Upload
iaeme
View
110
Download
0
Tags:
Embed Size (px)
Citation preview
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
21
CHARACTERIZATION OF TRANSIENTS AND FAULT DIAGNOSIS IN
TRANSFORMER BY DISCRETE WAVELET TRANSFORM AND
ARTIFICIAL NEURAL NETWORK
M. Mujtahid Ansari1, Nilesh S. Mahajan
2, Dr. M A Beg
3
1, 2
Department of Electrical Engineering, SSBT’s COET, Bambhori. (MS) India 3Department of Electrical Engineering, Al-Qassim University, Buraidah, Kingdom of Saudi Arabia
ABSTRACT
The transients to which the transformer is mainly subjected are impact of high voltage and
high frequency wave arising from various causes. Switching magnetization and inter-turn faults are
also responsible for transients phenomena. To characterize and discriminate the transient arising
from magnetization and inter-turn faults are presented here. This characterization will give value
added information for improving protection algorithm. The detection method can provide
information to predict fault ahead in time so as that necessary corrective actions are taken to prevent
outages and reduce down time. The data is taken from different test results like normal
(magnetization) and abnormal (inter-turn fault) in this work; Discrete Wavelet Transform concept is
used. Feature extraction and method of discrimination between transformer magnetization and fault
current is derived by Discrete Wavelet Transform (DWT) Tests are performed on 2KVA,
230/230Volt custom built single phase transformer. The results are found using Discrete and
conclusion presented. The spectral energies of the wavelet components are calculated and then
employed to train Artificial Neural Network (ANN) to discriminate an internal fault from the
magnetizing inrush current. The results presented clearly show that the proposed technique can
accurately discriminate between an internal fault and a magnetizing inrush current in transformer
protection.
Keywords: Inrush Current, Internal Fault, Transients, Second Harmonic Component in Transformer,
Wavelet Transform and Artificial Neural Network.
I. INTRODUCTION
To avoid the needless trip by magnetizing inrush current, the second harmonic component is
commonly used for blocking differential relay in power transformers. The major drawback of the
INTERNATIONAL JOURNAL OF ELECTRICAL ENGINEERING &
TECHNOLOGY (IJEET)
ISSN 0976 – 6545(Print) ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35
© IAEME: www.iaeme.com/IJEET.asp Journal Impact Factor (2014): 6.8310 (Calculated by GISI) www.jifactor.com
IJEET
© I A E M E
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
22
differential protection of power transformer is the possibility for false tripping caused by the
magnetizing inrush current during transformer energization. [5] In this situation, the second
harmonic component present in the inrush current is used as a discrimination factor between fault
and inrush currents. In general, the major sources of harmonics in the inrush currents are
nonlinearities of transformer core; saturation of current transformers; core residual magnetization;
and switching instant. This work proposes new wavelet-based and artificial neural network methods
to identify inrush current and to distinguish it from inter-urn faults. [1-2]
Recently, Artificial Neural Network (ANN) techniques have been applied to power
transformer protection to distinguish internal faults from magnetizing inrush currents [6]. The main
advantage of the ANN method over the conventional method is the non algorithmic parallel
distributed architecture for information processing. In references [5] and [6], neural network based
schemes for protection of a single phase transformer have been investigated. However, the ANNs in
these existing studies are specific to particular transformer systems, and would have to be retrained
again for other systems. Moreover, the employed feature extraction techniques are based on time or
frequency domain signals, and not both time and frequency features of the signal; this is very
important for accurately distinguishing between an internal fault and inrush current.
