polymers

Article
A Novel Universal Approach for Temperature
Correction on Frequency Domain Spectroscopy Curve
of Transformer Polymer Insulation
Jiefeng Liu 1 , Xianhao Fan 1 , Yiyi Zhang 1, * , Hanbo Zheng 1 , Huilu Yao 2, *, Chaohai Zhang 1 ,
Yubo Zhang 3 and Dajian Li 3
1 College of Electrical Engineering, Guangxi University, Nanning 530004, China
2 School of Physical and Technology, Guangxi University, Nanning 530004, China
3 Electric Power Research Institute of Guangxi Power Grid Co., Ltd., Nanning 530023, China
* Correspondence: yiyizhang@gxu.edu.cn (Y.Z.); huiluy@gmail.com (H.Y.)




Received: 22 May 2019; Accepted: 14 June 2019; Published: 2 July 2019


Abstract: It is a fact that the frequency domain spectroscopy (FDS) curve at different temperatures
can be corrected by the shift factor (αT ) extracted from the master curve. However, the αT and master
curve reported by previous works are distinctive due to the difference in the construction algorithm.
Therefore, it is of great significance to report a universal approach for extracting αT . In this work, the
unaged oil-immersed pressboards with different moisture content (mc%) are firstly prepared and
selected as the research specimen. Then, the αT of FDS curves on the above pressboard is extracted
based upon the master curve technique. The influence mechanism under the various test temperature
(T) and mc% is therefore analyzed so as to establish a universal model for predicting the αT . The
present findings reveal that the αT value extracted from FDS curves is both temperature-dependent
and moisture-dependent. In addition, the predicted αT is not only suitable for temperature correction
on FDS curve of same type pressboard with different insulation conditions (moisture contents and
aging degrees), but also maintains considerable accuracy when applied to different types of pressboard.
Therefore, the obtained conclusions will provide a universal method for temperature correction on
FDS curve of transformer polymer insulation.

Keywords: transformer polymer insulation; frequency domain spectroscopy (FDS); shift factor (αT );
master curve technique; temperature correction

1. Introduction
The performance of transformer oil–paper insulation is related to the stable operation of the entire
power system. It is a fact that the aging condition of the polymer (paper) insulation of the energized
transformer mainly determines the lifespan of the transformer oil–paper insulation system [1–5].
Thus, the condition evaluation of transformer polymer insulation has become a research hotspot. In
contrast with the traditional method based upon chemical and electric parameters, the frequency
domain spectroscopy (FDS) technique has received widespread attention, due to its advantages of
being sensitive to insulation conditions and more suitable for field testing [6–10]. In the last decades,
a large number of studies on the FDS technique were carried out [11–16] and the research findings
reveal that temperature is one of the crucial factors affecting dielectric response measurements [17–23].
Provided that the contribution of the temperature effect on the FDS curve is ignored, the obtained
response data are not reliable. The method for temperature correction on FDS curve is therefore of
great significance. In other words, it is an urgent issue to report a reliable approach for temperature
correction on the FDS curve.

Polymers 2019, 11, 1126; doi:10.3390/polym11071126 www.mdpi.com/journal/polymers

Polymers 2019, 11, 1126 2 of 11

In reviews of previous studies, it is pointed out that the temperature effect merely leads to the
FDS curve moving along the frequency axis and will not alter its shape [20–23]. In addition, it is
reported that the FDS curves at different test temperatures can be linked by shift factor (αT ) [20–23].
The general shape of the curve is often preserved if the data is plotted on a log–log scale. This feature
allows the formation of a master curve by shifting the curves along the frequency axis until they form
a continuous curve at a chosen reference temperature [23], and the above scale factor is defined as the
shift factor. A lot of work indicates that the shift factor extracted by the master curve technique can be
used for temperature correction on FDS curve. However, the master curve and shift factor obtained by
previous works are distinctive due to the difference in the construction algorithm. Therefore, it is of
great significance to report a universal approach for extracting the shift factor without constructing
the master curve. In view of this key issue, the author preliminarily reported a universal approach
for calculating the shift factor, which is realized by establishing a functional relationship between
the shift factor and various test temperatures [20]. The previous study indicates that the variation
law between the shift factor and test temperature can be described by the Arrhenius equation. It
is interesting to note that, as for the special operating environment of the transformer insulation
system, the dominant reaction mechanism of polymer aging should be hydrolysis when referring to the
Arrhenius process [24]. It seems likely that the value of the shift factor is both temperature-dependent
and moisture-dependent [23]. Provided that the synergy effect produced by moisture content and
temperature is ignored, the accuracy and versatility of the established model will be severely limited.
Therefore, the universal model [20] for temperature correction on FDS curve is not appropriate if the
synergy effect is not considered.
Given this consideration, this paper is devoted to understanding the mechanism of the synergy
effect on the αT so as to propose a universal approach for temperature correction. In the present work,
the unaged oil-immersed pressboards with different moisture content are firstly prepared under the
controlled laboratory conditions. Next, the variation law among the shift factor, moisture content,
and the temperature is deeply analyzed so as to establish a functional model for predicting the shift
factor. The present findings point out that the shift factor corresponding to any test temperature and
moisture content can be accurately predicted by the above functional model. Finally, the experiments
of the temperature correction are performed so as to verify the feasibility and accuracy of the proposed
approach. The conclusions reveal that the predicted shift factor is not only suitable for temperature
correction on FDS curve of same type pressboard with different insulation conditions (moisture content
and aging degree), but considerable accuracy is also maintained when it is applied to different types of
paperboard. Therefore, the conclusions will provide a universal idea for temperature correction on
FDS curve of transformer polymer insulation.

