parameter for most of faulted lines are
above 99%. Therefore, the arriving times for proposed fault distance estimation algorithm can be detected
from the component that is a combination of reconstructed signals by using detail coefficients level 1, level 2
and level 3 (D1+D2+D3) of db6 mother wavelet.
Figure 5. Performance of average FDEAC for all lines with different detail levels under different noise levels
4.5. Accuracy performance of proposed algorithm
Simulation results with additive noise of different signal-to-noise ratio are shown in Table 3. The
results compare the average accuracy of fault distance estimation for the travelling wave using different delay
time, including dq, d0 and modal transformation. Examples include all single line to ground faults.
According to the result in Table 3 in the case of phase ‘a’ to ground (AG) fault, higher accuracy can
be obtained from the time delay of the faulted phase, meaning that the algorithm will use the time delay from
𝛥𝑡𝛼 or 𝛥𝑡𝛽.The estimation result using the non-faulted phase, specifically 𝛥𝑡𝛾, g ive low accuracy percentage. In
addition, the estimation accuracy using time delay from the faulted phase is comparable to the time delay
from traditional dq and d0 transformation
In the case of phase ‘b’ to ground (BG) fault, a very good average accuracy can be found using the
faulted phase time delay (𝛥𝑡𝛼 or 𝛥𝑡𝛾) specifically above 99%. There is exception at 30 dB SNR cases. The
accuracy gets lower because of higher noise interference. Furthermore, if the conventional time delay, 𝛥𝑡𝑑, is
used, the estimation error is less accurate for the BG fault.
Table 3. Average fault distance estimation accuracy percentage of proposed fault localization algorithm for
fault occurred at L1
Fault type SNR (dB) dq
t
 d
t
 t
 t
 t

AG 30 99.88 99.96 99.49 99.98 72.63
AG 40 99.56 99.51 99.54 99.51 50.48
AG 50 99.28 99.28 99.28 99.28 32.31
AG 60 99.26 99.26 99.26 99.26 69.44
AG 70 99.26 99.26 99.26 99.26 55.33
BG 30 31.10 26.63 31.80 13.85 93.50
BG 40 98.05 47.93 94.05 13.85 99.15
BG 50 99.84 73.91 99.88 18.96 99.86
BG 60 99.63 91.27 99.66 89.67 99.66
BG 70 99.28 98.40 99.28 82.00 99.28
CG 30 97.87 99.56 40.04 99.64 98.48
CG 40 98.79 99.79 22.37 99.79 99.79
CG 50 99.10 99.26 61.25 99.26 99.26
CG 60 99.26 99.26 36.62 99.26 99.26
CG 70 99.26 99.26 49.10 99.26 99.26

21
Finally, in the case of phase ‘c’ to ground (CG) fault, the fault distance estimation result using the
information form the faulted phase (𝛥𝑡𝛽 or 𝛥𝑡𝛾) gives above 99% accuracy. In addition, the proposed
algorithm using the faulted phase time delay can get better accuracy than the conventional time delay under
low signal-to-noise ratio (30dB and 40dB). These results are similar to the (AG) fault cases.
According to the result shown in Table 3, the proposed algorithm has demonstrated that the time
delay of that related to the faulted phase of the transform current waveform using Karenbauer’s
transformation give the best accuracy. Furthermore, in case of the single phase to ground fault, two time-
delay parameters should be the very close, for example, 𝛥𝑡𝛼 and 𝛥𝑡𝛽 for phase ‘a’ to ground fault. Therefore,
the algorithm can use this information to verify the result as well.
To demonstrate an overall performance, the averaged fault distance accuracy percentage of all lines
are shown in Figure 6. The results show that the performance of average fault distance estimation accuracy
for all lines with additive noise at 70 dB SNR. The simulations were done for high fault current at 0.001 Ohm
fault resistance shown in Figure 6(a) and moderate fault resistance at 50 Ohms shown in Figure 6(b). Each
subfigure shows average fault distance estimation accuracy using three-time delays: d0 time delay, dq time
delay and faulted phase (FP) time delay to compare the overall accuracy.
According to the result in Figure 6, the proposed FP subscription time delay (𝛥𝑡𝛼, 𝛥𝑡𝛽, and 𝛥𝑡𝛾)
gives better averaged accuracy than the conventional time delays (𝛥𝑡𝑑 and 𝛥𝑡𝑑𝑞) for fault distance estimation
in case of faults in most lines. The averaged estimation accuracy for fault on line 13 is worse than the
algorithm that use the 𝛥𝑡𝑑𝑞time delay. However, this recording was done at Bus 1, which is far away.
Different point of observation can be done to double check or narrow down the fault location. The result on
line 4 at 50 Ohms fault resistance is also worse than the 0.001 Ohm fault resistance case. This may be due to
shorter distance and lower fault signal causing the wavelet to give the inaccurate time delay. However,
overall performance is good for fault on all lines in this network topology. Hence, the time delay of the
faulted phase should be used to calculate the fault distance.
(a)
(b)
Figure 6. Average fault distance estimation accuracy for all lines with SNR=70 dB for
(a) Rf =0.001 Ω and (b) Rf =50 Ω