II. NEED OF FREQUENCY INFORMATION
Often times, the information that cannot be readily seen in the time-domain can be seen in the
frequency domain like ECG signal (Electro Cardio Graph, graphical recording of heart's electrical
activity). The typical shape of a healthy ECG signal is well known to cardiologists. Any significant
deviation from that shape is usually considered to be a symptom of a pathological condition. This, of
course, is only one simple example why frequency content might be useful. Today Fourier
transforms are used in many different areas including all branches of engineering. Although FT is
probably the most popular transform being used (especially in electrical engineering), it is not the
only one. There are many other transforms that are used quite often by engineers and
mathematicians. Hilbert transform, short-time Fourier transform (more about this later), Wigner
distributions, the Radon Transform, and of course our featured transformation , the wavelet
transform, constitute only a small portion of a huge list of transforms that are available at engineer's
and mathematician's disposal. Every transformation technique has its own area of application, with
advantages and disadvantages, and the wavelet transform (WT) is no exception. For a better
understanding of the need for the WT let's look at the FT more closely. FT (as well as WT) is a
reversible transform, that is, it allows going back and forwarding between the raw and processed
(transformed) signals. However, only either of them is available at any given time. That is, no
frequency information is available in the time-domain signal, and no time information is available in
the Fourier transformed signal. The natural question that comes to mind is that is it necessary to have
both the time and the frequency information at the same time? The particular application and the
nature of the signal in hand. Over the years, various incipient fault detection techniques, such as
dissolved gas analysis and partial discharge analysis have been successfully applied to large power
transformer fault diagnosis. Since these techniques have high-cost and some are offline, a low-cost,
online internal fault detection technique for transformers using terminal measurements would be very
useful. [1-3]
A powerful method based on signal analysis should be used in monitoring. This method
should discriminate between normal and abnormal operating cases that occur transformers such as
internal faults, magnetizing inrush. There have been several methods, based on time domain
techniques, frequency domain techniques or time-frequency domain techniques. In previous studies,
researchers have used Fourier transform (FT) or windowed-Fourier transform. In recent studies,
wavelet transform based methods have been used for analysis of characteristics of terminal currents
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
23
and voltages .Traditional Fourier analysis, which deals with periodic signals and has been the main
frequency-domain analysis tool in many applications, fails in transient processes such as
magnetizing inrush and internal faults. The wavelet transform (WT), on the other hand, can be useful
in analyzing the transient phenomena associated with the transformer faults. Since the FT gives only
frequency information of a signal, time information is lost. Therefore, one technique known as
windowed FT or short-time FT (STFT) has been developed. However, the STFT has the limitation of
a fixed window width. So it does not provide good resolution in both time on other hand, WT
provide great resolution in time for high frequency component of signal and great resolution in
frequency for low frequency components of a signal. In a sense, wavelets have a window that
automatically adjusts to give the appropriate resolution. [2-3].
III. WAVELET APPLICATION
In recent years, researchers in applied mathematics and signal processing have developed
powerful wavelet techniques for the multi scale representation and analysis of Signals These new
methods differ from the traditional Fourier techniques Wavelets localize the information in the time-
frequency plane; in particular, they are capable of trading one type of resolution for another, which
makes them especially suitable for the analysis of non-stationary signals. One important area of
application where these properties have been found to be relevant is power engineering. Due to the
wide variety of signals and problems encountered in power engineering, there are various
applications of wavelet transform. These range from the analysis of the power quality disturbance
signals to, very recently, power system relaying and protection. The main difficulty in dealing with
power engineering phenomena is the extreme variability of the signals and the necessity to operate
on a case by case basis. Another important aspect of power disturbance signals is the fact that the
information of interest is often a combination of features that are well localized temporally or
spatially (e.g., transients in power systems). This requires the use of analysis methods sufficiently
which are versatile to handle signals in terms of their time-frequency localization. Our discussion is
organized into two main parts: (1) a discussion of the main properties of WT and their particular
relevance to power engineering problems and (2) a critical review of power engineering applications.
In Section II, we start by examining the properties of WT that are most relevant to power engineering
problems. We consider the primary power engineering applications, provide the reader with the
relevant background information, and review recent wavelet developments in these areas.
