[Downloaded from www.aece.ro on Wednesday, March 04, 2020 at 12:31:59 (UTC) by 178.171.28.63. Redistribution subject to AECE license or copyright.

Advances in Electrical and Computer Engineering Volume 20, Number 1, 2020

A New Approach for Estimating Insulation

Condition of Field Transformers Using FRA
Mohd Fairouz MOHD YOUSOF1, Salem AL-AMERI1, Hussein AHMAD2, Hazlee Azil ILLIAS3
and Syahrun Nizam Md ARSHAD4
1
Universiti Tun Hussein Onn Malaysia, Parit Raja, 86400, Malaysia
2
Universiti Teknologi Malaysia, Skudai, 81310, Malaysia
3
Universiti Malaya, Kuala Lumpur 50603, Malaysia
4
Universiti Malaysia Perlis, Arau, 02600, Malaysia
[email protected]

1
Abstract—Frequency response analysis (FRA) is a tool for transformer has been explored and discussed [3-5]. Previous
evaluating mechanical integrity of transformer’s core and studies have investigated the sensitivity of FRA
winding. However, several studies have reported that FRA is measurement on temperature, oil presence and moisture
also sensitive to other parameters such as temperature,
content. A work in [3] presented a case study on a dry
moisture content and presence of oil. Since some of these
parameters are critical in the insulation degradation process, it transformer by varying the surrounding temperature. The
is interesting to further investigate the possibility of using FRA results show that FRA measurement was slightly affected
for insulation assessment. In this paper, the insulation due to the change of temperature. It was also presented in
conditions of three field transformers are investigated using [5] and [6] that temperature variation in transformer can
FRA. This is performed after statistical analysis on the shift the frequency response horizontally. In the presence of
response suggested no mechanical damage on the transformers.
oil, a work in [7] presented two FRA measurements, which
The responses are further analyzed to determine the
percentage of change of winding capacitance. This is achieved were conducted before and after oil filling process. The
by comparing resonance frequencies between the responses. results show that the response is shifted towards lower
The percentage of change represents the amount of frequency after the presence of oil. The shift is due to the
degradation in the insulation. This is because as the insulation increase of insulation permittivity from the oil.
condition degrades, the response is shifted towards lower A laboratory study was conducted in [1] to determine the
frequencies. This new investigation demonstrates the
effect of moisture content in transformer oil-paper insulation
applicability of FRA to estimate the insulation degradation.
using FRA and FDS tests. The finding shows correlation
Index Terms—frequency response, insulation, insulation between the moisture content results estimated from the
testing, power transformers. FDS test and the variation of frequency response from the
FRA test. It was observed that as the moisture content
I. INTRODUCTION increases, the frequency response is shifted horizontally
Transformers are continuously exposed to thermal, towards lower frequency. This is due to the presence of
mechanical and electrical stresses throughout their moisture, which increases the relative permittivity of the oil-
operational lifetime. These stresses can affect the overall paper insulation. In another study presented in [8], the
condition of the unit, which consequently reduces its finding also indicated the influence of moisture content on
reliability to deliver electricity. The condition of insulation the frequency response. This can be seen from the shifting
is usually the main parameter to determine the aging status of peaks and valleys (resonances) in the response.
of transformer. For this reason, numerous studies were In this paper, further investigation on this matter is
conducted to understand the insulation condition. Besides presented. The paper aims to explore on the applicability of
tan delta test, which is commonly used to evaluate the FRA test to assess transformer insulation condition. This
insulation, there are other advance diagnostic methods such work is a continuation from the previous research, which
as return voltage measurement (RVM), frequency domain has been published in [1]. Previous work has established
spectroscopy (FDS) and polarization and depolarization that the percentage of moisture content in the insulation has
current (PDC), where for example were discussed in [1], [2]. a correlation with the percentage of change of capacitance
However, these techniques are operating based on slow ∆C%. This was determined from laboratory experiments on
polarization process of the dielectric material, thus they are a 5kVA transformer, which was placed under accelerated
very time-consuming. For an example, PDC measurement and controlled aging process. FDS and FRA tests were
requires 1.2 hour to obtain information on time domain up to performed on the unit for several times before and during
2000 seconds. the process.
Frequency response analysis (FRA) is a diagnostic tool, From the FRA test, the percentage of change of
which is used to assess the mechanical aspect of the capacitance ∆C% was achieved by comparing the frequency
transformer’s core and winding. However, recently FRA value of the peaks and valleys (resonances) in the response.
sensitivity towards non-mechanical parameters of the On the other hand, the FDS results were analyzed to
estimate the moisture content of the transformer’s insulation
1
This work was supported using Research Fund E15501, Research system.
Management Centre, UTHM.

