Distribution System Faults Classification And

Location Based On Wavelet Transform

MukeshThakre, Suresh Kumar Gawre & Mrityunjay Kumar Mishra

Electrical Engg.Deptt., MANIT, Bhopal.
E-mail : [email protected], [email protected] & [email protected]

Abstract- Electrical distribution system, which is the equipment failure or mis-operation which results in the
largest portion of networks, acts as a final power delivery loss of power, voltage dips on the power system.
path to the end users. Fault classification and location is
very important in power system engineering in order to Fault location and distance estimation is very important
clear fault quickly and restore power supply as soon as issue in power system engineering in order to clear fault
possible with minimum interruption. Hence, ensuring its quickly and restore power supply as soon as possible
efficient and reliable operation is an extremely important with minimum interruption. This is necessary for
and challenging task. With availability of inadequate reliable operation of power equipment and satisfaction
system information and presence of high impedance faults, of customer. The following definitions related to this
locating faults in a distribution system pose a major
challenge to the utility operators. Most of the previous
study are:
work on fault identification and location concentrated on Fault: A fault is an unpermitted deviation of at least one
estimating the status of circuit breakers and relays with characteristic property (feature) of the system from the
aid of some algorithmic approaches. In this paper, a fault
acceptable, usual, standard condition.
identification and location technique using wavelet multi-
resolution approach for radial distribution systems are Disturbance: undesired and uncontrollable interference
proposed. The current measurements at the substation, acting on the system.
available in the distribution network have been utilized
and the effectiveness of the proposed approach is Fault detection: Finding if there is any fault in the
demonstrated on 7-node three-phase test systems. Also in system and also the time of the fault.
this work distribution systems model was developed and
simulated using power system block set of MATLAB to Fault Isolation: Determining the type and location of
obtain fault current waveforms. The waveforms were the fault.
analyzed using the Discrete Wavelet Transform (DWT)
toolbox by selecting suitable wavelet family to obtain the
Fault identification: Determining the magnitude and
pre-fault, during-fault and post-fault coefficients for time-variant behavior of the fault.
estimating the fault classification. It was estimated and Fault diagnosis: Fault diagnosis is determining which
achieved using Daubechies ‘db5’ discrete wavelet
fault has occurred, in other words, determining the cause
transform.
of the observed out-of-control status.
Key Words-Distribution network, Fault identification, Multi-
resolution analysis, Power system faults, and Wavelet Monitoring: Observing and recording the progress of
transform. different variables in a process over a period of time.
Error: An error is the deviation of the measured value
I. INTRODUCTION from the actual or true value.
An important objective of all the power systems is Failure: A failure is a permanent interruption of a
to maintain a very high level of continuity of service, system„s ability to perform a required function under
and when abnormal conditions occur, to minimize the specified operating conditions.
outage times. It is practically impossible to avoid
consequences of natural events, physical accidents,