 ISSN: 2502-4752
22
The performance of overall average estimation accuracy for all specified SNR values is illustrated in
Figure 7(a) for 0.001 Ohm minimum fault resistance and Figure 7(b) for 50 Ohm maximum assumed fault
resistance respectively. In Figure 7, The results compare the fault distance estimation accuracy all lines with
the same SNR. Three lines represents the value estimated with d0 time delay, dq time delay and FP time
delay to demonstrate the advantage of the proposed algorithm.
(a)
(b)
Figure 7. Overall average estimation accuracy in all lines with SNR= 30 to 70dB for (a) Rf = 0.001 and
(b) Rf = 50 
According to the result in Figure 7, the averaged estimation accuracy of the proposed faulted phase
time delay demonstrates that it gives better accuracy than conventional time delay from aerial component
(𝛥𝑡𝑑) and aerial different component (𝛥𝑡𝑑𝑞) for all the noise case. This is due to the fact that the faulted phase
contains more information about the fault than the healthy line, which used in the conventional dq
transformation. Hence, the proposed algorithm has demonstrated that it can accurately locate the fault in a
loop transmission under the simulation even with additive noise.
5. CONCLUSION
This paper proposes a simple fault localization technique in loop distribution systems. The
proposed technique passed through band pass filter, Karenbauer’s transform, and wavelet transform criteria
to estimate for fault distance by using only three phase current signals at the source bus (B1) of the test loop
distribution network. The proposed algorithm uses the arriving time of faulted phase-based modal component
that can give higher accuracy than that of conventional methods with the consideration of noise effects, such