Digital Object Identifier 10.4316/AECE.2020.01005

35
1582-7445 © 2020 AECE
 [Downloaded from www.aece.ro on Wednesday, March 04, 2020 at 12:31:59 (UTC) by 178.171.28.63. Redistribution subject to AECE license or copyright.]

Advances in Electrical and Computer Engineering Volume 20, Number 1, 2020

For further investigation, the proposed method in [1] is The transformers expose to mechanical and electrical
tested on three transformers, which have been in the field for faults. Also, transformer faults could be due to insulation
several years. This is conducted to determine the degradation. The mineral oil is used in transformer as
applicability of the method on the actual transformers. This insulation material as well as cooling medium [12], [13].
paper is arranged as follows; the theory section introduces The main insulation material used for conductor is the paper
FRA, insulation degradation and change in winding and pressboard [14]. The transformer’s durability and
capacitance. In the third section, the proposed methodology performance rely heavily on its insulation system [13]. The
to obtain the percentage of change of capacitance from the insulation material itself can be degraded due to the aging
FRA measurement is presented. The fourth section presents process. During the process, the insulation material is
results and discussion of three case studies from on-site subjected to various electrical, mechanical and thermal
transformers. Finally, the last section concludes the work stresses.
and suggests possible future work on this topic. Transformers are also constantly exposed to voltage
surges. This might occur due to switching of a large
II. FREQUENCY RESPONSE ANALYSIS inductive load or lightning strike. Such overvoltage will
It was addressed in many literatures such as in [9] that produce electrical stress on the insulation system winding,
frequency response analysis (FRA) is the most suitable which ultimately could result to the failure of the
method for detecting internal mechanical faults in power transformer [15]. On the mechanical stress, this is controlled
transformers. FRA test is performed by injecting low by having the clamping structure to hold the core and
voltage AC signal U1(f) at one terminal of the winding and winding structure. However, if the clamping pressure
measures the returning output signal U2(f) from the other exceeds the specified value, this may cause damage to the
terminal of the same winding. FRA instrument obtains the insulation system [6], [16]. Additionally, constant vibration
voltage transfer function of the winding from 20 Hz to 2 of the transformer is also a form of mechanical stress on the
MHz. The ratio of the output and input signals, insulation system.
20log10|U2(f)/U1(f)| generates a unique frequency response, Additionally, there is environmental factor that can cause
which is used to analyze the condition of the transformer. If degradation on the insulation such as the moisture. If the
there is any abnormal occurrence in the transformer, the insulation material absorbs and contains large amount of
resistance, inductance and capacitance of the transformer moisture, this will decrease the dielectric strength of the
will change, thus affecting the original frequency response. material. This is because the moisture will occupy the voids
In other words, the method of analysis in FRA test is to use in the insulation material, thus providing low resistance path
the fingerprint (original response) of the transformer to for the current to travel. Consequently, this will lead to
identify any changes by comparing it with the latest dielectric failure.
response [10]. Both the temperature and moisture can affect the paper
insulation of the transformer. Hence, the insulation dielectric
property, which is the permittivity, is affected. Since
insulation’s capacitance is dependent on the permittivity, its
degradation can be determined based on the change of
winding capacitance.
According to [17], from the theory of electro-technical, a
transformer winding is equivalent to a linear, bilateral
passive network that consists of inductances and
capacitances. Typical core and winding structure of a
transformer can be seen as a complex electrical network of
resistances R, self-inductances L, ground capacitances Cg,
coupling or mutual inductances Lm, capacitance between
windings Cw and series capacitances Cs as shown in Fig. 2.
Figure 1. The principle process of FRA measurement [11].

The principle of FRA measurement is shown in Fig. 1.

The transfer function of a transformer is given by (1) [11].
The phase and magnitude are calculated using (2) and (3).
Both magnitude and phase response plots can be obtained in
FRA test. However, the magnitude response is normally
used in the analysis since it provides more information and
easier to understand compared with the phase plot.
U (f)
H( f )  2 (1)
U1 ( f )
U2 ( f )
K ( f 1)  20 log10 (2)
U1 ( f )
U1 ( f )
 ( f 1)  tan 1 (3) Figure 2. Simplified network and equivalent circuit diagram of a
U 2 ( f ) transformer winding [11].

36
 [Downloaded from www.aece.ro on Wednesday, March 04, 2020 at 12:31:59 (UTC) by 178.171.28.63. Redistribution subject to AECE license or copyright.]