Time-Frequency Localization Wavelets are families of functions generated from one single function, called an analyzing
wavelet or mother wavelet, by means of scaling and translating operations. Some mother wavelets
are shown in Fig.1. The difference between these wavelets is mainly due to the different lengths of
filters that define the wavelet and scaling functions. Wavelets must be oscillatory, must decay
quickly to zero (can only be non-zero for a short period), and must integrate to zero. The scaling
operation is nothing more than performing “stretching” and “compressing” operations on the mother
wavelet, which in turn can be used to obtain the different frequency information of the function to be
analyzed. The compressed version is used to satisfy the high frequency needs, and the dilated version
is used to meet low frequency requirements. Then, the translated version is used to obtain the time
information of the function to be analyzed. In this way, a family of scaled and translated wavelets is
created and serves as the base, the building blocks, for representing the function to be analyzed. The
scaled (dilated) and translated (shifted) versions of the Daubechies mother wavelet are shown in
Fig.2. Daubechies wavelets belong to a special class of mother wavelets and actually are used most
often for detection, localization, identification and classification of power disturbance.
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
24
Fig.1: Four mother wavelets often used in wavelet analysis
Fig.2: Scaled and translated version of D4 wavelet
IV. EXPERIMENTATION AND DATA COLLECTION
The setup for experiments has a custom built 230V/230V, 2KVA, 50Hz, single-phase
transformer with externally accessible taps on both primary and secondary to introduce faults. The
primary winding and secondary winding has 272 turns respectively. The load on the secondary
comprises of static and rotating elements.
Data acquisition card by Tektronix Instruments was used to capture the voltages and current
signals. The Tektronix DSO TPS2014B, with 100MHz bandwidth and adjustable sampling rate
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
25
1GHz is used to capture the currents and voltages. The Tektronix current probes of rating 100mV/A,
input range of 0 to 70Amps AC RMS, 100A peak and frequency range 0 to 100Khz are used. These
signals were recorded at a sample rate of 10,000 samples/sec.
Different cases of inter turn short circuit are staged, considering the effect of number of turns
shorted on primary and secondary on load condition. Experimental set up is as shown in fig .3 and
fig.4 The current and voltage signals were captured for inrush and faulted condition, The captured
data are stored in excel sheet with the notations Vp-Primary voltage, Ip-Primary current ,Vs-
Secondary voltage, Is- Secondary current, Ts- Sampling time and FEQ-frequency of supply voltage
at captured instant
Fig.3: Experimental set up
Procedure for data collection
1. The magnetization current is captured at primary side.
2. Inter-turn faults are done on primary and secondary winding through contractor under load
condition.
3. The difference of primary and secondary is done sample by sample. The fifth channel is set in
Math function which directly gives differential current.
4. Current transformer and Voltage transformer are used to capture the current and voltage, The
analog signals are sampled at rate of 10000sample/sec. by Tektronix Digital Oscilloscope
(DSO).
The data is stored in excel sheets using Data Acquisition Card by Tektronix DSO.
V. WAVELET ANALYSIS
At the first stage an original signal is divided in to two halves of the frequency bandwidth,
and sent to both Low Pass Filter (LPF) and High Pass Filter (HPF). The coefficients of filter pairs are
associated with the selection of mother wavelet, the Daubechies Db-4type wavelet is used as mother
wavelet. Then the output of LPF is further cut in half of the frequency bandwidth and then sent to the
second stage, this procedure is repeated until the signal is decomposed to a pre-defined certain level.
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
26
If the original signal were being sampled at Fs Hz, the highest frequency that the signal could
contain, from Nyquist’s theorem, would be Fs/2 Hz. This frequency would be seen at the output of
the high pass filter, which is the first detail 1; similarly, the band of frequencies between Fs/4 and
Fs/8 would be captured in detail 2, and so on. The sampling frequency is taken to be 10 kHz and
Table I shows the frequency levels of the wavelet function coefficients.