2. Sample Construction
In order to verify the versatility of the proposed method, two types of pressboard discs with
different moisture contents were utilized for performing the variable temperature experiment and
verification experiment. The diameter of the pressboard discs for both experiments were 160 mm. The
details are described in Table 1. The transformer oil is the Karamay No. 25 naphthenic mineral oil and
satisfies the standard of ASTM D3487-2000 (II).

Table 1. The comparison of the parameters of two pressboards.

Pressboard Types Type I Type II

Brand T4 transformer pressboard Common transformer pressboard
Taizhou Weidmann High Voltage Chongqing AEA Group
Manufacturer
Insulation Co., Ltd. Transformer Co., Ltd.
Thickness 0.5 mm 2 mm
Tensile strength MD: 98 MPa, CMD: 47 MPa MD: 150.04 MPa, CMD: 57.14 MPa
Density 0.96 g/cm3 1.17 g/cm3
Polymers 2019, 11, 1126 3 of 11

Polymers 2019, 11, x FOR PEER REVIEW 3 of 11

Polymers 2019, 11, x FOR PEER REVIEW
As shown in Figure 1, the insulating oil and the pressboard are vacuum dried in a ratio of3 20:1 of 11
at
◦
105 C, As 50 shown
Pa, andintheFigure 1, the insulating
oil-immersed oil and
pressboard the pressboard
is obtained are vacuum
by vacuum immersed ◦
driedatin60a ratioC, 50ofPa. 20:1The
As shown in Figure 1, the insulating oil and the pressboard are vacuum dried in a ratio of 20:1
at 105 °C,with
pressboard 50 Pa, and the
initial oil-immersed
moisture contentpressboard
a% is placed is obtained by vacuum
in a precision immersed
electronic balance at 60
and °C,its50quality
Pa.
at 105 °C, 50 Pa, and the oil-immersed pressboard is obtained by vacuum immersed at 60 °C, 50 Pa.
The
(m)The pressboard
is recorded; with initial moisture content a% is placed in a precision electronic balance and its
pressboardthe withnatural
initialmoisture
moisture absorption
content a% is is placed
later performed
in a precisionso as to obtain
electronic the pressboard
balance and its
withquality
various(m) is recorded; the natural moisture absorption ismeasured
later performed so as balance
to obtain the
quality (m)expected
is recorded;moisture contents.
the natural Provided
moisture that theis
absorption value of
later performed sotheas to obtainreaches the
m× pressboard
(1 + b%)/(1 with
+ various
a%), the expected
moisture moisture
content of contents.
the pressboardProvidedis that
regarded the
as measured
b% [25,26]. value
The of the
moisture
pressboard with various expected moisture contents. Provided that the measured value of the
balance
balance reaches m × (1 + b%)/(1
distribution is later++performed
a%), the moisture content
at 45 ◦ C, content ofSubsequently,
the pressboard isDPregarded as b%
balance reaches mprocess
× (1 + b%)/(1 a%), the moisture for 48 h.of the pressboard theis value (degree
regarded as b% of
[25,26]. The moisture balance distribution process is later performed at 45 °C, for 48 h.
polymerization) and moisture
[25,26]. The moisture balancecontent are obtained
distribution by a DP
process is tester
later (by means ofatviscosity
performed 45 °C, testing)
for 48 and h. a
Subsequently, the DP value (degree of polymerization) and moisture content are obtained by a DP
moisture tester (by
Subsequently, themeans
DP value of Karl Fischer
(degree Titration), as shown
of polymerization) in Figure
and moisture 1. In are
content thisobtained
paper, the by unaged
a DP
tester (by means of viscosity testing) and a moisture tester (by means of Karl Fischer Titration), as
oil-immersed
tester (by meanspressboards (typetesting)
of viscosity I) withanddifferent moisture
a moisture testercontents
(by means (1.3%, 2.3%,Fischer
of Karl 3.1% and 4.4%) were
Titration), as
shown in Figure 1. In this paper, the unaged oil-immersed pressboards (type I) with different
shown
firstly in Figure
prepared. The 1. In curves
tanδ this paper, the unaged
are tested by DIRANA,oil-immersed
where the pressboards (type I) with
dielectric response different is
measurement
moisture contents (1.3%, 2.3%, 3.1% and 4.4%) were firstly prepared. The tanδ curves are tested by
moisture
achieved by contents (1.3%, 2.3%,
three-electrode test 3.1%
cell, and 4.4%)
shown in were2;firstly
Figure the prepared.
test voltage The
is set tanδ
to curves
200 V, thearetesttested by
frequency
DIRANA, where the dielectric response measurement is achieved by three-electrode test cell, shown
is 2DIRANA,
× 10 −3 − where
5 × 103the dielectric response measurement is achieved by
Hz. three-electrode test cell, shown
in Figure 2; the test voltage is set to 200 V, the test frequency is 2 × 10−3 − 5 × 103 Hz.
in Figure 2; the test voltage is set to 200 V, the test frequency is 2 × 10−3 − 5 × 103 Hz.

Insulating
Insulating
Oil
Oil 105
50 Pa
105 60 Immersed
Cellulose 50 Pa 60
50Pa Immersed
48h
Cellulose
Pressboard 50Pa 48h Oil-immersed
Pressboard Oil-immersed
Pressboard
Vacuum Drying Vacuum Immersed
Vacuum Drying Vacuum Immersed Pressboard
Moisture
Moisture
Absorption
Absorption
Pressboard with
Tanδ Curves at Pressboard with
Different mc%
Tanδ Curves at Variable
Various Temp. Variable Different mc%
Temperature
Various Temp. Temperature
Moisture
Dierectric Response Tester Moisture
Equilibrium
Dierectric Response Tester Equilibrium
for 48h
for 48h
mc% and DP
mc% and DP

DP Tester Moisture Tester

DP Tester Moisture Tester

Figure 1. The experimental schedule of unaged pressboard with type I.