Advances in Electrical and Computer Engineering Volume 20, Number 1, 2020

When a fault occurs in a transformer, the inductance and

capacitance will change as well, thus its frequency response.
According to [4], [18], as the winding insulation changes,
resonances in the frequency response will also deviate by
either shifting to lower or higher frequencies. Resonances in
FRA are formed due to interaction between inductive and
capacitive reactance. Every resonance has its own
inductance and capacitance values. When the capacitance
and inductance change, the resonant frequency will also (a)
change. The capacitance is influenced by the insulation
degradation factor such as the temperature and moisture.
The ageing process of transformer can be observed using
FRA if it is conducted several times throughout its lifetime.
Such work has been presented in [1] by measuring the
percentage of change of the winding capacitance, which can
be computed from the frequency response.

III. METHODOLOGY (b)

Figure 4. FRA test configurations (a) End to end open circuit (b) End to end
The main purpose of FRA test is to determine mechanical short circuit [7].
damage on the transformer winding or core. However, this
study proposed a method to investigate the non-mechanical After performing FRA measurement, the statistical
loss in the transformer, which is the ageing of transformer indicators are applied to compare the difference between
winding insulation. The flow chart of the proposed method two frequency responses. The two responses are the
is shown in Fig. 3. reference response and the measured response. The
reference response could be the transformer fingerprint,
Start FRA from identical transformer or FRA from another limb.
This method is used to assist the interpretation of
FRA Measurement transformer condition. There are several statistical indicators
Statistical analysis
available such as root means square error (RMSE), standard
deviation (SD) and correlation coefficient (CC) [21]. The
most common indicators are CC and absolute sum of
Yes Mechanical No
damage? algorithmic error (ASLE) [22]. Other indicators were also
proposed by various studies such as comprehensive standard
Perform other test/ FRA Insulation ageing
inspection analysis
deviation (CSD) [23]. Nevertheless, CC is the most
preferable indicator in FRA analysis. The equations of CC is
Compare frequency peaks given in (4) as
between responses
 i 1   xi  x    yi  y  
N
Calculate ∆C% between CC  (4)
 i 1 xi  x 2   i 1  yi  y 2
responses N N

 Phase R more/less moisture content then Y.

 Phase Y more/less moisture content then B.
where xi and yi are the i-th values of data x and y while
 Phase B more/less moisture content then R. x and y are the mean values from the data. N is the total
data in x or y. Here, x and y are the 1st and 2nd responses to
End
be compared respectively. Based on [24], [25], an example
Figure 3. Flow chart of the proposed method.
for the meaning of CC in transformer winding condition is
For the FRA measurement, CIGRE WG A2.26 [7], IEEE given in Table I.
std. C57.149-2012 [19] and IEC 60076-18 [20], there are
TABLE I. CC BENCHMARK LIMITS FOR ESTIMATING TRANSFORMER
four different test configurations for connecting the FRA CONDITION FOR COMPARISON BETWEEN PHASES.
equipment to the transformer terminal. They are the open Condition CC
circuit end-to-end, short circuit end-to-end, capacitive inter- Good 0.95 - 1.00
winding and inductive inter-winding tests. However, the Marginal 0.90 - 0.94
Investigate < 0.90
open and short circuit end-to-end tests are commonly used.
This is because these test configurations can provide
Table I gives three CC benchmark limits for estimating
adequate information to analyze the transformer. The test
the transformer overall condition. This can be used for
configurations are shown in Fig. 4.
comparing between responses of windings from different
phases of similar transformer or from sister transformer. It
can be estimated that the transformer is in a good condition
when CC is between 0.95 and 1.0. It is in marginal condition
if CC is between 0.90 and 0.94. For less than 0.90, it is
treated that the transformer is in poor condition, thus

37
 [Downloaded from www.aece.ro on Wednesday, March 04, 2020 at 12:31:59 (UTC) by 178.171.28.63. Redistribution subject to AECE license or copyright.]

Advances in Electrical and Computer Engineering Volume 20, Number 1, 2020

requiring further investigation. However, it is important to 0

C original
mention that these limits could only serve as an estimation. C increased by 20%
In this work, the benchmark limits are used in the analysis. -20