23
as white noise. According to detail level analysis section, using the arriving times of the component
combined the three reconstructed signals using three detail level coefficients (D1, D2 and D3) can give best
accuracy. The proposed algorithm shows that the faulted phase time delay() can give better accuracy than
using conventional time delay() and aerial different time delay().
The algorithm performance is successfully tested with data obtained by simulations with 1650
different cases of sampled data at various situations, faulted feeders, faulted resistance, 30 to 70 dB SNR
values and fault locations. The proposed methodology is evaluated through MATLAB/Simulation single line
to ground fault in a 10 kV grounded loop distribution system. The proposed algorithm can give fault location
accuracy up to 99.7% with 30 dB contaminated SNR for the nearest lines from the measured terminal. Using
proposed algorithm, the overall fault information accuracy can get up to 83% and overall fault distance
estimation accuracy up to 93%. This proposed technique can be used to implement in real data with actual
fault records.
ACKNOWLEDGEMENTS
The authors would like to express her gratitude to their scholarship program, the Higher Education
Research Promotion and Thailand's Education Hub for Southern Region of ASEAN Countries (TEH-AC).
REFERENCES
[1] Y. Liao, “A novel method for locating faults on distribution systems,” Elect. Pow. Syst. Res., vol. 117, pp. 21–26, 2014, doi:
10.1016/j.epsr.2014.07.026.
[2] S. A. Al Kazzaz, I. Ismael, and K. K. Mohammed, “Fault detection and location of power transmission lines using intelligent
distance relay,” International Journal of Power Electronics and Driver System, vol. 11, no. 2, pp. 726-734, Jun. 2020, doi:
10.11591/ijpedsv11.i2pp726-734.
[3] A. C. Adewole, R. Tzoneva, and S. Behardien, “Distribution network fault section identification and fault location using wavelet
entropy and neural networks,” Applied Soft Computing, vol. 46, pp. 296–306, Sept. 2016, doi: 10.1016/j.asoc.2016.05.013.
[4] A. Swetapadma and A. Yadav, “A novel decision tree regression-based fault distance estimation scheme for transmission lines,”
IEEE Transactions on Power Delivery, vol. 32, no. 1, pp. 234-245, Feb. 2017, doi: 10.1109/TPWRD.2016.2598553.
[5] X. Deng, R. Yuan, Z. Xiao, T. Li, and K. L. L. Wang, “Fault location in loop distribution network using SVM technology,”
International Journal of Electrical Power & Energy System, vol. 65, pp. 254–261, Feb. 2015, doi: 10.1016/j.ijepes.2014.10.010.
[6] F. I. Jabbar, D. Soomro, A. H. Tawafan, M. N. Abdullah, N. H. Radzi, and M. H. Baloch, “Optimization of detection of a single
line to ground fault based on ABCNN algorithm,” International Journal of Power Electronics and Driver System, vol. 9, no. 4,
pp. 623-629, Dec. 2020, doi: 10.11591/ijai.v9.i4.pp623-629.
[7] S. S. D. Sahel and M. Boudour, “Wavelet energy moment and neural networks-based particle swarm optimisation for
transmission line protection,” Bulletin of Electrical Engineering and Informatics, vol. 8, no. 1, pp. 10-20, Mar. 2019, doi:
10.11591/eeiv8i1.1214.
[8] M. Daisy and R. Dashti, “Single phase fault location in electrical distribution feeder using hybrid method,” Journal Energy, vol.
103, pp. 356–368, May 2016, doi: 10.1016/j.energy.2016.02.097.
[9] S. H. Mortazavi, Z. Moravej, and S. M. Shahrtash, “A searching based method for locating high impedance arcing fault in
distribution networks,” IEEE Transactions on Power Delivery, vol. 34, no. 2, pp. 438-447, Apr. 2019, doi:
10.1109/TPWRD.2018.2874879.
[10] M. J. S. Ramos, A. S. Bretas, D. P. Bernardon, and L. L. Pfitscher, “Distribution network HIF location: A frequency domain
system model and WLS parameter estimation approach,” Electrical Power Systems Research, vol. 146, pp. 170-176, May 2017,
doi: 10.1016/j.epsr.2017.01.030.
[11] P. Cao, H. Shu, B. Yang, J. Dong, Y. Fang and T. Yu, “Speeded-up robust features based single-ended traveling wave fault
location: a practical case study in Yunnan power grid of China,” IET Generation Transmission & Distribution, vol. 12, no. 4, pp.
886-894, Feb. 2018, doi: 10.1049/iet-gtd.2017.0735.
[12] E. E. Aker, M. L. Othman, I. Aris, N. I. Abdul Wahab, H. Hizam, and O. Emmanuel, “Transmission line fault identification and
classification with integrated FACTS device using multiresolution analysis and naïve bayes classifier,” International Journal of
Power Electronics and Driver System, vol. 11, no. 2, pp. 907-913, Jun. 2020, doi: 10.11591/ijpedsv11.i2pp907-913.
[13] J. Ding, L. Li, Y. Zheng, C. Zhao, H. Chen, and X. Wang, “Distributed traveling-wave-based fault location without time
synchronization and wave velocity error,” IET Generation Transmission & Distribution, vol. 11, no. 8, pp. 2085-2093, Jun. 2017,
doi: 10.1049/iet-gtd.2016.1778.
[14] S. Shi, B. Zhu, A. Lei, and X. Dong, “Fault location for radial distribution network via topology and reclosure-generating
traveling waves,” IEEE Transactions on Smart Grid, vol. 10, no. 6, pp. 6404-6413, Nov. 2019, doi: 10.1109/TSG.2019.2904210.
[15] J. Sadeh, E. Bakhshizadeh, and R. Kazemzadeh, “A new fault location algorithm for radial distribution systems using modal
analysis,” International Journal of Power Electronics & Driver System, vol. 45, no. 1, pp. 271–278, Feb. 2013, doi:
10.1016/j.ijepes.2012.08.053.
[16] R. Liang, F. Wang, G. Fu, X. Xue, and R. Zhou, “A general fault location method in complex power grid based on wide-area
traveling wave data acquisition,” International Journal of Power Electronics & Driver System, vol. 83, pp. 213–218, Dec. 2016,
doi: 10.1016/j.ijepes.2016.04.021.
[17] E. Khoudry, A. Belfqih, J. Boukherouaa, and F. Elmariami, “Traveling wave-based fault location for power transmission lines
using morphological filters and clarke modal components International Journal of Electrical and Computer Engineering., vol. 10,
no. 2, pp. 1122-1134, Apr. 2020, doi: 10.11591/ijecev10i2.pp1122-1134.
[18] B. K. Chaitanya and A. Yadva, “Decision tree aided traveling wave-based fault section identification and location scheme for
multi-terminal transmission lines,” Journal Measurement, vol. 135, pp. 312-322, Mar. 2019, doi:
10.1016/j.measurement.2018.11.069.