Figure1.1.The
Figure Theexperimental
experimental schedule of unaged
schedule of unagedpressboard
pressboardwith
withtype
typeI. I.

Figure 2. The three-electrode test cell.

Figure2.
Figure 2. The
The three-electrode
three-electrode test
testcell.
cell.
3. The Extraction of the Shift Factor
3. The
3. The Extractionofofthe
Extraction theShift
ShiftFactor
Factor
In this paper, the unaged oil-immersed pressboards (type I) with various mc% are selected as
In In
thisthis paper,
paper, thethe unaged
unaged oil-immersed
oil-immersed pressboards(type
pressboards (typeI)I)with
withvarious
variousmc%
mc%are
areselected
the research object, the tanδ curves of above pressboards with various moisture contents are shown
selectedasasthe
the research
research object, object,
the the tanδ
tanδ curvescurves
of of above
above pressboards
pressboards with
with various
various moisturecontents
moisture contentsare
areshown
shown in
in Figure 3, and the test temperature is 45 °C. Then, the tanδ curves of the above pressboards at
in Figure 3, and the test temperature ◦is 45 °C. Then, the tanδ curves of the above pressboards
Figure 3, and the test temperature is 45 C. Then, the tanδ curves of the above pressboards at different at
test temperatures (45, 60, 75, and 90 ◦ C) were measured, respectively, which is displayed in Figure 4.
Polymers 2019, 11, x FOR PEER REVIEW 4 of 11
Polymers 2019, 11, x FOR PEER REVIEW 4 of 11

different test temperatures (45, 60, 75, and 90 °C) were measured, respectively, which is displayed in
different
Polymers test11,temperatures
2019, 1126 (45, 60, 75, and 90 °C) were measured, respectively, which is displayed
4 of in
11
Figure 4.
Figure 4.

mc%=1.3%
103 mc%=1.3%
mc%=2.3%
103 mc%=2.3%
mc%=3.1%
mc%=3.1%
mc%=4.4%
mc%=4.4%
102
102

tanδ（%）
tanδ（%）
101
101

100
100

10-1
10-1 10-2 10-1 100 101 102 103
10-2 10-1 100 101 102 103
f/HZ
f/HZ

Figure 3. The tanδ curves of unaged pressboards (type I) with various mc% (45 °C).
Figure 3. The tanδ curves of unaged pressboards (type I) with various mc% (45 ◦°C).
C).

103
45°C 103 45°C
45°C
60°C 45°C
C
60°
102
60°C
75°C 60°C
C
75°
102 75°C
C 75°C
C
90° 102 90°
90°C 102 90°C
tanδ（%）

tanδ（%）
tanδ（%）

101
tanδ（%）

101 101
101

100
100 100
100

10-1 10-1
10-1 10-2 10-1 100 101 102 103 10-1 10-2 10-1 100 101 102 103
10-2 10-1 100 101 102 103 10-2 10-1 100 101 102 103
f/HZ f/HZ
f/HZ f/HZ
(a) (b)
(a) (b)
45°C
103
45°C 45°C
C
45°C
C 60°
103 60° 103 60°C
60°C
C 103 75°C
75° 75°C
C
75°C
C 90°
90° 90°C
90°C
102
102
tanδ（%）
tanδ（%）

tanδ（%）
tanδ（%）

102
102

101
101

101
101
100
100

10-2 10-1 100 101 102 103 10-2 10-1 100 101 102 103
10-2 10-1 100 101 102 103 10-2 10-1 100 101 102 103
f/HZ f/HZ
f/HZ f/HZ
(c) (d)
(c) (d)
Figure 4.
4. The tanδ
tanδ curvesofofunaged
unaged pressboard with type
I atII various
at various
testtest temperatures.
(a) mc% = 1.3%,
Figure 4. The
Figure The tanδcurves
curves of unagedpressboard
pressboardwith
withtype
type at various temperatures.
test temperatures.
(a)
(b) mc%
mc% =
= 1.3%,
2.3%, (b)
(c) mc%
mc% =
= 2.3%,
3.1%, (c)
(d)mc%
mc% ==3.1%,
4.4%.(d) mc% = 4.4%.
(a) mc% = 1.3%, (b) mc% = 2.3%, (c) mc% = 3.1%, (d) mc% = 4.4%.

It is a fact that the increase in ambient temperature enhances the kinetic energy of the molecules
It is a fact that the increase in ambient temperature enhances the kinetic energy of the molecules
thepressboard
inside the pressboard samples,
samples, making
making it easier
it easier for the molecules
for the molecules to the
to complete complete the polarization
orientation orientation
inside the pressboard samples, making it easier for the molecules to complete the orientation
polarization
under under
the action the action
of external of external
electric electricThis
field forces. fieldmeans
forces. This
that themeans that the
pressboard pressboard
samples samples
with the same
polarization under the action of external electric field forces. This means that the pressboard samples
insulation condition (aging degree and mc%) will need less time to establish the process of relaxation
polarization under the high temperature. Therefore, the high temperature allows the relaxation
Polymers 2019, 11, x FOR PEER REVIEW 5 of 11