Magnitude (dB)
Additionally, if the limits suggest that the transformer is in
good or marginal conditions, this work proposes to further
-40
analyze on the insulation condition. This is because although
there is no mechanical fault in the transformer, small
difference between responses is most likely due to the -60
ageing of insulation. This can be estimated and can provide
valuable additional information. Changes in insulation -80 -1 0 1
condition due to winding insulation ageing can cause 10 10 10
Frequency (Hz)
variation in the winding capacitance. However, it is
Figure 5. Shifting of resonance due to increase of capacitance in LC circuit.
probably not possible to calculate the winding capacitance
from FRA measurement. Nevertheless, the ratio of IV. CASE STUDIES
capacitance can be computed from two frequency responses
and this has been presented in [1]. This is achievable by The proposed interpretation scheme is applied on three
comparing two resonances as shown in the following case studies, where all cases are based on the measured
equations. The equation of the resonant frequency for response of field transformers. The first response was taken
parallel LC circuit is given in (5). The ratio of two from a transformer owned by a local power utility. It was
frequencies is therefore can be written as in (6). manufactured by EWT. The measurement was conducted on
the HV winding using end-to-end open circuit test
configuration. The second and third case studies are from
1
f  (5) two transformers located at a substation in a university. The
(2 LC ) transformers were manufactured by ABB and Tyree. For
f1 2 L2C2 these transformers, the measurement was conducted using
 (6) end-to-end short circuit test configuration on the HV
f 2 2 L1C1
winding. All three transformers are in the same vector
f12C group, which is Dyn11. The transformer winding phases are
 2 (7) labelled as R, Y and B as shown in Fig. 6. The specifications
2C1
f2 of the transformers are presented in Table II.
 C  C1   C2 
C21 %   2   100%    1  100% (8)
 C1   C1 
C   f2 
C21 %   2  1  100%   1  1  100% (9)
 C1   2 
 f2 

The capacitance ratio can be obtained by assuming that

the winding inductance is unchanged, where L1 = L2. Figure 6. Three-phase winding connection Dyn11.
Therefore, the frequency ratio as shown in (7) can be
obtained. Authors in [1] suggested that C1 and C2 are the TABLE II. SPECIFICATIONS OF THREE FIELD TRANSFORMERS.
capacitances from two different frequency responses. Here, Case
Power HV
LV Rating Vector
f1 and f2 are the resonant frequencies from two responses, Manufacturer Rating Rating
Study (kV) group
(MVA) (kV)
which are being compared. Therefore, the percentage of 1 EWT 0.75 11 0.433 Dyn11
capacitance change ΔC% can be resolved from the 2 ABB 1 11.5 0.433 Dyn11
frequency ratio as given in (8) and (9). The calculated 3 Tyree Electrical 1 11.5 0.433 Dyn11
percentage is basically the frequency difference between
two resonances of different responses. Theoretically, as TABLE III. CORRELATION COEFFICIENTS (20 KHZ TO 2 MHZ) BETWEEN
PHASES FOR THREE CASE STUDIES.
mentioned in [1], [5], winding response will begin to shift Transformers R-Y Y-B R-B
towards lower frequency as the ageing activity increases. EWT 0.9990 0.9993 0.9999
This is due to the increasing of insulation permittivity, ABB 0.9998 0.9999 0.9999
Tyree 0.9985 0.9985 0.9997
which is mainly caused by the rising of moisture content.
The capacitance also changes as it is dependent on the As presented earlier in the methodology, the first step of
insulation permittivity [16], [18]. An example of shifting the analysis is to use statistical indicator, which is CC. The
frequency response is shown in Fig. 5. The two responses result of CC for all case studies is presented in Table III.
show the frequency deviation occurred due to the increase of Since there is no earlier FRA measurement from all
capacitance of a simple LC circuit while the inductance transformers, the comparisons are made between phases. It
remains unchanged. can be observed that the values in Table III by referring to
the limits in Table I indicated that all transformers are in
good condition. Hence, there is no mechanical damage in
the winding. Since there is no physical damage in the
winding, this work proposes to further analyze the response

38
 [Downloaded from www.aece.ro on Wednesday, March 04, 2020 at 12:31:59 (UTC) by 178.171.28.63. Redistribution subject to AECE license or copyright.]