Fig.4: Photo of practical setup
Fig.5: Implementation of DWT
Table I: Frequency levels of Wavelet Functions Coefficients
Decomposition
Level
Frequency
Components, Hz
D1 5000-2500
D2 2500-1250
D3 1250-625
D4 625-312.5
D5 312.5-156.25
A5 0-156.25
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
27
The waveforms of inrush and fault along with decomposition levels are shown.
Fig. 6: Wave form and decomposition Fig 7: Wave form and decomposition
levels of Inrush differential current levels of Primary fault differential current
Fig 8 (a) Wave form and decomposition levels of magnetization inrush differential Current along
original signal
Fig 8 (b) Wave form and decomposition levels of primary fault differential Current along original
signal
From visual inspection of fig. 6 and fig.7 characterize the transient and discriminate between
magnetization inrush and interturn fault. In these two figures d1 and d2 are nearly similar and
discrimination is difficult. By keen observation at decomposition level d3 and d4 of figure 6, the
wavelet coefficients are corresponds to magnetization peek. Whereas in fig 7, the large wavelet
coefficients for the decomposition level d3 to d5 appears at the instant of swiching and attenuates
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
28
with the length. Fig 8(a) and fig 8(b) shows the each decomposition level along with original signal
for readily justification of above lines.
Above observations are no longer valid because these decomposition levels signature are
changing with instant of switching as well as instant of fault. To arive some results, the energies of
d3, d4, and d5 decomposed levels for one cycle are calculated and ANN is used for classifications of
transients. Since the d1 and d2 decomposition level are very much mixed to each others, hence d1
and d2 levels are not considered.
Flow Chart for ANN
Fig 9: Flow chart
Description of Flow Chart
1. Captured primary and secondary current by Data acquisition card and Tektronix Instruments.
2. Calculate the differential current Id= Ip- Is
3. If the differential current value is less than threshold value, then take next samples otherwise
proceed with that vale.
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
29
4. Calculate DWT of differential current whose value is above threshold value.
5. Obtain the energies of decomposed levels from d3 to d5.
6. The energy of decomposed levels d3 to d5 is given to ANN as input data to discriminate the
fault and inrush that is healthy condition.
7. Energies are calculated by formula � � ∑ ����� ��
8. If ANN out put is discriminated as fault, then issue trip signal, otherwise proceed further.
Neural Network as a classifier Because of inherent capability of ANN to classify even though the data are non linearly
separable. Here the data are not classified by other method hence ANN, Multilayer Perceptron
(MLP) Neural Network is trained very systematic manner for classification. Energies of decomposed
level D3, D4, and D5 for one cycle were used as a input to the networks. The cluster diagram is
shown in Data fig. 10 It shows that this is non linearly separable data. For generalization the
randomized data is fed to the network and is trained for different hidden layers. Various training
methods are used for training the network and average minimum MSE on training and testing data is
measured. It is observed that MLP with single hidden layer gives better performance. The number of
Processing Elements (PEs) in the hidden layer is varied. The network is trained and minimum MSE
is obtained.
For all training methods it is assumed that learning rate LR=0.8, momentum MM=0.6, data
used for training purpose TR= 50 %, data used for cross validation CV = 20% and data used for
testing purpose TS= 30%. With these assumptions the variation of average mse and percent accuracy
for both Inrush and fault with respect to number of processing elements in the hidden layer is
obtained. Fig 7.9(a) Shows Variation of average minimum MSE with respect to number of
processing elements in the hidden layer, for one of the training method ‘ traincgf ’ of Conjugate
Gradient.
Fig 10: Cluster diagram, showing mixed energies of d3 to d5 levels.
Conjugate Gradient Method
METHOD: ‘ traincgf ’ (i) MSE Vs Number of PE’S
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
30
Fig 11 (a): Variation of average minimum MSE with number of processing elements in the
hidden layer
Also average classification accuracy with respect number of processing elements in the
hidden layer, for the same method is obtained and plotted in fig 11 (b).(ii) % ACCURACY Vs
Number of PE’S.