with the same insulation condition (aging degree and mc%) will need less time to establish the
process of relaxation polarization under the high temperature. Therefore, the high temperature
Polymers 2019, 11, 1126 5 of 11
allows the relaxation polarization process to be established at a higher frequency section than the
reference temperature. This interesting phenomenon can be observed on the variation law of the
tanδ curve process
polarization under to thebetemperature
established at effect,
a higherwhich
frequency is shown
section in thanFigure 4. Obviously,
the reference temperature.the high
This
temperature
interesting will cause the
phenomenon can tanδ curves to
be observed onmove toward law
the variation the ofright
the along the frequency
tanδ curve under the axis without
temperature
changing
effect, its shape.
which is shown This conclusion
in Figure is also consistent
4. Obviously, the high with the existing
temperature will research
cause thefindings
tanδ curves [17–23].
to move
towardProvided
the rightthat thethe
along FDS curve at axis
frequency different testchanging
without temperatures, T, is This
its shape. moved along the
conclusion is frequency axis
also consistent
to the
with thereference temperature
existing research (Tref)[17–23].
findings by a certain scale factor, the reference curve is defined as the
master curve and
Provided thatthe
theabove scale at
FDS curve factor is defined
different as the shift factor
test temperatures, αT, andalong
T, is moved the formula of the shift
the frequency axis
factor
to is as shown
the reference in Equation
temperature (1)) by
(Tref [20], where, scale
a certain fTref and fT are
factor, thethe frequencies
reference curvecorresponding
is defined as the to amaster
given
point and
curve on the
the above
tanδ scale
curvefactor
at the reference
is defined as the shift factor α
temperature and thethe
T , and test temperature,
formula of the shiftrespectively.
factor is as
Moreover,
shown the reference
in Equation temperature
(1) [20], where, fTref in this
and fT paper
are theisfrequencies
set to 45 °C. corresponding to a given point on the
tanδ curve at the reference temperature and the testf temperature, respectively. Moreover, the reference
temperature in this paper is set to 45 ◦ C.  T 
T (1)
fTref fT
αT = (1)
As mentioned in the introduction part, the temperature fTre f effect does not alter the shape of the
response curve. Consequently,
As mentioned the master
in the introduction part, curve obtained by
the temperature shifting
effect does along thethe
not alter frequency
shape ofaxis
the
should coincide with the measured curve at the reference temperature. The
response curve. Consequently, the master curve obtained by shifting along the frequency axisconstruction algorithm
should
moves the
coincide tanδ
with thecurve at different
measured curve test temperatures
at the along the frequency
reference temperature. axis to the
The construction left. In this
algorithm movespaper,
the
as long as the shifted curves and the measured curve at the reference temperature
tanδ curve at different test temperatures along the frequency axis to the left. In this paper, as long as thehave the
maximum
shifted coincidence
curves degree (i.e.,
and the measured theatsum
curve the of squaredtemperature
reference deviations reaches
have thethe minimum),
maximum the shift
coincidence
factor of the above test temperature is extracted immediately. Figure 5 presents the
degree (i.e., the sum of squared deviations reaches the minimum), the shift factor of the above test construction
results.
temperature is extracted immediately. Figure 5 presents the construction results.

before shift at 45°C
before shift at 60°C
102
before shift at 75°C
before shift at 90°C
tanδ（%）
103
before shift at 45°C
before shift at 60°C
before shift at 75°C
102 before shift at 90°C

tanδ（%）

101
101

master curve master curve

100
after shift at 45°C 100 after shift at 45°C
after shift at 60°C after shift at 60°C
after shift at 75°C after shift at 75°C
after shift at 90°C after shift at 90°C
10-1 10-1
10-3 10-2 10-1 100 101 102 103 10-4 10-3 10-2 10-1 100 101 102 103
f/HZ f/HZ

(a) (b)

before shift at 45°C before shift at 45°C

103
before shift at 60°C before shift at 60°C
before shift at 75°C before shift at 75°C
before shift at 90°C 103 before shift at 90°C
102
tanδ（%）

tanδ（%）

102
101

master curve master curve

after shift at 45°C after shift at 45°C
100 after shift at 60°C 101
after shift at 60°C
after shift at 75°C after shift at 75°C
after shift at 90°C after shift at 90°C
10-4 10-3 10-2 10-1 100 101 102 103 10-4 10-3 10-2 10-1 100 101 102 103
f/HZ f/HZ

(c) (d)

Figure 5. The construction results of the master curve of the unaged pressboard with type I.
(a) mc% = 1.3%, (b) mc% = 2.3%, (c) mc% = 3.1%, (d) mc% = 4.4%.
Polymers 2019, 11, 1126 6 of 11

In this paper, the fitting of the master curve is completed by least square estimation (LSE), provided
that the FDS curves at reference temperature (45 ◦ C) and various test temperatures can be represented
by tanδ(fTrefi ) and tanδ(fTi ), respectively. The fTrefi , fTi represents the i-th sampling frequency of the tanδ
curve at the reference temperature and test temperatures, respectively. It can be seen from Equation (1),
fTi /αT = fTrefi , provided that the tanδ curves at different test temperatures can perfectly coincide with the
master curve after translation, i.e., tanδ(fTrefi ) = tanδ(fTi / αT ). However, the results of translation are not
perfect and always bring deviations. Subsequently, the βi = [tanδ(fTrefi ) − tanδ(fTi / αT )]2 is defined as the
deviation of a single sample point in the fitting process of the master curve. The following boundary
conditions need to be mentioned, once (fTi / αT ) < min(fTrefi ) or max(fTi / αT ) < fTrefi , the above sampling
points are considered as invalid points, and the deviation βi is therefore assumed to be 0. The sum of
squared deviations ( βi ) is therefore defined as the objective function of the fitting process. Once the
P

βi reaches the minimum value, the calculated αT is exactly what we need. Moreover, the shift factor
P

values of pressboard 1 extracted by the above process are presented is Table 2. It is found that the shift
factor presented an increasing trend with increasing temperature and moisture content.

Table 2. The extracted shift factor of unaged pressboard with type I at different temperature.