Advances in Electrical and Computer Engineering Volume 20, Number 1, 2020

by evaluating the insulation. This can be done by computing The results for comparison between phases Y and B and
the percentage of change of winding capacitance. phases B and R are given in Table IV. The percentage of
In the first case study, a three-phase 750 kVA 11/0.415 change of capacitance is calculated from (8). Based on (8),
kV transformer manufactured by EWT is investigated. negative percentage is obtained if the second response is
Fig. 7 presents the end-to-end open circuit responses of all shifted towards lower frequency than the first response. In
HV winding phases. The figure also includes vertical lines, other words, the frequencies of the peaks (f2) in the second
which is referring to the peaks (resonances) of every response are lower than the frequencies of the peaks (f1) in
response. It maybe difficult to visually identify each peak in the first response.
the response to run the analysis. In this case, the peaks’
magnitude and frequency were automatically identified TABLE IV. PERCENTAGE OF CHANGE OF WINDING CAPACITANCE FOR THE
FIRST CASE STUDY.
using finding local maxima function in MATLAB software.
No. ∆CR-Y (%) ∆CY-B (%) ∆CB-R (%)
The calculation of percentage of capacitance change ΔC% 1 -6.3671 4.4835 2.2172
was also performed using similar software. This analysis is 2 -14.2304 9.1680 6.8000
only applied in the high frequency region where the winding 3 -6.3671 6.8000 0
4 -6.3671 4.4835 2.2172
influences FRA measurement, which is 20 kHz to 2 MHz.
5 4.4835 -4.2911 0
Therefore, the resonance located at about 400 Hz is 6 0 -2.1691 2.2172
neglected. Any resonance below 20 kHz is not taken into 7 2.5763 9.3111 -10.8157
consideration thus the winding position (lateral or central) Average -3.7531 3.9694 0.3766
will not affect the analysis.
From Table IV, the first comparison compares phases R
and Y. Here, phase R is the first response while phase Y is
the second response. All peaks show variation on the
percentages except for the 6th peak where the frequency
remains unchanged (0%). In average, the percentage of
change of capacitance ∆CR-Y is -3.75%, which indicates that
phase Y winding has higher capacitance than phase R
winding. This suggests that winding of phase Y experiences
more degradation than winding of phase R. The second
comparison compares phases Y and B. The average
percentage ∆CY-B is 3.97%, which indicates that most peaks
of phase B are located at higher frequencies than the peaks
Figure 7. End-to-end open circuit response of the HV winding. of phase Y. Based on this observation, it is believed that the
winding of phase B experiences less degradation than phase
Fig. 8 shows a closer view of the responses between 50 Y. Similar for ∆CB-R, the average percentage of 0.38%
kHz to 2 MHz with all peaks indicated by vertical lines. suggests that winding phase R experiences slightly less
There are only 7 peaks identified for comparison between degradation than phase B. In overall, the winding insulation
phases R and Y. This is because the 7th peak from phase R of phase Y has the most degradation. Insulation of phase B
does not have its matching peak in phase Y. The percentage winding receives moderate degradation while phase R
of change of capacitance between phase R and Y is winding has the least degradation.
computed based on the peak frequencies and the results are The second case study is conducted on a three-phase
shown in Table IV. For the case of phase B, 11 peaks where 1MVA transformer manufactured by ABB. The measured
identified but only 7 peaks are selected as in the other responses of phases R, Y and B are shown in Fig. 9. There
phases. are 8 peaks identified between 70 kHz and 2 MHz on all
three responses. From these peaks, the percentage of change
of winding capacitance is computed and the result is given
in Table V.
1 2 3 4 56 7 8

1 2 3 4 56 7 8

1 2 3 4 56 7 8

Figure 8. Frequency responses of phases R, Y and B with peaks marked by

vertical lines from first case study.
Figure 9. Frequency responses of phases R, Y and B with peaks marked by
vertical lines from second case study.

39
 [Downloaded from www.aece.ro on Wednesday, March 04, 2020 at 12:31:59 (UTC) by 178.171.28.63. Redistribution subject to AECE license or copyright.]