 ISSN: 2502-4752
24
[19] L. Ye, D. You, X. Yin, K. Wang, and J. Wu, “An improved fault-location method for distribution system using wavelets and
support vector regression,” International Journal of Power Electronics and Energy System, vol. 55, pp. 467–472, Feb. 2014, doi:
10.1016/j.ijepes.2013.09.027.
[20] Q. Jia, X. Dong, and S. Mirsaeidi, “A traveling-wave-based line protection strategy against single-line-to-ground faults in active
distribution networks,” International Journal of Power Electronics and Energy System, vol. 107, pp. 403-411, May 2019, doi:
10.1016/j.ijepes.2018.11.032.
[21] L. Rui, F. Guoqing, Z. Xueyuan, and X. Xue, “Fault location based on single terminal traveling wave analysis in radial
distribution network,” International Journal of Power Electronics and Energy System, vol. 66, pp. 160-165, Mar. 2015, doi:
10.1016/j.ijepes.2014.10.026.
[22] L. Rui, P. Nan, Y. Zhi, and F. Zare, “A novel single-phase-to-earth fault location method for distribution network based on zero-
sequence components distribution characteristics,” International Journal of Power Electronics and Energy System, vol. 102, pp.
11-22, Nov. 2018, doi: 10.1016/j.ijepes.2018.04.015.
[23] L Rui, L. Chenglei, P. Nan, C. Menghan, and W. Fei. “Fault location for power grid based on transient traveling wave data fusion
via asynchronous voltage measurements,” International Journal of Power Electronics and Energy System, vol. 93, pp. 426-439,
Dec. 2017, doi: 10.1016/j.ijepes.2017.06.002.
[24] C. Zhou, Q. Shu, and X. Han, “A single phase earth fault location scheme for distribution feeder on the basis of the difference of
zero mode traveling waves,” International Journal of Power Electronics and Energy System, pp. 1-9, Dec. 2016, doi:
10.1002/etep.2298.
[25] S. Myint and W. Wichakool, “A traveling wave-based fault section and fault distance estimation algorithm for grounded
distribution systems,” in 2019 IEEE PES GTD Grand International Conference and Exposition. Asia, Bangkok, Thailand, pp.
472-477, Mar. 2019, doi: 10.1109/GTDAsia.2019.8715933.
[26] S. Myint and W. Wichakool, "Fault type identification method based on wavelet detail coefficients of modal current components,"
in 2018 IEEE 5th
International Conference on Smart Instrumentation Measurement and Application, Songkhla, Thailand, pp. 1-6,
Nov. 2018, doi: 10.1109/ICSIMA.2018.8688796.
BIOGRAPHIES OF AUTHORS
Shwe Myint received the B.E and M.E degrees in Electrical Engineering from
Mandalay Technological University (MTU), Mandalay, Myanmar in 2006 and 2014
respectively. She is currently pursuing the Ph.D. degree program in Electrical Engineering
Department at Prince of Songkla University, Hatyai Campus, Songkhla, Thailand. Her
current research interests include power system analysis, renewable energy, signal
processing and machine learning. She can be contacted at email: 5910130043@psu.ac.th.
Warit Wichakool received the Ph.D. degree in Electrical Engineering and
Computer Science from the Massachusetts Institute of Technology, Cambridge, MA, USA,
in 2011. He is currently working at Electrical Engineering Department at Prince of Songkla
University, Hatyai Campus, Songkhla, Thailand. His current research interests include
power system analysis, power converter design, power electronics circuits and applications
and renewable energy. He can be contacted at email: warit.wi@psu.ac.th.

A simple faulted phase-based fault distance estimation algorithm for a loop distribution system

Recommended

More Related Content

Similar to A simple faulted phase-based fault distance estimation algorithm for a loop distribution system (20)

More from IJEECSIAES (17)

Recently uploaded (20)

A simple faulted phase-based fault distance estimation algorithm for a loop distribution system