Measured Value
No. mc%
318.15 K 333.15 K 348.15 K 363.15 K
1 1.3% 1.00 4.43 20.45 57.79
2 2.3% 1.00 4.51 20.33 60.23
3 3.1% 1.00 4.84 21.83 59.14
4 4.4% 1.00 5.17 22.98 63.25

4. A Universal Model for Extracting the Shift Factor

As mentioned above, the variation law between the shift factor and test temperature can be
accurately linked to the Arrhenius equation [20–24]. It is interesting to note that the Arrhenius equation
is an empirical formula used to describe the relationship between the chemical reaction rate constant k
and the ambient temperature. Their analysis revealed an equation in the form [20]
Ea
k = A · e− R·T , (2)

where, the A is the pre-exponential factor, which is related to the moisture content inside the polymer
insulation material. R is a gas constant, and R = 8.314 J/mol·K. T is the temperature during the chemical
reaction process. The time–temperature superposition theory [27,28] indicates that the effect generated
by increasing temperature and prolonging the reaction duration on the molecular motion is equivalent.
In other words, the variation law of the microstructure parameters of polymer insulation under high
temperature can be observed under other conditions (both a lower temperature and longer duration).
Thus, the FDS curve under the high temperature can be obtained by moving the curves under low
temperature according to a shift factor, and the calculation equation is shown in Equation (3).
Ea Ea 1
Ea − R·Tre ( Tre − T1 )
αT = A · e− R·T /A · e f =eR f (3)

From Equation (3), it is estimated that the value of the shift factor value is positively correlated
with the activation energy and the test temperature. It has been shown in another published study that
the value of the activation energy is related to the chemical reaction temperature and the concentration
of the reactants [24]. It is worth mentioning that, as for the special operating environment of the
transformer insulation system, the dominant reaction mechanism of the polymer aging should be
hydrolysis when referring to the Arrhenius equation [24,29,30]. Therefore, once the synergy effect
produced by moisture content and test temperature is ignored, the accuracy and versatility of the
established model will be severely limited. In order to explore the variation rule of the shift factor
Polymers 2019, 11, x FOR PEER REVIEW 7 of 11

polymer
Polymers aging
2019, should
11, 1126 be hydrolysis when referring to the Arrhenius equation [24,29,30]. Therefore, 7 of 11
once the synergy effect produced by moisture content and test temperature is ignored, the accuracy
and versatility of the established model will be severely limited. In order to explore the variation
under
rule ofthe
thesynergy effect,under
shift factor in thisthe
article, the temperature,
synergy effect, in thismoisture content,
article, the and shiftmoisture
temperature, factor (shown in
content,
Table 2) are set to x, y, and z values, respectively, and their distribution in three-dimensional coordinates
and shift factor (shown in Table 2) are set to x, y, and z values, respectively, and their distribution in
isthree-dimensional
plotted in Figure coordinates
6a. is plotted in Figure 6a.

(a) (b)

Figure6.6.The
Figure Thevariation
variationlawlaw
of of shift
shift factor.
factor. (a) The
(a) The scatter
scatter plot
plot of testoftemperature,
test temperature, mc%
mc% and and
shift shift
factor;
factor;
(b) The (b) Thesurface
fitting fitting surface of the
of the shift shift factor.
factor.

ItIt can
canbe
beseen
seenfrom
fromFigure
Figure6a,
6a,the
thevalue
valueof
ofthe
theshift
shiftfactor
factorshows
showsaaregular
regularvariation
variationtrend
trendwith
with
the altered moisture content and test temperature. In order to verify this viewpoint,
the altered moisture content and test temperature. In order to verify this viewpoint, the variation law the variation
law shown
shown in Figure
in Figure 6 is studied
6 is studied so establish
so as to as to establish a functional
a functional modelmodel for predicting
for predicting the
the shift shiftunder
factor factor
under
any testany test temperature
temperature and moisture
and moisture content,content, where
where the the surface
fitting fitting surface
is shownis shown
in Figurein Figure
6b. From6b.
From Figure
Figure 6b, each6b,point
eachon point on thesurface
the fitting fitting surface represents
represents a shiftunder
a shift factor factorthe
under
clearthe clear moisture
moisture content
content
and and test temperature.
test temperature. Provided Provided that
that the test the test temperature
temperature and moisture andcontent
moisture content are
of samples of samples
known,
are known, its shift factor can be predicted by using the equation
its shift factor can be predicted by using the equation shown in Table 3. shown in Table 3.

Table3.3. The
Table The universal
universal equation
equationfor
forpredicting
predictingthe
theshift
shiftfactor.
factor.
2
Zα0 +=A01Z0⋅+xA+01 ·xB+01B⋅01y·y+
+BB0202
·y2⋅+yC02+ C 02 ⋅ x ⋅ y
·x·y
αT = T 1+A1 ·x+B1 ·y+A2 ·x22+B2 ·y2 +C2 2·x·y
1 + A1 ⋅ x + B1 ⋅ y + A2 ⋅ x + B2 ⋅ y + C 2 ⋅ x ⋅ y
Z0 −342,488 A1 67.5388 Precision 10−15
A01 Z0−59.000
−342,488 AA1 2 67.5388 0.55934Precision
Degree of freedom
10−15 6
B01 2137.2 A3 −2.82072 Reduced Chi-Sqr 0.5229
B02 A01−3.3360
−59.000 AB2 1 0.55934 0.00694
Degree of freedomR-Square 6 0.9991
C02 −0.1576 B2 −0.18727 Fit Status succeeded
B01 2137.2 A3 −2.82072 Reduced Chi-Sqr 0.5229
In the equationBin02 Table 3, the x,By1represent
−3.3360 0.00694the moisture content and test0.9991
R-Square temperature, respectively.
Table 3 shows that the residual sum of squares (i.e., Reduced Chi-Sqr) of the fitting equation is less than
C02 −0.1576
1, and the fitting goodness B2 −0.18727
(i.e., R-Square) Fit Statusthe results
is close to 1. Therefore, succeeded
of mathematical statistics
preliminarily prove the feasibility of the proposed equation. Further, the shift factor predicted by the
universal equation
In the is shown
equation in Table
in Table 3, the4. x,In ysummary,
representthe
thepresent findings
moisture reveal
content andthat
testthetemperature,
shift factor
corresponding to test3 temperature
respectively. Table shows that theandresidual
moisturesum content can be accurately
of squares predicted
(i.e., Reduced by of
Chi-Sqr) thethe
reported
fitting
universal approach (shown in Table 3).
equation is less than 1, and the fitting goodness (i.e., R-Square) is close to 1. Therefore, the results of
mathematical statistics preliminarily prove the feasibility of the proposed equation. Further, the
shift factor predicted by the universal equation is shown in Table 4. In summary, the present
Polymers 2019, 11, 1126 8 of 11