Advances in Electrical and Computer Engineering Volume 20, Number 1, 2020

TABLE V. PERCENTAGE OF CHANGE OF WINDING CAPACITANCE FOR THE TABLE VI. PERCENTAGE OF CHANGE OF WINDING CAPACITANCE FOR THE
SECOND CASE STUDY. THIRD CASE STUDY.
No. ∆CR-Y (%) ∆CY-B (%) ∆CB-R (%) No. ∆CR-Y (%) ∆CY-B (%) ∆CB-R (%)
1 3.1178 -4.5008 1.5470 1 0 1.5470 -1.5235
2 1.1580 -1.1447 0 2 1.5468 -7.3880 6.3327
3 1.1577 -2.2761 1.1580 3 0 0 0
4 2.3294 0 -2.2763 4 0 0 0
5 2.3294 -1.1445 -1.1449 5 -2.2761 3.5141 0
6 3.5141 -2.2761 -1.1447 6 -3.3950 0 -1.1447
7 1.1577 0 -1.1445 7 -1.1447 4.7130 3.5143
8 7.1519 0 -6.6745 8 -1.1447 3.5143 -3.3950
Average 2.7395 -1.4178 -1.2100 9 -2.2763 4.7127 -2.2763
10 -1.1447 1.1580 -2.2761
11 -2.2761 2.3291 0
From Table V, all percentages of the change of 12 0 3.5143 0
capacitance between phases R and Y, ∆CR-Y are positive. 13 - - -3.3950
The average percentage of change of capacitance is 2.74%. Average -1.0092 1.4679 -0.3203
Positive value indicates that the peaks from the first
response, which is phase R, are located at higher frequencies The first comparison in Table VI is between winding
than the peaks from phase Y response. From the percentage, responses of phases R and Y. In average, the percentage of
it is believed that insulation in phase R experiences more change of capacitance, ∆CR-Y is -1.01%, which suggests that
degradation or ageing than insulation in phase Y. The insulation in the winding of phase Y has more degradation
second comparison is between the winding of phase Y and than the winding of phase R. In the second comparison
winding of phase B, ∆CY-B. The average percentage is - between phases Y and B, the average percentage is 1.47%.
1.42%, which indicates that most peaks of phase B response This indicates that the insulation in winding from phase B
are located at lower frequencies than the peaks of phase Y experiences less ageing than phase Y. Finally, the third
response. This suggests that the winding of phase B comparison, which is between phase B and phase R, shows
experiences more ageing than winding of phase Y. that the insulation of winding in phase R has more
Similarly, for ∆CB-R, with the average percentage of –1.21%, degradation than phase B. This is considered from the
it suggests that the winding of phase R suffers more negative percentage, which is -0.32%. Based from the
insulation degradation than the winding of phase B. In analysis on this transformer, insulation in the winding of
conclusion, for this particular transformer, insulation phase Y has the most degradation while insulation in phase
condition of phase R winding degraded the most while B winding has the least degradation.
phase Y winding has the least degradation of the insulation.
In the third case study, a three-phase 1MVA Tyree TABLE VII. AVERAGE PERCENTAGE OF CHANGE OF WINDING
CAPACITANCE FOR ALL CASE STUDIES.
electrical transformer is investigated. A comparison between
∆C (%)
responses of winding phase R and winding phase Y is Transformer
∆C R-Y (%) ∆C Y-B (%) ∆C B-R (%)
shown in Fig. 10 with vertical lines indicating the peaks of EWT -3.7531 3.9694 0.3766
both responses. There are 13 peaks identified on both phase ABB 2.7395 -1.4178 -1.2100
B and phase R responses while phase Y response has only Tyree -1.0092 1.4679 -0.3203
12 peaks. For this reason, only 12 peaks can be compared
between phase Y and other phases. Comparison between Table VII summarizes the percentage of change of
phase B and phase R can be made on all 13 peaks. The winding capacitance for all three case studies. Essentially,
calculated percentage of change of capacitance between negative percentage indicates that all or majority of the
each phase is shown in Table VI. resonances of the second response are located at lower
frequencies than the resonances of the first response. This is
1 2 3 4 5 6 7 8 910 11 12
vice versa for positive percentage. If both windings are in
perfect condition, there will be no difference between the
responses since both windings have identical structure with
the same insulation condition. Therefore, slight variation
between responses is most likely due to non-mechanical
1 2 3 4 5 6 7 8 910 11 12 changes on the winding i.e. ageing of insulation.
Non-mechanical effects such as moisture content and
temperature on frequency response have been documented
in- previous literatures such as in [1], [3], [26], [27].
Basically, the rise of moisture content and temperature
1 2 3 4 5 6 7 8 910 11 12 increases the insulation permittivity thus the winding
capacitance. For this reason, the changes on capacitance are
directly related to the movement of resonances.
In a three-phase transformer, the level of insulation
ageing of every winding can be estimated by comparing
with other windings to obtain the percentage of change of
Figure 10. Frequency responses of phases R, Y and B with peaks marked the capacitance. The highest percentage indicates the highest
by vertical lines from third case study.
degradation on the winding insulation and vice versa. As
can be seen in all three case studies, every winding

40
 [Downloaded from www.aece.ro on Wednesday, March 04, 2020 at 12:31:59 (UTC) by 178.171.28.63. Redistribution subject to AECE license or copyright.]