ISSN (Print) : 2319 – 2526, Volume-2, Issue-4, 2013

86
 International Journal on Advanced Computer Theory and Engineering (IJACTE)

Malfunction: A malfunction is an intermittent 4. Simplicity

irregularity in the fulfillment of a system„s desired
5. Economics
function.
Various factors like availability of circuit breakers
Natural events can cause short circuits i.e. faults
and fault indicators affect the protection system. The
which can either be single phase to ground or phase to
abnormalities, faults and intolerable conditions must
phase or phase to phase to ground or a three phase fault.
provide a distinguishable difference between from the
Most faults in an electrical system occur with a network
normal operating or tolerable conditions. In this works,
of overhead lines are single-phase to ground faults
voltage and current waveforms at both the sending and
caused due to lightning induced transient high voltage
receiving end are used for detection and classification of
and from falling trees. In the overhead lines, tree contact
the problem. Any significant changes in these quantities
caused by wind is a major cause for faults. The
provide a means to detect abnormal conditions and
appropriate percentages of occurrences various faults
hence employed for relay operation. So the important
are listed below [1]:
step is to first determine the quantity and the associated
Single line to ground fault – 70-80% difference value that would separate the faulted and
non-faulted waveforms. There could be two kinds of
Line-Line to ground fault - 10-17%
problems; a faulted waveform could be detected as a
Line-Line fault – 8-10% normal wave. This results in delivering the power
exceeding the currents and frequency limits to the
Three phase – 2-3%
customers leading to equipment failures. The other error
When faults occur in the power system, they could arise if a normal waveform is detected as a faulted
usually provide significant changes in the system wave. In this case, a certain portion of the power system
quantities like over-current, over or under-power, power is unnecessarily isolated from the system.
factor, impedance, frequency and power or current In general, the first step in the power system
direction. The most common and also the one used in
relaying algorithms is the detection of fault and the next
this work is the over-current and so over-current
step is classification. This work uses a combination of
protection is widely used.
Wavelet Transforms and multi-resolution detecting and
Power system protection is the art of applying and classifying power system faults. The Objective of this
setting up relays or fuses or both, to provide maximum work is to classify the faults according to the following
sensitivity to faults and abnormal conditions and also to parameters:
avoid false alarms during normal operating conditions.
1. Fault type
So it is desirable that a correct decision be made by the
protective device as to whether the trouble is an 2. Fault Location
abnormal condition or just a transient which the system
The fault cases are classified as one of ten different
can absorb and return to normal working condition. The
types of faults viz. three Single line to ground faults,
protective relays are more of a preventive device which
three Line to Line faults, three Line to Line to ground
operates after a fault has occurred and it helps in
faults and a three phase fault. The fault location is an
minimizing the duration of trouble and limiting the
important parameter especially in high voltage power
damage, outage time and related problems. For the
systems. The knowledge of fault location leads to high
system to operate properly, it is necessary to isolate the
speed fault clearance as well as improved transient
trouble area quickly with a minimum number of system
stability.
disturbances. Both failure to operate and incorrect
operation can result in major system upsets involving II. HISTORICAL REVIEW : FROM FOURIER
increased equipment damage, increased personnel ANALYSIS TO WAVELET ANALYSIS
hazards and possibly long interruption of service.
The history of wavelets begins with the
The fault detection and classification system which development of the traditional Fourier transform (FT)
is discussed in this work is a part of the protective (Donald and Walden, 2000), which is widely applied in
device. The five basic qualities of a protective relays [1] signal analysis and image processing. Fourier transform
are breaks down a signal into the sum of infinite series of
1. Reliability sines and cosines of different frequencies, in another
word; FT is a mathematical technique for transforming
2. Selectivity our view of the signal from time-based to frequency-
3. Speed of operation based. Fourier transform is very effective in problems
dealing with frequency location. However, time

ISSN (Print) : 2319 – 2526, Volume-2, Issue-4, 2013

87
 International Journal on Advanced Computer Theory and Engineering (IJACTE)

information is lost during the process of transforming to The application of wavelet transform in engineering
frequency domain. This means that although we might areas usually requires discrete wavelet transform
be able to determine all the frequencies present in a (DWT), which implies the discrete form of t, a, b eqn.
signal, we do not know when they are present. In the The representation of DWT can be written as:
time series process data, the most important part of the
signal is the transient characteristics: drift, trends, and
abrupt changes, and FT is not suited to detect them.
In an effort to improve the performance of the FT,
the short time Fourier transform (STFT) has been Where original a and b parameters in eqn are changed to
developed in signal analysis. STSF compromises be the functions of integers m, n. k is an integer variable
between the time and frequency based views of a signal and it refers to a sample number in an input signal.
by examining a signal under a fixed time window. The A very useful implementation of DWT, called multi-
drawback of STSF is that the time window is fixed and resolution analysis, is demonstrated in Fig. below. The
same for all the frequencies. Many signals require a original sampled signal x(n) is passed through a high
more flexible approach; the window size is required to pass filter h(n) and a low pass filter l(n). Then the
vary according to the frequency. outputs from both filters are decimated by 2to obtain the
detail coefficients and the approximation coefficients at
Wavelet analysis or wavelet transform is close in
level 1 (D1 andA1). The approximation coefficients are
spirit to the Fourier transform, but has a significant
then sent to the second stage to repeat the procedure.
advance. It applies a windowing technique with
Finally, the signal is decomposed at the expected level.
variable-sized regions, a shorter time interval is used to
analyze the high frequency components of a signal and a Detail1 (D2)

