More Info

                     Muhammad Azizi Abdul Rahman; Halimatun Hashim; P.S. Ghosh

          Electrical Engineering Department, College of Engineering, Universiti Tenaga Nasional


       Monitoring the health of power transformer is important for the reliability of electrical power
       supply. Conventional tests carried out on power transformers can only detect damage of
       permanent nature. Frequency Response Analysis (FRA) is found to be a useful tool for
       reliable detection of incipient mechanical fault in a transformer. There are various methods
       of evaluating the frequency spectrum to confirm the presence of an incipient fault. This
       paper aims to detect and identify the type of faults existing in the windings and core of
       132/33kV, 90MVA power transformer based on FRA data using signal processing
       techniques, taking into consideration the recommended frequency range specified by Doble
       Engineering Company. Thus, it is found that the proposed technique is very efficient and
       effective in analyzing and detecting accurately the type of mechanical fault existing in a
       power transformer.

1.       INTRODUCTION                                      reference     measurement.     The     reference
                                                           measurement can be from the same winding,
Frequency Response Analysis, normally known                another phase of the same transformer, or from a
as FRA, is a condition monitoring technique for            sister transformer. Any changes in resonant
expensive power equipment. It measures the                 frequencies may indicate the occurrence of
frequency response of passive element (RLC) of             faults. Fault such as short-circuited turns,
the apparatus; for example, in a power                     mechanical damage to winding and core, and
transformer, it measures the impedance of the              loose turns are detectable through this method
winding over a wide range of frequencies. The              [5,6].
main purpose of this test is to diagnose any
incipient fault in a power transformer. Results            Currently, the research on FRA has primarily
acquire from the apparatus are compared with a             directed into two. The first is to find the
reference data set. Differences may be used to             techniques or methods on how to obtain FRA
analyze the type and location of fault. FRA is a           result. This involves the development of
novel and reliable technique that has a number of          software and hardware in order to get the
very sensitive analyzing methods. It can detect            information. The second group of researchers is
minor and major faults. The main advantage of              working on analyzing techniques and
this method is its ability to find fault especially        interpretation from the result obtained. It
mechanical damage to the windings that cannot              involves proper mathematical and statistical
always be detected by other means. The                     evaluation. These techniques would help to
frequency behavior of RLC circuits can help                analyze different types of fault with respect to
describe the internal components found in a                different types of transformers.
                                                           The present study aims to find appropriate
Earlier work [1] was done on a three phase 2.5             analyzing technique(s) to detect mechanical
MVA power transformer with voltage level of                faults   on    three-phase    90MVA        power
33/11kV. Different kinds of minor and major                transformers with voltage level of 132/33kV.
faults were created and the diagnosis for each             Two processing techniques for the measurement
fault was done. Signal processing techniques               are used for the purpose of interpretation. They
such as comparison and signature were used with            are signature technique and comparison
selective frequency range to detect the kind of            techniques. Spectrums are categorized according
faults. Another work [2,3,4] was done that                 to Doble frequency analysis zones [7]. The two
concentrate more on the comparison method.                 techniques are applied on those frequencies.
The measurement was taken and compared with                Each transformer has its own signature.

However, a transformer that does not have                                        analysis from a power transformer. The test is
historical signature can use a sister transformer                                done on a transformer that has a healthy high
as a reference. This study shows that the above                                  voltage winding.
mentioned techniques are quite effective and
efficient in identifying the fault in transformer.

2.             MEASURING METHOD

The sweep frequency method is used in this
study. HP4395A is meanwhile used as network
analyzer for the purpose of generating the signal
and making the measurement. The frequency
range is from 5Hz to 10 MHz. The sweep                                                Figure 3: Signature of a healthy transformer
frequency data is collected in 5 subbands of 2
kHz, 20 kHz, 200 kHz, 2 MHz, and 10 MHz.                                         Numerical data obtained from FRA program is
Each subband consists of 401 equally spaced                                      manipulated through Matlab program by plotting
points with the total of 2005 points. Figure 1                                   a graph either in logarithmic or linear format.
shows block diagram of the test set-up, while                                    Data obtained from various measurements made
Figure 2 shows circuit diagram of the test set-up.                               on the same or different transformers can be
                                                                                 plotted on the same axes according to the need.
                                                                                 The plotting is banded according to Doble
        PC & GPIB                      ISOLATED                                  analysis zones that are low frequency for core
         CARD                             AC                                     and magnetic circuit (frequency < 10 kHz), mid
                                        SUPPLY                                   frequency for winding geometry (10 kHz <
                                                                        T        frequency < 600 kHz), and high frequency for
          HP 4395A                       SWITCH                                  inter-connections and test system (frequency >
        RF R A     B                      BOX                                    600 kHz).
                                         SWITCH                                  3.         ANALYZING TECHNIQUE

                                                                                 Two analyzing techniques are used for the
                                                                                 purpose of graphical interpretation. The methods
                     Figure 1: Block diagram                                     are signature and comparison.