Table 4. The predicted shift factor of unaged pressboard with type I.

Predicted Value
No. mc%
318.15 K 333.15 K 348.15 K 363.15 K
1 1.3% 0.983 4.439 20.210 58.185
2 2.3% 0.988 4.603 20.847 59.013
3 3.1% 0.998 4.772 21.522 60.154
4 4.4% 1.027 5.139 23.013 63.056

5. A Scheme for Temperature Correction on FDS Curve Using Reported Universal Approach
This article aims to introduce a scheme for temperature correction on FDS curve so as to verify the
feasibility and accuracy of the reported universal approach. Therefore, it is worth making the following
interpretations. (1) The shift factor can be accurately predicted by using the universal model shown
in Table 3. (2) The shifted FDS curves at different test temperatures were obtained by shifting the
measured curves (at 45 ◦ C) along the frequency axis according to a predicted shift factor. (3) Provided
that the reported method is valid, the shifted FDS curves at any test temperature should coincide with
the measured curves at 45 ◦ C.
Further, several pressboards were selected as the study objects in this article so as to construct the
FDS curve at the various test temperatures. In order to emphasize its universality, several pressboards
with different insulation conditions (mc% and DP) were utilized for feasibility verification of the
proposed approach, where, the aged pressboard with various DP is obtained by the process of
accelerated thermal aging at 130 ◦ C, and where the aging rate of the pressboard has been studied
previously, predicting the variation law between the DP value (aging degree) and aging duration [24].
The pressboard samples designated as 1, 2, 3, and 4 are common transformer pressboards with type II,
and prepared as described in a previous study [20]; the other T4 transformer pressboards with type I,
and newly prepared in this work, are designated 5 and 6. The details of the above various pressboards
are shown in Table 5.

Table 5. The predicted shift factor of various pressboards.

No. DP mc% 318.15 K 333.15 K 348.15 K 363.15 K

1 425 1.2% 0.983 4.428 20.158 58.137
2 994 2.8% 0.994 4.705 21.250 59.673
3 841 3.7% 1.010 4.926 22.143 61.318
4 1285 4.0% 1.016 5.012 22.495 62.010
5 663 0.9% 0.983 4.385 20.012 58.028
6 640 1.3% 0.983 4.439 20.210 58.185

Figure 7 presents the comparison of the shifted curve and measured curve at different temperatures.
Obviously, the shifted curve obtained by the reported approach in this paper is basically coincident with
the measured curve at any test temperature. Therefore, it is considered that the reported temperature
correction method is effective. The obtained findings reveal that the reported approach is not only
suitable for temperature correction on FDS curve of pressboard with different insulation conditions
(mc% and degree of polymerization), but also maintains considerable accuracy when applied to
different types of paperboard. The available study points out that the shift factor seems more likely
temperature-dependent and moisture-dependent than aging-dependent [23]. The above conclusion is
preliminarily confirmed in this work and described in Figures 6 and 7. In summary, the experimental
results of the temperature correction provide a basis for the feasibility verification of the proposed
universal approach.
Polymers 2019, 11, 1126 9 of 11
Polymers 2019, 11, x FOR PEER REVIEW 9 of 11

measured curve at 45°C measured curve at 45°C

measured curve at 60°C measured curve at 60°C
103
measured curve at 75°C 10 3 measured curve at 75°C
measured curve at 90°C measured curve at 90°C
102
102
tanδ (%)

tanδ (%)
101

101
100

predicted curve at 45°C predicted curve at 45°C

10-1 predicted curve at 60°C predicted curve at 60°C
100
predicted curve at 75°C predicted curve at 75°C
predicted curve at 90°C predicted curve at 90°C
10-2
10-3 10-2 10-1 100 101 102 103 10-3 10-2 10-1 100 101 102 103
f/Hz f/Hz

(a) (b)

measured curve at 45°C measured curve at 45°C

104 measured curve at 60°C measured curve at 60°C
measured curve at 75°C measured curve at 75°C
103
measured curve at 90°C measured curve at 90°C
103
tanδ (%)

tanδ (%)

102
2
10

101
101
predicted curve at 45°C predicted curve at 45°C
predicted curve at 60°C predicted curve at 60°C
predicted curve at 75°C predicted curve at 75°C
predicted curve at 90°C 100 predicted curve at 90°C
100
10-4 10 -3
10 -2
10 -1
100
10 1
10 2
103
10-3 10-2 10-1 100 101 102 103
f/Hz f/Hz

(c) (d)

measured curve at 45°C measured curve at 45°C

measured curve at 60°C measured curve at 60°C
102
measured curve at 75°C measured curve at 75°C
measured curve at 90°C 102 measured curve at 90°C

101
tanδ (%)
tanδ (%)