longer one to analyze the low frequency components of Detail 2 (D2)

h(n) 2
the signal. Wavelet analysis is very effective for dealing
x(n)
with local aspects of a signal, like trends, breakdown h(n) 2
Detail 3 (D3)

points, and self similarity. Furthermore, wavelet analysis

h(n) 2
is capable of removing noise from signal and compress l(n) 2

signal.
l(n) 2 Approximation 3 (A3)

III. WAVELET TRANSFORM l(n) 2

Wavelet analysis is a relatively new signal

processing tool and is applied recently by many
researchers in power systems due to its strong capability Fig.1. The idea of wavelet multi-resolution analysis
of time and frequency domain analysis. The two areas
with most applications are power quality analysis and
power system protection. In the case shown in Fig., if the original sampling
frequency is F, the signal information captured by D1 is
The definition of continuous wavelet transform (CWT) between F/4 and F/2 of the frequency band. D2 captures
for a given signal x(t)with respect to a mother wavelet the information between F/8 and F/4. D3 captures the
(t) is: information between F/16 and F/8, and A3retains the
rest of the information of original signal between 0 and
F/16. By such means, we can easily extract useful
information from the original signal into different
frequency bands and at the same time the information is
Where a, is the scale factor and b is the translation matched to the related time period. An example, given
factor. in Fig., illustrates the procedure. The originalsignal is
For CWT, t, a, b are all continuous. Unlike Fourier one cycle of a post-fault current signal, as shown in Fig.
transform, the wavelet transform requires selection of a We use db5wavelet to make a 5 level decomposition.
mother wavelet for different applications. One of the The reconstructed versions of each detail and the
most popular mother wavelets for power system approximation are shown in Fig. The information of
transient analysis found in the literature is Daubechies‟s original signal is clearly represented at each frequency
wavelet family. In the new scheme, the db5 wavelet is band. The original signal can be reconstructed by adding
selected as the mother wavelet for detecting the short up those wavelet signals at the same sample point. The
duration, fast decaying fault generated transient signals. wavelet tool box in MATLAB provides a lot of useful
techniques for wavelet analysis [8].

ISSN (Print) : 2319 – 2526, Volume-2, Issue-4, 2013

88
 International Journal on Advanced Computer Theory and Engineering (IJACTE)

The wavelet transform contains a low pass filter (only levels. Experimentally, it is found that different types of
obtaining low frequencies), which is denoted by l(n), faults generate different patterns of detail signals.
and a high pass filter (only obtaining high frequencies), Looking at the pattern of fault levels a set of rule base
which is denoted by h(n). At each level, the original have been worked out to identify ten different types of
signal, s, passes through two both low and high pass faults (i.e. single phase to ground faults, line to line
filters and emerges as two signals, which is detail faults, double line to ground to faults and three-phase
coefficients dn, and approximation coefficients an. The symmetrical fault). A simple rule base generated to
terms “approximation” and “detail” are named by the identify the nature of fault on the distribution line in
fact that an. is the approximation of an-1corresponding to case of 7-node system is presented below:
the “low frequencies” of an-1, whereas the detail dn takes
Rule 1: Set the absolute values of fault levels for phases
into account its “high frequencies”. Wavelet coefficients
a, b and c, are Ka, Kb, and Kc. And assuming the fault
at various frequencies reflect the signal variations at
values are K1, K2, K3.
those frequencies and corresponding times. Clearly, at
the kth step of this partitioning procedure, the original Rule 2: The notation for the system conditions are as:
signal, s, is expressed by Eq. “1”denotes faulty conditions in particular phases, and
“0” denotes no-faulty condition.
𝑠 = 𝑎𝑘 + 𝑑𝑘 + 𝑑𝑘−1 + … … + 𝑑1
Rule 3: Use Comparator for compare the fault levels in
With ak. representing the smooth signals referring to the
the digital form “0” or “1”
time scale 2k and dnis the detail of time series with the
time scale located in the interval [2k-1, 2k]. Rule 4: If Ka >K1then fault type is L-G.
Rule 5: If Ka >K1,Kb >K2 ,then fault type is LL-G.
Rule 6: If Ka >K1,Kb >K2 ,Kc>K3,then fault type is
LLL-G.
Rule 7: If Ka >K1,Kb >K2,and Ka − Kb >∈then
faulttype is AB (i.e.LL), and similarly for other LL
faults.