Fig 11 (b): Variation of % Accuracy with number of processing elements in the hidden layer
It is observed from fig 7.9(a) and fig 7.9 (b) that, for this training method eleven number of
processing elements in the hidden are required to get minimum MSE(0.0218) and maximum percent
accuracy for Inrush ( 95% ) ,for fault ( 99% ).Similarly Variation of average minimum MSE and
Variation of percent accuracy with respect to number of processing elements in the hidden layer for
another training method ‘traincgp’ of Conjugate Gradient (CG) obtained, as shown in fig 11 (a) and
fig 11 (b).
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
31
VI. COMPARISON OF ALL METHOD
Variation of % Accuracy for all training methods with particular number of processing
elements in the hidden layer are plotted from fig 12 to fig 14
Fig 12: Variation of % Accuracy with 13 number of processing elements in the hidden layer
Fig 13: Variation of % Accuracy with 14 number of processing elements in the hidden layer
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
32
Fig 14: Variation of % Accuracy with 11 number of processing elements in the hidden layer
VII. COMPARISON of Analysis
Table II: Comparison of MSE for Different Methods
PE'S
MSE FOR DIFFERENT METHODS
traincgf traincgp traincgb trainscg
traincgf
(QN) trainmoss trainlm
1 0.092 0.121 0.115 0.087 0.077 0.0042 0.70
2 0.072 0.244 0.096 0.025 0.112 0.0884 0.035
3 0.073 0.064 0.058 0.033 0.061 0.0938 0.003
4 0.118 0.087 0.044 0.020 0.047 0.0628 0.012
5 0.061 0.021 0.032 0.019 0.019 0.029 0.17
6 0.074 0.040 0.022 0.004 0.091 0.0229 0.006
7 0.094 0.054 0.026 0.050 0.043 0.0464 0.033
8 0.042 0.014 0.032 0.010 0.025 0.0217 0.006
9 0.091 0.03 0.014 0.030 0.022 0.0407 0.055
10 0.155 0.032 0.009 0.006 0.033 0.0396 0.001
11 0.021 0.026 0.031 0.036 0.037 0.0272 0.003
12 0.093 0.010 0.005 0.003 0.024 0.0391 0.0109
13 0.132 0.012 0.022 0.043 0.022 0.0336 0.0007
14 0.021 0.054 0.041 0.027 0.008 0.0147 0.021
15 0.039 0.023 0.029 0.013 0.050 0.0135 0.0001
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
33
Table III: Comparison of PE Vs Methods for % Accuracy
PE’S
PERCENT ACCURACY OF INRUSH AND FAULT FOR DIFFERENT
METHODS
traincgf traincgp traincgb Trainscg traincgf(
QN)
trainoss trainlm
In F In F In F In F In F In F In F
1 75 96 68 95 71 99 75 92 65 100 80 93 71 99
2 70 100 88 66 66 100 95 99 65 100 70 95 95 98
3 60 99 66 100 90 96 95 99 65 100 70 100 96 99
4 70 88 71 100 88 100 91 99 88 98 90 98 93 99
5 63 100 93 99 93 99 95 98 95 99 91 99 93 99
6 75 97 88 96 93 99 95 99 80 97 93 99 95 100
7 75 97 73 97 95 95 81 100 80 98 95 95 93 99
8 90 98 95 99 95 99 93 99 91 97 93 99 90 99
9 78 75 95 97 96 98 96 98 90 100 80 100 95 99
10 60 100 95 99 98 99 93 99 88 99 88 96 95 99
11 95 99 90 97 91 98 90 97 70 98 95 99 93 98
12 66 98 95 99 93 97 97 99 95 99 93 99 95 99
13 95 97 95 99 95 99 93 99 95 95 95 99 93 99
14 95 96 93 96 95 99 95 98 90 99 93 99 96 98
15 95 98 95 99 95 99 93 99 86 97 93 99 100 100
VIII. DISCUSSIONS ON RESULTS
It is observed from graphs, for every training method the number of Processing element (PE)
in the hidden layer are required to get minimum value of Mean Square Error(MSE) corresponds to
maximum percentage accuracy for discriminating the magnetization inrush and inter-turn fault.