                                                                                 3.1        Signature Technique
                    Figure 2: Circuit diagram
                    Zs                      Zt                                   This technique takes the measurement from all
                                                                                 the three phases of a transformer and creates a
                                                                                 single signal to represent the condition of the
                                                                                 transformer at the testing time.
S                           R    Zs                    T           Zs

                                                                                 Key equations:
                                                                                                         Ri + Yi + Bi
                                                                                                 Avi =
S - injected signal              Zs - impedance of measurement cables            where,
R - reference measurement        Zt - test impedance (transformer winding)
T - test measurement
                                                                                 Ri ≡ red phase measurement at the respective
A few fault cases involving winding and tap                                      frequency.
changer are selected. Data from healthy                                          Yi ≡ yellow phase measurement at the
transformers of different manufacturers are also                                 respective frequency.
taken as references. The main purpose of this                                    Bi ≡ blue phase measurement at the respective
study is to find how FRA diagnostic test can help                                frequency.
in monitoring the condition of healthy                                           Avi ≡ average value of red, yellow and blue
transformers. Figure 3 shows the outcome of an                                   phases at the respective frequency.

                                                           that is either high voltage (HV) or low voltage
Si = [(Ri − Avi )2 + (Y − Avi )2 + (Bi − Avi )2 ]
                       i                                   (LV) as the phase under study.

where,                                                     Key equations:

Si ≡ single value created at the respective                                  Di = Ri - REFi
frequency.                                                 where,

A single response is created out of the response           Ri ≡ phase measurement at the respective
signals from all the three phases to show the              frequency.
similarity of windings at the testing time. Figure         REFi ≡ reference measurement at the respective
4 shows signature spectrum of a healthy                    frequency.
transformer according to Doble analysis zones.             Di ≡ difference at the respective frequency.

                                                           Figure 5 shows difference spectrum of a same
                                                           phase for the same transformer before and after
                                                           faults according to Doble analysis zones.

      Figure 4: Signature spectrum of a healthy

This technique requires a transformer
characteristic. Measurement acquires from the               Figure 5: Comparison spectrum for before and
factory can be regarded as its birth certificate.                     after fault of a transformer
However, this technique does not require
historical measurement to interpret the condition          This technique needs reference data. It will be
of a transformer since comparison can also be              very useful if historical data is available so that
done with transformer of the same type known as            the differences can be analyzed. However,
sister unit.                                               comparison technique could analyze the
                                                           signature phase by phase, which is not possible
3.2       Comparison Technique                             in signature technique.

This technique compares a reading from one                 4.0      CASE STUDIES
phase of a transformer to a reference set of data.
The differences between the two readings                   4.1      Case Study 1 – Fault on HV winding
represent the condition of the transformer
winding as compared to the reference. The                  This study has been carried out on a 132/33kV,
reference spectrum can be obtained from the                90MVA power transformer, which has
healthy phase of either the same transformer or            experienced a short circuit failure. Both
the sister transformer. It is to be highlighted here       techniques, signature and comparison are used in
also that the signature of the healthy reference           this case study. We have tested both the HV and
phase must be obtained from the same winding               LV windings and it is found that the HV winding

signature has indicated some significant changes.        Comparison spectrum for A phase is shown in
So the analysis here has been discussed based on         Figure 8. The changes to the FRA measurement
the HV winding. Figures 6 and 7 show the                 subband especially for the frequency range up to
signature spectrum on the HV winding before              2kHz. The second subband shows that there is
and after the fault. Significant difference of the       almost no change in the winding geometry.
amplitude can be seen particularly in the first
subband. However, there is not much difference
in the second and third subbands. It is evident
from the spectrum that the problem is most
likely at the core and magnetic circuit of the
winding. Based on [1], major fault such as
winding fault can be identified by the low
frequency spectrum in the range of 2kHz. The
changes are obviously seen in phase A, but not in
phase B and C. Further investigation indicates
that there is HV mid winding earth fault occurs
in phase A of the transformer.

                                                          Figure 8: Comparison spectrum for before and
                                                                           after fault

                                                         4.2      Case Study 2 – Sister transformers
                                                                  from two different manufacturers

                                                         A study has been carried out on two 132/33kV,
                                                         90 MVA healthy power transformers. These two
                                                         transformers are of different manufacturers and
                                                         installed at two different substations.         In
                                                         addition, their production dates are also found to
Figure 6: Signature spectrum for the transformer         be different by few years.
                  before fault

Figure 7: Signature spectrum for the transformer         Figure 9: Signature spectrum for transformer by
                   after fault                                           manufacturer A

                                                        slightly different due to the factors explained in
                                                        the signature technique.