101

100
predicted curve at 45°C predicted curve at 45°C
100 predicted curve at 60°C
predicted curve at 60°C
predicted curve at 75°C predicted curve at 75°C
predicted curve at 90°C predicted curve at 90°C
10-1
10-3 10-2 10-1 100 101 102 103 10-3 10-2 10-1 100 101 102 103
f/Hz f/Hz

(e) (f)

Figure 7. The verification test of the reported universal approach for temperature correction.
Figure 7. The verification test of the reported universal approach for temperature correction.
(a) DP = 425, mc% = 1.2%, (b) DP = 994, mc% = 2.8%, (c) DP = 841, mc% = 3.7%, (d) DP = 1285,
(a) DP = 425, mc% = 1.2%, (b) DP = 994, mc% = 2.8%, (c) DP = 841, mc% = 3.7%, (d) DP = 1285,
mc% = 4.0%, (e) DP = 663, mc% = 0.9%, (f) DP = 640, mc% = 1.3%.
mc% = 4.0%, (e) DP = 663, mc% = 0.9%, (f) DP = 640, mc% = 1.3%.
6. Conclusions
6. Conclusions
In order to correct the temperature effect on FDS curve of transformer polymer insulation, the
In order to correct the temperature effect on FDS curve of transformer polymer insulation, the
knowledge of the shift factor is rather crucial. Therefore, a universal approach for temperature
knowledge of the shift factor is rather crucial. Therefore, a universal approach for temperature correction
on FDS curve is reported. The present analysis and findings have led to the following conclusions.
Polymers 2019, 11, 1126 10 of 11

I. The reported approach for predicting the shift factor is based upon the depth analysis of the
variation law among the shift factor, moisture content and test temperature. The findings
reveal that the shift factor is both temperature-dependent and moisture-dependent.
II. It is estimated that the value of the shift factor is positively correlated with the moisture content
and test temperature. Moreover, the temperature correction results indicate that the shift factor
mainly depends on the test temperature and moisture content rather than the aging degree.
III. The verification results preliminarily reveal that the reported approach is not only suitable for
temperature correction on FDS curve of pressboard with different insulation content, but also
maintains considerable accuracy when applied to different types of paperboard. Therefore, the
obtained conclusions will provide a universal idea for temperature correction on FDS curve of
transformer polymer insulation.

Author Contributions: In this research activity, J.L., X.F. designed the experiments and performed the writing;
Y.Z. (Yiyi Zhang) and H.Y. are both corresponding authors; all authors have analyzed the data and approved the
submitted manuscript.
Funding: This work was supported by the National Natural Science Foundation of China (61473272; 51867003),
the Natural Science Foundation of Guangxi (2018JJB160056; 2018JJB160064; 2018JJA160176), the Guangxi thousand
backbone teachers training program, the Boshike Award scheme for young innovative talents, the Basic Ability
Promotion Project for Yong Teachers in Universities of Guangxi (2019KY0046; 2019KY0022) and the Guangxi
Bagui young scholars special funding in support of this work.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Feng, D.W.; Yang, L.J.; Zhou, L.W.; Liao, R.J.; Chen, X. Effect of Oil–Paper–Pressboard Mass Ratio on Furfural
Content in Transformer Oil. IEEE Trans. Dielectr. Electr. Insul. 2019, 26, 1289–1296.
2. Feng, D.W.; Yang, L.J.; Liao, R.J.; Zhou, L.W. Effect of Moisture Content on the Production and Partitioning
of Furfural in Oil–Paper Insulation. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 2389–2397. [CrossRef]
3. Liu, J.; Zheng, H.; Zhang, Y.; Li, X.; Fang, J.; Liu, Y.; Liao, C.; Li, Y.; Zhao, J. Dissolved gases forecasting
based on wavelet least squares support vector regression and imperialist competition algorithm for assessing
incipient faults of transformer polymer insulation. Polymers 2019, 11, 85. [CrossRef] [PubMed]
4. Baral, A.; Chakravorti, S. Prediction of Moisture Present in Cellulosic Part of Power Transformer Insulation
using Transfer Function of Modified Debye Model. IEEE Trans. Dielectr. Electr. Insul. 2014, 21, 1368–1375.
[CrossRef]
5. Verma, H.C.; Baral, A.; Pradhan, A.K. A Method to Estimate Activation Energy of Power Transformer
Insulation Using Time Domain Spectroscopy Data. IEEE Trans. Dielectr. Electr. Insul. 2017, 24, 3245–3253.
[CrossRef]
6. Pradhan, A.K.; Chatterjee, B.; Chakravorti, S. Estimation of Dielectric Dissipation Factor of Cellulosic Parts
in Oil-paper Insulation by Frequency Domain Spectroscopy. IEEE Trans. Dielectr. Electr. Insul. 2016, 23,
2720–2729. [CrossRef]
7. Saha, T.K.; Purkait, P. Understanding the impacts of moisture and thermal aging on transformer’s insulation
by dielectric response and molecular weight measurements. IEEE Trans. Dielectr. Electr. Insul. 2008, 15,
568–582. [CrossRef]
8. Koch, M.; Prevost, T. Analysis of Dielectric Response Measurements for Condition Assessment of Oil-Paper
Transformer Insulation. IEEE Trans. Dielectr. Electr. Insul. 2012, 19, 1908–1915. [CrossRef]
9. Yang, L.; Zou, T.; Deng, B.; Zhang, H.; Mo, Y.; Peng, P. Assessment of Oil-Paper Insulation Aging Using
Frequency Domain Spectroscopy and Moisture Equilibrium Curves. IEEE Access 2019, 7, 45670–45678.
[CrossRef]
10. Pradhan, A.K.; Koley, C. Determination of optimized slope of triangular excitation for condition assessment
of oil-paper insulation by frequency domain spectroscopy. IEEE Trans. Dielectr. Electr. Insul. 2016, 23,
1303–1312. [CrossRef]
11. Jaya, M.; Geißler, D.; Leibfried, T. Accelerating dielectric response measurements on power transformers-part
I: A frequency-domain approach. IEEE Trans. Power Deliv. 2013, 28, 1469–1473. [CrossRef]
Polymers 2019, 11, 1126 11 of 11