Advances in Electrical and Computer Engineering Volume 20, Number 1, 2020

experienced different degree of degradation. This is due to [2] S. M. Gubanski et al., "Dielectric response methods for diagnostics of
power transformers," Electr. Insul. Mag. IEEE, vol. 19, no. 3, pp. 12-
various loading condition on every phase, resulting in 18, May/June 2003. https://doi.org/10.1109/MEI.2003.1203017
different heating and ageing effect on every winding. [3] A. A. Reykherdt and V. Davydov, “Case studies of factors influencing
frequency response analysis measurements and power transformer
TABLE VIII. THE SEVERITY OF INSULATION AGEING FOR ALL CASES. diagnostics,” Electr. Insul. Mag. IEEE, vol. 27, no. 1, pp. 22–30,
Severity of insulation Jan./Feb. 2011. https://doi.org/10.1109/MEI.2011.5699444
Case Study Case Study Case Study [4] K. G. N. B. Abeywickrama, Y. V. Serdyuk, and S. M. Gubanski,
ageing from most
1 2 3 “Exploring Possibilities for Characterization of Power Transformer
severe to least severe
Insulation by Frequency Response Analysis (FRA),” IEEE Trans.
1 Winding Y Winding R Winding Y
Power Del., vol. 21, no. 3, pp. 1375–1382, July 2006.
2 Winding B Winding B Winding R https://doi.org/10.1109/TPWRD.2006.875855
3 Winding R Winding Y Winding B [5] M. Bagheri, B. T. Phung, T. Blackburn, and A. Naderian, “Influence
of temperature on frequency response analysis of transformer
From the results in Table VII, the ranking of severity of winding,” in IEEE Electr. Insul. Conf. (EIC), 2013, pp. 40–44.
https://doi.org/10.1109/EIC.2013.6554198
insulation ageing is established in Table VIII. The most [6] C. Krause, W. Goetz, and B. Heinrich, “The impact of drying and oil
ageing or degradation experienced in a winding can be impregnation conditions and of temperature cycles on the clamping
determined from the highest value of negative percentage. force of power transformer windings,” in Conf. Rec. IEEE Int. Symp.
Electr. Insul. , 2002, pp. 350–353.
For first and third cases, the proposed method suggests that https://doi.org/10.1109/elinsl.2002.995948
phase Y winding has the highest degradation on the [7] P. Picher et al. “Mechanical-Condition Assessment of Transformer
insulation. This is most probably due to the winding that is Windings Using Frequency Response Analysis (FRA),” CIGRÉ
working group A 2, 2008. https://e-cigre.org/publication/342.
located between two other windings. The lack of space [8] N. Abeywickrama, C. Ekanayake, Y. V Serdyuk, and S. M. Gubanski,
surrounding the winding could have caused higher heating “Effects of the Insulation Quality on the Frequency Response of
condition, thus suffering more degradation compared to the Power Transformers,” J. Electr. Eng., vol. 1, no. 4, pp. 534–542,
2006. https://doi.org/10.5370/JEET.2006.1.4.534
other windings. However, the second case study shows that [9] M. Hartje, “A new way of FRA (frequency response analysis),”
phase R winding is the most severe. A possible reason for presented at the 19th Int. Symp. High Volt. Eng., Pilsen, Czech
this is the loading at phase R is much higher than the other Republic, 2015.
phases. Nevertheless, transformer loading data was not [10] N. Hashemnia, S. Islam, and M. A. S. Masoum, “Understanding
Power Transformer Frequency Response Analysis Signatures,” IEEE
available to verify this. Additionally, further test on the Electr. Insul. Mag., vol. 29, no. 3, pp. 48–56, May/June 2013.
insulation using suitable equipment such as FDS could https://doi.org/10.1109/MEI.2013.6507414.
reveal more information. [11] F. Predl, “Interpretation of Sweep Frequency Response Analysis
(SFRA) Measurement Results,” presented at the Techcon Asia-
Pacific,Sydney, NSW, 2016. https://www.ee.co.za/article/interpreting-
V. CONCLUSION sweep-frequency-response-analysis-measurements.html
[12] O. Särneroth. (2001, Jan. 1). Insulation Materials [Online]. Available:
This paper presented a study on the use of frequency library.e.abb.com
response analysis (FRA) to evaluate insulation condition of [13] W. Ziomek, “Transformer Electrical Insulation [Editorial],” IEEE
three field transformers. It was found that there are slight Trans. Dielectr. Electr. Insul., vol. 19, no. 6, pp. 1841–1842, Dec.
2012. https://doi.org/10.1109/TDEI.2012.6396938.
differences between responses of each of the transformers. [14] T. A. Prevost, “Degradation of Cellulose Insulation in Liquid-Filled
Based on the statistical analysis, it was concluded that there Power Transformers,” presented at the 4th Annual Technical
is no mechanical damage on the winding. Although there Conference New Diagnostic Concepts for Better Asset Management,
2005. https://digital-library.theiet.org › content › journals › ip-
was no suspected damage, the insulation condition was smt_20000021
investigated using the proposed method, which is the [15] P. Gill, Electrical power equipment maintenance and testing, 2nd ed.,
percentage of change of capacitance ΔC%. This was Florida, CRCPress, 2008.
https://www.academia.edu/33236618/Electrical_Power_Equipment_
achieved by comparing the peak frequencies between two Maintenance_and_Testing_-_2nd_Edition.
responses. The method works based on the fact that the [16] A. Marinescu, and C. L. Ungureanu, “About Axial Clamping Force
response moves horizontally due to the changes of Monitoring at Power Transformer Windings during their Active
temperature and moisture. From the presented cases, it was Lifetime,” Ann. Univ. Craiova, Electr. Eng. Series, no. 32, pp. 56–61,
2008. http://elth.ucv.ro/fisiere/anale/2008/2_5.pdf
observed that certain responses are shifted towards lower [17] T. Sano and K. Miyagi, “Influence factors on FRA waveforms”, in
frequency, suggesting degradation on the insulation. This IEEE International Conference on Condition Monitoring and
was indicated by negative value of the percentage of change Diagnosis (CMD), 2012, pp. 601–604.
https://doi.org/10.1109/CMD.2012.6416216.
of capacitance ΔC%. By comparing between different [18] M. F. M. Yousof, “Frequency Response Analysis for Transformer
phases, the severity of insulation ageing for every phase can Winding Condition Monitoring,” Ph.D. dissertation, Sch. Inf.
be estimated. Technol. Electr. Eng., The Univ. of Queensland, Brisbane, Australia,
2015. https://espace.library.uq.edu.au
[19] IEEE Guide for the Application and Interpretation of Frequency
ACKNOWLEDGMENT Response Analysis for Oil-Immersed Transformers, IEEE Standard
C57.149, 2012. https://standards.ieee.org › standard › C57_149-2012
The authors also gratefully acknowledge Tapan Kumar [20] Power transformers - Part 18: Measurement of frequency response,
Saha, Chandima Ekanayake and Hui Ma from the University IEC 60076-18, 2012. https://global.ihs.com › doc_detail ›
of Queensland, Australia for their contributions in the early document_name=IEC 60076-18.
[21] M. F. M. Yousof, C. Ekanayake, and T. K. Saha, “Study of
part of this project. transformer winding deformation by frequency response analysis,” in
IEEE Power Energy Soc. Gen. Meet. (PESGM), 2013, pp. 1-5.
REFERENCES https://doi.org/10.1109/PESMG.2013.6672525
[22] M. F. M. Yousof, S. Riang, and M. K. A. Uyup, “The influence of
[1] M. F. M. Yousof, C. Ekanayake, and T. K. Saha, "Examining the data size in statistical analysis of power transformer frequency
ageing of transformer insulation using FRA and FDS techniques," response,” in 2016 IEEE Int. Conf. Power Energy (PECon), 2016, pp.
IEEE Trans. Dielectr. Electr. Insul., vol. 22, no. 2, pp. 1258-1265, 595–599. https://doi.org/10.1109/PECON.2016.7951630
Apr. 2015. https://doi.org/10.1109/TDEI.2015.7076829