For each method Learning Rule(LR), Momentum(MM),Training data(TR),Cross Validation
data(CV) and Testing data(TS) are kept same. The values are LR= 0.8, MM = 0.6, TR = 50%, CV
=20% and TS = 30%. This strategy is applied to all method mentioned in case study. This strategy is
applied to Conjugate Gradient methods like Traincgf, Traincgp, Traincgb and Traincg. It is found
that up to 15 number of processing element none of method is able to characterize the desired result.
Same exercise is done for Quasi Newton method .The two methods of Quasi Newton such as
Traincgf and Trainnos, are not capable to discriminate the magnetization inrush and fault current
correctly. It is found that by Levenbreg – Marquardt Trainlm method gives 100% classification with
fifteen numbers of neurons or processing element in hidden layer for discrimination and characterize
the magnetization inrush and fault current.
The above discussions become clearer by table no.2 which shows the comparison of MSE for
different methods. It shows that for trainlm method with LR=0.8, MM=0.6, TR= 50 %, CV = 20%
and TS= 30% the MSE is minimum among all other methods for fifteen number of processing
element. The minimum value of MSE obtained is 0.0001.
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
34
Similarly table no. 3 shows the comparison of number of Processing Elements (PE) verses
methods for percentage accuracy. It also shows that that for trainlm method with LR=0.8, MM=0.6,
TR= 50 %, CV = 20% and TS= 30% the percent accuracy with respect to 15 number of processing
elements is 100%.
It is found that the accuracy of train network for correctly classifying the transient due to
magnetization inrush and fault is 93%.
IX. CONCLUSION
This paper discussed efforts to characterize transients for transformers, resulting from various
types of faults. The experiments were conducted on a single-phase transformer model. The data were
obtained from experiments for several cases related to the transformer operation such as magnetizing
inrush, external system short circuits, internal short circuits. The data were analyzed using discrete
wavelet transforms (DWTs). Same exercise is done for Quasi Newton method .The two methods of
Quasi Newton such as Traincgf and Trainnos, are not capable to discriminate the magnetization
inrush and fault current correctly. It is found that by Levenbreg – Marquardt Trainlm method gives
100% classification with fifteen numbers of neurons or processing element in hidden layer for
discrimination and characterize the magnetization inrush and fault current. The characteristics of the
cases and differences between cases are presented. The results show great potential for using this
method for predictive maintenance and maintaining reliability of transformers. Future work will
investigate using characteristics of fault data with an intelligent method such as neural networks for a
discrimination process and life estimation of transformer.
X. REFERENCES
[1] M Mujtahid Ansari, S R Paraskar and Dr G M Dhole, “Characterization of Transients and
Fault Diagnosis in Transformers Using Discrete Wavelet Transform” International Journal of
Electrical Engineering and Technology (IJEET), Vol. 4, Issue 5, September-October 2013.
Page 87-95.
[2] M Mujtahid Ansari, Patil B P and Dr G M Dhole, “Application of Signal Analysis for Fault
Diagnosis in Transformer by Discrete Wavelet Transform” International Journal of
Engineering Research & Technology (IJERT), Vol. 2 Issue 2, February- 2013. Page 648.
[3] Karen L. Butler-Purry and Mustafa Bagriyanik.” Characterization of Transients in
Transformers Using Discrete Wavelet Transforms” IEEE Transactions on Power system,
Vol. 18, No. 2, May 2003. Page 648.