12. Zhang, D.; Yun, H.; Zhan, J.; Sun, X.; He, W.; Niu, C.; Mu, H.; Zhang, G.-j. Insulation Condition Diagnosis of
Oil-Immersed Paper Insulation Based on Non-linear Frequency-Domain Dielectric Response. IEEE Trans.
Dielectr. Electr. Insul. 2018, 25, 1980–1988. [CrossRef]
13. Jadav, R.B.; Ekanayake, C.; Saha, T.K. Understanding the Impact of Moisture and Ageing of Transformer
Insulation on Frequency Domain Spectroscopy. IEEE Trans. Dielectr. Electr. Insul. 2014, 21, 369–379. [CrossRef]
14. Liu, J.; Zheng, H.; Zhang, Y.; Wei, H.; Liao, R. Grey Relational Analysis for Insulation Condition Assessment
of Power Transformers Based Upon Conventional Dielectric Response Measurement. Energies 2017, 10, 1526.
[CrossRef]
15. Yang, X.; Nielsen, S.; Ledwich, G. Frequency Domain Spectroscopy Measurements of Oil-paper Insulation
for Energized Transformers. IEEE Trans. Dielectr. Electr. Insul. 2017, 24, 1657–1664. [CrossRef]
16. Ojha, S.K.; Purkait, P. Modeling of Relaxation Phenomena in Transformer Oil-paper Insulation for
Understanding Dielectric Response Measurements. IEEE Trans. Dielectr. Electr. Insul. 2016, 23, 3190–3198.
[CrossRef]
17. Saha, T.K.; Purkait, P. Investigations of Temperature Effects on the Dielectric Response Measurements of
Transformer Oil-Paper Insulation System. IEEE Trans. Power Deliv. 2008, 23, 252–260. [CrossRef]
18. Baral, A.; Chakravorti, S. Compensating the Effect of Temperature Variation on Dielectric Response of
Oil-paper Insulation used in Power Transformers. IEEE Trans. Dielectr. Electr. Insul. 2016, 23, 2462–2474.
[CrossRef]
19. Dutta, S.; Basalt, A.; Pradhan, A.K.; Chakravorti, S. Effect of measurement temperature on power transformer
insulation diagnosis using frequency-domain spectroscopy. IET Sci. Meas. Technol. 2017, 11, 773–779.
[CrossRef]
20. Zhang, Y.; Liu, J.; Zheng, H.; Wang, K. Feasibility of a universal approach for temperature correction in
frequency domain spectroscopy of transformer insulation. IEEE Trans. Dielectr. Electr. Insul. 2018, 25,
1766–1773. [CrossRef]
21. Liao, R.J.; Liu, J.F.; Yang, L.J. Quantitative analysis of insulation condition of oil-paper insulation based on
frequency domain spectroscopy. IEEE Trans. Dielectr. Electr. Insul. 2015, 22, 322–334. [CrossRef]
22. Gao, J.; Yang, L.J.; Wang, Y.Y. Effect of moisture and thermal degradation on the activation energy of oil-paper
insulation in frequency domain spectroscope measurement. IET Gener. Transm. Distrib. 2016, 10, 2042–2049.
[CrossRef]
23. Linhjell, D.; Lundgaard, L.E.; Gafvert, U. Dielectric response of mineral oil impregnated cellulose and the
impact of aging. IEEE Trans. Dielectr. Electr. Insul. 2007, 14, 156–169. [CrossRef]
24. Liu, J.; Fan, X.; Zheng, H.; Zhang, Y.; Zhang, C.; Lai, B.; Wang, J.; Ren, G.; Zhang, E. Aging condition
assessment of transformer oil-immersed cellulosic insulation based upon the average activation energy
method. Cellulose 2019, 26, 3891–3908. [CrossRef]
25. Liu, J.; Zheng, H.; Zhang, Y.; Zhou, T.; Zhao, J.; Li, J.; Liu, J.; Li, J. Comparative investigation on the
performance of modified system poles and traditional system poles obtained from PDC data for diagnosing
ageing condition of transformers polymer insulation materials. Polymers 2018, 10, 191. [CrossRef] [PubMed]
26. Zhang, Y.; Liu, J.; Zheng, H.; Wei, H.; Liao, R. Study on Quantitative Correlations between the Ageing
Condition of Transformer Cellulose Insulation and the Large Time Constant Obtained from the Extended
Debye Model. Energies 2017, 10, 1842. [CrossRef]
27. Gillen, K.T.; Celina, M. The wear-out approach for predicting the remaining lifetime of materials. Polym.
Degrad. Stab. 2000, 71, 15–30. [CrossRef]
28. Urzhumtsev, Y.S. Time-temperature superposition. Rev. Mech. Compos. Mater. 1975, 11, 57–72.
29. Lelekakis, N.; Martin, D.; Wijaya, J. Ageing rate of paper insulation used in power transformers Part 1:
Oil/paper system with low oxygen concentration. IEEE Trans. Dielectr. Electr. Insul. 2012, 19, 1999–2008.
[CrossRef]
30. Lelekakis, N.; Martin, D.; Wijaya, J. Ageing rate of paper insulation used in power transformers Part 2:
Oil/paper system with medium and high oxygen concentration. IEEE Trans. Dielectr. Electr. Insul. 2012, 19,
2009–2018. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).