                                                        This case study aims to prove that if standard
                                                        signature of a certain transformer is not
                                                        available, then signature of the same rating
                                                        transformer, but from different manufacturer can
                                                        be used as reference, which is known as sister

                                                        4.3      Case Study 3 – Fault on tap changer

                                                        This case study has been carried out on a
                                                        132/33kV, 90MVA power transformer that has
                                                        failed during operation. Signature technique is
                                                        preferred in this analysis since there is no
                                                        historical data available. Figure 12 shows the
Figure 10: Signature spectrum for transformer by        signature spectrum on the HV winding of the
                manufacturer B                          transformer after the fault.

Figures 9 and 10 show the signature spectrum on
the HV winding of the transformers. The first
subbands have almost identical spectrum. It can
be observed that the transformers are healthy and
there is no indication of cores and magnetic
circuit problem. The second and third subbands
are slightly different. These differences can be
attributed as the results of settling of their
windings after manufacturing and installation at
site and interconnections of the winding.

                                                        Figure 12: Signature spectrum of the transformer
                                                                    with tap changer problem

                                                        The result is compared with those of Figures 9
                                                        and 10 of healthy transformers from Case Study
                                                        2. Difference in amplitude can be observed at the
                                                        early stage of the first subband. The problem has
                                                        occurred most probably at the core and magnetic
                                                        circuit of the winding. However, the spectrum
                                                        indicates that the fault is not likely to have
                                                        occurred on the transformer main winding as in
Figure 11: Comparison spectrum for transformer          Case Study 1. The pattern of the signature in
                   A and B                              Figure 12 is not similar to the pattern in Figure 7,
                                                        which is meant for the main winding fault.
Comparison spectrum from the same phase of              Further investigation on the transformer revealed
both the sister transformers is shown in Figure         that the tap changer was found to be open-
11. The first subband shows there is almost no          circuited. After doing the repair work, the
difference between the two transformers.                transformer has been put back into operation
However, the second and third subbands are              again.

This case study aims to show the reliability of                Power Transformer”, Electricity Today,
signature technique when historical data is not                vol.13, no.6, pp.14-19, 2001.
available by using sister transformers as
references to detect and identify the fault.             [4]   S. A. Ryder, “Method For Comparing
                                                               Frequency      Response       Analysis
5.       CONCLUSION                                            Measurements”, in Proc. 2002 IEEE Int.
                                                               Symp. Electrical Insulation, Boston,
Every transformer winding has a unique                         MA, 2002, pp.187-190.
signature that is sensitive to changes in the
parameters of the winding, namely resistance,            [5]   S.M Islam, “Detection of Shorted Turns
inductance,      and    capacitance.   Frequency               and Winding Movements in Large
spectrum of a transformer is very sensitive to any             Power Transformers Using Frequency
deformation or displacement of the winding.                    Response Analysis”, IEEE Power
Frequency response analysis is a very effective                Society, Winter Meeting, Singapore,
tool for diagnosing transformer condition. It is               2000, vol.3, pp.2233-2238.
particularly useful in detecting any fault that is
due to mechanical damage to the winding. The             [6]   J. Bak-Jensen, B. Bak-Jensen, and S. D.
technique is also very reliable for detecting any              Mikkelsen, “Detection of Faults and
short circuit to the winding.                                  Aging Phenomena in Transformers by
                                                               Transfer Functions”, IEEE Transactions
Results from a measurement can be analyzed                     on Power Delivery, vol.10, no.1,
through several techniques via graphical                       January 1999.
presentation. However, reference is needed for
better interpretation. The reference can either be       [7]
from historical data of the same transformer or                transformers/info/FRA/FRA_Tutorial_
from sister transformers. In many cases,                       Charles_Sweester.ppt
historical data for transformers already in
operation is difficult to get owing to shutdown
requirement. Sister transformers are used in
those cases as the reference measurement. The
interpretation of the results is meanwhile a great
help in determining further action to be taken
especially for suspected transformers. FRA can
be a very effective tool for condition monitoring.
It can avoid catastrophic failure in transformers
and also help maintenance engineer to estimate
time and cost for repairing the transformer after
the fault before undertaking maintenance.


[1]      S. Birlasekaran, F. Fetherston, “Off/On-
         Line FRA Condition Monitoring
         Technique For Power Transformer”,
         IEEE Power Engineering Review,
         pp.54-56, August 1999.

[2]      S. A. Ryder, “Diagnosing Transformer
         Faults Using Frequency Response
         Analysis”, IEEE Electrical Insulation
         Magazine, vol.19, no.2, pp.16-22,
         March/April 2003.

[3]      S. A. Ryder, “Frequency Response
         Analysis For Diagnostic Testing of


To top