[4] Jawad Faiz, and S. Lotfi-Fard,” A Novel Wavelet-Based Algorithm for Discrimination of
Internal Faults From Magnetizing Inrush Currents in Power Transformers” IEEE
Transactions on Power Delivery, Vol. 21, No. 4, October 2006 Page 1989.
[5] Omer A S Youssef, A Wavelet Base Technique for Discrimination between fault and
magnetization inrush current in transformer”, IEEE Transactions on Power Delivery, Vol. 18
No.1 January2003. Page 170.
[6] S A Saleh & M A Rahman.”, Modeling and protection of three phase transformer using
wavelet packet transform,” IEEE Transactions on Power Delivery, Vol 20 No.2 April 2005
Page 1273.
[7] Peilin Mao and Raj K Aggarwal ,” A noval approach to classification of the transient
phenomena in power transformers using combine Wavelet Transform and Neural Network,”
IEEE Transactions on Power Delivery , Vol 16 No.4 October 2001 Page 654.
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 – 6545(Print),
ISSN 0976 – 6553(Online) Volume 5, Issue 8, August (2014), pp. 21-35 © IAEME
35
[8] Sami G Abdulsalam,” Analitical study of transformer inrush current transient and it’s
application,” International conference on power transient in Montreal, Canada on June
19-23, 2005 paper PST05-140.
[9] Jialong Wang and Randy Hamilton, Analysis of transformer inrush current and comparison
of harmonic restrain method in transformer protection, Basler Electric Company,
978-1-4244-4108( c ), 2008 IEEE.
[10] H. L. Willis, G. V. Welch, and R. R. Schrieber, Aging Power Delivery Infrastructures. New
York: Marcel Dekker, 2001.
[11] H. Wang and K. L. Butler, “Modeling transformers with internal incipient faults,” IEEE
Trans. Power Delivery, vol. 17, pp. 500–509, Apr. 2002.
[12] A.V.Padmaja and V. Usha Reddy, “Application of Wavelet Transform for Monitoring Short
Duration Disturbances in Distribution Systems” International Journal of Electrical
Engineering and Technology (IJEET), Vol. 3, Issue 1, January-June 2012. Page 112 - 122.
AUTHOR’S DETAIL
M. Mujtahid Ansari received his B.E.( Electrical Engg) and M.E.(Electrical Power
System) from the S.G.B.University, Amravati, India in year 1996 and 2009
respectively . He is member of IE, ISTE and IACSIT. He has joined SSBT’s College
of Engineering and Technology, Jalgaon in the year 2000 where he is working as
Assistant Professor in Electrical Engineering Department. He has published three
papers in international journal and one text book. He is chairman Board of studies in
Electrical Engineering and Instrumentation Engineering under the faculty of Engineering and
Technology, North Maharashtra University, Jalgaon (MS) India.
Nilesh S. Mahajan had completed his M. E. in Electrical Power Systems from
Government College of Engineering, Aurangabad. He is working as an Assistant
Professor in the Department of Electrical Engineering in SSBT’s College of
Engineering & Technology, Bambhori, Jalgaon. He has published two papers in
international journal.
Dr. M. A. Beg received his B.E. and M.E. from the S.G.B.Amravati University
,Amravati, India in year 1992 and 2001 respectively in Electrical Power system
Engineering and completed his Ph.D. from the S.G.B.University, Amravati, in Power
Quality. In the year 1990, he joined S.S.G.M. College of Engg. Shegaon, where he
had worked Assistant Professor in Electrical Engg. He also served as head and
academic dean in the same college. Presently he is working as professor in department
of Electrical Engineering, Al-Qassim University, Buraidah, Kingdom of Saudi Arabia, His present
research interest includes power quality monitoring and signal processing technique applications in
power systems. He is member of IACSIT and ISTE. He has published research papers in different
International and National Journals and Conference Proceedings He is reviewer of International
Journal “Electrical Power Components & System (EPCS), Taylor & Fransis Group Publication,
England, International Journal EPSR (Elsevier).