V. PROPOSED FAULT LOCATION ALGORITHM

The objective of the fault location algorithm is

to identify the faulty section by comparing the sharp
variation values of the detail signals in different phases.
The method proposed for fault location assumes that the
fault is identified using the method discussed in
previous section. Prior knowledge of the type of fault is
a requirement of the algorithm. The fault phenomena
involve large disturbances in the system and the
parameter selected for the fault location is summation of
Fig.2. Wavelet multi-resolution analysis for a fault third level output of the wavelet transform. Hence, in the
signal presence of the 'noisy data' found in the real time
recordings, there will be only minor variations in the
IV. PROPOSED FAULT IDENTIFICATION SCHEME
summation value of the third level output as the
Fault identification is to determine the process magnitude of disturbances created by faults are always
variables that are responsible for the fault and thus the large as compared to the noisy data. Also, noise consists
information can be used to diagnose the fault. It will be of high frequency signals, whereas the third level output
helpful to focus on the subsystems where the fault of the wavelet transforms contain signals only in the
occurred in a large plant. It not only helps to diagnose range of 60 - 200 Hz i.e. it includes second and third
the fault more efficiently but also saves time and money. harmonic contents, which are predominant in case of
faults. Therefore, the noise signals get filtered out from
For identification of fault type, several test cases for the third level output. This makes the fault location
both 7-node systems are simulated. To generate a setof technique very effective. Variable calculated and used
rule base for each of the cases considered, the MRA of for fault location are:
the current signature obtained from main substation
feeder are used to calculate the absolute values of fault

ISSN (Print) : 2319 – 2526, Volume-2, Issue-4, 2013

89
 International Journal on Advanced Computer Theory and Engineering (IJACTE)

𝑁
Ta,b,c= 𝑘 =1 𝑑𝑚 (𝑘) -4
Phase "b" Reconstructed detail at level 1
x 10
2

VI. SIMULATION AND RESULTS 1.5

For identification and location of faulty section, 1

several cases covering all the line sections (6 in case of 0.5

7-node system), different type of faults are studied. For
each case, the summation of third level coefficients of 0

MRA viz. Ta,b,c, are determined. It is found that for each -0.5
section, these values are different. However, Ta,b,c values
are find for different fault type (LG, LL, LLG or -1

LLLG). The average value of Ta,b,c, obtained for the 7- -1.5

node systems are given in Tables I.
200 400 600 800 1000 1200 1400 1600 1800 2000
The effectiveness of the proposed algorithm is also
tested for the fault cases in the line section or at buses (b)
for which the a , band c coefficients are not determined
for the rule generation. It is found that for both 7-node x 10
-4 Phase "c" Reconstructed detail at level 1

systems the method successfully locates the faulty 1.5

section in all the cases except for a case when the fault
1
occurs in the section (bus no. 3 and 4 in 7-node).
0.5
B3 B5