41
 [Downloaded from www.aece.ro on Wednesday, March 04, 2020 at 12:31:59 (UTC) by 178.171.28.63. Redistribution subject to AECE license or copyright.]

Advances in Electrical and Computer Engineering Volume 20, Number 1, 2020

[23] K. P. Badgujar, M. Maoyafikuddin, and S. V. Kulkarni, “Alternative [26] A. Akbaril, H. Firoozil, H. Borsi, and M. Kharezi, “Assessment of
statistical techniques for aiding SFRA diagnostics in transformers,” Drying Quality for Power Transformer during Manufacturing Process
IET Gener. Transm. Distrib., vol. 6, no. 3, p. 189, March 2012. Using Variation of Transfer Function,” in IEEE Conf. on Electr. Insul.
https://doi.org/10.1049/iet-gtd.2011.0268 and Dielectr. Phenom. (CEIDP), 2006, pp. 113–116.
[24] G. M. Kennedy, A. J. McGrail, and J. A. Lapworth, “Using cross- https://doi.org/10.1109/CEIDP.2006.312074
correlation coefficients to analyze transformer sweep frequency [27] C. Ekanayake, S. M. Gubanski, A. Graczkowski, and K. Walczak,
response analysis (SFRA) traces,” in IEEE PES Conf. Expo. “Frequency response of oil impregnated pressboard and paper
(PowerAfrica), 2007, pp. 16–20. samples for estimating moisture in transformer insulation,” IEEE
https://doi.org/10.1109/PESAFR.2007.4498059. Trans. Power Del., vol. 21, no. 3, pp. 1309–1317, July 2006.
[25] G. Kennedy, A. McGrail, and J. Lapworth, “Transformer sweep https://doi.org/10.1109/TPWRD.2006.877632.
frequency response analysis (SFRA),” Energize, pp. 28–33, Oct.
2007.http://www.ee.co.za/wpcontent/uploads/legacy/Transformer.

42