B1 B2
-0.5

B4 B6 -1

Substation
-1.5

200 400 600 800 1000 1200 1400 1600 1800 2000
B7

(c)
Fig. 3. Single line diagram of the 7-node distribution Fig. 4. Wavelet coefficients obtained from MRA of 3-
system phaes currents of LG ('AG fault) for (a) Phase A, (b)
Table-I Fault identification and Localizations Phase B, and (c) for Phase C, respectively.
Bus No. Fault Type Distance from Fault Currents with LG Fault
substation (in Km) 18

B-3 A-G (i.e LG) 9 16

B-4 A-G (i.e LG) 8 14

Phase 'a'
Phase 'b'
12 phase 'c'
Phase "a" Reconstructed detail at level 1
10
0.3
8

0.2 6

0.1 4

2
0
0

-0.1
-2
200 400 600 800 1000 1200 1400 1600 1800 2000
-0.2

-0.3 Fig. 5.The fault current value with LG (i.e. AG faults).

200 400 600 800 1000 1200 1400 1600 1800 2000

(a)

ISSN (Print) : 2319 – 2526, Volume-2, Issue-4, 2013

90
 International Journal on Advanced Computer Theory and Engineering (IJACTE)

IV. CONCLUSIONS
[5] S. Santoso, E. J. Powers, W. M. Grady, and P.
In this paper, the wavelet Multi-Resolution Hofmann, “Power quality assessment via wavelet
Analysis(MRA) has been applied to identify and locate transform analysis,” IEEE Trans. Power
the fault in the radial distribution system and is tested on Delivery, vol. 11, no. 2, pp. 924–930, Apr. 1996.
7-nodedistribution systems. Wavelet Transform was
[6] A. H. Osman and O. P. Malik, “Transmission line
used to extract distinctive features in the input signals.
distance protection based on wavelet transform,”
The MRA approach is found to be very effective in
IEEE Trans. Power Delivery, vol. 19, no. 2, pp.
identifying various types of fault (LG, LL, LLG, and
515–523, Apr. 2004.
LLLG) and also in locating the faulty section. The
method utilizes a set of quite simple rule base approach [7] O. Youssef, “New algorithm to phase selection
for identification of faults and only the average value of based on wavelet transforms,” IEEE Trans.
resolution of the substation current signal. Hence, the Power Delivery, vol. 17, no. 4, pp. 908–914, Oct.
method is quite simple to adopt and extremely fast for 2002.
the fault identification and location.
[8] Wavelet Toolbox User‟s Guide, The Mathworks
Inc., Natick, MA, June 2005.
V. REFERENCES
[9] M. M. Saba, R. Das, P. Verho, and D. Novosel,
[1] J. Lewis Blackburn, Protective Relaying
"Review of Fault Location Techniques for
Principles and Applications, Second Edition.
Distrbution Systems," Power Systems
[2] U. D. Dwivedi, s. N. Singh, S. C. Srivastava, “A andCommunication Infrastructure for the future,
Wavelet Based Approach for Classification and Beijing, Sep 2002.
Location of Faults in Distribution Systems”,
[10] S. MalIat, "A theory for multiresolution signal
IEEE, 2008, pp.-1-6.
decomposition: the wavelet representation," IEEE
[3] C. H. Kim and R. Aggarwal, “Wavelet transforms Trans on Pattern Analysis and
in power systems Part 1: General introduction to MachineIntelligence, vol. 11, no. 7, pp. 674-693,
the wavelet transforms,” Power Engineering 1989.
Journal, vol. 14, no. 2, pp. 81–88, Apr. 2000.
[11] O. Chaari, M. Meunier, and F. Brouaye,
[4] C. H. Kim and R. Aggarwal, “Wavelet transforms "Wavelets: a new tool for resonant grounded
in power systems Part 2: Examples of application power distribution systems relaying," IEEE Trans
to actual power system transients,” Power on Power Delivery, vol. 11, no. 3, pp. 1301-1308,
Engineering Journal, vol. 15, no. 4, pp. 193–202, 1996.
Aug. 2001.



ISSN (Print) : 2319 – 2526, Volume-2, Issue-4, 2013

91